From af73257b093737838d6086890c91f6ec13291ea7 Mon Sep 17 00:00:00 2001 From: Conrad Meyer Date: Sat, 29 Dec 2018 06:51:10 +0000 Subject: [PATCH] import zstd 1.3.8 --- Makefile | 10 +- NEWS | 402 ---- README.md | 54 +- appveyor.yml | 2 + contrib/adaptive-compression/Makefile | 4 +- contrib/adaptive-compression/adapt.c | 2 +- contrib/adaptive-compression/datagencli.c | 2 +- contrib/meson/README | 3 - contrib/meson/meson.build | 144 -- contrib/meson/meson_options.txt | 3 - contrib/pzstd/Makefile | 2 +- .../examples/parallel_processing.c | 16 +- contrib/seekable_format/zstdseek_compress.c | 35 +- contrib/seekable_format/zstdseek_decompress.c | 31 +- doc/zstd_compression_format.md | 37 +- doc/zstd_manual.html | 1370 ++++++------ lib/BUCK | 17 +- lib/Makefile | 39 +- lib/README.md | 108 +- lib/common/bitstream.h | 4 +- lib/common/compiler.h | 31 +- lib/common/cpu.h | 2 +- lib/common/debug.h | 33 +- lib/common/error_private.c | 6 + lib/common/fse.h | 4 +- lib/common/huf.h | 26 +- lib/common/pool.c | 2 +- lib/common/zstd_common.c | 4 +- lib/common/zstd_errors.h | 1 + lib/common/zstd_internal.h | 13 +- lib/compress/fse_compress.c | 6 +- lib/compress/hist.c | 30 +- lib/compress/hist.h | 19 +- lib/compress/huf_compress.c | 64 +- lib/compress/zstd_compress.c | 990 +++++---- lib/compress/zstd_compress_internal.h | 118 +- lib/compress/zstd_double_fast.c | 8 +- lib/compress/zstd_fast.c | 30 +- lib/compress/zstd_lazy.c | 43 +- lib/compress/zstd_ldm.c | 85 +- lib/compress/zstd_ldm.h | 8 +- lib/compress/zstd_opt.c | 181 +- lib/compress/zstd_opt.h | 8 + lib/compress/zstdmt_compress.c | 302 ++- lib/compress/zstdmt_compress.h | 28 +- lib/decompress/huf_decompress.c | 220 +- lib/decompress/zstd_ddict.c | 240 ++ lib/decompress/zstd_ddict.h | 44 + lib/decompress/zstd_decompress.c | 1924 +++-------------- lib/decompress/zstd_decompress_block.c | 1307 +++++++++++ lib/decompress/zstd_decompress_block.h | 59 + lib/decompress/zstd_decompress_internal.h | 168 ++ lib/dictBuilder/cover.c | 24 +- lib/dictBuilder/fastcover.c | 30 +- lib/dictBuilder/zdict.c | 56 +- lib/legacy/zstd_v04.c | 10 - lib/legacy/zstd_v05.c | 30 +- lib/zstd.h | 1564 ++++++++------ programs/Makefile | 45 +- programs/README.md | 5 + programs/benchfn.c | 263 +++ programs/benchfn.h | 167 ++ programs/{bench.c => benchzstd.c} | 420 +--- programs/{bench.h => benchzstd.h} | 119 +- programs/datagen.c | 14 +- programs/dibio.c | 6 +- programs/fileio.c | 262 ++- programs/fileio.h | 4 +- programs/platform.h | 1 + programs/util.c | 673 ++++++ programs/util.h | 631 +----- programs/zstd.1 | 34 +- programs/zstd.1.md | 52 +- programs/zstdcli.c | 100 +- programs/zstdgrep | 141 +- programs/zstdgrep.1 | 2 +- programs/zstdless.1 | 2 +- tests/Makefile | 30 +- tests/README.md | 18 +- tests/datagencli.c | 2 +- tests/decodecorpus.c | 88 +- tests/fullbench.c | 120 +- tests/fuzz/Makefile | 6 +- tests/fuzz/simple_round_trip.c | 4 +- tests/fuzz/stream_round_trip.c | 12 +- tests/fuzz/zstd_helpers.c | 42 +- tests/fuzzer.c | 576 +++-- tests/libzstd_partial_builds.sh | 89 +- tests/longmatch.c | 2 +- tests/paramgrill.c | 791 ++++--- tests/playTests.sh | 55 +- tests/poolTests.c | 12 +- tests/regression/Makefile | 58 + tests/regression/config.c | 230 ++ tests/regression/config.h | 86 + tests/regression/data.c | 617 ++++++ tests/regression/data.h | 140 ++ tests/regression/levels.h | 44 + tests/regression/method.c | 566 +++++ tests/regression/method.h | 65 + tests/regression/result.c | 28 + tests/regression/result.h | 103 + tests/regression/results.csv | 448 ++++ tests/regression/test.c | 362 ++++ tests/roundTripCrash.c | 10 +- tests/seqgen.c | 2 +- tests/symbols.c | 1 - tests/zstreamtest.c | 274 +-- zlibWrapper/Makefile | 16 +- zlibWrapper/examples/zwrapbench.c | 25 +- zlibWrapper/zstd_zlibwrapper.c | 6 +- 111 files changed, 11292 insertions(+), 6580 deletions(-) delete mode 100644 NEWS delete mode 100644 contrib/meson/README delete mode 100644 contrib/meson/meson.build delete mode 100644 contrib/meson/meson_options.txt create mode 100644 lib/decompress/zstd_ddict.c create mode 100644 lib/decompress/zstd_ddict.h create mode 100644 lib/decompress/zstd_decompress_block.c create mode 100644 lib/decompress/zstd_decompress_block.h create mode 100644 lib/decompress/zstd_decompress_internal.h create mode 100644 programs/benchfn.c create mode 100644 programs/benchfn.h rename programs/{bench.c => benchzstd.c} (69%) rename programs/{bench.h => benchzstd.h} (61%) create mode 100644 programs/util.c create mode 100644 tests/regression/Makefile create mode 100644 tests/regression/config.c create mode 100644 tests/regression/config.h create mode 100644 tests/regression/data.c create mode 100644 tests/regression/data.h create mode 100644 tests/regression/levels.h create mode 100644 tests/regression/method.c create mode 100644 tests/regression/method.h create mode 100644 tests/regression/result.c create mode 100644 tests/regression/result.h create mode 100644 tests/regression/results.csv create mode 100644 tests/regression/test.c diff --git a/Makefile b/Makefile index c63db80e9e0c..ed78d1956135 100644 --- a/Makefile +++ b/Makefile @@ -64,7 +64,8 @@ zlibwrapper: lib ## test: run long-duration tests .PHONY: test -test: MOREFLAGS += -g -DDEBUGLEVEL=1 -Werror +DEBUGLEVEL ?= 1 +test: MOREFLAGS += -g -DDEBUGLEVEL=$(DEBUGLEVEL) -Werror test: MOREFLAGS="$(MOREFLAGS)" $(MAKE) -j -C $(PRGDIR) allVariants $(MAKE) -C $(TESTDIR) $@ @@ -129,7 +130,12 @@ ifneq (,$(filter $(shell uname),Linux Darwin GNU/kFreeBSD GNU OpenBSD FreeBSD Dr HOST_OS = POSIX CMAKE_PARAMS = -DZSTD_BUILD_CONTRIB:BOOL=ON -DZSTD_BUILD_STATIC:BOOL=ON -DZSTD_BUILD_TESTS:BOOL=ON -DZSTD_ZLIB_SUPPORT:BOOL=ON -DZSTD_LZMA_SUPPORT:BOOL=ON -DCMAKE_BUILD_TYPE=Release -EGREP = egrep --color=never +HAVE_COLORNEVER = $(shell echo a | egrep --color=never a > /dev/null 2> /dev/null && echo 1 || echo 0) +EGREP_OPTIONS ?= +ifeq ($HAVE_COLORNEVER, 1) +EGREP_OPTIONS += --color=never +endif +EGREP = egrep $(EGREP_OPTIONS) # Print a two column output of targets and their description. To add a target description, put a # comment in the Makefile with the format "## : ". For example: diff --git a/NEWS b/NEWS deleted file mode 100644 index 637bd93038e7..000000000000 --- a/NEWS +++ /dev/null @@ -1,402 +0,0 @@ -v1.3.7 -perf: slightly better decompression speed on clang (depending on hardware target) -fix : performance of dictionary compression for small input < 4 KB at levels 9 and 10 -build: no longer build backtrace by default in release mode; restrict further automatic mode -build: control backtrace support through build macro BACKTRACE -misc: added man pages for zstdless and zstdgrep, by @samrussell - -v1.3.6 -perf: much faster dictionary builder, by @jenniferliu -perf: faster dictionary compression on small data when using multiple contexts, by @felixhandte -perf: faster dictionary decompression when using a very large number of dictionaries simultaneously -cli : fix : does no longer overwrite destination when source does not exist (#1082) -cli : new command --adapt, for automatic compression level adaptation -api : fix : block api can be streamed with > 4 GB, reported by @catid -api : reduced ZSTD_DDict size by 2 KB -api : minimum negative compression level is defined, and can be queried using ZSTD_minCLevel(). -build: support Haiku target, by @korli -build: Read Legacy format is limited to v0.5+ by default. Can be changed at compile time with macro ZSTD_LEGACY_SUPPORT. -doc : zstd_compression_format.md updated to match wording in IETF RFC 8478 -misc: tests/paramgrill, a parameter optimizer, by @GeorgeLu97 - -v1.3.5 -perf: much faster dictionary compression, by @felixhandte -perf: small quality improvement for dictionary generation, by @terrelln -perf: slightly improved high compression levels (notably level 19) -mem : automatic memory release for long duration contexts -cli : fix : overlapLog can be manually set -cli : fix : decoding invalid lz4 frames -api : fix : performance degradation for dictionary compression when using advanced API, by @terrelln -api : change : clarify ZSTD_CCtx_reset() vs ZSTD_CCtx_resetParameters(), by @terrelln -build: select custom libzstd scope through control macros, by @GeorgeLu97 -build: OpenBSD patch, by @bket -build: make and make all are compatible with -j -doc : clarify zstd_compression_format.md, updated for IETF RFC process -misc: pzstd compatible with reproducible compilation, by @lamby - -v1.3.4 -perf: faster speed (especially decoding speed) on recent cpus (haswell+) -perf: much better performance associating --long with multi-threading, by @terrelln -perf: better compression at levels 13-15 -cli : asynchronous compression by default, for faster experience (use --single-thread for former behavior) -cli : smoother status report in multi-threading mode -cli : added command --fast=#, for faster compression modes -cli : fix crash when not overwriting existing files, by Pádraig Brady (@pixelb) -api : `nbThreads` becomes `nbWorkers` : 1 triggers asynchronous mode -api : compression levels can be negative, for even more speed -api : ZSTD_getFrameProgression() : get precise progress status of ZSTDMT anytime -api : ZSTDMT can accept new compression parameters during compression -api : implemented all advanced dictionary decompression prototypes -build: improved meson recipe, by Shawn Landden (@shawnl) -build: VS2017 scripts, by @HaydnTrigg -misc: all /contrib projects fixed -misc: added /contrib/docker script by @gyscos - -v1.3.3 -perf: faster zstd_opt strategy (levels 16-19) -fix : bug #944 : multithreading with shared ditionary and large data, reported by @gsliepen -cli : fix : content size written in header by default -cli : fix : improved LZ4 format support, by @felixhandte -cli : new : hidden command `-S`, to benchmark multiple files while generating one result per file -api : fix : support large skippable frames, by @terrelln -api : fix : streaming interface was adding a useless 3-bytes null block to small frames -api : change : when setting `pledgedSrcSize`, use `ZSTD_CONTENTSIZE_UNKNOWN` macro value to mean "unknown" -build: fix : compilation under rhel6 and centos6, reported by @pixelb -build: added `check` target - -v1.3.2 -new : long range mode, using --long command, by Stella Lau (@stellamplau) -new : ability to generate and decode magicless frames (#591) -changed : maximum nb of threads reduced to 200, to avoid address space exhaustion in 32-bits mode -fix : multi-threading compression works with custom allocators -fix : ZSTD_sizeof_CStream() was over-evaluating memory usage -fix : a rare compression bug when compression generates very large distances and bunch of other conditions (only possible at --ultra -22) -fix : 32-bits build can now decode large offsets (levels 21+) -cli : added LZ4 frame support by default, by Felix Handte (@felixhandte) -cli : improved --list output -cli : new : can split input file for dictionary training, using command -B# -cli : new : clean operation artefact on Ctrl-C interruption -cli : fix : do not change /dev/null permissions when using command -t with root access, reported by @mike155 (#851) -cli : fix : write file size in header in multiple-files mode -api : added macro ZSTD_COMPRESSBOUND() for static allocation -api : experimental : new advanced decompression API -api : fix : sizeof_CCtx() used to over-estimate -build: fix : no-multithread variant compiles without pool.c dependency, reported by Mitchell Blank Jr (@mitchblank) (#819) -build: better compatibility with reproducible builds, by Bernhard M. Wiedemann (@bmwiedemann) (#818) -example : added streaming_memory_usage -license : changed /examples license to BSD + GPLv2 -license : fix a few header files to reflect new license (#825) - -v1.3.1 -New license : BSD + GPLv2 -perf: substantially decreased memory usage in Multi-threading mode, thanks to reports by Tino Reichardt (@mcmilk) -perf: Multi-threading supports up to 256 threads. Cap at 256 when more are requested (#760) -cli : improved and fixed --list command, by @ib (#772) -cli : command -vV to list supported formats, by @ib (#771) -build : fixed binary variants, reported by @svenha (#788) -build : fix Visual compilation for non x86/x64 targets, reported by Greg Slazinski (@GregSlazinski) (#718) -API exp : breaking change : ZSTD_getframeHeader() provides more information -API exp : breaking change : pinned down values of error codes -doc : fixed huffman example, by Ulrich Kunitz (@ulikunitz) -new : contrib/adaptive-compression, I/O driven compression strength, by Paul Cruz (@paulcruz74) -new : contrib/long_distance_matching, statistics by Stella Lau (@stellamplau) -updated : contrib/linux-kernel, by Nick Terrell (@terrelln) - -v1.3.0 -cli : new : `--list` command, by Paul Cruz -cli : changed : xz/lzma support enabled by default -cli : changed : `-t *` continue processing list after a decompression error -API : added : ZSTD_versionString() -API : promoted to stable status : ZSTD_getFrameContentSize(), by Sean Purcell -API exp : new advanced API : ZSTD_compress_generic(), ZSTD_CCtx_setParameter() -API exp : new : API for static or external allocation : ZSTD_initStatic?Ctx() -API exp : added : ZSTD_decompressBegin_usingDDict(), requested by Guy Riddle (#700) -API exp : clarified memory estimation / measurement functions. -API exp : changed : strongest strategy renamed ZSTD_btultra, fastest strategy ZSTD_fast set to 1 -tools : decodecorpus can generate random dictionary-compressed samples, by Paul Cruz -new : contrib/seekable_format, demo and API, by Sean Purcell -changed : contrib/linux-kernel, updated version and license, by Nick Terrell - -v1.2.0 -cli : changed : Multithreading enabled by default (use target zstd-nomt or HAVE_THREAD=0 to disable) -cli : new : command -T0 means "detect and use nb of cores", by Sean Purcell -cli : new : zstdmt symlink hardwired to `zstd -T0` -cli : new : command --threads=# (#671) -cli : changed : cover dictionary builder by default, for improved quality, by Nick Terrell -cli : new : commands --train-cover and --train-legacy, to select dictionary algorithm and parameters -cli : experimental targets `zstd4` and `xzstd4`, with support for lz4 format, by Sean Purcell -cli : fix : does not output compressed data on console -cli : fix : ignore symbolic links unless --force specified, -API : breaking change : ZSTD_createCDict_advanced(), only use compressionParameters as argument -API : added : prototypes ZSTD_*_usingCDict_advanced(), for direct control over frameParameters. -API : improved: ZSTDMT_compressCCtx() reduced memory usage -API : fix : ZSTDMT_compressCCtx() now provides srcSize in header (#634) -API : fix : src size stored in frame header is controlled at end of frame -API : fix : enforced consistent rules for pledgedSrcSize==0 (#641) -API : fix : error code "GENERIC" replaced by "dstSizeTooSmall" when appropriate -build: improved cmake script, by @Majlen -build: enabled Multi-threading support for *BSD, by Baptiste Daroussin -tools: updated Paramgrill. Command -O# provides best parameters for sample and speed target. -new : contrib/linux-kernel version, by Nick Terrell - -v1.1.4 -cli : new : can compress in *.gz format, using --format=gzip command, by Przemyslaw Skibinski -cli : new : advanced benchmark command --priority=rt -cli : fix : write on sparse-enabled file systems in 32-bits mode, by @ds77 -cli : fix : --rm remains silent when input is stdin -cli : experimental : xzstd, with support for xz/lzma decoding, by Przemyslaw Skibinski -speed : improved decompression speed in streaming mode for single shot scenarios (+5%) -memory: DDict (decompression dictionary) memory usage down from 150 KB to 20 KB -arch: 32-bits variant able to generate and decode very long matches (>32 MB), by Sean Purcell -API : new : ZSTD_findFrameCompressedSize(), ZSTD_getFrameContentSize(), ZSTD_findDecompressedSize() -API : changed : dropped support of legacy versions <= v0.3 (can be changed by modifying ZSTD_LEGACY_SUPPORT value) -build : new: meson build system in contrib/meson, by Dima Krasner -build : improved cmake script, by @Majlen -build : added -Wformat-security flag, as recommended by Padraig Brady -doc : new : educational decoder, by Sean Purcell - -v1.1.3 -cli : zstd can decompress .gz files (can be disabled with `make zstd-nogz` or `make HAVE_ZLIB=0`) -cli : new : experimental target `make zstdmt`, with multi-threading support -cli : new : improved dictionary builder "cover" (experimental), by Nick Terrell, based on prior work by Giuseppe Ottaviano. -cli : new : advanced commands for detailed parameters, by Przemyslaw Skibinski -cli : fix zstdless on Mac OS-X, by Andrew Janke -cli : fix #232 "compress non-files" -dictBuilder : improved dictionary generation quality, thanks to Nick Terrell -API : new : lib/compress/ZSTDMT_compress.h multithreading API (experimental) -API : new : ZSTD_create?Dict_byReference(), requested by Bartosz Taudul -API : new : ZDICT_finalizeDictionary() -API : fix : ZSTD_initCStream_usingCDict() properly writes dictID into frame header, by Gregory Szorc (#511) -API : fix : all symbols properly exposed in libzstd, by Nick Terrell -build : support for Solaris target, by Przemyslaw Skibinski -doc : clarified specification, by Sean Purcell - -v1.1.2 -API : streaming : decompression : changed : automatic implicit reset when chain-decoding new frames without init -API : experimental : added : dictID retrieval functions, and ZSTD_initCStream_srcSize() -API : zbuff : changed : prototypes now generate deprecation warnings -lib : improved : faster decompression speed at ultra compression settings and 32-bits mode -lib : changed : only public ZSTD_ symbols are now exposed -lib : changed : reduced usage of stack memory -lib : fixed : several corner case bugs, by Nick Terrell -cli : new : gzstd, experimental version able to decode .gz files, by Przemyslaw Skibinski -cli : new : preserve file attributes -cli : new : added zstdless and zstdgrep tools -cli : fixed : status displays total amount decoded, even for file consisting of multiple frames (like pzstd) -cli : fixed : zstdcat -zlib_wrapper : added support for gz* functions, by Przemyslaw Skibinski -install : better compatibility with FreeBSD, by Dimitry Andric -source tree : changed : zbuff source files moved to lib/deprecated - -v1.1.1 -New : command -M#, --memory=, --memlimit=, --memlimit-decompress= to limit allowed memory consumption -New : doc/zstd_manual.html, by Przemyslaw Skibinski -Improved : slightly better compression ratio at --ultra levels (>= 20) -Improved : better memory usage when using streaming compression API, thanks to @Rogier-5 report -Added : API : ZSTD_initCStream_usingCDict(), ZSTD_initDStream_usingDDict() (experimental section) -Added : example/multiple_streaming_compression.c -Changed : zstd_errors.h is now installed within /include (and replaces errors_public.h) -Updated man page -Fixed : zstd-small, zstd-compress and zstd-decompress compilation targets - -v1.1.0 -New : contrib/pzstd, parallel version of zstd, by Nick Terrell -added : NetBSD install target (#338) -Improved : speed for batches of small files -Improved : speed of zlib wrapper, by Przemyslaw Skibinski -Changed : libzstd on Windows supports legacy formats, by Christophe Chevalier -Fixed : CLI -d output to stdout by default when input is stdin (#322) -Fixed : CLI correctly detects console on Mac OS-X -Fixed : CLI supports recursive mode `-r` on Mac OS-X -Fixed : Legacy decoders use unified error codes, reported by benrg (#341), fixed by Przemyslaw Skibinski -Fixed : compatibility with OpenBSD, reported by Juan Francisco Cantero Hurtado (#319) -Fixed : compatibility with Hurd, by Przemyslaw Skibinski (#365) -Fixed : zstd-pgo, reported by octoploid (#329) - -v1.0.0 -Change Licensing, all project is now BSD, Copyright Facebook -Small decompression speed improvement -API : Streaming API supports legacy format -API : ZDICT_getDictID(), ZSTD_sizeof_{CCtx, DCtx, CStream, DStream}(), ZSTD_setDStreamParamter() -CLI supports legacy formats v0.4+ -Fixed : compression fails on certain huge files, reported by Jesse McGrew -Enhanced documentation, by Przemyslaw Skibinski - -v0.8.1 -New streaming API -Changed : --ultra now enables levels beyond 19 -Changed : -i# now selects benchmark time in second -Fixed : ZSTD_compress* can now compress > 4 GB in a single pass, reported by Nick Terrell -Fixed : speed regression on specific patterns (#272) -Fixed : support for Z_SYNC_FLUSH, by Dmitry Krot (#291) -Fixed : ICC compilation, by Przemyslaw Skibinski - -v0.8.0 -Improved : better speed on clang and gcc -O2, thanks to Eric Biggers -New : Build on FreeBSD and DragonFly, thanks to JrMarino -Changed : modified API : ZSTD_compressEnd() -Fixed : legacy mode with ZSTD_HEAPMODE=0, by Christopher Bergqvist -Fixed : premature end of frame when zero-sized raw block, reported by Eric Biggers -Fixed : large dictionaries (> 384 KB), reported by Ilona Papava -Fixed : checksum correctly checked in single-pass mode -Fixed : combined --test amd --rm, reported by Andreas M. Nilsson -Modified : minor compression level adaptations -Updated : compression format specification to v0.2.0 -changed : zstd.h moved to /lib directory - -v0.7.5 -Transition version, supporting decoding of v0.8.x - -v0.7.4 -Added : homebrew for Mac, by Daniel Cade -Added : more examples -Fixed : segfault when using small dictionaries, reported by Felix Handte -Modified : default compression level for CLI is now 3 -Updated : specification, to v0.1.1 - -v0.7.3 -New : compression format specification -New : `--` separator, stating that all following arguments are file names. Suggested by Chip Turner. -New : `ZSTD_getDecompressedSize()` -New : OpenBSD target, by Juan Francisco Cantero Hurtado -New : `examples` directory -fixed : dictBuilder using HC levels, reported by Bartosz Taudul -fixed : legacy support from ZSTD_decompress_usingDDict(), reported by Felix Handte -fixed : multi-blocks decoding with intermediate uncompressed blocks, reported by Greg Slazinski -modified : removed "mem.h" and "error_public.h" dependencies from "zstd.h" (experimental section) -modified : legacy functions no longer need magic number - -v0.7.2 -fixed : ZSTD_decompressBlock() using multiple consecutive blocks. Reported by Greg Slazinski. -fixed : potential segfault on very large files (many gigabytes). Reported by Chip Turner. -fixed : CLI displays system error message when destination file cannot be created (#231). Reported by Chip Turner. - -v0.7.1 -fixed : ZBUFF_compressEnd() called multiple times with too small `dst` buffer, reported by Christophe Chevalier -fixed : dictBuilder fails if first sample is too small, reported by Руслан Ковалёв -fixed : corruption issue, reported by cj -modified : checksum enabled by default in command line mode - -v0.7.0 -New : Support for directory compression, using `-r`, thanks to Przemyslaw Skibinski -New : Command `--rm`, to remove source file after successful de/compression -New : Visual build scripts, by Christophe Chevalier -New : Support for Sparse File-systems (do not use space for zero-filled sectors) -New : Frame checksum support -New : Support pass-through mode (when using `-df`) -API : more efficient Dictionary API : `ZSTD_compress_usingCDict()`, `ZSTD_decompress_usingDDict()` -API : create dictionary files from custom content, by Giuseppe Ottaviano -API : support for custom malloc/free functions -New : controllable Dictionary ID -New : Support for skippable frames - -v0.6.1 -New : zlib wrapper API, thanks to Przemyslaw Skibinski -New : Ability to compile compressor / decompressor separately -Changed : new lib directory structure -Fixed : Legacy codec v0.5 compatible with dictionary decompression -Fixed : Decoder corruption error (#173) -Fixed : null-string roundtrip (#176) -New : benchmark mode can select directory as input -Experimental : midipix support, VMS support - -v0.6.0 -Stronger high compression modes, thanks to Przemyslaw Skibinski -API : ZSTD_getFrameParams() provides size of decompressed content -New : highest compression modes require `--ultra` command to fully unleash their capacity -Fixed : zstd cli return error code > 0 and removes dst file artifact when decompression fails, thanks to Chip Turner - -v0.5.1 -New : Optimal parsing => Very high compression modes, thanks to Przemyslaw Skibinski -Changed : Dictionary builder integrated into libzstd and zstd cli -Changed (!) : zstd cli now uses "multiple input files" as default mode. See `zstd -h`. -Fix : high compression modes for big-endian platforms -New : zstd cli : `-t` | `--test` command - -v0.5.0 -New : dictionary builder utility -Changed : streaming & dictionary API -Improved : better compression of small data - -v0.4.7 -Improved : small compression speed improvement in HC mode -Changed : `zstd_decompress.c` has ZSTD_LEGACY_SUPPORT to 0 by default -fix : bt search bug - -v0.4.6 -fix : fast compression mode on Windows -New : cmake configuration file, thanks to Artyom Dymchenko -Improved : high compression mode on repetitive data -New : block-level API -New : ZSTD_duplicateCCtx() - -v0.4.5 -new : -m/--multiple : compress/decompress multiple files - -v0.4.4 -Fixed : high compression modes for Windows 32 bits -new : external dictionary API extended to buffered mode and accessible through command line -new : windows DLL project, thanks to Christophe Chevalier - -v0.4.3 : -new : external dictionary API -new : zstd-frugal - -v0.4.2 : -Generic minor improvements for small blocks -Fixed : big-endian compatibility, by Peter Harris (#85) - -v0.4.1 -Fixed : ZSTD_LEGACY_SUPPORT=0 build mode (reported by Luben) -removed `zstd.c` - -v0.4.0 -Command line utility compatible with high compression levels -Removed zstdhc => merged into zstd -Added : ZBUFF API (see zstd_buffered.h) -Rolling buffer support - -v0.3.6 -small blocks params - -v0.3.5 -minor generic compression improvements - -v0.3.4 -Faster fast cLevels - -v0.3.3 -Small compression ratio improvement - -v0.3.2 -Fixed Visual Studio - -v0.3.1 : -Small compression ratio improvement - -v0.3 -HC mode : compression levels 2-26 - -v0.2.2 -Fix : Visual Studio 2013 & 2015 release compilation, by Christophe Chevalier - -v0.2.1 -Fix : Read errors, advanced fuzzer tests, by Hanno Böck - -v0.2.0 -**Breaking format change** -Faster decompression speed -Can still decode v0.1 format - -v0.1.3 -fix uninitialization warning, reported by Evan Nemerson - -v0.1.2 -frame concatenation support - -v0.1.1 -fix compression bug -detects write-flush errors - -v0.1.0 -first release diff --git a/README.md b/README.md index dc99dc0fd30a..4b6d19e7e72d 100644 --- a/README.md +++ b/README.md @@ -9,7 +9,11 @@ and a command line utility producing and decoding `.zst`, `.gz`, `.xz` and `.lz4 Should your project require another programming language, a list of known ports and bindings is provided on [Zstandard homepage](http://www.zstd.net/#other-languages). -Development branch status : [![Build Status][travisDevBadge]][travisLink] [![Build status][AppveyorDevBadge]][AppveyorLink] [![Build status][CircleDevBadge]][CircleLink] +**Development branch status:** + +[![Build Status][travisDevBadge]][travisLink] +[![Build status][AppveyorDevBadge]][AppveyorLink] +[![Build status][CircleDevBadge]][CircleLink] [travisDevBadge]: https://travis-ci.org/facebook/zstd.svg?branch=dev "Continuous Integration test suite" [travisLink]: https://travis-ci.org/facebook/zstd @@ -18,7 +22,7 @@ Development branch status : [![Build Status][travisDevBadge]][travisLink] [![B [CircleDevBadge]: https://circleci.com/gh/facebook/zstd/tree/dev.svg?style=shield "Short test suite" [CircleLink]: https://circleci.com/gh/facebook/zstd -### Benchmarks +## Benchmarks For reference, several fast compression algorithms were tested and compared on a server running Linux Debian (`Linux version 4.14.0-3-amd64`), @@ -42,7 +46,7 @@ on the [Silesia compression corpus]. | snappy 1.1.4 | 2.091 | 530 MB/s | 1800 MB/s | | lzf 3.6 -1 | 2.077 | 400 MB/s | 860 MB/s | -[zlib]:http://www.zlib.net/ +[zlib]: http://www.zlib.net/ [LZ4]: http://www.lz4.org/ Zstd can also offer stronger compression ratios at the cost of compression speed. @@ -65,7 +69,7 @@ A few other algorithms can produce higher compression ratios at slower speeds, f For a larger picture including slow modes, [click on this link](doc/images/DCspeed5.png). -### The case for Small Data compression +## The case for Small Data compression Previous charts provide results applicable to typical file and stream scenarios (several MB). Small data comes with different perspectives. @@ -89,24 +93,24 @@ Training works if there is some correlation in a family of small data samples. T Hence, deploying one dictionary per type of data will provide the greatest benefits. Dictionary gains are mostly effective in the first few KB. Then, the compression algorithm will gradually use previously decoded content to better compress the rest of the file. -#### Dictionary compression How To: +### Dictionary compression How To: -1) Create the dictionary +1. Create the dictionary -`zstd --train FullPathToTrainingSet/* -o dictionaryName` + `zstd --train FullPathToTrainingSet/* -o dictionaryName` -2) Compress with dictionary +2. Compress with dictionary -`zstd -D dictionaryName FILE` + `zstd -D dictionaryName FILE` -3) Decompress with dictionary +3. Decompress with dictionary -`zstd -D dictionaryName --decompress FILE.zst` + `zstd -D dictionaryName --decompress FILE.zst` -### Build instructions +## Build instructions -#### Makefile +### Makefile If your system is compatible with standard `make` (or `gmake`), invoking `make` in root directory will generate `zstd` cli in root directory. @@ -115,7 +119,7 @@ Other available options include: - `make install` : create and install zstd cli, library and man pages - `make check` : create and run `zstd`, tests its behavior on local platform -#### cmake +### cmake A `cmake` project generator is provided within `build/cmake`. It can generate Makefiles or other build scripts @@ -123,11 +127,17 @@ to create `zstd` binary, and `libzstd` dynamic and static libraries. By default, `CMAKE_BUILD_TYPE` is set to `Release`. -#### Meson +### Meson -A Meson project is provided within `contrib/meson`. +A Meson project is provided within [`build/meson`](build/meson). Follow +build instructions in that directory. -#### Visual Studio (Windows) +You can also take a look at [`.travis.yml`](.travis.yml) file for an +example about how Meson is used to build this project. + +Note that default build type is **release**. + +### Visual Studio (Windows) Going into `build` directory, you will find additional possibilities: - Projects for Visual Studio 2005, 2008 and 2010. @@ -135,17 +145,21 @@ Going into `build` directory, you will find additional possibilities: - Automated build scripts for Visual compiler by [@KrzysFR](https://github.com/KrzysFR), in `build/VS_scripts`, which will build `zstd` cli and `libzstd` library without any need to open Visual Studio solution. +### Buck -### Status +You can build the zstd binary via buck by executing: `buck build programs:zstd` from the root of the repo. +The output binary will be in `buck-out/gen/programs/`. + +## Status Zstandard is currently deployed within Facebook. It is used continuously to compress large amounts of data in multiple formats and use cases. Zstandard is considered safe for production environments. -### License +## License Zstandard is dual-licensed under [BSD](LICENSE) and [GPLv2](COPYING). -### Contributing +## Contributing The "dev" branch is the one where all contributions are merged before reaching "master". If you plan to propose a patch, please commit into the "dev" branch, or its own feature branch. diff --git a/appveyor.yml b/appveyor.yml index 2b674ce3ca1c..35f019dd6b09 100644 --- a/appveyor.yml +++ b/appveyor.yml @@ -3,6 +3,8 @@ branches: only: - master + - appveyorTest + - /visual*/ environment: matrix: - COMPILER: "gcc" diff --git a/contrib/adaptive-compression/Makefile b/contrib/adaptive-compression/Makefile index 5a746dcd42da..c26efcd2815b 100644 --- a/contrib/adaptive-compression/Makefile +++ b/contrib/adaptive-compression/Makefile @@ -22,10 +22,10 @@ FLAGS = $(CPPFLAGS) $(CFLAGS) $(LDFLAGS) $(MULTITHREAD_LDFLAGS) all: adapt datagen -adapt: $(ZSTD_FILES) adapt.c +adapt: $(ZSTD_FILES) $(PRGDIR)/util.c adapt.c $(CC) $(FLAGS) $^ -o $@ -adapt-debug: $(ZSTD_FILES) adapt.c +adapt-debug: $(ZSTD_FILES) $(PRGDIR)/util.c adapt.c $(CC) $(FLAGS) -DDEBUG_MODE=2 $^ -o adapt datagen : $(PRGDIR)/datagen.c datagencli.c diff --git a/contrib/adaptive-compression/adapt.c b/contrib/adaptive-compression/adapt.c index 49f1d4d57530..524d1378a80f 100644 --- a/contrib/adaptive-compression/adapt.c +++ b/contrib/adaptive-compression/adapt.c @@ -579,7 +579,7 @@ static void* compressionThread(void* arg) params.cParams.windowLog = 23; { size_t const initError = ZSTD_compressBegin_advanced(ctx->cctx, job->src.start + job->dictSize - useDictSize, useDictSize, params, 0); - size_t const windowSizeError = ZSTD_CCtx_setParameter(ctx->cctx, ZSTD_p_forceMaxWindow, 1); + size_t const windowSizeError = ZSTD_CCtx_setParameter(ctx->cctx, ZSTD_c_forceMaxWindow, 1); if (ZSTD_isError(initError) || ZSTD_isError(windowSizeError)) { DISPLAY("Error: something went wrong while starting compression\n"); signalErrorToThreads(ctx); diff --git a/contrib/adaptive-compression/datagencli.c b/contrib/adaptive-compression/datagencli.c index bf9601f20976..5f3c93140493 100644 --- a/contrib/adaptive-compression/datagencli.c +++ b/contrib/adaptive-compression/datagencli.c @@ -120,7 +120,7 @@ int main(int argc, const char** argv) DISPLAYLEVEL(4, "Compressible data Generator \n"); if (probaU32!=COMPRESSIBILITY_DEFAULT) DISPLAYLEVEL(3, "Compressibility : %i%%\n", probaU32); - DISPLAYLEVEL(3, "Seed = %u \n", seed); + DISPLAYLEVEL(3, "Seed = %u \n", (unsigned)seed); RDG_genStdout(size, (double)probaU32/100, litProba, seed); DISPLAYLEVEL(1, "\n"); diff --git a/contrib/meson/README b/contrib/meson/README deleted file mode 100644 index 0b5331e6dbfc..000000000000 --- a/contrib/meson/README +++ /dev/null @@ -1,3 +0,0 @@ -This Meson project is provided with no guarantee and maintained by Dima Krasner . - -It outputs one libzstd, either shared or static, depending on default_library. diff --git a/contrib/meson/meson.build b/contrib/meson/meson.build deleted file mode 100644 index 98c9b0293007..000000000000 --- a/contrib/meson/meson.build +++ /dev/null @@ -1,144 +0,0 @@ -project('zstd', 'c', license: 'BSD') - -libm = meson.get_compiler('c').find_library('m', required: true) - -lib_dir = join_paths('..', '..', 'lib') -common_dir = join_paths(lib_dir, 'common') -compress_dir = join_paths(lib_dir, 'compress') -decompress_dir = join_paths(lib_dir, 'decompress') -dictbuilder_dir = join_paths(lib_dir, 'dictBuilder') -deprecated_dir = join_paths(lib_dir, 'deprecated') - -libzstd_srcs = [ - join_paths(common_dir, 'entropy_common.c'), - join_paths(common_dir, 'fse_decompress.c'), - join_paths(common_dir, 'threading.c'), - join_paths(common_dir, 'pool.c'), - join_paths(common_dir, 'zstd_common.c'), - join_paths(common_dir, 'error_private.c'), - join_paths(common_dir, 'xxhash.c'), - join_paths(compress_dir, 'fse_compress.c'), - join_paths(compress_dir, 'hist.c'), - join_paths(compress_dir, 'huf_compress.c'), - join_paths(compress_dir, 'zstd_compress.c'), - join_paths(compress_dir, 'zstd_fast.c'), - join_paths(compress_dir, 'zstd_double_fast.c'), - join_paths(compress_dir, 'zstd_lazy.c'), - join_paths(compress_dir, 'zstd_opt.c'), - join_paths(compress_dir, 'zstd_ldm.c'), - join_paths(compress_dir, 'zstdmt_compress.c'), - join_paths(decompress_dir, 'huf_decompress.c'), - join_paths(decompress_dir, 'zstd_decompress.c'), - join_paths(dictbuilder_dir, 'cover.c'), - join_paths(dictbuilder_dir, 'divsufsort.c'), - join_paths(dictbuilder_dir, 'zdict.c'), - join_paths(deprecated_dir, 'zbuff_common.c'), - join_paths(deprecated_dir, 'zbuff_compress.c'), - join_paths(deprecated_dir, 'zbuff_decompress.c') -] - -libzstd_includes = [include_directories(common_dir, dictbuilder_dir, compress_dir, lib_dir)] - -legacy = get_option('legacy_support') -if legacy == '0' - legacy = 'false' -endif -if legacy != 'false' - if legacy == 'true' - legacy = '1' - endif - #See ZSTD_LEGACY_SUPPORT of programs/README.md - message('Enabling legacy support back to version 0.' + legacy) - legacy_int = legacy.to_int() - if legacy_int > 7 - legacy_int = 7 - endif - libzstd_cflags = ['-DZSTD_LEGACY_SUPPORT=' + legacy] - - legacy_dir = join_paths(lib_dir, 'legacy') - libzstd_includes += [include_directories(legacy_dir)] - if legacy_int <= 1 - libzstd_srcs += join_paths(legacy_dir, 'zstd_v01.c') - endif - if legacy_int <= 2 - libzstd_srcs += join_paths(legacy_dir, 'zstd_v02.c') - endif - if legacy_int <= 3 - libzstd_srcs += join_paths(legacy_dir, 'zstd_v03.c') - endif - if legacy_int <= 4 - libzstd_srcs += join_paths(legacy_dir, 'zstd_v04.c') - endif - if legacy_int <= 5 - libzstd_srcs += join_paths(legacy_dir, 'zstd_v05.c') - endif - if legacy_int <= 6 - libzstd_srcs += join_paths(legacy_dir, 'zstd_v06.c') - endif - if legacy_int <= 7 - libzstd_srcs += join_paths(legacy_dir, 'zstd_v07.c') - endif -else - libzstd_cflags = [] -endif - -if get_option('multithread') - message('Enabling multi-threading support') - add_global_arguments('-DZSTD_MULTITHREAD', language: 'c') - libzstd_deps = [dependency('threads')] -else - libzstd_deps = [] -endif - -libzstd = library('zstd', - libzstd_srcs, - include_directories: libzstd_includes, - c_args: libzstd_cflags, - dependencies: libzstd_deps, - install: true, - soversion: '1', - ) - -programs_dir = join_paths('..', '..', 'programs') - -zstd = executable('zstd', - join_paths(programs_dir, 'bench.c'), - join_paths(programs_dir, 'datagen.c'), - join_paths(programs_dir, 'dibio.c'), - join_paths(programs_dir, 'fileio.c'), - join_paths(programs_dir, 'zstdcli.c'), - include_directories: libzstd_includes, - c_args: ['-DZSTD_NODICT', '-DZSTD_NOBENCH'], - link_with: libzstd, - install: true) - -tests_dir = join_paths('..', '..', 'tests') -datagen_c = join_paths(programs_dir, 'datagen.c') -test_includes = libzstd_includes + [include_directories(programs_dir)] - -fullbench = executable('fullbench', - datagen_c, join_paths(tests_dir, 'fullbench.c'), - include_directories: test_includes, - link_with: libzstd) -test('fullbench', fullbench) - -fuzzer = executable('fuzzer', - datagen_c, join_paths(tests_dir, 'fuzzer.c'), - include_directories: test_includes, - link_with: libzstd) -test('fuzzer', fuzzer) - -if target_machine.system() != 'windows' - paramgrill = executable('paramgrill', - datagen_c, join_paths(tests_dir, 'paramgrill.c'), - join_paths(programs_dir, 'bench.c'), - include_directories: test_includes, - link_with: libzstd, - dependencies: libm) - test('paramgrill', paramgrill) - - datagen = executable('datagen', - datagen_c, join_paths(tests_dir, 'datagencli.c'), - include_directories: test_includes, - link_with: libzstd) -endif diff --git a/contrib/meson/meson_options.txt b/contrib/meson/meson_options.txt deleted file mode 100644 index 99845c8aa513..000000000000 --- a/contrib/meson/meson_options.txt +++ /dev/null @@ -1,3 +0,0 @@ -option('multithread', type: 'boolean', value: false) -option('legacy_support', type: 'string', value: '4', - description: 'True or false, or 7 to 1 for v0.7+ to v0.1+.') diff --git a/contrib/pzstd/Makefile b/contrib/pzstd/Makefile index 14b932297714..1d434dbcd458 100644 --- a/contrib/pzstd/Makefile +++ b/contrib/pzstd/Makefile @@ -171,7 +171,7 @@ roundtripcheck: roundtrip check $(TESTPROG) ./test/RoundTripTest$(EXT) $(TESTFLAGS) # Build the main binary -pzstd$(EXT): main.o Options.o Pzstd.o SkippableFrame.o $(ZSTDDIR)/libzstd.a +pzstd$(EXT): main.o $(PROGDIR)/util.o Options.o Pzstd.o SkippableFrame.o $(ZSTDDIR)/libzstd.a $(LD_COMMAND) # Target that depends on all the tests diff --git a/contrib/seekable_format/examples/parallel_processing.c b/contrib/seekable_format/examples/parallel_processing.c index da3477632e2c..36226b49fd3c 100644 --- a/contrib/seekable_format/examples/parallel_processing.c +++ b/contrib/seekable_format/examples/parallel_processing.c @@ -148,20 +148,20 @@ static void sumFile_orDie(const char* fname, int nbThreads) size_t const initResult = ZSTD_seekable_initFile(seekable, fin); if (ZSTD_isError(initResult)) { fprintf(stderr, "ZSTD_seekable_init() error : %s \n", ZSTD_getErrorName(initResult)); exit(11); } - size_t const numFrames = ZSTD_seekable_getNumFrames(seekable); + unsigned const numFrames = ZSTD_seekable_getNumFrames(seekable); struct sum_job* jobs = (struct sum_job*)malloc(numFrames * sizeof(struct sum_job)); - size_t i; - for (i = 0; i < numFrames; i++) { - jobs[i] = (struct sum_job){ fname, 0, i, 0 }; - POOL_add(pool, sumFrame, &jobs[i]); + unsigned fnb; + for (fnb = 0; fnb < numFrames; fnb++) { + jobs[fnb] = (struct sum_job){ fname, 0, fnb, 0 }; + POOL_add(pool, sumFrame, &jobs[fnb]); } unsigned long long total = 0; - for (i = 0; i < numFrames; i++) { - while (!jobs[i].done) SLEEP(5); /* wake up every 5 milliseconds to check */ - total += jobs[i].sum; + for (fnb = 0; fnb < numFrames; fnb++) { + while (!jobs[fnb].done) SLEEP(5); /* wake up every 5 milliseconds to check */ + total += jobs[fnb].sum; } printf("Sum: %llu\n", total); diff --git a/contrib/seekable_format/zstdseek_compress.c b/contrib/seekable_format/zstdseek_compress.c index 4ab666bdbbac..5a75714fac5b 100644 --- a/contrib/seekable_format/zstdseek_compress.c +++ b/contrib/seekable_format/zstdseek_compress.c @@ -8,6 +8,8 @@ */ #include /* malloc, free */ +#include /* UINT_MAX */ +#include #define XXH_STATIC_LINKING_ONLY #define XXH_NAMESPACE ZSTD_ @@ -139,7 +141,7 @@ size_t ZSTD_seekable_freeCStream(ZSTD_seekable_CStream* zcs) size_t ZSTD_seekable_initCStream(ZSTD_seekable_CStream* zcs, int compressionLevel, int checksumFlag, - U32 maxFrameSize) + unsigned maxFrameSize) { zcs->framelog.size = 0; zcs->frameCSize = 0; @@ -167,9 +169,9 @@ size_t ZSTD_seekable_initCStream(ZSTD_seekable_CStream* zcs, } size_t ZSTD_seekable_logFrame(ZSTD_frameLog* fl, - unsigned compressedSize, - unsigned decompressedSize, - unsigned checksum) + unsigned compressedSize, + unsigned decompressedSize, + unsigned checksum) { if (fl->size == ZSTD_SEEKABLE_MAXFRAMES) return ERROR(frameIndex_tooLarge); @@ -184,7 +186,8 @@ size_t ZSTD_seekable_logFrame(ZSTD_frameLog* fl, if (newEntries == NULL) return ERROR(memory_allocation); fl->entries = newEntries; - fl->capacity = newCapacity; + assert(newCapacity <= UINT_MAX); + fl->capacity = (U32)newCapacity; } fl->entries[fl->size] = (framelogEntry_t){ @@ -268,7 +271,7 @@ size_t ZSTD_seekable_compressStream(ZSTD_seekable_CStream* zcs, ZSTD_outBuffer* static inline size_t ZSTD_seekable_seekTableSize(const ZSTD_frameLog* fl) { size_t const sizePerFrame = 8 + (fl->checksumFlag?4:0); - size_t const seekTableLen = ZSTD_skippableHeaderSize + + size_t const seekTableLen = ZSTD_SKIPPABLEHEADERSIZE + sizePerFrame * fl->size + ZSTD_seekTableFooterSize; @@ -307,32 +310,32 @@ size_t ZSTD_seekable_writeSeekTable(ZSTD_frameLog* fl, ZSTD_outBuffer* output) size_t const seekTableLen = ZSTD_seekable_seekTableSize(fl); CHECK_Z(ZSTD_stwrite32(fl, output, ZSTD_MAGIC_SKIPPABLE_START | 0xE, 0)); - CHECK_Z(ZSTD_stwrite32(fl, output, seekTableLen - ZSTD_skippableHeaderSize, - 4)); + assert(seekTableLen <= (size_t)UINT_MAX); + CHECK_Z(ZSTD_stwrite32(fl, output, (U32)seekTableLen - ZSTD_SKIPPABLEHEADERSIZE, 4)); while (fl->seekTableIndex < fl->size) { + unsigned long long const start = ZSTD_SKIPPABLEHEADERSIZE + sizePerFrame * fl->seekTableIndex; + assert(start + 8 <= UINT_MAX); CHECK_Z(ZSTD_stwrite32(fl, output, fl->entries[fl->seekTableIndex].cSize, - ZSTD_skippableHeaderSize + - sizePerFrame * fl->seekTableIndex + 0)); + (U32)start + 0)); CHECK_Z(ZSTD_stwrite32(fl, output, fl->entries[fl->seekTableIndex].dSize, - ZSTD_skippableHeaderSize + - sizePerFrame * fl->seekTableIndex + 4)); + (U32)start + 4)); if (fl->checksumFlag) { CHECK_Z(ZSTD_stwrite32( fl, output, fl->entries[fl->seekTableIndex].checksum, - ZSTD_skippableHeaderSize + - sizePerFrame * fl->seekTableIndex + 8)); + (U32)start + 8)); } fl->seekTableIndex++; } + assert(seekTableLen <= UINT_MAX); CHECK_Z(ZSTD_stwrite32(fl, output, fl->size, - seekTableLen - ZSTD_seekTableFooterSize)); + (U32)seekTableLen - ZSTD_seekTableFooterSize)); if (output->size - output->pos < 1) return seekTableLen - fl->seekTablePos; if (fl->seekTablePos < seekTableLen - 4) { @@ -345,7 +348,7 @@ size_t ZSTD_seekable_writeSeekTable(ZSTD_frameLog* fl, ZSTD_outBuffer* output) } CHECK_Z(ZSTD_stwrite32(fl, output, ZSTD_SEEKABLE_MAGICNUMBER, - seekTableLen - 4)); + (U32)seekTableLen - 4)); if (fl->seekTablePos != seekTableLen) return ERROR(GENERIC); return 0; diff --git a/contrib/seekable_format/zstdseek_decompress.c b/contrib/seekable_format/zstdseek_decompress.c index b4c48754ef87..a0206592299f 100644 --- a/contrib/seekable_format/zstdseek_decompress.c +++ b/contrib/seekable_format/zstdseek_decompress.c @@ -54,8 +54,9 @@ # define LONG_SEEK fseek #endif -#include /* malloc, free */ -#include /* FILE* */ +#include /* malloc, free */ +#include /* FILE* */ +#include /* UNIT_MAX */ #include #define XXH_STATIC_LINKING_ONLY @@ -200,13 +201,14 @@ size_t ZSTD_seekable_free(ZSTD_seekable* zs) * Performs a binary search to find the last frame with a decompressed offset * <= pos * @return : the frame's index */ -U32 ZSTD_seekable_offsetToFrameIndex(ZSTD_seekable* const zs, unsigned long long pos) +unsigned ZSTD_seekable_offsetToFrameIndex(ZSTD_seekable* const zs, unsigned long long pos) { U32 lo = 0; - U32 hi = zs->seekTable.tableLen; + U32 hi = (U32)zs->seekTable.tableLen; + assert(zs->seekTable.tableLen <= UINT_MAX); if (pos >= zs->seekTable.entries[zs->seekTable.tableLen].dOffset) { - return zs->seekTable.tableLen; + return (U32)zs->seekTable.tableLen; } while (lo + 1 < hi) { @@ -220,31 +222,32 @@ U32 ZSTD_seekable_offsetToFrameIndex(ZSTD_seekable* const zs, unsigned long long return lo; } -U32 ZSTD_seekable_getNumFrames(ZSTD_seekable* const zs) +unsigned ZSTD_seekable_getNumFrames(ZSTD_seekable* const zs) { - return zs->seekTable.tableLen; + assert(zs->seekTable.tableLen <= UINT_MAX); + return (unsigned)zs->seekTable.tableLen; } -unsigned long long ZSTD_seekable_getFrameCompressedOffset(ZSTD_seekable* const zs, U32 frameIndex) +unsigned long long ZSTD_seekable_getFrameCompressedOffset(ZSTD_seekable* const zs, unsigned frameIndex) { if (frameIndex >= zs->seekTable.tableLen) return ZSTD_SEEKABLE_FRAMEINDEX_TOOLARGE; return zs->seekTable.entries[frameIndex].cOffset; } -unsigned long long ZSTD_seekable_getFrameDecompressedOffset(ZSTD_seekable* const zs, U32 frameIndex) +unsigned long long ZSTD_seekable_getFrameDecompressedOffset(ZSTD_seekable* const zs, unsigned frameIndex) { if (frameIndex >= zs->seekTable.tableLen) return ZSTD_SEEKABLE_FRAMEINDEX_TOOLARGE; return zs->seekTable.entries[frameIndex].dOffset; } -size_t ZSTD_seekable_getFrameCompressedSize(ZSTD_seekable* const zs, U32 frameIndex) +size_t ZSTD_seekable_getFrameCompressedSize(ZSTD_seekable* const zs, unsigned frameIndex) { if (frameIndex >= zs->seekTable.tableLen) return ERROR(frameIndex_tooLarge); return zs->seekTable.entries[frameIndex + 1].cOffset - zs->seekTable.entries[frameIndex].cOffset; } -size_t ZSTD_seekable_getFrameDecompressedSize(ZSTD_seekable* const zs, U32 frameIndex) +size_t ZSTD_seekable_getFrameDecompressedSize(ZSTD_seekable* const zs, unsigned frameIndex) { if (frameIndex > zs->seekTable.tableLen) return ERROR(frameIndex_tooLarge); return zs->seekTable.entries[frameIndex + 1].dOffset - @@ -275,7 +278,7 @@ static size_t ZSTD_seekable_loadSeekTable(ZSTD_seekable* zs) { U32 const numFrames = MEM_readLE32(zs->inBuff); U32 const sizePerEntry = 8 + (checksumFlag?4:0); U32 const tableSize = sizePerEntry * numFrames; - U32 const frameSize = tableSize + ZSTD_seekTableFooterSize + ZSTD_skippableHeaderSize; + U32 const frameSize = tableSize + ZSTD_seekTableFooterSize + ZSTD_SKIPPABLEHEADERSIZE; U32 remaining = frameSize - ZSTD_seekTableFooterSize; /* don't need to re-read footer */ { @@ -290,7 +293,7 @@ static size_t ZSTD_seekable_loadSeekTable(ZSTD_seekable* zs) if (MEM_readLE32(zs->inBuff) != (ZSTD_MAGIC_SKIPPABLE_START | 0xE)) { return ERROR(prefix_unknown); } - if (MEM_readLE32(zs->inBuff+4) + ZSTD_skippableHeaderSize != frameSize) { + if (MEM_readLE32(zs->inBuff+4) + ZSTD_SKIPPABLEHEADERSIZE != frameSize) { return ERROR(prefix_unknown); } @@ -444,7 +447,7 @@ size_t ZSTD_seekable_decompress(ZSTD_seekable* zs, void* dst, size_t len, unsign return len; } -size_t ZSTD_seekable_decompressFrame(ZSTD_seekable* zs, void* dst, size_t dstSize, U32 frameIndex) +size_t ZSTD_seekable_decompressFrame(ZSTD_seekable* zs, void* dst, size_t dstSize, unsigned frameIndex) { if (frameIndex >= zs->seekTable.tableLen) { return ERROR(frameIndex_tooLarge); diff --git a/doc/zstd_compression_format.md b/doc/zstd_compression_format.md index e562e628bc9c..ed758cf598ac 100644 --- a/doc/zstd_compression_format.md +++ b/doc/zstd_compression_format.md @@ -16,7 +16,7 @@ Distribution of this document is unlimited. ### Version -0.3.0 (25/09/18) +0.3.1 (25/10/18) Introduction @@ -913,11 +913,37 @@ Note that blocks which are not `Compressed_Block` are skipped, they do not contr ###### Offset updates rules The newest offset takes the lead in offset history, -shifting others back (up to its previous place if it was already present). +shifting others back by one rank, +up to the previous rank of the new offset _if it was present in history_. + +__Examples__ : + +In the common case, when new offset is not part of history : +`Repeated_Offset3` = `Repeated_Offset2` +`Repeated_Offset2` = `Repeated_Offset1` +`Repeated_Offset1` = `NewOffset` + +When the new offset _is_ part of history, there may be specific adjustments. + +When `NewOffset` == `Repeated_Offset1`, offset history remains actually unmodified. + +When `NewOffset` == `Repeated_Offset2`, +`Repeated_Offset1` and `Repeated_Offset2` ranks are swapped. +`Repeated_Offset3` is unmodified. + +When `NewOffset` == `Repeated_Offset3`, +there is actually no difference with the common case : +all offsets are shifted by one rank, +`NewOffset` (== `Repeated_Offset3`) becomes the new `Repeated_Offset1`. + +Also worth mentioning, the specific corner case when `offset_value` == 3, +and the literal length of the current sequence is zero. +In which case , `NewOffset` = `Repeated_Offset1` - 1_byte. +Here also, from an offset history update perspective, it's just a common case : +`Repeated_Offset3` = `Repeated_Offset2` +`Repeated_Offset2` = `Repeated_Offset1` +`Repeated_Offset1` = `NewOffset` ( == `Repeated_Offset1` - 1_byte ) -This means that when `Repeated_Offset1` (most recent) is used, history is unmodified. -When `Repeated_Offset2` is used, it's swapped with `Repeated_Offset1`. -If any other offset is used, it becomes `Repeated_Offset1` and the rest are shift back by one. Skippable Frames @@ -1629,6 +1655,7 @@ or at least provide a meaningful error code explaining for which reason it canno Version changes --------------- +- 0.3.1 : minor clarification regarding offset history update rules - 0.3.0 : minor edits to match RFC8478 - 0.2.9 : clarifications for huffman weights direct representation, by Ulrich Kunitz - 0.2.8 : clarifications for IETF RFC discuss diff --git a/doc/zstd_manual.html b/doc/zstd_manual.html index f9b1daa8a28c..6dfa6d997cb9 100644 --- a/doc/zstd_manual.html +++ b/doc/zstd_manual.html @@ -1,10 +1,10 @@ -zstd 1.3.7 Manual +zstd 1.3.8 Manual -

zstd 1.3.7 Manual

+

zstd 1.3.8 Manual

Contents

    @@ -19,16 +19,19 @@
  1. Streaming compression - HowTo
  2. Streaming decompression - HowTo
  3. ADVANCED AND EXPERIMENTAL FUNCTIONS
    4. -
  4. Frame size functions
    5. -
  5. Memory management
    6. -
  6. Advanced compression functions
    7. -
  7. Advanced decompression functions
    8. -
  8. Advanced streaming functions
    9. -
  9. Buffer-less and synchronous inner streaming functions
    10. -
  10. Buffer-less streaming compression (synchronous mode)
    11. -
  11. Buffer-less streaming decompression (synchronous mode)
    12. -
  12. New advanced API (experimental)
    13. -
  13. Block level API
    14. +
  14. Candidate API for promotion to stable status
    15. +
  15. Advanced compression API
    16. +
  16. experimental API (static linking only)
    17. +
  17. Frame size functions
    18. +
  18. Memory management
    19. +
  19. Advanced compression functions
    20. +
  20. Advanced decompression functions
    21. +
  21. Advanced streaming functions
    22. +
  22. Buffer-less and synchronous inner streaming functions
    23. +
  23. Buffer-less streaming compression (synchronous mode)
    24. +
  24. Buffer-less streaming decompression (synchronous mode)
    25. +
  25. ZSTD_getFrameHeader() :
    26. +
  26. Block level API

Introduction

@@ -64,7 +67,7 @@
 
 
Version

 
-
unsigned ZSTD_versionNumber(void);   /**< useful to check dll version */
+
unsigned ZSTD_versionNumber(void);   /**< to check runtime library version */
 


 
Default constant

 
@@ -139,11 +142,13 @@ int         ZSTD_maxCLevel(void);               /*!< maximum compression lev
 ZSTD_CCtx* ZSTD_createCCtx(void);
 size_t     ZSTD_freeCCtx(ZSTD_CCtx* cctx);
 


-
size_t ZSTD_compressCCtx(ZSTD_CCtx* ctx,
+
size_t ZSTD_compressCCtx(ZSTD_CCtx* cctx,
                          void* dst, size_t dstCapacity,
                    const void* src, size_t srcSize,
                          int compressionLevel);
-
  Same as ZSTD_compress(), requires an allocated ZSTD_CCtx (see ZSTD_createCCtx()). 
+
  Same as ZSTD_compress(), using an explicit ZSTD_CCtx
+  The function will compress at requested compression level,
+  ignoring any other parameter 
 


 
 
Decompression context
  When decompressing many times,
@@ -155,10 +160,13 @@ size_t     ZSTD_freeCCtx(ZSTD_CCtx* cctx);
 ZSTD_DCtx* ZSTD_createDCtx(void);
 size_t     ZSTD_freeDCtx(ZSTD_DCtx* dctx);
 


-
size_t ZSTD_decompressDCtx(ZSTD_DCtx* ctx,
+
size_t ZSTD_decompressDCtx(ZSTD_DCtx* dctx,
                            void* dst, size_t dstCapacity,
                      const void* src, size_t srcSize);
-
  Same as ZSTD_decompress(), requires an allocated ZSTD_DCtx (see ZSTD_createDCtx()) 
+
  Same as ZSTD_decompress(),
+  requires an allocated ZSTD_DCtx.
+  Compatible with sticky parameters.
+ 
 


 
 
Simple dictionary API

@@ -168,18 +176,22 @@ size_t     ZSTD_freeDCtx(ZSTD_DCtx* dctx);
                          const void* src, size_t srcSize,
                          const void* dict,size_t dictSize,
                                int compressionLevel);
-
  Compression using a predefined Dictionary (see dictBuilder/zdict.h).
+
  Compression at an explicit compression level using a Dictionary.
+  A dictionary can be any arbitrary data segment (also called a prefix),
+  or a buffer with specified information (see dictBuilder/zdict.h).
   Note : This function loads the dictionary, resulting in significant startup delay.
-  Note : When `dict == NULL || dictSize < 8` no dictionary is used. 
+         It's intended for a dictionary used only once.
+  Note 2 : When `dict == NULL || dictSize < 8` no dictionary is used. 
 


 
 
size_t ZSTD_decompress_usingDict(ZSTD_DCtx* dctx,
                                  void* dst, size_t dstCapacity,
                            const void* src, size_t srcSize,
                            const void* dict,size_t dictSize);
-
  Decompression using a predefined Dictionary (see dictBuilder/zdict.h).
+
  Decompression using a known Dictionary.
   Dictionary must be identical to the one used during compression.
   Note : This function loads the dictionary, resulting in significant startup delay.
+         It's intended for a dictionary used only once.
   Note : When `dict == NULL || dictSize < 8` no dictionary is used. 
 


 
@@ -187,11 +199,12 @@ size_t     ZSTD_freeDCtx(ZSTD_DCtx* dctx);
 
 
ZSTD_CDict* ZSTD_createCDict(const void* dictBuffer, size_t dictSize,
                              int compressionLevel);
-
  When compressing multiple messages / blocks with the same dictionary, it's recommended to load it just once.
-  ZSTD_createCDict() will create a digested dictionary, ready to start future compression operations without startup delay.
+
  When compressing multiple messages / blocks using the same dictionary, it's recommended to load it only once.
+  ZSTD_createCDict() will create a digested dictionary, ready to start future compression operations without startup cost.
   ZSTD_CDict can be created once and shared by multiple threads concurrently, since its usage is read-only.
-  `dictBuffer` can be released after ZSTD_CDict creation, since its content is copied within CDict
-  Note : A ZSTD_CDict can be created with an empty dictionary, but it is inefficient for small data. 
+ `dictBuffer` can be released after ZSTD_CDict creation, because its content is copied within CDict.
+  Consider experimental function `ZSTD_createCDict_byReference()` if you prefer to not duplicate `dictBuffer` content.
+  Note : A ZSTD_CDict can be created from an empty dictBuffer, but it is inefficient when used to compress small data. 
 


 
 
size_t      ZSTD_freeCDict(ZSTD_CDict* CDict);
@@ -203,16 +216,14 @@ size_t     ZSTD_freeDCtx(ZSTD_DCtx* dctx);
                           const void* src, size_t srcSize,
                           const ZSTD_CDict* cdict);
 
  Compression using a digested Dictionary.
-  Faster startup than ZSTD_compress_usingDict(), recommended when same dictionary is used multiple times.
-  Note that compression level is decided during dictionary creation.
-  Frame parameters are hardcoded (dictID=yes, contentSize=yes, checksum=no)
-  Note : ZSTD_compress_usingCDict() can be used with a ZSTD_CDict created from an empty dictionary.
-         But it is inefficient for small data, and it is recommended to use ZSTD_compressCCtx(). 
+  Recommended when same dictionary is used multiple times.
+  Note : compression level is _decided at dictionary creation time_,
+     and frame parameters are hardcoded (dictID=yes, contentSize=yes, checksum=no) 
 


 
 
ZSTD_DDict* ZSTD_createDDict(const void* dictBuffer, size_t dictSize);
 
  Create a digested dictionary, ready to start decompression operation without startup delay.
-  dictBuffer can be released after DDict creation, as its content is copied inside DDict 
+  dictBuffer can be released after DDict creation, as its content is copied inside DDict. 
 


 
 
size_t      ZSTD_freeDDict(ZSTD_DDict* ddict);
@@ -224,7 +235,7 @@ size_t     ZSTD_freeDCtx(ZSTD_DCtx* dctx);
                             const void* src, size_t srcSize,
                             const ZSTD_DDict* ddict);
 
  Decompression using a digested Dictionary.
-  Faster startup than ZSTD_decompress_usingDict(), recommended when same dictionary is used multiple times. 
+  Recommended when same dictionary is used multiple times. 
 


 
 
Streaming

@@ -245,13 +256,17 @@ size_t     ZSTD_freeDCtx(ZSTD_DCtx* dctx);
   A ZSTD_CStream object is required to track streaming operation.
   Use ZSTD_createCStream() and ZSTD_freeCStream() to create/release resources.
   ZSTD_CStream objects can be reused multiple times on consecutive compression operations.
-  It is recommended to re-use ZSTD_CStream in situations where many streaming operations will be achieved consecutively,
-  since it will play nicer with system's memory, by re-using already allocated memory.
-  Use one separate ZSTD_CStream per thread for parallel execution.
+  It is recommended to re-use ZSTD_CStream since it will play nicer with system's memory, by re-using already allocated memory.
 
-  Start a new compression by initializing ZSTD_CStream context.
-  Use ZSTD_initCStream() to start a new compression operation.
-  Use variants ZSTD_initCStream_usingDict() or ZSTD_initCStream_usingCDict() for streaming with dictionary (experimental section)
+  For parallel execution, use one separate ZSTD_CStream per thread.
+
+  note : since v1.3.0, ZSTD_CStream and ZSTD_CCtx are the same thing.
+
+  Parameters are sticky : when starting a new compression on the same context,
+  it will re-use the same sticky parameters as previous compression session.
+  When in doubt, it's recommended to fully initialize the context before usage.
+  Use ZSTD_initCStream() to set the parameter to a selected compression level.
+  Use advanced API (ZSTD_CCtx_setParameter(), etc.) to set more specific parameters.
 
   Use ZSTD_compressStream() as many times as necessary to consume input stream.
   The function will automatically update both `pos` fields within `input` and `output`.
@@ -260,12 +275,11 @@ size_t     ZSTD_freeDCtx(ZSTD_DCtx* dctx);
   in which case `input.pos < input.size`.
   The caller must check if input has been entirely consumed.
   If not, the caller must make some room to receive more compressed data,
-  typically by emptying output buffer, or allocating a new output buffer,
   and then present again remaining input data.
-  @return : a size hint, preferred nb of bytes to use as input for next function call
-            or an error code, which can be tested using ZSTD_isError().
-            Note 1 : it's just a hint, to help latency a little, any other value will work fine.
-            Note 2 : size hint is guaranteed to be <= ZSTD_CStreamInSize()
+ @return : a size hint, preferred nb of bytes to use as input for next function call
+           or an error code, which can be tested using ZSTD_isError().
+           Note 1 : it's just a hint, to help latency a little, any value will work fine.
+           Note 2 : size hint is guaranteed to be <= ZSTD_CStreamInSize()
 
   At any moment, it's possible to flush whatever data might remain stuck within internal buffer,
   using ZSTD_flushStream(). `output->pos` will be updated.
@@ -305,25 +319,24 @@ size_t ZSTD_endStream(ZSTD_CStream* zcs, ZSTD_outBuffer* output);
   Use ZSTD_createDStream() and ZSTD_freeDStream() to create/release resources.
   ZSTD_DStream objects can be re-used multiple times.
 
-  Use ZSTD_initDStream() to start a new decompression operation,
-   or ZSTD_initDStream_usingDict() if decompression requires a dictionary.
-   @return : recommended first input size
+  Use ZSTD_initDStream() to start a new decompression operation.
+ @return : recommended first input size
+  Alternatively, use advanced API to set specific properties.
 
   Use ZSTD_decompressStream() repetitively to consume your input.
   The function will update both `pos` fields.
   If `input.pos < input.size`, some input has not been consumed.
   It's up to the caller to present again remaining data.
-  The function tries to flush all data decoded immediately, repecting buffer sizes.
+  The function tries to flush all data decoded immediately, respecting output buffer size.
   If `output.pos < output.size`, decoder has flushed everything it could.
-  But if `output.pos == output.size`, there is no such guarantee,
-  it's likely that some decoded data was not flushed and still remains within internal buffers.
+  But if `output.pos == output.size`, there might be some data left within internal buffers.,
   In which case, call ZSTD_decompressStream() again to flush whatever remains in the buffer.
-  When no additional input is provided, amount of data flushed is necessarily <= ZSTD_BLOCKSIZE_MAX.
+  Note : with no additional input provided, amount of data flushed is necessarily <= ZSTD_BLOCKSIZE_MAX.
  @return : 0 when a frame is completely decoded and fully flushed,
         or an error code, which can be tested using ZSTD_isError(),
         or any other value > 0, which means there is still some decoding or flushing to do to complete current frame :
-                                the return value is a suggested next input size (a hint for better latency)
-                                that will never load more than the current frame.
+                                the return value is a suggested next input size (just a hint for better latency)
+                                that will never request more than the remaining frame size.
  
 

 
@@ -340,32 +353,477 @@ size_t ZSTD_decompressStream(ZSTD_DStream* zds, ZSTD_outBuffer* output, ZSTD_inB
 
size_t ZSTD_DStreamOutSize(void);   /*!< recommended size for output buffer. Guarantee to successfully flush at least one complete block in all circumstances. */
 


 
ADVANCED AND EXPERIMENTAL FUNCTIONS
- The definitions in this section are considered experimental.
- They should never be used with a dynamic library, as prototypes may change in the future.
+ The definitions in the following section are considered experimental.
  They are provided for advanced scenarios.
+ They should never be used with a dynamic library, as prototypes may change in the future.
  Use them only in association with static linking.
  
 

 
+
Candidate API for promotion to stable status
+ The following symbols and constants form the "staging area" :
+ they are considered to join "stable API" by v1.4.0.
+ The proposal is written so that it can be made stable "as is",
+ though it's still possible to suggest improvements.
+ Staging is in fact last chance for changes,
+ the API is locked once reaching "stable" status.
+ 
+

+
 
int ZSTD_minCLevel(void);  /*!< minimum negative compression level allowed */
 


-
typedef enum { ZSTD_fast=1, ZSTD_dfast, ZSTD_greedy, ZSTD_lazy, ZSTD_lazy2,
-               ZSTD_btlazy2, ZSTD_btopt, ZSTD_btultra } ZSTD_strategy;   /* from faster to stronger */
+
size_t ZSTD_findFrameCompressedSize(const void* src, size_t srcSize);
+
 `src` should point to the start of a ZSTD frame or skippable frame.
+ `srcSize` must be >= first frame size
+ @return : the compressed size of the first frame starting at `src`,
+           suitable to pass as `srcSize` to `ZSTD_decompress` or similar,
+        or an error code if input is invalid 
+


+
+
size_t ZSTD_sizeof_CCtx(const ZSTD_CCtx* cctx);
+size_t ZSTD_sizeof_DCtx(const ZSTD_DCtx* dctx);
+size_t ZSTD_sizeof_CStream(const ZSTD_CStream* zcs);
+size_t ZSTD_sizeof_DStream(const ZSTD_DStream* zds);
+size_t ZSTD_sizeof_CDict(const ZSTD_CDict* cdict);
+size_t ZSTD_sizeof_DDict(const ZSTD_DDict* ddict);
+
  These functions give the _current_ memory usage of selected object.
+  Note that object memory usage can evolve (increase or decrease) over time. 
+


+
+
Advanced compression API

+
+
typedef enum { ZSTD_fast=1,
+               ZSTD_dfast=2,
+               ZSTD_greedy=3,
+               ZSTD_lazy=4,
+               ZSTD_lazy2=5,
+               ZSTD_btlazy2=6,
+               ZSTD_btopt=7,
+               ZSTD_btultra=8,
+               ZSTD_btultra2=9
+               /* note : new strategies _might_ be added in the future.
+                         Only the order (from fast to strong) is guaranteed */
+} ZSTD_strategy;
+


+
typedef enum {
+
+    /* compression parameters */
+    ZSTD_c_compressionLevel=100, /* Update all compression parameters according to pre-defined cLevel table
+                              * Default level is ZSTD_CLEVEL_DEFAULT==3.
+                              * Special: value 0 means default, which is controlled by ZSTD_CLEVEL_DEFAULT.
+                              * Note 1 : it's possible to pass a negative compression level.
+                              * Note 2 : setting a level sets all default values of other compression parameters */
+    ZSTD_c_windowLog=101,    /* Maximum allowed back-reference distance, expressed as power of 2.
+                              * Must be clamped between ZSTD_WINDOWLOG_MIN and ZSTD_WINDOWLOG_MAX.
+                              * Special: value 0 means "use default windowLog".
+                              * Note: Using a windowLog greater than ZSTD_WINDOWLOG_LIMIT_DEFAULT
+                              *       requires explicitly allowing such window size at decompression stage if using streaming. */
+    ZSTD_c_hashLog=102,      /* Size of the initial probe table, as a power of 2.
+                              * Resulting memory usage is (1 << (hashLog+2)).
+                              * Must be clamped between ZSTD_HASHLOG_MIN and ZSTD_HASHLOG_MAX.
+                              * Larger tables improve compression ratio of strategies <= dFast,
+                              * and improve speed of strategies > dFast.
+                              * Special: value 0 means "use default hashLog". */
+    ZSTD_c_chainLog=103,     /* Size of the multi-probe search table, as a power of 2.
+                              * Resulting memory usage is (1 << (chainLog+2)).
+                              * Must be clamped between ZSTD_CHAINLOG_MIN and ZSTD_CHAINLOG_MAX.
+                              * Larger tables result in better and slower compression.
+                              * This parameter is useless when using "fast" strategy.
+                              * It's still useful when using "dfast" strategy,
+                              * in which case it defines a secondary probe table.
+                              * Special: value 0 means "use default chainLog". */
+    ZSTD_c_searchLog=104,    /* Number of search attempts, as a power of 2.
+                              * More attempts result in better and slower compression.
+                              * This parameter is useless when using "fast" and "dFast" strategies.
+                              * Special: value 0 means "use default searchLog". */
+    ZSTD_c_minMatch=105,     /* Minimum size of searched matches.
+                              * Note that Zstandard can still find matches of smaller size,
+                              * it just tweaks its search algorithm to look for this size and larger.
+                              * Larger values increase compression and decompression speed, but decrease ratio.
+                              * Must be clamped between ZSTD_MINMATCH_MIN and ZSTD_MINMATCH_MAX.
+                              * Note that currently, for all strategies < btopt, effective minimum is 4.
+                              *                    , for all strategies > fast, effective maximum is 6.
+                              * Special: value 0 means "use default minMatchLength". */
+    ZSTD_c_targetLength=106, /* Impact of this field depends on strategy.
+                              * For strategies btopt, btultra & btultra2:
+                              *     Length of Match considered "good enough" to stop search.
+                              *     Larger values make compression stronger, and slower.
+                              * For strategy fast:
+                              *     Distance between match sampling.
+                              *     Larger values make compression faster, and weaker.
+                              * Special: value 0 means "use default targetLength". */
+    ZSTD_c_strategy=107,     /* See ZSTD_strategy enum definition.
+                              * The higher the value of selected strategy, the more complex it is,
+                              * resulting in stronger and slower compression.
+                              * Special: value 0 means "use default strategy". */
+
+    /* LDM mode parameters */
+    ZSTD_c_enableLongDistanceMatching=160, /* Enable long distance matching.
+                                     * This parameter is designed to improve compression ratio
+                                     * for large inputs, by finding large matches at long distance.
+                                     * It increases memory usage and window size.
+                                     * Note: enabling this parameter increases default ZSTD_c_windowLog to 128 MB
+                                     * except when expressly set to a different value. */
+    ZSTD_c_ldmHashLog=161,   /* Size of the table for long distance matching, as a power of 2.
+                              * Larger values increase memory usage and compression ratio,
+                              * but decrease compression speed.
+                              * Must be clamped between ZSTD_HASHLOG_MIN and ZSTD_HASHLOG_MAX
+                              * default: windowlog - 7.
+                              * Special: value 0 means "automatically determine hashlog". */
+    ZSTD_c_ldmMinMatch=162,  /* Minimum match size for long distance matcher.
+                              * Larger/too small values usually decrease compression ratio.
+                              * Must be clamped between ZSTD_LDM_MINMATCH_MIN and ZSTD_LDM_MINMATCH_MAX.
+                              * Special: value 0 means "use default value" (default: 64). */
+    ZSTD_c_ldmBucketSizeLog=163, /* Log size of each bucket in the LDM hash table for collision resolution.
+                              * Larger values improve collision resolution but decrease compression speed.
+                              * The maximum value is ZSTD_LDM_BUCKETSIZELOG_MAX.
+                              * Special: value 0 means "use default value" (default: 3). */
+    ZSTD_c_ldmHashRateLog=164, /* Frequency of inserting/looking up entries into the LDM hash table.
+                              * Must be clamped between 0 and (ZSTD_WINDOWLOG_MAX - ZSTD_HASHLOG_MIN).
+                              * Default is MAX(0, (windowLog - ldmHashLog)), optimizing hash table usage.
+                              * Larger values improve compression speed.
+                              * Deviating far from default value will likely result in a compression ratio decrease.
+                              * Special: value 0 means "automatically determine hashRateLog". */
+
+    /* frame parameters */
+    ZSTD_c_contentSizeFlag=200, /* Content size will be written into frame header _whenever known_ (default:1)
+                              * Content size must be known at the beginning of compression.
+                              * This is automatically the case when using ZSTD_compress2(),
+                              * For streaming variants, content size must be provided with ZSTD_CCtx_setPledgedSrcSize() */
+    ZSTD_c_checksumFlag=201, /* A 32-bits checksum of content is written at end of frame (default:0) */
+    ZSTD_c_dictIDFlag=202,   /* When applicable, dictionary's ID is written into frame header (default:1) */
+
+    /* multi-threading parameters */
+    /* These parameters are only useful if multi-threading is enabled (compiled with build macro ZSTD_MULTITHREAD).
+     * They return an error otherwise. */
+    ZSTD_c_nbWorkers=400,    /* Select how many threads will be spawned to compress in parallel.
+                              * When nbWorkers >= 1, triggers asynchronous mode when used with ZSTD_compressStream*() :
+                              * ZSTD_compressStream*() consumes input and flush output if possible, but immediately gives back control to caller,
+                              * while compression work is performed in parallel, within worker threads.
+                              * (note : a strong exception to this rule is when first invocation of ZSTD_compressStream2() sets ZSTD_e_end :
+                              *  in which case, ZSTD_compressStream2() delegates to ZSTD_compress2(), which is always a blocking call).
+                              * More workers improve speed, but also increase memory usage.
+                              * Default value is `0`, aka "single-threaded mode" : no worker is spawned, compression is performed inside Caller's thread, all invocations are blocking */
+    ZSTD_c_jobSize=401,      /* Size of a compression job. This value is enforced only when nbWorkers >= 1.
+                              * Each compression job is completed in parallel, so this value can indirectly impact the nb of active threads.
+                              * 0 means default, which is dynamically determined based on compression parameters.
+                              * Job size must be a minimum of overlap size, or 1 MB, whichever is largest.
+                              * The minimum size is automatically and transparently enforced */
+    ZSTD_c_overlapLog=402,   /* Control the overlap size, as a fraction of window size.
+                              * The overlap size is an amount of data reloaded from previous job at the beginning of a new job.
+                              * It helps preserve compression ratio, while each job is compressed in parallel.
+                              * This value is enforced only when nbWorkers >= 1.
+                              * Larger values increase compression ratio, but decrease speed.
+                              * Possible values range from 0 to 9 :
+                              * - 0 means "default" : value will be determined by the library, depending on strategy
+                              * - 1 means "no overlap"
+                              * - 9 means "full overlap", using a full window size.
+                              * Each intermediate rank increases/decreases load size by a factor 2 :
+                              * 9: full window;  8: w/2;  7: w/4;  6: w/8;  5:w/16;  4: w/32;  3:w/64;  2:w/128;  1:no overlap;  0:default
+                              * default value varies between 6 and 9, depending on strategy */
+
+    /* note : additional experimental parameters are also available
+     * within the experimental section of the API.
+     * At the time of this writing, they include :
+     * ZSTD_c_rsyncable
+     * ZSTD_c_format
+     * ZSTD_c_forceMaxWindow
+     * ZSTD_c_forceAttachDict
+     * Because they are not stable, it's necessary to define ZSTD_STATIC_LINKING_ONLY to access them.
+     * note : never ever use experimentalParam? names directly;
+     *        also, the enums values themselves are unstable and can still change.
+     */
+     ZSTD_c_experimentalParam1=500,
+     ZSTD_c_experimentalParam2=10,
+     ZSTD_c_experimentalParam3=1000,
+     ZSTD_c_experimentalParam4=1001
+} ZSTD_cParameter;
 


 
typedef struct {
-    unsigned windowLog;      /**< largest match distance : larger == more compression, more memory needed during decompression */
-    unsigned chainLog;       /**< fully searched segment : larger == more compression, slower, more memory (useless for fast) */
-    unsigned hashLog;        /**< dispatch table : larger == faster, more memory */
-    unsigned searchLog;      /**< nb of searches : larger == more compression, slower */
-    unsigned searchLength;   /**< match length searched : larger == faster decompression, sometimes less compression */
-    unsigned targetLength;   /**< acceptable match size for optimal parser (only) : larger == more compression, slower */
-    ZSTD_strategy strategy;
+    size_t error;
+    int lowerBound;
+    int upperBound;
+} ZSTD_bounds;
+


+
ZSTD_bounds ZSTD_cParam_getBounds(ZSTD_cParameter cParam);
+
  All parameters must belong to an interval with lower and upper bounds,
+  otherwise they will either trigger an error or be automatically clamped.
+ @return : a structure, ZSTD_bounds, which contains
+         - an error status field, which must be tested using ZSTD_isError()
+         - lower and upper bounds, both inclusive
+ 
+


+
+
size_t ZSTD_CCtx_setParameter(ZSTD_CCtx* cctx, ZSTD_cParameter param, int value);
+
  Set one compression parameter, selected by enum ZSTD_cParameter.
+  All parameters have valid bounds. Bounds can be queried using ZSTD_cParam_getBounds().
+  Providing a value beyond bound will either clamp it, or trigger an error (depending on parameter).
+  Setting a parameter is generally only possible during frame initialization (before starting compression).
+  Exception : when using multi-threading mode (nbWorkers >= 1),
+              the following parameters can be updated _during_ compression (within same frame):
+              => compressionLevel, hashLog, chainLog, searchLog, minMatch, targetLength and strategy.
+              new parameters will be active for next job only (after a flush()).
+ @return : an error code (which can be tested using ZSTD_isError()).
+ 
+


+
+
size_t ZSTD_CCtx_setPledgedSrcSize(ZSTD_CCtx* cctx, unsigned long long pledgedSrcSize);
+
  Total input data size to be compressed as a single frame.
+  Value will be written in frame header, unless if explicitly forbidden using ZSTD_c_contentSizeFlag.
+  This value will also be controlled at end of frame, and trigger an error if not respected.
+ @result : 0, or an error code (which can be tested with ZSTD_isError()).
+  Note 1 : pledgedSrcSize==0 actually means zero, aka an empty frame.
+           In order to mean "unknown content size", pass constant ZSTD_CONTENTSIZE_UNKNOWN.
+           ZSTD_CONTENTSIZE_UNKNOWN is default value for any new frame.
+  Note 2 : pledgedSrcSize is only valid once, for the next frame.
+           It's discarded at the end of the frame, and replaced by ZSTD_CONTENTSIZE_UNKNOWN.
+  Note 3 : Whenever all input data is provided and consumed in a single round,
+           for example with ZSTD_compress2(),
+           or invoking immediately ZSTD_compressStream2(,,,ZSTD_e_end),
+           this value is automatically overriden by srcSize instead.
+ 
+


+
+
size_t ZSTD_CCtx_loadDictionary(ZSTD_CCtx* cctx, const void* dict, size_t dictSize);
+
  Create an internal CDict from `dict` buffer.
+  Decompression will have to use same dictionary.
+ @result : 0, or an error code (which can be tested with ZSTD_isError()).
+  Special: Loading a NULL (or 0-size) dictionary invalidates previous dictionary,
+           meaning "return to no-dictionary mode".
+  Note 1 : Dictionary is sticky, it will be used for all future compressed frames.
+           To return to "no-dictionary" situation, load a NULL dictionary (or reset parameters).
+  Note 2 : Loading a dictionary involves building tables.
+           It's also a CPU consuming operation, with non-negligible impact on latency.
+           Tables are dependent on compression parameters, and for this reason,
+           compression parameters can no longer be changed after loading a dictionary.
+  Note 3 :`dict` content will be copied internally.
+           Use experimental ZSTD_CCtx_loadDictionary_byReference() to reference content instead.
+           In such a case, dictionary buffer must outlive its users.
+  Note 4 : Use ZSTD_CCtx_loadDictionary_advanced()
+           to precisely select how dictionary content must be interpreted. 
+


+
+
size_t ZSTD_CCtx_refCDict(ZSTD_CCtx* cctx, const ZSTD_CDict* cdict);
+
  Reference a prepared dictionary, to be used for all next compressed frames.
+  Note that compression parameters are enforced from within CDict,
+  and supercede any compression parameter previously set within CCtx.
+  The dictionary will remain valid for future compressed frames using same CCtx.
+ @result : 0, or an error code (which can be tested with ZSTD_isError()).
+  Special : Referencing a NULL CDict means "return to no-dictionary mode".
+  Note 1 : Currently, only one dictionary can be managed.
+           Referencing a new dictionary effectively "discards" any previous one.
+  Note 2 : CDict is just referenced, its lifetime must outlive its usage within CCtx. 
+


+
+
size_t ZSTD_CCtx_refPrefix(ZSTD_CCtx* cctx,
+                     const void* prefix, size_t prefixSize);
+
  Reference a prefix (single-usage dictionary) for next compressed frame.
+  A prefix is **only used once**. Tables are discarded at end of frame (ZSTD_e_end).
+  Decompression will need same prefix to properly regenerate data.
+  Compressing with a prefix is similar in outcome as performing a diff and compressing it,
+  but performs much faster, especially during decompression (compression speed is tunable with compression level).
+ @result : 0, or an error code (which can be tested with ZSTD_isError()).
+  Special: Adding any prefix (including NULL) invalidates any previous prefix or dictionary
+  Note 1 : Prefix buffer is referenced. It **must** outlive compression.
+           Its content must remain unmodified during compression.
+  Note 2 : If the intention is to diff some large src data blob with some prior version of itself,
+           ensure that the window size is large enough to contain the entire source.
+           See ZSTD_c_windowLog.
+  Note 3 : Referencing a prefix involves building tables, which are dependent on compression parameters.
+           It's a CPU consuming operation, with non-negligible impact on latency.
+           If there is a need to use the same prefix multiple times, consider loadDictionary instead.
+  Note 4 : By default, the prefix is interpreted as raw content (ZSTD_dm_rawContent).
+           Use experimental ZSTD_CCtx_refPrefix_advanced() to alter dictionary interpretation. 
+


+
+
typedef enum {
+    ZSTD_reset_session_only = 1,
+    ZSTD_reset_parameters = 2,
+    ZSTD_reset_session_and_parameters = 3
+} ZSTD_ResetDirective;
+


+
size_t ZSTD_CCtx_reset(ZSTD_CCtx* cctx, ZSTD_ResetDirective reset);
+
  There are 2 different things that can be reset, independently or jointly :
+  - The session : will stop compressing current frame, and make CCtx ready to start a new one.
+                  Useful after an error, or to interrupt any ongoing compression.
+                  Any internal data not yet flushed is cancelled.
+                  Compression parameters and dictionary remain unchanged.
+                  They will be used to compress next frame.
+                  Resetting session never fails.
+  - The parameters : changes all parameters back to "default".
+                  This removes any reference to any dictionary too.
+                  Parameters can only be changed between 2 sessions (i.e. no compression is currently ongoing)
+                  otherwise the reset fails, and function returns an error value (which can be tested using ZSTD_isError())
+  - Both : similar to resetting the session, followed by resetting parameters.
+ 
+


+
+
size_t ZSTD_compress2( ZSTD_CCtx* cctx,
+                       void* dst, size_t dstCapacity,
+                 const void* src, size_t srcSize);
+
  Behave the same as ZSTD_compressCCtx(), but compression parameters are set using the advanced API.
+  ZSTD_compress2() always starts a new frame.
+  Should cctx hold data from a previously unfinished frame, everything about it is forgotten.
+  - Compression parameters are pushed into CCtx before starting compression, using ZSTD_CCtx_set*()
+  - The function is always blocking, returns when compression is completed.
+  Hint : compression runs faster if `dstCapacity` >=  `ZSTD_compressBound(srcSize)`.
+ @return : compressed size written into `dst` (<= `dstCapacity),
+           or an error code if it fails (which can be tested using ZSTD_isError()).
+ 
+


+
+
typedef enum {
+    ZSTD_e_continue=0, /* collect more data, encoder decides when to output compressed result, for optimal compression ratio */
+    ZSTD_e_flush=1,    /* flush any data provided so far,
+                        * it creates (at least) one new block, that can be decoded immediately on reception;
+                        * frame will continue: any future data can still reference previously compressed data, improving compression. */
+    ZSTD_e_end=2       /* flush any remaining data _and_ close current frame.
+                        * note that frame is only closed after compressed data is fully flushed (return value == 0).
+                        * After that point, any additional data starts a new frame.
+                        * note : each frame is independent (does not reference any content from previous frame). */
+} ZSTD_EndDirective;
+


+
size_t ZSTD_compressStream2( ZSTD_CCtx* cctx,
+                             ZSTD_outBuffer* output,
+                             ZSTD_inBuffer* input,
+                             ZSTD_EndDirective endOp);
+
  Behaves about the same as ZSTD_compressStream, with additional control on end directive.
+  - Compression parameters are pushed into CCtx before starting compression, using ZSTD_CCtx_set*()
+  - Compression parameters cannot be changed once compression is started (save a list of exceptions in multi-threading mode)
+  - outpot->pos must be <= dstCapacity, input->pos must be <= srcSize
+  - outpot->pos and input->pos will be updated. They are guaranteed to remain below their respective limit.
+  - When nbWorkers==0 (default), function is blocking : it completes its job before returning to caller.
+  - When nbWorkers>=1, function is non-blocking : it just acquires a copy of input, and distributes jobs to internal worker threads, flush whatever is available,
+                                                  and then immediately returns, just indicating that there is some data remaining to be flushed.
+                                                  The function nonetheless guarantees forward progress : it will return only after it reads or write at least 1+ byte.
+  - Exception : if the first call requests a ZSTD_e_end directive and provides enough dstCapacity, the function delegates to ZSTD_compress2() which is always blocking.
+  - @return provides a minimum amount of data remaining to be flushed from internal buffers
+            or an error code, which can be tested using ZSTD_isError().
+            if @return != 0, flush is not fully completed, there is still some data left within internal buffers.
+            This is useful for ZSTD_e_flush, since in this case more flushes are necessary to empty all buffers.
+            For ZSTD_e_end, @return == 0 when internal buffers are fully flushed and frame is completed.
+  - after a ZSTD_e_end directive, if internal buffer is not fully flushed (@return != 0),
+            only ZSTD_e_end or ZSTD_e_flush operations are allowed.
+            Before starting a new compression job, or changing compression parameters,
+            it is required to fully flush internal buffers.
+ 
+


+
+
typedef enum {
+
+    ZSTD_d_windowLogMax=100, /* Select a size limit (in power of 2) beyond which
+                              * the streaming API will refuse to allocate memory buffer
+                              * in order to protect the host from unreasonable memory requirements.
+                              * This parameter is only useful in streaming mode, since no internal buffer is allocated in single-pass mode.
+                              * By default, a decompression context accepts window sizes <= (1 << ZSTD_WINDOWLOG_LIMIT_DEFAULT) */
+
+    /* note : additional experimental parameters are also available
+     * within the experimental section of the API.
+     * At the time of this writing, they include :
+     * ZSTD_c_format
+     * Because they are not stable, it's necessary to define ZSTD_STATIC_LINKING_ONLY to access them.
+     * note : never ever use experimentalParam? names directly
+     */
+     ZSTD_d_experimentalParam1=1000
+
+} ZSTD_dParameter;
+


+
ZSTD_bounds ZSTD_dParam_getBounds(ZSTD_dParameter dParam);
+
  All parameters must belong to an interval with lower and upper bounds,
+  otherwise they will either trigger an error or be automatically clamped.
+ @return : a structure, ZSTD_bounds, which contains
+         - an error status field, which must be tested using ZSTD_isError()
+         - both lower and upper bounds, inclusive
+ 
+


+
+
size_t ZSTD_DCtx_setParameter(ZSTD_DCtx* dctx, ZSTD_dParameter param, int value);
+
  Set one compression parameter, selected by enum ZSTD_dParameter.
+  All parameters have valid bounds. Bounds can be queried using ZSTD_dParam_getBounds().
+  Providing a value beyond bound will either clamp it, or trigger an error (depending on parameter).
+  Setting a parameter is only possible during frame initialization (before starting decompression).
+ @return : 0, or an error code (which can be tested using ZSTD_isError()).
+ 
+


+
+
size_t ZSTD_DCtx_loadDictionary(ZSTD_DCtx* dctx, const void* dict, size_t dictSize);
+
  Create an internal DDict from dict buffer,
+  to be used to decompress next frames.
+  The dictionary remains valid for all future frames, until explicitly invalidated.
+ @result : 0, or an error code (which can be tested with ZSTD_isError()).
+  Special : Adding a NULL (or 0-size) dictionary invalidates any previous dictionary,
+            meaning "return to no-dictionary mode".
+  Note 1 : Loading a dictionary involves building tables,
+           which has a non-negligible impact on CPU usage and latency.
+           It's recommended to "load once, use many times", to amortize the cost
+  Note 2 :`dict` content will be copied internally, so `dict` can be released after loading.
+           Use ZSTD_DCtx_loadDictionary_byReference() to reference dictionary content instead.
+  Note 3 : Use ZSTD_DCtx_loadDictionary_advanced() to take control of
+           how dictionary content is loaded and interpreted.
+ 
+


+
+
size_t ZSTD_DCtx_refDDict(ZSTD_DCtx* dctx, const ZSTD_DDict* ddict);
+
  Reference a prepared dictionary, to be used to decompress next frames.
+  The dictionary remains active for decompression of future frames using same DCtx.
+ @result : 0, or an error code (which can be tested with ZSTD_isError()).
+  Note 1 : Currently, only one dictionary can be managed.
+           Referencing a new dictionary effectively "discards" any previous one.
+  Special: referencing a NULL DDict means "return to no-dictionary mode".
+  Note 2 : DDict is just referenced, its lifetime must outlive its usage from DCtx.
+ 
+


+
+
size_t ZSTD_DCtx_refPrefix(ZSTD_DCtx* dctx,
+                     const void* prefix, size_t prefixSize);
+
  Reference a prefix (single-usage dictionary) to decompress next frame.
+  This is the reverse operation of ZSTD_CCtx_refPrefix(),
+  and must use the same prefix as the one used during compression.
+  Prefix is **only used once**. Reference is discarded at end of frame.
+  End of frame is reached when ZSTD_decompressStream() returns 0.
+ @result : 0, or an error code (which can be tested with ZSTD_isError()).
+  Note 1 : Adding any prefix (including NULL) invalidates any previously set prefix or dictionary
+  Note 2 : Prefix buffer is referenced. It **must** outlive decompression.
+           Prefix buffer must remain unmodified up to the end of frame,
+           reached when ZSTD_decompressStream() returns 0.
+  Note 3 : By default, the prefix is treated as raw content (ZSTD_dm_rawContent).
+           Use ZSTD_CCtx_refPrefix_advanced() to alter dictMode (Experimental section)
+  Note 4 : Referencing a raw content prefix has almost no cpu nor memory cost.
+           A full dictionary is more costly, as it requires building tables.
+ 
+


+
+
size_t ZSTD_DCtx_reset(ZSTD_DCtx* dctx, ZSTD_ResetDirective reset);
+
  Return a DCtx to clean state.
+  Session and parameters can be reset jointly or separately.
+  Parameters can only be reset when no active frame is being decompressed.
+ @return : 0, or an error code, which can be tested with ZSTD_isError()
+ 
+


+
+
experimental API (static linking only)
+ The following symbols and constants
+ are not planned to join "stable API" status in the near future.
+ They can still change in future versions.
+ Some of them are planned to remain in the static_only section indefinitely.
+ Some of them might be removed in the future (especially when redundant with existing stable functions)
+ 
+

+
+
typedef struct {
+    unsigned windowLog;       /**< largest match distance : larger == more compression, more memory needed during decompression */
+    unsigned chainLog;        /**< fully searched segment : larger == more compression, slower, more memory (useless for fast) */
+    unsigned hashLog;         /**< dispatch table : larger == faster, more memory */
+    unsigned searchLog;       /**< nb of searches : larger == more compression, slower */
+    unsigned minMatch;        /**< match length searched : larger == faster decompression, sometimes less compression */
+    unsigned targetLength;    /**< acceptable match size for optimal parser (only) : larger == more compression, slower */
+    ZSTD_strategy strategy;   /**< see ZSTD_strategy definition above */
 } ZSTD_compressionParameters;
 


 
typedef struct {
-    unsigned contentSizeFlag; /**< 1: content size will be in frame header (when known) */
-    unsigned checksumFlag;    /**< 1: generate a 32-bits checksum at end of frame, for error detection */
-    unsigned noDictIDFlag;    /**< 1: no dictID will be saved into frame header (if dictionary compression) */
+    int contentSizeFlag; /**< 1: content size will be in frame header (when known) */
+    int checksumFlag;    /**< 1: generate a 32-bits checksum using XXH64 algorithm at end of frame, for error detection */
+    int noDictIDFlag;    /**< 1: no dictID will be saved into frame header (dictID is only useful for dictionary compression) */
 } ZSTD_frameParameters;
 


 
typedef struct {
@@ -374,25 +832,65 @@ size_t ZSTD_decompressStream(ZSTD_DStream* zds, ZSTD_outBuffer* output, ZSTD_inB
 } ZSTD_parameters;
 


 
typedef enum {
-    ZSTD_dct_auto=0,      /* dictionary is "full" when starting with ZSTD_MAGIC_DICTIONARY, otherwise it is "rawContent" */
-    ZSTD_dct_rawContent,  /* ensures dictionary is always loaded as rawContent, even if it starts with ZSTD_MAGIC_DICTIONARY */
-    ZSTD_dct_fullDict     /* refuses to load a dictionary if it does not respect Zstandard's specification */
+    ZSTD_dct_auto = 0,       /* dictionary is "full" when starting with ZSTD_MAGIC_DICTIONARY, otherwise it is "rawContent" */
+    ZSTD_dct_rawContent = 1, /* ensures dictionary is always loaded as rawContent, even if it starts with ZSTD_MAGIC_DICTIONARY */
+    ZSTD_dct_fullDict = 2    /* refuses to load a dictionary if it does not respect Zstandard's specification, starting with ZSTD_MAGIC_DICTIONARY */
 } ZSTD_dictContentType_e;
 


 
typedef enum {
-    ZSTD_dlm_byCopy = 0, /**< Copy dictionary content internally */
-    ZSTD_dlm_byRef,      /**< Reference dictionary content -- the dictionary buffer must outlive its users. */
+    ZSTD_dlm_byCopy = 0,  /**< Copy dictionary content internally */
+    ZSTD_dlm_byRef = 1,   /**< Reference dictionary content -- the dictionary buffer must outlive its users. */
 } ZSTD_dictLoadMethod_e;
 


-
Frame size functions

-
-
size_t ZSTD_findFrameCompressedSize(const void* src, size_t srcSize);
-
  `src` should point to the start of a ZSTD encoded frame or skippable frame
-  `srcSize` must be >= first frame size
-  @return : the compressed size of the first frame starting at `src`,
-            suitable to pass to `ZSTD_decompress` or similar,
-            or an error code if input is invalid 
-


+
typedef enum {
+    /* Opened question : should we have a format ZSTD_f_auto ?
+     * Today, it would mean exactly the same as ZSTD_f_zstd1.
+     * But, in the future, should several formats become supported,
+     * on the compression side, it would mean "default format".
+     * On the decompression side, it would mean "automatic format detection",
+     * so that ZSTD_f_zstd1 would mean "accept *only* zstd frames".
+     * Since meaning is a little different, another option could be to define different enums for compression and decompression.
+     * This question could be kept for later, when there are actually multiple formats to support,
+     * but there is also the question of pinning enum values, and pinning value `0` is especially important */
+    ZSTD_f_zstd1 = 0,           /* zstd frame format, specified in zstd_compression_format.md (default) */
+    ZSTD_f_zstd1_magicless = 1, /* Variant of zstd frame format, without initial 4-bytes magic number.
+                                 * Useful to save 4 bytes per generated frame.
+                                 * Decoder cannot recognise automatically this format, requiring this instruction. */
+} ZSTD_format_e;
+


+
typedef enum {
+    /* Note: this enum and the behavior it controls are effectively internal
+     * implementation details of the compressor. They are expected to continue
+     * to evolve and should be considered only in the context of extremely
+     * advanced performance tuning.
+     *
+     * Zstd currently supports the use of a CDict in two ways:
+     *
+     * - The contents of the CDict can be copied into the working context. This
+     *   means that the compression can search both the dictionary and input
+     *   while operating on a single set of internal tables. This makes
+     *   the compression faster per-byte of input. However, the initial copy of
+     *   the CDict's tables incurs a fixed cost at the beginning of the
+     *   compression. For small compressions (< 8 KB), that copy can dominate
+     *   the cost of the compression.
+     *
+     * - The CDict's tables can be used in-place. In this model, compression is
+     *   slower per input byte, because the compressor has to search two sets of
+     *   tables. However, this model incurs no start-up cost (as long as the
+     *   working context's tables can be reused). For small inputs, this can be
+     *   faster than copying the CDict's tables.
+     *
+     * Zstd has a simple internal heuristic that selects which strategy to use
+     * at the beginning of a compression. However, if experimentation shows that
+     * Zstd is making poor choices, it is possible to override that choice with
+     * this enum.
+     */
+    ZSTD_dictDefaultAttach = 0, /* Use the default heuristic. */
+    ZSTD_dictForceAttach   = 1, /* Never copy the dictionary. */
+    ZSTD_dictForceCopy     = 2, /* Always copy the dictionary. */
+} ZSTD_dictAttachPref_e;
+


+
Frame size functions

 
 
unsigned long long ZSTD_findDecompressedSize(const void* src, size_t srcSize);
 
  `src` should point the start of a series of ZSTD encoded and/or skippable frames
@@ -418,22 +916,12 @@ size_t ZSTD_decompressStream(ZSTD_DStream* zds, ZSTD_outBuffer* output, ZSTD_inB
 


 
 
size_t ZSTD_frameHeaderSize(const void* src, size_t srcSize);
-
  srcSize must be >= ZSTD_frameHeaderSize_prefix.
+
  srcSize must be >= ZSTD_FRAMEHEADERSIZE_PREFIX.
  @return : size of the Frame Header,
            or an error code (if srcSize is too small) 
 


 
-
Memory management

-
-
size_t ZSTD_sizeof_CCtx(const ZSTD_CCtx* cctx);
-size_t ZSTD_sizeof_DCtx(const ZSTD_DCtx* dctx);
-size_t ZSTD_sizeof_CStream(const ZSTD_CStream* zcs);
-size_t ZSTD_sizeof_DStream(const ZSTD_DStream* zds);
-size_t ZSTD_sizeof_CDict(const ZSTD_CDict* cdict);
-size_t ZSTD_sizeof_DDict(const ZSTD_DDict* ddict);
-
  These functions give the current memory usage of selected object.
-  Object memory usage can evolve when re-used. 
-


+
Memory management

 
 
size_t ZSTD_estimateCCtxSize(int compressionLevel);
 size_t ZSTD_estimateCCtxSize_usingCParams(ZSTD_compressionParameters cParams);
@@ -445,7 +933,7 @@ size_t ZSTD_estimateDCtxSize(void);
   It will also consider src size to be arbitrarily "large", which is worst case.
   If srcSize is known to always be small, ZSTD_estimateCCtxSize_usingCParams() can provide a tighter estimation.
   ZSTD_estimateCCtxSize_usingCParams() can be used in tandem with ZSTD_getCParams() to create cParams from compressionLevel.
-  ZSTD_estimateCCtxSize_usingCCtxParams() can be used in tandem with ZSTD_CCtxParam_setParameter(). Only single-threaded compression is supported. This function will return an error code if ZSTD_p_nbWorkers is >= 1.
+  ZSTD_estimateCCtxSize_usingCCtxParams() can be used in tandem with ZSTD_CCtxParam_setParameter(). Only single-threaded compression is supported. This function will return an error code if ZSTD_c_nbWorkers is >= 1.
   Note : CCtx size estimation is only correct for single-threaded compression. 
 


 
@@ -458,7 +946,7 @@ size_t ZSTD_estimateDStreamSize_fromFrame(const void* src, size_t srcSize);
   It will also consider src size to be arbitrarily "large", which is worst case.
   If srcSize is known to always be small, ZSTD_estimateCStreamSize_usingCParams() can provide a tighter estimation.
   ZSTD_estimateCStreamSize_usingCParams() can be used in tandem with ZSTD_getCParams() to create cParams from compressionLevel.
-  ZSTD_estimateCStreamSize_usingCCtxParams() can be used in tandem with ZSTD_CCtxParam_setParameter(). Only single-threaded compression is supported. This function will return an error code if ZSTD_p_nbWorkers is >= 1.
+  ZSTD_estimateCStreamSize_usingCCtxParams() can be used in tandem with ZSTD_CCtxParam_setParameter(). Only single-threaded compression is supported. This function will return an error code if ZSTD_c_nbWorkers is >= 1.
   Note : CStream size estimation is only correct for single-threaded compression.
   ZSTD_DStream memory budget depends on window Size.
   This information can be passed manually, using ZSTD_estimateDStreamSize,
@@ -513,12 +1001,13 @@ static ZSTD_customMem const ZSTD_defaultCMem = { NULL, NULL, NULL };  /**< t
  
 


 
-
Advanced compression functions

+
Advanced compression functions

 
 
ZSTD_CDict* ZSTD_createCDict_byReference(const void* dictBuffer, size_t dictSize, int compressionLevel);
 
  Create a digested dictionary for compression
-  Dictionary content is simply referenced, and therefore stays in dictBuffer.
-  It is important that dictBuffer outlives CDict, it must remain read accessible throughout the lifetime of CDict 
+  Dictionary content is just referenced, not duplicated.
+  As a consequence, `dictBuffer` **must** outlive CDict,
+  and its content must remain unmodified throughout the lifetime of CDict. 
 


 
 
ZSTD_compressionParameters ZSTD_getCParams(int compressionLevel, unsigned long long estimatedSrcSize, size_t dictSize);
@@ -540,22 +1029,120 @@ static ZSTD_customMem const ZSTD_defaultCMem = { NULL, NULL, NULL };  /**< t
   both values are optional, select `0` if unknown. 
 


 
-
size_t ZSTD_compress_advanced (ZSTD_CCtx* cctx,
-                      void* dst, size_t dstCapacity,
-                const void* src, size_t srcSize,
-                const void* dict,size_t dictSize,
-                      ZSTD_parameters params);
-
   Same as ZSTD_compress_usingDict(), with fine-tune control over each compression parameter 
+
size_t ZSTD_compress_advanced(ZSTD_CCtx* cctx,
+                              void* dst, size_t dstCapacity,
+                        const void* src, size_t srcSize,
+                        const void* dict,size_t dictSize,
+                              ZSTD_parameters params);
+
  Same as ZSTD_compress_usingDict(), with fine-tune control over compression parameters (by structure) 
 


 
 
size_t ZSTD_compress_usingCDict_advanced(ZSTD_CCtx* cctx,
-                      void* dst, size_t dstCapacity,
-                const void* src, size_t srcSize,
-                const ZSTD_CDict* cdict, ZSTD_frameParameters fParams);
-
   Same as ZSTD_compress_usingCDict(), with fine-tune control over frame parameters 
+                                  void* dst, size_t dstCapacity,
+                            const void* src, size_t srcSize,
+                            const ZSTD_CDict* cdict,
+                                  ZSTD_frameParameters fParams);
+
  Same as ZSTD_compress_usingCDict(), with fine-tune control over frame parameters 
 


 
-
Advanced decompression functions

+
size_t ZSTD_CCtx_loadDictionary_byReference(ZSTD_CCtx* cctx, const void* dict, size_t dictSize);
+
  Same as ZSTD_CCtx_loadDictionary(), but dictionary content is referenced, instead of being copied into CCtx.
+  It saves some memory, but also requires that `dict` outlives its usage within `cctx` 
+


+
+
size_t ZSTD_CCtx_loadDictionary_advanced(ZSTD_CCtx* cctx, const void* dict, size_t dictSize, ZSTD_dictLoadMethod_e dictLoadMethod, ZSTD_dictContentType_e dictContentType);
+
  Same as ZSTD_CCtx_loadDictionary(), but gives finer control over
+  how to load the dictionary (by copy ? by reference ?)
+  and how to interpret it (automatic ? force raw mode ? full mode only ?) 
+


+
+
size_t ZSTD_CCtx_refPrefix_advanced(ZSTD_CCtx* cctx, const void* prefix, size_t prefixSize, ZSTD_dictContentType_e dictContentType);
+
  Same as ZSTD_CCtx_refPrefix(), but gives finer control over
+  how to interpret prefix content (automatic ? force raw mode (default) ? full mode only ?) 
+


+
+
size_t ZSTD_CCtx_getParameter(ZSTD_CCtx* cctx, ZSTD_cParameter param, int* value);
+
  Get the requested compression parameter value, selected by enum ZSTD_cParameter,
+  and store it into int* value.
+ @return : 0, or an error code (which can be tested with ZSTD_isError()).
+ 
+


+
+
ZSTD_CCtx_params* ZSTD_createCCtxParams(void);
+size_t ZSTD_freeCCtxParams(ZSTD_CCtx_params* params);
+
  Quick howto :
+  - ZSTD_createCCtxParams() : Create a ZSTD_CCtx_params structure
+  - ZSTD_CCtxParam_setParameter() : Push parameters one by one into
+                                    an existing ZSTD_CCtx_params structure.
+                                    This is similar to
+                                    ZSTD_CCtx_setParameter().
+  - ZSTD_CCtx_setParametersUsingCCtxParams() : Apply parameters to
+                                    an existing CCtx.
+                                    These parameters will be applied to
+                                    all subsequent frames.
+  - ZSTD_compressStream2() : Do compression using the CCtx.
+  - ZSTD_freeCCtxParams() : Free the memory.
+
+  This can be used with ZSTD_estimateCCtxSize_advanced_usingCCtxParams()
+  for static allocation of CCtx for single-threaded compression.
+ 
+


+
+
size_t ZSTD_CCtxParams_reset(ZSTD_CCtx_params* params);
+
  Reset params to default values.
+ 
+


+
+
size_t ZSTD_CCtxParams_init(ZSTD_CCtx_params* cctxParams, int compressionLevel);
+
  Initializes the compression parameters of cctxParams according to
+  compression level. All other parameters are reset to their default values.
+ 
+


+
+
size_t ZSTD_CCtxParams_init_advanced(ZSTD_CCtx_params* cctxParams, ZSTD_parameters params);
+
  Initializes the compression and frame parameters of cctxParams according to
+  params. All other parameters are reset to their default values.
+ 
+


+
+
size_t ZSTD_CCtxParam_setParameter(ZSTD_CCtx_params* params, ZSTD_cParameter param, int value);
+
  Similar to ZSTD_CCtx_setParameter.
+  Set one compression parameter, selected by enum ZSTD_cParameter.
+  Parameters must be applied to a ZSTD_CCtx using ZSTD_CCtx_setParametersUsingCCtxParams().
+ @result : 0, or an error code (which can be tested with ZSTD_isError()).
+ 
+


+
+
size_t ZSTD_CCtxParam_getParameter(ZSTD_CCtx_params* params, ZSTD_cParameter param, int* value);
+
 Similar to ZSTD_CCtx_getParameter.
+ Get the requested value of one compression parameter, selected by enum ZSTD_cParameter.
+ @result : 0, or an error code (which can be tested with ZSTD_isError()).
+ 
+


+
+
size_t ZSTD_CCtx_setParametersUsingCCtxParams(
+        ZSTD_CCtx* cctx, const ZSTD_CCtx_params* params);
+
  Apply a set of ZSTD_CCtx_params to the compression context.
+  This can be done even after compression is started,
+    if nbWorkers==0, this will have no impact until a new compression is started.
+    if nbWorkers>=1, new parameters will be picked up at next job,
+       with a few restrictions (windowLog, pledgedSrcSize, nbWorkers, jobSize, and overlapLog are not updated).
+ 
+


+
+
size_t ZSTD_compressStream2_simpleArgs (
+                ZSTD_CCtx* cctx,
+                void* dst, size_t dstCapacity, size_t* dstPos,
+          const void* src, size_t srcSize, size_t* srcPos,
+                ZSTD_EndDirective endOp);
+
  Same as ZSTD_compressStream2(),
+  but using only integral types as arguments.
+  This variant might be helpful for binders from dynamic languages
+  which have troubles handling structures containing memory pointers.
+ 
+


+
+
Advanced decompression functions

 
 
unsigned ZSTD_isFrame(const void* buffer, size_t size);
 
  Tells if the content of `buffer` starts with a valid Frame Identifier.
@@ -595,7 +1182,56 @@ static ZSTD_customMem const ZSTD_defaultCMem = { NULL, NULL, NULL };  /**< t
   When identifying the exact failure cause, it's possible to use ZSTD_getFrameHeader(), which will provide a more precise error code. 
 


 
-
Advanced streaming functions

+
size_t ZSTD_DCtx_loadDictionary_byReference(ZSTD_DCtx* dctx, const void* dict, size_t dictSize);
+
  Same as ZSTD_DCtx_loadDictionary(),
+  but references `dict` content instead of copying it into `dctx`.
+  This saves memory if `dict` remains around.,
+  However, it's imperative that `dict` remains accessible (and unmodified) while being used, so it must outlive decompression. 
+


+
+
size_t ZSTD_DCtx_loadDictionary_advanced(ZSTD_DCtx* dctx, const void* dict, size_t dictSize, ZSTD_dictLoadMethod_e dictLoadMethod, ZSTD_dictContentType_e dictContentType);
+
  Same as ZSTD_DCtx_loadDictionary(),
+  but gives direct control over
+  how to load the dictionary (by copy ? by reference ?)
+  and how to interpret it (automatic ? force raw mode ? full mode only ?). 
+


+
+
size_t ZSTD_DCtx_refPrefix_advanced(ZSTD_DCtx* dctx, const void* prefix, size_t prefixSize, ZSTD_dictContentType_e dictContentType);
+
  Same as ZSTD_DCtx_refPrefix(), but gives finer control over
+  how to interpret prefix content (automatic ? force raw mode (default) ? full mode only ?) 
+


+
+
size_t ZSTD_DCtx_setMaxWindowSize(ZSTD_DCtx* dctx, size_t maxWindowSize);
+
  Refuses allocating internal buffers for frames requiring a window size larger than provided limit.
+  This protects a decoder context from reserving too much memory for itself (potential attack scenario).
+  This parameter is only useful in streaming mode, since no internal buffer is allocated in single-pass mode.
+  By default, a decompression context accepts all window sizes <= (1 << ZSTD_WINDOWLOG_LIMIT_DEFAULT)
+ @return : 0, or an error code (which can be tested using ZSTD_isError()).
+ 
+


+
+
size_t ZSTD_DCtx_setFormat(ZSTD_DCtx* dctx, ZSTD_format_e format);
+
  Instruct the decoder context about what kind of data to decode next.
+  This instruction is mandatory to decode data without a fully-formed header,
+  such ZSTD_f_zstd1_magicless for example.
+ @return : 0, or an error code (which can be tested using ZSTD_isError()). 
+


+
+
size_t ZSTD_decompressStream_simpleArgs (
+                ZSTD_DCtx* dctx,
+                void* dst, size_t dstCapacity, size_t* dstPos,
+          const void* src, size_t srcSize, size_t* srcPos);
+
  Same as ZSTD_decompressStream(),
+  but using only integral types as arguments.
+  This can be helpful for binders from dynamic languages
+  which have troubles handling structures containing memory pointers.
+ 
+


+
+
Advanced streaming functions
  Warning : most of these functions are now redundant with the Advanced API.
+  Once Advanced API reaches "stable" status,
+  redundant functions will be deprecated, and then at some point removed.
+

 
 
Advanced Streaming compression functions
size_t ZSTD_initCStream_srcSize(ZSTD_CStream* zcs, int compressionLevel, unsigned long long pledgedSrcSize);   /**< pledgedSrcSize must be correct. If it is not known at init time, use ZSTD_CONTENTSIZE_UNKNOWN. Note that, for compatibility with older programs, "0" also disables frame content size field. It may be enabled in the future. */
 size_t ZSTD_initCStream_usingDict(ZSTD_CStream* zcs, const void* dict, size_t dictSize, int compressionLevel); /**< creates of an internal CDict (incompatible with static CCtx), except if dict == NULL or dictSize < 8, in which case no dict is used. Note: dict is loaded with ZSTD_dm_auto (treated as a full zstd dictionary if it begins with ZSTD_MAGIC_DICTIONARY, else as raw content) and ZSTD_dlm_byCopy.*/
@@ -605,7 +1241,7 @@ size_t ZSTD_initCStream_usingCDict(ZSTD_CStream* zcs, const ZSTD_CDict* cdict);
 size_t ZSTD_initCStream_usingCDict_advanced(ZSTD_CStream* zcs, const ZSTD_CDict* cdict, ZSTD_frameParameters fParams, unsigned long long pledgedSrcSize);  /**< same as ZSTD_initCStream_usingCDict(), with control over frame parameters. pledgedSrcSize must be correct. If srcSize is not known at init time, use value ZSTD_CONTENTSIZE_UNKNOWN. */
 


 
size_t ZSTD_resetCStream(ZSTD_CStream* zcs, unsigned long long pledgedSrcSize);
-
  start a new compression job, using same parameters from previous job.
+
  start a new frame, using same parameters from previous frame.
   This is typically useful to skip dictionary loading stage, since it will re-use it in-place.
   Note that zcs must be init at least once before using ZSTD_resetCStream().
   If pledgedSrcSize is not known at reset time, use macro ZSTD_CONTENTSIZE_UNKNOWN.
@@ -635,25 +1271,23 @@ size_t ZSTD_initCStream_usingCDict_advanced(ZSTD_CStream* zcs, const ZSTD_CDict*
   + there is no active job (could be checked with ZSTD_frameProgression()), or
   + oldest job is still actively compressing data,
     but everything it has produced has also been flushed so far,
-    therefore flushing speed is currently limited by production speed of oldest job
-    irrespective of the speed of concurrent newer jobs.
+    therefore flush speed is limited by production speed of oldest job
+    irrespective of the speed of concurrent (and newer) jobs.
  
 


 
-
Advanced Streaming decompression functions
typedef enum { DStream_p_maxWindowSize } ZSTD_DStreamParameter_e;
-size_t ZSTD_setDStreamParameter(ZSTD_DStream* zds, ZSTD_DStreamParameter_e paramType, unsigned paramValue);   /* obsolete : this API will be removed in a future version */
-size_t ZSTD_initDStream_usingDict(ZSTD_DStream* zds, const void* dict, size_t dictSize); /**< note: no dictionary will be used if dict == NULL or dictSize < 8 */
+
Advanced Streaming decompression functions
size_t ZSTD_initDStream_usingDict(ZSTD_DStream* zds, const void* dict, size_t dictSize); /**< note: no dictionary will be used if dict == NULL or dictSize < 8 */
 size_t ZSTD_initDStream_usingDDict(ZSTD_DStream* zds, const ZSTD_DDict* ddict);  /**< note : ddict is referenced, it must outlive decompression session */
 size_t ZSTD_resetDStream(ZSTD_DStream* zds);  /**< re-use decompression parameters from previous init; saves dictionary loading */
 


-
Buffer-less and synchronous inner streaming functions
+
Buffer-less and synchronous inner streaming functions
   This is an advanced API, giving full control over buffer management, for users which need direct control over memory.
   But it's also a complex one, with several restrictions, documented below.
   Prefer normal streaming API for an easier experience.
  
 

 
-
Buffer-less streaming compression (synchronous mode)
+
Buffer-less streaming compression (synchronous mode)
   A ZSTD_CCtx object is required to track streaming operations.
   Use ZSTD_createCCtx() / ZSTD_freeCCtx() to manage resource.
   ZSTD_CCtx object can be re-used multiple times within successive compression operations.
@@ -689,7 +1323,7 @@ size_t ZSTD_compressBegin_usingCDict(ZSTD_CCtx* cctx, const ZSTD_CDict* cdict);
 size_t ZSTD_compressBegin_usingCDict_advanced(ZSTD_CCtx* const cctx, const ZSTD_CDict* const cdict, ZSTD_frameParameters const fParams, unsigned long long const pledgedSrcSize);   /* compression parameters are already set within cdict. pledgedSrcSize must be correct. If srcSize is not known, use macro ZSTD_CONTENTSIZE_UNKNOWN */
 size_t ZSTD_copyCCtx(ZSTD_CCtx* cctx, const ZSTD_CCtx* preparedCCtx, unsigned long long pledgedSrcSize); /**<  note: if pledgedSrcSize is not known, use ZSTD_CONTENTSIZE_UNKNOWN */
 


-
Buffer-less streaming decompression (synchronous mode)
+
Buffer-less streaming decompression (synchronous mode)
   A ZSTD_DCtx object is required to track streaming operations.
   Use ZSTD_createDCtx() / ZSTD_freeDCtx() to manage it.
   A ZSTD_DCtx object can be re-used multiple times.
@@ -770,496 +1404,24 @@ typedef struct {
     unsigned dictID;
     unsigned checksumFlag;
 } ZSTD_frameHeader;
-/** ZSTD_getFrameHeader() :
- *  decode Frame Header, or requires larger `srcSize`.
- * @return : 0, `zfhPtr` is correctly filled,
- *          >0, `srcSize` is too small, value is wanted `srcSize` amount,
- *           or an error code, which can be tested using ZSTD_isError() */
-size_t ZSTD_getFrameHeader(ZSTD_frameHeader* zfhPtr, const void* src, size_t srcSize);   /**< doesn't consume input */
+


+
ZSTD_getFrameHeader() :
  decode Frame Header, or requires larger `srcSize`.
+ @return : 0, `zfhPtr` is correctly filled,
+          >0, `srcSize` is too small, value is wanted `srcSize` amount,
+           or an error code, which can be tested using ZSTD_isError() 
+

+
+
size_t ZSTD_getFrameHeader(ZSTD_frameHeader* zfhPtr, const void* src, size_t srcSize);   /**< doesn't consume input */
+


+
size_t ZSTD_getFrameHeader_advanced(ZSTD_frameHeader* zfhPtr, const void* src, size_t srcSize, ZSTD_format_e format);
 size_t ZSTD_decodingBufferSize_min(unsigned long long windowSize, unsigned long long frameContentSize);  /**< when frame content size is not known, pass in frameContentSize == ZSTD_CONTENTSIZE_UNKNOWN */
-


-
typedef enum { ZSTDnit_frameHeader, ZSTDnit_blockHeader, ZSTDnit_block, ZSTDnit_lastBlock, ZSTDnit_checksum, ZSTDnit_skippableFrame } ZSTD_nextInputType_e;
-


-
New advanced API (experimental)

-
-
typedef enum {
-    /* Opened question : should we have a format ZSTD_f_auto ?
-     * Today, it would mean exactly the same as ZSTD_f_zstd1.
-     * But, in the future, should several formats become supported,
-     * on the compression side, it would mean "default format".
-     * On the decompression side, it would mean "automatic format detection",
-     * so that ZSTD_f_zstd1 would mean "accept *only* zstd frames".
-     * Since meaning is a little different, another option could be to define different enums for compression and decompression.
-     * This question could be kept for later, when there are actually multiple formats to support,
-     * but there is also the question of pinning enum values, and pinning value `0` is especially important */
-    ZSTD_f_zstd1 = 0,        /* zstd frame format, specified in zstd_compression_format.md (default) */
-    ZSTD_f_zstd1_magicless,  /* Variant of zstd frame format, without initial 4-bytes magic number.
-                              * Useful to save 4 bytes per generated frame.
-                              * Decoder cannot recognise automatically this format, requiring instructions. */
-} ZSTD_format_e;
-


-
typedef enum {
-    /* compression format */
-    ZSTD_p_format = 10,      /* See ZSTD_format_e enum definition.
-                              * Cast selected format as unsigned for ZSTD_CCtx_setParameter() compatibility. */
-
-    /* compression parameters */
-    ZSTD_p_compressionLevel=100, /* Update all compression parameters according to pre-defined cLevel table
-                              * Default level is ZSTD_CLEVEL_DEFAULT==3.
-                              * Special: value 0 means default, which is controlled by ZSTD_CLEVEL_DEFAULT.
-                              * Note 1 : it's possible to pass a negative compression level by casting it to unsigned type.
-                              * Note 2 : setting a level sets all default values of other compression parameters.
-                              * Note 3 : setting compressionLevel automatically updates ZSTD_p_compressLiterals. */
-    ZSTD_p_windowLog,        /* Maximum allowed back-reference distance, expressed as power of 2.
-                              * Must be clamped between ZSTD_WINDOWLOG_MIN and ZSTD_WINDOWLOG_MAX.
-                              * Special: value 0 means "use default windowLog".
-                              * Note: Using a window size greater than ZSTD_MAXWINDOWSIZE_DEFAULT (default: 2^27)
-                              *       requires explicitly allowing such window size during decompression stage. */
-    ZSTD_p_hashLog,          /* Size of the initial probe table, as a power of 2.
-                              * Resulting table size is (1 << (hashLog+2)).
-                              * Must be clamped between ZSTD_HASHLOG_MIN and ZSTD_HASHLOG_MAX.
-                              * Larger tables improve compression ratio of strategies <= dFast,
-                              * and improve speed of strategies > dFast.
-                              * Special: value 0 means "use default hashLog". */
-    ZSTD_p_chainLog,         /* Size of the multi-probe search table, as a power of 2.
-                              * Resulting table size is (1 << (chainLog+2)).
-                              * Must be clamped between ZSTD_CHAINLOG_MIN and ZSTD_CHAINLOG_MAX.
-                              * Larger tables result in better and slower compression.
-                              * This parameter is useless when using "fast" strategy.
-                              * Note it's still useful when using "dfast" strategy,
-                              * in which case it defines a secondary probe table.
-                              * Special: value 0 means "use default chainLog". */
-    ZSTD_p_searchLog,        /* Number of search attempts, as a power of 2.
-                              * More attempts result in better and slower compression.
-                              * This parameter is useless when using "fast" and "dFast" strategies.
-                              * Special: value 0 means "use default searchLog". */
-    ZSTD_p_minMatch,         /* Minimum size of searched matches (note : repCode matches can be smaller).
-                              * Larger values make faster compression and decompression, but decrease ratio.
-                              * Must be clamped between ZSTD_SEARCHLENGTH_MIN and ZSTD_SEARCHLENGTH_MAX.
-                              * Note that currently, for all strategies < btopt, effective minimum is 4.
-                              *                    , for all strategies > fast, effective maximum is 6.
-                              * Special: value 0 means "use default minMatchLength". */
-    ZSTD_p_targetLength,     /* Impact of this field depends on strategy.
-                              * For strategies btopt & btultra:
-                              *     Length of Match considered "good enough" to stop search.
-                              *     Larger values make compression stronger, and slower.
-                              * For strategy fast:
-                              *     Distance between match sampling.
-                              *     Larger values make compression faster, and weaker.
-                              * Special: value 0 means "use default targetLength". */
-    ZSTD_p_compressionStrategy, /* See ZSTD_strategy enum definition.
-                              * Cast selected strategy as unsigned for ZSTD_CCtx_setParameter() compatibility.
-                              * The higher the value of selected strategy, the more complex it is,
-                              * resulting in stronger and slower compression.
-                              * Special: value 0 means "use default strategy". */
-
-    ZSTD_p_enableLongDistanceMatching=160, /* Enable long distance matching.
-                                         * This parameter is designed to improve compression ratio
-                                         * for large inputs, by finding large matches at long distance.
-                                         * It increases memory usage and window size.
-                                         * Note: enabling this parameter increases ZSTD_p_windowLog to 128 MB
-                                         * except when expressly set to a different value. */
-    ZSTD_p_ldmHashLog,       /* Size of the table for long distance matching, as a power of 2.
-                              * Larger values increase memory usage and compression ratio,
-                              * but decrease compression speed.
-                              * Must be clamped between ZSTD_HASHLOG_MIN and ZSTD_HASHLOG_MAX
-                              * default: windowlog - 7.
-                              * Special: value 0 means "automatically determine hashlog". */
-    ZSTD_p_ldmMinMatch,      /* Minimum match size for long distance matcher.
-                              * Larger/too small values usually decrease compression ratio.
-                              * Must be clamped between ZSTD_LDM_MINMATCH_MIN and ZSTD_LDM_MINMATCH_MAX.
-                              * Special: value 0 means "use default value" (default: 64). */
-    ZSTD_p_ldmBucketSizeLog, /* Log size of each bucket in the LDM hash table for collision resolution.
-                              * Larger values improve collision resolution but decrease compression speed.
-                              * The maximum value is ZSTD_LDM_BUCKETSIZELOG_MAX .
-                              * Special: value 0 means "use default value" (default: 3). */
-    ZSTD_p_ldmHashEveryLog,  /* Frequency of inserting/looking up entries in the LDM hash table.
-                              * Must be clamped between 0 and (ZSTD_WINDOWLOG_MAX - ZSTD_HASHLOG_MIN).
-                              * Default is MAX(0, (windowLog - ldmHashLog)), optimizing hash table usage.
-                              * Larger values improve compression speed.
-                              * Deviating far from default value will likely result in a compression ratio decrease.
-                              * Special: value 0 means "automatically determine hashEveryLog". */
-
-    /* frame parameters */
-    ZSTD_p_contentSizeFlag=200, /* Content size will be written into frame header _whenever known_ (default:1)
-                              * Content size must be known at the beginning of compression,
-                              * it is provided using ZSTD_CCtx_setPledgedSrcSize() */
-    ZSTD_p_checksumFlag,     /* A 32-bits checksum of content is written at end of frame (default:0) */
-    ZSTD_p_dictIDFlag,       /* When applicable, dictionary's ID is written into frame header (default:1) */
-
-    /* multi-threading parameters */
-    /* These parameters are only useful if multi-threading is enabled (ZSTD_MULTITHREAD).
-     * They return an error otherwise. */
-    ZSTD_p_nbWorkers=400,    /* Select how many threads will be spawned to compress in parallel.
-                              * When nbWorkers >= 1, triggers asynchronous mode :
-                              * ZSTD_compress_generic() consumes some input, flush some output if possible, and immediately gives back control to caller,
-                              * while compression work is performed in parallel, within worker threads.
-                              * (note : a strong exception to this rule is when first invocation sets ZSTD_e_end : it becomes a blocking call).
-                              * More workers improve speed, but also increase memory usage.
-                              * Default value is `0`, aka "single-threaded mode" : no worker is spawned, compression is performed inside Caller's thread, all invocations are blocking */
-    ZSTD_p_jobSize,          /* Size of a compression job. This value is enforced only in non-blocking mode.
-                              * Each compression job is completed in parallel, so this value indirectly controls the nb of active threads.
-                              * 0 means default, which is dynamically determined based on compression parameters.
-                              * Job size must be a minimum of overlapSize, or 1 MB, whichever is largest.
-                              * The minimum size is automatically and transparently enforced */
-    ZSTD_p_overlapSizeLog,   /* Size of previous input reloaded at the beginning of each job.
-                              * 0 => no overlap, 6(default) => use 1/8th of windowSize, >=9 => use full windowSize */
-
-    /* =================================================================== */
-    /* experimental parameters - no stability guaranteed                   */
-    /* =================================================================== */
-
-    ZSTD_p_forceMaxWindow=1100, /* Force back-reference distances to remain < windowSize,
-                              * even when referencing into Dictionary content (default:0) */
-    ZSTD_p_forceAttachDict,  /* ZSTD supports usage of a CDict in-place
-                              * (avoiding having to copy the compression tables
-                              * from the CDict into the working context). Using
-                              * a CDict in this way saves an initial setup step,
-                              * but comes at the cost of more work per byte of
-                              * input. ZSTD has a simple internal heuristic that
-                              * guesses which strategy will be faster. You can
-                              * use this flag to override that guess.
-                              *
-                              * Note that the by-reference, in-place strategy is
-                              * only used when reusing a compression context
-                              * with compatible compression parameters. (If
-                              * incompatible / uninitialized, the working
-                              * context needs to be cleared anyways, which is
-                              * about as expensive as overwriting it with the
-                              * dictionary context, so there's no savings in
-                              * using the CDict by-ref.)
-                              *
-                              * Values greater than 0 force attaching the dict.
-                              * Values less than 0 force copying the dict.
-                              * 0 selects the default heuristic-guided behavior.
-                              */
-
-} ZSTD_cParameter;
-


-
size_t ZSTD_CCtx_setParameter(ZSTD_CCtx* cctx, ZSTD_cParameter param, unsigned value);
-
  Set one compression parameter, selected by enum ZSTD_cParameter.
-  Setting a parameter is generally only possible during frame initialization (before starting compression).
-  Exception : when using multi-threading mode (nbThreads >= 1),
-              following parameters can be updated _during_ compression (within same frame):
-              => compressionLevel, hashLog, chainLog, searchLog, minMatch, targetLength and strategy.
-              new parameters will be active on next job, or after a flush().
-  Note : when `value` type is not unsigned (int, or enum), cast it to unsigned for proper type checking.
-  @result : informational value (typically, value being set, correctly clamped),
-            or an error code (which can be tested with ZSTD_isError()). 
-


-
-
size_t ZSTD_CCtx_getParameter(ZSTD_CCtx* cctx, ZSTD_cParameter param, unsigned* value);
-
 Get the requested value of one compression parameter, selected by enum ZSTD_cParameter.
- @result : 0, or an error code (which can be tested with ZSTD_isError()).
- 
-


-
-
size_t ZSTD_CCtx_setPledgedSrcSize(ZSTD_CCtx* cctx, unsigned long long pledgedSrcSize);
-
  Total input data size to be compressed as a single frame.
-  This value will be controlled at the end, and result in error if not respected.
- @result : 0, or an error code (which can be tested with ZSTD_isError()).
-  Note 1 : 0 means zero, empty.
-           In order to mean "unknown content size", pass constant ZSTD_CONTENTSIZE_UNKNOWN.
-           ZSTD_CONTENTSIZE_UNKNOWN is default value for any new compression job.
-  Note 2 : If all data is provided and consumed in a single round,
-           this value is overriden by srcSize instead. 
-


-
-
size_t ZSTD_CCtx_loadDictionary(ZSTD_CCtx* cctx, const void* dict, size_t dictSize);
-size_t ZSTD_CCtx_loadDictionary_byReference(ZSTD_CCtx* cctx, const void* dict, size_t dictSize);
-size_t ZSTD_CCtx_loadDictionary_advanced(ZSTD_CCtx* cctx, const void* dict, size_t dictSize, ZSTD_dictLoadMethod_e dictLoadMethod, ZSTD_dictContentType_e dictContentType);
-
  Create an internal CDict from `dict` buffer.
-  Decompression will have to use same dictionary.
- @result : 0, or an error code (which can be tested with ZSTD_isError()).
-  Special: Adding a NULL (or 0-size) dictionary invalidates previous dictionary,
-           meaning "return to no-dictionary mode".
-  Note 1 : Dictionary will be used for all future compression jobs.
-           To return to "no-dictionary" situation, load a NULL dictionary
-  Note 2 : Loading a dictionary involves building tables, which are dependent on compression parameters.
-           For this reason, compression parameters cannot be changed anymore after loading a dictionary.
-           It's also a CPU consuming operation, with non-negligible impact on latency.
-  Note 3 :`dict` content will be copied internally.
-           Use ZSTD_CCtx_loadDictionary_byReference() to reference dictionary content instead.
-           In such a case, dictionary buffer must outlive its users.
-  Note 4 : Use ZSTD_CCtx_loadDictionary_advanced()
-           to precisely select how dictionary content must be interpreted. 
-


-
-
size_t ZSTD_CCtx_refCDict(ZSTD_CCtx* cctx, const ZSTD_CDict* cdict);
-
  Reference a prepared dictionary, to be used for all next compression jobs.
-  Note that compression parameters are enforced from within CDict,
-  and supercede any compression parameter previously set within CCtx.
-  The dictionary will remain valid for future compression jobs using same CCtx.
- @result : 0, or an error code (which can be tested with ZSTD_isError()).
-  Special : adding a NULL CDict means "return to no-dictionary mode".
-  Note 1 : Currently, only one dictionary can be managed.
-           Adding a new dictionary effectively "discards" any previous one.
-  Note 2 : CDict is just referenced, its lifetime must outlive CCtx. 
-


-
-
size_t ZSTD_CCtx_refPrefix(ZSTD_CCtx* cctx,
-                           const void* prefix, size_t prefixSize);
-size_t ZSTD_CCtx_refPrefix_advanced(ZSTD_CCtx* cctx,
-                           const void* prefix, size_t prefixSize,
-                           ZSTD_dictContentType_e dictContentType);
-
  Reference a prefix (single-usage dictionary) for next compression job.
-  Decompression will need same prefix to properly regenerate data.
-  Compressing with a prefix is similar in outcome as performing a diff and compressing it,
-  but performs much faster, especially during decompression (compression speed is tunable with compression level).
-  Note that prefix is **only used once**. Tables are discarded at end of compression job (ZSTD_e_end).
- @result : 0, or an error code (which can be tested with ZSTD_isError()).
-  Special: Adding any prefix (including NULL) invalidates any previous prefix or dictionary
-  Note 1 : Prefix buffer is referenced. It **must** outlive compression job.
-           Its contain must remain unmodified up to end of compression (ZSTD_e_end).
-  Note 2 : If the intention is to diff some large src data blob with some prior version of itself,
-           ensure that the window size is large enough to contain the entire source.
-           See ZSTD_p_windowLog.
-  Note 3 : Referencing a prefix involves building tables, which are dependent on compression parameters.
-           It's a CPU consuming operation, with non-negligible impact on latency.
-           If there is a need to use same prefix multiple times, consider loadDictionary instead.
-  Note 4 : By default, the prefix is treated as raw content (ZSTD_dm_rawContent).
-           Use ZSTD_CCtx_refPrefix_advanced() to alter dictMode. 
-


-
-
void ZSTD_CCtx_reset(ZSTD_CCtx* cctx);
-
  Return a CCtx to clean state.
-  Useful after an error, or to interrupt an ongoing compression job and start a new one.
-  Any internal data not yet flushed is cancelled.
-  The parameters and dictionary are kept unchanged, to reset them use ZSTD_CCtx_resetParameters().
- 
-


-
-
size_t ZSTD_CCtx_resetParameters(ZSTD_CCtx* cctx);
-
  All parameters are back to default values (compression level is ZSTD_CLEVEL_DEFAULT).
-  Dictionary (if any) is dropped.
-  Resetting parameters is only possible during frame initialization (before starting compression).
-  To reset the context use ZSTD_CCtx_reset().
-  @return 0 or an error code (which can be checked with ZSTD_isError()).
- 
-


-
-
typedef enum {
-    ZSTD_e_continue=0, /* collect more data, encoder decides when to output compressed result, for optimal compression ratio */
-    ZSTD_e_flush,      /* flush any data provided so far,
-                        * it creates (at least) one new block, that can be decoded immediately on reception;
-                        * frame will continue: any future data can still reference previously compressed data, improving compression. */
-    ZSTD_e_end         /* flush any remaining data and close current frame.
-                        * any additional data starts a new frame.
-                        * each frame is independent (does not reference any content from previous frame). */
-} ZSTD_EndDirective;
-


-
size_t ZSTD_compress_generic (ZSTD_CCtx* cctx,
-                              ZSTD_outBuffer* output,
-                              ZSTD_inBuffer* input,
-                              ZSTD_EndDirective endOp);
-
  Behave about the same as ZSTD_compressStream. To note :
-  - Compression parameters are pushed into CCtx before starting compression, using ZSTD_CCtx_setParameter()
-  - Compression parameters cannot be changed once compression is started.
-  - outpot->pos must be <= dstCapacity, input->pos must be <= srcSize
-  - outpot->pos and input->pos will be updated. They are guaranteed to remain below their respective limit.
-  - In single-thread mode (default), function is blocking : it completed its job before returning to caller.
-  - In multi-thread mode, function is non-blocking : it just acquires a copy of input, and distribute job to internal worker threads,
-                                                     and then immediately returns, just indicating that there is some data remaining to be flushed.
-                                                     The function nonetheless guarantees forward progress : it will return only after it reads or write at least 1+ byte.
-  - Exception : in multi-threading mode, if the first call requests a ZSTD_e_end directive, it is blocking : it will complete compression before giving back control to caller.
-  - @return provides a minimum amount of data remaining to be flushed from internal buffers
-            or an error code, which can be tested using ZSTD_isError().
-            if @return != 0, flush is not fully completed, there is still some data left within internal buffers.
-            This is useful for ZSTD_e_flush, since in this case more flushes are necessary to empty all buffers.
-            For ZSTD_e_end, @return == 0 when internal buffers are fully flushed and frame is completed.
-  - after a ZSTD_e_end directive, if internal buffer is not fully flushed (@return != 0),
-            only ZSTD_e_end or ZSTD_e_flush operations are allowed.
-            Before starting a new compression job, or changing compression parameters,
-            it is required to fully flush internal buffers.
- 
-


-
-
size_t ZSTD_compress_generic_simpleArgs (
-                ZSTD_CCtx* cctx,
-                void* dst, size_t dstCapacity, size_t* dstPos,
-          const void* src, size_t srcSize, size_t* srcPos,
-                ZSTD_EndDirective endOp);
-
  Same as ZSTD_compress_generic(),
-  but using only integral types as arguments.
-  Argument list is larger than ZSTD_{in,out}Buffer,
-  but can be helpful for binders from dynamic languages
-  which have troubles handling structures containing memory pointers.
- 
-


-
-
ZSTD_CCtx_params* ZSTD_createCCtxParams(void);
-size_t ZSTD_freeCCtxParams(ZSTD_CCtx_params* params);
-
  Quick howto :
-  - ZSTD_createCCtxParams() : Create a ZSTD_CCtx_params structure
-  - ZSTD_CCtxParam_setParameter() : Push parameters one by one into
-                                    an existing ZSTD_CCtx_params structure.
-                                    This is similar to
-                                    ZSTD_CCtx_setParameter().
-  - ZSTD_CCtx_setParametersUsingCCtxParams() : Apply parameters to
-                                    an existing CCtx.
-                                    These parameters will be applied to
-                                    all subsequent compression jobs.
-  - ZSTD_compress_generic() : Do compression using the CCtx.
-  - ZSTD_freeCCtxParams() : Free the memory.
-
-  This can be used with ZSTD_estimateCCtxSize_advanced_usingCCtxParams()
-  for static allocation for single-threaded compression.
- 
-


-
-
size_t ZSTD_CCtxParams_reset(ZSTD_CCtx_params* params);
-
  Reset params to default values.
- 
-


-
-
size_t ZSTD_CCtxParams_init(ZSTD_CCtx_params* cctxParams, int compressionLevel);
-
  Initializes the compression parameters of cctxParams according to
-  compression level. All other parameters are reset to their default values.
- 
-


-
-
size_t ZSTD_CCtxParams_init_advanced(ZSTD_CCtx_params* cctxParams, ZSTD_parameters params);
-
  Initializes the compression and frame parameters of cctxParams according to
-  params. All other parameters are reset to their default values.
- 
-


-
-
size_t ZSTD_CCtxParam_setParameter(ZSTD_CCtx_params* params, ZSTD_cParameter param, unsigned value);
-
  Similar to ZSTD_CCtx_setParameter.
-  Set one compression parameter, selected by enum ZSTD_cParameter.
-  Parameters must be applied to a ZSTD_CCtx using ZSTD_CCtx_setParametersUsingCCtxParams().
-  Note : when `value` is an enum, cast it to unsigned for proper type checking.
- @result : 0, or an error code (which can be tested with ZSTD_isError()).
- 
-


-
-
size_t ZSTD_CCtxParam_getParameter(ZSTD_CCtx_params* params, ZSTD_cParameter param, unsigned* value);
-
 Similar to ZSTD_CCtx_getParameter.
- Get the requested value of one compression parameter, selected by enum ZSTD_cParameter.
- @result : 0, or an error code (which can be tested with ZSTD_isError()).
- 
-


-
-
size_t ZSTD_CCtx_setParametersUsingCCtxParams(
-        ZSTD_CCtx* cctx, const ZSTD_CCtx_params* params);
-
  Apply a set of ZSTD_CCtx_params to the compression context.
-  This can be done even after compression is started,
-    if nbWorkers==0, this will have no impact until a new compression is started.
-    if nbWorkers>=1, new parameters will be picked up at next job,
-       with a few restrictions (windowLog, pledgedSrcSize, nbWorkers, jobSize, and overlapLog are not updated).
- 
-


-
-
Advanced decompression API
/* ==================================== */
-


-
size_t ZSTD_DCtx_loadDictionary(ZSTD_DCtx* dctx, const void* dict, size_t dictSize);
-size_t ZSTD_DCtx_loadDictionary_byReference(ZSTD_DCtx* dctx, const void* dict, size_t dictSize);
-size_t ZSTD_DCtx_loadDictionary_advanced(ZSTD_DCtx* dctx, const void* dict, size_t dictSize, ZSTD_dictLoadMethod_e dictLoadMethod, ZSTD_dictContentType_e dictContentType);
-
  Create an internal DDict from dict buffer,
-  to be used to decompress next frames.
- @result : 0, or an error code (which can be tested with ZSTD_isError()).
-  Special : Adding a NULL (or 0-size) dictionary invalidates any previous dictionary,
-            meaning "return to no-dictionary mode".
-  Note 1 : `dict` content will be copied internally.
-            Use ZSTD_DCtx_loadDictionary_byReference()
-            to reference dictionary content instead.
-            In which case, the dictionary buffer must outlive its users.
-  Note 2 : Loading a dictionary involves building tables,
-           which has a non-negligible impact on CPU usage and latency.
-  Note 3 : Use ZSTD_DCtx_loadDictionary_advanced() to select
-           how dictionary content will be interpreted and loaded.
- 
-


-
-
size_t ZSTD_DCtx_refDDict(ZSTD_DCtx* dctx, const ZSTD_DDict* ddict);
-
  Reference a prepared dictionary, to be used to decompress next frames.
-  The dictionary remains active for decompression of future frames using same DCtx.
- @result : 0, or an error code (which can be tested with ZSTD_isError()).
-  Note 1 : Currently, only one dictionary can be managed.
-           Referencing a new dictionary effectively "discards" any previous one.
-  Special : adding a NULL DDict means "return to no-dictionary mode".
-  Note 2 : DDict is just referenced, its lifetime must outlive its usage from DCtx.
- 
-


-
-
size_t ZSTD_DCtx_refPrefix(ZSTD_DCtx* dctx,
-                        const void* prefix, size_t prefixSize);
-size_t ZSTD_DCtx_refPrefix_advanced(ZSTD_DCtx* dctx,
-                        const void* prefix, size_t prefixSize,
-                        ZSTD_dictContentType_e dictContentType);
-
  Reference a prefix (single-usage dictionary) for next compression job.
-  This is the reverse operation of ZSTD_CCtx_refPrefix(),
-  and must use the same prefix as the one used during compression.
-  Prefix is **only used once**. Reference is discarded at end of frame.
-  End of frame is reached when ZSTD_DCtx_decompress_generic() returns 0.
- @result : 0, or an error code (which can be tested with ZSTD_isError()).
-  Note 1 : Adding any prefix (including NULL) invalidates any previously set prefix or dictionary
-  Note 2 : Prefix buffer is referenced. It **must** outlive decompression job.
-           Prefix buffer must remain unmodified up to the end of frame,
-           reached when ZSTD_DCtx_decompress_generic() returns 0.
-  Note 3 : By default, the prefix is treated as raw content (ZSTD_dm_rawContent).
-           Use ZSTD_CCtx_refPrefix_advanced() to alter dictMode.
-  Note 4 : Referencing a raw content prefix has almost no cpu nor memory cost.
-           A fulldict prefix is more costly though.
- 
-


-
-
size_t ZSTD_DCtx_setMaxWindowSize(ZSTD_DCtx* dctx, size_t maxWindowSize);
-
  Refuses allocating internal buffers for frames requiring a window size larger than provided limit.
-  This is useful to prevent a decoder context from reserving too much memory for itself (potential attack scenario).
-  This parameter is only useful in streaming mode, since no internal buffer is allocated in direct mode.
-  By default, a decompression context accepts all window sizes <= (1 << ZSTD_WINDOWLOG_MAX)
- @return : 0, or an error code (which can be tested using ZSTD_isError()).
- 
-


-
-
size_t ZSTD_DCtx_setFormat(ZSTD_DCtx* dctx, ZSTD_format_e format);
-
  Instruct the decoder context about what kind of data to decode next.
-  This instruction is mandatory to decode data without a fully-formed header,
-  such ZSTD_f_zstd1_magicless for example.
- @return : 0, or an error code (which can be tested using ZSTD_isError()).
- 
-


-
-
size_t ZSTD_getFrameHeader_advanced(ZSTD_frameHeader* zfhPtr,
-            const void* src, size_t srcSize, ZSTD_format_e format);
 
  same as ZSTD_getFrameHeader(),
   with added capability to select a format (like ZSTD_f_zstd1_magicless) 
 


 
-
size_t ZSTD_decompress_generic(ZSTD_DCtx* dctx,
-                               ZSTD_outBuffer* output,
-                               ZSTD_inBuffer* input);
-
  Behave the same as ZSTD_decompressStream.
-  Decompression parameters cannot be changed once decompression is started.
- @return : an error code, which can be tested using ZSTD_isError()
-           if >0, a hint, nb of expected input bytes for next invocation.
-           `0` means : a frame has just been fully decoded and flushed.
- 
-


-
-
size_t ZSTD_decompress_generic_simpleArgs (
-                ZSTD_DCtx* dctx,
-                void* dst, size_t dstCapacity, size_t* dstPos,
-          const void* src, size_t srcSize, size_t* srcPos);
-
  Same as ZSTD_decompress_generic(),
-  but using only integral types as arguments.
-  Argument list is larger than ZSTD_{in,out}Buffer,
-  but can be helpful for binders from dynamic languages
-  which have troubles handling structures containing memory pointers.
- 
-


-
-
void ZSTD_DCtx_reset(ZSTD_DCtx* dctx);
-
  Return a DCtx to clean state.
-  If a decompression was ongoing, any internal data not yet flushed is cancelled.
-  All parameters are back to default values, including sticky ones.
-  Dictionary (if any) is dropped.
-  Parameters can be modified again after a reset.
- 
-


-
-
Block level API

+
typedef enum { ZSTDnit_frameHeader, ZSTDnit_blockHeader, ZSTDnit_block, ZSTDnit_lastBlock, ZSTDnit_checksum, ZSTDnit_skippableFrame } ZSTD_nextInputType_e;
+


+
Block level API

 
 
    Frame metadata cost is typically ~18 bytes, which can be non-negligible for very small blocks (< 100 bytes).
     User will have to take in charge required information to regenerate data, such as compressed and content sizes.
@@ -1273,10 +1435,10 @@ size_t ZSTD_DCtx_refPrefix_advanced(ZSTD_DCtx* dctx,
       + copyCCtx() and copyDCtx() can be used too
     - Block size is limited, it must be <= ZSTD_getBlockSize() <= ZSTD_BLOCKSIZE_MAX == 128 KB
       + If input is larger than a block size, it's necessary to split input data into multiple blocks
-      + For inputs larger than a single block size, consider using the regular ZSTD_compress() instead.
+      + For inputs larger than a single block, really consider using regular ZSTD_compress() instead.
         Frame metadata is not that costly, and quickly becomes negligible as source size grows larger.
     - When a block is considered not compressible enough, ZSTD_compressBlock() result will be zero.
-      In which case, nothing is produced into `dst`.
+      In which case, nothing is produced into `dst` !
       + User must test for such outcome and deal directly with uncompressed data
       + ZSTD_decompressBlock() doesn't accept uncompressed data as input !!!
       + In case of multiple successive blocks, should some of them be uncompressed,
diff --git a/lib/BUCK b/lib/BUCK
index bd93b082a347..637c20d667c6 100644
--- a/lib/BUCK
+++ b/lib/BUCK
@@ -1,6 +1,7 @@
 cxx_library(
     name='zstd',
     header_namespace='',
+    exported_headers=['zstd.h'],
     visibility=['PUBLIC'],
     deps=[
         ':common',
@@ -17,7 +18,7 @@ cxx_library(
     exported_headers=subdir_glob([
         ('compress', 'zstd*.h'),
     ]),
-    srcs=glob(['compress/zstd*.c']),
+    srcs=glob(['compress/zstd*.c', 'compress/hist.c']),
     deps=[':common'],
 )
 
@@ -40,7 +41,7 @@ cxx_library(
     header_namespace='',
     visibility=['PUBLIC'],
     exported_headers=subdir_glob([
-        ('decprecated', '*.h'),
+        ('deprecated', '*.h'),
     ]),
     srcs=glob(['deprecated/*.c']),
     deps=[':common'],
@@ -118,6 +119,7 @@ cxx_library(
         'decompress/huf_decompress.c',
     ],
     deps=[
+        ':debug',
         ':bitstream',
         ':compiler',
         ':errors',
@@ -204,9 +206,20 @@ cxx_library(
     ],
 )
 
+cxx_library(
+    name='debug',
+    header_namespace='',
+    visibility=['PUBLIC'],
+    exported_headers=subdir_glob([
+        ('common', 'debug.h'),
+    ]),
+    srcs=['common/debug.c'],
+)
+
 cxx_library(
     name='common',
     deps=[
+        ':debug',
         ':bitstream',
         ':compiler',
         ':cpu',
diff --git a/lib/Makefile b/lib/Makefile
index 9711f75eea87..386583e728b9 100644
--- a/lib/Makefile
+++ b/lib/Makefile
@@ -27,11 +27,16 @@ DEBUGFLAGS= -Wall -Wextra -Wcast-qual -Wcast-align -Wshadow \
             -Wstrict-aliasing=1 -Wswitch-enum -Wdeclaration-after-statement \
             -Wstrict-prototypes -Wundef -Wpointer-arith -Wformat-security \
             -Wvla -Wformat=2 -Winit-self -Wfloat-equal -Wwrite-strings \
-            -Wredundant-decls -Wmissing-prototypes
+            -Wredundant-decls -Wmissing-prototypes -Wc++-compat
 CFLAGS  += $(DEBUGFLAGS) $(MOREFLAGS)
 FLAGS    = $(CPPFLAGS) $(CFLAGS)
 
-GREP = grep --color=never
+HAVE_COLORNEVER = $(shell echo a | grep --color=never a > /dev/null 2> /dev/null && echo 1 || echo 0)
+GREP_OPTIONS ?=
+ifeq ($HAVE_COLORNEVER, 1)
+GREP_OPTIONS += --color=never
+endif
+GREP = grep $(GREP_OPTIONS)
 
 ZSTDCOMMON_FILES := $(sort $(wildcard common/*.c))
 ZSTDCOMP_FILES := $(sort $(wildcard compress/*.c))
@@ -45,6 +50,12 @@ ZSTD_LIB_COMPRESSION ?= 1
 ZSTD_LIB_DECOMPRESSION ?= 1
 ZSTD_LIB_DICTBUILDER ?= 1
 ZSTD_LIB_DEPRECATED ?= 1
+HUF_FORCE_DECOMPRESS_X1 ?= 0
+HUF_FORCE_DECOMPRESS_X2 ?= 0
+ZSTD_FORCE_DECOMPRESS_SHORT ?= 0
+ZSTD_FORCE_DECOMPRESS_LONG ?= 0
+ZSTD_NO_INLINE ?= 0
+ZSTD_STRIP_ERROR_STRINGS ?= 0
 
 ifeq ($(ZSTD_LIB_COMPRESSION), 0)
 	ZSTD_LIB_DICTBUILDER = 0
@@ -72,6 +83,30 @@ ifneq ($(ZSTD_LIB_DICTBUILDER), 0)
 	ZSTD_FILES += $(ZDICT_FILES)
 endif
 
+ifneq ($(HUF_FORCE_DECOMPRESS_X1), 0)
+	CFLAGS += -DHUF_FORCE_DECOMPRESS_X1
+endif
+
+ifneq ($(HUF_FORCE_DECOMPRESS_X2), 0)
+	CFLAGS += -DHUF_FORCE_DECOMPRESS_X2
+endif
+
+ifneq ($(ZSTD_FORCE_DECOMPRESS_SHORT), 0)
+	CFLAGS += -DZSTD_FORCE_DECOMPRESS_SHORT
+endif
+
+ifneq ($(ZSTD_FORCE_DECOMPRESS_LONG), 0)
+	CFLAGS += -DZSTD_FORCE_DECOMPRESS_LONG
+endif
+
+ifneq ($(ZSTD_NO_INLINE), 0)
+	CFLAGS += -DZSTD_NO_INLINE
+endif
+
+ifneq ($(ZSTD_STRIP_ERROR_STRINGS), 0)
+	CFLAGS += -DZSTD_STRIP_ERROR_STRINGS
+endif
+
 ifneq ($(ZSTD_LEGACY_SUPPORT), 0)
 ifeq ($(shell test $(ZSTD_LEGACY_SUPPORT) -lt 8; echo $$?), 0)
 	ZSTD_FILES += $(shell ls legacy/*.c | $(GREP) 'v0[$(ZSTD_LEGACY_SUPPORT)-7]')
diff --git a/lib/README.md b/lib/README.md
index 0966c7aef497..0c9cd6d58f3f 100644
--- a/lib/README.md
+++ b/lib/README.md
@@ -7,13 +7,32 @@ in order to make it easier to select or exclude features.
 
 #### Building
 
-`Makefile` script is provided, supporting all standard [Makefile conventions](https://www.gnu.org/prep/standards/html_node/Makefile-Conventions.html#Makefile-Conventions),
+`Makefile` script is provided, supporting [Makefile conventions](https://www.gnu.org/prep/standards/html_node/Makefile-Conventions.html#Makefile-Conventions),
 including commands variables, staged install, directory variables and standard targets.
 - `make` : generates both static and dynamic libraries
-- `make install` : install libraries in default system directories
+- `make install` : install libraries and headers in target system directories
 
-`libzstd` default scope includes compression, decompression, dictionary building,
-and decoding support for legacy formats >= v0.5.0.
+`libzstd` default scope is pretty large, including compression, decompression, dictionary builder,
+and support for decoding legacy formats >= v0.5.0.
+The scope can be reduced on demand (see paragraph _modular build_).
+
+
+#### Multithreading support
+
+Multithreading is disabled by default when building with `make`.
+Enabling multithreading requires 2 conditions :
+- set build macro `ZSTD_MULTITHREAD` (`-DZSTD_MULTITHREAD` for `gcc`)
+- for POSIX systems : compile with pthread (`-pthread` compilation flag for `gcc`)
+
+Both conditions are automatically applied when invoking `make lib-mt` target.
+
+When linking a POSIX program with a multithreaded version of `libzstd`,
+note that it's necessary to request the `-pthread` flag during link stage.
+
+Multithreading capabilities are exposed
+via the [advanced API defined in `lib/zstd.h`](https://github.com/facebook/zstd/blob/v1.3.8/lib/zstd.h#L592).
+This API is still labelled experimental,
+but is expected to become "stable" in the near future.
 
 
 #### API
@@ -26,63 +45,70 @@ Zstandard's stable API is exposed within [lib/zstd.h](zstd.h).
 Optional advanced features are exposed via :
 
 - `lib/common/zstd_errors.h` : translates `size_t` function results
-                              into an `ZSTD_ErrorCode`, for accurate error handling.
+                               into a `ZSTD_ErrorCode`, for accurate error handling.
+
 - `ZSTD_STATIC_LINKING_ONLY` : if this macro is defined _before_ including `zstd.h`,
-                          it unlocks access to advanced experimental API,
-                          exposed in second part of `zstd.h`.
-                          These APIs are not "stable", their definition may change in the future.
-                          As a consequence, it shall ___never be used with dynamic library___ !
+                          it unlocks access to the experimental API,
+                          exposed in the second part of `zstd.h`.
+                          All definitions in the experimental APIs are unstable,
+                          they may still change in the future, or even be removed.
+                          As a consequence, experimental definitions shall ___never be used with dynamic library___ !
                           Only static linking is allowed.
 
 
 #### Modular build
 
-It's possible to compile only a limited set of features.
+It's possible to compile only a limited set of features within `libzstd`.
+The file structure is designed to make this selection manually achievable for any build system :
 
 - Directory `lib/common` is always required, for all variants.
+
 - Compression source code lies in `lib/compress`
+
 - Decompression source code lies in `lib/decompress`
+
 - It's possible to include only `compress` or only `decompress`, they don't depend on each other.
+
 - `lib/dictBuilder` : makes it possible to generate dictionaries from a set of samples.
         The API is exposed in `lib/dictBuilder/zdict.h`.
         This module depends on both `lib/common` and `lib/compress` .
-- `lib/legacy` : source code to decompress legacy zstd formats, starting from `v0.1.0`.
+
+- `lib/legacy` : makes it possible to decompress legacy zstd formats, starting from `v0.1.0`.
         This module depends on `lib/common` and `lib/decompress`.
         To enable this feature, define `ZSTD_LEGACY_SUPPORT` during compilation.
         Specifying a number limits versions supported to that version onward.
         For example, `ZSTD_LEGACY_SUPPORT=2` means : "support legacy formats >= v0.2.0".
-        `ZSTD_LEGACY_SUPPORT=3` means : "support legacy formats >= v0.3.0", and so on.
-        Currently, the default library setting is `ZST_LEGACY_SUPPORT=5`.
-        It can be changed at build by any other value.
-        Note that any number >= 8 translates into "do __not__ support legacy formats",
-        since all versions of `zstd` >= v0.8 are compatible with v1+ specification.
-        `ZSTD_LEGACY_SUPPORT=0` also means "do __not__ support legacy formats".
-        Once enabled, this capability is transparently triggered within decompression functions.
-        It's also possible to invoke directly legacy API, as exposed in `lib/legacy/zstd_legacy.h`.
-        Each version also provides an additional dedicated set of advanced API.
+        Conversely, `ZSTD_LEGACY_SUPPORT=0` means "do __not__ support legacy formats".
+        By default, this build macro is set as `ZSTD_LEGACY_SUPPORT=5`.
+        Decoding supported legacy format is a transparent capability triggered within decompression functions.
+        It's also allowed to invoke legacy API directly, exposed in `lib/legacy/zstd_legacy.h`.
+        Each version does also provide its own set of advanced API.
         For example, advanced API for version `v0.4` is exposed in `lib/legacy/zstd_v04.h` .
-        Note : `lib/legacy` only supports _decoding_ legacy formats.
-- Similarly, you can define `ZSTD_LIB_COMPRESSION, ZSTD_LIB_DECOMPRESSION`, `ZSTD_LIB_DICTBUILDER`,
-        and `ZSTD_LIB_DEPRECATED` as 0 to forgo compilation of the corresponding features. This will
-        also disable compilation of all dependencies (eg. `ZSTD_LIB_COMPRESSION=0` will also disable
-        dictBuilder).
 
+- While invoking `make libzstd`, it's possible to define build macros
+        `ZSTD_LIB_COMPRESSION, ZSTD_LIB_DECOMPRESSION`, `ZSTD_LIB_DICTBUILDER`,
+        and `ZSTD_LIB_DEPRECATED` as `0` to forgo compilation of the corresponding features.
+        This will also disable compilation of all dependencies
+        (eg. `ZSTD_LIB_COMPRESSION=0` will also disable dictBuilder).
 
-#### Multithreading support
+- There are some additional build macros that can be used to minify the decoder.
 
-Multithreading is disabled by default when building with `make`.
-Enabling multithreading requires 2 conditions :
-- set macro `ZSTD_MULTITHREAD`
-- on POSIX systems : compile with pthread (`-pthread` compilation flag for `gcc`)
+  Zstandard often has more than one implementation of a piece of functionality,
+  where each implementation optimizes for different scenarios. For example, the
+  Huffman decoder has complementary implementations that decode the stream one
+  symbol at a time or two symbols at a time. Zstd normally includes both (and
+  dispatches between them at runtime), but by defining `HUF_FORCE_DECOMPRESS_X1`
+  or `HUF_FORCE_DECOMPRESS_X2`, you can force the use of one or the other, avoiding
+  compilation of the other. Similarly, `ZSTD_FORCE_DECOMPRESS_SEQUENCES_SHORT`
+  and `ZSTD_FORCE_DECOMPRESS_SEQUENCES_LONG` force the compilation and use of
+  only one or the other of two decompression implementations. The smallest
+  binary is achieved by using `HUF_FORCE_DECOMPRESS_X1` and
+  `ZSTD_FORCE_DECOMPRESS_SEQUENCES_SHORT`.
 
-Both conditions are automatically triggered by invoking `make lib-mt` target.
-Note that, when linking a POSIX program with a multithreaded version of `libzstd`,
-it's necessary to trigger `-pthread` flag during link stage.
-
-Multithreading capabilities are exposed
-via [advanced API `ZSTD_compress_generic()` defined in `lib/zstd.h`](https://github.com/facebook/zstd/blob/dev/lib/zstd.h#L919).
-This API is still considered experimental,
-but is expected to become "stable" at some point in the future.
+  For squeezing the last ounce of size out, you can also define
+  `ZSTD_NO_INLINE`, which disables inlining, and `ZSTD_STRIP_ERROR_STRINGS`,
+  which removes the error messages that are otherwise returned by
+  `ZSTD_getErrorName`.
 
 
 #### Windows : using MinGW+MSYS to create DLL
@@ -113,8 +139,8 @@ Consider migrating code towards supported streaming API exposed in `zstd.h`.
 
 The other files are not source code. There are :
 
- - `LICENSE` : contains the BSD license text
- - `Makefile` : `make` script to build and install zstd library (static and dynamic)
  - `BUCK` : support for `buck` build system (https://buckbuild.com/)
- - `libzstd.pc.in` : for `pkg-config` (used in `make install`)
+ - `Makefile` : `make` script to build and install zstd library (static and dynamic)
  - `README.md` : this file
+ - `dll/` : resources directory for Windows compilation
+ - `libzstd.pc.in` : script for `pkg-config` (used in `make install`)
diff --git a/lib/common/bitstream.h b/lib/common/bitstream.h
index ef89b9878e22..d955bd677b5b 100644
--- a/lib/common/bitstream.h
+++ b/lib/common/bitstream.h
@@ -389,7 +389,7 @@ MEM_STATIC void BIT_skipBits(BIT_DStream_t* bitD, U32 nbBits)
  *  Read (consume) next n bits from local register and update.
  *  Pay attention to not read more than nbBits contained into local register.
  * @return : extracted value. */
-MEM_STATIC size_t BIT_readBits(BIT_DStream_t* bitD, U32 nbBits)
+MEM_STATIC size_t BIT_readBits(BIT_DStream_t* bitD, unsigned nbBits)
 {
     size_t const value = BIT_lookBits(bitD, nbBits);
     BIT_skipBits(bitD, nbBits);
@@ -398,7 +398,7 @@ MEM_STATIC size_t BIT_readBits(BIT_DStream_t* bitD, U32 nbBits)
 
 /*! BIT_readBitsFast() :
  *  unsafe version; only works only if nbBits >= 1 */
-MEM_STATIC size_t BIT_readBitsFast(BIT_DStream_t* bitD, U32 nbBits)
+MEM_STATIC size_t BIT_readBitsFast(BIT_DStream_t* bitD, unsigned nbBits)
 {
     size_t const value = BIT_lookBitsFast(bitD, nbBits);
     assert(nbBits >= 1);
diff --git a/lib/common/compiler.h b/lib/common/compiler.h
index 07f875e4d38e..7f561282ca35 100644
--- a/lib/common/compiler.h
+++ b/lib/common/compiler.h
@@ -15,6 +15,8 @@
 *  Compiler specifics
 *********************************************************/
 /* force inlining */
+
+#if !defined(ZSTD_NO_INLINE)
 #if defined (__GNUC__) || defined(__cplusplus) || defined(__STDC_VERSION__) && __STDC_VERSION__ >= 199901L   /* C99 */
 #  define INLINE_KEYWORD inline
 #else
@@ -29,6 +31,13 @@
 #  define FORCE_INLINE_ATTR
 #endif
 
+#else
+
+#define INLINE_KEYWORD
+#define FORCE_INLINE_ATTR
+
+#endif
+
 /**
  * FORCE_INLINE_TEMPLATE is used to define C "templates", which take constant
  * parameters. They must be inlined for the compiler to elimininate the constant
@@ -89,23 +98,21 @@
 #endif
 
 /* prefetch
- * can be disabled, by declaring NO_PREFETCH macro
- * All prefetch invocations use a single default locality 2,
- * generating instruction prefetcht1,
- * which, according to Intel, means "load data into L2 cache".
- * This is a good enough "middle ground" for the time being,
- * though in theory, it would be better to specialize locality depending on data being prefetched.
- * Tests could not determine any sensible difference based on locality value. */
+ * can be disabled, by declaring NO_PREFETCH build macro */
 #if defined(NO_PREFETCH)
-#  define PREFETCH(ptr)     (void)(ptr)  /* disabled */
+#  define PREFETCH_L1(ptr)  (void)(ptr)  /* disabled */
+#  define PREFETCH_L2(ptr)  (void)(ptr)  /* disabled */
 #else
 #  if defined(_MSC_VER) && (defined(_M_X64) || defined(_M_I86))  /* _mm_prefetch() is not defined outside of x86/x64 */
 #    include    /* https://msdn.microsoft.com/fr-fr/library/84szxsww(v=vs.90).aspx */
-#    define PREFETCH(ptr)   _mm_prefetch((const char*)(ptr), _MM_HINT_T1)
+#    define PREFETCH_L1(ptr)  _mm_prefetch((const char*)(ptr), _MM_HINT_T0)
+#    define PREFETCH_L2(ptr)  _mm_prefetch((const char*)(ptr), _MM_HINT_T1)
 #  elif defined(__GNUC__) && ( (__GNUC__ >= 4) || ( (__GNUC__ == 3) && (__GNUC_MINOR__ >= 1) ) )
-#    define PREFETCH(ptr)   __builtin_prefetch((ptr), 0 /* rw==read */, 2 /* locality */)
+#    define PREFETCH_L1(ptr)  __builtin_prefetch((ptr), 0 /* rw==read */, 3 /* locality */)
+#    define PREFETCH_L2(ptr)  __builtin_prefetch((ptr), 0 /* rw==read */, 2 /* locality */)
 #  else
-#    define PREFETCH(ptr)   (void)(ptr)  /* disabled */
+#    define PREFETCH_L1(ptr) (void)(ptr)  /* disabled */
+#    define PREFETCH_L2(ptr) (void)(ptr)  /* disabled */
 #  endif
 #endif  /* NO_PREFETCH */
 
@@ -116,7 +123,7 @@
     size_t const _size = (size_t)(s);     \
     size_t _pos;                          \
     for (_pos=0; _pos<_size; _pos+=CACHELINE_SIZE) {  \
-        PREFETCH(_ptr + _pos);            \
+        PREFETCH_L2(_ptr + _pos);         \
     }                                     \
 }
 
diff --git a/lib/common/cpu.h b/lib/common/cpu.h
index eeb428ad5f6a..5f0923fc9289 100644
--- a/lib/common/cpu.h
+++ b/lib/common/cpu.h
@@ -78,7 +78,7 @@ MEM_STATIC ZSTD_cpuid_t ZSTD_cpuid(void) {
       __asm__(
           "pushl %%ebx\n\t"
           "cpuid\n\t"
-          "movl %%ebx, %%eax\n\r"
+          "movl %%ebx, %%eax\n\t"
           "popl %%ebx"
           : "=a"(f7b), "=c"(f7c)
           : "a"(7), "c"(0)
diff --git a/lib/common/debug.h b/lib/common/debug.h
index 0c04ad2cc98c..b4fc89d49741 100644
--- a/lib/common/debug.h
+++ b/lib/common/debug.h
@@ -57,9 +57,9 @@ extern "C" {
 #endif
 
 
-/* static assert is triggered at compile time, leaving no runtime artefact,
- * but can only work with compile-time constants.
- * This variant can only be used inside a function. */
+/* static assert is triggered at compile time, leaving no runtime artefact.
+ * static assert only works with compile-time constants.
+ * Also, this variant can only be used inside a function. */
 #define DEBUG_STATIC_ASSERT(c) (void)sizeof(char[(c) ? 1 : -1])
 
 
@@ -70,9 +70,19 @@ extern "C" {
 #  define DEBUGLEVEL 0
 #endif
 
+
+/* DEBUGFILE can be defined externally,
+ * typically through compiler command line.
+ * note : currently useless.
+ * Value must be stderr or stdout */
+#ifndef DEBUGFILE
+#  define DEBUGFILE stderr
+#endif
+
+
 /* recommended values for DEBUGLEVEL :
- * 0 : no debug, all run-time functions disabled
- * 1 : no display, enables assert() only
+ * 0 : release mode, no debug, all run-time checks disabled
+ * 1 : enables assert() only, no display
  * 2 : reserved, for currently active debug path
  * 3 : events once per object lifetime (CCtx, CDict, etc.)
  * 4 : events once per frame
@@ -81,7 +91,7 @@ extern "C" {
  * 7+: events at every position (*very* verbose)
  *
  * It's generally inconvenient to output traces > 5.
- * In which case, it's possible to selectively enable higher verbosity levels
+ * In which case, it's possible to selectively trigger high verbosity levels
  * by modifying g_debug_level.
  */
 
@@ -95,11 +105,12 @@ extern "C" {
 
 #if (DEBUGLEVEL>=2)
 #  include 
-extern int g_debuglevel; /* here, this variable is only declared,
-                           it actually lives in debug.c,
-                           and is shared by the whole process.
-                           It's typically used to enable very verbose levels
-                           on selective conditions (such as position in src) */
+extern int g_debuglevel; /* the variable is only declared,
+                            it actually lives in debug.c,
+                            and is shared by the whole process.
+                            It's not thread-safe.
+                            It's useful when enabling very verbose levels
+                            on selective conditions (such as position in src) */
 
 #  define RAWLOG(l, ...) {                                      \
                 if (l<=g_debuglevel) {                          \
diff --git a/lib/common/error_private.c b/lib/common/error_private.c
index d004ee636c67..7c1bb67a23f4 100644
--- a/lib/common/error_private.c
+++ b/lib/common/error_private.c
@@ -14,6 +14,10 @@
 
 const char* ERR_getErrorString(ERR_enum code)
 {
+#ifdef ZSTD_STRIP_ERROR_STRINGS
+    (void)code;
+    return "Error strings stripped";
+#else
     static const char* const notErrorCode = "Unspecified error code";
     switch( code )
     {
@@ -39,10 +43,12 @@ const char* ERR_getErrorString(ERR_enum code)
     case PREFIX(dictionaryCreation_failed): return "Cannot create Dictionary from provided samples";
     case PREFIX(dstSize_tooSmall): return "Destination buffer is too small";
     case PREFIX(srcSize_wrong): return "Src size is incorrect";
+    case PREFIX(dstBuffer_null): return "Operation on NULL destination buffer";
         /* following error codes are not stable and may be removed or changed in a future version */
     case PREFIX(frameIndex_tooLarge): return "Frame index is too large";
     case PREFIX(seekableIO): return "An I/O error occurred when reading/seeking";
     case PREFIX(maxCode):
     default: return notErrorCode;
     }
+#endif
 }
diff --git a/lib/common/fse.h b/lib/common/fse.h
index a5a6b6d4db70..f72c519b259d 100644
--- a/lib/common/fse.h
+++ b/lib/common/fse.h
@@ -512,7 +512,7 @@ MEM_STATIC void FSE_initCState(FSE_CState_t* statePtr, const FSE_CTable* ct)
     const U32 tableLog = MEM_read16(ptr);
     statePtr->value = (ptrdiff_t)1<stateTable = u16ptr+2;
-    statePtr->symbolTT = ((const U32*)ct + 1 + (tableLog ? (1<<(tableLog-1)) : 1));
+    statePtr->symbolTT = ct + 1 + (tableLog ? (1<<(tableLog-1)) : 1);
     statePtr->stateLog = tableLog;
 }
 
@@ -531,7 +531,7 @@ MEM_STATIC void FSE_initCState2(FSE_CState_t* statePtr, const FSE_CTable* ct, U3
     }
 }
 
-MEM_STATIC void FSE_encodeSymbol(BIT_CStream_t* bitC, FSE_CState_t* statePtr, U32 symbol)
+MEM_STATIC void FSE_encodeSymbol(BIT_CStream_t* bitC, FSE_CState_t* statePtr, unsigned symbol)
 {
     FSE_symbolCompressionTransform const symbolTT = ((const FSE_symbolCompressionTransform*)(statePtr->symbolTT))[symbol];
     const U16* const stateTable = (const U16*)(statePtr->stateTable);
diff --git a/lib/common/huf.h b/lib/common/huf.h
index de9464111064..6b572c448d9d 100644
--- a/lib/common/huf.h
+++ b/lib/common/huf.h
@@ -173,15 +173,19 @@ typedef U32 HUF_DTable;
 *  Advanced decompression functions
 ******************************************/
 size_t HUF_decompress4X1 (void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize);   /**< single-symbol decoder */
+#ifndef HUF_FORCE_DECOMPRESS_X1
 size_t HUF_decompress4X2 (void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize);   /**< double-symbols decoder */
+#endif
 
 size_t HUF_decompress4X_DCtx (HUF_DTable* dctx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize);   /**< decodes RLE and uncompressed */
 size_t HUF_decompress4X_hufOnly(HUF_DTable* dctx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize); /**< considers RLE and uncompressed as errors */
 size_t HUF_decompress4X_hufOnly_wksp(HUF_DTable* dctx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize, void* workSpace, size_t wkspSize); /**< considers RLE and uncompressed as errors */
 size_t HUF_decompress4X1_DCtx(HUF_DTable* dctx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize);   /**< single-symbol decoder */
 size_t HUF_decompress4X1_DCtx_wksp(HUF_DTable* dctx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize, void* workSpace, size_t wkspSize);   /**< single-symbol decoder */
+#ifndef HUF_FORCE_DECOMPRESS_X1
 size_t HUF_decompress4X2_DCtx(HUF_DTable* dctx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize);   /**< double-symbols decoder */
 size_t HUF_decompress4X2_DCtx_wksp(HUF_DTable* dctx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize, void* workSpace, size_t wkspSize);   /**< double-symbols decoder */
+#endif
 
 
 /* ****************************************
@@ -228,7 +232,7 @@ size_t HUF_compress4X_repeat(void* dst, size_t dstSize,
 #define HUF_CTABLE_WORKSPACE_SIZE_U32 (2*HUF_SYMBOLVALUE_MAX +1 +1)
 #define HUF_CTABLE_WORKSPACE_SIZE (HUF_CTABLE_WORKSPACE_SIZE_U32 * sizeof(unsigned))
 size_t HUF_buildCTable_wksp (HUF_CElt* tree,
-                       const U32* count, U32 maxSymbolValue, U32 maxNbBits,
+                       const unsigned* count, U32 maxSymbolValue, U32 maxNbBits,
                              void* workSpace, size_t wkspSize);
 
 /*! HUF_readStats() :
@@ -277,14 +281,22 @@ U32 HUF_selectDecoder (size_t dstSize, size_t cSrcSize);
 #define HUF_DECOMPRESS_WORKSPACE_SIZE (2 << 10)
 #define HUF_DECOMPRESS_WORKSPACE_SIZE_U32 (HUF_DECOMPRESS_WORKSPACE_SIZE / sizeof(U32))
 
+#ifndef HUF_FORCE_DECOMPRESS_X2
 size_t HUF_readDTableX1 (HUF_DTable* DTable, const void* src, size_t srcSize);
 size_t HUF_readDTableX1_wksp (HUF_DTable* DTable, const void* src, size_t srcSize, void* workSpace, size_t wkspSize);
+#endif
+#ifndef HUF_FORCE_DECOMPRESS_X1
 size_t HUF_readDTableX2 (HUF_DTable* DTable, const void* src, size_t srcSize);
 size_t HUF_readDTableX2_wksp (HUF_DTable* DTable, const void* src, size_t srcSize, void* workSpace, size_t wkspSize);
+#endif
 
 size_t HUF_decompress4X_usingDTable(void* dst, size_t maxDstSize, const void* cSrc, size_t cSrcSize, const HUF_DTable* DTable);
+#ifndef HUF_FORCE_DECOMPRESS_X2
 size_t HUF_decompress4X1_usingDTable(void* dst, size_t maxDstSize, const void* cSrc, size_t cSrcSize, const HUF_DTable* DTable);
+#endif
+#ifndef HUF_FORCE_DECOMPRESS_X1
 size_t HUF_decompress4X2_usingDTable(void* dst, size_t maxDstSize, const void* cSrc, size_t cSrcSize, const HUF_DTable* DTable);
+#endif
 
 
 /* ====================== */
@@ -306,24 +318,36 @@ size_t HUF_compress1X_repeat(void* dst, size_t dstSize,
                        HUF_CElt* hufTable, HUF_repeat* repeat, int preferRepeat, int bmi2);
 
 size_t HUF_decompress1X1 (void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize);   /* single-symbol decoder */
+#ifndef HUF_FORCE_DECOMPRESS_X1
 size_t HUF_decompress1X2 (void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize);   /* double-symbol decoder */
+#endif
 
 size_t HUF_decompress1X_DCtx (HUF_DTable* dctx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize);
 size_t HUF_decompress1X_DCtx_wksp (HUF_DTable* dctx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize, void* workSpace, size_t wkspSize);
+#ifndef HUF_FORCE_DECOMPRESS_X2
 size_t HUF_decompress1X1_DCtx(HUF_DTable* dctx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize);   /**< single-symbol decoder */
 size_t HUF_decompress1X1_DCtx_wksp(HUF_DTable* dctx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize, void* workSpace, size_t wkspSize);   /**< single-symbol decoder */
+#endif
+#ifndef HUF_FORCE_DECOMPRESS_X1
 size_t HUF_decompress1X2_DCtx(HUF_DTable* dctx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize);   /**< double-symbols decoder */
 size_t HUF_decompress1X2_DCtx_wksp(HUF_DTable* dctx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize, void* workSpace, size_t wkspSize);   /**< double-symbols decoder */
+#endif
 
 size_t HUF_decompress1X_usingDTable(void* dst, size_t maxDstSize, const void* cSrc, size_t cSrcSize, const HUF_DTable* DTable);   /**< automatic selection of sing or double symbol decoder, based on DTable */
+#ifndef HUF_FORCE_DECOMPRESS_X2
 size_t HUF_decompress1X1_usingDTable(void* dst, size_t maxDstSize, const void* cSrc, size_t cSrcSize, const HUF_DTable* DTable);
+#endif
+#ifndef HUF_FORCE_DECOMPRESS_X1
 size_t HUF_decompress1X2_usingDTable(void* dst, size_t maxDstSize, const void* cSrc, size_t cSrcSize, const HUF_DTable* DTable);
+#endif
 
 /* BMI2 variants.
  * If the CPU has BMI2 support, pass bmi2=1, otherwise pass bmi2=0.
  */
 size_t HUF_decompress1X_usingDTable_bmi2(void* dst, size_t maxDstSize, const void* cSrc, size_t cSrcSize, const HUF_DTable* DTable, int bmi2);
+#ifndef HUF_FORCE_DECOMPRESS_X2
 size_t HUF_decompress1X1_DCtx_wksp_bmi2(HUF_DTable* dctx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize, void* workSpace, size_t wkspSize, int bmi2);
+#endif
 size_t HUF_decompress4X_usingDTable_bmi2(void* dst, size_t maxDstSize, const void* cSrc, size_t cSrcSize, const HUF_DTable* DTable, int bmi2);
 size_t HUF_decompress4X_hufOnly_wksp_bmi2(HUF_DTable* dctx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize, void* workSpace, size_t wkspSize, int bmi2);
 
diff --git a/lib/common/pool.c b/lib/common/pool.c
index 281b3824ac4d..7a8294543280 100644
--- a/lib/common/pool.c
+++ b/lib/common/pool.c
@@ -88,8 +88,8 @@ static void* POOL_thread(void* opaque) {
             ctx->numThreadsBusy++;
             ctx->queueEmpty = ctx->queueHead == ctx->queueTail;
             /* Unlock the mutex, signal a pusher, and run the job */
-            ZSTD_pthread_mutex_unlock(&ctx->queueMutex);
             ZSTD_pthread_cond_signal(&ctx->queuePushCond);
+            ZSTD_pthread_mutex_unlock(&ctx->queueMutex);
 
             job.function(job.opaque);
 
diff --git a/lib/common/zstd_common.c b/lib/common/zstd_common.c
index 6f05d240e43c..667f4a27fc69 100644
--- a/lib/common/zstd_common.c
+++ b/lib/common/zstd_common.c
@@ -30,8 +30,10 @@ const char* ZSTD_versionString(void) { return ZSTD_VERSION_STRING; }
 /*-****************************************
 *  ZSTD Error Management
 ******************************************/
+#undef ZSTD_isError   /* defined within zstd_internal.h */
 /*! ZSTD_isError() :
- *  tells if a return value is an error code */
+ *  tells if a return value is an error code
+ *  symbol is required for external callers */
 unsigned ZSTD_isError(size_t code) { return ERR_isError(code); }
 
 /*! ZSTD_getErrorName() :
diff --git a/lib/common/zstd_errors.h b/lib/common/zstd_errors.h
index 57533f28696b..92a3433896c5 100644
--- a/lib/common/zstd_errors.h
+++ b/lib/common/zstd_errors.h
@@ -72,6 +72,7 @@ typedef enum {
   ZSTD_error_workSpace_tooSmall= 66,
   ZSTD_error_dstSize_tooSmall = 70,
   ZSTD_error_srcSize_wrong    = 72,
+  ZSTD_error_dstBuffer_null   = 74,
   /* following error codes are __NOT STABLE__, they can be removed or changed in future versions */
   ZSTD_error_frameIndex_tooLarge = 100,
   ZSTD_error_seekableIO          = 102,
diff --git a/lib/common/zstd_internal.h b/lib/common/zstd_internal.h
index e75adfa61323..edeb74b9c3a2 100644
--- a/lib/common/zstd_internal.h
+++ b/lib/common/zstd_internal.h
@@ -41,6 +41,9 @@ extern "C" {
 
 /* ---- static assert (debug) --- */
 #define ZSTD_STATIC_ASSERT(c) DEBUG_STATIC_ASSERT(c)
+#define ZSTD_isError ERR_isError   /* for inlining */
+#define FSE_isError  ERR_isError
+#define HUF_isError  ERR_isError
 
 
 /*-*************************************
@@ -75,7 +78,6 @@ static const U32 repStartValue[ZSTD_REP_NUM] = { 1, 4, 8 };
 #define BIT0   1
 
 #define ZSTD_WINDOWLOG_ABSOLUTEMIN 10
-#define ZSTD_WINDOWLOG_DEFAULTMAX 27 /* Default maximum allowed window log */
 static const size_t ZSTD_fcs_fieldSize[4] = { 0, 2, 4, 8 };
 static const size_t ZSTD_did_fieldSize[4] = { 0, 1, 2, 4 };
 
@@ -242,7 +244,7 @@ typedef struct {
     blockType_e blockType;
     U32 lastBlock;
     U32 origSize;
-} blockProperties_t;
+} blockProperties_t;   /* declared here for decompress and fullbench */
 
 /*! ZSTD_getcBlockSize() :
  *  Provides the size of compressed block from block header `src` */
@@ -250,6 +252,13 @@ typedef struct {
 size_t ZSTD_getcBlockSize(const void* src, size_t srcSize,
                           blockProperties_t* bpPtr);
 
+/*! ZSTD_decodeSeqHeaders() :
+ *  decode sequence header from src */
+/* Used by: decompress, fullbench (does not get its definition from here) */
+size_t ZSTD_decodeSeqHeaders(ZSTD_DCtx* dctx, int* nbSeqPtr,
+                       const void* src, size_t srcSize);
+
+
 #if defined (__cplusplus)
 }
 #endif
diff --git a/lib/compress/fse_compress.c b/lib/compress/fse_compress.c
index 4408f0ed5b50..60f357bbd24e 100644
--- a/lib/compress/fse_compress.c
+++ b/lib/compress/fse_compress.c
@@ -115,7 +115,7 @@ size_t FSE_buildCTable_wksp(FSE_CTable* ct,
     /* symbol start positions */
     {   U32 u;
         cumul[0] = 0;
-        for (u=1; u<=maxSymbolValue+1; u++) {
+        for (u=1; u <= maxSymbolValue+1; u++) {
             if (normalizedCounter[u-1]==-1) {  /* Low proba symbol */
                 cumul[u] = cumul[u-1] + 1;
                 tableSymbol[highThreshold--] = (FSE_FUNCTION_TYPE)(u-1);
@@ -658,7 +658,7 @@ size_t FSE_compress_wksp (void* dst, size_t dstSize, const void* src, size_t src
     BYTE* op = ostart;
     BYTE* const oend = ostart + dstSize;
 
-    U32   count[FSE_MAX_SYMBOL_VALUE+1];
+    unsigned count[FSE_MAX_SYMBOL_VALUE+1];
     S16   norm[FSE_MAX_SYMBOL_VALUE+1];
     FSE_CTable* CTable = (FSE_CTable*)workSpace;
     size_t const CTableSize = FSE_CTABLE_SIZE_U32(tableLog, maxSymbolValue);
@@ -672,7 +672,7 @@ size_t FSE_compress_wksp (void* dst, size_t dstSize, const void* src, size_t src
     if (!tableLog) tableLog = FSE_DEFAULT_TABLELOG;
 
     /* Scan input and build symbol stats */
-    {   CHECK_V_F(maxCount, HIST_count_wksp(count, &maxSymbolValue, src, srcSize, (unsigned*)scratchBuffer) );
+    {   CHECK_V_F(maxCount, HIST_count_wksp(count, &maxSymbolValue, src, srcSize, scratchBuffer, scratchBufferSize) );
         if (maxCount == srcSize) return 1;   /* only a single symbol in src : rle */
         if (maxCount == 1) return 0;         /* each symbol present maximum once => not compressible */
         if (maxCount < (srcSize >> 7)) return 0;   /* Heuristic : not compressible enough */
diff --git a/lib/compress/hist.c b/lib/compress/hist.c
index 16524756b8dc..45b7babc1e23 100644
--- a/lib/compress/hist.c
+++ b/lib/compress/hist.c
@@ -73,6 +73,7 @@ unsigned HIST_count_simple(unsigned* count, unsigned* maxSymbolValuePtr,
     return largestCount;
 }
 
+typedef enum { trustInput, checkMaxSymbolValue } HIST_checkInput_e;
 
 /* HIST_count_parallel_wksp() :
  * store histogram into 4 intermediate tables, recombined at the end.
@@ -85,8 +86,8 @@ unsigned HIST_count_simple(unsigned* count, unsigned* maxSymbolValuePtr,
 static size_t HIST_count_parallel_wksp(
                                 unsigned* count, unsigned* maxSymbolValuePtr,
                                 const void* source, size_t sourceSize,
-                                unsigned checkMax,
-                                unsigned* const workSpace)
+                                HIST_checkInput_e check,
+                                U32* const workSpace)
 {
     const BYTE* ip = (const BYTE*)source;
     const BYTE* const iend = ip+sourceSize;
@@ -137,7 +138,7 @@ static size_t HIST_count_parallel_wksp(
     /* finish last symbols */
     while (ipmaxSymbolValue; s--) {
             Counting1[s] += Counting2[s] + Counting3[s] + Counting4[s];
             if (Counting1[s]) return ERROR(maxSymbolValue_tooSmall);
@@ -157,14 +158,18 @@ static size_t HIST_count_parallel_wksp(
 
 /* HIST_countFast_wksp() :
  * Same as HIST_countFast(), but using an externally provided scratch buffer.
- * `workSpace` size must be table of >= HIST_WKSP_SIZE_U32 unsigned */
+ * `workSpace` is a writable buffer which must be 4-bytes aligned,
+ * `workSpaceSize` must be >= HIST_WKSP_SIZE
+ */
 size_t HIST_countFast_wksp(unsigned* count, unsigned* maxSymbolValuePtr,
                           const void* source, size_t sourceSize,
-                          unsigned* workSpace)
+                          void* workSpace, size_t workSpaceSize)
 {
     if (sourceSize < 1500) /* heuristic threshold */
         return HIST_count_simple(count, maxSymbolValuePtr, source, sourceSize);
-    return HIST_count_parallel_wksp(count, maxSymbolValuePtr, source, sourceSize, 0, workSpace);
+    if ((size_t)workSpace & 3) return ERROR(GENERIC);  /* must be aligned on 4-bytes boundaries */
+    if (workSpaceSize < HIST_WKSP_SIZE) return ERROR(workSpace_tooSmall);
+    return HIST_count_parallel_wksp(count, maxSymbolValuePtr, source, sourceSize, trustInput, (U32*)workSpace);
 }
 
 /* fast variant (unsafe : won't check if src contains values beyond count[] limit) */
@@ -172,24 +177,27 @@ size_t HIST_countFast(unsigned* count, unsigned* maxSymbolValuePtr,
                      const void* source, size_t sourceSize)
 {
     unsigned tmpCounters[HIST_WKSP_SIZE_U32];
-    return HIST_countFast_wksp(count, maxSymbolValuePtr, source, sourceSize, tmpCounters);
+    return HIST_countFast_wksp(count, maxSymbolValuePtr, source, sourceSize, tmpCounters, sizeof(tmpCounters));
 }
 
 /* HIST_count_wksp() :
  * Same as HIST_count(), but using an externally provided scratch buffer.
  * `workSpace` size must be table of >= HIST_WKSP_SIZE_U32 unsigned */
 size_t HIST_count_wksp(unsigned* count, unsigned* maxSymbolValuePtr,
-                 const void* source, size_t sourceSize, unsigned* workSpace)
+                       const void* source, size_t sourceSize,
+                       void* workSpace, size_t workSpaceSize)
 {
+    if ((size_t)workSpace & 3) return ERROR(GENERIC);  /* must be aligned on 4-bytes boundaries */
+    if (workSpaceSize < HIST_WKSP_SIZE) return ERROR(workSpace_tooSmall);
     if (*maxSymbolValuePtr < 255)
-        return HIST_count_parallel_wksp(count, maxSymbolValuePtr, source, sourceSize, 1, workSpace);
+        return HIST_count_parallel_wksp(count, maxSymbolValuePtr, source, sourceSize, checkMaxSymbolValue, (U32*)workSpace);
     *maxSymbolValuePtr = 255;
-    return HIST_countFast_wksp(count, maxSymbolValuePtr, source, sourceSize, workSpace);
+    return HIST_countFast_wksp(count, maxSymbolValuePtr, source, sourceSize, workSpace, workSpaceSize);
 }
 
 size_t HIST_count(unsigned* count, unsigned* maxSymbolValuePtr,
                  const void* src, size_t srcSize)
 {
     unsigned tmpCounters[HIST_WKSP_SIZE_U32];
-    return HIST_count_wksp(count, maxSymbolValuePtr, src, srcSize, tmpCounters);
+    return HIST_count_wksp(count, maxSymbolValuePtr, src, srcSize, tmpCounters, sizeof(tmpCounters));
 }
diff --git a/lib/compress/hist.h b/lib/compress/hist.h
index 8b1991a90bd3..8b389358dc10 100644
--- a/lib/compress/hist.h
+++ b/lib/compress/hist.h
@@ -41,11 +41,11 @@
 
 /*! HIST_count():
  *  Provides the precise count of each byte within a table 'count'.
- *  'count' is a table of unsigned int, of minimum size (*maxSymbolValuePtr+1).
+ * 'count' is a table of unsigned int, of minimum size (*maxSymbolValuePtr+1).
  *  Updates *maxSymbolValuePtr with actual largest symbol value detected.
- *  @return : count of the most frequent symbol (which isn't identified).
- *            or an error code, which can be tested using HIST_isError().
- *            note : if return == srcSize, there is only one symbol.
+ * @return : count of the most frequent symbol (which isn't identified).
+ *           or an error code, which can be tested using HIST_isError().
+ *           note : if return == srcSize, there is only one symbol.
  */
 size_t HIST_count(unsigned* count, unsigned* maxSymbolValuePtr,
                   const void* src, size_t srcSize);
@@ -56,14 +56,16 @@ unsigned HIST_isError(size_t code);  /**< tells if a return value is an error co
 /* --- advanced histogram functions --- */
 
 #define HIST_WKSP_SIZE_U32 1024
+#define HIST_WKSP_SIZE    (HIST_WKSP_SIZE_U32 * sizeof(unsigned))
 /** HIST_count_wksp() :
  *  Same as HIST_count(), but using an externally provided scratch buffer.
  *  Benefit is this function will use very little stack space.
- * `workSpace` must be a table of unsigned of size >= HIST_WKSP_SIZE_U32
+ * `workSpace` is a writable buffer which must be 4-bytes aligned,
+ * `workSpaceSize` must be >= HIST_WKSP_SIZE
  */
 size_t HIST_count_wksp(unsigned* count, unsigned* maxSymbolValuePtr,
                        const void* src, size_t srcSize,
-                       unsigned* workSpace);
+                       void* workSpace, size_t workSpaceSize);
 
 /** HIST_countFast() :
  *  same as HIST_count(), but blindly trusts that all byte values within src are <= *maxSymbolValuePtr.
@@ -74,11 +76,12 @@ size_t HIST_countFast(unsigned* count, unsigned* maxSymbolValuePtr,
 
 /** HIST_countFast_wksp() :
  *  Same as HIST_countFast(), but using an externally provided scratch buffer.
- * `workSpace` must be a table of unsigned of size >= HIST_WKSP_SIZE_U32
+ * `workSpace` is a writable buffer which must be 4-bytes aligned,
+ * `workSpaceSize` must be >= HIST_WKSP_SIZE
  */
 size_t HIST_countFast_wksp(unsigned* count, unsigned* maxSymbolValuePtr,
                            const void* src, size_t srcSize,
-                           unsigned* workSpace);
+                           void* workSpace, size_t workSpaceSize);
 
 /*! HIST_count_simple() :
  *  Same as HIST_countFast(), this function is unsafe,
diff --git a/lib/compress/huf_compress.c b/lib/compress/huf_compress.c
index 4c40572f2284..f074f1e0a95d 100644
--- a/lib/compress/huf_compress.c
+++ b/lib/compress/huf_compress.c
@@ -88,13 +88,13 @@ static size_t HUF_compressWeights (void* dst, size_t dstSize, const void* weight
     BYTE* op = ostart;
     BYTE* const oend = ostart + dstSize;
 
-    U32 maxSymbolValue = HUF_TABLELOG_MAX;
+    unsigned maxSymbolValue = HUF_TABLELOG_MAX;
     U32 tableLog = MAX_FSE_TABLELOG_FOR_HUFF_HEADER;
 
     FSE_CTable CTable[FSE_CTABLE_SIZE_U32(MAX_FSE_TABLELOG_FOR_HUFF_HEADER, HUF_TABLELOG_MAX)];
     BYTE scratchBuffer[1< HUF_BLOCKSIZE_MAX) return ERROR(srcSize_wrong);   /* current block size limit */
@@ -664,11 +665,11 @@ static size_t HUF_compress_internal (
     if (preferRepeat && repeat && *repeat == HUF_repeat_valid) {
         return HUF_compressCTable_internal(ostart, op, oend,
                                            src, srcSize,
-                                           singleStream, oldHufTable, bmi2);
+                                           nbStreams, oldHufTable, bmi2);
     }
 
     /* Scan input and build symbol stats */
-    {   CHECK_V_F(largest, HIST_count_wksp (table->count, &maxSymbolValue, (const BYTE*)src, srcSize, table->count) );
+    {   CHECK_V_F(largest, HIST_count_wksp (table->count, &maxSymbolValue, (const BYTE*)src, srcSize, workSpace, wkspSize) );
         if (largest == srcSize) { *ostart = ((const BYTE*)src)[0]; return 1; }   /* single symbol, rle */
         if (largest <= (srcSize >> 7)+4) return 0;   /* heuristic : probably not compressible enough */
     }
@@ -683,14 +684,15 @@ static size_t HUF_compress_internal (
     if (preferRepeat && repeat && *repeat != HUF_repeat_none) {
         return HUF_compressCTable_internal(ostart, op, oend,
                                            src, srcSize,
-                                           singleStream, oldHufTable, bmi2);
+                                           nbStreams, oldHufTable, bmi2);
     }
 
     /* Build Huffman Tree */
     huffLog = HUF_optimalTableLog(huffLog, srcSize, maxSymbolValue);
-    {   CHECK_V_F(maxBits, HUF_buildCTable_wksp(table->CTable, table->count,
-                                                maxSymbolValue, huffLog,
-                                                table->nodeTable, sizeof(table->nodeTable)) );
+    {   size_t const maxBits = HUF_buildCTable_wksp(table->CTable, table->count,
+                                            maxSymbolValue, huffLog,
+                                            table->nodeTable, sizeof(table->nodeTable));
+        CHECK_F(maxBits);
         huffLog = (U32)maxBits;
         /* Zero unused symbols in CTable, so we can check it for validity */
         memset(table->CTable + (maxSymbolValue + 1), 0,
@@ -706,7 +708,7 @@ static size_t HUF_compress_internal (
             if (oldSize <= hSize + newSize || hSize + 12 >= srcSize) {
                 return HUF_compressCTable_internal(ostart, op, oend,
                                                    src, srcSize,
-                                                   singleStream, oldHufTable, bmi2);
+                                                   nbStreams, oldHufTable, bmi2);
         }   }
 
         /* Use the new huffman table */
@@ -718,7 +720,7 @@ static size_t HUF_compress_internal (
     }
     return HUF_compressCTable_internal(ostart, op, oend,
                                        src, srcSize,
-                                       singleStream, table->CTable, bmi2);
+                                       nbStreams, table->CTable, bmi2);
 }
 
 
@@ -728,7 +730,7 @@ size_t HUF_compress1X_wksp (void* dst, size_t dstSize,
                       void* workSpace, size_t wkspSize)
 {
     return HUF_compress_internal(dst, dstSize, src, srcSize,
-                                 maxSymbolValue, huffLog, 1 /*single stream*/,
+                                 maxSymbolValue, huffLog, HUF_singleStream,
                                  workSpace, wkspSize,
                                  NULL, NULL, 0, 0 /*bmi2*/);
 }
@@ -740,7 +742,7 @@ size_t HUF_compress1X_repeat (void* dst, size_t dstSize,
                       HUF_CElt* hufTable, HUF_repeat* repeat, int preferRepeat, int bmi2)
 {
     return HUF_compress_internal(dst, dstSize, src, srcSize,
-                                 maxSymbolValue, huffLog, 1 /*single stream*/,
+                                 maxSymbolValue, huffLog, HUF_singleStream,
                                  workSpace, wkspSize, hufTable,
                                  repeat, preferRepeat, bmi2);
 }
@@ -762,7 +764,7 @@ size_t HUF_compress4X_wksp (void* dst, size_t dstSize,
                       void* workSpace, size_t wkspSize)
 {
     return HUF_compress_internal(dst, dstSize, src, srcSize,
-                                 maxSymbolValue, huffLog, 0 /*4 streams*/,
+                                 maxSymbolValue, huffLog, HUF_fourStreams,
                                  workSpace, wkspSize,
                                  NULL, NULL, 0, 0 /*bmi2*/);
 }
@@ -777,7 +779,7 @@ size_t HUF_compress4X_repeat (void* dst, size_t dstSize,
                       HUF_CElt* hufTable, HUF_repeat* repeat, int preferRepeat, int bmi2)
 {
     return HUF_compress_internal(dst, dstSize, src, srcSize,
-                                 maxSymbolValue, huffLog, 0 /* 4 streams */,
+                                 maxSymbolValue, huffLog, HUF_fourStreams,
                                  workSpace, wkspSize,
                                  hufTable, repeat, preferRepeat, bmi2);
 }
diff --git a/lib/compress/zstd_compress.c b/lib/compress/zstd_compress.c
index 5f6280a8f7cc..c2c9d3bc553d 100644
--- a/lib/compress/zstd_compress.c
+++ b/lib/compress/zstd_compress.c
@@ -11,6 +11,7 @@
 /*-*************************************
 *  Dependencies
 ***************************************/
+#include          /* INT_MAX */
 #include          /* memset */
 #include "cpu.h"
 #include "mem.h"
@@ -61,7 +62,7 @@ static void ZSTD_initCCtx(ZSTD_CCtx* cctx, ZSTD_customMem memManager)
     memset(cctx, 0, sizeof(*cctx));
     cctx->customMem = memManager;
     cctx->bmi2 = ZSTD_cpuid_bmi2(ZSTD_cpuid());
-    {   size_t const err = ZSTD_CCtx_resetParameters(cctx);
+    {   size_t const err = ZSTD_CCtx_reset(cctx, ZSTD_reset_parameters);
         assert(!ZSTD_isError(err));
         (void)err;
     }
@@ -128,7 +129,7 @@ static size_t ZSTD_sizeof_mtctx(const ZSTD_CCtx* cctx)
 #ifdef ZSTD_MULTITHREAD
     return ZSTDMT_sizeof_CCtx(cctx->mtctx);
 #else
-    (void) cctx;
+    (void)cctx;
     return 0;
 #endif
 }
@@ -226,9 +227,160 @@ static ZSTD_CCtx_params ZSTD_assignParamsToCCtxParams(
     return ret;
 }
 
-#define CLAMPCHECK(val,min,max) {            \
-    if (((val)<(min)) | ((val)>(max))) {     \
-        return ERROR(parameter_outOfBound);  \
+ZSTD_bounds ZSTD_cParam_getBounds(ZSTD_cParameter param)
+{
+    ZSTD_bounds bounds = { 0, 0, 0 };
+
+    switch(param)
+    {
+    case ZSTD_c_compressionLevel:
+        bounds.lowerBound = ZSTD_minCLevel();
+        bounds.upperBound = ZSTD_maxCLevel();
+        return bounds;
+
+    case ZSTD_c_windowLog:
+        bounds.lowerBound = ZSTD_WINDOWLOG_MIN;
+        bounds.upperBound = ZSTD_WINDOWLOG_MAX;
+        return bounds;
+
+    case ZSTD_c_hashLog:
+        bounds.lowerBound = ZSTD_HASHLOG_MIN;
+        bounds.upperBound = ZSTD_HASHLOG_MAX;
+        return bounds;
+
+    case ZSTD_c_chainLog:
+        bounds.lowerBound = ZSTD_CHAINLOG_MIN;
+        bounds.upperBound = ZSTD_CHAINLOG_MAX;
+        return bounds;
+
+    case ZSTD_c_searchLog:
+        bounds.lowerBound = ZSTD_SEARCHLOG_MIN;
+        bounds.upperBound = ZSTD_SEARCHLOG_MAX;
+        return bounds;
+
+    case ZSTD_c_minMatch:
+        bounds.lowerBound = ZSTD_MINMATCH_MIN;
+        bounds.upperBound = ZSTD_MINMATCH_MAX;
+        return bounds;
+
+    case ZSTD_c_targetLength:
+        bounds.lowerBound = ZSTD_TARGETLENGTH_MIN;
+        bounds.upperBound = ZSTD_TARGETLENGTH_MAX;
+        return bounds;
+
+    case ZSTD_c_strategy:
+        bounds.lowerBound = ZSTD_STRATEGY_MIN;
+        bounds.upperBound = ZSTD_STRATEGY_MAX;
+        return bounds;
+
+    case ZSTD_c_contentSizeFlag:
+        bounds.lowerBound = 0;
+        bounds.upperBound = 1;
+        return bounds;
+
+    case ZSTD_c_checksumFlag:
+        bounds.lowerBound = 0;
+        bounds.upperBound = 1;
+        return bounds;
+
+    case ZSTD_c_dictIDFlag:
+        bounds.lowerBound = 0;
+        bounds.upperBound = 1;
+        return bounds;
+
+    case ZSTD_c_nbWorkers:
+        bounds.lowerBound = 0;
+#ifdef ZSTD_MULTITHREAD
+        bounds.upperBound = ZSTDMT_NBWORKERS_MAX;
+#else
+        bounds.upperBound = 0;
+#endif
+        return bounds;
+
+    case ZSTD_c_jobSize:
+        bounds.lowerBound = 0;
+#ifdef ZSTD_MULTITHREAD
+        bounds.upperBound = ZSTDMT_JOBSIZE_MAX;
+#else
+        bounds.upperBound = 0;
+#endif
+        return bounds;
+
+    case ZSTD_c_overlapLog:
+        bounds.lowerBound = ZSTD_OVERLAPLOG_MIN;
+        bounds.upperBound = ZSTD_OVERLAPLOG_MAX;
+        return bounds;
+
+    case ZSTD_c_enableLongDistanceMatching:
+        bounds.lowerBound = 0;
+        bounds.upperBound = 1;
+        return bounds;
+
+    case ZSTD_c_ldmHashLog:
+        bounds.lowerBound = ZSTD_LDM_HASHLOG_MIN;
+        bounds.upperBound = ZSTD_LDM_HASHLOG_MAX;
+        return bounds;
+
+    case ZSTD_c_ldmMinMatch:
+        bounds.lowerBound = ZSTD_LDM_MINMATCH_MIN;
+        bounds.upperBound = ZSTD_LDM_MINMATCH_MAX;
+        return bounds;
+
+    case ZSTD_c_ldmBucketSizeLog:
+        bounds.lowerBound = ZSTD_LDM_BUCKETSIZELOG_MIN;
+        bounds.upperBound = ZSTD_LDM_BUCKETSIZELOG_MAX;
+        return bounds;
+
+    case ZSTD_c_ldmHashRateLog:
+        bounds.lowerBound = ZSTD_LDM_HASHRATELOG_MIN;
+        bounds.upperBound = ZSTD_LDM_HASHRATELOG_MAX;
+        return bounds;
+
+    /* experimental parameters */
+    case ZSTD_c_rsyncable:
+        bounds.lowerBound = 0;
+        bounds.upperBound = 1;
+        return bounds;
+
+    case ZSTD_c_forceMaxWindow :
+        bounds.lowerBound = 0;
+        bounds.upperBound = 1;
+        return bounds;
+
+    case ZSTD_c_format:
+        ZSTD_STATIC_ASSERT(ZSTD_f_zstd1 < ZSTD_f_zstd1_magicless);
+        bounds.lowerBound = ZSTD_f_zstd1;
+        bounds.upperBound = ZSTD_f_zstd1_magicless;   /* note : how to ensure at compile time that this is the highest value enum ? */
+        return bounds;
+
+    case ZSTD_c_forceAttachDict:
+        ZSTD_STATIC_ASSERT(ZSTD_dictDefaultAttach < ZSTD_dictForceCopy);
+        bounds.lowerBound = ZSTD_dictDefaultAttach;
+        bounds.upperBound = ZSTD_dictForceCopy;       /* note : how to ensure at compile time that this is the highest value enum ? */
+        return bounds;
+
+    default:
+        {   ZSTD_bounds const boundError = { ERROR(parameter_unsupported), 0, 0 };
+            return boundError;
+        }
+    }
+}
+
+/* ZSTD_cParam_withinBounds:
+ * @return 1 if value is within cParam bounds,
+ * 0 otherwise */
+static int ZSTD_cParam_withinBounds(ZSTD_cParameter cParam, int value)
+{
+    ZSTD_bounds const bounds = ZSTD_cParam_getBounds(cParam);
+    if (ZSTD_isError(bounds.error)) return 0;
+    if (value < bounds.lowerBound) return 0;
+    if (value > bounds.upperBound) return 0;
+    return 1;
+}
+
+#define BOUNDCHECK(cParam, val) {                  \
+    if (!ZSTD_cParam_withinBounds(cParam,val)) {   \
+        return ERROR(parameter_outOfBound);        \
 }   }
 
 
@@ -236,38 +388,39 @@ static int ZSTD_isUpdateAuthorized(ZSTD_cParameter param)
 {
     switch(param)
     {
-    case ZSTD_p_compressionLevel:
-    case ZSTD_p_hashLog:
-    case ZSTD_p_chainLog:
-    case ZSTD_p_searchLog:
-    case ZSTD_p_minMatch:
-    case ZSTD_p_targetLength:
-    case ZSTD_p_compressionStrategy:
+    case ZSTD_c_compressionLevel:
+    case ZSTD_c_hashLog:
+    case ZSTD_c_chainLog:
+    case ZSTD_c_searchLog:
+    case ZSTD_c_minMatch:
+    case ZSTD_c_targetLength:
+    case ZSTD_c_strategy:
         return 1;
 
-    case ZSTD_p_format:
-    case ZSTD_p_windowLog:
-    case ZSTD_p_contentSizeFlag:
-    case ZSTD_p_checksumFlag:
-    case ZSTD_p_dictIDFlag:
-    case ZSTD_p_forceMaxWindow :
-    case ZSTD_p_nbWorkers:
-    case ZSTD_p_jobSize:
-    case ZSTD_p_overlapSizeLog:
-    case ZSTD_p_enableLongDistanceMatching:
-    case ZSTD_p_ldmHashLog:
-    case ZSTD_p_ldmMinMatch:
-    case ZSTD_p_ldmBucketSizeLog:
-    case ZSTD_p_ldmHashEveryLog:
-    case ZSTD_p_forceAttachDict:
+    case ZSTD_c_format:
+    case ZSTD_c_windowLog:
+    case ZSTD_c_contentSizeFlag:
+    case ZSTD_c_checksumFlag:
+    case ZSTD_c_dictIDFlag:
+    case ZSTD_c_forceMaxWindow :
+    case ZSTD_c_nbWorkers:
+    case ZSTD_c_jobSize:
+    case ZSTD_c_overlapLog:
+    case ZSTD_c_rsyncable:
+    case ZSTD_c_enableLongDistanceMatching:
+    case ZSTD_c_ldmHashLog:
+    case ZSTD_c_ldmMinMatch:
+    case ZSTD_c_ldmBucketSizeLog:
+    case ZSTD_c_ldmHashRateLog:
+    case ZSTD_c_forceAttachDict:
     default:
         return 0;
     }
 }
 
-size_t ZSTD_CCtx_setParameter(ZSTD_CCtx* cctx, ZSTD_cParameter param, unsigned value)
+size_t ZSTD_CCtx_setParameter(ZSTD_CCtx* cctx, ZSTD_cParameter param, int value)
 {
-    DEBUGLOG(4, "ZSTD_CCtx_setParameter (%u, %u)", (U32)param, value);
+    DEBUGLOG(4, "ZSTD_CCtx_setParameter (%i, %i)", (int)param, value);
     if (cctx->streamStage != zcss_init) {
         if (ZSTD_isUpdateAuthorized(param)) {
             cctx->cParamsChanged = 1;
@@ -277,51 +430,52 @@ size_t ZSTD_CCtx_setParameter(ZSTD_CCtx* cctx, ZSTD_cParameter param, unsigned v
 
     switch(param)
     {
-    case ZSTD_p_format :
+    case ZSTD_c_format :
         return ZSTD_CCtxParam_setParameter(&cctx->requestedParams, param, value);
 
-    case ZSTD_p_compressionLevel:
+    case ZSTD_c_compressionLevel:
         if (cctx->cdict) return ERROR(stage_wrong);
         return ZSTD_CCtxParam_setParameter(&cctx->requestedParams, param, value);
 
-    case ZSTD_p_windowLog:
-    case ZSTD_p_hashLog:
-    case ZSTD_p_chainLog:
-    case ZSTD_p_searchLog:
-    case ZSTD_p_minMatch:
-    case ZSTD_p_targetLength:
-    case ZSTD_p_compressionStrategy:
+    case ZSTD_c_windowLog:
+    case ZSTD_c_hashLog:
+    case ZSTD_c_chainLog:
+    case ZSTD_c_searchLog:
+    case ZSTD_c_minMatch:
+    case ZSTD_c_targetLength:
+    case ZSTD_c_strategy:
         if (cctx->cdict) return ERROR(stage_wrong);
         return ZSTD_CCtxParam_setParameter(&cctx->requestedParams, param, value);
 
-    case ZSTD_p_contentSizeFlag:
-    case ZSTD_p_checksumFlag:
-    case ZSTD_p_dictIDFlag:
+    case ZSTD_c_contentSizeFlag:
+    case ZSTD_c_checksumFlag:
+    case ZSTD_c_dictIDFlag:
         return ZSTD_CCtxParam_setParameter(&cctx->requestedParams, param, value);
 
-    case ZSTD_p_forceMaxWindow :  /* Force back-references to remain < windowSize,
+    case ZSTD_c_forceMaxWindow :  /* Force back-references to remain < windowSize,
                                    * even when referencing into Dictionary content.
                                    * default : 0 when using a CDict, 1 when using a Prefix */
         return ZSTD_CCtxParam_setParameter(&cctx->requestedParams, param, value);
 
-    case ZSTD_p_forceAttachDict:
+    case ZSTD_c_forceAttachDict:
         return ZSTD_CCtxParam_setParameter(&cctx->requestedParams, param, value);
 
-    case ZSTD_p_nbWorkers:
-        if ((value>0) && cctx->staticSize) {
+    case ZSTD_c_nbWorkers:
+        if ((value!=0) && cctx->staticSize) {
             return ERROR(parameter_unsupported);  /* MT not compatible with static alloc */
         }
         return ZSTD_CCtxParam_setParameter(&cctx->requestedParams, param, value);
 
-    case ZSTD_p_jobSize:
-    case ZSTD_p_overlapSizeLog:
+    case ZSTD_c_jobSize:
+    case ZSTD_c_overlapLog:
+    case ZSTD_c_rsyncable:
         return ZSTD_CCtxParam_setParameter(&cctx->requestedParams, param, value);
 
-    case ZSTD_p_enableLongDistanceMatching:
-    case ZSTD_p_ldmHashLog:
-    case ZSTD_p_ldmMinMatch:
-    case ZSTD_p_ldmBucketSizeLog:
-    case ZSTD_p_ldmHashEveryLog:
+    case ZSTD_c_enableLongDistanceMatching:
+    case ZSTD_c_ldmHashLog:
+    case ZSTD_c_ldmMinMatch:
+    case ZSTD_c_ldmBucketSizeLog:
+    case ZSTD_c_ldmHashRateLog:
         if (cctx->cdict) return ERROR(stage_wrong);
         return ZSTD_CCtxParam_setParameter(&cctx->requestedParams, param, value);
 
@@ -329,21 +483,21 @@ size_t ZSTD_CCtx_setParameter(ZSTD_CCtx* cctx, ZSTD_cParameter param, unsigned v
     }
 }
 
-size_t ZSTD_CCtxParam_setParameter(
-        ZSTD_CCtx_params* CCtxParams, ZSTD_cParameter param, unsigned value)
+size_t ZSTD_CCtxParam_setParameter(ZSTD_CCtx_params* CCtxParams,
+                                   ZSTD_cParameter param, int value)
 {
-    DEBUGLOG(4, "ZSTD_CCtxParam_setParameter (%u, %u)", (U32)param, value);
+    DEBUGLOG(4, "ZSTD_CCtxParam_setParameter (%i, %i)", (int)param, value);
     switch(param)
     {
-    case ZSTD_p_format :
-        if (value > (unsigned)ZSTD_f_zstd1_magicless)
-            return ERROR(parameter_unsupported);
+    case ZSTD_c_format :
+        BOUNDCHECK(ZSTD_c_format, value);
         CCtxParams->format = (ZSTD_format_e)value;
         return (size_t)CCtxParams->format;
 
-    case ZSTD_p_compressionLevel : {
-        int cLevel = (int)value;  /* cast expected to restore negative sign */
+    case ZSTD_c_compressionLevel : {
+        int cLevel = value;
         if (cLevel > ZSTD_maxCLevel()) cLevel = ZSTD_maxCLevel();
+        if (cLevel < ZSTD_minCLevel()) cLevel = ZSTD_minCLevel();
         if (cLevel) {  /* 0 : does not change current level */
             CCtxParams->compressionLevel = cLevel;
         }
@@ -351,213 +505,229 @@ size_t ZSTD_CCtxParam_setParameter(
         return 0;  /* return type (size_t) cannot represent negative values */
     }
 
-    case ZSTD_p_windowLog :
-        if (value>0)   /* 0 => use default */
-            CLAMPCHECK(value, ZSTD_WINDOWLOG_MIN, ZSTD_WINDOWLOG_MAX);
+    case ZSTD_c_windowLog :
+        if (value!=0)   /* 0 => use default */
+            BOUNDCHECK(ZSTD_c_windowLog, value);
         CCtxParams->cParams.windowLog = value;
         return CCtxParams->cParams.windowLog;
 
-    case ZSTD_p_hashLog :
-        if (value>0)   /* 0 => use default */
-            CLAMPCHECK(value, ZSTD_HASHLOG_MIN, ZSTD_HASHLOG_MAX);
+    case ZSTD_c_hashLog :
+        if (value!=0)   /* 0 => use default */
+            BOUNDCHECK(ZSTD_c_hashLog, value);
         CCtxParams->cParams.hashLog = value;
         return CCtxParams->cParams.hashLog;
 
-    case ZSTD_p_chainLog :
-        if (value>0)   /* 0 => use default */
-            CLAMPCHECK(value, ZSTD_CHAINLOG_MIN, ZSTD_CHAINLOG_MAX);
+    case ZSTD_c_chainLog :
+        if (value!=0)   /* 0 => use default */
+            BOUNDCHECK(ZSTD_c_chainLog, value);
         CCtxParams->cParams.chainLog = value;
         return CCtxParams->cParams.chainLog;
 
-    case ZSTD_p_searchLog :
-        if (value>0)   /* 0 => use default */
-            CLAMPCHECK(value, ZSTD_SEARCHLOG_MIN, ZSTD_SEARCHLOG_MAX);
+    case ZSTD_c_searchLog :
+        if (value!=0)   /* 0 => use default */
+            BOUNDCHECK(ZSTD_c_searchLog, value);
         CCtxParams->cParams.searchLog = value;
         return value;
 
-    case ZSTD_p_minMatch :
-        if (value>0)   /* 0 => use default */
-            CLAMPCHECK(value, ZSTD_SEARCHLENGTH_MIN, ZSTD_SEARCHLENGTH_MAX);
-        CCtxParams->cParams.searchLength = value;
-        return CCtxParams->cParams.searchLength;
+    case ZSTD_c_minMatch :
+        if (value!=0)   /* 0 => use default */
+            BOUNDCHECK(ZSTD_c_minMatch, value);
+        CCtxParams->cParams.minMatch = value;
+        return CCtxParams->cParams.minMatch;
 
-    case ZSTD_p_targetLength :
-        /* all values are valid. 0 => use default */
+    case ZSTD_c_targetLength :
+        BOUNDCHECK(ZSTD_c_targetLength, value);
         CCtxParams->cParams.targetLength = value;
         return CCtxParams->cParams.targetLength;
 
-    case ZSTD_p_compressionStrategy :
-        if (value>0)   /* 0 => use default */
-            CLAMPCHECK(value, (unsigned)ZSTD_fast, (unsigned)ZSTD_btultra);
+    case ZSTD_c_strategy :
+        if (value!=0)   /* 0 => use default */
+            BOUNDCHECK(ZSTD_c_strategy, value);
         CCtxParams->cParams.strategy = (ZSTD_strategy)value;
         return (size_t)CCtxParams->cParams.strategy;
 
-    case ZSTD_p_contentSizeFlag :
+    case ZSTD_c_contentSizeFlag :
         /* Content size written in frame header _when known_ (default:1) */
-        DEBUGLOG(4, "set content size flag = %u", (value>0));
-        CCtxParams->fParams.contentSizeFlag = value > 0;
+        DEBUGLOG(4, "set content size flag = %u", (value!=0));
+        CCtxParams->fParams.contentSizeFlag = value != 0;
         return CCtxParams->fParams.contentSizeFlag;
 
-    case ZSTD_p_checksumFlag :
+    case ZSTD_c_checksumFlag :
         /* A 32-bits content checksum will be calculated and written at end of frame (default:0) */
-        CCtxParams->fParams.checksumFlag = value > 0;
+        CCtxParams->fParams.checksumFlag = value != 0;
         return CCtxParams->fParams.checksumFlag;
 
-    case ZSTD_p_dictIDFlag : /* When applicable, dictionary's dictID is provided in frame header (default:1) */
-        DEBUGLOG(4, "set dictIDFlag = %u", (value>0));
+    case ZSTD_c_dictIDFlag : /* When applicable, dictionary's dictID is provided in frame header (default:1) */
+        DEBUGLOG(4, "set dictIDFlag = %u", (value!=0));
         CCtxParams->fParams.noDictIDFlag = !value;
         return !CCtxParams->fParams.noDictIDFlag;
 
-    case ZSTD_p_forceMaxWindow :
-        CCtxParams->forceWindow = (value > 0);
+    case ZSTD_c_forceMaxWindow :
+        CCtxParams->forceWindow = (value != 0);
         return CCtxParams->forceWindow;
 
-    case ZSTD_p_forceAttachDict :
-        CCtxParams->attachDictPref = value ?
-                                    (value > 0 ? ZSTD_dictForceAttach : ZSTD_dictForceCopy) :
-                                     ZSTD_dictDefaultAttach;
+    case ZSTD_c_forceAttachDict : {
+        const ZSTD_dictAttachPref_e pref = (ZSTD_dictAttachPref_e)value;
+        BOUNDCHECK(ZSTD_c_forceAttachDict, pref);
+        CCtxParams->attachDictPref = pref;
         return CCtxParams->attachDictPref;
+    }
 
-    case ZSTD_p_nbWorkers :
+    case ZSTD_c_nbWorkers :
 #ifndef ZSTD_MULTITHREAD
-        if (value>0) return ERROR(parameter_unsupported);
+        if (value!=0) return ERROR(parameter_unsupported);
         return 0;
 #else
         return ZSTDMT_CCtxParam_setNbWorkers(CCtxParams, value);
 #endif
 
-    case ZSTD_p_jobSize :
+    case ZSTD_c_jobSize :
 #ifndef ZSTD_MULTITHREAD
         return ERROR(parameter_unsupported);
 #else
         return ZSTDMT_CCtxParam_setMTCtxParameter(CCtxParams, ZSTDMT_p_jobSize, value);
 #endif
 
-    case ZSTD_p_overlapSizeLog :
+    case ZSTD_c_overlapLog :
 #ifndef ZSTD_MULTITHREAD
         return ERROR(parameter_unsupported);
 #else
-        return ZSTDMT_CCtxParam_setMTCtxParameter(CCtxParams, ZSTDMT_p_overlapSectionLog, value);
+        return ZSTDMT_CCtxParam_setMTCtxParameter(CCtxParams, ZSTDMT_p_overlapLog, value);
 #endif
 
-    case ZSTD_p_enableLongDistanceMatching :
-        CCtxParams->ldmParams.enableLdm = (value>0);
+    case ZSTD_c_rsyncable :
+#ifndef ZSTD_MULTITHREAD
+        return ERROR(parameter_unsupported);
+#else
+        return ZSTDMT_CCtxParam_setMTCtxParameter(CCtxParams, ZSTDMT_p_rsyncable, value);
+#endif
+
+    case ZSTD_c_enableLongDistanceMatching :
+        CCtxParams->ldmParams.enableLdm = (value!=0);
         return CCtxParams->ldmParams.enableLdm;
 
-    case ZSTD_p_ldmHashLog :
-        if (value>0)   /* 0 ==> auto */
-            CLAMPCHECK(value, ZSTD_HASHLOG_MIN, ZSTD_HASHLOG_MAX);
+    case ZSTD_c_ldmHashLog :
+        if (value!=0)   /* 0 ==> auto */
+            BOUNDCHECK(ZSTD_c_ldmHashLog, value);
         CCtxParams->ldmParams.hashLog = value;
         return CCtxParams->ldmParams.hashLog;
 
-    case ZSTD_p_ldmMinMatch :
-        if (value>0)   /* 0 ==> default */
-            CLAMPCHECK(value, ZSTD_LDM_MINMATCH_MIN, ZSTD_LDM_MINMATCH_MAX);
+    case ZSTD_c_ldmMinMatch :
+        if (value!=0)   /* 0 ==> default */
+            BOUNDCHECK(ZSTD_c_ldmMinMatch, value);
         CCtxParams->ldmParams.minMatchLength = value;
         return CCtxParams->ldmParams.minMatchLength;
 
-    case ZSTD_p_ldmBucketSizeLog :
-        if (value > ZSTD_LDM_BUCKETSIZELOG_MAX)
-            return ERROR(parameter_outOfBound);
+    case ZSTD_c_ldmBucketSizeLog :
+        if (value!=0)   /* 0 ==> default */
+            BOUNDCHECK(ZSTD_c_ldmBucketSizeLog, value);
         CCtxParams->ldmParams.bucketSizeLog = value;
         return CCtxParams->ldmParams.bucketSizeLog;
 
-    case ZSTD_p_ldmHashEveryLog :
+    case ZSTD_c_ldmHashRateLog :
         if (value > ZSTD_WINDOWLOG_MAX - ZSTD_HASHLOG_MIN)
             return ERROR(parameter_outOfBound);
-        CCtxParams->ldmParams.hashEveryLog = value;
-        return CCtxParams->ldmParams.hashEveryLog;
+        CCtxParams->ldmParams.hashRateLog = value;
+        return CCtxParams->ldmParams.hashRateLog;
 
     default: return ERROR(parameter_unsupported);
     }
 }
 
-size_t ZSTD_CCtx_getParameter(ZSTD_CCtx* cctx, ZSTD_cParameter param, unsigned* value)
+size_t ZSTD_CCtx_getParameter(ZSTD_CCtx* cctx, ZSTD_cParameter param, int* value)
 {
     return ZSTD_CCtxParam_getParameter(&cctx->requestedParams, param, value);
 }
 
 size_t ZSTD_CCtxParam_getParameter(
-        ZSTD_CCtx_params* CCtxParams, ZSTD_cParameter param, unsigned* value)
+        ZSTD_CCtx_params* CCtxParams, ZSTD_cParameter param, int* value)
 {
     switch(param)
     {
-    case ZSTD_p_format :
+    case ZSTD_c_format :
         *value = CCtxParams->format;
         break;
-    case ZSTD_p_compressionLevel :
+    case ZSTD_c_compressionLevel :
         *value = CCtxParams->compressionLevel;
         break;
-    case ZSTD_p_windowLog :
+    case ZSTD_c_windowLog :
         *value = CCtxParams->cParams.windowLog;
         break;
-    case ZSTD_p_hashLog :
+    case ZSTD_c_hashLog :
         *value = CCtxParams->cParams.hashLog;
         break;
-    case ZSTD_p_chainLog :
+    case ZSTD_c_chainLog :
         *value = CCtxParams->cParams.chainLog;
         break;
-    case ZSTD_p_searchLog :
+    case ZSTD_c_searchLog :
         *value = CCtxParams->cParams.searchLog;
         break;
-    case ZSTD_p_minMatch :
-        *value = CCtxParams->cParams.searchLength;
+    case ZSTD_c_minMatch :
+        *value = CCtxParams->cParams.minMatch;
         break;
-    case ZSTD_p_targetLength :
+    case ZSTD_c_targetLength :
         *value = CCtxParams->cParams.targetLength;
         break;
-    case ZSTD_p_compressionStrategy :
+    case ZSTD_c_strategy :
         *value = (unsigned)CCtxParams->cParams.strategy;
         break;
-    case ZSTD_p_contentSizeFlag :
+    case ZSTD_c_contentSizeFlag :
         *value = CCtxParams->fParams.contentSizeFlag;
         break;
-    case ZSTD_p_checksumFlag :
+    case ZSTD_c_checksumFlag :
         *value = CCtxParams->fParams.checksumFlag;
         break;
-    case ZSTD_p_dictIDFlag :
+    case ZSTD_c_dictIDFlag :
         *value = !CCtxParams->fParams.noDictIDFlag;
         break;
-    case ZSTD_p_forceMaxWindow :
+    case ZSTD_c_forceMaxWindow :
         *value = CCtxParams->forceWindow;
         break;
-    case ZSTD_p_forceAttachDict :
+    case ZSTD_c_forceAttachDict :
         *value = CCtxParams->attachDictPref;
         break;
-    case ZSTD_p_nbWorkers :
+    case ZSTD_c_nbWorkers :
 #ifndef ZSTD_MULTITHREAD
         assert(CCtxParams->nbWorkers == 0);
 #endif
         *value = CCtxParams->nbWorkers;
         break;
-    case ZSTD_p_jobSize :
+    case ZSTD_c_jobSize :
 #ifndef ZSTD_MULTITHREAD
         return ERROR(parameter_unsupported);
 #else
-        *value = CCtxParams->jobSize;
+        assert(CCtxParams->jobSize <= INT_MAX);
+        *value = (int)CCtxParams->jobSize;
         break;
 #endif
-    case ZSTD_p_overlapSizeLog :
+    case ZSTD_c_overlapLog :
 #ifndef ZSTD_MULTITHREAD
         return ERROR(parameter_unsupported);
 #else
-        *value = CCtxParams->overlapSizeLog;
+        *value = CCtxParams->overlapLog;
         break;
 #endif
-    case ZSTD_p_enableLongDistanceMatching :
+    case ZSTD_c_rsyncable :
+#ifndef ZSTD_MULTITHREAD
+        return ERROR(parameter_unsupported);
+#else
+        *value = CCtxParams->rsyncable;
+        break;
+#endif
+    case ZSTD_c_enableLongDistanceMatching :
         *value = CCtxParams->ldmParams.enableLdm;
         break;
-    case ZSTD_p_ldmHashLog :
+    case ZSTD_c_ldmHashLog :
         *value = CCtxParams->ldmParams.hashLog;
         break;
-    case ZSTD_p_ldmMinMatch :
+    case ZSTD_c_ldmMinMatch :
         *value = CCtxParams->ldmParams.minMatchLength;
         break;
-    case ZSTD_p_ldmBucketSizeLog :
+    case ZSTD_c_ldmBucketSizeLog :
         *value = CCtxParams->ldmParams.bucketSizeLog;
         break;
-    case ZSTD_p_ldmHashEveryLog :
-        *value = CCtxParams->ldmParams.hashEveryLog;
+    case ZSTD_c_ldmHashRateLog :
+        *value = CCtxParams->ldmParams.hashRateLog;
         break;
     default: return ERROR(parameter_unsupported);
     }
@@ -655,34 +825,35 @@ size_t ZSTD_CCtx_refPrefix_advanced(
 
 /*! ZSTD_CCtx_reset() :
  *  Also dumps dictionary */
-void ZSTD_CCtx_reset(ZSTD_CCtx* cctx)
+size_t ZSTD_CCtx_reset(ZSTD_CCtx* cctx, ZSTD_ResetDirective reset)
 {
-    cctx->streamStage = zcss_init;
-    cctx->pledgedSrcSizePlusOne = 0;
+    if ( (reset == ZSTD_reset_session_only)
+      || (reset == ZSTD_reset_session_and_parameters) ) {
+        cctx->streamStage = zcss_init;
+        cctx->pledgedSrcSizePlusOne = 0;
+    }
+    if ( (reset == ZSTD_reset_parameters)
+      || (reset == ZSTD_reset_session_and_parameters) ) {
+        if (cctx->streamStage != zcss_init) return ERROR(stage_wrong);
+        cctx->cdict = NULL;
+        return ZSTD_CCtxParams_reset(&cctx->requestedParams);
+    }
+    return 0;
 }
 
-size_t ZSTD_CCtx_resetParameters(ZSTD_CCtx* cctx)
-{
-    if (cctx->streamStage != zcss_init) return ERROR(stage_wrong);
-    cctx->cdict = NULL;
-    return ZSTD_CCtxParams_reset(&cctx->requestedParams);
-}
 
 /** ZSTD_checkCParams() :
     control CParam values remain within authorized range.
     @return : 0, or an error code if one value is beyond authorized range */
 size_t ZSTD_checkCParams(ZSTD_compressionParameters cParams)
 {
-    CLAMPCHECK(cParams.windowLog, ZSTD_WINDOWLOG_MIN, ZSTD_WINDOWLOG_MAX);
-    CLAMPCHECK(cParams.chainLog, ZSTD_CHAINLOG_MIN, ZSTD_CHAINLOG_MAX);
-    CLAMPCHECK(cParams.hashLog, ZSTD_HASHLOG_MIN, ZSTD_HASHLOG_MAX);
-    CLAMPCHECK(cParams.searchLog, ZSTD_SEARCHLOG_MIN, ZSTD_SEARCHLOG_MAX);
-    CLAMPCHECK(cParams.searchLength, ZSTD_SEARCHLENGTH_MIN, ZSTD_SEARCHLENGTH_MAX);
-    ZSTD_STATIC_ASSERT(ZSTD_TARGETLENGTH_MIN == 0);
-    if (cParams.targetLength > ZSTD_TARGETLENGTH_MAX)
-        return ERROR(parameter_outOfBound);
-    if ((U32)(cParams.strategy) > (U32)ZSTD_btultra)
-        return ERROR(parameter_unsupported);
+    BOUNDCHECK(ZSTD_c_windowLog, cParams.windowLog);
+    BOUNDCHECK(ZSTD_c_chainLog,  cParams.chainLog);
+    BOUNDCHECK(ZSTD_c_hashLog,   cParams.hashLog);
+    BOUNDCHECK(ZSTD_c_searchLog, cParams.searchLog);
+    BOUNDCHECK(ZSTD_c_minMatch,  cParams.minMatch);
+    BOUNDCHECK(ZSTD_c_targetLength,cParams.targetLength);
+    BOUNDCHECK(ZSTD_c_strategy,  cParams.strategy);
     return 0;
 }
 
@@ -692,19 +863,19 @@ size_t ZSTD_checkCParams(ZSTD_compressionParameters cParams)
 static ZSTD_compressionParameters
 ZSTD_clampCParams(ZSTD_compressionParameters cParams)
 {
-#   define CLAMP(val,min,max) {      \
-        if (valmax) val=max;   \
+#   define CLAMP_TYPE(cParam, val, type) {                                \
+        ZSTD_bounds const bounds = ZSTD_cParam_getBounds(cParam);         \
+        if ((int)valbounds.upperBound) val=(type)bounds.upperBound; \
     }
-    CLAMP(cParams.windowLog, ZSTD_WINDOWLOG_MIN, ZSTD_WINDOWLOG_MAX);
-    CLAMP(cParams.chainLog, ZSTD_CHAINLOG_MIN, ZSTD_CHAINLOG_MAX);
-    CLAMP(cParams.hashLog, ZSTD_HASHLOG_MIN, ZSTD_HASHLOG_MAX);
-    CLAMP(cParams.searchLog, ZSTD_SEARCHLOG_MIN, ZSTD_SEARCHLOG_MAX);
-    CLAMP(cParams.searchLength, ZSTD_SEARCHLENGTH_MIN, ZSTD_SEARCHLENGTH_MAX);
-    ZSTD_STATIC_ASSERT(ZSTD_TARGETLENGTH_MIN == 0);
-    if (cParams.targetLength > ZSTD_TARGETLENGTH_MAX)
-        cParams.targetLength = ZSTD_TARGETLENGTH_MAX;
-    CLAMP(cParams.strategy, ZSTD_fast, ZSTD_btultra);
+#   define CLAMP(cParam, val) CLAMP_TYPE(cParam, val, int)
+    CLAMP(ZSTD_c_windowLog, cParams.windowLog);
+    CLAMP(ZSTD_c_chainLog,  cParams.chainLog);
+    CLAMP(ZSTD_c_hashLog,   cParams.hashLog);
+    CLAMP(ZSTD_c_searchLog, cParams.searchLog);
+    CLAMP(ZSTD_c_minMatch,  cParams.minMatch);
+    CLAMP(ZSTD_c_targetLength,cParams.targetLength);
+    CLAMP_TYPE(ZSTD_c_strategy,cParams.strategy, ZSTD_strategy);
     return cParams;
 }
 
@@ -774,7 +945,7 @@ ZSTD_compressionParameters ZSTD_getCParamsFromCCtxParams(
     if (CCtxParams->cParams.hashLog) cParams.hashLog = CCtxParams->cParams.hashLog;
     if (CCtxParams->cParams.chainLog) cParams.chainLog = CCtxParams->cParams.chainLog;
     if (CCtxParams->cParams.searchLog) cParams.searchLog = CCtxParams->cParams.searchLog;
-    if (CCtxParams->cParams.searchLength) cParams.searchLength = CCtxParams->cParams.searchLength;
+    if (CCtxParams->cParams.minMatch) cParams.minMatch = CCtxParams->cParams.minMatch;
     if (CCtxParams->cParams.targetLength) cParams.targetLength = CCtxParams->cParams.targetLength;
     if (CCtxParams->cParams.strategy) cParams.strategy = CCtxParams->cParams.strategy;
     assert(!ZSTD_checkCParams(cParams));
@@ -787,13 +958,12 @@ ZSTD_sizeof_matchState(const ZSTD_compressionParameters* const cParams,
 {
     size_t const chainSize = (cParams->strategy == ZSTD_fast) ? 0 : ((size_t)1 << cParams->chainLog);
     size_t const hSize = ((size_t)1) << cParams->hashLog;
-    U32    const hashLog3 = (forCCtx && cParams->searchLength==3) ? MIN(ZSTD_HASHLOG3_MAX, cParams->windowLog) : 0;
+    U32    const hashLog3 = (forCCtx && cParams->minMatch==3) ? MIN(ZSTD_HASHLOG3_MAX, cParams->windowLog) : 0;
     size_t const h3Size = ((size_t)1) << hashLog3;
     size_t const tableSpace = (chainSize + hSize + h3Size) * sizeof(U32);
     size_t const optPotentialSpace = ((MaxML+1) + (MaxLL+1) + (MaxOff+1) + (1<strategy == ZSTD_btopt) ||
-                                         (cParams->strategy == ZSTD_btultra)))
+    size_t const optSpace = (forCCtx && (cParams->strategy >= ZSTD_btopt))
                                 ? optPotentialSpace
                                 : 0;
     DEBUGLOG(4, "chainSize: %u - hSize: %u - h3Size: %u",
@@ -808,7 +978,7 @@ size_t ZSTD_estimateCCtxSize_usingCCtxParams(const ZSTD_CCtx_params* params)
     {   ZSTD_compressionParameters const cParams =
                 ZSTD_getCParamsFromCCtxParams(params, 0, 0);
         size_t const blockSize = MIN(ZSTD_BLOCKSIZE_MAX, (size_t)1 << cParams.windowLog);
-        U32    const divider = (cParams.searchLength==3) ? 3 : 4;
+        U32    const divider = (cParams.minMatch==3) ? 3 : 4;
         size_t const maxNbSeq = blockSize / divider;
         size_t const tokenSpace = WILDCOPY_OVERLENGTH + blockSize + 11*maxNbSeq;
         size_t const entropySpace = HUF_WORKSPACE_SIZE;
@@ -843,7 +1013,7 @@ size_t ZSTD_estimateCCtxSize(int compressionLevel)
 {
     int level;
     size_t memBudget = 0;
-    for (level=1; level<=compressionLevel; level++) {
+    for (level=MIN(compressionLevel, 1); level<=compressionLevel; level++) {
         size_t const newMB = ZSTD_estimateCCtxSize_internal(level);
         if (newMB > memBudget) memBudget = newMB;
     }
@@ -879,7 +1049,7 @@ size_t ZSTD_estimateCStreamSize(int compressionLevel)
 {
     int level;
     size_t memBudget = 0;
-    for (level=1; level<=compressionLevel; level++) {
+    for (level=MIN(compressionLevel, 1); level<=compressionLevel; level++) {
         size_t const newMB = ZSTD_estimateCStreamSize_internal(level);
         if (newMB > memBudget) memBudget = newMB;
     }
@@ -933,7 +1103,7 @@ static U32 ZSTD_equivalentCParams(ZSTD_compressionParameters cParams1,
     return (cParams1.hashLog  == cParams2.hashLog)
          & (cParams1.chainLog == cParams2.chainLog)
          & (cParams1.strategy == cParams2.strategy)   /* opt parser space */
-         & ((cParams1.searchLength==3) == (cParams2.searchLength==3));  /* hashlog3 space */
+         & ((cParams1.minMatch==3) == (cParams2.minMatch==3));  /* hashlog3 space */
 }
 
 static void ZSTD_assertEqualCParams(ZSTD_compressionParameters cParams1,
@@ -945,7 +1115,7 @@ static void ZSTD_assertEqualCParams(ZSTD_compressionParameters cParams1,
     assert(cParams1.chainLog     == cParams2.chainLog);
     assert(cParams1.hashLog      == cParams2.hashLog);
     assert(cParams1.searchLog    == cParams2.searchLog);
-    assert(cParams1.searchLength == cParams2.searchLength);
+    assert(cParams1.minMatch     == cParams2.minMatch);
     assert(cParams1.targetLength == cParams2.targetLength);
     assert(cParams1.strategy     == cParams2.strategy);
 }
@@ -960,7 +1130,7 @@ static U32 ZSTD_equivalentLdmParams(ldmParams_t ldmParams1,
             ldmParams1.hashLog == ldmParams2.hashLog &&
             ldmParams1.bucketSizeLog == ldmParams2.bucketSizeLog &&
             ldmParams1.minMatchLength == ldmParams2.minMatchLength &&
-            ldmParams1.hashEveryLog == ldmParams2.hashEveryLog);
+            ldmParams1.hashRateLog == ldmParams2.hashRateLog);
 }
 
 typedef enum { ZSTDb_not_buffered, ZSTDb_buffered } ZSTD_buffered_policy_e;
@@ -976,7 +1146,7 @@ static U32 ZSTD_sufficientBuff(size_t bufferSize1, size_t maxNbSeq1,
 {
     size_t const windowSize2 = MAX(1, (size_t)MIN(((U64)1 << cParams2.windowLog), pledgedSrcSize));
     size_t const blockSize2 = MIN(ZSTD_BLOCKSIZE_MAX, windowSize2);
-    size_t const maxNbSeq2 = blockSize2 / ((cParams2.searchLength == 3) ? 3 : 4);
+    size_t const maxNbSeq2 = blockSize2 / ((cParams2.minMatch == 3) ? 3 : 4);
     size_t const maxNbLit2 = blockSize2;
     size_t const neededBufferSize2 = (buffPol2==ZSTDb_buffered) ? windowSize2 + blockSize2 : 0;
     DEBUGLOG(4, "ZSTD_sufficientBuff: is neededBufferSize2=%u <= bufferSize1=%u",
@@ -1034,8 +1204,8 @@ static void ZSTD_invalidateMatchState(ZSTD_matchState_t* ms)
 {
     ZSTD_window_clear(&ms->window);
 
-    ms->nextToUpdate = ms->window.dictLimit + 1;
-    ms->nextToUpdate3 = ms->window.dictLimit + 1;
+    ms->nextToUpdate = ms->window.dictLimit;
+    ms->nextToUpdate3 = ms->window.dictLimit;
     ms->loadedDictEnd = 0;
     ms->opt.litLengthSum = 0;  /* force reset of btopt stats */
     ms->dictMatchState = NULL;
@@ -1080,7 +1250,7 @@ ZSTD_reset_matchState(ZSTD_matchState_t* ms,
 {
     size_t const chainSize = (cParams->strategy == ZSTD_fast) ? 0 : ((size_t)1 << cParams->chainLog);
     size_t const hSize = ((size_t)1) << cParams->hashLog;
-    U32    const hashLog3 = (forCCtx && cParams->searchLength==3) ? MIN(ZSTD_HASHLOG3_MAX, cParams->windowLog) : 0;
+    U32    const hashLog3 = (forCCtx && cParams->minMatch==3) ? MIN(ZSTD_HASHLOG3_MAX, cParams->windowLog) : 0;
     size_t const h3Size = ((size_t)1) << hashLog3;
     size_t const tableSpace = (chainSize + hSize + h3Size) * sizeof(U32);
 
@@ -1094,9 +1264,9 @@ ZSTD_reset_matchState(ZSTD_matchState_t* ms,
     ZSTD_invalidateMatchState(ms);
 
     /* opt parser space */
-    if (forCCtx && ((cParams->strategy == ZSTD_btopt) | (cParams->strategy == ZSTD_btultra))) {
+    if (forCCtx && (cParams->strategy >= ZSTD_btopt)) {
         DEBUGLOG(4, "reserving optimal parser space");
-        ms->opt.litFreq = (U32*)ptr;
+        ms->opt.litFreq = (unsigned*)ptr;
         ms->opt.litLengthFreq = ms->opt.litFreq + (1<opt.matchLengthFreq = ms->opt.litLengthFreq + (MaxLL+1);
         ms->opt.offCodeFreq = ms->opt.matchLengthFreq + (MaxML+1);
@@ -1158,13 +1328,13 @@ static size_t ZSTD_resetCCtx_internal(ZSTD_CCtx* zc,
         /* Adjust long distance matching parameters */
         ZSTD_ldm_adjustParameters(¶ms.ldmParams, ¶ms.cParams);
         assert(params.ldmParams.hashLog >= params.ldmParams.bucketSizeLog);
-        assert(params.ldmParams.hashEveryLog < 32);
-        zc->ldmState.hashPower = ZSTD_ldm_getHashPower(params.ldmParams.minMatchLength);
+        assert(params.ldmParams.hashRateLog < 32);
+        zc->ldmState.hashPower = ZSTD_rollingHash_primePower(params.ldmParams.minMatchLength);
     }
 
     {   size_t const windowSize = MAX(1, (size_t)MIN(((U64)1 << params.cParams.windowLog), pledgedSrcSize));
         size_t const blockSize = MIN(ZSTD_BLOCKSIZE_MAX, windowSize);
-        U32    const divider = (params.cParams.searchLength==3) ? 3 : 4;
+        U32    const divider = (params.cParams.minMatch==3) ? 3 : 4;
         size_t const maxNbSeq = blockSize / divider;
         size_t const tokenSpace = WILDCOPY_OVERLENGTH + blockSize + 11*maxNbSeq;
         size_t const buffOutSize = (zbuff==ZSTDb_buffered) ? ZSTD_compressBound(blockSize)+1 : 0;
@@ -1227,7 +1397,7 @@ static size_t ZSTD_resetCCtx_internal(ZSTD_CCtx* zc,
         if (pledgedSrcSize == ZSTD_CONTENTSIZE_UNKNOWN)
             zc->appliedParams.fParams.contentSizeFlag = 0;
         DEBUGLOG(4, "pledged content size : %u ; flag : %u",
-            (U32)pledgedSrcSize, zc->appliedParams.fParams.contentSizeFlag);
+            (unsigned)pledgedSrcSize, zc->appliedParams.fParams.contentSizeFlag);
         zc->blockSize = blockSize;
 
         XXH64_reset(&zc->xxhState, 0);
@@ -1306,16 +1476,17 @@ void ZSTD_invalidateRepCodes(ZSTD_CCtx* cctx) {
  * dictionary tables into the working context is faster than using them
  * in-place.
  */
-static const size_t attachDictSizeCutoffs[(unsigned)ZSTD_btultra+1] = {
-    8 KB, /* unused */
-    8 KB, /* ZSTD_fast */
+static const size_t attachDictSizeCutoffs[ZSTD_STRATEGY_MAX+1] = {
+    8 KB,  /* unused */
+    8 KB,  /* ZSTD_fast */
     16 KB, /* ZSTD_dfast */
     32 KB, /* ZSTD_greedy */
     32 KB, /* ZSTD_lazy */
     32 KB, /* ZSTD_lazy2 */
     32 KB, /* ZSTD_btlazy2 */
     32 KB, /* ZSTD_btopt */
-    8 KB /* ZSTD_btultra */
+    8 KB,  /* ZSTD_btultra */
+    8 KB   /* ZSTD_btultra2 */
 };
 
 static int ZSTD_shouldAttachDict(const ZSTD_CDict* cdict,
@@ -1447,7 +1618,8 @@ static size_t ZSTD_resetCCtx_usingCDict(ZSTD_CCtx* cctx,
                             ZSTD_buffered_policy_e zbuff)
 {
 
-    DEBUGLOG(4, "ZSTD_resetCCtx_usingCDict (pledgedSrcSize=%u)", (U32)pledgedSrcSize);
+    DEBUGLOG(4, "ZSTD_resetCCtx_usingCDict (pledgedSrcSize=%u)",
+                (unsigned)pledgedSrcSize);
 
     if (ZSTD_shouldAttachDict(cdict, params, pledgedSrcSize)) {
         return ZSTD_resetCCtx_byAttachingCDict(
@@ -1670,7 +1842,9 @@ static size_t ZSTD_compressRleLiteralsBlock (void* dst, size_t dstCapacity, cons
  * note : use same formula for both situations */
 static size_t ZSTD_minGain(size_t srcSize, ZSTD_strategy strat)
 {
-    U32 const minlog = (strat==ZSTD_btultra) ? 7 : 6;
+    U32 const minlog = (strat>=ZSTD_btultra) ? (U32)(strat) - 1 : 6;
+    ZSTD_STATIC_ASSERT(ZSTD_btultra == 8);
+    assert(ZSTD_cParam_withinBounds(ZSTD_c_strategy, strat));
     return (srcSize >> minlog) + 2;
 }
 
@@ -1679,7 +1853,8 @@ static size_t ZSTD_compressLiterals (ZSTD_hufCTables_t const* prevHuf,
                                      ZSTD_strategy strategy, int disableLiteralCompression,
                                      void* dst, size_t dstCapacity,
                                const void* src, size_t srcSize,
-                                     U32* workspace, const int bmi2)
+                                     void* workspace, size_t wkspSize,
+                               const int bmi2)
 {
     size_t const minGain = ZSTD_minGain(srcSize, strategy);
     size_t const lhSize = 3 + (srcSize >= 1 KB) + (srcSize >= 16 KB);
@@ -1708,9 +1883,9 @@ static size_t ZSTD_compressLiterals (ZSTD_hufCTables_t const* prevHuf,
         int const preferRepeat = strategy < ZSTD_lazy ? srcSize <= 1024 : 0;
         if (repeat == HUF_repeat_valid && lhSize == 3) singleStream = 1;
         cLitSize = singleStream ? HUF_compress1X_repeat(ostart+lhSize, dstCapacity-lhSize, src, srcSize, 255, 11,
-                                      workspace, HUF_WORKSPACE_SIZE, (HUF_CElt*)nextHuf->CTable, &repeat, preferRepeat, bmi2)
+                                      workspace, wkspSize, (HUF_CElt*)nextHuf->CTable, &repeat, preferRepeat, bmi2)
                                 : HUF_compress4X_repeat(ostart+lhSize, dstCapacity-lhSize, src, srcSize, 255, 11,
-                                      workspace, HUF_WORKSPACE_SIZE, (HUF_CElt*)nextHuf->CTable, &repeat, preferRepeat, bmi2);
+                                      workspace, wkspSize, (HUF_CElt*)nextHuf->CTable, &repeat, preferRepeat, bmi2);
         if (repeat != HUF_repeat_none) {
             /* reused the existing table */
             hType = set_repeat;
@@ -1977,7 +2152,7 @@ ZSTD_selectEncodingType(
         assert(!ZSTD_isError(NCountCost));
         assert(compressedCost < ERROR(maxCode));
         DEBUGLOG(5, "Estimated bit costs: basic=%u\trepeat=%u\tcompressed=%u",
-                    (U32)basicCost, (U32)repeatCost, (U32)compressedCost);
+                    (unsigned)basicCost, (unsigned)repeatCost, (unsigned)compressedCost);
         if (basicCost <= repeatCost && basicCost <= compressedCost) {
             DEBUGLOG(5, "Selected set_basic");
             assert(isDefaultAllowed);
@@ -1999,7 +2174,7 @@ ZSTD_selectEncodingType(
 MEM_STATIC size_t
 ZSTD_buildCTable(void* dst, size_t dstCapacity,
                 FSE_CTable* nextCTable, U32 FSELog, symbolEncodingType_e type,
-                U32* count, U32 max,
+                unsigned* count, U32 max,
                 const BYTE* codeTable, size_t nbSeq,
                 const S16* defaultNorm, U32 defaultNormLog, U32 defaultMax,
                 const FSE_CTable* prevCTable, size_t prevCTableSize,
@@ -2007,11 +2182,13 @@ ZSTD_buildCTable(void* dst, size_t dstCapacity,
 {
     BYTE* op = (BYTE*)dst;
     const BYTE* const oend = op + dstCapacity;
+    DEBUGLOG(6, "ZSTD_buildCTable (dstCapacity=%u)", (unsigned)dstCapacity);
 
     switch (type) {
     case set_rle:
-        *op = codeTable[0];
         CHECK_F(FSE_buildCTable_rle(nextCTable, (BYTE)max));
+        if (dstCapacity==0) return ERROR(dstSize_tooSmall);
+        *op = codeTable[0];
         return 1;
     case set_repeat:
         memcpy(nextCTable, prevCTable, prevCTableSize);
@@ -2053,6 +2230,9 @@ ZSTD_encodeSequences_body(
     FSE_CState_t  stateLitLength;
 
     CHECK_E(BIT_initCStream(&blockStream, dst, dstCapacity), dstSize_tooSmall); /* not enough space remaining */
+    DEBUGLOG(6, "available space for bitstream : %i  (dstCapacity=%u)",
+                (int)(blockStream.endPtr - blockStream.startPtr),
+                (unsigned)dstCapacity);
 
     /* first symbols */
     FSE_initCState2(&stateMatchLength, CTable_MatchLength, mlCodeTable[nbSeq-1]);
@@ -2085,9 +2265,9 @@ ZSTD_encodeSequences_body(
             U32  const ofBits = ofCode;
             U32  const mlBits = ML_bits[mlCode];
             DEBUGLOG(6, "encoding: litlen:%2u - matchlen:%2u - offCode:%7u",
-                        sequences[n].litLength,
-                        sequences[n].matchLength + MINMATCH,
-                        sequences[n].offset);
+                        (unsigned)sequences[n].litLength,
+                        (unsigned)sequences[n].matchLength + MINMATCH,
+                        (unsigned)sequences[n].offset);
                                                                             /* 32b*/  /* 64b*/
                                                                             /* (7)*/  /* (7)*/
             FSE_encodeSymbol(&blockStream, &stateOffsetBits, ofCode);       /* 15 */  /* 15 */
@@ -2112,6 +2292,7 @@ ZSTD_encodeSequences_body(
                 BIT_addBits(&blockStream, sequences[n].offset, ofBits);     /* 31 */
             }
             BIT_flushBits(&blockStream);                                    /* (7)*/
+            DEBUGLOG(7, "remaining space : %i", (int)(blockStream.endPtr - blockStream.ptr));
     }   }
 
     DEBUGLOG(6, "ZSTD_encodeSequences: flushing ML state with %u bits", stateMatchLength.stateLog);
@@ -2169,6 +2350,7 @@ static size_t ZSTD_encodeSequences(
             FSE_CTable const* CTable_LitLength, BYTE const* llCodeTable,
             seqDef const* sequences, size_t nbSeq, int longOffsets, int bmi2)
 {
+    DEBUGLOG(5, "ZSTD_encodeSequences: dstCapacity = %u", (unsigned)dstCapacity);
 #if DYNAMIC_BMI2
     if (bmi2) {
         return ZSTD_encodeSequences_bmi2(dst, dstCapacity,
@@ -2186,16 +2368,20 @@ static size_t ZSTD_encodeSequences(
                                         sequences, nbSeq, longOffsets);
 }
 
-MEM_STATIC size_t ZSTD_compressSequences_internal(seqStore_t* seqStorePtr,
-                              ZSTD_entropyCTables_t const* prevEntropy,
-                              ZSTD_entropyCTables_t* nextEntropy,
-                              ZSTD_CCtx_params const* cctxParams,
-                              void* dst, size_t dstCapacity, U32* workspace,
-                              const int bmi2)
+/* ZSTD_compressSequences_internal():
+ * actually compresses both literals and sequences */
+MEM_STATIC size_t
+ZSTD_compressSequences_internal(seqStore_t* seqStorePtr,
+                          const ZSTD_entropyCTables_t* prevEntropy,
+                                ZSTD_entropyCTables_t* nextEntropy,
+                          const ZSTD_CCtx_params* cctxParams,
+                                void* dst, size_t dstCapacity,
+                                void* workspace, size_t wkspSize,
+                          const int bmi2)
 {
     const int longOffsets = cctxParams->cParams.windowLog > STREAM_ACCUMULATOR_MIN;
     ZSTD_strategy const strategy = cctxParams->cParams.strategy;
-    U32 count[MaxSeq+1];
+    unsigned count[MaxSeq+1];
     FSE_CTable* CTable_LitLength = nextEntropy->fse.litlengthCTable;
     FSE_CTable* CTable_OffsetBits = nextEntropy->fse.offcodeCTable;
     FSE_CTable* CTable_MatchLength = nextEntropy->fse.matchlengthCTable;
@@ -2212,6 +2398,7 @@ MEM_STATIC size_t ZSTD_compressSequences_internal(seqStore_t* seqStorePtr,
     BYTE* lastNCount = NULL;
 
     ZSTD_STATIC_ASSERT(HUF_WORKSPACE_SIZE >= (1<litStart;
@@ -2222,7 +2409,8 @@ MEM_STATIC size_t ZSTD_compressSequences_internal(seqStore_t* seqStorePtr,
                                     cctxParams->cParams.strategy, disableLiteralCompression,
                                     op, dstCapacity,
                                     literals, litSize,
-                                    workspace, bmi2);
+                                    workspace, wkspSize,
+                                    bmi2);
         if (ZSTD_isError(cSize))
           return cSize;
         assert(cSize <= dstCapacity);
@@ -2249,51 +2437,63 @@ MEM_STATIC size_t ZSTD_compressSequences_internal(seqStore_t* seqStorePtr,
     /* convert length/distances into codes */
     ZSTD_seqToCodes(seqStorePtr);
     /* build CTable for Literal Lengths */
-    {   U32 max = MaxLL;
-        size_t const mostFrequent = HIST_countFast_wksp(count, &max, llCodeTable, nbSeq, workspace);   /* can't fail */
+    {   unsigned max = MaxLL;
+        size_t const mostFrequent = HIST_countFast_wksp(count, &max, llCodeTable, nbSeq, workspace, wkspSize);   /* can't fail */
         DEBUGLOG(5, "Building LL table");
         nextEntropy->fse.litlength_repeatMode = prevEntropy->fse.litlength_repeatMode;
-        LLtype = ZSTD_selectEncodingType(&nextEntropy->fse.litlength_repeatMode, count, max, mostFrequent, nbSeq, LLFSELog, prevEntropy->fse.litlengthCTable, LL_defaultNorm, LL_defaultNormLog, ZSTD_defaultAllowed, strategy);
+        LLtype = ZSTD_selectEncodingType(&nextEntropy->fse.litlength_repeatMode,
+                                        count, max, mostFrequent, nbSeq,
+                                        LLFSELog, prevEntropy->fse.litlengthCTable,
+                                        LL_defaultNorm, LL_defaultNormLog,
+                                        ZSTD_defaultAllowed, strategy);
         assert(set_basic < set_compressed && set_rle < set_compressed);
         assert(!(LLtype < set_compressed && nextEntropy->fse.litlength_repeatMode != FSE_repeat_none)); /* We don't copy tables */
         {   size_t const countSize = ZSTD_buildCTable(op, oend - op, CTable_LitLength, LLFSELog, (symbolEncodingType_e)LLtype,
                                                     count, max, llCodeTable, nbSeq, LL_defaultNorm, LL_defaultNormLog, MaxLL,
                                                     prevEntropy->fse.litlengthCTable, sizeof(prevEntropy->fse.litlengthCTable),
-                                                    workspace, HUF_WORKSPACE_SIZE);
+                                                    workspace, wkspSize);
             if (ZSTD_isError(countSize)) return countSize;
             if (LLtype == set_compressed)
                 lastNCount = op;
             op += countSize;
     }   }
     /* build CTable for Offsets */
-    {   U32 max = MaxOff;
-        size_t const mostFrequent = HIST_countFast_wksp(count, &max, ofCodeTable, nbSeq, workspace);  /* can't fail */
+    {   unsigned max = MaxOff;
+        size_t const mostFrequent = HIST_countFast_wksp(count, &max, ofCodeTable, nbSeq, workspace, wkspSize);  /* can't fail */
         /* We can only use the basic table if max <= DefaultMaxOff, otherwise the offsets are too large */
         ZSTD_defaultPolicy_e const defaultPolicy = (max <= DefaultMaxOff) ? ZSTD_defaultAllowed : ZSTD_defaultDisallowed;
         DEBUGLOG(5, "Building OF table");
         nextEntropy->fse.offcode_repeatMode = prevEntropy->fse.offcode_repeatMode;
-        Offtype = ZSTD_selectEncodingType(&nextEntropy->fse.offcode_repeatMode, count, max, mostFrequent, nbSeq, OffFSELog, prevEntropy->fse.offcodeCTable, OF_defaultNorm, OF_defaultNormLog, defaultPolicy, strategy);
+        Offtype = ZSTD_selectEncodingType(&nextEntropy->fse.offcode_repeatMode,
+                                        count, max, mostFrequent, nbSeq,
+                                        OffFSELog, prevEntropy->fse.offcodeCTable,
+                                        OF_defaultNorm, OF_defaultNormLog,
+                                        defaultPolicy, strategy);
         assert(!(Offtype < set_compressed && nextEntropy->fse.offcode_repeatMode != FSE_repeat_none)); /* We don't copy tables */
         {   size_t const countSize = ZSTD_buildCTable(op, oend - op, CTable_OffsetBits, OffFSELog, (symbolEncodingType_e)Offtype,
                                                     count, max, ofCodeTable, nbSeq, OF_defaultNorm, OF_defaultNormLog, DefaultMaxOff,
                                                     prevEntropy->fse.offcodeCTable, sizeof(prevEntropy->fse.offcodeCTable),
-                                                    workspace, HUF_WORKSPACE_SIZE);
+                                                    workspace, wkspSize);
             if (ZSTD_isError(countSize)) return countSize;
             if (Offtype == set_compressed)
                 lastNCount = op;
             op += countSize;
     }   }
     /* build CTable for MatchLengths */
-    {   U32 max = MaxML;
-        size_t const mostFrequent = HIST_countFast_wksp(count, &max, mlCodeTable, nbSeq, workspace);   /* can't fail */
-        DEBUGLOG(5, "Building ML table");
+    {   unsigned max = MaxML;
+        size_t const mostFrequent = HIST_countFast_wksp(count, &max, mlCodeTable, nbSeq, workspace, wkspSize);   /* can't fail */
+        DEBUGLOG(5, "Building ML table (remaining space : %i)", (int)(oend-op));
         nextEntropy->fse.matchlength_repeatMode = prevEntropy->fse.matchlength_repeatMode;
-        MLtype = ZSTD_selectEncodingType(&nextEntropy->fse.matchlength_repeatMode, count, max, mostFrequent, nbSeq, MLFSELog, prevEntropy->fse.matchlengthCTable, ML_defaultNorm, ML_defaultNormLog, ZSTD_defaultAllowed, strategy);
+        MLtype = ZSTD_selectEncodingType(&nextEntropy->fse.matchlength_repeatMode,
+                                        count, max, mostFrequent, nbSeq,
+                                        MLFSELog, prevEntropy->fse.matchlengthCTable,
+                                        ML_defaultNorm, ML_defaultNormLog,
+                                        ZSTD_defaultAllowed, strategy);
         assert(!(MLtype < set_compressed && nextEntropy->fse.matchlength_repeatMode != FSE_repeat_none)); /* We don't copy tables */
         {   size_t const countSize = ZSTD_buildCTable(op, oend - op, CTable_MatchLength, MLFSELog, (symbolEncodingType_e)MLtype,
                                                     count, max, mlCodeTable, nbSeq, ML_defaultNorm, ML_defaultNormLog, MaxML,
                                                     prevEntropy->fse.matchlengthCTable, sizeof(prevEntropy->fse.matchlengthCTable),
-                                                    workspace, HUF_WORKSPACE_SIZE);
+                                                    workspace, wkspSize);
             if (ZSTD_isError(countSize)) return countSize;
             if (MLtype == set_compressed)
                 lastNCount = op;
@@ -2328,19 +2528,24 @@ MEM_STATIC size_t ZSTD_compressSequences_internal(seqStore_t* seqStorePtr,
         }
     }
 
+    DEBUGLOG(5, "compressed block size : %u", (unsigned)(op - ostart));
     return op - ostart;
 }
 
-MEM_STATIC size_t ZSTD_compressSequences(seqStore_t* seqStorePtr,
-                        const ZSTD_entropyCTables_t* prevEntropy,
-                              ZSTD_entropyCTables_t* nextEntropy,
-                        const ZSTD_CCtx_params* cctxParams,
-                              void* dst, size_t dstCapacity,
-                              size_t srcSize, U32* workspace, int bmi2)
+MEM_STATIC size_t
+ZSTD_compressSequences(seqStore_t* seqStorePtr,
+                       const ZSTD_entropyCTables_t* prevEntropy,
+                             ZSTD_entropyCTables_t* nextEntropy,
+                       const ZSTD_CCtx_params* cctxParams,
+                             void* dst, size_t dstCapacity,
+                             size_t srcSize,
+                             void* workspace, size_t wkspSize,
+                             int bmi2)
 {
     size_t const cSize = ZSTD_compressSequences_internal(
-            seqStorePtr, prevEntropy, nextEntropy, cctxParams, dst, dstCapacity,
-            workspace, bmi2);
+                            seqStorePtr, prevEntropy, nextEntropy, cctxParams,
+                            dst, dstCapacity,
+                            workspace, wkspSize, bmi2);
     if (cSize == 0) return 0;
     /* When srcSize <= dstCapacity, there is enough space to write a raw uncompressed block.
      * Since we ran out of space, block must be not compressible, so fall back to raw uncompressed block.
@@ -2362,7 +2567,7 @@ MEM_STATIC size_t ZSTD_compressSequences(seqStore_t* seqStorePtr,
  * assumption : strat is a valid strategy */
 ZSTD_blockCompressor ZSTD_selectBlockCompressor(ZSTD_strategy strat, ZSTD_dictMode_e dictMode)
 {
-    static const ZSTD_blockCompressor blockCompressor[3][(unsigned)ZSTD_btultra+1] = {
+    static const ZSTD_blockCompressor blockCompressor[3][ZSTD_STRATEGY_MAX+1] = {
         { ZSTD_compressBlock_fast  /* default for 0 */,
           ZSTD_compressBlock_fast,
           ZSTD_compressBlock_doubleFast,
@@ -2371,7 +2576,8 @@ ZSTD_blockCompressor ZSTD_selectBlockCompressor(ZSTD_strategy strat, ZSTD_dictMo
           ZSTD_compressBlock_lazy2,
           ZSTD_compressBlock_btlazy2,
           ZSTD_compressBlock_btopt,
-          ZSTD_compressBlock_btultra },
+          ZSTD_compressBlock_btultra,
+          ZSTD_compressBlock_btultra2 },
         { ZSTD_compressBlock_fast_extDict  /* default for 0 */,
           ZSTD_compressBlock_fast_extDict,
           ZSTD_compressBlock_doubleFast_extDict,
@@ -2380,6 +2586,7 @@ ZSTD_blockCompressor ZSTD_selectBlockCompressor(ZSTD_strategy strat, ZSTD_dictMo
           ZSTD_compressBlock_lazy2_extDict,
           ZSTD_compressBlock_btlazy2_extDict,
           ZSTD_compressBlock_btopt_extDict,
+          ZSTD_compressBlock_btultra_extDict,
           ZSTD_compressBlock_btultra_extDict },
         { ZSTD_compressBlock_fast_dictMatchState  /* default for 0 */,
           ZSTD_compressBlock_fast_dictMatchState,
@@ -2389,14 +2596,14 @@ ZSTD_blockCompressor ZSTD_selectBlockCompressor(ZSTD_strategy strat, ZSTD_dictMo
           ZSTD_compressBlock_lazy2_dictMatchState,
           ZSTD_compressBlock_btlazy2_dictMatchState,
           ZSTD_compressBlock_btopt_dictMatchState,
+          ZSTD_compressBlock_btultra_dictMatchState,
           ZSTD_compressBlock_btultra_dictMatchState }
     };
     ZSTD_blockCompressor selectedCompressor;
     ZSTD_STATIC_ASSERT((unsigned)ZSTD_fast == 1);
 
-    assert((U32)strat >= (U32)ZSTD_fast);
-    assert((U32)strat <= (U32)ZSTD_btultra);
-    selectedCompressor = blockCompressor[(int)dictMode][(U32)strat];
+    assert(ZSTD_cParam_withinBounds(ZSTD_c_strategy, strat));
+    selectedCompressor = blockCompressor[(int)dictMode][(int)strat];
     assert(selectedCompressor != NULL);
     return selectedCompressor;
 }
@@ -2421,15 +2628,15 @@ static size_t ZSTD_compressBlock_internal(ZSTD_CCtx* zc,
 {
     ZSTD_matchState_t* const ms = &zc->blockState.matchState;
     size_t cSize;
-    DEBUGLOG(5, "ZSTD_compressBlock_internal (dstCapacity=%zu, dictLimit=%u, nextToUpdate=%u)",
-                dstCapacity, ms->window.dictLimit, ms->nextToUpdate);
+    DEBUGLOG(5, "ZSTD_compressBlock_internal (dstCapacity=%u, dictLimit=%u, nextToUpdate=%u)",
+                (unsigned)dstCapacity, (unsigned)ms->window.dictLimit, (unsigned)ms->nextToUpdate);
     assert(srcSize <= ZSTD_BLOCKSIZE_MAX);
 
     /* Assert that we have correctly flushed the ctx params into the ms's copy */
     ZSTD_assertEqualCParams(zc->appliedParams.cParams, ms->cParams);
 
     if (srcSize < MIN_CBLOCK_SIZE+ZSTD_blockHeaderSize+1) {
-        ZSTD_ldm_skipSequences(&zc->externSeqStore, srcSize, zc->appliedParams.cParams.searchLength);
+        ZSTD_ldm_skipSequences(&zc->externSeqStore, srcSize, zc->appliedParams.cParams.minMatch);
         cSize = 0;
         goto out;  /* don't even attempt compression below a certain srcSize */
     }
@@ -2437,8 +2644,8 @@ static size_t ZSTD_compressBlock_internal(ZSTD_CCtx* zc,
     ms->opt.symbolCosts = &zc->blockState.prevCBlock->entropy;   /* required for optimal parser to read stats from dictionary */
 
     /* a gap between an attached dict and the current window is not safe,
-     * they must remain adjacent, and when that stops being the case, the dict
-     * must be unset */
+     * they must remain adjacent,
+     * and when that stops being the case, the dict must be unset */
     assert(ms->dictMatchState == NULL || ms->loadedDictEnd == ms->window.dictLimit);
 
     /* limited update after a very long match */
@@ -2495,7 +2702,9 @@ static size_t ZSTD_compressBlock_internal(ZSTD_CCtx* zc,
             &zc->blockState.prevCBlock->entropy, &zc->blockState.nextCBlock->entropy,
             &zc->appliedParams,
             dst, dstCapacity,
-            srcSize, zc->entropyWorkspace, zc->bmi2);
+            srcSize,
+            zc->entropyWorkspace, HUF_WORKSPACE_SIZE /* statically allocated in resetCCtx */,
+            zc->bmi2);
 
 out:
     if (!ZSTD_isError(cSize) && cSize != 0) {
@@ -2535,7 +2744,7 @@ static size_t ZSTD_compress_frameChunk (ZSTD_CCtx* cctx,
     U32 const maxDist = (U32)1 << cctx->appliedParams.cParams.windowLog;
     assert(cctx->appliedParams.cParams.windowLog <= 31);
 
-    DEBUGLOG(5, "ZSTD_compress_frameChunk (blockSize=%u)", (U32)blockSize);
+    DEBUGLOG(5, "ZSTD_compress_frameChunk (blockSize=%u)", (unsigned)blockSize);
     if (cctx->appliedParams.fParams.checksumFlag && srcSize)
         XXH64_update(&cctx->xxhState, src, srcSize);
 
@@ -2583,7 +2792,7 @@ static size_t ZSTD_compress_frameChunk (ZSTD_CCtx* cctx,
             assert(dstCapacity >= cSize);
             dstCapacity -= cSize;
             DEBUGLOG(5, "ZSTD_compress_frameChunk: adding a block of size %u",
-                        (U32)cSize);
+                        (unsigned)cSize);
     }   }
 
     if (lastFrameChunk && (op>ostart)) cctx->stage = ZSTDcs_ending;
@@ -2606,9 +2815,9 @@ static size_t ZSTD_writeFrameHeader(void* dst, size_t dstCapacity,
     size_t pos=0;
 
     assert(!(params.fParams.contentSizeFlag && pledgedSrcSize == ZSTD_CONTENTSIZE_UNKNOWN));
-    if (dstCapacity < ZSTD_frameHeaderSize_max) return ERROR(dstSize_tooSmall);
+    if (dstCapacity < ZSTD_FRAMEHEADERSIZE_MAX) return ERROR(dstSize_tooSmall);
     DEBUGLOG(4, "ZSTD_writeFrameHeader : dictIDFlag : %u ; dictID : %u ; dictIDSizeCode : %u",
-                !params.fParams.noDictIDFlag, dictID,  dictIDSizeCode);
+                !params.fParams.noDictIDFlag, (unsigned)dictID, (unsigned)dictIDSizeCode);
 
     if (params.format == ZSTD_f_zstd1) {
         MEM_writeLE32(dst, ZSTD_MAGICNUMBER);
@@ -2672,7 +2881,7 @@ static size_t ZSTD_compressContinue_internal (ZSTD_CCtx* cctx,
     size_t fhSize = 0;
 
     DEBUGLOG(5, "ZSTD_compressContinue_internal, stage: %u, srcSize: %u",
-                cctx->stage, (U32)srcSize);
+                cctx->stage, (unsigned)srcSize);
     if (cctx->stage==ZSTDcs_created) return ERROR(stage_wrong);   /* missing init (ZSTD_compressBegin) */
 
     if (frame && (cctx->stage==ZSTDcs_init)) {
@@ -2709,7 +2918,7 @@ static size_t ZSTD_compressContinue_internal (ZSTD_CCtx* cctx,
         }
     }
 
-    DEBUGLOG(5, "ZSTD_compressContinue_internal (blockSize=%u)", (U32)cctx->blockSize);
+    DEBUGLOG(5, "ZSTD_compressContinue_internal (blockSize=%u)", (unsigned)cctx->blockSize);
     {   size_t const cSize = frame ?
                              ZSTD_compress_frameChunk (cctx, dst, dstCapacity, src, srcSize, lastFrameChunk) :
                              ZSTD_compressBlock_internal (cctx, dst, dstCapacity, src, srcSize);
@@ -2721,7 +2930,7 @@ static size_t ZSTD_compressContinue_internal (ZSTD_CCtx* cctx,
             ZSTD_STATIC_ASSERT(ZSTD_CONTENTSIZE_UNKNOWN == (unsigned long long)-1);
             if (cctx->consumedSrcSize+1 > cctx->pledgedSrcSizePlusOne) {
                 DEBUGLOG(4, "error : pledgedSrcSize = %u, while realSrcSize >= %u",
-                    (U32)cctx->pledgedSrcSizePlusOne-1, (U32)cctx->consumedSrcSize);
+                    (unsigned)cctx->pledgedSrcSizePlusOne-1, (unsigned)cctx->consumedSrcSize);
                 return ERROR(srcSize_wrong);
             }
         }
@@ -2733,7 +2942,7 @@ size_t ZSTD_compressContinue (ZSTD_CCtx* cctx,
                               void* dst, size_t dstCapacity,
                         const void* src, size_t srcSize)
 {
-    DEBUGLOG(5, "ZSTD_compressContinue (srcSize=%u)", (U32)srcSize);
+    DEBUGLOG(5, "ZSTD_compressContinue (srcSize=%u)", (unsigned)srcSize);
     return ZSTD_compressContinue_internal(cctx, dst, dstCapacity, src, srcSize, 1 /* frame mode */, 0 /* last chunk */);
 }
 
@@ -2791,6 +3000,7 @@ static size_t ZSTD_loadDictionaryContent(ZSTD_matchState_t* ms,
     case ZSTD_btlazy2:   /* we want the dictionary table fully sorted */
     case ZSTD_btopt:
     case ZSTD_btultra:
+    case ZSTD_btultra2:
         if (srcSize >= HASH_READ_SIZE)
             ZSTD_updateTree(ms, iend-HASH_READ_SIZE, iend);
         break;
@@ -2861,7 +3071,9 @@ static size_t ZSTD_loadZstdDictionary(ZSTD_compressedBlockState_t* bs,
         if (offcodeLog > OffFSELog) return ERROR(dictionary_corrupted);
         /* Defer checking offcodeMaxValue because we need to know the size of the dictionary content */
         /* fill all offset symbols to avoid garbage at end of table */
-        CHECK_E( FSE_buildCTable_wksp(bs->entropy.fse.offcodeCTable, offcodeNCount, MaxOff, offcodeLog, workspace, HUF_WORKSPACE_SIZE),
+        CHECK_E( FSE_buildCTable_wksp(bs->entropy.fse.offcodeCTable,
+                                    offcodeNCount, MaxOff, offcodeLog,
+                                    workspace, HUF_WORKSPACE_SIZE),
                  dictionary_corrupted);
         dictPtr += offcodeHeaderSize;
     }
@@ -2873,7 +3085,9 @@ static size_t ZSTD_loadZstdDictionary(ZSTD_compressedBlockState_t* bs,
         if (matchlengthLog > MLFSELog) return ERROR(dictionary_corrupted);
         /* Every match length code must have non-zero probability */
         CHECK_F( ZSTD_checkDictNCount(matchlengthNCount, matchlengthMaxValue, MaxML));
-        CHECK_E( FSE_buildCTable_wksp(bs->entropy.fse.matchlengthCTable, matchlengthNCount, matchlengthMaxValue, matchlengthLog, workspace, HUF_WORKSPACE_SIZE),
+        CHECK_E( FSE_buildCTable_wksp(bs->entropy.fse.matchlengthCTable,
+                                    matchlengthNCount, matchlengthMaxValue, matchlengthLog,
+                                    workspace, HUF_WORKSPACE_SIZE),
                  dictionary_corrupted);
         dictPtr += matchlengthHeaderSize;
     }
@@ -2885,7 +3099,9 @@ static size_t ZSTD_loadZstdDictionary(ZSTD_compressedBlockState_t* bs,
         if (litlengthLog > LLFSELog) return ERROR(dictionary_corrupted);
         /* Every literal length code must have non-zero probability */
         CHECK_F( ZSTD_checkDictNCount(litlengthNCount, litlengthMaxValue, MaxLL));
-        CHECK_E( FSE_buildCTable_wksp(bs->entropy.fse.litlengthCTable, litlengthNCount, litlengthMaxValue, litlengthLog, workspace, HUF_WORKSPACE_SIZE),
+        CHECK_E( FSE_buildCTable_wksp(bs->entropy.fse.litlengthCTable,
+                                    litlengthNCount, litlengthMaxValue, litlengthLog,
+                                    workspace, HUF_WORKSPACE_SIZE),
                  dictionary_corrupted);
         dictPtr += litlengthHeaderSize;
     }
@@ -3023,7 +3239,7 @@ size_t ZSTD_compressBegin_usingDict(ZSTD_CCtx* cctx, const void* dict, size_t di
     ZSTD_parameters const params = ZSTD_getParams(compressionLevel, ZSTD_CONTENTSIZE_UNKNOWN, dictSize);
     ZSTD_CCtx_params const cctxParams =
             ZSTD_assignParamsToCCtxParams(cctx->requestedParams, params);
-    DEBUGLOG(4, "ZSTD_compressBegin_usingDict (dictSize=%u)", (U32)dictSize);
+    DEBUGLOG(4, "ZSTD_compressBegin_usingDict (dictSize=%u)", (unsigned)dictSize);
     return ZSTD_compressBegin_internal(cctx, dict, dictSize, ZSTD_dct_auto, ZSTD_dtlm_fast, NULL,
                                        cctxParams, ZSTD_CONTENTSIZE_UNKNOWN, ZSTDb_not_buffered);
 }
@@ -3067,7 +3283,7 @@ static size_t ZSTD_writeEpilogue(ZSTD_CCtx* cctx, void* dst, size_t dstCapacity)
     if (cctx->appliedParams.fParams.checksumFlag) {
         U32 const checksum = (U32) XXH64_digest(&cctx->xxhState);
         if (dstCapacity<4) return ERROR(dstSize_tooSmall);
-        DEBUGLOG(4, "ZSTD_writeEpilogue: write checksum : %08X", checksum);
+        DEBUGLOG(4, "ZSTD_writeEpilogue: write checksum : %08X", (unsigned)checksum);
         MEM_writeLE32(op, checksum);
         op += 4;
     }
@@ -3093,7 +3309,7 @@ size_t ZSTD_compressEnd (ZSTD_CCtx* cctx,
         DEBUGLOG(4, "end of frame : controlling src size");
         if (cctx->pledgedSrcSizePlusOne != cctx->consumedSrcSize+1) {
             DEBUGLOG(4, "error : pledgedSrcSize = %u, while realSrcSize = %u",
-                (U32)cctx->pledgedSrcSizePlusOne-1, (U32)cctx->consumedSrcSize);
+                (unsigned)cctx->pledgedSrcSizePlusOne-1, (unsigned)cctx->consumedSrcSize);
             return ERROR(srcSize_wrong);
     }   }
     return cSize + endResult;
@@ -3139,7 +3355,7 @@ size_t ZSTD_compress_advanced_internal(
         const void* dict,size_t dictSize,
         ZSTD_CCtx_params params)
 {
-    DEBUGLOG(4, "ZSTD_compress_advanced_internal (srcSize:%u)", (U32)srcSize);
+    DEBUGLOG(4, "ZSTD_compress_advanced_internal (srcSize:%u)", (unsigned)srcSize);
     CHECK_F( ZSTD_compressBegin_internal(cctx,
                          dict, dictSize, ZSTD_dct_auto, ZSTD_dtlm_fast, NULL,
                          params, srcSize, ZSTDb_not_buffered) );
@@ -3163,7 +3379,7 @@ size_t ZSTD_compressCCtx(ZSTD_CCtx* cctx,
                    const void* src, size_t srcSize,
                          int compressionLevel)
 {
-    DEBUGLOG(4, "ZSTD_compressCCtx (srcSize=%u)", (U32)srcSize);
+    DEBUGLOG(4, "ZSTD_compressCCtx (srcSize=%u)", (unsigned)srcSize);
     assert(cctx != NULL);
     return ZSTD_compress_usingDict(cctx, dst, dstCapacity, src, srcSize, NULL, 0, compressionLevel);
 }
@@ -3189,7 +3405,7 @@ size_t ZSTD_estimateCDictSize_advanced(
         size_t dictSize, ZSTD_compressionParameters cParams,
         ZSTD_dictLoadMethod_e dictLoadMethod)
 {
-    DEBUGLOG(5, "sizeof(ZSTD_CDict) : %u", (U32)sizeof(ZSTD_CDict));
+    DEBUGLOG(5, "sizeof(ZSTD_CDict) : %u", (unsigned)sizeof(ZSTD_CDict));
     return sizeof(ZSTD_CDict) + HUF_WORKSPACE_SIZE + ZSTD_sizeof_matchState(&cParams, /* forCCtx */ 0)
            + (dictLoadMethod == ZSTD_dlm_byRef ? 0 : dictSize);
 }
@@ -3203,7 +3419,7 @@ size_t ZSTD_estimateCDictSize(size_t dictSize, int compressionLevel)
 size_t ZSTD_sizeof_CDict(const ZSTD_CDict* cdict)
 {
     if (cdict==NULL) return 0;   /* support sizeof on NULL */
-    DEBUGLOG(5, "sizeof(*cdict) : %u", (U32)sizeof(*cdict));
+    DEBUGLOG(5, "sizeof(*cdict) : %u", (unsigned)sizeof(*cdict));
     return cdict->workspaceSize + (cdict->dictBuffer ? cdict->dictContentSize : 0) + sizeof(*cdict);
 }
 
@@ -3214,7 +3430,7 @@ static size_t ZSTD_initCDict_internal(
                     ZSTD_dictContentType_e dictContentType,
                     ZSTD_compressionParameters cParams)
 {
-    DEBUGLOG(3, "ZSTD_initCDict_internal (dictContentType:%u)", (U32)dictContentType);
+    DEBUGLOG(3, "ZSTD_initCDict_internal (dictContentType:%u)", (unsigned)dictContentType);
     assert(!ZSTD_checkCParams(cParams));
     cdict->matchState.cParams = cParams;
     if ((dictLoadMethod == ZSTD_dlm_byRef) || (!dictBuffer) || (!dictSize)) {
@@ -3264,7 +3480,7 @@ ZSTD_CDict* ZSTD_createCDict_advanced(const void* dictBuffer, size_t dictSize,
                                       ZSTD_dictContentType_e dictContentType,
                                       ZSTD_compressionParameters cParams, ZSTD_customMem customMem)
 {
-    DEBUGLOG(3, "ZSTD_createCDict_advanced, mode %u", (U32)dictContentType);
+    DEBUGLOG(3, "ZSTD_createCDict_advanced, mode %u", (unsigned)dictContentType);
     if (!customMem.customAlloc ^ !customMem.customFree) return NULL;
 
     {   ZSTD_CDict* const cdict = (ZSTD_CDict*)ZSTD_malloc(sizeof(ZSTD_CDict), customMem);
@@ -3345,7 +3561,7 @@ const ZSTD_CDict* ZSTD_initStaticCDict(
     void* ptr;
     if ((size_t)workspace & 7) return NULL;  /* 8-aligned */
     DEBUGLOG(4, "(workspaceSize < neededSize) : (%u < %u) => %u",
-        (U32)workspaceSize, (U32)neededSize, (U32)(workspaceSize < neededSize));
+        (unsigned)workspaceSize, (unsigned)neededSize, (unsigned)(workspaceSize < neededSize));
     if (workspaceSize < neededSize) return NULL;
 
     if (dictLoadMethod == ZSTD_dlm_byCopy) {
@@ -3505,7 +3721,7 @@ static size_t ZSTD_resetCStream_internal(ZSTD_CStream* cctx,
 size_t ZSTD_resetCStream(ZSTD_CStream* zcs, unsigned long long pledgedSrcSize)
 {
     ZSTD_CCtx_params params = zcs->requestedParams;
-    DEBUGLOG(4, "ZSTD_resetCStream: pledgedSrcSize = %u", (U32)pledgedSrcSize);
+    DEBUGLOG(4, "ZSTD_resetCStream: pledgedSrcSize = %u", (unsigned)pledgedSrcSize);
     if (pledgedSrcSize==0) pledgedSrcSize = ZSTD_CONTENTSIZE_UNKNOWN;
     params.fParams.contentSizeFlag = 1;
     return ZSTD_resetCStream_internal(zcs, NULL, 0, ZSTD_dct_auto, zcs->cdict, params, pledgedSrcSize);
@@ -3525,7 +3741,7 @@ size_t ZSTD_initCStream_internal(ZSTD_CStream* zcs,
     assert(!((dict) && (cdict)));  /* either dict or cdict, not both */
 
     if (dict && dictSize >= 8) {
-        DEBUGLOG(4, "loading dictionary of size %u", (U32)dictSize);
+        DEBUGLOG(4, "loading dictionary of size %u", (unsigned)dictSize);
         if (zcs->staticSize) {   /* static CCtx : never uses malloc */
             /* incompatible with internal cdict creation */
             return ERROR(memory_allocation);
@@ -3584,7 +3800,7 @@ size_t ZSTD_initCStream_advanced(ZSTD_CStream* zcs,
                                  ZSTD_parameters params, unsigned long long pledgedSrcSize)
 {
     DEBUGLOG(4, "ZSTD_initCStream_advanced: pledgedSrcSize=%u, flag=%u",
-                (U32)pledgedSrcSize, params.fParams.contentSizeFlag);
+                (unsigned)pledgedSrcSize, params.fParams.contentSizeFlag);
     CHECK_F( ZSTD_checkCParams(params.cParams) );
     if ((pledgedSrcSize==0) && (params.fParams.contentSizeFlag==0)) pledgedSrcSize = ZSTD_CONTENTSIZE_UNKNOWN;  /* for compatibility with older programs relying on this behavior. Users should now specify ZSTD_CONTENTSIZE_UNKNOWN. This line will be removed in the future. */
     zcs->requestedParams = ZSTD_assignParamsToCCtxParams(zcs->requestedParams, params);
@@ -3612,8 +3828,15 @@ size_t ZSTD_initCStream(ZSTD_CStream* zcs, int compressionLevel)
 
 /*======   Compression   ======*/
 
-MEM_STATIC size_t ZSTD_limitCopy(void* dst, size_t dstCapacity,
-                           const void* src, size_t srcSize)
+static size_t ZSTD_nextInputSizeHint(const ZSTD_CCtx* cctx)
+{
+    size_t hintInSize = cctx->inBuffTarget - cctx->inBuffPos;
+    if (hintInSize==0) hintInSize = cctx->blockSize;
+    return hintInSize;
+}
+
+static size_t ZSTD_limitCopy(void* dst, size_t dstCapacity,
+                       const void* src, size_t srcSize)
 {
     size_t const length = MIN(dstCapacity, srcSize);
     if (length) memcpy(dst, src, length);
@@ -3621,7 +3844,7 @@ MEM_STATIC size_t ZSTD_limitCopy(void* dst, size_t dstCapacity,
 }
 
 /** ZSTD_compressStream_generic():
- *  internal function for all *compressStream*() variants and *compress_generic()
+ *  internal function for all *compressStream*() variants
  *  non-static, because can be called from zstdmt_compress.c
  * @return : hint size for next input */
 size_t ZSTD_compressStream_generic(ZSTD_CStream* zcs,
@@ -3638,7 +3861,7 @@ size_t ZSTD_compressStream_generic(ZSTD_CStream* zcs,
     U32 someMoreWork = 1;
 
     /* check expectations */
-    DEBUGLOG(5, "ZSTD_compressStream_generic, flush=%u", (U32)flushMode);
+    DEBUGLOG(5, "ZSTD_compressStream_generic, flush=%u", (unsigned)flushMode);
     assert(zcs->inBuff != NULL);
     assert(zcs->inBuffSize > 0);
     assert(zcs->outBuff !=  NULL);
@@ -3660,12 +3883,12 @@ size_t ZSTD_compressStream_generic(ZSTD_CStream* zcs,
                 /* shortcut to compression pass directly into output buffer */
                 size_t const cSize = ZSTD_compressEnd(zcs,
                                                 op, oend-op, ip, iend-ip);
-                DEBUGLOG(4, "ZSTD_compressEnd : %u", (U32)cSize);
+                DEBUGLOG(4, "ZSTD_compressEnd : cSize=%u", (unsigned)cSize);
                 if (ZSTD_isError(cSize)) return cSize;
                 ip = iend;
                 op += cSize;
                 zcs->frameEnded = 1;
-                ZSTD_CCtx_reset(zcs);
+                ZSTD_CCtx_reset(zcs, ZSTD_reset_session_only);
                 someMoreWork = 0; break;
             }
             /* complete loading into inBuffer */
@@ -3709,7 +3932,7 @@ size_t ZSTD_compressStream_generic(ZSTD_CStream* zcs,
                 if (zcs->inBuffTarget > zcs->inBuffSize)
                     zcs->inBuffPos = 0, zcs->inBuffTarget = zcs->blockSize;
                 DEBUGLOG(5, "inBuffTarget:%u / inBuffSize:%u",
-                         (U32)zcs->inBuffTarget, (U32)zcs->inBuffSize);
+                         (unsigned)zcs->inBuffTarget, (unsigned)zcs->inBuffSize);
                 if (!lastBlock)
                     assert(zcs->inBuffTarget <= zcs->inBuffSize);
                 zcs->inToCompress = zcs->inBuffPos;
@@ -3718,7 +3941,7 @@ size_t ZSTD_compressStream_generic(ZSTD_CStream* zcs,
                     if (zcs->frameEnded) {
                         DEBUGLOG(5, "Frame completed directly in outBuffer");
                         someMoreWork = 0;
-                        ZSTD_CCtx_reset(zcs);
+                        ZSTD_CCtx_reset(zcs, ZSTD_reset_session_only);
                     }
                     break;
                 }
@@ -3733,7 +3956,7 @@ size_t ZSTD_compressStream_generic(ZSTD_CStream* zcs,
                 size_t const flushed = ZSTD_limitCopy(op, oend-op,
                             zcs->outBuff + zcs->outBuffFlushedSize, toFlush);
                 DEBUGLOG(5, "toFlush: %u into %u ==> flushed: %u",
-                            (U32)toFlush, (U32)(oend-op), (U32)flushed);
+                            (unsigned)toFlush, (unsigned)(oend-op), (unsigned)flushed);
                 op += flushed;
                 zcs->outBuffFlushedSize += flushed;
                 if (toFlush!=flushed) {
@@ -3746,7 +3969,7 @@ size_t ZSTD_compressStream_generic(ZSTD_CStream* zcs,
                 if (zcs->frameEnded) {
                     DEBUGLOG(5, "Frame completed on flush");
                     someMoreWork = 0;
-                    ZSTD_CCtx_reset(zcs);
+                    ZSTD_CCtx_reset(zcs, ZSTD_reset_session_only);
                     break;
                 }
                 zcs->streamStage = zcss_load;
@@ -3761,28 +3984,34 @@ size_t ZSTD_compressStream_generic(ZSTD_CStream* zcs,
     input->pos = ip - istart;
     output->pos = op - ostart;
     if (zcs->frameEnded) return 0;
-    {   size_t hintInSize = zcs->inBuffTarget - zcs->inBuffPos;
-        if (hintInSize==0) hintInSize = zcs->blockSize;
-        return hintInSize;
+    return ZSTD_nextInputSizeHint(zcs);
+}
+
+static size_t ZSTD_nextInputSizeHint_MTorST(const ZSTD_CCtx* cctx)
+{
+#ifdef ZSTD_MULTITHREAD
+    if (cctx->appliedParams.nbWorkers >= 1) {
+        assert(cctx->mtctx != NULL);
+        return ZSTDMT_nextInputSizeHint(cctx->mtctx);
     }
+#endif
+    return ZSTD_nextInputSizeHint(cctx);
+
 }
 
 size_t ZSTD_compressStream(ZSTD_CStream* zcs, ZSTD_outBuffer* output, ZSTD_inBuffer* input)
 {
-    /* check conditions */
-    if (output->pos > output->size) return ERROR(GENERIC);
-    if (input->pos  > input->size)  return ERROR(GENERIC);
-
-    return ZSTD_compressStream_generic(zcs, output, input, ZSTD_e_continue);
+    CHECK_F( ZSTD_compressStream2(zcs, output, input, ZSTD_e_continue) );
+    return ZSTD_nextInputSizeHint_MTorST(zcs);
 }
 
 
-size_t ZSTD_compress_generic (ZSTD_CCtx* cctx,
-                              ZSTD_outBuffer* output,
-                              ZSTD_inBuffer* input,
-                              ZSTD_EndDirective endOp)
+size_t ZSTD_compressStream2( ZSTD_CCtx* cctx,
+                             ZSTD_outBuffer* output,
+                             ZSTD_inBuffer* input,
+                             ZSTD_EndDirective endOp)
 {
-    DEBUGLOG(5, "ZSTD_compress_generic, endOp=%u ", (U32)endOp);
+    DEBUGLOG(5, "ZSTD_compressStream2, endOp=%u ", (unsigned)endOp);
     /* check conditions */
     if (output->pos > output->size) return ERROR(GENERIC);
     if (input->pos  > input->size)  return ERROR(GENERIC);
@@ -3792,9 +4021,9 @@ size_t ZSTD_compress_generic (ZSTD_CCtx* cctx,
     if (cctx->streamStage == zcss_init) {
         ZSTD_CCtx_params params = cctx->requestedParams;
         ZSTD_prefixDict const prefixDict = cctx->prefixDict;
-        memset(&cctx->prefixDict, 0, sizeof(cctx->prefixDict));  /* single usage */
-        assert(prefixDict.dict==NULL || cctx->cdict==NULL);   /* only one can be set */
-        DEBUGLOG(4, "ZSTD_compress_generic : transparent init stage");
+        memset(&cctx->prefixDict, 0, sizeof(cctx->prefixDict));   /* single usage */
+        assert(prefixDict.dict==NULL || cctx->cdict==NULL);    /* only one can be set */
+        DEBUGLOG(4, "ZSTD_compressStream2 : transparent init stage");
         if (endOp == ZSTD_e_end) cctx->pledgedSrcSizePlusOne = input->size + 1;  /* auto-fix pledgedSrcSize */
         params.cParams = ZSTD_getCParamsFromCCtxParams(
                 &cctx->requestedParams, cctx->pledgedSrcSizePlusOne-1, 0 /*dictSize*/);
@@ -3807,7 +4036,7 @@ size_t ZSTD_compress_generic (ZSTD_CCtx* cctx,
         if (params.nbWorkers > 0) {
             /* mt context creation */
             if (cctx->mtctx == NULL) {
-                DEBUGLOG(4, "ZSTD_compress_generic: creating new mtctx for nbWorkers=%u",
+                DEBUGLOG(4, "ZSTD_compressStream2: creating new mtctx for nbWorkers=%u",
                             params.nbWorkers);
                 cctx->mtctx = ZSTDMT_createCCtx_advanced(params.nbWorkers, cctx->customMem);
                 if (cctx->mtctx == NULL) return ERROR(memory_allocation);
@@ -3829,6 +4058,7 @@ size_t ZSTD_compress_generic (ZSTD_CCtx* cctx,
             assert(cctx->streamStage == zcss_load);
             assert(cctx->appliedParams.nbWorkers == 0);
     }   }
+    /* end of transparent initialization stage */
 
     /* compression stage */
 #ifdef ZSTD_MULTITHREAD
@@ -3840,18 +4070,18 @@ size_t ZSTD_compress_generic (ZSTD_CCtx* cctx,
         {   size_t const flushMin = ZSTDMT_compressStream_generic(cctx->mtctx, output, input, endOp);
             if ( ZSTD_isError(flushMin)
               || (endOp == ZSTD_e_end && flushMin == 0) ) { /* compression completed */
-                ZSTD_CCtx_reset(cctx);
+                ZSTD_CCtx_reset(cctx, ZSTD_reset_session_only);
             }
-            DEBUGLOG(5, "completed ZSTD_compress_generic delegating to ZSTDMT_compressStream_generic");
+            DEBUGLOG(5, "completed ZSTD_compressStream2 delegating to ZSTDMT_compressStream_generic");
             return flushMin;
     }   }
 #endif
     CHECK_F( ZSTD_compressStream_generic(cctx, output, input, endOp) );
-    DEBUGLOG(5, "completed ZSTD_compress_generic");
+    DEBUGLOG(5, "completed ZSTD_compressStream2");
     return cctx->outBuffContentSize - cctx->outBuffFlushedSize; /* remaining to flush */
 }
 
-size_t ZSTD_compress_generic_simpleArgs (
+size_t ZSTD_compressStream2_simpleArgs (
                             ZSTD_CCtx* cctx,
                             void* dst, size_t dstCapacity, size_t* dstPos,
                       const void* src, size_t srcSize, size_t* srcPos,
@@ -3859,13 +4089,33 @@ size_t ZSTD_compress_generic_simpleArgs (
 {
     ZSTD_outBuffer output = { dst, dstCapacity, *dstPos };
     ZSTD_inBuffer  input  = { src, srcSize, *srcPos };
-    /* ZSTD_compress_generic() will check validity of dstPos and srcPos */
-    size_t const cErr = ZSTD_compress_generic(cctx, &output, &input, endOp);
+    /* ZSTD_compressStream2() will check validity of dstPos and srcPos */
+    size_t const cErr = ZSTD_compressStream2(cctx, &output, &input, endOp);
     *dstPos = output.pos;
     *srcPos = input.pos;
     return cErr;
 }
 
+size_t ZSTD_compress2(ZSTD_CCtx* cctx,
+                      void* dst, size_t dstCapacity,
+                      const void* src, size_t srcSize)
+{
+    ZSTD_CCtx_reset(cctx, ZSTD_reset_session_only);
+    {   size_t oPos = 0;
+        size_t iPos = 0;
+        size_t const result = ZSTD_compressStream2_simpleArgs(cctx,
+                                        dst, dstCapacity, &oPos,
+                                        src, srcSize, &iPos,
+                                        ZSTD_e_end);
+        if (ZSTD_isError(result)) return result;
+        if (result != 0) {  /* compression not completed, due to lack of output space */
+            assert(oPos == dstCapacity);
+            return ERROR(dstSize_tooSmall);
+        }
+        assert(iPos == srcSize);   /* all input is expected consumed */
+        return oPos;
+    }
+}
 
 /*======   Finalize   ======*/
 
@@ -3874,21 +4124,21 @@ size_t ZSTD_compress_generic_simpleArgs (
 size_t ZSTD_flushStream(ZSTD_CStream* zcs, ZSTD_outBuffer* output)
 {
     ZSTD_inBuffer input = { NULL, 0, 0 };
-    if (output->pos > output->size) return ERROR(GENERIC);
-    CHECK_F( ZSTD_compressStream_generic(zcs, output, &input, ZSTD_e_flush) );
-    return zcs->outBuffContentSize - zcs->outBuffFlushedSize;  /* remaining to flush */
+    return ZSTD_compressStream2(zcs, output, &input, ZSTD_e_flush);
 }
 
 
 size_t ZSTD_endStream(ZSTD_CStream* zcs, ZSTD_outBuffer* output)
 {
     ZSTD_inBuffer input = { NULL, 0, 0 };
-    if (output->pos > output->size) return ERROR(GENERIC);
-    CHECK_F( ZSTD_compressStream_generic(zcs, output, &input, ZSTD_e_end) );
+    size_t const remainingToFlush = ZSTD_compressStream2(zcs, output, &input, ZSTD_e_end);
+    CHECK_F( remainingToFlush );
+    if (zcs->appliedParams.nbWorkers > 0) return remainingToFlush;   /* minimal estimation */
+    /* single thread mode : attempt to calculate remaining to flush more precisely */
     {   size_t const lastBlockSize = zcs->frameEnded ? 0 : ZSTD_BLOCKHEADERSIZE;
         size_t const checksumSize = zcs->frameEnded ? 0 : zcs->appliedParams.fParams.checksumFlag * 4;
-        size_t const toFlush = zcs->outBuffContentSize - zcs->outBuffFlushedSize + lastBlockSize + checksumSize;
-        DEBUGLOG(4, "ZSTD_endStream : remaining to flush : %u", (U32)toFlush);
+        size_t const toFlush = remainingToFlush + lastBlockSize + checksumSize;
+        DEBUGLOG(4, "ZSTD_endStream : remaining to flush : %u", (unsigned)toFlush);
         return toFlush;
     }
 }
@@ -3905,27 +4155,27 @@ static const ZSTD_compressionParameters ZSTD_defaultCParameters[4][ZSTD_MAX_CLEV
     /* W,  C,  H,  S,  L, TL, strat */
     { 19, 12, 13,  1,  6,  1, ZSTD_fast    },  /* base for negative levels */
     { 19, 13, 14,  1,  7,  0, ZSTD_fast    },  /* level  1 */
-    { 19, 15, 16,  1,  6,  0, ZSTD_fast    },  /* level  2 */
-    { 20, 16, 17,  1,  5,  1, ZSTD_dfast   },  /* level  3 */
-    { 20, 18, 18,  1,  5,  1, ZSTD_dfast   },  /* level  4 */
-    { 20, 18, 18,  2,  5,  2, ZSTD_greedy  },  /* level  5 */
-    { 21, 18, 19,  2,  5,  4, ZSTD_lazy    },  /* level  6 */
-    { 21, 18, 19,  3,  5,  8, ZSTD_lazy2   },  /* level  7 */
+    { 20, 15, 16,  1,  6,  0, ZSTD_fast    },  /* level  2 */
+    { 21, 16, 17,  1,  5,  1, ZSTD_dfast   },  /* level  3 */
+    { 21, 18, 18,  1,  5,  1, ZSTD_dfast   },  /* level  4 */
+    { 21, 18, 19,  2,  5,  2, ZSTD_greedy  },  /* level  5 */
+    { 21, 19, 19,  3,  5,  4, ZSTD_greedy  },  /* level  6 */
+    { 21, 19, 19,  3,  5,  8, ZSTD_lazy    },  /* level  7 */
     { 21, 19, 19,  3,  5, 16, ZSTD_lazy2   },  /* level  8 */
     { 21, 19, 20,  4,  5, 16, ZSTD_lazy2   },  /* level  9 */
-    { 21, 20, 21,  4,  5, 16, ZSTD_lazy2   },  /* level 10 */
-    { 21, 21, 22,  4,  5, 16, ZSTD_lazy2   },  /* level 11 */
-    { 22, 20, 22,  5,  5, 16, ZSTD_lazy2   },  /* level 12 */
-    { 22, 21, 22,  4,  5, 32, ZSTD_btlazy2 },  /* level 13 */
-    { 22, 21, 22,  5,  5, 32, ZSTD_btlazy2 },  /* level 14 */
-    { 22, 22, 22,  6,  5, 32, ZSTD_btlazy2 },  /* level 15 */
-    { 22, 21, 22,  4,  5, 48, ZSTD_btopt   },  /* level 16 */
-    { 23, 22, 22,  4,  4, 64, ZSTD_btopt   },  /* level 17 */
-    { 23, 23, 22,  6,  3,256, ZSTD_btopt   },  /* level 18 */
-    { 23, 24, 22,  7,  3,256, ZSTD_btultra },  /* level 19 */
-    { 25, 25, 23,  7,  3,256, ZSTD_btultra },  /* level 20 */
-    { 26, 26, 24,  7,  3,512, ZSTD_btultra },  /* level 21 */
-    { 27, 27, 25,  9,  3,999, ZSTD_btultra },  /* level 22 */
+    { 22, 20, 21,  4,  5, 16, ZSTD_lazy2   },  /* level 10 */
+    { 22, 21, 22,  4,  5, 16, ZSTD_lazy2   },  /* level 11 */
+    { 22, 21, 22,  5,  5, 16, ZSTD_lazy2   },  /* level 12 */
+    { 22, 21, 22,  5,  5, 32, ZSTD_btlazy2 },  /* level 13 */
+    { 22, 22, 23,  5,  5, 32, ZSTD_btlazy2 },  /* level 14 */
+    { 22, 23, 23,  6,  5, 32, ZSTD_btlazy2 },  /* level 15 */
+    { 22, 22, 22,  5,  5, 48, ZSTD_btopt   },  /* level 16 */
+    { 23, 23, 22,  5,  4, 64, ZSTD_btopt   },  /* level 17 */
+    { 23, 23, 22,  6,  3, 64, ZSTD_btultra },  /* level 18 */
+    { 23, 24, 22,  7,  3,256, ZSTD_btultra2},  /* level 19 */
+    { 25, 25, 23,  7,  3,256, ZSTD_btultra2},  /* level 20 */
+    { 26, 26, 24,  7,  3,512, ZSTD_btultra2},  /* level 21 */
+    { 27, 27, 25,  9,  3,999, ZSTD_btultra2},  /* level 22 */
 },
 {   /* for srcSize <= 256 KB */
     /* W,  C,  H,  S,  L,  T, strat */
@@ -3940,18 +4190,18 @@ static const ZSTD_compressionParameters ZSTD_defaultCParameters[4][ZSTD_MAX_CLEV
     { 18, 18, 19,  4,  4,  8, ZSTD_lazy2   },  /* level  8 */
     { 18, 18, 19,  5,  4,  8, ZSTD_lazy2   },  /* level  9 */
     { 18, 18, 19,  6,  4,  8, ZSTD_lazy2   },  /* level 10 */
-    { 18, 18, 19,  5,  4, 16, ZSTD_btlazy2 },  /* level 11.*/
-    { 18, 19, 19,  6,  4, 16, ZSTD_btlazy2 },  /* level 12.*/
-    { 18, 19, 19,  8,  4, 16, ZSTD_btlazy2 },  /* level 13 */
-    { 18, 18, 19,  4,  4, 24, ZSTD_btopt   },  /* level 14.*/
-    { 18, 18, 19,  4,  3, 24, ZSTD_btopt   },  /* level 15.*/
-    { 18, 19, 19,  6,  3, 64, ZSTD_btopt   },  /* level 16.*/
-    { 18, 19, 19,  8,  3,128, ZSTD_btopt   },  /* level 17.*/
-    { 18, 19, 19, 10,  3,256, ZSTD_btopt   },  /* level 18.*/
-    { 18, 19, 19, 10,  3,256, ZSTD_btultra },  /* level 19.*/
-    { 18, 19, 19, 11,  3,512, ZSTD_btultra },  /* level 20.*/
-    { 18, 19, 19, 12,  3,512, ZSTD_btultra },  /* level 21.*/
-    { 18, 19, 19, 13,  3,999, ZSTD_btultra },  /* level 22.*/
+    { 18, 18, 19,  5,  4, 12, ZSTD_btlazy2 },  /* level 11.*/
+    { 18, 19, 19,  7,  4, 12, ZSTD_btlazy2 },  /* level 12.*/
+    { 18, 18, 19,  4,  4, 16, ZSTD_btopt   },  /* level 13 */
+    { 18, 18, 19,  4,  3, 32, ZSTD_btopt   },  /* level 14.*/
+    { 18, 18, 19,  6,  3,128, ZSTD_btopt   },  /* level 15.*/
+    { 18, 19, 19,  6,  3,128, ZSTD_btultra },  /* level 16.*/
+    { 18, 19, 19,  8,  3,256, ZSTD_btultra },  /* level 17.*/
+    { 18, 19, 19,  6,  3,128, ZSTD_btultra2},  /* level 18.*/
+    { 18, 19, 19,  8,  3,256, ZSTD_btultra2},  /* level 19.*/
+    { 18, 19, 19, 10,  3,512, ZSTD_btultra2},  /* level 20.*/
+    { 18, 19, 19, 12,  3,512, ZSTD_btultra2},  /* level 21.*/
+    { 18, 19, 19, 13,  3,999, ZSTD_btultra2},  /* level 22.*/
 },
 {   /* for srcSize <= 128 KB */
     /* W,  C,  H,  S,  L,  T, strat */
@@ -3966,26 +4216,26 @@ static const ZSTD_compressionParameters ZSTD_defaultCParameters[4][ZSTD_MAX_CLEV
     { 17, 17, 17,  4,  4,  8, ZSTD_lazy2   },  /* level  8 */
     { 17, 17, 17,  5,  4,  8, ZSTD_lazy2   },  /* level  9 */
     { 17, 17, 17,  6,  4,  8, ZSTD_lazy2   },  /* level 10 */
-    { 17, 17, 17,  7,  4,  8, ZSTD_lazy2   },  /* level 11 */
-    { 17, 18, 17,  6,  4, 16, ZSTD_btlazy2 },  /* level 12 */
-    { 17, 18, 17,  8,  4, 16, ZSTD_btlazy2 },  /* level 13.*/
-    { 17, 18, 17,  4,  4, 32, ZSTD_btopt   },  /* level 14.*/
-    { 17, 18, 17,  6,  3, 64, ZSTD_btopt   },  /* level 15.*/
-    { 17, 18, 17,  7,  3,128, ZSTD_btopt   },  /* level 16.*/
-    { 17, 18, 17,  7,  3,256, ZSTD_btopt   },  /* level 17.*/
-    { 17, 18, 17,  8,  3,256, ZSTD_btopt   },  /* level 18.*/
-    { 17, 18, 17,  8,  3,256, ZSTD_btultra },  /* level 19.*/
-    { 17, 18, 17,  9,  3,256, ZSTD_btultra },  /* level 20.*/
-    { 17, 18, 17, 10,  3,256, ZSTD_btultra },  /* level 21.*/
-    { 17, 18, 17, 11,  3,512, ZSTD_btultra },  /* level 22.*/
+    { 17, 17, 17,  5,  4,  8, ZSTD_btlazy2 },  /* level 11 */
+    { 17, 18, 17,  7,  4, 12, ZSTD_btlazy2 },  /* level 12 */
+    { 17, 18, 17,  3,  4, 12, ZSTD_btopt   },  /* level 13.*/
+    { 17, 18, 17,  4,  3, 32, ZSTD_btopt   },  /* level 14.*/
+    { 17, 18, 17,  6,  3,256, ZSTD_btopt   },  /* level 15.*/
+    { 17, 18, 17,  6,  3,128, ZSTD_btultra },  /* level 16.*/
+    { 17, 18, 17,  8,  3,256, ZSTD_btultra },  /* level 17.*/
+    { 17, 18, 17, 10,  3,512, ZSTD_btultra },  /* level 18.*/
+    { 17, 18, 17,  5,  3,256, ZSTD_btultra2},  /* level 19.*/
+    { 17, 18, 17,  7,  3,512, ZSTD_btultra2},  /* level 20.*/
+    { 17, 18, 17,  9,  3,512, ZSTD_btultra2},  /* level 21.*/
+    { 17, 18, 17, 11,  3,999, ZSTD_btultra2},  /* level 22.*/
 },
 {   /* for srcSize <= 16 KB */
     /* W,  C,  H,  S,  L,  T, strat */
     { 14, 12, 13,  1,  5,  1, ZSTD_fast    },  /* base for negative levels */
     { 14, 14, 15,  1,  5,  0, ZSTD_fast    },  /* level  1 */
     { 14, 14, 15,  1,  4,  0, ZSTD_fast    },  /* level  2 */
-    { 14, 14, 14,  2,  4,  1, ZSTD_dfast   },  /* level  3.*/
-    { 14, 14, 14,  4,  4,  2, ZSTD_greedy  },  /* level  4.*/
+    { 14, 14, 15,  2,  4,  1, ZSTD_dfast   },  /* level  3 */
+    { 14, 14, 14,  4,  4,  2, ZSTD_greedy  },  /* level  4 */
     { 14, 14, 14,  3,  4,  4, ZSTD_lazy    },  /* level  5.*/
     { 14, 14, 14,  4,  4,  8, ZSTD_lazy2   },  /* level  6 */
     { 14, 14, 14,  6,  4,  8, ZSTD_lazy2   },  /* level  7 */
@@ -3993,17 +4243,17 @@ static const ZSTD_compressionParameters ZSTD_defaultCParameters[4][ZSTD_MAX_CLEV
     { 14, 15, 14,  5,  4,  8, ZSTD_btlazy2 },  /* level  9.*/
     { 14, 15, 14,  9,  4,  8, ZSTD_btlazy2 },  /* level 10.*/
     { 14, 15, 14,  3,  4, 12, ZSTD_btopt   },  /* level 11.*/
-    { 14, 15, 14,  6,  3, 16, ZSTD_btopt   },  /* level 12.*/
-    { 14, 15, 14,  6,  3, 24, ZSTD_btopt   },  /* level 13.*/
-    { 14, 15, 15,  6,  3, 48, ZSTD_btopt   },  /* level 14.*/
-    { 14, 15, 15,  6,  3, 64, ZSTD_btopt   },  /* level 15.*/
-    { 14, 15, 15,  6,  3, 96, ZSTD_btopt   },  /* level 16.*/
-    { 14, 15, 15,  6,  3,128, ZSTD_btopt   },  /* level 17.*/
-    { 14, 15, 15,  8,  3,256, ZSTD_btopt   },  /* level 18.*/
-    { 14, 15, 15,  6,  3,256, ZSTD_btultra },  /* level 19.*/
-    { 14, 15, 15,  8,  3,256, ZSTD_btultra },  /* level 20.*/
-    { 14, 15, 15,  9,  3,256, ZSTD_btultra },  /* level 21.*/
-    { 14, 15, 15, 10,  3,512, ZSTD_btultra },  /* level 22.*/
+    { 14, 15, 14,  4,  3, 24, ZSTD_btopt   },  /* level 12.*/
+    { 14, 15, 14,  5,  3, 32, ZSTD_btultra },  /* level 13.*/
+    { 14, 15, 15,  6,  3, 64, ZSTD_btultra },  /* level 14.*/
+    { 14, 15, 15,  7,  3,256, ZSTD_btultra },  /* level 15.*/
+    { 14, 15, 15,  5,  3, 48, ZSTD_btultra2},  /* level 16.*/
+    { 14, 15, 15,  6,  3,128, ZSTD_btultra2},  /* level 17.*/
+    { 14, 15, 15,  7,  3,256, ZSTD_btultra2},  /* level 18.*/
+    { 14, 15, 15,  8,  3,256, ZSTD_btultra2},  /* level 19.*/
+    { 14, 15, 15,  8,  3,512, ZSTD_btultra2},  /* level 20.*/
+    { 14, 15, 15,  9,  3,512, ZSTD_btultra2},  /* level 21.*/
+    { 14, 15, 15, 10,  3,999, ZSTD_btultra2},  /* level 22.*/
 },
 };
 
@@ -4022,8 +4272,8 @@ ZSTD_compressionParameters ZSTD_getCParams(int compressionLevel, unsigned long l
     if (compressionLevel > ZSTD_MAX_CLEVEL) row = ZSTD_MAX_CLEVEL;
     {   ZSTD_compressionParameters cp = ZSTD_defaultCParameters[tableID][row];
         if (compressionLevel < 0) cp.targetLength = (unsigned)(-compressionLevel);   /* acceleration factor */
-        return ZSTD_adjustCParams_internal(cp, srcSizeHint, dictSize); }
-
+        return ZSTD_adjustCParams_internal(cp, srcSizeHint, dictSize);
+    }
 }
 
 /*! ZSTD_getParams() :
diff --git a/lib/compress/zstd_compress_internal.h b/lib/compress/zstd_compress_internal.h
index 43f7c1486a9c..29bca59859ea 100644
--- a/lib/compress/zstd_compress_internal.h
+++ b/lib/compress/zstd_compress_internal.h
@@ -48,12 +48,6 @@ extern "C" {
 typedef enum { ZSTDcs_created=0, ZSTDcs_init, ZSTDcs_ongoing, ZSTDcs_ending } ZSTD_compressionStage_e;
 typedef enum { zcss_init=0, zcss_load, zcss_flush } ZSTD_cStreamStage;
 
-typedef enum {
-    ZSTD_dictDefaultAttach = 0,
-    ZSTD_dictForceAttach = 1,
-    ZSTD_dictForceCopy = -1,
-} ZSTD_dictAttachPref_e;
-
 typedef struct ZSTD_prefixDict_s {
     const void* dict;
     size_t dictSize;
@@ -96,10 +90,10 @@ typedef enum { zop_dynamic=0, zop_predef } ZSTD_OptPrice_e;
 
 typedef struct {
     /* All tables are allocated inside cctx->workspace by ZSTD_resetCCtx_internal() */
-    U32* litFreq;                /* table of literals statistics, of size 256 */
-    U32* litLengthFreq;          /* table of litLength statistics, of size (MaxLL+1) */
-    U32* matchLengthFreq;        /* table of matchLength statistics, of size (MaxML+1) */
-    U32* offCodeFreq;            /* table of offCode statistics, of size (MaxOff+1) */
+    unsigned* litFreq;           /* table of literals statistics, of size 256 */
+    unsigned* litLengthFreq;     /* table of litLength statistics, of size (MaxLL+1) */
+    unsigned* matchLengthFreq;   /* table of matchLength statistics, of size (MaxML+1) */
+    unsigned* offCodeFreq;       /* table of offCode statistics, of size (MaxOff+1) */
     ZSTD_match_t* matchTable;    /* list of found matches, of size ZSTD_OPT_NUM+1 */
     ZSTD_optimal_t* priceTable;  /* All positions tracked by optimal parser, of size ZSTD_OPT_NUM+1 */
 
@@ -139,7 +133,7 @@ struct ZSTD_matchState_t {
     U32* hashTable3;
     U32* chainTable;
     optState_t opt;         /* optimal parser state */
-    const ZSTD_matchState_t *dictMatchState;
+    const ZSTD_matchState_t * dictMatchState;
     ZSTD_compressionParameters cParams;
 };
 
@@ -167,7 +161,7 @@ typedef struct {
     U32 hashLog;            /* Log size of hashTable */
     U32 bucketSizeLog;      /* Log bucket size for collision resolution, at most 8 */
     U32 minMatchLength;     /* Minimum match length */
-    U32 hashEveryLog;       /* Log number of entries to skip */
+    U32 hashRateLog;       /* Log number of entries to skip */
     U32 windowLog;          /* Window log for the LDM */
 } ldmParams_t;
 
@@ -196,9 +190,10 @@ struct ZSTD_CCtx_params_s {
     ZSTD_dictAttachPref_e attachDictPref;
 
     /* Multithreading: used to pass parameters to mtctx */
-    unsigned nbWorkers;
-    unsigned jobSize;
-    unsigned overlapSizeLog;
+    int nbWorkers;
+    size_t jobSize;
+    int overlapLog;
+    int rsyncable;
 
     /* Long distance matching parameters */
     ldmParams_t ldmParams;
@@ -498,6 +493,64 @@ MEM_STATIC size_t ZSTD_hashPtr(const void* p, U32 hBits, U32 mls)
     }
 }
 
+/** ZSTD_ipow() :
+ * Return base^exponent.
+ */
+static U64 ZSTD_ipow(U64 base, U64 exponent)
+{
+    U64 power = 1;
+    while (exponent) {
+      if (exponent & 1) power *= base;
+      exponent >>= 1;
+      base *= base;
+    }
+    return power;
+}
+
+#define ZSTD_ROLL_HASH_CHAR_OFFSET 10
+
+/** ZSTD_rollingHash_append() :
+ * Add the buffer to the hash value.
+ */
+static U64 ZSTD_rollingHash_append(U64 hash, void const* buf, size_t size)
+{
+    BYTE const* istart = (BYTE const*)buf;
+    size_t pos;
+    for (pos = 0; pos < size; ++pos) {
+        hash *= prime8bytes;
+        hash += istart[pos] + ZSTD_ROLL_HASH_CHAR_OFFSET;
+    }
+    return hash;
+}
+
+/** ZSTD_rollingHash_compute() :
+ * Compute the rolling hash value of the buffer.
+ */
+MEM_STATIC U64 ZSTD_rollingHash_compute(void const* buf, size_t size)
+{
+    return ZSTD_rollingHash_append(0, buf, size);
+}
+
+/** ZSTD_rollingHash_primePower() :
+ * Compute the primePower to be passed to ZSTD_rollingHash_rotate() for a hash
+ * over a window of length bytes.
+ */
+MEM_STATIC U64 ZSTD_rollingHash_primePower(U32 length)
+{
+    return ZSTD_ipow(prime8bytes, length - 1);
+}
+
+/** ZSTD_rollingHash_rotate() :
+ * Rotate the rolling hash by one byte.
+ */
+MEM_STATIC U64 ZSTD_rollingHash_rotate(U64 hash, BYTE toRemove, BYTE toAdd, U64 primePower)
+{
+    hash -= (toRemove + ZSTD_ROLL_HASH_CHAR_OFFSET) * primePower;
+    hash *= prime8bytes;
+    hash += toAdd + ZSTD_ROLL_HASH_CHAR_OFFSET;
+    return hash;
+}
+
 /*-*************************************
 *  Round buffer management
 ***************************************/
@@ -626,20 +679,23 @@ MEM_STATIC U32 ZSTD_window_correctOverflow(ZSTD_window_t* window, U32 cycleLog,
  * dictMatchState mode, lowLimit and dictLimit are the same, and the dictionary
  * is below them. forceWindow and dictMatchState are therefore incompatible.
  */
-MEM_STATIC void ZSTD_window_enforceMaxDist(ZSTD_window_t* window,
-                                           void const* srcEnd, U32 maxDist,
-                                           U32* loadedDictEndPtr,
-                                           const ZSTD_matchState_t** dictMatchStatePtr)
+MEM_STATIC void
+ZSTD_window_enforceMaxDist(ZSTD_window_t* window,
+                           void const* srcEnd,
+                           U32 maxDist,
+                           U32* loadedDictEndPtr,
+                     const ZSTD_matchState_t** dictMatchStatePtr)
 {
-    U32 const current = (U32)((BYTE const*)srcEnd - window->base);
-    U32 loadedDictEnd = loadedDictEndPtr != NULL ? *loadedDictEndPtr : 0;
-    DEBUGLOG(5, "ZSTD_window_enforceMaxDist: current=%u, maxDist=%u", current, maxDist);
-    if (current > maxDist + loadedDictEnd) {
-        U32 const newLowLimit = current - maxDist;
+    U32 const blockEndIdx = (U32)((BYTE const*)srcEnd - window->base);
+    U32 loadedDictEnd = (loadedDictEndPtr != NULL) ? *loadedDictEndPtr : 0;
+    DEBUGLOG(5, "ZSTD_window_enforceMaxDist: blockEndIdx=%u, maxDist=%u",
+                (unsigned)blockEndIdx, (unsigned)maxDist);
+    if (blockEndIdx > maxDist + loadedDictEnd) {
+        U32 const newLowLimit = blockEndIdx - maxDist;
         if (window->lowLimit < newLowLimit) window->lowLimit = newLowLimit;
         if (window->dictLimit < window->lowLimit) {
             DEBUGLOG(5, "Update dictLimit to match lowLimit, from %u to %u",
-                        window->dictLimit, window->lowLimit);
+                        (unsigned)window->dictLimit, (unsigned)window->lowLimit);
             window->dictLimit = window->lowLimit;
         }
         if (loadedDictEndPtr)
@@ -690,20 +746,23 @@ MEM_STATIC U32 ZSTD_window_update(ZSTD_window_t* window,
 
 
 /* debug functions */
+#if (DEBUGLEVEL>=2)
 
 MEM_STATIC double ZSTD_fWeight(U32 rawStat)
 {
     U32 const fp_accuracy = 8;
     U32 const fp_multiplier = (1 << fp_accuracy);
-    U32 const stat = rawStat + 1;
-    U32 const hb = ZSTD_highbit32(stat);
+    U32 const newStat = rawStat + 1;
+    U32 const hb = ZSTD_highbit32(newStat);
     U32 const BWeight = hb * fp_multiplier;
-    U32 const FWeight = (stat << fp_accuracy) >> hb;
+    U32 const FWeight = (newStat << fp_accuracy) >> hb;
     U32 const weight = BWeight + FWeight;
     assert(hb + fp_accuracy < 31);
     return (double)weight / fp_multiplier;
 }
 
+/* display a table content,
+ * listing each element, its frequency, and its predicted bit cost */
 MEM_STATIC void ZSTD_debugTable(const U32* table, U32 max)
 {
     unsigned u, sum;
@@ -715,6 +774,9 @@ MEM_STATIC void ZSTD_debugTable(const U32* table, U32 max)
     }
 }
 
+#endif
+
+
 #if defined (__cplusplus)
 }
 #endif
diff --git a/lib/compress/zstd_double_fast.c b/lib/compress/zstd_double_fast.c
index 7b9e18e7ed9d..47faf6d641db 100644
--- a/lib/compress/zstd_double_fast.c
+++ b/lib/compress/zstd_double_fast.c
@@ -18,7 +18,7 @@ void ZSTD_fillDoubleHashTable(ZSTD_matchState_t* ms,
     const ZSTD_compressionParameters* const cParams = &ms->cParams;
     U32* const hashLarge = ms->hashTable;
     U32  const hBitsL = cParams->hashLog;
-    U32  const mls = cParams->searchLength;
+    U32  const mls = cParams->minMatch;
     U32* const hashSmall = ms->chainTable;
     U32  const hBitsS = cParams->chainLog;
     const BYTE* const base = ms->window.base;
@@ -309,7 +309,7 @@ size_t ZSTD_compressBlock_doubleFast(
         ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
         void const* src, size_t srcSize)
 {
-    const U32 mls = ms->cParams.searchLength;
+    const U32 mls = ms->cParams.minMatch;
     switch(mls)
     {
     default: /* includes case 3 */
@@ -329,7 +329,7 @@ size_t ZSTD_compressBlock_doubleFast_dictMatchState(
         ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
         void const* src, size_t srcSize)
 {
-    const U32 mls = ms->cParams.searchLength;
+    const U32 mls = ms->cParams.minMatch;
     switch(mls)
     {
     default: /* includes case 3 */
@@ -483,7 +483,7 @@ size_t ZSTD_compressBlock_doubleFast_extDict(
         ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
         void const* src, size_t srcSize)
 {
-    U32 const mls = ms->cParams.searchLength;
+    U32 const mls = ms->cParams.minMatch;
     switch(mls)
     {
     default: /* includes case 3 */
diff --git a/lib/compress/zstd_fast.c b/lib/compress/zstd_fast.c
index 247746517cd5..40ba0f73e65a 100644
--- a/lib/compress/zstd_fast.c
+++ b/lib/compress/zstd_fast.c
@@ -18,7 +18,7 @@ void ZSTD_fillHashTable(ZSTD_matchState_t* ms,
     const ZSTD_compressionParameters* const cParams = &ms->cParams;
     U32* const hashTable = ms->hashTable;
     U32  const hBits = cParams->hashLog;
-    U32  const mls = cParams->searchLength;
+    U32  const mls = cParams->minMatch;
     const BYTE* const base = ms->window.base;
     const BYTE* ip = base + ms->nextToUpdate;
     const BYTE* const iend = ((const BYTE*)end) - HASH_READ_SIZE;
@@ -27,18 +27,18 @@ void ZSTD_fillHashTable(ZSTD_matchState_t* ms,
     /* Always insert every fastHashFillStep position into the hash table.
      * Insert the other positions if their hash entry is empty.
      */
-    for (; ip + fastHashFillStep - 1 <= iend; ip += fastHashFillStep) {
+    for ( ; ip + fastHashFillStep < iend + 2; ip += fastHashFillStep) {
         U32 const current = (U32)(ip - base);
-        U32 i;
-        for (i = 0; i < fastHashFillStep; ++i) {
-            size_t const hash = ZSTD_hashPtr(ip + i, hBits, mls);
-            if (i == 0 || hashTable[hash] == 0)
-                hashTable[hash] = current + i;
-            /* Only load extra positions for ZSTD_dtlm_full */
-            if (dtlm == ZSTD_dtlm_fast)
-                break;
-        }
-    }
+        size_t const hash0 = ZSTD_hashPtr(ip, hBits, mls);
+        hashTable[hash0] = current;
+        if (dtlm == ZSTD_dtlm_fast) continue;
+        /* Only load extra positions for ZSTD_dtlm_full */
+        {   U32 p;
+            for (p = 1; p < fastHashFillStep; ++p) {
+                size_t const hash = ZSTD_hashPtr(ip + p, hBits, mls);
+                if (hashTable[hash] == 0) {  /* not yet filled */
+                    hashTable[hash] = current + p;
+    }   }   }   }
 }
 
 FORCE_INLINE_TEMPLATE
@@ -235,7 +235,7 @@ size_t ZSTD_compressBlock_fast(
         void const* src, size_t srcSize)
 {
     ZSTD_compressionParameters const* cParams = &ms->cParams;
-    U32 const mls = cParams->searchLength;
+    U32 const mls = cParams->minMatch;
     assert(ms->dictMatchState == NULL);
     switch(mls)
     {
@@ -256,7 +256,7 @@ size_t ZSTD_compressBlock_fast_dictMatchState(
         void const* src, size_t srcSize)
 {
     ZSTD_compressionParameters const* cParams = &ms->cParams;
-    U32 const mls = cParams->searchLength;
+    U32 const mls = cParams->minMatch;
     assert(ms->dictMatchState != NULL);
     switch(mls)
     {
@@ -375,7 +375,7 @@ size_t ZSTD_compressBlock_fast_extDict(
         void const* src, size_t srcSize)
 {
     ZSTD_compressionParameters const* cParams = &ms->cParams;
-    U32 const mls = cParams->searchLength;
+    U32 const mls = cParams->minMatch;
     switch(mls)
     {
     default: /* includes case 3 */
diff --git a/lib/compress/zstd_lazy.c b/lib/compress/zstd_lazy.c
index af615e07763d..53f998a4374b 100644
--- a/lib/compress/zstd_lazy.c
+++ b/lib/compress/zstd_lazy.c
@@ -63,12 +63,13 @@ ZSTD_updateDUBT(ZSTD_matchState_t* ms,
 static void
 ZSTD_insertDUBT1(ZSTD_matchState_t* ms,
                  U32 current, const BYTE* inputEnd,
-                 U32 nbCompares, U32 btLow, const ZSTD_dictMode_e dictMode)
+                 U32 nbCompares, U32 btLow,
+                 const ZSTD_dictMode_e dictMode)
 {
     const ZSTD_compressionParameters* const cParams = &ms->cParams;
-    U32*   const bt = ms->chainTable;
-    U32    const btLog  = cParams->chainLog - 1;
-    U32    const btMask = (1 << btLog) - 1;
+    U32* const bt = ms->chainTable;
+    U32  const btLog  = cParams->chainLog - 1;
+    U32  const btMask = (1 << btLog) - 1;
     size_t commonLengthSmaller=0, commonLengthLarger=0;
     const BYTE* const base = ms->window.base;
     const BYTE* const dictBase = ms->window.dictBase;
@@ -80,7 +81,7 @@ ZSTD_insertDUBT1(ZSTD_matchState_t* ms,
     const BYTE* match;
     U32* smallerPtr = bt + 2*(current&btMask);
     U32* largerPtr  = smallerPtr + 1;
-    U32 matchIndex = *smallerPtr;
+    U32 matchIndex = *smallerPtr;   /* this candidate is unsorted : next sorted candidate is reached through *smallerPtr, while *largerPtr contains previous unsorted candidate (which is already saved and can be overwritten) */
     U32 dummy32;   /* to be nullified at the end */
     U32 const windowLow = ms->window.lowLimit;
 
@@ -93,6 +94,9 @@ ZSTD_insertDUBT1(ZSTD_matchState_t* ms,
         U32* const nextPtr = bt + 2*(matchIndex & btMask);
         size_t matchLength = MIN(commonLengthSmaller, commonLengthLarger);   /* guaranteed minimum nb of common bytes */
         assert(matchIndex < current);
+        /* note : all candidates are now supposed sorted,
+         * but it's still possible to have nextPtr[1] == ZSTD_DUBT_UNSORTED_MARK
+         * when a real index has the same value as ZSTD_DUBT_UNSORTED_MARK */
 
         if ( (dictMode != ZSTD_extDict)
           || (matchIndex+matchLength >= dictLimit)  /* both in current segment*/
@@ -108,7 +112,7 @@ ZSTD_insertDUBT1(ZSTD_matchState_t* ms,
             match = dictBase + matchIndex;
             matchLength += ZSTD_count_2segments(ip+matchLength, match+matchLength, iend, dictEnd, prefixStart);
             if (matchIndex+matchLength >= dictLimit)
-                match = base + matchIndex;   /* to prepare for next usage of match[matchLength] */
+                match = base + matchIndex;   /* preparation for next read of match[matchLength] */
         }
 
         DEBUGLOG(8, "ZSTD_insertDUBT1: comparing %u with %u : found %u common bytes ",
@@ -258,7 +262,7 @@ ZSTD_DUBT_findBestMatch(ZSTD_matchState_t* ms,
          && (nbCandidates > 1) ) {
         DEBUGLOG(8, "ZSTD_DUBT_findBestMatch: candidate %u is unsorted",
                     matchIndex);
-        *unsortedMark = previousCandidate;
+        *unsortedMark = previousCandidate;  /* the unsortedMark becomes a reversed chain, to move up back to original position */
         previousCandidate = matchIndex;
         matchIndex = *nextCandidate;
         nextCandidate = bt + 2*(matchIndex&btMask);
@@ -266,11 +270,13 @@ ZSTD_DUBT_findBestMatch(ZSTD_matchState_t* ms,
         nbCandidates --;
     }
 
+    /* nullify last candidate if it's still unsorted
+     * simplification, detrimental to compression ratio, beneficial for speed */
     if ( (matchIndex > unsortLimit)
       && (*unsortedMark==ZSTD_DUBT_UNSORTED_MARK) ) {
         DEBUGLOG(7, "ZSTD_DUBT_findBestMatch: nullify last unsorted candidate %u",
                     matchIndex);
-        *nextCandidate = *unsortedMark = 0;   /* nullify next candidate if it's still unsorted (note : simplification, detrimental to compression ratio, beneficial for speed) */
+        *nextCandidate = *unsortedMark = 0;
     }
 
     /* batch sort stacked candidates */
@@ -285,14 +291,14 @@ ZSTD_DUBT_findBestMatch(ZSTD_matchState_t* ms,
     }
 
     /* find longest match */
-    {   size_t commonLengthSmaller=0, commonLengthLarger=0;
+    {   size_t commonLengthSmaller = 0, commonLengthLarger = 0;
         const BYTE* const dictBase = ms->window.dictBase;
         const U32 dictLimit = ms->window.dictLimit;
         const BYTE* const dictEnd = dictBase + dictLimit;
         const BYTE* const prefixStart = base + dictLimit;
         U32* smallerPtr = bt + 2*(current&btMask);
         U32* largerPtr  = bt + 2*(current&btMask) + 1;
-        U32 matchEndIdx = current+8+1;
+        U32 matchEndIdx = current + 8 + 1;
         U32 dummy32;   /* to be nullified at the end */
         size_t bestLength = 0;
 
@@ -386,7 +392,7 @@ ZSTD_BtFindBestMatch_selectMLS (  ZSTD_matchState_t* ms,
                             const BYTE* ip, const BYTE* const iLimit,
                                   size_t* offsetPtr)
 {
-    switch(ms->cParams.searchLength)
+    switch(ms->cParams.minMatch)
     {
     default : /* includes case 3 */
     case 4 : return ZSTD_BtFindBestMatch(ms, ip, iLimit, offsetPtr, 4, ZSTD_noDict);
@@ -402,7 +408,7 @@ static size_t ZSTD_BtFindBestMatch_dictMatchState_selectMLS (
                         const BYTE* ip, const BYTE* const iLimit,
                         size_t* offsetPtr)
 {
-    switch(ms->cParams.searchLength)
+    switch(ms->cParams.minMatch)
     {
     default : /* includes case 3 */
     case 4 : return ZSTD_BtFindBestMatch(ms, ip, iLimit, offsetPtr, 4, ZSTD_dictMatchState);
@@ -418,7 +424,7 @@ static size_t ZSTD_BtFindBestMatch_extDict_selectMLS (
                         const BYTE* ip, const BYTE* const iLimit,
                         size_t* offsetPtr)
 {
-    switch(ms->cParams.searchLength)
+    switch(ms->cParams.minMatch)
     {
     default : /* includes case 3 */
     case 4 : return ZSTD_BtFindBestMatch(ms, ip, iLimit, offsetPtr, 4, ZSTD_extDict);
@@ -433,7 +439,7 @@ static size_t ZSTD_BtFindBestMatch_extDict_selectMLS (
 /* *********************************
 *  Hash Chain
 ***********************************/
-#define NEXT_IN_CHAIN(d, mask)   chainTable[(d) & mask]
+#define NEXT_IN_CHAIN(d, mask)   chainTable[(d) & (mask)]
 
 /* Update chains up to ip (excluded)
    Assumption : always within prefix (i.e. not within extDict) */
@@ -463,7 +469,7 @@ static U32 ZSTD_insertAndFindFirstIndex_internal(
 
 U32 ZSTD_insertAndFindFirstIndex(ZSTD_matchState_t* ms, const BYTE* ip) {
     const ZSTD_compressionParameters* const cParams = &ms->cParams;
-    return ZSTD_insertAndFindFirstIndex_internal(ms, cParams, ip, ms->cParams.searchLength);
+    return ZSTD_insertAndFindFirstIndex_internal(ms, cParams, ip, ms->cParams.minMatch);
 }
 
 
@@ -497,6 +503,7 @@ size_t ZSTD_HcFindBestMatch_generic (
         size_t currentMl=0;
         if ((dictMode != ZSTD_extDict) || matchIndex >= dictLimit) {
             const BYTE* const match = base + matchIndex;
+            assert(matchIndex >= dictLimit);   /* ensures this is true if dictMode != ZSTD_extDict */
             if (match[ml] == ip[ml])   /* potentially better */
                 currentMl = ZSTD_count(ip, match, iLimit);
         } else {
@@ -559,7 +566,7 @@ FORCE_INLINE_TEMPLATE size_t ZSTD_HcFindBestMatch_selectMLS (
                         const BYTE* ip, const BYTE* const iLimit,
                         size_t* offsetPtr)
 {
-    switch(ms->cParams.searchLength)
+    switch(ms->cParams.minMatch)
     {
     default : /* includes case 3 */
     case 4 : return ZSTD_HcFindBestMatch_generic(ms, ip, iLimit, offsetPtr, 4, ZSTD_noDict);
@@ -575,7 +582,7 @@ static size_t ZSTD_HcFindBestMatch_dictMatchState_selectMLS (
                         const BYTE* ip, const BYTE* const iLimit,
                         size_t* offsetPtr)
 {
-    switch(ms->cParams.searchLength)
+    switch(ms->cParams.minMatch)
     {
     default : /* includes case 3 */
     case 4 : return ZSTD_HcFindBestMatch_generic(ms, ip, iLimit, offsetPtr, 4, ZSTD_dictMatchState);
@@ -591,7 +598,7 @@ FORCE_INLINE_TEMPLATE size_t ZSTD_HcFindBestMatch_extDict_selectMLS (
                         const BYTE* ip, const BYTE* const iLimit,
                         size_t* offsetPtr)
 {
-    switch(ms->cParams.searchLength)
+    switch(ms->cParams.minMatch)
     {
     default : /* includes case 3 */
     case 4 : return ZSTD_HcFindBestMatch_generic(ms, ip, iLimit, offsetPtr, 4, ZSTD_extDict);
diff --git a/lib/compress/zstd_ldm.c b/lib/compress/zstd_ldm.c
index 6238ddecf24f..58eb2ffe4db3 100644
--- a/lib/compress/zstd_ldm.c
+++ b/lib/compress/zstd_ldm.c
@@ -37,8 +37,8 @@ void ZSTD_ldm_adjustParameters(ldmParams_t* params,
         params->hashLog = MAX(ZSTD_HASHLOG_MIN, params->windowLog - LDM_HASH_RLOG);
         assert(params->hashLog <= ZSTD_HASHLOG_MAX);
     }
-    if (params->hashEveryLog == 0) {
-        params->hashEveryLog = params->windowLog < params->hashLog
+    if (params->hashRateLog == 0) {
+        params->hashRateLog = params->windowLog < params->hashLog
                                    ? 0
                                    : params->windowLog - params->hashLog;
     }
@@ -119,20 +119,20 @@ static void ZSTD_ldm_insertEntry(ldmState_t* ldmState,
  *
  *  Gets the small hash, checksum, and tag from the rollingHash.
  *
- *  If the tag matches (1 << ldmParams.hashEveryLog)-1, then
+ *  If the tag matches (1 << ldmParams.hashRateLog)-1, then
  *  creates an ldmEntry from the offset, and inserts it into the hash table.
  *
  *  hBits is the length of the small hash, which is the most significant hBits
  *  of rollingHash. The checksum is the next 32 most significant bits, followed
- *  by ldmParams.hashEveryLog bits that make up the tag. */
+ *  by ldmParams.hashRateLog bits that make up the tag. */
 static void ZSTD_ldm_makeEntryAndInsertByTag(ldmState_t* ldmState,
                                              U64 const rollingHash,
                                              U32 const hBits,
                                              U32 const offset,
                                              ldmParams_t const ldmParams)
 {
-    U32 const tag = ZSTD_ldm_getTag(rollingHash, hBits, ldmParams.hashEveryLog);
-    U32 const tagMask = ((U32)1 << ldmParams.hashEveryLog) - 1;
+    U32 const tag = ZSTD_ldm_getTag(rollingHash, hBits, ldmParams.hashRateLog);
+    U32 const tagMask = ((U32)1 << ldmParams.hashRateLog) - 1;
     if (tag == tagMask) {
         U32 const hash = ZSTD_ldm_getSmallHash(rollingHash, hBits);
         U32 const checksum = ZSTD_ldm_getChecksum(rollingHash, hBits);
@@ -143,56 +143,6 @@ static void ZSTD_ldm_makeEntryAndInsertByTag(ldmState_t* ldmState,
     }
 }
 
-/** ZSTD_ldm_getRollingHash() :
- *  Get a 64-bit hash using the first len bytes from buf.
- *
- *  Giving bytes s = s_1, s_2, ... s_k, the hash is defined to be
- *  H(s) = s_1*(a^(k-1)) + s_2*(a^(k-2)) + ... + s_k*(a^0)
- *
- *  where the constant a is defined to be prime8bytes.
- *
- *  The implementation adds an offset to each byte, so
- *  H(s) = (s_1 + HASH_CHAR_OFFSET)*(a^(k-1)) + ... */
-static U64 ZSTD_ldm_getRollingHash(const BYTE* buf, U32 len)
-{
-    U64 ret = 0;
-    U32 i;
-    for (i = 0; i < len; i++) {
-        ret *= prime8bytes;
-        ret += buf[i] + LDM_HASH_CHAR_OFFSET;
-    }
-    return ret;
-}
-
-/** ZSTD_ldm_ipow() :
- *  Return base^exp. */
-static U64 ZSTD_ldm_ipow(U64 base, U64 exp)
-{
-    U64 ret = 1;
-    while (exp) {
-        if (exp & 1) { ret *= base; }
-        exp >>= 1;
-        base *= base;
-    }
-    return ret;
-}
-
-U64 ZSTD_ldm_getHashPower(U32 minMatchLength) {
-    DEBUGLOG(4, "ZSTD_ldm_getHashPower: mml=%u", minMatchLength);
-    assert(minMatchLength >= ZSTD_LDM_MINMATCH_MIN);
-    return ZSTD_ldm_ipow(prime8bytes, minMatchLength - 1);
-}
-
-/** ZSTD_ldm_updateHash() :
- *  Updates hash by removing toRemove and adding toAdd. */
-static U64 ZSTD_ldm_updateHash(U64 hash, BYTE toRemove, BYTE toAdd, U64 hashPower)
-{
-    hash -= ((toRemove + LDM_HASH_CHAR_OFFSET) * hashPower);
-    hash *= prime8bytes;
-    hash += toAdd + LDM_HASH_CHAR_OFFSET;
-    return hash;
-}
-
 /** ZSTD_ldm_countBackwardsMatch() :
  *  Returns the number of bytes that match backwards before pIn and pMatch.
  *
@@ -238,6 +188,7 @@ static size_t ZSTD_ldm_fillFastTables(ZSTD_matchState_t* ms,
     case ZSTD_btlazy2:
     case ZSTD_btopt:
     case ZSTD_btultra:
+    case ZSTD_btultra2:
         break;
     default:
         assert(0);  /* not possible : not a valid strategy id */
@@ -261,9 +212,9 @@ static U64 ZSTD_ldm_fillLdmHashTable(ldmState_t* state,
     const BYTE* cur = lastHashed + 1;
 
     while (cur < iend) {
-        rollingHash = ZSTD_ldm_updateHash(rollingHash, cur[-1],
-                                          cur[ldmParams.minMatchLength-1],
-                                          state->hashPower);
+        rollingHash = ZSTD_rollingHash_rotate(rollingHash, cur[-1],
+                                              cur[ldmParams.minMatchLength-1],
+                                              state->hashPower);
         ZSTD_ldm_makeEntryAndInsertByTag(state,
                                          rollingHash, hBits,
                                          (U32)(cur - base), ldmParams);
@@ -297,8 +248,8 @@ static size_t ZSTD_ldm_generateSequences_internal(
     U64 const hashPower = ldmState->hashPower;
     U32 const hBits = params->hashLog - params->bucketSizeLog;
     U32 const ldmBucketSize = 1U << params->bucketSizeLog;
-    U32 const hashEveryLog = params->hashEveryLog;
-    U32 const ldmTagMask = (1U << params->hashEveryLog) - 1;
+    U32 const hashRateLog = params->hashRateLog;
+    U32 const ldmTagMask = (1U << params->hashRateLog) - 1;
     /* Prefix and extDict parameters */
     U32 const dictLimit = ldmState->window.dictLimit;
     U32 const lowestIndex = extDict ? ldmState->window.lowLimit : dictLimit;
@@ -324,16 +275,16 @@ static size_t ZSTD_ldm_generateSequences_internal(
         size_t forwardMatchLength = 0, backwardMatchLength = 0;
         ldmEntry_t* bestEntry = NULL;
         if (ip != istart) {
-            rollingHash = ZSTD_ldm_updateHash(rollingHash, lastHashed[0],
-                                              lastHashed[minMatchLength],
-                                              hashPower);
+            rollingHash = ZSTD_rollingHash_rotate(rollingHash, lastHashed[0],
+                                                  lastHashed[minMatchLength],
+                                                  hashPower);
         } else {
-            rollingHash = ZSTD_ldm_getRollingHash(ip, minMatchLength);
+            rollingHash = ZSTD_rollingHash_compute(ip, minMatchLength);
         }
         lastHashed = ip;
 
         /* Do not insert and do not look for a match */
-        if (ZSTD_ldm_getTag(rollingHash, hBits, hashEveryLog) != ldmTagMask) {
+        if (ZSTD_ldm_getTag(rollingHash, hBits, hashRateLog) != ldmTagMask) {
            ip++;
            continue;
         }
@@ -593,7 +544,7 @@ size_t ZSTD_ldm_blockCompress(rawSeqStore_t* rawSeqStore,
     void const* src, size_t srcSize)
 {
     const ZSTD_compressionParameters* const cParams = &ms->cParams;
-    unsigned const minMatch = cParams->searchLength;
+    unsigned const minMatch = cParams->minMatch;
     ZSTD_blockCompressor const blockCompressor =
         ZSTD_selectBlockCompressor(cParams->strategy, ZSTD_matchState_dictMode(ms));
     /* Input bounds */
diff --git a/lib/compress/zstd_ldm.h b/lib/compress/zstd_ldm.h
index 21fba4d591a4..a47846128b20 100644
--- a/lib/compress/zstd_ldm.h
+++ b/lib/compress/zstd_ldm.h
@@ -21,7 +21,7 @@ extern "C" {
 *  Long distance matching
 ***************************************/
 
-#define ZSTD_LDM_DEFAULT_WINDOW_LOG ZSTD_WINDOWLOG_DEFAULTMAX
+#define ZSTD_LDM_DEFAULT_WINDOW_LOG ZSTD_WINDOWLOG_LIMIT_DEFAULT
 
 /**
  * ZSTD_ldm_generateSequences():
@@ -86,12 +86,8 @@ size_t ZSTD_ldm_getTableSize(ldmParams_t params);
  */
 size_t ZSTD_ldm_getMaxNbSeq(ldmParams_t params, size_t maxChunkSize);
 
-/** ZSTD_ldm_getTableSize() :
- *  Return prime8bytes^(minMatchLength-1) */
-U64 ZSTD_ldm_getHashPower(U32 minMatchLength);
-
 /** ZSTD_ldm_adjustParameters() :
- *  If the params->hashEveryLog is not set, set it to its default value based on
+ *  If the params->hashRateLog is not set, set it to its default value based on
  *  windowLog and params->hashLog.
  *
  *  Ensures that params->bucketSizeLog is <= params->hashLog (setting it to
diff --git a/lib/compress/zstd_opt.c b/lib/compress/zstd_opt.c
index 8af69a91d46e..44de6e97fd20 100644
--- a/lib/compress/zstd_opt.c
+++ b/lib/compress/zstd_opt.c
@@ -17,6 +17,8 @@
 #define ZSTD_FREQ_DIV       4   /* log factor when using previous stats to init next stats */
 #define ZSTD_MAX_PRICE     (1<<30)
 
+#define ZSTD_PREDEF_THRESHOLD 1024   /* if srcSize < ZSTD_PREDEF_THRESHOLD, symbols' cost is assumed static, directly determined by pre-defined distributions */
+
 
 /*-*************************************
 *  Price functions for optimal parser
@@ -52,11 +54,15 @@ MEM_STATIC U32 ZSTD_fracWeight(U32 rawStat)
     return weight;
 }
 
-/* debugging function, @return price in bytes */
+#if (DEBUGLEVEL>=2)
+/* debugging function,
+ * @return price in bytes as fractional value
+ * for debug messages only */
 MEM_STATIC double ZSTD_fCost(U32 price)
 {
     return (double)price / (BITCOST_MULTIPLIER*8);
 }
+#endif
 
 static void ZSTD_setBasePrices(optState_t* optPtr, int optLevel)
 {
@@ -67,29 +73,44 @@ static void ZSTD_setBasePrices(optState_t* optPtr, int optLevel)
 }
 
 
-static U32 ZSTD_downscaleStat(U32* table, U32 lastEltIndex, int malus)
+/* ZSTD_downscaleStat() :
+ * reduce all elements in table by a factor 2^(ZSTD_FREQ_DIV+malus)
+ * return the resulting sum of elements */
+static U32 ZSTD_downscaleStat(unsigned* table, U32 lastEltIndex, int malus)
 {
     U32 s, sum=0;
+    DEBUGLOG(5, "ZSTD_downscaleStat (nbElts=%u)", (unsigned)lastEltIndex+1);
     assert(ZSTD_FREQ_DIV+malus > 0 && ZSTD_FREQ_DIV+malus < 31);
-    for (s=0; s<=lastEltIndex; s++) {
+    for (s=0; s> (ZSTD_FREQ_DIV+malus));
         sum += table[s];
     }
     return sum;
 }
 
-static void ZSTD_rescaleFreqs(optState_t* const optPtr,
-                              const BYTE* const src, size_t const srcSize,
-                              int optLevel)
+/* ZSTD_rescaleFreqs() :
+ * if first block (detected by optPtr->litLengthSum == 0) : init statistics
+ *    take hints from dictionary if there is one
+ *    or init from zero, using src for literals stats, or flat 1 for match symbols
+ * otherwise downscale existing stats, to be used as seed for next block.
+ */
+static void
+ZSTD_rescaleFreqs(optState_t* const optPtr,
+            const BYTE* const src, size_t const srcSize,
+                  int const optLevel)
 {
+    DEBUGLOG(5, "ZSTD_rescaleFreqs (srcSize=%u)", (unsigned)srcSize);
     optPtr->priceType = zop_dynamic;
 
     if (optPtr->litLengthSum == 0) {  /* first block : init */
-        if (srcSize <= 1024)   /* heuristic */
+        if (srcSize <= ZSTD_PREDEF_THRESHOLD) {  /* heuristic */
+            DEBUGLOG(5, "(srcSize <= ZSTD_PREDEF_THRESHOLD) => zop_predef");
             optPtr->priceType = zop_predef;
+        }
 
         assert(optPtr->symbolCosts != NULL);
-        if (optPtr->symbolCosts->huf.repeatMode == HUF_repeat_valid) { /* huffman table presumed generated by dictionary */
+        if (optPtr->symbolCosts->huf.repeatMode == HUF_repeat_valid) {
+            /* huffman table presumed generated by dictionary */
             optPtr->priceType = zop_dynamic;
 
             assert(optPtr->litFreq != NULL);
@@ -208,7 +229,9 @@ static U32 ZSTD_litLengthPrice(U32 const litLength, const optState_t* const optP
 
     /* dynamic statistics */
     {   U32 const llCode = ZSTD_LLcode(litLength);
-        return (LL_bits[llCode] * BITCOST_MULTIPLIER) + (optPtr->litLengthSumBasePrice - WEIGHT(optPtr->litLengthFreq[llCode], optLevel));
+        return (LL_bits[llCode] * BITCOST_MULTIPLIER)
+             + optPtr->litLengthSumBasePrice
+             - WEIGHT(optPtr->litLengthFreq[llCode], optLevel);
     }
 }
 
@@ -253,7 +276,7 @@ static int ZSTD_literalsContribution(const BYTE* const literals, U32 const litLe
 FORCE_INLINE_TEMPLATE U32
 ZSTD_getMatchPrice(U32 const offset,
                    U32 const matchLength,
-                   const optState_t* const optPtr,
+             const optState_t* const optPtr,
                    int const optLevel)
 {
     U32 price;
@@ -385,7 +408,6 @@ static U32 ZSTD_insertBt1(
     U32* largerPtr  = smallerPtr + 1;
     U32 dummy32;   /* to be nullified at the end */
     U32 const windowLow = ms->window.lowLimit;
-    U32 const matchLow = windowLow ? windowLow : 1;
     U32 matchEndIdx = current+8+1;
     size_t bestLength = 8;
     U32 nbCompares = 1U << cParams->searchLog;
@@ -401,7 +423,8 @@ static U32 ZSTD_insertBt1(
     assert(ip <= iend-8);   /* required for h calculation */
     hashTable[h] = current;   /* Update Hash Table */
 
-    while (nbCompares-- && (matchIndex >= matchLow)) {
+    assert(windowLow > 0);
+    while (nbCompares-- && (matchIndex >= windowLow)) {
         U32* const nextPtr = bt + 2*(matchIndex & btMask);
         size_t matchLength = MIN(commonLengthSmaller, commonLengthLarger);   /* guaranteed minimum nb of common bytes */
         assert(matchIndex < current);
@@ -479,7 +502,7 @@ void ZSTD_updateTree_internal(
     const BYTE* const base = ms->window.base;
     U32 const target = (U32)(ip - base);
     U32 idx = ms->nextToUpdate;
-    DEBUGLOG(5, "ZSTD_updateTree_internal, from %u to %u  (dictMode:%u)",
+    DEBUGLOG(6, "ZSTD_updateTree_internal, from %u to %u  (dictMode:%u)",
                 idx, target, dictMode);
 
     while(idx < target)
@@ -488,15 +511,18 @@ void ZSTD_updateTree_internal(
 }
 
 void ZSTD_updateTree(ZSTD_matchState_t* ms, const BYTE* ip, const BYTE* iend) {
-    ZSTD_updateTree_internal(ms, ip, iend, ms->cParams.searchLength, ZSTD_noDict);
+    ZSTD_updateTree_internal(ms, ip, iend, ms->cParams.minMatch, ZSTD_noDict);
 }
 
 FORCE_INLINE_TEMPLATE
 U32 ZSTD_insertBtAndGetAllMatches (
                     ZSTD_matchState_t* ms,
                     const BYTE* const ip, const BYTE* const iLimit, const ZSTD_dictMode_e dictMode,
-                    U32 rep[ZSTD_REP_NUM], U32 const ll0,
-                    ZSTD_match_t* matches, const U32 lengthToBeat, U32 const mls /* template */)
+                    U32 rep[ZSTD_REP_NUM],
+                    U32 const ll0,   /* tells if associated literal length is 0 or not. This value must be 0 or 1 */
+                    ZSTD_match_t* matches,
+                    const U32 lengthToBeat,
+                    U32 const mls /* template */)
 {
     const ZSTD_compressionParameters* const cParams = &ms->cParams;
     U32 const sufficient_len = MIN(cParams->targetLength, ZSTD_OPT_NUM -1);
@@ -542,6 +568,7 @@ U32 ZSTD_insertBtAndGetAllMatches (
     DEBUGLOG(8, "ZSTD_insertBtAndGetAllMatches: current=%u", current);
 
     /* check repCode */
+    assert(ll0 <= 1);   /* necessarily 1 or 0 */
     {   U32 const lastR = ZSTD_REP_NUM + ll0;
         U32 repCode;
         for (repCode = ll0; repCode < lastR; repCode++) {
@@ -724,7 +751,7 @@ FORCE_INLINE_TEMPLATE U32 ZSTD_BtGetAllMatches (
                         ZSTD_match_t* matches, U32 const lengthToBeat)
 {
     const ZSTD_compressionParameters* const cParams = &ms->cParams;
-    U32 const matchLengthSearch = cParams->searchLength;
+    U32 const matchLengthSearch = cParams->minMatch;
     DEBUGLOG(8, "ZSTD_BtGetAllMatches");
     if (ip < ms->window.base + ms->nextToUpdate) return 0;   /* skipped area */
     ZSTD_updateTree_internal(ms, ip, iHighLimit, matchLengthSearch, dictMode);
@@ -774,12 +801,30 @@ static U32 ZSTD_totalLen(ZSTD_optimal_t sol)
     return sol.litlen + sol.mlen;
 }
 
+#if 0 /* debug */
+
+static void
+listStats(const U32* table, int lastEltID)
+{
+    int const nbElts = lastEltID + 1;
+    int enb;
+    for (enb=0; enb < nbElts; enb++) {
+        (void)table;
+        //RAWLOG(2, "%3i:%3i,  ", enb, table[enb]);
+        RAWLOG(2, "%4i,", table[enb]);
+    }
+    RAWLOG(2, " \n");
+}
+
+#endif
+
 FORCE_INLINE_TEMPLATE size_t
 ZSTD_compressBlock_opt_generic(ZSTD_matchState_t* ms,
                                seqStore_t* seqStore,
                                U32 rep[ZSTD_REP_NUM],
-                               const void* src, size_t srcSize,
-                               const int optLevel, const ZSTD_dictMode_e dictMode)
+                         const void* src, size_t srcSize,
+                         const int optLevel,
+                         const ZSTD_dictMode_e dictMode)
 {
     optState_t* const optStatePtr = &ms->opt;
     const BYTE* const istart = (const BYTE*)src;
@@ -792,14 +837,15 @@ ZSTD_compressBlock_opt_generic(ZSTD_matchState_t* ms,
     const ZSTD_compressionParameters* const cParams = &ms->cParams;
 
     U32 const sufficient_len = MIN(cParams->targetLength, ZSTD_OPT_NUM -1);
-    U32 const minMatch = (cParams->searchLength == 3) ? 3 : 4;
+    U32 const minMatch = (cParams->minMatch == 3) ? 3 : 4;
 
     ZSTD_optimal_t* const opt = optStatePtr->priceTable;
     ZSTD_match_t* const matches = optStatePtr->matchTable;
     ZSTD_optimal_t lastSequence;
 
     /* init */
-    DEBUGLOG(5, "ZSTD_compressBlock_opt_generic");
+    DEBUGLOG(5, "ZSTD_compressBlock_opt_generic: current=%u, prefix=%u, nextToUpdate=%u",
+                (U32)(ip - base), ms->window.dictLimit, ms->nextToUpdate);
     assert(optLevel <= 2);
     ms->nextToUpdate3 = ms->nextToUpdate;
     ZSTD_rescaleFreqs(optStatePtr, (const BYTE*)src, srcSize, optLevel);
@@ -999,7 +1045,7 @@ ZSTD_compressBlock_opt_generic(ZSTD_matchState_t* ms,
                     U32 const offCode = opt[storePos].off;
                     U32 const advance = llen + mlen;
                     DEBUGLOG(6, "considering seq starting at %zi, llen=%u, mlen=%u",
-                                anchor - istart, llen, mlen);
+                                anchor - istart, (unsigned)llen, (unsigned)mlen);
 
                     if (mlen==0) {  /* only literals => must be last "sequence", actually starting a new stream of sequences */
                         assert(storePos == storeEnd);   /* must be last sequence */
@@ -1047,11 +1093,11 @@ size_t ZSTD_compressBlock_btopt(
 
 
 /* used in 2-pass strategy */
-static U32 ZSTD_upscaleStat(U32* table, U32 lastEltIndex, int bonus)
+static U32 ZSTD_upscaleStat(unsigned* table, U32 lastEltIndex, int bonus)
 {
     U32 s, sum=0;
-    assert(ZSTD_FREQ_DIV+bonus > 0);
-    for (s=0; s<=lastEltIndex; s++) {
+    assert(ZSTD_FREQ_DIV+bonus >= 0);
+    for (s=0; slitSum = ZSTD_upscaleStat(optPtr->litFreq, MaxLit, 0);
-    optPtr->litLengthSum = ZSTD_upscaleStat(optPtr->litLengthFreq, MaxLL, 1);
-    optPtr->matchLengthSum = ZSTD_upscaleStat(optPtr->matchLengthFreq, MaxML, 1);
-    optPtr->offCodeSum = ZSTD_upscaleStat(optPtr->offCodeFreq, MaxOff, 1);
+    optPtr->litLengthSum = ZSTD_upscaleStat(optPtr->litLengthFreq, MaxLL, 0);
+    optPtr->matchLengthSum = ZSTD_upscaleStat(optPtr->matchLengthFreq, MaxML, 0);
+    optPtr->offCodeSum = ZSTD_upscaleStat(optPtr->offCodeFreq, MaxOff, 0);
+}
+
+/* ZSTD_initStats_ultra():
+ * make a first compression pass, just to seed stats with more accurate starting values.
+ * only works on first block, with no dictionary and no ldm.
+ * this function cannot error, hence its constract must be respected.
+ */
+static void
+ZSTD_initStats_ultra(ZSTD_matchState_t* ms,
+                     seqStore_t* seqStore,
+                     U32 rep[ZSTD_REP_NUM],
+               const void* src, size_t srcSize)
+{
+    U32 tmpRep[ZSTD_REP_NUM];  /* updated rep codes will sink here */
+    memcpy(tmpRep, rep, sizeof(tmpRep));
+
+    DEBUGLOG(4, "ZSTD_initStats_ultra (srcSize=%zu)", srcSize);
+    assert(ms->opt.litLengthSum == 0);    /* first block */
+    assert(seqStore->sequences == seqStore->sequencesStart);   /* no ldm */
+    assert(ms->window.dictLimit == ms->window.lowLimit);   /* no dictionary */
+    assert(ms->window.dictLimit - ms->nextToUpdate <= 1);  /* no prefix (note: intentional overflow, defined as 2-complement) */
+
+    ZSTD_compressBlock_opt_generic(ms, seqStore, tmpRep, src, srcSize, 2 /*optLevel*/, ZSTD_noDict);   /* generate stats into ms->opt*/
+
+    /* invalidate first scan from history */
+    ZSTD_resetSeqStore(seqStore);
+    ms->window.base -= srcSize;
+    ms->window.dictLimit += (U32)srcSize;
+    ms->window.lowLimit = ms->window.dictLimit;
+    ms->nextToUpdate = ms->window.dictLimit;
+    ms->nextToUpdate3 = ms->window.dictLimit;
+
+    /* re-inforce weight of collected statistics */
+    ZSTD_upscaleStats(&ms->opt);
 }
 
 size_t ZSTD_compressBlock_btultra(
@@ -1073,33 +1153,34 @@ size_t ZSTD_compressBlock_btultra(
         const void* src, size_t srcSize)
 {
     DEBUGLOG(5, "ZSTD_compressBlock_btultra (srcSize=%zu)", srcSize);
-#if 0
-    /* 2-pass strategy (disabled)
+    return ZSTD_compressBlock_opt_generic(ms, seqStore, rep, src, srcSize, 2 /*optLevel*/, ZSTD_noDict);
+}
+
+size_t ZSTD_compressBlock_btultra2(
+        ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+        const void* src, size_t srcSize)
+{
+    U32 const current = (U32)((const BYTE*)src - ms->window.base);
+    DEBUGLOG(5, "ZSTD_compressBlock_btultra2 (srcSize=%zu)", srcSize);
+
+    /* 2-pass strategy:
      * this strategy makes a first pass over first block to collect statistics
      * and seed next round's statistics with it.
+     * After 1st pass, function forgets everything, and starts a new block.
+     * Consequently, this can only work if no data has been previously loaded in tables,
+     * aka, no dictionary, no prefix, no ldm preprocessing.
      * The compression ratio gain is generally small (~0.5% on first block),
      * the cost is 2x cpu time on first block. */
     assert(srcSize <= ZSTD_BLOCKSIZE_MAX);
     if ( (ms->opt.litLengthSum==0)   /* first block */
-      && (seqStore->sequences == seqStore->sequencesStart)   /* no ldm */
-      && (ms->window.dictLimit == ms->window.lowLimit) ) {   /* no dictionary */
-        U32 tmpRep[ZSTD_REP_NUM];
-        DEBUGLOG(5, "ZSTD_compressBlock_btultra: first block: collecting statistics");
-        assert(ms->nextToUpdate >= ms->window.dictLimit
-            && ms->nextToUpdate <= ms->window.dictLimit + 1);
-        memcpy(tmpRep, rep, sizeof(tmpRep));
-        ZSTD_compressBlock_opt_generic(ms, seqStore, tmpRep, src, srcSize, 2 /*optLevel*/, ZSTD_noDict);   /* generate stats into ms->opt*/
-        ZSTD_resetSeqStore(seqStore);
-        /* invalidate first scan from history */
-        ms->window.base -= srcSize;
-        ms->window.dictLimit += (U32)srcSize;
-        ms->window.lowLimit = ms->window.dictLimit;
-        ms->nextToUpdate = ms->window.dictLimit;
-        ms->nextToUpdate3 = ms->window.dictLimit;
-        /* re-inforce weight of collected statistics */
-        ZSTD_upscaleStats(&ms->opt);
+      && (seqStore->sequences == seqStore->sequencesStart)  /* no ldm */
+      && (ms->window.dictLimit == ms->window.lowLimit)   /* no dictionary */
+      && (current == ms->window.dictLimit)   /* start of frame, nothing already loaded nor skipped */
+      && (srcSize > ZSTD_PREDEF_THRESHOLD)
+      ) {
+        ZSTD_initStats_ultra(ms, seqStore, rep, src, srcSize);
     }
-#endif
+
     return ZSTD_compressBlock_opt_generic(ms, seqStore, rep, src, srcSize, 2 /*optLevel*/, ZSTD_noDict);
 }
 
@@ -1130,3 +1211,7 @@ size_t ZSTD_compressBlock_btultra_extDict(
 {
     return ZSTD_compressBlock_opt_generic(ms, seqStore, rep, src, srcSize, 2 /*optLevel*/, ZSTD_extDict);
 }
+
+/* note : no btultra2 variant for extDict nor dictMatchState,
+ * because btultra2 is not meant to work with dictionaries
+ * and is only specific for the first block (no prefix) */
diff --git a/lib/compress/zstd_opt.h b/lib/compress/zstd_opt.h
index eeadb604c6a8..094f7476650e 100644
--- a/lib/compress/zstd_opt.h
+++ b/lib/compress/zstd_opt.h
@@ -26,6 +26,10 @@ size_t ZSTD_compressBlock_btopt(
 size_t ZSTD_compressBlock_btultra(
         ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
         void const* src, size_t srcSize);
+size_t ZSTD_compressBlock_btultra2(
+        ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+        void const* src, size_t srcSize);
+
 
 size_t ZSTD_compressBlock_btopt_dictMatchState(
         ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
@@ -41,6 +45,10 @@ size_t ZSTD_compressBlock_btultra_extDict(
         ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
         void const* src, size_t srcSize);
 
+        /* note : no btultra2 variant for extDict nor dictMatchState,
+         * because btultra2 is not meant to work with dictionaries
+         * and is only specific for the first block (no prefix) */
+
 #if defined (__cplusplus)
 }
 #endif
diff --git a/lib/compress/zstdmt_compress.c b/lib/compress/zstdmt_compress.c
index f4aba1d2c494..2cbd6ffade38 100644
--- a/lib/compress/zstdmt_compress.c
+++ b/lib/compress/zstdmt_compress.c
@@ -9,21 +9,19 @@
  */
 
 
-/* ======   Tuning parameters   ====== */
-#define ZSTDMT_NBWORKERS_MAX 200
-#define ZSTDMT_JOBSIZE_MAX  (MEM_32bits() ? (512 MB) : (2 GB))  /* note : limited by `jobSize` type, which is `unsigned` */
-#define ZSTDMT_OVERLAPLOG_DEFAULT 6
-
-
 /* ======   Compiler specifics   ====== */
 #if defined(_MSC_VER)
 #  pragma warning(disable : 4204)   /* disable: C4204: non-constant aggregate initializer */
 #endif
 
 
+/* ======   Constants   ====== */
+#define ZSTDMT_OVERLAPLOG_DEFAULT 0
+
+
 /* ======   Dependencies   ====== */
 #include       /* memcpy, memset */
-#include       /* INT_MAX */
+#include       /* INT_MAX, UINT_MAX */
 #include "pool.h"        /* threadpool */
 #include "threading.h"   /* mutex */
 #include "zstd_compress_internal.h"  /* MIN, ERROR, ZSTD_*, ZSTD_highbit32 */
@@ -57,9 +55,9 @@ static unsigned long long GetCurrentClockTimeMicroseconds(void)
    static clock_t _ticksPerSecond = 0;
    if (_ticksPerSecond <= 0) _ticksPerSecond = sysconf(_SC_CLK_TCK);
 
-   { struct tms junk; clock_t newTicks = (clock_t) times(&junk);
-     return ((((unsigned long long)newTicks)*(1000000))/_ticksPerSecond); }
-}
+   {   struct tms junk; clock_t newTicks = (clock_t) times(&junk);
+       return ((((unsigned long long)newTicks)*(1000000))/_ticksPerSecond);
+}  }
 
 #define MUTEX_WAIT_TIME_DLEVEL 6
 #define ZSTD_PTHREAD_MUTEX_LOCK(mutex) {          \
@@ -342,8 +340,8 @@ static ZSTDMT_seqPool* ZSTDMT_expandSeqPool(ZSTDMT_seqPool* pool, U32 nbWorkers)
 
 typedef struct {
     ZSTD_pthread_mutex_t poolMutex;
-    unsigned totalCCtx;
-    unsigned availCCtx;
+    int totalCCtx;
+    int availCCtx;
     ZSTD_customMem cMem;
     ZSTD_CCtx* cctx[1];   /* variable size */
 } ZSTDMT_CCtxPool;
@@ -351,16 +349,16 @@ typedef struct {
 /* note : all CCtx borrowed from the pool should be released back to the pool _before_ freeing the pool */
 static void ZSTDMT_freeCCtxPool(ZSTDMT_CCtxPool* pool)
 {
-    unsigned u;
-    for (u=0; utotalCCtx; u++)
-        ZSTD_freeCCtx(pool->cctx[u]);  /* note : compatible with free on NULL */
+    int cid;
+    for (cid=0; cidtotalCCtx; cid++)
+        ZSTD_freeCCtx(pool->cctx[cid]);  /* note : compatible with free on NULL */
     ZSTD_pthread_mutex_destroy(&pool->poolMutex);
     ZSTD_free(pool, pool->cMem);
 }
 
 /* ZSTDMT_createCCtxPool() :
  * implies nbWorkers >= 1 , checked by caller ZSTDMT_createCCtx() */
-static ZSTDMT_CCtxPool* ZSTDMT_createCCtxPool(unsigned nbWorkers,
+static ZSTDMT_CCtxPool* ZSTDMT_createCCtxPool(int nbWorkers,
                                               ZSTD_customMem cMem)
 {
     ZSTDMT_CCtxPool* const cctxPool = (ZSTDMT_CCtxPool*) ZSTD_calloc(
@@ -381,7 +379,7 @@ static ZSTDMT_CCtxPool* ZSTDMT_createCCtxPool(unsigned nbWorkers,
 }
 
 static ZSTDMT_CCtxPool* ZSTDMT_expandCCtxPool(ZSTDMT_CCtxPool* srcPool,
-                                              unsigned nbWorkers)
+                                              int nbWorkers)
 {
     if (srcPool==NULL) return NULL;
     if (nbWorkers <= srcPool->totalCCtx) return srcPool;   /* good enough */
@@ -469,9 +467,9 @@ static int ZSTDMT_serialState_reset(serialState_t* serialState, ZSTDMT_seqPool*
         DEBUGLOG(4, "LDM window size = %u KB", (1U << params.cParams.windowLog) >> 10);
         ZSTD_ldm_adjustParameters(¶ms.ldmParams, ¶ms.cParams);
         assert(params.ldmParams.hashLog >= params.ldmParams.bucketSizeLog);
-        assert(params.ldmParams.hashEveryLog < 32);
+        assert(params.ldmParams.hashRateLog < 32);
         serialState->ldmState.hashPower =
-                ZSTD_ldm_getHashPower(params.ldmParams.minMatchLength);
+                ZSTD_rollingHash_primePower(params.ldmParams.minMatchLength);
     } else {
         memset(¶ms.ldmParams, 0, sizeof(params.ldmParams));
     }
@@ -674,7 +672,7 @@ static void ZSTDMT_compressionJob(void* jobDescription)
         if (ZSTD_isError(initError)) JOB_ERROR(initError);
     } else {  /* srcStart points at reloaded section */
         U64 const pledgedSrcSize = job->firstJob ? job->fullFrameSize : job->src.size;
-        {   size_t const forceWindowError = ZSTD_CCtxParam_setParameter(&jobParams, ZSTD_p_forceMaxWindow, !job->firstJob);
+        {   size_t const forceWindowError = ZSTD_CCtxParam_setParameter(&jobParams, ZSTD_c_forceMaxWindow, !job->firstJob);
             if (ZSTD_isError(forceWindowError)) JOB_ERROR(forceWindowError);
         }
         {   size_t const initError = ZSTD_compressBegin_advanced_internal(cctx,
@@ -777,6 +775,14 @@ typedef struct {
 
 static const roundBuff_t kNullRoundBuff = {NULL, 0, 0};
 
+#define RSYNC_LENGTH 32
+
+typedef struct {
+  U64 hash;
+  U64 hitMask;
+  U64 primePower;
+} rsyncState_t;
+
 struct ZSTDMT_CCtx_s {
     POOL_ctx* factory;
     ZSTDMT_jobDescription* jobs;
@@ -790,6 +796,7 @@ struct ZSTDMT_CCtx_s {
     inBuff_t inBuff;
     roundBuff_t roundBuff;
     serialState_t serial;
+    rsyncState_t rsync;
     unsigned singleBlockingThread;
     unsigned jobIDMask;
     unsigned doneJobID;
@@ -859,7 +866,7 @@ size_t ZSTDMT_CCtxParam_setNbWorkers(ZSTD_CCtx_params* params, unsigned nbWorker
 {
     if (nbWorkers > ZSTDMT_NBWORKERS_MAX) nbWorkers = ZSTDMT_NBWORKERS_MAX;
     params->nbWorkers = nbWorkers;
-    params->overlapSizeLog = ZSTDMT_OVERLAPLOG_DEFAULT;
+    params->overlapLog = ZSTDMT_OVERLAPLOG_DEFAULT;
     params->jobSize = 0;
     return nbWorkers;
 }
@@ -969,52 +976,59 @@ size_t ZSTDMT_sizeof_CCtx(ZSTDMT_CCtx* mtctx)
 }
 
 /* Internal only */
-size_t ZSTDMT_CCtxParam_setMTCtxParameter(ZSTD_CCtx_params* params,
-                                ZSTDMT_parameter parameter, unsigned value) {
+size_t
+ZSTDMT_CCtxParam_setMTCtxParameter(ZSTD_CCtx_params* params,
+                                   ZSTDMT_parameter parameter,
+                                   int value)
+{
     DEBUGLOG(4, "ZSTDMT_CCtxParam_setMTCtxParameter");
     switch(parameter)
     {
     case ZSTDMT_p_jobSize :
-        DEBUGLOG(4, "ZSTDMT_CCtxParam_setMTCtxParameter : set jobSize to %u", value);
-        if ( (value > 0)  /* value==0 => automatic job size */
-           & (value < ZSTDMT_JOBSIZE_MIN) )
+        DEBUGLOG(4, "ZSTDMT_CCtxParam_setMTCtxParameter : set jobSize to %i", value);
+        if ( value != 0  /* default */
+          && value < ZSTDMT_JOBSIZE_MIN)
             value = ZSTDMT_JOBSIZE_MIN;
-        if (value > ZSTDMT_JOBSIZE_MAX)
-            value = ZSTDMT_JOBSIZE_MAX;
+        assert(value >= 0);
+        if (value > ZSTDMT_JOBSIZE_MAX) value = ZSTDMT_JOBSIZE_MAX;
         params->jobSize = value;
         return value;
-    case ZSTDMT_p_overlapSectionLog :
-        if (value > 9) value = 9;
-        DEBUGLOG(4, "ZSTDMT_p_overlapSectionLog : %u", value);
-        params->overlapSizeLog = (value >= 9) ? 9 : value;
+
+    case ZSTDMT_p_overlapLog :
+        DEBUGLOG(4, "ZSTDMT_p_overlapLog : %i", value);
+        if (value < ZSTD_OVERLAPLOG_MIN) value = ZSTD_OVERLAPLOG_MIN;
+        if (value > ZSTD_OVERLAPLOG_MAX) value = ZSTD_OVERLAPLOG_MAX;
+        params->overlapLog = value;
         return value;
+
+    case ZSTDMT_p_rsyncable :
+        value = (value != 0);
+        params->rsyncable = value;
+        return value;
+
     default :
         return ERROR(parameter_unsupported);
     }
 }
 
-size_t ZSTDMT_setMTCtxParameter(ZSTDMT_CCtx* mtctx, ZSTDMT_parameter parameter, unsigned value)
+size_t ZSTDMT_setMTCtxParameter(ZSTDMT_CCtx* mtctx, ZSTDMT_parameter parameter, int value)
 {
     DEBUGLOG(4, "ZSTDMT_setMTCtxParameter");
-    switch(parameter)
-    {
-    case ZSTDMT_p_jobSize :
-        return ZSTDMT_CCtxParam_setMTCtxParameter(&mtctx->params, parameter, value);
-    case ZSTDMT_p_overlapSectionLog :
-        return ZSTDMT_CCtxParam_setMTCtxParameter(&mtctx->params, parameter, value);
-    default :
-        return ERROR(parameter_unsupported);
-    }
+    return ZSTDMT_CCtxParam_setMTCtxParameter(&mtctx->params, parameter, value);
 }
 
-size_t ZSTDMT_getMTCtxParameter(ZSTDMT_CCtx* mtctx, ZSTDMT_parameter parameter, unsigned* value)
+size_t ZSTDMT_getMTCtxParameter(ZSTDMT_CCtx* mtctx, ZSTDMT_parameter parameter, int* value)
 {
     switch (parameter) {
     case ZSTDMT_p_jobSize:
-        *value = mtctx->params.jobSize;
+        assert(mtctx->params.jobSize <= INT_MAX);
+        *value = (int)(mtctx->params.jobSize);
         break;
-    case ZSTDMT_p_overlapSectionLog:
-        *value = mtctx->params.overlapSizeLog;
+    case ZSTDMT_p_overlapLog:
+        *value = mtctx->params.overlapLog;
+        break;
+    case ZSTDMT_p_rsyncable:
+        *value = mtctx->params.rsyncable;
         break;
     default:
         return ERROR(parameter_unsupported);
@@ -1140,22 +1154,66 @@ size_t ZSTDMT_toFlushNow(ZSTDMT_CCtx* mtctx)
 /* =====   Multi-threaded compression   ===== */
 /* ------------------------------------------ */
 
-static size_t ZSTDMT_computeTargetJobLog(ZSTD_CCtx_params const params)
+static unsigned ZSTDMT_computeTargetJobLog(ZSTD_CCtx_params const params)
 {
     if (params.ldmParams.enableLdm)
+        /* In Long Range Mode, the windowLog is typically oversized.
+         * In which case, it's preferable to determine the jobSize
+         * based on chainLog instead. */
         return MAX(21, params.cParams.chainLog + 4);
     return MAX(20, params.cParams.windowLog + 2);
 }
 
-static size_t ZSTDMT_computeOverlapLog(ZSTD_CCtx_params const params)
+static int ZSTDMT_overlapLog_default(ZSTD_strategy strat)
 {
-    unsigned const overlapRLog = (params.overlapSizeLog>9) ? 0 : 9-params.overlapSizeLog;
-    if (params.ldmParams.enableLdm)
-        return (MIN(params.cParams.windowLog, ZSTDMT_computeTargetJobLog(params) - 2) - overlapRLog);
-    return overlapRLog >= 9 ? 0 : (params.cParams.windowLog - overlapRLog);
+    switch(strat)
+    {
+        case ZSTD_btultra2:
+            return 9;
+        case ZSTD_btultra:
+        case ZSTD_btopt:
+            return 8;
+        case ZSTD_btlazy2:
+        case ZSTD_lazy2:
+            return 7;
+        case ZSTD_lazy:
+        case ZSTD_greedy:
+        case ZSTD_dfast:
+        case ZSTD_fast:
+        default:;
+    }
+    return 6;
 }
 
-static unsigned ZSTDMT_computeNbJobs(ZSTD_CCtx_params params, size_t srcSize, unsigned nbWorkers) {
+static int ZSTDMT_overlapLog(int ovlog, ZSTD_strategy strat)
+{
+    assert(0 <= ovlog && ovlog <= 9);
+    if (ovlog == 0) return ZSTDMT_overlapLog_default(strat);
+    return ovlog;
+}
+
+static size_t ZSTDMT_computeOverlapSize(ZSTD_CCtx_params const params)
+{
+    int const overlapRLog = 9 - ZSTDMT_overlapLog(params.overlapLog, params.cParams.strategy);
+    int ovLog = (overlapRLog >= 8) ? 0 : (params.cParams.windowLog - overlapRLog);
+    assert(0 <= overlapRLog && overlapRLog <= 8);
+    if (params.ldmParams.enableLdm) {
+        /* In Long Range Mode, the windowLog is typically oversized.
+         * In which case, it's preferable to determine the jobSize
+         * based on chainLog instead.
+         * Then, ovLog becomes a fraction of the jobSize, rather than windowSize */
+        ovLog = MIN(params.cParams.windowLog, ZSTDMT_computeTargetJobLog(params) - 2)
+                - overlapRLog;
+    }
+    assert(0 <= ovLog && ovLog <= 30);
+    DEBUGLOG(4, "overlapLog : %i", params.overlapLog);
+    DEBUGLOG(4, "overlap size : %i", 1 << ovLog);
+    return (ovLog==0) ? 0 : (size_t)1 << ovLog;
+}
+
+static unsigned
+ZSTDMT_computeNbJobs(ZSTD_CCtx_params params, size_t srcSize, unsigned nbWorkers)
+{
     assert(nbWorkers>0);
     {   size_t const jobSizeTarget = (size_t)1 << ZSTDMT_computeTargetJobLog(params);
         size_t const jobMaxSize = jobSizeTarget << 2;
@@ -1178,7 +1236,7 @@ static size_t ZSTDMT_compress_advanced_internal(
                 ZSTD_CCtx_params params)
 {
     ZSTD_CCtx_params const jobParams = ZSTDMT_initJobCCtxParams(params);
-    size_t const overlapSize = (size_t)1 << ZSTDMT_computeOverlapLog(params);
+    size_t const overlapSize = ZSTDMT_computeOverlapSize(params);
     unsigned const nbJobs = ZSTDMT_computeNbJobs(params, srcSize, params.nbWorkers);
     size_t const proposedJobSize = (srcSize + (nbJobs-1)) / nbJobs;
     size_t const avgJobSize = (((proposedJobSize-1) & 0x1FFFF) < 0x7FFF) ? proposedJobSize + 0xFFFF : proposedJobSize;   /* avoid too small last block */
@@ -1289,16 +1347,17 @@ static size_t ZSTDMT_compress_advanced_internal(
 }
 
 size_t ZSTDMT_compress_advanced(ZSTDMT_CCtx* mtctx,
-                               void* dst, size_t dstCapacity,
-                         const void* src, size_t srcSize,
-                         const ZSTD_CDict* cdict,
-                               ZSTD_parameters params,
-                               unsigned overlapLog)
+                                void* dst, size_t dstCapacity,
+                          const void* src, size_t srcSize,
+                          const ZSTD_CDict* cdict,
+                                ZSTD_parameters params,
+                                int overlapLog)
 {
     ZSTD_CCtx_params cctxParams = mtctx->params;
     cctxParams.cParams = params.cParams;
     cctxParams.fParams = params.fParams;
-    cctxParams.overlapSizeLog = overlapLog;
+    assert(ZSTD_OVERLAPLOG_MIN <= overlapLog && overlapLog <= ZSTD_OVERLAPLOG_MAX);
+    cctxParams.overlapLog = overlapLog;
     return ZSTDMT_compress_advanced_internal(mtctx,
                                              dst, dstCapacity,
                                              src, srcSize,
@@ -1311,8 +1370,8 @@ size_t ZSTDMT_compressCCtx(ZSTDMT_CCtx* mtctx,
                      const void* src, size_t srcSize,
                            int compressionLevel)
 {
-    U32 const overlapLog = (compressionLevel >= ZSTD_maxCLevel()) ? 9 : ZSTDMT_OVERLAPLOG_DEFAULT;
     ZSTD_parameters params = ZSTD_getParams(compressionLevel, srcSize, 0);
+    int const overlapLog = ZSTDMT_overlapLog_default(params.cParams.strategy);
     params.fParams.contentSizeFlag = 1;
     return ZSTDMT_compress_advanced(mtctx, dst, dstCapacity, src, srcSize, NULL, params, overlapLog);
 }
@@ -1339,8 +1398,8 @@ size_t ZSTDMT_initCStream_internal(
     if (params.nbWorkers != mtctx->params.nbWorkers)
         CHECK_F( ZSTDMT_resize(mtctx, params.nbWorkers) );
 
-    if (params.jobSize > 0 && params.jobSize < ZSTDMT_JOBSIZE_MIN) params.jobSize = ZSTDMT_JOBSIZE_MIN;
-    if (params.jobSize > ZSTDMT_JOBSIZE_MAX) params.jobSize = ZSTDMT_JOBSIZE_MAX;
+    if (params.jobSize != 0 && params.jobSize < ZSTDMT_JOBSIZE_MIN) params.jobSize = ZSTDMT_JOBSIZE_MIN;
+    if (params.jobSize > (size_t)ZSTDMT_JOBSIZE_MAX) params.jobSize = ZSTDMT_JOBSIZE_MAX;
 
     mtctx->singleBlockingThread = (pledgedSrcSize <= ZSTDMT_JOBSIZE_MIN);  /* do not trigger multi-threading when srcSize is too small */
     if (mtctx->singleBlockingThread) {
@@ -1375,14 +1434,24 @@ size_t ZSTDMT_initCStream_internal(
         mtctx->cdict = cdict;
     }
 
-    mtctx->targetPrefixSize = (size_t)1 << ZSTDMT_computeOverlapLog(params);
-    DEBUGLOG(4, "overlapLog=%u => %u KB", params.overlapSizeLog, (U32)(mtctx->targetPrefixSize>>10));
+    mtctx->targetPrefixSize = ZSTDMT_computeOverlapSize(params);
+    DEBUGLOG(4, "overlapLog=%i => %u KB", params.overlapLog, (U32)(mtctx->targetPrefixSize>>10));
     mtctx->targetSectionSize = params.jobSize;
     if (mtctx->targetSectionSize == 0) {
         mtctx->targetSectionSize = 1ULL << ZSTDMT_computeTargetJobLog(params);
     }
+    if (params.rsyncable) {
+        /* Aim for the targetsectionSize as the average job size. */
+        U32 const jobSizeMB = (U32)(mtctx->targetSectionSize >> 20);
+        U32 const rsyncBits = ZSTD_highbit32(jobSizeMB) + 20;
+        assert(jobSizeMB >= 1);
+        DEBUGLOG(4, "rsyncLog = %u", rsyncBits);
+        mtctx->rsync.hash = 0;
+        mtctx->rsync.hitMask = (1ULL << rsyncBits) - 1;
+        mtctx->rsync.primePower = ZSTD_rollingHash_primePower(RSYNC_LENGTH);
+    }
     if (mtctx->targetSectionSize < mtctx->targetPrefixSize) mtctx->targetSectionSize = mtctx->targetPrefixSize;  /* job size must be >= overlap size */
-    DEBUGLOG(4, "Job Size : %u KB (note : set to %u)", (U32)(mtctx->targetSectionSize>>10), params.jobSize);
+    DEBUGLOG(4, "Job Size : %u KB (note : set to %u)", (U32)(mtctx->targetSectionSize>>10), (U32)params.jobSize);
     DEBUGLOG(4, "inBuff Size : %u KB", (U32)(mtctx->targetSectionSize>>10));
     ZSTDMT_setBufferSize(mtctx->bufPool, ZSTD_compressBound(mtctx->targetSectionSize));
     {
@@ -1818,6 +1887,89 @@ static int ZSTDMT_tryGetInputRange(ZSTDMT_CCtx* mtctx)
     return 1;
 }
 
+typedef struct {
+  size_t toLoad;  /* The number of bytes to load from the input. */
+  int flush;      /* Boolean declaring if we must flush because we found a synchronization point. */
+} syncPoint_t;
+
+/**
+ * Searches through the input for a synchronization point. If one is found, we
+ * will instruct the caller to flush, and return the number of bytes to load.
+ * Otherwise, we will load as many bytes as possible and instruct the caller
+ * to continue as normal.
+ */
+static syncPoint_t
+findSynchronizationPoint(ZSTDMT_CCtx const* mtctx, ZSTD_inBuffer const input)
+{
+    BYTE const* const istart = (BYTE const*)input.src + input.pos;
+    U64 const primePower = mtctx->rsync.primePower;
+    U64 const hitMask = mtctx->rsync.hitMask;
+
+    syncPoint_t syncPoint;
+    U64 hash;
+    BYTE const* prev;
+    size_t pos;
+
+    syncPoint.toLoad = MIN(input.size - input.pos, mtctx->targetSectionSize - mtctx->inBuff.filled);
+    syncPoint.flush = 0;
+    if (!mtctx->params.rsyncable)
+        /* Rsync is disabled. */
+        return syncPoint;
+    if (mtctx->inBuff.filled + syncPoint.toLoad < RSYNC_LENGTH)
+        /* Not enough to compute the hash.
+         * We will miss any synchronization points in this RSYNC_LENGTH byte
+         * window. However, since it depends only in the internal buffers, if the
+         * state is already synchronized, we will remain synchronized.
+         * Additionally, the probability that we miss a synchronization point is
+         * low: RSYNC_LENGTH / targetSectionSize.
+         */
+        return syncPoint;
+    /* Initialize the loop variables. */
+    if (mtctx->inBuff.filled >= RSYNC_LENGTH) {
+        /* We have enough bytes buffered to initialize the hash.
+         * Start scanning at the beginning of the input.
+         */
+        pos = 0;
+        prev = (BYTE const*)mtctx->inBuff.buffer.start + mtctx->inBuff.filled - RSYNC_LENGTH;
+        hash = ZSTD_rollingHash_compute(prev, RSYNC_LENGTH);
+    } else {
+        /* We don't have enough bytes buffered to initialize the hash, but
+         * we know we have at least RSYNC_LENGTH bytes total.
+         * Start scanning after the first RSYNC_LENGTH bytes less the bytes
+         * already buffered.
+         */
+        pos = RSYNC_LENGTH - mtctx->inBuff.filled;
+        prev = (BYTE const*)mtctx->inBuff.buffer.start - pos;
+        hash = ZSTD_rollingHash_compute(mtctx->inBuff.buffer.start, mtctx->inBuff.filled);
+        hash = ZSTD_rollingHash_append(hash, istart, pos);
+    }
+    /* Starting with the hash of the previous RSYNC_LENGTH bytes, roll
+     * through the input. If we hit a synchronization point, then cut the
+     * job off, and tell the compressor to flush the job. Otherwise, load
+     * all the bytes and continue as normal.
+     * If we go too long without a synchronization point (targetSectionSize)
+     * then a block will be emitted anyways, but this is okay, since if we
+     * are already synchronized we will remain synchronized.
+     */
+    for (; pos < syncPoint.toLoad; ++pos) {
+        BYTE const toRemove = pos < RSYNC_LENGTH ? prev[pos] : istart[pos - RSYNC_LENGTH];
+        /* if (pos >= RSYNC_LENGTH) assert(ZSTD_rollingHash_compute(istart + pos - RSYNC_LENGTH, RSYNC_LENGTH) == hash); */
+        hash = ZSTD_rollingHash_rotate(hash, toRemove, istart[pos], primePower);
+        if ((hash & hitMask) == hitMask) {
+            syncPoint.toLoad = pos + 1;
+            syncPoint.flush = 1;
+            break;
+        }
+    }
+    return syncPoint;
+}
+
+size_t ZSTDMT_nextInputSizeHint(const ZSTDMT_CCtx* mtctx)
+{
+    size_t hintInSize = mtctx->targetSectionSize - mtctx->inBuff.filled;
+    if (hintInSize==0) hintInSize = mtctx->targetSectionSize;
+    return hintInSize;
+}
 
 /** ZSTDMT_compressStream_generic() :
  *  internal use only - exposed to be invoked from zstd_compress.c
@@ -1844,7 +1996,8 @@ size_t ZSTDMT_compressStream_generic(ZSTDMT_CCtx* mtctx,
     }
 
     /* single-pass shortcut (note : synchronous-mode) */
-    if ( (mtctx->nextJobID == 0)      /* just started */
+    if ( (!mtctx->params.rsyncable)   /* rsyncable mode is disabled */
+      && (mtctx->nextJobID == 0)      /* just started */
       && (mtctx->inBuff.filled == 0)  /* nothing buffered */
       && (!mtctx->jobReady)           /* no job already created */
       && (endOp == ZSTD_e_end)        /* end order */
@@ -1876,14 +2029,17 @@ size_t ZSTDMT_compressStream_generic(ZSTDMT_CCtx* mtctx,
                 DEBUGLOG(5, "ZSTDMT_tryGetInputRange completed successfully : mtctx->inBuff.buffer.start = %p", mtctx->inBuff.buffer.start);
         }
         if (mtctx->inBuff.buffer.start != NULL) {
-            size_t const toLoad = MIN(input->size - input->pos, mtctx->targetSectionSize - mtctx->inBuff.filled);
+            syncPoint_t const syncPoint = findSynchronizationPoint(mtctx, *input);
+            if (syncPoint.flush && endOp == ZSTD_e_continue) {
+                endOp = ZSTD_e_flush;
+            }
             assert(mtctx->inBuff.buffer.capacity >= mtctx->targetSectionSize);
             DEBUGLOG(5, "ZSTDMT_compressStream_generic: adding %u bytes on top of %u to buffer of size %u",
-                        (U32)toLoad, (U32)mtctx->inBuff.filled, (U32)mtctx->targetSectionSize);
-            memcpy((char*)mtctx->inBuff.buffer.start + mtctx->inBuff.filled, (const char*)input->src + input->pos, toLoad);
-            input->pos += toLoad;
-            mtctx->inBuff.filled += toLoad;
-            forwardInputProgress = toLoad>0;
+                        (U32)syncPoint.toLoad, (U32)mtctx->inBuff.filled, (U32)mtctx->targetSectionSize);
+            memcpy((char*)mtctx->inBuff.buffer.start + mtctx->inBuff.filled, (const char*)input->src + input->pos, syncPoint.toLoad);
+            input->pos += syncPoint.toLoad;
+            mtctx->inBuff.filled += syncPoint.toLoad;
+            forwardInputProgress = syncPoint.toLoad>0;
         }
         if ((input->pos < input->size) && (endOp == ZSTD_e_end))
             endOp = ZSTD_e_flush;   /* can't end now : not all input consumed */
diff --git a/lib/compress/zstdmt_compress.h b/lib/compress/zstdmt_compress.h
index 12ad9f899b57..ee771681fa59 100644
--- a/lib/compress/zstdmt_compress.h
+++ b/lib/compress/zstdmt_compress.h
@@ -28,6 +28,16 @@
 #include "zstd.h"            /* ZSTD_inBuffer, ZSTD_outBuffer, ZSTDLIB_API */
 
 
+/* ===   Constants   === */
+#ifndef ZSTDMT_NBWORKERS_MAX
+#  define ZSTDMT_NBWORKERS_MAX 200
+#endif
+#ifndef ZSTDMT_JOBSIZE_MIN
+#  define ZSTDMT_JOBSIZE_MIN (1 MB)
+#endif
+#define ZSTDMT_JOBSIZE_MAX  (MEM_32bits() ? (512 MB) : (1024 MB))
+
+
 /* ===   Memory management   === */
 typedef struct ZSTDMT_CCtx_s ZSTDMT_CCtx;
 ZSTDLIB_API ZSTDMT_CCtx* ZSTDMT_createCCtx(unsigned nbWorkers);
@@ -52,6 +62,7 @@ ZSTDLIB_API size_t ZSTDMT_compressCCtx(ZSTDMT_CCtx* mtctx,
 ZSTDLIB_API size_t ZSTDMT_initCStream(ZSTDMT_CCtx* mtctx, int compressionLevel);
 ZSTDLIB_API size_t ZSTDMT_resetCStream(ZSTDMT_CCtx* mtctx, unsigned long long pledgedSrcSize);  /**< if srcSize is not known at reset time, use ZSTD_CONTENTSIZE_UNKNOWN. Note: for compatibility with older programs, 0 means the same as ZSTD_CONTENTSIZE_UNKNOWN, but it will change in the future to mean "empty" */
 
+ZSTDLIB_API size_t ZSTDMT_nextInputSizeHint(const ZSTDMT_CCtx* mtctx);
 ZSTDLIB_API size_t ZSTDMT_compressStream(ZSTDMT_CCtx* mtctx, ZSTD_outBuffer* output, ZSTD_inBuffer* input);
 
 ZSTDLIB_API size_t ZSTDMT_flushStream(ZSTDMT_CCtx* mtctx, ZSTD_outBuffer* output);   /**< @return : 0 == all flushed; >0 : still some data to be flushed; or an error code (ZSTD_isError()) */
@@ -60,16 +71,12 @@ ZSTDLIB_API size_t ZSTDMT_endStream(ZSTDMT_CCtx* mtctx, ZSTD_outBuffer* output);
 
 /* ===   Advanced functions and parameters  === */
 
-#ifndef ZSTDMT_JOBSIZE_MIN
-#  define ZSTDMT_JOBSIZE_MIN (1U << 20)   /* 1 MB - Minimum size of each compression job */
-#endif
-
 ZSTDLIB_API size_t ZSTDMT_compress_advanced(ZSTDMT_CCtx* mtctx,
                                            void* dst, size_t dstCapacity,
                                      const void* src, size_t srcSize,
                                      const ZSTD_CDict* cdict,
                                            ZSTD_parameters params,
-                                           unsigned overlapLog);
+                                           int overlapLog);
 
 ZSTDLIB_API size_t ZSTDMT_initCStream_advanced(ZSTDMT_CCtx* mtctx,
                                         const void* dict, size_t dictSize,   /* dict can be released after init, a local copy is preserved within zcs */
@@ -84,8 +91,9 @@ ZSTDLIB_API size_t ZSTDMT_initCStream_usingCDict(ZSTDMT_CCtx* mtctx,
 /* ZSTDMT_parameter :
  * List of parameters that can be set using ZSTDMT_setMTCtxParameter() */
 typedef enum {
-    ZSTDMT_p_jobSize,           /* Each job is compressed in parallel. By default, this value is dynamically determined depending on compression parameters. Can be set explicitly here. */
-    ZSTDMT_p_overlapSectionLog  /* Each job may reload a part of previous job to enhance compressionr ratio; 0 == no overlap, 6(default) == use 1/8th of window, >=9 == use full window. This is a "sticky" parameter : its value will be re-used on next compression job */
+    ZSTDMT_p_jobSize,     /* Each job is compressed in parallel. By default, this value is dynamically determined depending on compression parameters. Can be set explicitly here. */
+    ZSTDMT_p_overlapLog,  /* Each job may reload a part of previous job to enhance compressionr ratio; 0 == no overlap, 6(default) == use 1/8th of window, >=9 == use full window. This is a "sticky" parameter : its value will be re-used on next compression job */
+    ZSTDMT_p_rsyncable    /* Enables rsyncable mode. */
 } ZSTDMT_parameter;
 
 /* ZSTDMT_setMTCtxParameter() :
@@ -93,12 +101,12 @@ typedef enum {
  * The function must be called typically after ZSTD_createCCtx() but __before ZSTDMT_init*() !__
  * Parameters not explicitly reset by ZSTDMT_init*() remain the same in consecutive compression sessions.
  * @return : 0, or an error code (which can be tested using ZSTD_isError()) */
-ZSTDLIB_API size_t ZSTDMT_setMTCtxParameter(ZSTDMT_CCtx* mtctx, ZSTDMT_parameter parameter, unsigned value);
+ZSTDLIB_API size_t ZSTDMT_setMTCtxParameter(ZSTDMT_CCtx* mtctx, ZSTDMT_parameter parameter, int value);
 
 /* ZSTDMT_getMTCtxParameter() :
  * Query the ZSTDMT_CCtx for a parameter value.
  * @return : 0, or an error code (which can be tested using ZSTD_isError()) */
-ZSTDLIB_API size_t ZSTDMT_getMTCtxParameter(ZSTDMT_CCtx* mtctx, ZSTDMT_parameter parameter, unsigned* value);
+ZSTDLIB_API size_t ZSTDMT_getMTCtxParameter(ZSTDMT_CCtx* mtctx, ZSTDMT_parameter parameter, int* value);
 
 
 /*! ZSTDMT_compressStream_generic() :
@@ -129,7 +137,7 @@ size_t ZSTDMT_toFlushNow(ZSTDMT_CCtx* mtctx);
 
 /*! ZSTDMT_CCtxParam_setMTCtxParameter()
  *  like ZSTDMT_setMTCtxParameter(), but into a ZSTD_CCtx_Params */
-size_t ZSTDMT_CCtxParam_setMTCtxParameter(ZSTD_CCtx_params* params, ZSTDMT_parameter parameter, unsigned value);
+size_t ZSTDMT_CCtxParam_setMTCtxParameter(ZSTD_CCtx_params* params, ZSTDMT_parameter parameter, int value);
 
 /*! ZSTDMT_CCtxParam_setNbWorkers()
  *  Set nbWorkers, and clamp it.
diff --git a/lib/decompress/huf_decompress.c b/lib/decompress/huf_decompress.c
index 83ecaff01e84..3f8bd2973207 100644
--- a/lib/decompress/huf_decompress.c
+++ b/lib/decompress/huf_decompress.c
@@ -43,6 +43,19 @@
 #include "huf.h"
 #include "error_private.h"
 
+/* **************************************************************
+*  Macros
+****************************************************************/
+
+/* These two optional macros force the use one way or another of the two
+ * Huffman decompression implementations. You can't force in both directions
+ * at the same time.
+ */
+#if defined(HUF_FORCE_DECOMPRESS_X1) && \
+    defined(HUF_FORCE_DECOMPRESS_X2)
+#error "Cannot force the use of the X1 and X2 decoders at the same time!"
+#endif
+
 
 /* **************************************************************
 *  Error Management
@@ -58,6 +71,51 @@
 #define HUF_ALIGN_MASK(x, mask) (((x) + (mask)) & ~(mask))
 
 
+/* **************************************************************
+*  BMI2 Variant Wrappers
+****************************************************************/
+#if DYNAMIC_BMI2
+
+#define HUF_DGEN(fn)                                                        \
+                                                                            \
+    static size_t fn##_default(                                             \
+                  void* dst,  size_t dstSize,                               \
+            const void* cSrc, size_t cSrcSize,                              \
+            const HUF_DTable* DTable)                                       \
+    {                                                                       \
+        return fn##_body(dst, dstSize, cSrc, cSrcSize, DTable);             \
+    }                                                                       \
+                                                                            \
+    static TARGET_ATTRIBUTE("bmi2") size_t fn##_bmi2(                       \
+                  void* dst,  size_t dstSize,                               \
+            const void* cSrc, size_t cSrcSize,                              \
+            const HUF_DTable* DTable)                                       \
+    {                                                                       \
+        return fn##_body(dst, dstSize, cSrc, cSrcSize, DTable);             \
+    }                                                                       \
+                                                                            \
+    static size_t fn(void* dst, size_t dstSize, void const* cSrc,           \
+                     size_t cSrcSize, HUF_DTable const* DTable, int bmi2)   \
+    {                                                                       \
+        if (bmi2) {                                                         \
+            return fn##_bmi2(dst, dstSize, cSrc, cSrcSize, DTable);         \
+        }                                                                   \
+        return fn##_default(dst, dstSize, cSrc, cSrcSize, DTable);          \
+    }
+
+#else
+
+#define HUF_DGEN(fn)                                                        \
+    static size_t fn(void* dst, size_t dstSize, void const* cSrc,           \
+                     size_t cSrcSize, HUF_DTable const* DTable, int bmi2)   \
+    {                                                                       \
+        (void)bmi2;                                                         \
+        return fn##_body(dst, dstSize, cSrc, cSrcSize, DTable);             \
+    }
+
+#endif
+
+
 /*-***************************/
 /*  generic DTableDesc       */
 /*-***************************/
@@ -71,6 +129,8 @@ static DTableDesc HUF_getDTableDesc(const HUF_DTable* table)
 }
 
 
+#ifndef HUF_FORCE_DECOMPRESS_X2
+
 /*-***************************/
 /*  single-symbol decoding   */
 /*-***************************/
@@ -307,46 +367,6 @@ typedef size_t (*HUF_decompress_usingDTable_t)(void *dst, size_t dstSize,
                                                const void *cSrc,
                                                size_t cSrcSize,
                                                const HUF_DTable *DTable);
-#if DYNAMIC_BMI2
-
-#define HUF_DGEN(fn)                                                               \
-                                                                            \
-    static size_t fn##_default(                                             \
-                  void* dst,  size_t dstSize,                               \
-            const void* cSrc, size_t cSrcSize,                              \
-            const HUF_DTable* DTable)                                       \
-    {                                                                       \
-        return fn##_body(dst, dstSize, cSrc, cSrcSize, DTable);             \
-    }                                                                       \
-                                                                            \
-    static TARGET_ATTRIBUTE("bmi2") size_t fn##_bmi2(                       \
-                  void* dst,  size_t dstSize,                               \
-            const void* cSrc, size_t cSrcSize,                              \
-            const HUF_DTable* DTable)                                       \
-    {                                                                       \
-        return fn##_body(dst, dstSize, cSrc, cSrcSize, DTable);             \
-    }                                                                       \
-                                                                            \
-    static size_t fn(void* dst, size_t dstSize, void const* cSrc,           \
-                     size_t cSrcSize, HUF_DTable const* DTable, int bmi2)   \
-    {                                                                       \
-        if (bmi2) {                                                         \
-            return fn##_bmi2(dst, dstSize, cSrc, cSrcSize, DTable);         \
-        }                                                                   \
-        return fn##_default(dst, dstSize, cSrc, cSrcSize, DTable);          \
-    }
-
-#else
-
-#define HUF_DGEN(fn)                                                               \
-    static size_t fn(void* dst, size_t dstSize, void const* cSrc,           \
-                     size_t cSrcSize, HUF_DTable const* DTable, int bmi2)   \
-    {                                                                       \
-        (void)bmi2;                                                         \
-        return fn##_body(dst, dstSize, cSrc, cSrcSize, DTable);             \
-    }
-
-#endif
 
 HUF_DGEN(HUF_decompress1X1_usingDTable_internal)
 HUF_DGEN(HUF_decompress4X1_usingDTable_internal)
@@ -437,6 +457,10 @@ size_t HUF_decompress4X1 (void* dst, size_t dstSize, const void* cSrc, size_t cS
     return HUF_decompress4X1_DCtx(DTable, dst, dstSize, cSrc, cSrcSize);
 }
 
+#endif /* HUF_FORCE_DECOMPRESS_X2 */
+
+
+#ifndef HUF_FORCE_DECOMPRESS_X1
 
 /* *************************/
 /* double-symbols decoding */
@@ -911,6 +935,8 @@ size_t HUF_decompress4X2 (void* dst, size_t dstSize, const void* cSrc, size_t cS
     return HUF_decompress4X2_DCtx(DTable, dst, dstSize, cSrc, cSrcSize);
 }
 
+#endif /* HUF_FORCE_DECOMPRESS_X1 */
+
 
 /* ***********************************/
 /* Universal decompression selectors */
@@ -921,8 +947,18 @@ size_t HUF_decompress1X_usingDTable(void* dst, size_t maxDstSize,
                                     const HUF_DTable* DTable)
 {
     DTableDesc const dtd = HUF_getDTableDesc(DTable);
+#if defined(HUF_FORCE_DECOMPRESS_X1)
+    (void)dtd;
+    assert(dtd.tableType == 0);
+    return HUF_decompress1X1_usingDTable_internal(dst, maxDstSize, cSrc, cSrcSize, DTable, /* bmi2 */ 0);
+#elif defined(HUF_FORCE_DECOMPRESS_X2)
+    (void)dtd;
+    assert(dtd.tableType == 1);
+    return HUF_decompress1X2_usingDTable_internal(dst, maxDstSize, cSrc, cSrcSize, DTable, /* bmi2 */ 0);
+#else
     return dtd.tableType ? HUF_decompress1X2_usingDTable_internal(dst, maxDstSize, cSrc, cSrcSize, DTable, /* bmi2 */ 0) :
                            HUF_decompress1X1_usingDTable_internal(dst, maxDstSize, cSrc, cSrcSize, DTable, /* bmi2 */ 0);
+#endif
 }
 
 size_t HUF_decompress4X_usingDTable(void* dst, size_t maxDstSize,
@@ -930,11 +966,22 @@ size_t HUF_decompress4X_usingDTable(void* dst, size_t maxDstSize,
                                     const HUF_DTable* DTable)
 {
     DTableDesc const dtd = HUF_getDTableDesc(DTable);
+#if defined(HUF_FORCE_DECOMPRESS_X1)
+    (void)dtd;
+    assert(dtd.tableType == 0);
+    return HUF_decompress4X1_usingDTable_internal(dst, maxDstSize, cSrc, cSrcSize, DTable, /* bmi2 */ 0);
+#elif defined(HUF_FORCE_DECOMPRESS_X2)
+    (void)dtd;
+    assert(dtd.tableType == 1);
+    return HUF_decompress4X2_usingDTable_internal(dst, maxDstSize, cSrc, cSrcSize, DTable, /* bmi2 */ 0);
+#else
     return dtd.tableType ? HUF_decompress4X2_usingDTable_internal(dst, maxDstSize, cSrc, cSrcSize, DTable, /* bmi2 */ 0) :
                            HUF_decompress4X1_usingDTable_internal(dst, maxDstSize, cSrc, cSrcSize, DTable, /* bmi2 */ 0);
+#endif
 }
 
 
+#if !defined(HUF_FORCE_DECOMPRESS_X1) && !defined(HUF_FORCE_DECOMPRESS_X2)
 typedef struct { U32 tableTime; U32 decode256Time; } algo_time_t;
 static const algo_time_t algoTime[16 /* Quantization */][3 /* single, double, quad */] =
 {
@@ -956,6 +1003,7 @@ static const algo_time_t algoTime[16 /* Quantization */][3 /* single, double, qu
     {{1455,128}, {2422,124}, {4174,124}},   /* Q ==14 : 87-93% */
     {{ 722,128}, {1891,145}, {1936,146}},   /* Q ==15 : 93-99% */
 };
+#endif
 
 /** HUF_selectDecoder() :
  *  Tells which decoder is likely to decode faster,
@@ -966,6 +1014,15 @@ U32 HUF_selectDecoder (size_t dstSize, size_t cSrcSize)
 {
     assert(dstSize > 0);
     assert(dstSize <= 128*1024);
+#if defined(HUF_FORCE_DECOMPRESS_X1)
+    (void)dstSize;
+    (void)cSrcSize;
+    return 0;
+#elif defined(HUF_FORCE_DECOMPRESS_X2)
+    (void)dstSize;
+    (void)cSrcSize;
+    return 1;
+#else
     /* decoder timing evaluation */
     {   U32 const Q = (cSrcSize >= dstSize) ? 15 : (U32)(cSrcSize * 16 / dstSize);   /* Q < 16 */
         U32 const D256 = (U32)(dstSize >> 8);
@@ -973,14 +1030,18 @@ U32 HUF_selectDecoder (size_t dstSize, size_t cSrcSize)
         U32 DTime1 = algoTime[Q][1].tableTime + (algoTime[Q][1].decode256Time * D256);
         DTime1 += DTime1 >> 3;  /* advantage to algorithm using less memory, to reduce cache eviction */
         return DTime1 < DTime0;
-}   }
+    }
+#endif
+}
 
 
 typedef size_t (*decompressionAlgo)(void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize);
 
 size_t HUF_decompress (void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize)
 {
+#if !defined(HUF_FORCE_DECOMPRESS_X1) && !defined(HUF_FORCE_DECOMPRESS_X2)
     static const decompressionAlgo decompress[2] = { HUF_decompress4X1, HUF_decompress4X2 };
+#endif
 
     /* validation checks */
     if (dstSize == 0) return ERROR(dstSize_tooSmall);
@@ -989,7 +1050,17 @@ size_t HUF_decompress (void* dst, size_t dstSize, const void* cSrc, size_t cSrcS
     if (cSrcSize == 1) { memset(dst, *(const BYTE*)cSrc, dstSize); return dstSize; }   /* RLE */
 
     {   U32 const algoNb = HUF_selectDecoder(dstSize, cSrcSize);
+#if defined(HUF_FORCE_DECOMPRESS_X1)
+        (void)algoNb;
+        assert(algoNb == 0);
+        return HUF_decompress4X1(dst, dstSize, cSrc, cSrcSize);
+#elif defined(HUF_FORCE_DECOMPRESS_X2)
+        (void)algoNb;
+        assert(algoNb == 1);
+        return HUF_decompress4X2(dst, dstSize, cSrc, cSrcSize);
+#else
         return decompress[algoNb](dst, dstSize, cSrc, cSrcSize);
+#endif
     }
 }
 
@@ -1002,8 +1073,18 @@ size_t HUF_decompress4X_DCtx (HUF_DTable* dctx, void* dst, size_t dstSize, const
     if (cSrcSize == 1) { memset(dst, *(const BYTE*)cSrc, dstSize); return dstSize; }   /* RLE */
 
     {   U32 const algoNb = HUF_selectDecoder(dstSize, cSrcSize);
+#if defined(HUF_FORCE_DECOMPRESS_X1)
+        (void)algoNb;
+        assert(algoNb == 0);
+        return HUF_decompress4X1_DCtx(dctx, dst, dstSize, cSrc, cSrcSize);
+#elif defined(HUF_FORCE_DECOMPRESS_X2)
+        (void)algoNb;
+        assert(algoNb == 1);
+        return HUF_decompress4X2_DCtx(dctx, dst, dstSize, cSrc, cSrcSize);
+#else
         return algoNb ? HUF_decompress4X2_DCtx(dctx, dst, dstSize, cSrc, cSrcSize) :
                         HUF_decompress4X1_DCtx(dctx, dst, dstSize, cSrc, cSrcSize) ;
+#endif
     }
 }
 
@@ -1025,8 +1106,19 @@ size_t HUF_decompress4X_hufOnly_wksp(HUF_DTable* dctx, void* dst,
     if (cSrcSize == 0) return ERROR(corruption_detected);
 
     {   U32 const algoNb = HUF_selectDecoder(dstSize, cSrcSize);
-        return algoNb ? HUF_decompress4X2_DCtx_wksp(dctx, dst, dstSize, cSrc, cSrcSize, workSpace, wkspSize):
+#if defined(HUF_FORCE_DECOMPRESS_X1)
+        (void)algoNb;
+        assert(algoNb == 0);
+        return HUF_decompress4X1_DCtx_wksp(dctx, dst, dstSize, cSrc, cSrcSize, workSpace, wkspSize);
+#elif defined(HUF_FORCE_DECOMPRESS_X2)
+        (void)algoNb;
+        assert(algoNb == 1);
+        return HUF_decompress4X2_DCtx_wksp(dctx, dst, dstSize, cSrc, cSrcSize, workSpace, wkspSize);
+#else
+        return algoNb ? HUF_decompress4X2_DCtx_wksp(dctx, dst, dstSize, cSrc,
+                            cSrcSize, workSpace, wkspSize):
                         HUF_decompress4X1_DCtx_wksp(dctx, dst, dstSize, cSrc, cSrcSize, workSpace, wkspSize);
+#endif
     }
 }
 
@@ -1041,10 +1133,22 @@ size_t HUF_decompress1X_DCtx_wksp(HUF_DTable* dctx, void* dst, size_t dstSize,
     if (cSrcSize == 1) { memset(dst, *(const BYTE*)cSrc, dstSize); return dstSize; }   /* RLE */
 
     {   U32 const algoNb = HUF_selectDecoder(dstSize, cSrcSize);
+#if defined(HUF_FORCE_DECOMPRESS_X1)
+        (void)algoNb;
+        assert(algoNb == 0);
+        return HUF_decompress1X1_DCtx_wksp(dctx, dst, dstSize, cSrc,
+                                cSrcSize, workSpace, wkspSize);
+#elif defined(HUF_FORCE_DECOMPRESS_X2)
+        (void)algoNb;
+        assert(algoNb == 1);
+        return HUF_decompress1X2_DCtx_wksp(dctx, dst, dstSize, cSrc,
+                                cSrcSize, workSpace, wkspSize);
+#else
         return algoNb ? HUF_decompress1X2_DCtx_wksp(dctx, dst, dstSize, cSrc,
                                 cSrcSize, workSpace, wkspSize):
                         HUF_decompress1X1_DCtx_wksp(dctx, dst, dstSize, cSrc,
                                 cSrcSize, workSpace, wkspSize);
+#endif
     }
 }
 
@@ -1060,10 +1164,21 @@ size_t HUF_decompress1X_DCtx(HUF_DTable* dctx, void* dst, size_t dstSize,
 size_t HUF_decompress1X_usingDTable_bmi2(void* dst, size_t maxDstSize, const void* cSrc, size_t cSrcSize, const HUF_DTable* DTable, int bmi2)
 {
     DTableDesc const dtd = HUF_getDTableDesc(DTable);
+#if defined(HUF_FORCE_DECOMPRESS_X1)
+    (void)dtd;
+    assert(dtd.tableType == 0);
+    return HUF_decompress1X1_usingDTable_internal(dst, maxDstSize, cSrc, cSrcSize, DTable, bmi2);
+#elif defined(HUF_FORCE_DECOMPRESS_X2)
+    (void)dtd;
+    assert(dtd.tableType == 1);
+    return HUF_decompress1X2_usingDTable_internal(dst, maxDstSize, cSrc, cSrcSize, DTable, bmi2);
+#else
     return dtd.tableType ? HUF_decompress1X2_usingDTable_internal(dst, maxDstSize, cSrc, cSrcSize, DTable, bmi2) :
                            HUF_decompress1X1_usingDTable_internal(dst, maxDstSize, cSrc, cSrcSize, DTable, bmi2);
+#endif
 }
 
+#ifndef HUF_FORCE_DECOMPRESS_X2
 size_t HUF_decompress1X1_DCtx_wksp_bmi2(HUF_DTable* dctx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize, void* workSpace, size_t wkspSize, int bmi2)
 {
     const BYTE* ip = (const BYTE*) cSrc;
@@ -1075,12 +1190,23 @@ size_t HUF_decompress1X1_DCtx_wksp_bmi2(HUF_DTable* dctx, void* dst, size_t dstS
 
     return HUF_decompress1X1_usingDTable_internal(dst, dstSize, ip, cSrcSize, dctx, bmi2);
 }
+#endif
 
 size_t HUF_decompress4X_usingDTable_bmi2(void* dst, size_t maxDstSize, const void* cSrc, size_t cSrcSize, const HUF_DTable* DTable, int bmi2)
 {
     DTableDesc const dtd = HUF_getDTableDesc(DTable);
+#if defined(HUF_FORCE_DECOMPRESS_X1)
+    (void)dtd;
+    assert(dtd.tableType == 0);
+    return HUF_decompress4X1_usingDTable_internal(dst, maxDstSize, cSrc, cSrcSize, DTable, bmi2);
+#elif defined(HUF_FORCE_DECOMPRESS_X2)
+    (void)dtd;
+    assert(dtd.tableType == 1);
+    return HUF_decompress4X2_usingDTable_internal(dst, maxDstSize, cSrc, cSrcSize, DTable, bmi2);
+#else
     return dtd.tableType ? HUF_decompress4X2_usingDTable_internal(dst, maxDstSize, cSrc, cSrcSize, DTable, bmi2) :
                            HUF_decompress4X1_usingDTable_internal(dst, maxDstSize, cSrc, cSrcSize, DTable, bmi2);
+#endif
 }
 
 size_t HUF_decompress4X_hufOnly_wksp_bmi2(HUF_DTable* dctx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize, void* workSpace, size_t wkspSize, int bmi2)
@@ -1090,7 +1216,17 @@ size_t HUF_decompress4X_hufOnly_wksp_bmi2(HUF_DTable* dctx, void* dst, size_t ds
     if (cSrcSize == 0) return ERROR(corruption_detected);
 
     {   U32 const algoNb = HUF_selectDecoder(dstSize, cSrcSize);
+#if defined(HUF_FORCE_DECOMPRESS_X1)
+        (void)algoNb;
+        assert(algoNb == 0);
+        return HUF_decompress4X1_DCtx_wksp_bmi2(dctx, dst, dstSize, cSrc, cSrcSize, workSpace, wkspSize, bmi2);
+#elif defined(HUF_FORCE_DECOMPRESS_X2)
+        (void)algoNb;
+        assert(algoNb == 1);
+        return HUF_decompress4X2_DCtx_wksp_bmi2(dctx, dst, dstSize, cSrc, cSrcSize, workSpace, wkspSize, bmi2);
+#else
         return algoNb ? HUF_decompress4X2_DCtx_wksp_bmi2(dctx, dst, dstSize, cSrc, cSrcSize, workSpace, wkspSize, bmi2) :
                         HUF_decompress4X1_DCtx_wksp_bmi2(dctx, dst, dstSize, cSrc, cSrcSize, workSpace, wkspSize, bmi2);
+#endif
     }
 }
diff --git a/lib/decompress/zstd_ddict.c b/lib/decompress/zstd_ddict.c
new file mode 100644
index 000000000000..2ad0440684ca
--- /dev/null
+++ b/lib/decompress/zstd_ddict.c
@@ -0,0 +1,240 @@
+/*
+ * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
+ * You may select, at your option, one of the above-listed licenses.
+ */
+
+/* zstd_ddict.c :
+ * concentrates all logic that needs to know the internals of ZSTD_DDict object */
+
+/*-*******************************************************
+*  Dependencies
+*********************************************************/
+#include       /* memcpy, memmove, memset */
+#include "cpu.h"         /* bmi2 */
+#include "mem.h"         /* low level memory routines */
+#define FSE_STATIC_LINKING_ONLY
+#include "fse.h"
+#define HUF_STATIC_LINKING_ONLY
+#include "huf.h"
+#include "zstd_decompress_internal.h"
+#include "zstd_ddict.h"
+
+#if defined(ZSTD_LEGACY_SUPPORT) && (ZSTD_LEGACY_SUPPORT>=1)
+#  include "zstd_legacy.h"
+#endif
+
+
+
+/*-*******************************************************
+*  Types
+*********************************************************/
+struct ZSTD_DDict_s {
+    void* dictBuffer;
+    const void* dictContent;
+    size_t dictSize;
+    ZSTD_entropyDTables_t entropy;
+    U32 dictID;
+    U32 entropyPresent;
+    ZSTD_customMem cMem;
+};  /* typedef'd to ZSTD_DDict within "zstd.h" */
+
+const void* ZSTD_DDict_dictContent(const ZSTD_DDict* ddict)
+{
+    assert(ddict != NULL);
+    return ddict->dictContent;
+}
+
+size_t ZSTD_DDict_dictSize(const ZSTD_DDict* ddict)
+{
+    assert(ddict != NULL);
+    return ddict->dictSize;
+}
+
+void ZSTD_copyDDictParameters(ZSTD_DCtx* dctx, const ZSTD_DDict* ddict)
+{
+    DEBUGLOG(4, "ZSTD_copyDDictParameters");
+    assert(dctx != NULL);
+    assert(ddict != NULL);
+    dctx->dictID = ddict->dictID;
+    dctx->prefixStart = ddict->dictContent;
+    dctx->virtualStart = ddict->dictContent;
+    dctx->dictEnd = (const BYTE*)ddict->dictContent + ddict->dictSize;
+    dctx->previousDstEnd = dctx->dictEnd;
+    if (ddict->entropyPresent) {
+        dctx->litEntropy = 1;
+        dctx->fseEntropy = 1;
+        dctx->LLTptr = ddict->entropy.LLTable;
+        dctx->MLTptr = ddict->entropy.MLTable;
+        dctx->OFTptr = ddict->entropy.OFTable;
+        dctx->HUFptr = ddict->entropy.hufTable;
+        dctx->entropy.rep[0] = ddict->entropy.rep[0];
+        dctx->entropy.rep[1] = ddict->entropy.rep[1];
+        dctx->entropy.rep[2] = ddict->entropy.rep[2];
+    } else {
+        dctx->litEntropy = 0;
+        dctx->fseEntropy = 0;
+    }
+}
+
+
+static size_t
+ZSTD_loadEntropy_intoDDict(ZSTD_DDict* ddict,
+                           ZSTD_dictContentType_e dictContentType)
+{
+    ddict->dictID = 0;
+    ddict->entropyPresent = 0;
+    if (dictContentType == ZSTD_dct_rawContent) return 0;
+
+    if (ddict->dictSize < 8) {
+        if (dictContentType == ZSTD_dct_fullDict)
+            return ERROR(dictionary_corrupted);   /* only accept specified dictionaries */
+        return 0;   /* pure content mode */
+    }
+    {   U32 const magic = MEM_readLE32(ddict->dictContent);
+        if (magic != ZSTD_MAGIC_DICTIONARY) {
+            if (dictContentType == ZSTD_dct_fullDict)
+                return ERROR(dictionary_corrupted);   /* only accept specified dictionaries */
+            return 0;   /* pure content mode */
+        }
+    }
+    ddict->dictID = MEM_readLE32((const char*)ddict->dictContent + ZSTD_FRAMEIDSIZE);
+
+    /* load entropy tables */
+    CHECK_E( ZSTD_loadDEntropy(&ddict->entropy,
+                                ddict->dictContent, ddict->dictSize),
+             dictionary_corrupted );
+    ddict->entropyPresent = 1;
+    return 0;
+}
+
+
+static size_t ZSTD_initDDict_internal(ZSTD_DDict* ddict,
+                                      const void* dict, size_t dictSize,
+                                      ZSTD_dictLoadMethod_e dictLoadMethod,
+                                      ZSTD_dictContentType_e dictContentType)
+{
+    if ((dictLoadMethod == ZSTD_dlm_byRef) || (!dict) || (!dictSize)) {
+        ddict->dictBuffer = NULL;
+        ddict->dictContent = dict;
+        if (!dict) dictSize = 0;
+    } else {
+        void* const internalBuffer = ZSTD_malloc(dictSize, ddict->cMem);
+        ddict->dictBuffer = internalBuffer;
+        ddict->dictContent = internalBuffer;
+        if (!internalBuffer) return ERROR(memory_allocation);
+        memcpy(internalBuffer, dict, dictSize);
+    }
+    ddict->dictSize = dictSize;
+    ddict->entropy.hufTable[0] = (HUF_DTable)((HufLog)*0x1000001);  /* cover both little and big endian */
+
+    /* parse dictionary content */
+    CHECK_F( ZSTD_loadEntropy_intoDDict(ddict, dictContentType) );
+
+    return 0;
+}
+
+ZSTD_DDict* ZSTD_createDDict_advanced(const void* dict, size_t dictSize,
+                                      ZSTD_dictLoadMethod_e dictLoadMethod,
+                                      ZSTD_dictContentType_e dictContentType,
+                                      ZSTD_customMem customMem)
+{
+    if (!customMem.customAlloc ^ !customMem.customFree) return NULL;
+
+    {   ZSTD_DDict* const ddict = (ZSTD_DDict*) ZSTD_malloc(sizeof(ZSTD_DDict), customMem);
+        if (ddict == NULL) return NULL;
+        ddict->cMem = customMem;
+        {   size_t const initResult = ZSTD_initDDict_internal(ddict,
+                                            dict, dictSize,
+                                            dictLoadMethod, dictContentType);
+            if (ZSTD_isError(initResult)) {
+                ZSTD_freeDDict(ddict);
+                return NULL;
+        }   }
+        return ddict;
+    }
+}
+
+/*! ZSTD_createDDict() :
+*   Create a digested dictionary, to start decompression without startup delay.
+*   `dict` content is copied inside DDict.
+*   Consequently, `dict` can be released after `ZSTD_DDict` creation */
+ZSTD_DDict* ZSTD_createDDict(const void* dict, size_t dictSize)
+{
+    ZSTD_customMem const allocator = { NULL, NULL, NULL };
+    return ZSTD_createDDict_advanced(dict, dictSize, ZSTD_dlm_byCopy, ZSTD_dct_auto, allocator);
+}
+
+/*! ZSTD_createDDict_byReference() :
+ *  Create a digested dictionary, to start decompression without startup delay.
+ *  Dictionary content is simply referenced, it will be accessed during decompression.
+ *  Warning : dictBuffer must outlive DDict (DDict must be freed before dictBuffer) */
+ZSTD_DDict* ZSTD_createDDict_byReference(const void* dictBuffer, size_t dictSize)
+{
+    ZSTD_customMem const allocator = { NULL, NULL, NULL };
+    return ZSTD_createDDict_advanced(dictBuffer, dictSize, ZSTD_dlm_byRef, ZSTD_dct_auto, allocator);
+}
+
+
+const ZSTD_DDict* ZSTD_initStaticDDict(
+                                void* sBuffer, size_t sBufferSize,
+                                const void* dict, size_t dictSize,
+                                ZSTD_dictLoadMethod_e dictLoadMethod,
+                                ZSTD_dictContentType_e dictContentType)
+{
+    size_t const neededSpace = sizeof(ZSTD_DDict)
+                             + (dictLoadMethod == ZSTD_dlm_byRef ? 0 : dictSize);
+    ZSTD_DDict* const ddict = (ZSTD_DDict*)sBuffer;
+    assert(sBuffer != NULL);
+    assert(dict != NULL);
+    if ((size_t)sBuffer & 7) return NULL;   /* 8-aligned */
+    if (sBufferSize < neededSpace) return NULL;
+    if (dictLoadMethod == ZSTD_dlm_byCopy) {
+        memcpy(ddict+1, dict, dictSize);  /* local copy */
+        dict = ddict+1;
+    }
+    if (ZSTD_isError( ZSTD_initDDict_internal(ddict,
+                                              dict, dictSize,
+                                              ZSTD_dlm_byRef, dictContentType) ))
+        return NULL;
+    return ddict;
+}
+
+
+size_t ZSTD_freeDDict(ZSTD_DDict* ddict)
+{
+    if (ddict==NULL) return 0;   /* support free on NULL */
+    {   ZSTD_customMem const cMem = ddict->cMem;
+        ZSTD_free(ddict->dictBuffer, cMem);
+        ZSTD_free(ddict, cMem);
+        return 0;
+    }
+}
+
+/*! ZSTD_estimateDDictSize() :
+ *  Estimate amount of memory that will be needed to create a dictionary for decompression.
+ *  Note : dictionary created by reference using ZSTD_dlm_byRef are smaller */
+size_t ZSTD_estimateDDictSize(size_t dictSize, ZSTD_dictLoadMethod_e dictLoadMethod)
+{
+    return sizeof(ZSTD_DDict) + (dictLoadMethod == ZSTD_dlm_byRef ? 0 : dictSize);
+}
+
+size_t ZSTD_sizeof_DDict(const ZSTD_DDict* ddict)
+{
+    if (ddict==NULL) return 0;   /* support sizeof on NULL */
+    return sizeof(*ddict) + (ddict->dictBuffer ? ddict->dictSize : 0) ;
+}
+
+/*! ZSTD_getDictID_fromDDict() :
+ *  Provides the dictID of the dictionary loaded into `ddict`.
+ *  If @return == 0, the dictionary is not conformant to Zstandard specification, or empty.
+ *  Non-conformant dictionaries can still be loaded, but as content-only dictionaries. */
+unsigned ZSTD_getDictID_fromDDict(const ZSTD_DDict* ddict)
+{
+    if (ddict==NULL) return 0;
+    return ZSTD_getDictID_fromDict(ddict->dictContent, ddict->dictSize);
+}
diff --git a/lib/decompress/zstd_ddict.h b/lib/decompress/zstd_ddict.h
new file mode 100644
index 000000000000..0479d11bb032
--- /dev/null
+++ b/lib/decompress/zstd_ddict.h
@@ -0,0 +1,44 @@
+/*
+ * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
+ * You may select, at your option, one of the above-listed licenses.
+ */
+
+
+#ifndef ZSTD_DDICT_H
+#define ZSTD_DDICT_H
+
+/*-*******************************************************
+ *  Dependencies
+ *********************************************************/
+#include    /* size_t */
+#include "zstd.h"     /* ZSTD_DDict, and several public functions */
+
+
+/*-*******************************************************
+ *  Interface
+ *********************************************************/
+
+/* note: several prototypes are already published in `zstd.h` :
+ * ZSTD_createDDict()
+ * ZSTD_createDDict_byReference()
+ * ZSTD_createDDict_advanced()
+ * ZSTD_freeDDict()
+ * ZSTD_initStaticDDict()
+ * ZSTD_sizeof_DDict()
+ * ZSTD_estimateDDictSize()
+ * ZSTD_getDictID_fromDict()
+ */
+
+const void* ZSTD_DDict_dictContent(const ZSTD_DDict* ddict);
+size_t ZSTD_DDict_dictSize(const ZSTD_DDict* ddict);
+
+void ZSTD_copyDDictParameters(ZSTD_DCtx* dctx, const ZSTD_DDict* ddict);
+
+
+
+#endif /* ZSTD_DDICT_H */
diff --git a/lib/decompress/zstd_decompress.c b/lib/decompress/zstd_decompress.c
index 711b5b6d7aca..feef1ef67abd 100644
--- a/lib/decompress/zstd_decompress.c
+++ b/lib/decompress/zstd_decompress.c
@@ -37,12 +37,12 @@
  *  It's possible to set a different limit using ZSTD_DCtx_setMaxWindowSize().
  */
 #ifndef ZSTD_MAXWINDOWSIZE_DEFAULT
-#  define ZSTD_MAXWINDOWSIZE_DEFAULT (((U32)1 << ZSTD_WINDOWLOG_DEFAULTMAX) + 1)
+#  define ZSTD_MAXWINDOWSIZE_DEFAULT (((U32)1 << ZSTD_WINDOWLOG_LIMIT_DEFAULT) + 1)
 #endif
 
 /*!
  *  NO_FORWARD_PROGRESS_MAX :
- *  maximum allowed nb of calls to ZSTD_decompressStream() and ZSTD_decompress_generic()
+ *  maximum allowed nb of calls to ZSTD_decompressStream()
  *  without any forward progress
  *  (defined as: no byte read from input, and no byte flushed to output)
  *  before triggering an error.
@@ -56,128 +56,25 @@
 *  Dependencies
 *********************************************************/
 #include       /* memcpy, memmove, memset */
-#include "compiler.h"    /* prefetch */
 #include "cpu.h"         /* bmi2 */
 #include "mem.h"         /* low level memory routines */
 #define FSE_STATIC_LINKING_ONLY
 #include "fse.h"
 #define HUF_STATIC_LINKING_ONLY
 #include "huf.h"
-#include "zstd_internal.h"
+#include "zstd_internal.h"  /* blockProperties_t */
+#include "zstd_decompress_internal.h"   /* ZSTD_DCtx */
+#include "zstd_ddict.h"  /* ZSTD_DDictDictContent */
+#include "zstd_decompress_block.h"   /* ZSTD_decompressBlock_internal */
 
 #if defined(ZSTD_LEGACY_SUPPORT) && (ZSTD_LEGACY_SUPPORT>=1)
 #  include "zstd_legacy.h"
 #endif
 
-static const void* ZSTD_DDictDictContent(const ZSTD_DDict* ddict);
-static size_t ZSTD_DDictDictSize(const ZSTD_DDict* ddict);
-
-
-/*-*************************************
-*  Errors
-***************************************/
-#define ZSTD_isError ERR_isError   /* for inlining */
-#define FSE_isError  ERR_isError
-#define HUF_isError  ERR_isError
-
-
-/*_*******************************************************
-*  Memory operations
-**********************************************************/
-static void ZSTD_copy4(void* dst, const void* src) { memcpy(dst, src, 4); }
-
 
 /*-*************************************************************
 *   Context management
 ***************************************************************/
-typedef enum { ZSTDds_getFrameHeaderSize, ZSTDds_decodeFrameHeader,
-               ZSTDds_decodeBlockHeader, ZSTDds_decompressBlock,
-               ZSTDds_decompressLastBlock, ZSTDds_checkChecksum,
-               ZSTDds_decodeSkippableHeader, ZSTDds_skipFrame } ZSTD_dStage;
-
-typedef enum { zdss_init=0, zdss_loadHeader,
-               zdss_read, zdss_load, zdss_flush } ZSTD_dStreamStage;
-
-
-typedef struct {
-    U32 fastMode;
-    U32 tableLog;
-} ZSTD_seqSymbol_header;
-
-typedef struct {
-    U16  nextState;
-    BYTE nbAdditionalBits;
-    BYTE nbBits;
-    U32  baseValue;
-} ZSTD_seqSymbol;
-
-#define SEQSYMBOL_TABLE_SIZE(log)   (1 + (1 << (log)))
-
-typedef struct {
-    ZSTD_seqSymbol LLTable[SEQSYMBOL_TABLE_SIZE(LLFSELog)];    /* Note : Space reserved for FSE Tables */
-    ZSTD_seqSymbol OFTable[SEQSYMBOL_TABLE_SIZE(OffFSELog)];   /* is also used as temporary workspace while building hufTable during DDict creation */
-    ZSTD_seqSymbol MLTable[SEQSYMBOL_TABLE_SIZE(MLFSELog)];    /* and therefore must be at least HUF_DECOMPRESS_WORKSPACE_SIZE large */
-    HUF_DTable hufTable[HUF_DTABLE_SIZE(HufLog)];  /* can accommodate HUF_decompress4X */
-    U32 rep[ZSTD_REP_NUM];
-} ZSTD_entropyDTables_t;
-
-struct ZSTD_DCtx_s
-{
-    const ZSTD_seqSymbol* LLTptr;
-    const ZSTD_seqSymbol* MLTptr;
-    const ZSTD_seqSymbol* OFTptr;
-    const HUF_DTable* HUFptr;
-    ZSTD_entropyDTables_t entropy;
-    U32 workspace[HUF_DECOMPRESS_WORKSPACE_SIZE_U32];   /* space needed when building huffman tables */
-    const void* previousDstEnd;   /* detect continuity */
-    const void* prefixStart;      /* start of current segment */
-    const void* virtualStart;     /* virtual start of previous segment if it was just before current one */
-    const void* dictEnd;          /* end of previous segment */
-    size_t expected;
-    ZSTD_frameHeader fParams;
-    U64 decodedSize;
-    blockType_e bType;            /* used in ZSTD_decompressContinue(), store blockType between block header decoding and block decompression stages */
-    ZSTD_dStage stage;
-    U32 litEntropy;
-    U32 fseEntropy;
-    XXH64_state_t xxhState;
-    size_t headerSize;
-    ZSTD_format_e format;
-    const BYTE* litPtr;
-    ZSTD_customMem customMem;
-    size_t litSize;
-    size_t rleSize;
-    size_t staticSize;
-    int bmi2;                     /* == 1 if the CPU supports BMI2 and 0 otherwise. CPU support is determined dynamically once per context lifetime. */
-
-    /* dictionary */
-    ZSTD_DDict* ddictLocal;
-    const ZSTD_DDict* ddict;     /* set by ZSTD_initDStream_usingDDict(), or ZSTD_DCtx_refDDict() */
-    U32 dictID;
-    int ddictIsCold;             /* if == 1 : dictionary is "new" for working context, and presumed "cold" (not in cpu cache) */
-
-    /* streaming */
-    ZSTD_dStreamStage streamStage;
-    char*  inBuff;
-    size_t inBuffSize;
-    size_t inPos;
-    size_t maxWindowSize;
-    char*  outBuff;
-    size_t outBuffSize;
-    size_t outStart;
-    size_t outEnd;
-    size_t lhSize;
-    void* legacyContext;
-    U32 previousLegacyVersion;
-    U32 legacyVersion;
-    U32 hostageByte;
-    int noForwardProgress;
-
-    /* workspace */
-    BYTE litBuffer[ZSTD_BLOCKSIZE_MAX + WILDCOPY_OVERLENGTH];
-    BYTE headerBuffer[ZSTD_FRAMEHEADERSIZE_MAX];
-};  /* typedef'd to ZSTD_DCtx within "zstd.h" */
-
 size_t ZSTD_sizeof_DCtx (const ZSTD_DCtx* dctx)
 {
     if (dctx==NULL) return 0;   /* support sizeof NULL */
@@ -192,8 +89,8 @@ size_t ZSTD_estimateDCtxSize(void) { return sizeof(ZSTD_DCtx); }
 static size_t ZSTD_startingInputLength(ZSTD_format_e format)
 {
     size_t const startingInputLength = (format==ZSTD_f_zstd1_magicless) ?
-                    ZSTD_frameHeaderSize_prefix - ZSTD_FRAMEIDSIZE :
-                    ZSTD_frameHeaderSize_prefix;
+                    ZSTD_FRAMEHEADERSIZE_PREFIX - ZSTD_FRAMEIDSIZE :
+                    ZSTD_FRAMEHEADERSIZE_PREFIX;
     ZSTD_STATIC_ASSERT(ZSTD_FRAMEHEADERSIZE_PREFIX >= ZSTD_FRAMEIDSIZE);
     /* only supports formats ZSTD_f_zstd1 and ZSTD_f_zstd1_magicless */
     assert( (format == ZSTD_f_zstd1) || (format == ZSTD_f_zstd1_magicless) );
@@ -290,7 +187,7 @@ unsigned ZSTD_isFrame(const void* buffer, size_t size)
     if (size < ZSTD_FRAMEIDSIZE) return 0;
     {   U32 const magic = MEM_readLE32(buffer);
         if (magic == ZSTD_MAGICNUMBER) return 1;
-        if ((magic & 0xFFFFFFF0U) == ZSTD_MAGIC_SKIPPABLE_START) return 1;
+        if ((magic & ZSTD_MAGIC_SKIPPABLE_MASK) == ZSTD_MAGIC_SKIPPABLE_START) return 1;
     }
 #if defined(ZSTD_LEGACY_SUPPORT) && (ZSTD_LEGACY_SUPPORT >= 1)
     if (ZSTD_isLegacy(buffer, size)) return 1;
@@ -345,10 +242,10 @@ size_t ZSTD_getFrameHeader_advanced(ZSTD_frameHeader* zfhPtr, const void* src, s
 
     if ( (format != ZSTD_f_zstd1_magicless)
       && (MEM_readLE32(src) != ZSTD_MAGICNUMBER) ) {
-        if ((MEM_readLE32(src) & 0xFFFFFFF0U) == ZSTD_MAGIC_SKIPPABLE_START) {
+        if ((MEM_readLE32(src) & ZSTD_MAGIC_SKIPPABLE_MASK) == ZSTD_MAGIC_SKIPPABLE_START) {
             /* skippable frame */
-            if (srcSize < ZSTD_skippableHeaderSize)
-                return ZSTD_skippableHeaderSize; /* magic number + frame length */
+            if (srcSize < ZSTD_SKIPPABLEHEADERSIZE)
+                return ZSTD_SKIPPABLEHEADERSIZE; /* magic number + frame length */
             memset(zfhPtr, 0, sizeof(*zfhPtr));
             zfhPtr->frameContentSize = MEM_readLE32((const char *)src + ZSTD_FRAMEIDSIZE);
             zfhPtr->frameType = ZSTD_skippableFrame;
@@ -446,6 +343,21 @@ unsigned long long ZSTD_getFrameContentSize(const void *src, size_t srcSize)
     }   }
 }
 
+static size_t readSkippableFrameSize(void const* src, size_t srcSize)
+{
+    size_t const skippableHeaderSize = ZSTD_SKIPPABLEHEADERSIZE;
+    U32 sizeU32;
+
+    if (srcSize < ZSTD_SKIPPABLEHEADERSIZE)
+        return ERROR(srcSize_wrong);
+
+    sizeU32 = MEM_readLE32((BYTE const*)src + ZSTD_FRAMEIDSIZE);
+    if ((U32)(sizeU32 + ZSTD_SKIPPABLEHEADERSIZE) < sizeU32)
+        return ERROR(frameParameter_unsupported);
+
+    return skippableHeaderSize + sizeU32;
+}
+
 /** ZSTD_findDecompressedSize() :
  *  compatible with legacy mode
  *  `srcSize` must be the exact length of some number of ZSTD compressed and/or
@@ -455,15 +367,13 @@ unsigned long long ZSTD_findDecompressedSize(const void* src, size_t srcSize)
 {
     unsigned long long totalDstSize = 0;
 
-    while (srcSize >= ZSTD_frameHeaderSize_prefix) {
+    while (srcSize >= ZSTD_FRAMEHEADERSIZE_PREFIX) {
         U32 const magicNumber = MEM_readLE32(src);
 
-        if ((magicNumber & 0xFFFFFFF0U) == ZSTD_MAGIC_SKIPPABLE_START) {
-            size_t skippableSize;
-            if (srcSize < ZSTD_skippableHeaderSize)
-                return ERROR(srcSize_wrong);
-            skippableSize = MEM_readLE32((const BYTE *)src + ZSTD_FRAMEIDSIZE)
-                          + ZSTD_skippableHeaderSize;
+        if ((magicNumber & ZSTD_MAGIC_SKIPPABLE_MASK) == ZSTD_MAGIC_SKIPPABLE_START) {
+            size_t const skippableSize = readSkippableFrameSize(src, srcSize);
+            if (ZSTD_isError(skippableSize))
+                return skippableSize;
             if (srcSize < skippableSize) {
                 return ZSTD_CONTENTSIZE_ERROR;
             }
@@ -496,9 +406,9 @@ unsigned long long ZSTD_findDecompressedSize(const void* src, size_t srcSize)
 }
 
 /** ZSTD_getDecompressedSize() :
-*   compatible with legacy mode
-*   @return : decompressed size if known, 0 otherwise
-              note : 0 can mean any of the following :
+ *  compatible with legacy mode
+ * @return : decompressed size if known, 0 otherwise
+             note : 0 can mean any of the following :
                    - frame content is empty
                    - decompressed size field is not present in frame header
                    - frame header unknown / not supported
@@ -512,8 +422,8 @@ unsigned long long ZSTD_getDecompressedSize(const void* src, size_t srcSize)
 
 
 /** ZSTD_decodeFrameHeader() :
-*   `headerSize` must be the size provided by ZSTD_frameHeaderSize().
-*   @return : 0 if success, or an error code, which can be tested using ZSTD_isError() */
+ * `headerSize` must be the size provided by ZSTD_frameHeaderSize().
+ * @return : 0 if success, or an error code, which can be tested using ZSTD_isError() */
 static size_t ZSTD_decodeFrameHeader(ZSTD_DCtx* dctx, const void* src, size_t headerSize)
 {
     size_t const result = ZSTD_getFrameHeader_advanced(&(dctx->fParams), src, headerSize, dctx->format);
@@ -526,1275 +436,6 @@ static size_t ZSTD_decodeFrameHeader(ZSTD_DCtx* dctx, const void* src, size_t he
 }
 
 
-/*-*************************************************************
- *   Block decoding
- ***************************************************************/
-
-/*! ZSTD_getcBlockSize() :
-*   Provides the size of compressed block from block header `src` */
-size_t ZSTD_getcBlockSize(const void* src, size_t srcSize,
-                          blockProperties_t* bpPtr)
-{
-    if (srcSize < ZSTD_blockHeaderSize) return ERROR(srcSize_wrong);
-    {   U32 const cBlockHeader = MEM_readLE24(src);
-        U32 const cSize = cBlockHeader >> 3;
-        bpPtr->lastBlock = cBlockHeader & 1;
-        bpPtr->blockType = (blockType_e)((cBlockHeader >> 1) & 3);
-        bpPtr->origSize = cSize;   /* only useful for RLE */
-        if (bpPtr->blockType == bt_rle) return 1;
-        if (bpPtr->blockType == bt_reserved) return ERROR(corruption_detected);
-        return cSize;
-    }
-}
-
-
-static size_t ZSTD_copyRawBlock(void* dst, size_t dstCapacity,
-                          const void* src, size_t srcSize)
-{
-    if (dst==NULL) return ERROR(dstSize_tooSmall);
-    if (srcSize > dstCapacity) return ERROR(dstSize_tooSmall);
-    memcpy(dst, src, srcSize);
-    return srcSize;
-}
-
-
-static size_t ZSTD_setRleBlock(void* dst, size_t dstCapacity,
-                         const void* src, size_t srcSize,
-                               size_t regenSize)
-{
-    if (srcSize != 1) return ERROR(srcSize_wrong);
-    if (regenSize > dstCapacity) return ERROR(dstSize_tooSmall);
-    memset(dst, *(const BYTE*)src, regenSize);
-    return regenSize;
-}
-
-/* Hidden declaration for fullbench */
-size_t ZSTD_decodeLiteralsBlock(ZSTD_DCtx* dctx,
-                          const void* src, size_t srcSize);
-/*! ZSTD_decodeLiteralsBlock() :
- * @return : nb of bytes read from src (< srcSize )
- *  note : symbol not declared but exposed for fullbench */
-size_t ZSTD_decodeLiteralsBlock(ZSTD_DCtx* dctx,
-                          const void* src, size_t srcSize)   /* note : srcSize < BLOCKSIZE */
-{
-    if (srcSize < MIN_CBLOCK_SIZE) return ERROR(corruption_detected);
-
-    {   const BYTE* const istart = (const BYTE*) src;
-        symbolEncodingType_e const litEncType = (symbolEncodingType_e)(istart[0] & 3);
-
-        switch(litEncType)
-        {
-        case set_repeat:
-            if (dctx->litEntropy==0) return ERROR(dictionary_corrupted);
-            /* fall-through */
-
-        case set_compressed:
-            if (srcSize < 5) return ERROR(corruption_detected);   /* srcSize >= MIN_CBLOCK_SIZE == 3; here we need up to 5 for case 3 */
-            {   size_t lhSize, litSize, litCSize;
-                U32 singleStream=0;
-                U32 const lhlCode = (istart[0] >> 2) & 3;
-                U32 const lhc = MEM_readLE32(istart);
-                switch(lhlCode)
-                {
-                case 0: case 1: default:   /* note : default is impossible, since lhlCode into [0..3] */
-                    /* 2 - 2 - 10 - 10 */
-                    singleStream = !lhlCode;
-                    lhSize = 3;
-                    litSize  = (lhc >> 4) & 0x3FF;
-                    litCSize = (lhc >> 14) & 0x3FF;
-                    break;
-                case 2:
-                    /* 2 - 2 - 14 - 14 */
-                    lhSize = 4;
-                    litSize  = (lhc >> 4) & 0x3FFF;
-                    litCSize = lhc >> 18;
-                    break;
-                case 3:
-                    /* 2 - 2 - 18 - 18 */
-                    lhSize = 5;
-                    litSize  = (lhc >> 4) & 0x3FFFF;
-                    litCSize = (lhc >> 22) + (istart[4] << 10);
-                    break;
-                }
-                if (litSize > ZSTD_BLOCKSIZE_MAX) return ERROR(corruption_detected);
-                if (litCSize + lhSize > srcSize) return ERROR(corruption_detected);
-
-                /* prefetch huffman table if cold */
-                if (dctx->ddictIsCold && (litSize > 768 /* heuristic */)) {
-                    PREFETCH_AREA(dctx->HUFptr, sizeof(dctx->entropy.hufTable));
-                }
-
-                if (HUF_isError((litEncType==set_repeat) ?
-                                    ( singleStream ?
-                                        HUF_decompress1X_usingDTable_bmi2(dctx->litBuffer, litSize, istart+lhSize, litCSize, dctx->HUFptr, dctx->bmi2) :
-                                        HUF_decompress4X_usingDTable_bmi2(dctx->litBuffer, litSize, istart+lhSize, litCSize, dctx->HUFptr, dctx->bmi2) ) :
-                                    ( singleStream ?
-                                        HUF_decompress1X1_DCtx_wksp_bmi2(dctx->entropy.hufTable, dctx->litBuffer, litSize, istart+lhSize, litCSize,
-                                                                         dctx->workspace, sizeof(dctx->workspace), dctx->bmi2) :
-                                        HUF_decompress4X_hufOnly_wksp_bmi2(dctx->entropy.hufTable, dctx->litBuffer, litSize, istart+lhSize, litCSize,
-                                                                           dctx->workspace, sizeof(dctx->workspace), dctx->bmi2))))
-                    return ERROR(corruption_detected);
-
-                dctx->litPtr = dctx->litBuffer;
-                dctx->litSize = litSize;
-                dctx->litEntropy = 1;
-                if (litEncType==set_compressed) dctx->HUFptr = dctx->entropy.hufTable;
-                memset(dctx->litBuffer + dctx->litSize, 0, WILDCOPY_OVERLENGTH);
-                return litCSize + lhSize;
-            }
-
-        case set_basic:
-            {   size_t litSize, lhSize;
-                U32 const lhlCode = ((istart[0]) >> 2) & 3;
-                switch(lhlCode)
-                {
-                case 0: case 2: default:   /* note : default is impossible, since lhlCode into [0..3] */
-                    lhSize = 1;
-                    litSize = istart[0] >> 3;
-                    break;
-                case 1:
-                    lhSize = 2;
-                    litSize = MEM_readLE16(istart) >> 4;
-                    break;
-                case 3:
-                    lhSize = 3;
-                    litSize = MEM_readLE24(istart) >> 4;
-                    break;
-                }
-
-                if (lhSize+litSize+WILDCOPY_OVERLENGTH > srcSize) {  /* risk reading beyond src buffer with wildcopy */
-                    if (litSize+lhSize > srcSize) return ERROR(corruption_detected);
-                    memcpy(dctx->litBuffer, istart+lhSize, litSize);
-                    dctx->litPtr = dctx->litBuffer;
-                    dctx->litSize = litSize;
-                    memset(dctx->litBuffer + dctx->litSize, 0, WILDCOPY_OVERLENGTH);
-                    return lhSize+litSize;
-                }
-                /* direct reference into compressed stream */
-                dctx->litPtr = istart+lhSize;
-                dctx->litSize = litSize;
-                return lhSize+litSize;
-            }
-
-        case set_rle:
-            {   U32 const lhlCode = ((istart[0]) >> 2) & 3;
-                size_t litSize, lhSize;
-                switch(lhlCode)
-                {
-                case 0: case 2: default:   /* note : default is impossible, since lhlCode into [0..3] */
-                    lhSize = 1;
-                    litSize = istart[0] >> 3;
-                    break;
-                case 1:
-                    lhSize = 2;
-                    litSize = MEM_readLE16(istart) >> 4;
-                    break;
-                case 3:
-                    lhSize = 3;
-                    litSize = MEM_readLE24(istart) >> 4;
-                    if (srcSize<4) return ERROR(corruption_detected);   /* srcSize >= MIN_CBLOCK_SIZE == 3; here we need lhSize+1 = 4 */
-                    break;
-                }
-                if (litSize > ZSTD_BLOCKSIZE_MAX) return ERROR(corruption_detected);
-                memset(dctx->litBuffer, istart[lhSize], litSize + WILDCOPY_OVERLENGTH);
-                dctx->litPtr = dctx->litBuffer;
-                dctx->litSize = litSize;
-                return lhSize+1;
-            }
-        default:
-            return ERROR(corruption_detected);   /* impossible */
-        }
-    }
-}
-
-/* Default FSE distribution tables.
- * These are pre-calculated FSE decoding tables using default distributions as defined in specification :
- * https://github.com/facebook/zstd/blob/master/doc/zstd_compression_format.md#default-distributions
- * They were generated programmatically with following method :
- * - start from default distributions, present in /lib/common/zstd_internal.h
- * - generate tables normally, using ZSTD_buildFSETable()
- * - printout the content of tables
- * - pretify output, report below, test with fuzzer to ensure it's correct */
-
-/* Default FSE distribution table for Literal Lengths */
-static const ZSTD_seqSymbol LL_defaultDTable[(1<tableLog = 0;
-    DTableH->fastMode = 0;
-
-    cell->nbBits = 0;
-    cell->nextState = 0;
-    assert(nbAddBits < 255);
-    cell->nbAdditionalBits = (BYTE)nbAddBits;
-    cell->baseValue = baseValue;
-}
-
-
-/* ZSTD_buildFSETable() :
- * generate FSE decoding table for one symbol (ll, ml or off) */
-static void
-ZSTD_buildFSETable(ZSTD_seqSymbol* dt,
-    const short* normalizedCounter, unsigned maxSymbolValue,
-    const U32* baseValue, const U32* nbAdditionalBits,
-    unsigned tableLog)
-{
-    ZSTD_seqSymbol* const tableDecode = dt+1;
-    U16 symbolNext[MaxSeq+1];
-
-    U32 const maxSV1 = maxSymbolValue + 1;
-    U32 const tableSize = 1 << tableLog;
-    U32 highThreshold = tableSize-1;
-
-    /* Sanity Checks */
-    assert(maxSymbolValue <= MaxSeq);
-    assert(tableLog <= MaxFSELog);
-
-    /* Init, lay down lowprob symbols */
-    {   ZSTD_seqSymbol_header DTableH;
-        DTableH.tableLog = tableLog;
-        DTableH.fastMode = 1;
-        {   S16 const largeLimit= (S16)(1 << (tableLog-1));
-            U32 s;
-            for (s=0; s= largeLimit) DTableH.fastMode=0;
-                    symbolNext[s] = normalizedCounter[s];
-        }   }   }
-        memcpy(dt, &DTableH, sizeof(DTableH));
-    }
-
-    /* Spread symbols */
-    {   U32 const tableMask = tableSize-1;
-        U32 const step = FSE_TABLESTEP(tableSize);
-        U32 s, position = 0;
-        for (s=0; s highThreshold) position = (position + step) & tableMask;   /* lowprob area */
-        }   }
-        assert(position == 0); /* position must reach all cells once, otherwise normalizedCounter is incorrect */
-    }
-
-    /* Build Decoding table */
-    {   U32 u;
-        for (u=0; u max) return ERROR(corruption_detected);
-        {   U32 const symbol = *(const BYTE*)src;
-            U32 const baseline = baseValue[symbol];
-            U32 const nbBits = nbAdditionalBits[symbol];
-            ZSTD_buildSeqTable_rle(DTableSpace, baseline, nbBits);
-        }
-        *DTablePtr = DTableSpace;
-        return 1;
-    case set_basic :
-        *DTablePtr = defaultTable;
-        return 0;
-    case set_repeat:
-        if (!flagRepeatTable) return ERROR(corruption_detected);
-        /* prefetch FSE table if used */
-        if (ddictIsCold && (nbSeq > 24 /* heuristic */)) {
-            const void* const pStart = *DTablePtr;
-            size_t const pSize = sizeof(ZSTD_seqSymbol) * (SEQSYMBOL_TABLE_SIZE(maxLog));
-            PREFETCH_AREA(pStart, pSize);
-        }
-        return 0;
-    case set_compressed :
-        {   U32 tableLog;
-            S16 norm[MaxSeq+1];
-            size_t const headerSize = FSE_readNCount(norm, &max, &tableLog, src, srcSize);
-            if (FSE_isError(headerSize)) return ERROR(corruption_detected);
-            if (tableLog > maxLog) return ERROR(corruption_detected);
-            ZSTD_buildFSETable(DTableSpace, norm, max, baseValue, nbAdditionalBits, tableLog);
-            *DTablePtr = DTableSpace;
-            return headerSize;
-        }
-    default :   /* impossible */
-        assert(0);
-        return ERROR(GENERIC);
-    }
-}
-
-static const U32 LL_base[MaxLL+1] = {
-                 0,    1,    2,     3,     4,     5,     6,      7,
-                 8,    9,   10,    11,    12,    13,    14,     15,
-                16,   18,   20,    22,    24,    28,    32,     40,
-                48,   64, 0x80, 0x100, 0x200, 0x400, 0x800, 0x1000,
-                0x2000, 0x4000, 0x8000, 0x10000 };
-
-static const U32 OF_base[MaxOff+1] = {
-                 0,        1,       1,       5,     0xD,     0x1D,     0x3D,     0x7D,
-                 0xFD,   0x1FD,   0x3FD,   0x7FD,   0xFFD,   0x1FFD,   0x3FFD,   0x7FFD,
-                 0xFFFD, 0x1FFFD, 0x3FFFD, 0x7FFFD, 0xFFFFD, 0x1FFFFD, 0x3FFFFD, 0x7FFFFD,
-                 0xFFFFFD, 0x1FFFFFD, 0x3FFFFFD, 0x7FFFFFD, 0xFFFFFFD, 0x1FFFFFFD, 0x3FFFFFFD, 0x7FFFFFFD };
-
-static const U32 OF_bits[MaxOff+1] = {
-                     0,  1,  2,  3,  4,  5,  6,  7,
-                     8,  9, 10, 11, 12, 13, 14, 15,
-                    16, 17, 18, 19, 20, 21, 22, 23,
-                    24, 25, 26, 27, 28, 29, 30, 31 };
-
-static const U32 ML_base[MaxML+1] = {
-                     3,  4,  5,    6,     7,     8,     9,    10,
-                    11, 12, 13,   14,    15,    16,    17,    18,
-                    19, 20, 21,   22,    23,    24,    25,    26,
-                    27, 28, 29,   30,    31,    32,    33,    34,
-                    35, 37, 39,   41,    43,    47,    51,    59,
-                    67, 83, 99, 0x83, 0x103, 0x203, 0x403, 0x803,
-                    0x1003, 0x2003, 0x4003, 0x8003, 0x10003 };
-
-/* Hidden delcaration for fullbench */
-size_t ZSTD_decodeSeqHeaders(ZSTD_DCtx* dctx, int* nbSeqPtr,
-                             const void* src, size_t srcSize);
-
-size_t ZSTD_decodeSeqHeaders(ZSTD_DCtx* dctx, int* nbSeqPtr,
-                             const void* src, size_t srcSize)
-{
-    const BYTE* const istart = (const BYTE* const)src;
-    const BYTE* const iend = istart + srcSize;
-    const BYTE* ip = istart;
-    int nbSeq;
-    DEBUGLOG(5, "ZSTD_decodeSeqHeaders");
-
-    /* check */
-    if (srcSize < MIN_SEQUENCES_SIZE) return ERROR(srcSize_wrong);
-
-    /* SeqHead */
-    nbSeq = *ip++;
-    if (!nbSeq) { *nbSeqPtr=0; return 1; }
-    if (nbSeq > 0x7F) {
-        if (nbSeq == 0xFF) {
-            if (ip+2 > iend) return ERROR(srcSize_wrong);
-            nbSeq = MEM_readLE16(ip) + LONGNBSEQ, ip+=2;
-        } else {
-            if (ip >= iend) return ERROR(srcSize_wrong);
-            nbSeq = ((nbSeq-0x80)<<8) + *ip++;
-        }
-    }
-    *nbSeqPtr = nbSeq;
-
-    /* FSE table descriptors */
-    if (ip+4 > iend) return ERROR(srcSize_wrong); /* minimum possible size */
-    {   symbolEncodingType_e const LLtype = (symbolEncodingType_e)(*ip >> 6);
-        symbolEncodingType_e const OFtype = (symbolEncodingType_e)((*ip >> 4) & 3);
-        symbolEncodingType_e const MLtype = (symbolEncodingType_e)((*ip >> 2) & 3);
-        ip++;
-
-        /* Build DTables */
-        {   size_t const llhSize = ZSTD_buildSeqTable(dctx->entropy.LLTable, &dctx->LLTptr,
-                                                      LLtype, MaxLL, LLFSELog,
-                                                      ip, iend-ip,
-                                                      LL_base, LL_bits,
-                                                      LL_defaultDTable, dctx->fseEntropy,
-                                                      dctx->ddictIsCold, nbSeq);
-            if (ZSTD_isError(llhSize)) return ERROR(corruption_detected);
-            ip += llhSize;
-        }
-
-        {   size_t const ofhSize = ZSTD_buildSeqTable(dctx->entropy.OFTable, &dctx->OFTptr,
-                                                      OFtype, MaxOff, OffFSELog,
-                                                      ip, iend-ip,
-                                                      OF_base, OF_bits,
-                                                      OF_defaultDTable, dctx->fseEntropy,
-                                                      dctx->ddictIsCold, nbSeq);
-            if (ZSTD_isError(ofhSize)) return ERROR(corruption_detected);
-            ip += ofhSize;
-        }
-
-        {   size_t const mlhSize = ZSTD_buildSeqTable(dctx->entropy.MLTable, &dctx->MLTptr,
-                                                      MLtype, MaxML, MLFSELog,
-                                                      ip, iend-ip,
-                                                      ML_base, ML_bits,
-                                                      ML_defaultDTable, dctx->fseEntropy,
-                                                      dctx->ddictIsCold, nbSeq);
-            if (ZSTD_isError(mlhSize)) return ERROR(corruption_detected);
-            ip += mlhSize;
-        }
-    }
-
-    /* prefetch dictionary content */
-    if (dctx->ddictIsCold) {
-        size_t const dictSize = (const char*)dctx->prefixStart - (const char*)dctx->virtualStart;
-        size_t const psmin = MIN(dictSize, (size_t)(64*nbSeq) /* heuristic */ );
-        size_t const pSize = MIN(psmin, 128 KB /* protection */ );
-        const void* const pStart = (const char*)dctx->dictEnd - pSize;
-        PREFETCH_AREA(pStart, pSize);
-        dctx->ddictIsCold = 0;
-    }
-
-    return ip-istart;
-}
-
-
-typedef struct {
-    size_t litLength;
-    size_t matchLength;
-    size_t offset;
-    const BYTE* match;
-} seq_t;
-
-typedef struct {
-    size_t state;
-    const ZSTD_seqSymbol* table;
-} ZSTD_fseState;
-
-typedef struct {
-    BIT_DStream_t DStream;
-    ZSTD_fseState stateLL;
-    ZSTD_fseState stateOffb;
-    ZSTD_fseState stateML;
-    size_t prevOffset[ZSTD_REP_NUM];
-    const BYTE* prefixStart;
-    const BYTE* dictEnd;
-    size_t pos;
-} seqState_t;
-
-
-FORCE_NOINLINE
-size_t ZSTD_execSequenceLast7(BYTE* op,
-                              BYTE* const oend, seq_t sequence,
-                              const BYTE** litPtr, const BYTE* const litLimit,
-                              const BYTE* const base, const BYTE* const vBase, const BYTE* const dictEnd)
-{
-    BYTE* const oLitEnd = op + sequence.litLength;
-    size_t const sequenceLength = sequence.litLength + sequence.matchLength;
-    BYTE* const oMatchEnd = op + sequenceLength;   /* risk : address space overflow (32-bits) */
-    BYTE* const oend_w = oend - WILDCOPY_OVERLENGTH;
-    const BYTE* const iLitEnd = *litPtr + sequence.litLength;
-    const BYTE* match = oLitEnd - sequence.offset;
-
-    /* check */
-    if (oMatchEnd>oend) return ERROR(dstSize_tooSmall); /* last match must start at a minimum distance of WILDCOPY_OVERLENGTH from oend */
-    if (iLitEnd > litLimit) return ERROR(corruption_detected);   /* over-read beyond lit buffer */
-    if (oLitEnd <= oend_w) return ERROR(GENERIC);   /* Precondition */
-
-    /* copy literals */
-    if (op < oend_w) {
-        ZSTD_wildcopy(op, *litPtr, oend_w - op);
-        *litPtr += oend_w - op;
-        op = oend_w;
-    }
-    while (op < oLitEnd) *op++ = *(*litPtr)++;
-
-    /* copy Match */
-    if (sequence.offset > (size_t)(oLitEnd - base)) {
-        /* offset beyond prefix */
-        if (sequence.offset > (size_t)(oLitEnd - vBase)) return ERROR(corruption_detected);
-        match = dictEnd - (base-match);
-        if (match + sequence.matchLength <= dictEnd) {
-            memmove(oLitEnd, match, sequence.matchLength);
-            return sequenceLength;
-        }
-        /* span extDict & currentPrefixSegment */
-        {   size_t const length1 = dictEnd - match;
-            memmove(oLitEnd, match, length1);
-            op = oLitEnd + length1;
-            sequence.matchLength -= length1;
-            match = base;
-    }   }
-    while (op < oMatchEnd) *op++ = *match++;
-    return sequenceLength;
-}
-
-
-HINT_INLINE
-size_t ZSTD_execSequence(BYTE* op,
-                         BYTE* const oend, seq_t sequence,
-                         const BYTE** litPtr, const BYTE* const litLimit,
-                         const BYTE* const prefixStart, const BYTE* const virtualStart, const BYTE* const dictEnd)
-{
-    BYTE* const oLitEnd = op + sequence.litLength;
-    size_t const sequenceLength = sequence.litLength + sequence.matchLength;
-    BYTE* const oMatchEnd = op + sequenceLength;   /* risk : address space overflow (32-bits) */
-    BYTE* const oend_w = oend - WILDCOPY_OVERLENGTH;
-    const BYTE* const iLitEnd = *litPtr + sequence.litLength;
-    const BYTE* match = oLitEnd - sequence.offset;
-
-    /* check */
-    if (oMatchEnd>oend) return ERROR(dstSize_tooSmall); /* last match must start at a minimum distance of WILDCOPY_OVERLENGTH from oend */
-    if (iLitEnd > litLimit) return ERROR(corruption_detected);   /* over-read beyond lit buffer */
-    if (oLitEnd>oend_w) return ZSTD_execSequenceLast7(op, oend, sequence, litPtr, litLimit, prefixStart, virtualStart, dictEnd);
-
-    /* copy Literals */
-    ZSTD_copy8(op, *litPtr);
-    if (sequence.litLength > 8)
-        ZSTD_wildcopy(op+8, (*litPtr)+8, sequence.litLength - 8);   /* note : since oLitEnd <= oend-WILDCOPY_OVERLENGTH, no risk of overwrite beyond oend */
-    op = oLitEnd;
-    *litPtr = iLitEnd;   /* update for next sequence */
-
-    /* copy Match */
-    if (sequence.offset > (size_t)(oLitEnd - prefixStart)) {
-        /* offset beyond prefix -> go into extDict */
-        if (sequence.offset > (size_t)(oLitEnd - virtualStart))
-            return ERROR(corruption_detected);
-        match = dictEnd + (match - prefixStart);
-        if (match + sequence.matchLength <= dictEnd) {
-            memmove(oLitEnd, match, sequence.matchLength);
-            return sequenceLength;
-        }
-        /* span extDict & currentPrefixSegment */
-        {   size_t const length1 = dictEnd - match;
-            memmove(oLitEnd, match, length1);
-            op = oLitEnd + length1;
-            sequence.matchLength -= length1;
-            match = prefixStart;
-            if (op > oend_w || sequence.matchLength < MINMATCH) {
-              U32 i;
-              for (i = 0; i < sequence.matchLength; ++i) op[i] = match[i];
-              return sequenceLength;
-            }
-    }   }
-    /* Requirement: op <= oend_w && sequence.matchLength >= MINMATCH */
-
-    /* match within prefix */
-    if (sequence.offset < 8) {
-        /* close range match, overlap */
-        static const U32 dec32table[] = { 0, 1, 2, 1, 4, 4, 4, 4 };   /* added */
-        static const int dec64table[] = { 8, 8, 8, 7, 8, 9,10,11 };   /* subtracted */
-        int const sub2 = dec64table[sequence.offset];
-        op[0] = match[0];
-        op[1] = match[1];
-        op[2] = match[2];
-        op[3] = match[3];
-        match += dec32table[sequence.offset];
-        ZSTD_copy4(op+4, match);
-        match -= sub2;
-    } else {
-        ZSTD_copy8(op, match);
-    }
-    op += 8; match += 8;
-
-    if (oMatchEnd > oend-(16-MINMATCH)) {
-        if (op < oend_w) {
-            ZSTD_wildcopy(op, match, oend_w - op);
-            match += oend_w - op;
-            op = oend_w;
-        }
-        while (op < oMatchEnd) *op++ = *match++;
-    } else {
-        ZSTD_wildcopy(op, match, (ptrdiff_t)sequence.matchLength-8);   /* works even if matchLength < 8 */
-    }
-    return sequenceLength;
-}
-
-
-HINT_INLINE
-size_t ZSTD_execSequenceLong(BYTE* op,
-                             BYTE* const oend, seq_t sequence,
-                             const BYTE** litPtr, const BYTE* const litLimit,
-                             const BYTE* const prefixStart, const BYTE* const dictStart, const BYTE* const dictEnd)
-{
-    BYTE* const oLitEnd = op + sequence.litLength;
-    size_t const sequenceLength = sequence.litLength + sequence.matchLength;
-    BYTE* const oMatchEnd = op + sequenceLength;   /* risk : address space overflow (32-bits) */
-    BYTE* const oend_w = oend - WILDCOPY_OVERLENGTH;
-    const BYTE* const iLitEnd = *litPtr + sequence.litLength;
-    const BYTE* match = sequence.match;
-
-    /* check */
-    if (oMatchEnd > oend) return ERROR(dstSize_tooSmall); /* last match must start at a minimum distance of WILDCOPY_OVERLENGTH from oend */
-    if (iLitEnd > litLimit) return ERROR(corruption_detected);   /* over-read beyond lit buffer */
-    if (oLitEnd > oend_w) return ZSTD_execSequenceLast7(op, oend, sequence, litPtr, litLimit, prefixStart, dictStart, dictEnd);
-
-    /* copy Literals */
-    ZSTD_copy8(op, *litPtr);  /* note : op <= oLitEnd <= oend_w == oend - 8 */
-    if (sequence.litLength > 8)
-        ZSTD_wildcopy(op+8, (*litPtr)+8, sequence.litLength - 8);   /* note : since oLitEnd <= oend-WILDCOPY_OVERLENGTH, no risk of overwrite beyond oend */
-    op = oLitEnd;
-    *litPtr = iLitEnd;   /* update for next sequence */
-
-    /* copy Match */
-    if (sequence.offset > (size_t)(oLitEnd - prefixStart)) {
-        /* offset beyond prefix */
-        if (sequence.offset > (size_t)(oLitEnd - dictStart)) return ERROR(corruption_detected);
-        if (match + sequence.matchLength <= dictEnd) {
-            memmove(oLitEnd, match, sequence.matchLength);
-            return sequenceLength;
-        }
-        /* span extDict & currentPrefixSegment */
-        {   size_t const length1 = dictEnd - match;
-            memmove(oLitEnd, match, length1);
-            op = oLitEnd + length1;
-            sequence.matchLength -= length1;
-            match = prefixStart;
-            if (op > oend_w || sequence.matchLength < MINMATCH) {
-              U32 i;
-              for (i = 0; i < sequence.matchLength; ++i) op[i] = match[i];
-              return sequenceLength;
-            }
-    }   }
-    assert(op <= oend_w);
-    assert(sequence.matchLength >= MINMATCH);
-
-    /* match within prefix */
-    if (sequence.offset < 8) {
-        /* close range match, overlap */
-        static const U32 dec32table[] = { 0, 1, 2, 1, 4, 4, 4, 4 };   /* added */
-        static const int dec64table[] = { 8, 8, 8, 7, 8, 9,10,11 };   /* subtracted */
-        int const sub2 = dec64table[sequence.offset];
-        op[0] = match[0];
-        op[1] = match[1];
-        op[2] = match[2];
-        op[3] = match[3];
-        match += dec32table[sequence.offset];
-        ZSTD_copy4(op+4, match);
-        match -= sub2;
-    } else {
-        ZSTD_copy8(op, match);
-    }
-    op += 8; match += 8;
-
-    if (oMatchEnd > oend-(16-MINMATCH)) {
-        if (op < oend_w) {
-            ZSTD_wildcopy(op, match, oend_w - op);
-            match += oend_w - op;
-            op = oend_w;
-        }
-        while (op < oMatchEnd) *op++ = *match++;
-    } else {
-        ZSTD_wildcopy(op, match, (ptrdiff_t)sequence.matchLength-8);   /* works even if matchLength < 8 */
-    }
-    return sequenceLength;
-}
-
-static void
-ZSTD_initFseState(ZSTD_fseState* DStatePtr, BIT_DStream_t* bitD, const ZSTD_seqSymbol* dt)
-{
-    const void* ptr = dt;
-    const ZSTD_seqSymbol_header* const DTableH = (const ZSTD_seqSymbol_header*)ptr;
-    DStatePtr->state = BIT_readBits(bitD, DTableH->tableLog);
-    DEBUGLOG(6, "ZSTD_initFseState : val=%u using %u bits",
-                (U32)DStatePtr->state, DTableH->tableLog);
-    BIT_reloadDStream(bitD);
-    DStatePtr->table = dt + 1;
-}
-
-FORCE_INLINE_TEMPLATE void
-ZSTD_updateFseState(ZSTD_fseState* DStatePtr, BIT_DStream_t* bitD)
-{
-    ZSTD_seqSymbol const DInfo = DStatePtr->table[DStatePtr->state];
-    U32 const nbBits = DInfo.nbBits;
-    size_t const lowBits = BIT_readBits(bitD, nbBits);
-    DStatePtr->state = DInfo.nextState + lowBits;
-}
-
-/* We need to add at most (ZSTD_WINDOWLOG_MAX_32 - 1) bits to read the maximum
- * offset bits. But we can only read at most (STREAM_ACCUMULATOR_MIN_32 - 1)
- * bits before reloading. This value is the maximum number of bytes we read
- * after reloading when we are decoding long offets.
- */
-#define LONG_OFFSETS_MAX_EXTRA_BITS_32                       \
-    (ZSTD_WINDOWLOG_MAX_32 > STREAM_ACCUMULATOR_MIN_32       \
-        ? ZSTD_WINDOWLOG_MAX_32 - STREAM_ACCUMULATOR_MIN_32  \
-        : 0)
-
-typedef enum { ZSTD_lo_isRegularOffset, ZSTD_lo_isLongOffset=1 } ZSTD_longOffset_e;
-
-FORCE_INLINE_TEMPLATE seq_t
-ZSTD_decodeSequence(seqState_t* seqState, const ZSTD_longOffset_e longOffsets)
-{
-    seq_t seq;
-    U32 const llBits = seqState->stateLL.table[seqState->stateLL.state].nbAdditionalBits;
-    U32 const mlBits = seqState->stateML.table[seqState->stateML.state].nbAdditionalBits;
-    U32 const ofBits = seqState->stateOffb.table[seqState->stateOffb.state].nbAdditionalBits;
-    U32 const totalBits = llBits+mlBits+ofBits;
-    U32 const llBase = seqState->stateLL.table[seqState->stateLL.state].baseValue;
-    U32 const mlBase = seqState->stateML.table[seqState->stateML.state].baseValue;
-    U32 const ofBase = seqState->stateOffb.table[seqState->stateOffb.state].baseValue;
-
-    /* sequence */
-    {   size_t offset;
-        if (!ofBits)
-            offset = 0;
-        else {
-            ZSTD_STATIC_ASSERT(ZSTD_lo_isLongOffset == 1);
-            ZSTD_STATIC_ASSERT(LONG_OFFSETS_MAX_EXTRA_BITS_32 == 5);
-            assert(ofBits <= MaxOff);
-            if (MEM_32bits() && longOffsets && (ofBits >= STREAM_ACCUMULATOR_MIN_32)) {
-                U32 const extraBits = ofBits - MIN(ofBits, 32 - seqState->DStream.bitsConsumed);
-                offset = ofBase + (BIT_readBitsFast(&seqState->DStream, ofBits - extraBits) << extraBits);
-                BIT_reloadDStream(&seqState->DStream);
-                if (extraBits) offset += BIT_readBitsFast(&seqState->DStream, extraBits);
-                assert(extraBits <= LONG_OFFSETS_MAX_EXTRA_BITS_32);   /* to avoid another reload */
-            } else {
-                offset = ofBase + BIT_readBitsFast(&seqState->DStream, ofBits/*>0*/);   /* <=  (ZSTD_WINDOWLOG_MAX-1) bits */
-                if (MEM_32bits()) BIT_reloadDStream(&seqState->DStream);
-            }
-        }
-
-        if (ofBits <= 1) {
-            offset += (llBase==0);
-            if (offset) {
-                size_t temp = (offset==3) ? seqState->prevOffset[0] - 1 : seqState->prevOffset[offset];
-                temp += !temp;   /* 0 is not valid; input is corrupted; force offset to 1 */
-                if (offset != 1) seqState->prevOffset[2] = seqState->prevOffset[1];
-                seqState->prevOffset[1] = seqState->prevOffset[0];
-                seqState->prevOffset[0] = offset = temp;
-            } else {  /* offset == 0 */
-                offset = seqState->prevOffset[0];
-            }
-        } else {
-            seqState->prevOffset[2] = seqState->prevOffset[1];
-            seqState->prevOffset[1] = seqState->prevOffset[0];
-            seqState->prevOffset[0] = offset;
-        }
-        seq.offset = offset;
-    }
-
-    seq.matchLength = mlBase
-                    + ((mlBits>0) ? BIT_readBitsFast(&seqState->DStream, mlBits/*>0*/) : 0);  /* <=  16 bits */
-    if (MEM_32bits() && (mlBits+llBits >= STREAM_ACCUMULATOR_MIN_32-LONG_OFFSETS_MAX_EXTRA_BITS_32))
-        BIT_reloadDStream(&seqState->DStream);
-    if (MEM_64bits() && (totalBits >= STREAM_ACCUMULATOR_MIN_64-(LLFSELog+MLFSELog+OffFSELog)))
-        BIT_reloadDStream(&seqState->DStream);
-    /* Ensure there are enough bits to read the rest of data in 64-bit mode. */
-    ZSTD_STATIC_ASSERT(16+LLFSELog+MLFSELog+OffFSELog < STREAM_ACCUMULATOR_MIN_64);
-
-    seq.litLength = llBase
-                  + ((llBits>0) ? BIT_readBitsFast(&seqState->DStream, llBits/*>0*/) : 0);    /* <=  16 bits */
-    if (MEM_32bits())
-        BIT_reloadDStream(&seqState->DStream);
-
-    DEBUGLOG(6, "seq: litL=%u, matchL=%u, offset=%u",
-                (U32)seq.litLength, (U32)seq.matchLength, (U32)seq.offset);
-
-    /* ANS state update */
-    ZSTD_updateFseState(&seqState->stateLL, &seqState->DStream);    /* <=  9 bits */
-    ZSTD_updateFseState(&seqState->stateML, &seqState->DStream);    /* <=  9 bits */
-    if (MEM_32bits()) BIT_reloadDStream(&seqState->DStream);    /* <= 18 bits */
-    ZSTD_updateFseState(&seqState->stateOffb, &seqState->DStream);  /* <=  8 bits */
-
-    return seq;
-}
-
-FORCE_INLINE_TEMPLATE size_t
-ZSTD_decompressSequences_body( ZSTD_DCtx* dctx,
-                               void* dst, size_t maxDstSize,
-                         const void* seqStart, size_t seqSize, int nbSeq,
-                         const ZSTD_longOffset_e isLongOffset)
-{
-    const BYTE* ip = (const BYTE*)seqStart;
-    const BYTE* const iend = ip + seqSize;
-    BYTE* const ostart = (BYTE* const)dst;
-    BYTE* const oend = ostart + maxDstSize;
-    BYTE* op = ostart;
-    const BYTE* litPtr = dctx->litPtr;
-    const BYTE* const litEnd = litPtr + dctx->litSize;
-    const BYTE* const prefixStart = (const BYTE*) (dctx->prefixStart);
-    const BYTE* const vBase = (const BYTE*) (dctx->virtualStart);
-    const BYTE* const dictEnd = (const BYTE*) (dctx->dictEnd);
-    DEBUGLOG(5, "ZSTD_decompressSequences_body");
-
-    /* Regen sequences */
-    if (nbSeq) {
-        seqState_t seqState;
-        dctx->fseEntropy = 1;
-        { U32 i; for (i=0; ientropy.rep[i]; }
-        CHECK_E(BIT_initDStream(&seqState.DStream, ip, iend-ip), corruption_detected);
-        ZSTD_initFseState(&seqState.stateLL, &seqState.DStream, dctx->LLTptr);
-        ZSTD_initFseState(&seqState.stateOffb, &seqState.DStream, dctx->OFTptr);
-        ZSTD_initFseState(&seqState.stateML, &seqState.DStream, dctx->MLTptr);
-
-        for ( ; (BIT_reloadDStream(&(seqState.DStream)) <= BIT_DStream_completed) && nbSeq ; ) {
-            nbSeq--;
-            {   seq_t const sequence = ZSTD_decodeSequence(&seqState, isLongOffset);
-                size_t const oneSeqSize = ZSTD_execSequence(op, oend, sequence, &litPtr, litEnd, prefixStart, vBase, dictEnd);
-                DEBUGLOG(6, "regenerated sequence size : %u", (U32)oneSeqSize);
-                if (ZSTD_isError(oneSeqSize)) return oneSeqSize;
-                op += oneSeqSize;
-        }   }
-
-        /* check if reached exact end */
-        DEBUGLOG(5, "ZSTD_decompressSequences_body: after decode loop, remaining nbSeq : %i", nbSeq);
-        if (nbSeq) return ERROR(corruption_detected);
-        /* save reps for next block */
-        { U32 i; for (i=0; ientropy.rep[i] = (U32)(seqState.prevOffset[i]); }
-    }
-
-    /* last literal segment */
-    {   size_t const lastLLSize = litEnd - litPtr;
-        if (lastLLSize > (size_t)(oend-op)) return ERROR(dstSize_tooSmall);
-        memcpy(op, litPtr, lastLLSize);
-        op += lastLLSize;
-    }
-
-    return op-ostart;
-}
-
-static size_t
-ZSTD_decompressSequences_default(ZSTD_DCtx* dctx,
-                                 void* dst, size_t maxDstSize,
-                           const void* seqStart, size_t seqSize, int nbSeq,
-                           const ZSTD_longOffset_e isLongOffset)
-{
-    return ZSTD_decompressSequences_body(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset);
-}
-
-
-
-FORCE_INLINE_TEMPLATE seq_t
-ZSTD_decodeSequenceLong(seqState_t* seqState, ZSTD_longOffset_e const longOffsets)
-{
-    seq_t seq;
-    U32 const llBits = seqState->stateLL.table[seqState->stateLL.state].nbAdditionalBits;
-    U32 const mlBits = seqState->stateML.table[seqState->stateML.state].nbAdditionalBits;
-    U32 const ofBits = seqState->stateOffb.table[seqState->stateOffb.state].nbAdditionalBits;
-    U32 const totalBits = llBits+mlBits+ofBits;
-    U32 const llBase = seqState->stateLL.table[seqState->stateLL.state].baseValue;
-    U32 const mlBase = seqState->stateML.table[seqState->stateML.state].baseValue;
-    U32 const ofBase = seqState->stateOffb.table[seqState->stateOffb.state].baseValue;
-
-    /* sequence */
-    {   size_t offset;
-        if (!ofBits)
-            offset = 0;
-        else {
-            ZSTD_STATIC_ASSERT(ZSTD_lo_isLongOffset == 1);
-            ZSTD_STATIC_ASSERT(LONG_OFFSETS_MAX_EXTRA_BITS_32 == 5);
-            assert(ofBits <= MaxOff);
-            if (MEM_32bits() && longOffsets) {
-                U32 const extraBits = ofBits - MIN(ofBits, STREAM_ACCUMULATOR_MIN_32-1);
-                offset = ofBase + (BIT_readBitsFast(&seqState->DStream, ofBits - extraBits) << extraBits);
-                if (MEM_32bits() || extraBits) BIT_reloadDStream(&seqState->DStream);
-                if (extraBits) offset += BIT_readBitsFast(&seqState->DStream, extraBits);
-            } else {
-                offset = ofBase + BIT_readBitsFast(&seqState->DStream, ofBits);   /* <=  (ZSTD_WINDOWLOG_MAX-1) bits */
-                if (MEM_32bits()) BIT_reloadDStream(&seqState->DStream);
-            }
-        }
-
-        if (ofBits <= 1) {
-            offset += (llBase==0);
-            if (offset) {
-                size_t temp = (offset==3) ? seqState->prevOffset[0] - 1 : seqState->prevOffset[offset];
-                temp += !temp;   /* 0 is not valid; input is corrupted; force offset to 1 */
-                if (offset != 1) seqState->prevOffset[2] = seqState->prevOffset[1];
-                seqState->prevOffset[1] = seqState->prevOffset[0];
-                seqState->prevOffset[0] = offset = temp;
-            } else {
-                offset = seqState->prevOffset[0];
-            }
-        } else {
-            seqState->prevOffset[2] = seqState->prevOffset[1];
-            seqState->prevOffset[1] = seqState->prevOffset[0];
-            seqState->prevOffset[0] = offset;
-        }
-        seq.offset = offset;
-    }
-
-    seq.matchLength = mlBase + ((mlBits>0) ? BIT_readBitsFast(&seqState->DStream, mlBits) : 0);  /* <=  16 bits */
-    if (MEM_32bits() && (mlBits+llBits >= STREAM_ACCUMULATOR_MIN_32-LONG_OFFSETS_MAX_EXTRA_BITS_32))
-        BIT_reloadDStream(&seqState->DStream);
-    if (MEM_64bits() && (totalBits >= STREAM_ACCUMULATOR_MIN_64-(LLFSELog+MLFSELog+OffFSELog)))
-        BIT_reloadDStream(&seqState->DStream);
-    /* Verify that there is enough bits to read the rest of the data in 64-bit mode. */
-    ZSTD_STATIC_ASSERT(16+LLFSELog+MLFSELog+OffFSELog < STREAM_ACCUMULATOR_MIN_64);
-
-    seq.litLength = llBase + ((llBits>0) ? BIT_readBitsFast(&seqState->DStream, llBits) : 0);    /* <=  16 bits */
-    if (MEM_32bits())
-        BIT_reloadDStream(&seqState->DStream);
-
-    {   size_t const pos = seqState->pos + seq.litLength;
-        const BYTE* const matchBase = (seq.offset > pos) ? seqState->dictEnd : seqState->prefixStart;
-        seq.match = matchBase + pos - seq.offset;  /* note : this operation can overflow when seq.offset is really too large, which can only happen when input is corrupted.
-                                                    * No consequence though : no memory access will occur, overly large offset will be detected in ZSTD_execSequenceLong() */
-        seqState->pos = pos + seq.matchLength;
-    }
-
-    /* ANS state update */
-    ZSTD_updateFseState(&seqState->stateLL, &seqState->DStream);    /* <=  9 bits */
-    ZSTD_updateFseState(&seqState->stateML, &seqState->DStream);    /* <=  9 bits */
-    if (MEM_32bits()) BIT_reloadDStream(&seqState->DStream);    /* <= 18 bits */
-    ZSTD_updateFseState(&seqState->stateOffb, &seqState->DStream);  /* <=  8 bits */
-
-    return seq;
-}
-
-FORCE_INLINE_TEMPLATE size_t
-ZSTD_decompressSequencesLong_body(
-                               ZSTD_DCtx* dctx,
-                               void* dst, size_t maxDstSize,
-                         const void* seqStart, size_t seqSize, int nbSeq,
-                         const ZSTD_longOffset_e isLongOffset)
-{
-    const BYTE* ip = (const BYTE*)seqStart;
-    const BYTE* const iend = ip + seqSize;
-    BYTE* const ostart = (BYTE* const)dst;
-    BYTE* const oend = ostart + maxDstSize;
-    BYTE* op = ostart;
-    const BYTE* litPtr = dctx->litPtr;
-    const BYTE* const litEnd = litPtr + dctx->litSize;
-    const BYTE* const prefixStart = (const BYTE*) (dctx->prefixStart);
-    const BYTE* const dictStart = (const BYTE*) (dctx->virtualStart);
-    const BYTE* const dictEnd = (const BYTE*) (dctx->dictEnd);
-
-    /* Regen sequences */
-    if (nbSeq) {
-#define STORED_SEQS 4
-#define STOSEQ_MASK (STORED_SEQS-1)
-#define ADVANCED_SEQS 4
-        seq_t sequences[STORED_SEQS];
-        int const seqAdvance = MIN(nbSeq, ADVANCED_SEQS);
-        seqState_t seqState;
-        int seqNb;
-        dctx->fseEntropy = 1;
-        { U32 i; for (i=0; ientropy.rep[i]; }
-        seqState.prefixStart = prefixStart;
-        seqState.pos = (size_t)(op-prefixStart);
-        seqState.dictEnd = dictEnd;
-        CHECK_E(BIT_initDStream(&seqState.DStream, ip, iend-ip), corruption_detected);
-        ZSTD_initFseState(&seqState.stateLL, &seqState.DStream, dctx->LLTptr);
-        ZSTD_initFseState(&seqState.stateOffb, &seqState.DStream, dctx->OFTptr);
-        ZSTD_initFseState(&seqState.stateML, &seqState.DStream, dctx->MLTptr);
-
-        /* prepare in advance */
-        for (seqNb=0; (BIT_reloadDStream(&seqState.DStream) <= BIT_DStream_completed) && (seqNbentropy.rep[i] = (U32)(seqState.prevOffset[i]); }
-#undef STORED_SEQS
-#undef STOSEQ_MASK
-#undef ADVANCED_SEQS
-    }
-
-    /* last literal segment */
-    {   size_t const lastLLSize = litEnd - litPtr;
-        if (lastLLSize > (size_t)(oend-op)) return ERROR(dstSize_tooSmall);
-        memcpy(op, litPtr, lastLLSize);
-        op += lastLLSize;
-    }
-
-    return op-ostart;
-}
-
-static size_t
-ZSTD_decompressSequencesLong_default(ZSTD_DCtx* dctx,
-                                 void* dst, size_t maxDstSize,
-                           const void* seqStart, size_t seqSize, int nbSeq,
-                           const ZSTD_longOffset_e isLongOffset)
-{
-    return ZSTD_decompressSequencesLong_body(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset);
-}
-
-
-
-#if DYNAMIC_BMI2
-
-static TARGET_ATTRIBUTE("bmi2") size_t
-ZSTD_decompressSequences_bmi2(ZSTD_DCtx* dctx,
-                                 void* dst, size_t maxDstSize,
-                           const void* seqStart, size_t seqSize, int nbSeq,
-                           const ZSTD_longOffset_e isLongOffset)
-{
-    return ZSTD_decompressSequences_body(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset);
-}
-
-static TARGET_ATTRIBUTE("bmi2") size_t
-ZSTD_decompressSequencesLong_bmi2(ZSTD_DCtx* dctx,
-                                 void* dst, size_t maxDstSize,
-                           const void* seqStart, size_t seqSize, int nbSeq,
-                           const ZSTD_longOffset_e isLongOffset)
-{
-    return ZSTD_decompressSequencesLong_body(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset);
-}
-
-#endif
-
-typedef size_t (*ZSTD_decompressSequences_t)(
-    ZSTD_DCtx *dctx, void *dst, size_t maxDstSize,
-    const void *seqStart, size_t seqSize, int nbSeq,
-    const ZSTD_longOffset_e isLongOffset);
-
-static size_t ZSTD_decompressSequences(ZSTD_DCtx* dctx, void* dst, size_t maxDstSize,
-                                const void* seqStart, size_t seqSize, int nbSeq,
-                                const ZSTD_longOffset_e isLongOffset)
-{
-    DEBUGLOG(5, "ZSTD_decompressSequences");
-#if DYNAMIC_BMI2
-    if (dctx->bmi2) {
-        return ZSTD_decompressSequences_bmi2(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset);
-    }
-#endif
-  return ZSTD_decompressSequences_default(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset);
-}
-
-static size_t ZSTD_decompressSequencesLong(ZSTD_DCtx* dctx,
-                                void* dst, size_t maxDstSize,
-                                const void* seqStart, size_t seqSize, int nbSeq,
-                                const ZSTD_longOffset_e isLongOffset)
-{
-    DEBUGLOG(5, "ZSTD_decompressSequencesLong");
-#if DYNAMIC_BMI2
-    if (dctx->bmi2) {
-        return ZSTD_decompressSequencesLong_bmi2(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset);
-    }
-#endif
-  return ZSTD_decompressSequencesLong_default(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset);
-}
-
-/* ZSTD_getLongOffsetsShare() :
- * condition : offTable must be valid
- * @return : "share" of long offsets (arbitrarily defined as > (1<<23))
- *           compared to maximum possible of (1< 22) total += 1;
-    }
-
-    assert(tableLog <= OffFSELog);
-    total <<= (OffFSELog - tableLog);  /* scale to OffFSELog */
-
-    return total;
-}
-
-
-static size_t ZSTD_decompressBlock_internal(ZSTD_DCtx* dctx,
-                            void* dst, size_t dstCapacity,
-                      const void* src, size_t srcSize, const int frame)
-{   /* blockType == blockCompressed */
-    const BYTE* ip = (const BYTE*)src;
-    /* isLongOffset must be true if there are long offsets.
-     * Offsets are long if they are larger than 2^STREAM_ACCUMULATOR_MIN.
-     * We don't expect that to be the case in 64-bit mode.
-     * In block mode, window size is not known, so we have to be conservative.
-     * (note: but it could be evaluated from current-lowLimit)
-     */
-    ZSTD_longOffset_e const isLongOffset = (ZSTD_longOffset_e)(MEM_32bits() && (!frame || dctx->fParams.windowSize > (1ULL << STREAM_ACCUMULATOR_MIN)));
-    DEBUGLOG(5, "ZSTD_decompressBlock_internal (size : %u)", (U32)srcSize);
-
-    if (srcSize >= ZSTD_BLOCKSIZE_MAX) return ERROR(srcSize_wrong);
-
-    /* Decode literals section */
-    {   size_t const litCSize = ZSTD_decodeLiteralsBlock(dctx, src, srcSize);
-        DEBUGLOG(5, "ZSTD_decodeLiteralsBlock : %u", (U32)litCSize);
-        if (ZSTD_isError(litCSize)) return litCSize;
-        ip += litCSize;
-        srcSize -= litCSize;
-    }
-
-    /* Build Decoding Tables */
-    {   int nbSeq;
-        size_t const seqHSize = ZSTD_decodeSeqHeaders(dctx, &nbSeq, ip, srcSize);
-        if (ZSTD_isError(seqHSize)) return seqHSize;
-        ip += seqHSize;
-        srcSize -= seqHSize;
-
-        if ( (!frame || dctx->fParams.windowSize > (1<<24))
-          && (nbSeq>0) ) {  /* could probably use a larger nbSeq limit */
-            U32 const shareLongOffsets = ZSTD_getLongOffsetsShare(dctx->OFTptr);
-            U32 const minShare = MEM_64bits() ? 7 : 20; /* heuristic values, correspond to 2.73% and 7.81% */
-            if (shareLongOffsets >= minShare)
-                return ZSTD_decompressSequencesLong(dctx, dst, dstCapacity, ip, srcSize, nbSeq, isLongOffset);
-        }
-
-        return ZSTD_decompressSequences(dctx, dst, dstCapacity, ip, srcSize, nbSeq, isLongOffset);
-    }
-}
-
-
-static void ZSTD_checkContinuity(ZSTD_DCtx* dctx, const void* dst)
-{
-    if (dst != dctx->previousDstEnd) {   /* not contiguous */
-        dctx->dictEnd = dctx->previousDstEnd;
-        dctx->virtualStart = (const char*)dst - ((const char*)(dctx->previousDstEnd) - (const char*)(dctx->prefixStart));
-        dctx->prefixStart = dst;
-        dctx->previousDstEnd = dst;
-    }
-}
-
-size_t ZSTD_decompressBlock(ZSTD_DCtx* dctx,
-                            void* dst, size_t dstCapacity,
-                      const void* src, size_t srcSize)
-{
-    size_t dSize;
-    ZSTD_checkContinuity(dctx, dst);
-    dSize = ZSTD_decompressBlock_internal(dctx, dst, dstCapacity, src, srcSize, /* frame */ 0);
-    dctx->previousDstEnd = (char*)dst + dSize;
-    return dSize;
-}
-
-
-/** ZSTD_insertBlock() :
-    insert `src` block into `dctx` history. Useful to track uncompressed blocks. */
-ZSTDLIB_API size_t ZSTD_insertBlock(ZSTD_DCtx* dctx, const void* blockStart, size_t blockSize)
-{
-    ZSTD_checkContinuity(dctx, blockStart);
-    dctx->previousDstEnd = (const char*)blockStart + blockSize;
-    return blockSize;
-}
-
-
-static size_t ZSTD_generateNxBytes(void* dst, size_t dstCapacity, BYTE value, size_t length)
-{
-    if (length > dstCapacity) return ERROR(dstSize_tooSmall);
-    memset(dst, value, length);
-    return length;
-}
-
 /** ZSTD_findFrameCompressedSize() :
  *  compatible with legacy mode
  *  `src` must point to the start of a ZSTD frame, ZSTD legacy frame, or skippable frame
@@ -1806,9 +447,9 @@ size_t ZSTD_findFrameCompressedSize(const void *src, size_t srcSize)
     if (ZSTD_isLegacy(src, srcSize))
         return ZSTD_findFrameCompressedSizeLegacy(src, srcSize);
 #endif
-    if ( (srcSize >= ZSTD_skippableHeaderSize)
-      && (MEM_readLE32(src) & 0xFFFFFFF0U) == ZSTD_MAGIC_SKIPPABLE_START ) {
-        return ZSTD_skippableHeaderSize + MEM_readLE32((const BYTE*)src + ZSTD_FRAMEIDSIZE);
+    if ( (srcSize >= ZSTD_SKIPPABLEHEADERSIZE)
+      && (MEM_readLE32(src) & ZSTD_MAGIC_SKIPPABLE_MASK) == ZSTD_MAGIC_SKIPPABLE_START ) {
+        return readSkippableFrameSize(src, srcSize);
     } else {
         const BYTE* ip = (const BYTE*)src;
         const BYTE* const ipstart = ip;
@@ -1848,8 +489,64 @@ size_t ZSTD_findFrameCompressedSize(const void *src, size_t srcSize)
     }
 }
 
+
+
+/*-*************************************************************
+ *   Frame decoding
+ ***************************************************************/
+
+
+void ZSTD_checkContinuity(ZSTD_DCtx* dctx, const void* dst)
+{
+    if (dst != dctx->previousDstEnd) {   /* not contiguous */
+        dctx->dictEnd = dctx->previousDstEnd;
+        dctx->virtualStart = (const char*)dst - ((const char*)(dctx->previousDstEnd) - (const char*)(dctx->prefixStart));
+        dctx->prefixStart = dst;
+        dctx->previousDstEnd = dst;
+    }
+}
+
+/** ZSTD_insertBlock() :
+    insert `src` block into `dctx` history. Useful to track uncompressed blocks. */
+size_t ZSTD_insertBlock(ZSTD_DCtx* dctx, const void* blockStart, size_t blockSize)
+{
+    ZSTD_checkContinuity(dctx, blockStart);
+    dctx->previousDstEnd = (const char*)blockStart + blockSize;
+    return blockSize;
+}
+
+
+static size_t ZSTD_copyRawBlock(void* dst, size_t dstCapacity,
+                          const void* src, size_t srcSize)
+{
+    DEBUGLOG(5, "ZSTD_copyRawBlock");
+    if (dst == NULL) {
+        if (srcSize == 0) return 0;
+        return ERROR(dstBuffer_null);
+    }
+    if (srcSize > dstCapacity) return ERROR(dstSize_tooSmall);
+    memcpy(dst, src, srcSize);
+    return srcSize;
+}
+
+static size_t ZSTD_setRleBlock(void* dst, size_t dstCapacity,
+                               BYTE b,
+                               size_t regenSize)
+{
+    if (dst == NULL) {
+        if (regenSize == 0) return 0;
+        return ERROR(dstBuffer_null);
+    }
+    if (regenSize > dstCapacity) return ERROR(dstSize_tooSmall);
+    memset(dst, b, regenSize);
+    return regenSize;
+}
+
+
 /*! ZSTD_decompressFrame() :
-*   @dctx must be properly initialized */
+ * @dctx must be properly initialized
+ *  will update *srcPtr and *srcSizePtr,
+ *  to make *srcPtr progress by one frame. */
 static size_t ZSTD_decompressFrame(ZSTD_DCtx* dctx,
                                    void* dst, size_t dstCapacity,
                              const void** srcPtr, size_t *srcSizePtr)
@@ -1858,31 +555,33 @@ static size_t ZSTD_decompressFrame(ZSTD_DCtx* dctx,
     BYTE* const ostart = (BYTE* const)dst;
     BYTE* const oend = ostart + dstCapacity;
     BYTE* op = ostart;
-    size_t remainingSize = *srcSizePtr;
+    size_t remainingSrcSize = *srcSizePtr;
+
+    DEBUGLOG(4, "ZSTD_decompressFrame (srcSize:%i)", (int)*srcSizePtr);
 
     /* check */
-    if (remainingSize < ZSTD_frameHeaderSize_min+ZSTD_blockHeaderSize)
+    if (remainingSrcSize < ZSTD_FRAMEHEADERSIZE_MIN+ZSTD_blockHeaderSize)
         return ERROR(srcSize_wrong);
 
     /* Frame Header */
-    {   size_t const frameHeaderSize = ZSTD_frameHeaderSize(ip, ZSTD_frameHeaderSize_prefix);
+    {   size_t const frameHeaderSize = ZSTD_frameHeaderSize(ip, ZSTD_FRAMEHEADERSIZE_PREFIX);
         if (ZSTD_isError(frameHeaderSize)) return frameHeaderSize;
-        if (remainingSize < frameHeaderSize+ZSTD_blockHeaderSize)
+        if (remainingSrcSize < frameHeaderSize+ZSTD_blockHeaderSize)
             return ERROR(srcSize_wrong);
         CHECK_F( ZSTD_decodeFrameHeader(dctx, ip, frameHeaderSize) );
-        ip += frameHeaderSize; remainingSize -= frameHeaderSize;
+        ip += frameHeaderSize; remainingSrcSize -= frameHeaderSize;
     }
 
     /* Loop on each block */
     while (1) {
         size_t decodedSize;
         blockProperties_t blockProperties;
-        size_t const cBlockSize = ZSTD_getcBlockSize(ip, remainingSize, &blockProperties);
+        size_t const cBlockSize = ZSTD_getcBlockSize(ip, remainingSrcSize, &blockProperties);
         if (ZSTD_isError(cBlockSize)) return cBlockSize;
 
         ip += ZSTD_blockHeaderSize;
-        remainingSize -= ZSTD_blockHeaderSize;
-        if (cBlockSize > remainingSize) return ERROR(srcSize_wrong);
+        remainingSrcSize -= ZSTD_blockHeaderSize;
+        if (cBlockSize > remainingSrcSize) return ERROR(srcSize_wrong);
 
         switch(blockProperties.blockType)
         {
@@ -1893,7 +592,7 @@ static size_t ZSTD_decompressFrame(ZSTD_DCtx* dctx,
             decodedSize = ZSTD_copyRawBlock(op, oend-op, ip, cBlockSize);
             break;
         case bt_rle :
-            decodedSize = ZSTD_generateNxBytes(op, oend-op, *ip, blockProperties.origSize);
+            decodedSize = ZSTD_setRleBlock(op, oend-op, *ip, blockProperties.origSize);
             break;
         case bt_reserved :
         default:
@@ -1905,7 +604,7 @@ static size_t ZSTD_decompressFrame(ZSTD_DCtx* dctx,
             XXH64_update(&dctx->xxhState, op, decodedSize);
         op += decodedSize;
         ip += cBlockSize;
-        remainingSize -= cBlockSize;
+        remainingSrcSize -= cBlockSize;
         if (blockProperties.lastBlock) break;
     }
 
@@ -1916,16 +615,16 @@ static size_t ZSTD_decompressFrame(ZSTD_DCtx* dctx,
     if (dctx->fParams.checksumFlag) { /* Frame content checksum verification */
         U32 const checkCalc = (U32)XXH64_digest(&dctx->xxhState);
         U32 checkRead;
-        if (remainingSize<4) return ERROR(checksum_wrong);
+        if (remainingSrcSize<4) return ERROR(checksum_wrong);
         checkRead = MEM_readLE32(ip);
         if (checkRead != checkCalc) return ERROR(checksum_wrong);
         ip += 4;
-        remainingSize -= 4;
+        remainingSrcSize -= 4;
     }
 
     /* Allow caller to get size read */
     *srcPtr = ip;
-    *srcSizePtr = remainingSize;
+    *srcSizePtr = remainingSrcSize;
     return op-ostart;
 }
 
@@ -1942,11 +641,11 @@ static size_t ZSTD_decompressMultiFrame(ZSTD_DCtx* dctx,
     assert(dict==NULL || ddict==NULL);  /* either dict or ddict set, not both */
 
     if (ddict) {
-        dict = ZSTD_DDictDictContent(ddict);
-        dictSize = ZSTD_DDictDictSize(ddict);
+        dict = ZSTD_DDict_dictContent(ddict);
+        dictSize = ZSTD_DDict_dictSize(ddict);
     }
 
-    while (srcSize >= ZSTD_frameHeaderSize_prefix) {
+    while (srcSize >= ZSTD_FRAMEHEADERSIZE_PREFIX) {
 
 #if defined(ZSTD_LEGACY_SUPPORT) && (ZSTD_LEGACY_SUPPORT >= 1)
         if (ZSTD_isLegacy(src, srcSize)) {
@@ -1957,7 +656,9 @@ static size_t ZSTD_decompressMultiFrame(ZSTD_DCtx* dctx,
             if (dctx->staticSize) return ERROR(memory_allocation);
 
             decodedSize = ZSTD_decompressLegacy(dst, dstCapacity, src, frameSize, dict, dictSize);
+            if (ZSTD_isError(decodedSize)) return decodedSize;
 
+            assert(decodedSize <=- dstCapacity);
             dst = (BYTE*)dst + decodedSize;
             dstCapacity -= decodedSize;
 
@@ -1970,13 +671,11 @@ static size_t ZSTD_decompressMultiFrame(ZSTD_DCtx* dctx,
 
         {   U32 const magicNumber = MEM_readLE32(src);
             DEBUGLOG(4, "reading magic number %08X (expecting %08X)",
-                        (U32)magicNumber, (U32)ZSTD_MAGICNUMBER);
-            if ((magicNumber & 0xFFFFFFF0U) == ZSTD_MAGIC_SKIPPABLE_START) {
-                size_t skippableSize;
-                if (srcSize < ZSTD_skippableHeaderSize)
-                    return ERROR(srcSize_wrong);
-                skippableSize = MEM_readLE32((const BYTE*)src + ZSTD_FRAMEIDSIZE)
-                              + ZSTD_skippableHeaderSize;
+                        (unsigned)magicNumber, ZSTD_MAGICNUMBER);
+            if ((magicNumber & ZSTD_MAGIC_SKIPPABLE_MASK) == ZSTD_MAGIC_SKIPPABLE_START) {
+                size_t const skippableSize = readSkippableFrameSize(src, srcSize);
+                if (ZSTD_isError(skippableSize))
+                    return skippableSize;
                 if (srcSize < skippableSize) return ERROR(srcSize_wrong);
 
                 src = (const BYTE *)src + skippableSize;
@@ -2010,7 +709,7 @@ static size_t ZSTD_decompressMultiFrame(ZSTD_DCtx* dctx,
                 return ERROR(srcSize_wrong);
             }
             if (ZSTD_isError(res)) return res;
-            /* no need to bound check, ZSTD_decompressFrame already has */
+            assert(res <= dstCapacity);
             dst = (BYTE*)dst + res;
             dstCapacity -= res;
         }
@@ -2090,9 +789,10 @@ static int ZSTD_isSkipFrame(ZSTD_DCtx* dctx) { return dctx->stage == ZSTDds_skip
  *            or an error code, which can be tested using ZSTD_isError() */
 size_t ZSTD_decompressContinue(ZSTD_DCtx* dctx, void* dst, size_t dstCapacity, const void* src, size_t srcSize)
 {
-    DEBUGLOG(5, "ZSTD_decompressContinue (srcSize:%u)", (U32)srcSize);
+    DEBUGLOG(5, "ZSTD_decompressContinue (srcSize:%u)", (unsigned)srcSize);
     /* Sanity check */
-    if (srcSize != dctx->expected) return ERROR(srcSize_wrong);  /* not allowed */
+    if (srcSize != dctx->expected)
+        return ERROR(srcSize_wrong);  /* not allowed */
     if (dstCapacity) ZSTD_checkContinuity(dctx, dst);
 
     switch (dctx->stage)
@@ -2101,9 +801,9 @@ size_t ZSTD_decompressContinue(ZSTD_DCtx* dctx, void* dst, size_t dstCapacity, c
         assert(src != NULL);
         if (dctx->format == ZSTD_f_zstd1) {  /* allows header */
             assert(srcSize >= ZSTD_FRAMEIDSIZE);  /* to read skippable magic number */
-            if ((MEM_readLE32(src) & 0xFFFFFFF0U) == ZSTD_MAGIC_SKIPPABLE_START) {        /* skippable frame */
+            if ((MEM_readLE32(src) & ZSTD_MAGIC_SKIPPABLE_MASK) == ZSTD_MAGIC_SKIPPABLE_START) {        /* skippable frame */
                 memcpy(dctx->headerBuffer, src, srcSize);
-                dctx->expected = ZSTD_skippableHeaderSize - srcSize;  /* remaining to load to get full skippable frame header */
+                dctx->expected = ZSTD_SKIPPABLEHEADERSIZE - srcSize;  /* remaining to load to get full skippable frame header */
                 dctx->stage = ZSTDds_decodeSkippableHeader;
                 return 0;
         }   }
@@ -2163,19 +863,19 @@ size_t ZSTD_decompressContinue(ZSTD_DCtx* dctx, void* dst, size_t dstCapacity, c
                 rSize = ZSTD_copyRawBlock(dst, dstCapacity, src, srcSize);
                 break;
             case bt_rle :
-                rSize = ZSTD_setRleBlock(dst, dstCapacity, src, srcSize, dctx->rleSize);
+                rSize = ZSTD_setRleBlock(dst, dstCapacity, *(const BYTE*)src, dctx->rleSize);
                 break;
             case bt_reserved :   /* should never happen */
             default:
                 return ERROR(corruption_detected);
             }
             if (ZSTD_isError(rSize)) return rSize;
-            DEBUGLOG(5, "ZSTD_decompressContinue: decoded size from block : %u", (U32)rSize);
+            DEBUGLOG(5, "ZSTD_decompressContinue: decoded size from block : %u", (unsigned)rSize);
             dctx->decodedSize += rSize;
             if (dctx->fParams.checksumFlag) XXH64_update(&dctx->xxhState, dst, rSize);
 
             if (dctx->stage == ZSTDds_decompressLastBlock) {   /* end of frame */
-                DEBUGLOG(4, "ZSTD_decompressContinue: decoded size from frame : %u", (U32)dctx->decodedSize);
+                DEBUGLOG(4, "ZSTD_decompressContinue: decoded size from frame : %u", (unsigned)dctx->decodedSize);
                 if (dctx->fParams.frameContentSize != ZSTD_CONTENTSIZE_UNKNOWN) {
                     if (dctx->decodedSize != dctx->fParams.frameContentSize) {
                         return ERROR(corruption_detected);
@@ -2199,7 +899,7 @@ size_t ZSTD_decompressContinue(ZSTD_DCtx* dctx, void* dst, size_t dstCapacity, c
         assert(srcSize == 4);  /* guaranteed by dctx->expected */
         {   U32 const h32 = (U32)XXH64_digest(&dctx->xxhState);
             U32 const check32 = MEM_readLE32(src);
-            DEBUGLOG(4, "ZSTD_decompressContinue: checksum : calculated %08X :: %08X read", h32, check32);
+            DEBUGLOG(4, "ZSTD_decompressContinue: checksum : calculated %08X :: %08X read", (unsigned)h32, (unsigned)check32);
             if (check32 != h32) return ERROR(checksum_wrong);
             dctx->expected = 0;
             dctx->stage = ZSTDds_getFrameHeaderSize;
@@ -2208,8 +908,8 @@ size_t ZSTD_decompressContinue(ZSTD_DCtx* dctx, void* dst, size_t dstCapacity, c
 
     case ZSTDds_decodeSkippableHeader:
         assert(src != NULL);
-        assert(srcSize <= ZSTD_skippableHeaderSize);
-        memcpy(dctx->headerBuffer + (ZSTD_skippableHeaderSize - srcSize), src, srcSize);   /* complete skippable header */
+        assert(srcSize <= ZSTD_SKIPPABLEHEADERSIZE);
+        memcpy(dctx->headerBuffer + (ZSTD_SKIPPABLEHEADERSIZE - srcSize), src, srcSize);   /* complete skippable header */
         dctx->expected = MEM_readLE32(dctx->headerBuffer + ZSTD_FRAMEIDSIZE);   /* note : dctx->expected can grow seriously large, beyond local buffer size */
         dctx->stage = ZSTDds_skipFrame;
         return 0;
@@ -2220,7 +920,8 @@ size_t ZSTD_decompressContinue(ZSTD_DCtx* dctx, void* dst, size_t dstCapacity, c
         return 0;
 
     default:
-        return ERROR(GENERIC);   /* impossible */
+        assert(0);   /* impossible */
+        return ERROR(GENERIC);   /* some compiler require default to do something */
     }
 }
 
@@ -2234,11 +935,12 @@ static size_t ZSTD_refDictContent(ZSTD_DCtx* dctx, const void* dict, size_t dict
     return 0;
 }
 
-/*! ZSTD_loadEntropy() :
+/*! ZSTD_loadDEntropy() :
  *  dict : must point at beginning of a valid zstd dictionary.
  * @return : size of entropy tables read */
-static size_t ZSTD_loadEntropy(ZSTD_entropyDTables_t* entropy,
-                         const void* const dict, size_t const dictSize)
+size_t
+ZSTD_loadDEntropy(ZSTD_entropyDTables_t* entropy,
+                  const void* const dict, size_t const dictSize)
 {
     const BYTE* dictPtr = (const BYTE*)dict;
     const BYTE* const dictEnd = dictPtr + dictSize;
@@ -2252,15 +954,22 @@ static size_t ZSTD_loadEntropy(ZSTD_entropyDTables_t* entropy,
     ZSTD_STATIC_ASSERT(sizeof(entropy->LLTable) + sizeof(entropy->OFTable) + sizeof(entropy->MLTable) >= HUF_DECOMPRESS_WORKSPACE_SIZE);
     {   void* const workspace = &entropy->LLTable;   /* use fse tables as temporary workspace; implies fse tables are grouped together */
         size_t const workspaceSize = sizeof(entropy->LLTable) + sizeof(entropy->OFTable) + sizeof(entropy->MLTable);
+#ifdef HUF_FORCE_DECOMPRESS_X1
+        /* in minimal huffman, we always use X1 variants */
+        size_t const hSize = HUF_readDTableX1_wksp(entropy->hufTable,
+                                                dictPtr, dictEnd - dictPtr,
+                                                workspace, workspaceSize);
+#else
         size_t const hSize = HUF_readDTableX2_wksp(entropy->hufTable,
                                                 dictPtr, dictEnd - dictPtr,
                                                 workspace, workspaceSize);
+#endif
         if (HUF_isError(hSize)) return ERROR(dictionary_corrupted);
         dictPtr += hSize;
     }
 
     {   short offcodeNCount[MaxOff+1];
-        U32 offcodeMaxValue = MaxOff, offcodeLog;
+        unsigned offcodeMaxValue = MaxOff, offcodeLog;
         size_t const offcodeHeaderSize = FSE_readNCount(offcodeNCount, &offcodeMaxValue, &offcodeLog, dictPtr, dictEnd-dictPtr);
         if (FSE_isError(offcodeHeaderSize)) return ERROR(dictionary_corrupted);
         if (offcodeMaxValue > MaxOff) return ERROR(dictionary_corrupted);
@@ -2320,7 +1029,7 @@ static size_t ZSTD_decompress_insertDictionary(ZSTD_DCtx* dctx, const void* dict
     dctx->dictID = MEM_readLE32((const char*)dict + ZSTD_FRAMEIDSIZE);
 
     /* load entropy tables */
-    {   size_t const eSize = ZSTD_loadEntropy(&dctx->entropy, dict, dictSize);
+    {   size_t const eSize = ZSTD_loadDEntropy(&dctx->entropy, dict, dictSize);
         if (ZSTD_isError(eSize)) return ERROR(dictionary_corrupted);
         dict = (const char*)dict + eSize;
         dictSize -= eSize;
@@ -2364,209 +1073,25 @@ size_t ZSTD_decompressBegin_usingDict(ZSTD_DCtx* dctx, const void* dict, size_t
 
 /* ======   ZSTD_DDict   ====== */
 
-struct ZSTD_DDict_s {
-    void* dictBuffer;
-    const void* dictContent;
-    size_t dictSize;
-    ZSTD_entropyDTables_t entropy;
-    U32 dictID;
-    U32 entropyPresent;
-    ZSTD_customMem cMem;
-};  /* typedef'd to ZSTD_DDict within "zstd.h" */
-
-static const void* ZSTD_DDictDictContent(const ZSTD_DDict* ddict)
-{
-    assert(ddict != NULL);
-    return ddict->dictContent;
-}
-
-static size_t ZSTD_DDictDictSize(const ZSTD_DDict* ddict)
-{
-    assert(ddict != NULL);
-    return ddict->dictSize;
-}
-
 size_t ZSTD_decompressBegin_usingDDict(ZSTD_DCtx* dctx, const ZSTD_DDict* ddict)
 {
     DEBUGLOG(4, "ZSTD_decompressBegin_usingDDict");
     assert(dctx != NULL);
     if (ddict) {
-        dctx->ddictIsCold = (dctx->dictEnd != (const char*)ddict->dictContent + ddict->dictSize);
+        const char* const dictStart = (const char*)ZSTD_DDict_dictContent(ddict);
+        size_t const dictSize = ZSTD_DDict_dictSize(ddict);
+        const void* const dictEnd = dictStart + dictSize;
+        dctx->ddictIsCold = (dctx->dictEnd != dictEnd);
         DEBUGLOG(4, "DDict is %s",
                     dctx->ddictIsCold ? "~cold~" : "hot!");
     }
     CHECK_F( ZSTD_decompressBegin(dctx) );
     if (ddict) {   /* NULL ddict is equivalent to no dictionary */
-        dctx->dictID = ddict->dictID;
-        dctx->prefixStart = ddict->dictContent;
-        dctx->virtualStart = ddict->dictContent;
-        dctx->dictEnd = (const BYTE*)ddict->dictContent + ddict->dictSize;
-        dctx->previousDstEnd = dctx->dictEnd;
-        if (ddict->entropyPresent) {
-            dctx->litEntropy = 1;
-            dctx->fseEntropy = 1;
-            dctx->LLTptr = ddict->entropy.LLTable;
-            dctx->MLTptr = ddict->entropy.MLTable;
-            dctx->OFTptr = ddict->entropy.OFTable;
-            dctx->HUFptr = ddict->entropy.hufTable;
-            dctx->entropy.rep[0] = ddict->entropy.rep[0];
-            dctx->entropy.rep[1] = ddict->entropy.rep[1];
-            dctx->entropy.rep[2] = ddict->entropy.rep[2];
-        } else {
-            dctx->litEntropy = 0;
-            dctx->fseEntropy = 0;
-        }
+        ZSTD_copyDDictParameters(dctx, ddict);
     }
     return 0;
 }
 
-static size_t
-ZSTD_loadEntropy_inDDict(ZSTD_DDict* ddict,
-                         ZSTD_dictContentType_e dictContentType)
-{
-    ddict->dictID = 0;
-    ddict->entropyPresent = 0;
-    if (dictContentType == ZSTD_dct_rawContent) return 0;
-
-    if (ddict->dictSize < 8) {
-        if (dictContentType == ZSTD_dct_fullDict)
-            return ERROR(dictionary_corrupted);   /* only accept specified dictionaries */
-        return 0;   /* pure content mode */
-    }
-    {   U32 const magic = MEM_readLE32(ddict->dictContent);
-        if (magic != ZSTD_MAGIC_DICTIONARY) {
-            if (dictContentType == ZSTD_dct_fullDict)
-                return ERROR(dictionary_corrupted);   /* only accept specified dictionaries */
-            return 0;   /* pure content mode */
-        }
-    }
-    ddict->dictID = MEM_readLE32((const char*)ddict->dictContent + ZSTD_FRAMEIDSIZE);
-
-    /* load entropy tables */
-    CHECK_E( ZSTD_loadEntropy(&ddict->entropy,
-                              ddict->dictContent, ddict->dictSize),
-             dictionary_corrupted );
-    ddict->entropyPresent = 1;
-    return 0;
-}
-
-
-static size_t ZSTD_initDDict_internal(ZSTD_DDict* ddict,
-                                      const void* dict, size_t dictSize,
-                                      ZSTD_dictLoadMethod_e dictLoadMethod,
-                                      ZSTD_dictContentType_e dictContentType)
-{
-    if ((dictLoadMethod == ZSTD_dlm_byRef) || (!dict) || (!dictSize)) {
-        ddict->dictBuffer = NULL;
-        ddict->dictContent = dict;
-        if (!dict) dictSize = 0;
-    } else {
-        void* const internalBuffer = ZSTD_malloc(dictSize, ddict->cMem);
-        ddict->dictBuffer = internalBuffer;
-        ddict->dictContent = internalBuffer;
-        if (!internalBuffer) return ERROR(memory_allocation);
-        memcpy(internalBuffer, dict, dictSize);
-    }
-    ddict->dictSize = dictSize;
-    ddict->entropy.hufTable[0] = (HUF_DTable)((HufLog)*0x1000001);  /* cover both little and big endian */
-
-    /* parse dictionary content */
-    CHECK_F( ZSTD_loadEntropy_inDDict(ddict, dictContentType) );
-
-    return 0;
-}
-
-ZSTD_DDict* ZSTD_createDDict_advanced(const void* dict, size_t dictSize,
-                                      ZSTD_dictLoadMethod_e dictLoadMethod,
-                                      ZSTD_dictContentType_e dictContentType,
-                                      ZSTD_customMem customMem)
-{
-    if (!customMem.customAlloc ^ !customMem.customFree) return NULL;
-
-    {   ZSTD_DDict* const ddict = (ZSTD_DDict*) ZSTD_malloc(sizeof(ZSTD_DDict), customMem);
-        if (ddict == NULL) return NULL;
-        ddict->cMem = customMem;
-        {   size_t const initResult = ZSTD_initDDict_internal(ddict,
-                                            dict, dictSize,
-                                            dictLoadMethod, dictContentType);
-            if (ZSTD_isError(initResult)) {
-                ZSTD_freeDDict(ddict);
-                return NULL;
-        }   }
-        return ddict;
-    }
-}
-
-/*! ZSTD_createDDict() :
-*   Create a digested dictionary, to start decompression without startup delay.
-*   `dict` content is copied inside DDict.
-*   Consequently, `dict` can be released after `ZSTD_DDict` creation */
-ZSTD_DDict* ZSTD_createDDict(const void* dict, size_t dictSize)
-{
-    ZSTD_customMem const allocator = { NULL, NULL, NULL };
-    return ZSTD_createDDict_advanced(dict, dictSize, ZSTD_dlm_byCopy, ZSTD_dct_auto, allocator);
-}
-
-/*! ZSTD_createDDict_byReference() :
- *  Create a digested dictionary, to start decompression without startup delay.
- *  Dictionary content is simply referenced, it will be accessed during decompression.
- *  Warning : dictBuffer must outlive DDict (DDict must be freed before dictBuffer) */
-ZSTD_DDict* ZSTD_createDDict_byReference(const void* dictBuffer, size_t dictSize)
-{
-    ZSTD_customMem const allocator = { NULL, NULL, NULL };
-    return ZSTD_createDDict_advanced(dictBuffer, dictSize, ZSTD_dlm_byRef, ZSTD_dct_auto, allocator);
-}
-
-
-const ZSTD_DDict* ZSTD_initStaticDDict(
-                                void* sBuffer, size_t sBufferSize,
-                                const void* dict, size_t dictSize,
-                                ZSTD_dictLoadMethod_e dictLoadMethod,
-                                ZSTD_dictContentType_e dictContentType)
-{
-    size_t const neededSpace = sizeof(ZSTD_DDict)
-                             + (dictLoadMethod == ZSTD_dlm_byRef ? 0 : dictSize);
-    ZSTD_DDict* const ddict = (ZSTD_DDict*)sBuffer;
-    assert(sBuffer != NULL);
-    assert(dict != NULL);
-    if ((size_t)sBuffer & 7) return NULL;   /* 8-aligned */
-    if (sBufferSize < neededSpace) return NULL;
-    if (dictLoadMethod == ZSTD_dlm_byCopy) {
-        memcpy(ddict+1, dict, dictSize);  /* local copy */
-        dict = ddict+1;
-    }
-    if (ZSTD_isError( ZSTD_initDDict_internal(ddict,
-                                              dict, dictSize,
-                                              ZSTD_dlm_byRef, dictContentType) ))
-        return NULL;
-    return ddict;
-}
-
-
-size_t ZSTD_freeDDict(ZSTD_DDict* ddict)
-{
-    if (ddict==NULL) return 0;   /* support free on NULL */
-    {   ZSTD_customMem const cMem = ddict->cMem;
-        ZSTD_free(ddict->dictBuffer, cMem);
-        ZSTD_free(ddict, cMem);
-        return 0;
-    }
-}
-
-/*! ZSTD_estimateDDictSize() :
- *  Estimate amount of memory that will be needed to create a dictionary for decompression.
- *  Note : dictionary created by reference using ZSTD_dlm_byRef are smaller */
-size_t ZSTD_estimateDDictSize(size_t dictSize, ZSTD_dictLoadMethod_e dictLoadMethod)
-{
-    return sizeof(ZSTD_DDict) + (dictLoadMethod == ZSTD_dlm_byRef ? 0 : dictSize);
-}
-
-size_t ZSTD_sizeof_DDict(const ZSTD_DDict* ddict)
-{
-    if (ddict==NULL) return 0;   /* support sizeof on NULL */
-    return sizeof(*ddict) + (ddict->dictBuffer ? ddict->dictSize : 0) ;
-}
-
 /*! ZSTD_getDictID_fromDict() :
  *  Provides the dictID stored within dictionary.
  *  if @return == 0, the dictionary is not conformant with Zstandard specification.
@@ -2578,16 +1103,6 @@ unsigned ZSTD_getDictID_fromDict(const void* dict, size_t dictSize)
     return MEM_readLE32((const char*)dict + ZSTD_FRAMEIDSIZE);
 }
 
-/*! ZSTD_getDictID_fromDDict() :
- *  Provides the dictID of the dictionary loaded into `ddict`.
- *  If @return == 0, the dictionary is not conformant to Zstandard specification, or empty.
- *  Non-conformant dictionaries can still be loaded, but as content-only dictionaries. */
-unsigned ZSTD_getDictID_fromDDict(const ZSTD_DDict* ddict)
-{
-    if (ddict==NULL) return 0;
-    return ZSTD_getDictID_fromDict(ddict->dictContent, ddict->dictSize);
-}
-
 /*! ZSTD_getDictID_fromFrame() :
  *  Provides the dictID required to decompresse frame stored within `src`.
  *  If @return == 0, the dictID could not be decoded.
@@ -2695,7 +1210,7 @@ size_t ZSTD_DCtx_refPrefix(ZSTD_DCtx* dctx, const void* prefix, size_t prefixSiz
 
 
 /* ZSTD_initDStream_usingDict() :
- * return : expected size, aka ZSTD_frameHeaderSize_prefix.
+ * return : expected size, aka ZSTD_FRAMEHEADERSIZE_PREFIX.
  * this function cannot fail */
 size_t ZSTD_initDStream_usingDict(ZSTD_DStream* zds, const void* dict, size_t dictSize)
 {
@@ -2703,7 +1218,7 @@ size_t ZSTD_initDStream_usingDict(ZSTD_DStream* zds, const void* dict, size_t di
     zds->streamStage = zdss_init;
     zds->noForwardProgress = 0;
     CHECK_F( ZSTD_DCtx_loadDictionary(zds, dict, dictSize) );
-    return ZSTD_frameHeaderSize_prefix;
+    return ZSTD_FRAMEHEADERSIZE_PREFIX;
 }
 
 /* note : this variant can't fail */
@@ -2724,7 +1239,7 @@ size_t ZSTD_initDStream_usingDDict(ZSTD_DStream* dctx, const ZSTD_DDict* ddict)
 }
 
 /* ZSTD_resetDStream() :
- * return : expected size, aka ZSTD_frameHeaderSize_prefix.
+ * return : expected size, aka ZSTD_FRAMEHEADERSIZE_PREFIX.
  * this function cannot fail */
 size_t ZSTD_resetDStream(ZSTD_DStream* dctx)
 {
@@ -2733,23 +1248,9 @@ size_t ZSTD_resetDStream(ZSTD_DStream* dctx)
     dctx->lhSize = dctx->inPos = dctx->outStart = dctx->outEnd = 0;
     dctx->legacyVersion = 0;
     dctx->hostageByte = 0;
-    return ZSTD_frameHeaderSize_prefix;
+    return ZSTD_FRAMEHEADERSIZE_PREFIX;
 }
 
-size_t ZSTD_setDStreamParameter(ZSTD_DStream* dctx,
-                                ZSTD_DStreamParameter_e paramType, unsigned paramValue)
-{
-    if (dctx->streamStage != zdss_init) return ERROR(stage_wrong);
-    switch(paramType)
-    {
-        default : return ERROR(parameter_unsupported);
-        case DStream_p_maxWindowSize :
-            DEBUGLOG(4, "setting maxWindowSize = %u KB", paramValue >> 10);
-            dctx->maxWindowSize = paramValue ? paramValue : (U32)(-1);
-            break;
-    }
-    return 0;
-}
 
 size_t ZSTD_DCtx_refDDict(ZSTD_DCtx* dctx, const ZSTD_DDict* ddict)
 {
@@ -2758,18 +1259,92 @@ size_t ZSTD_DCtx_refDDict(ZSTD_DCtx* dctx, const ZSTD_DDict* ddict)
     return 0;
 }
 
+/* ZSTD_DCtx_setMaxWindowSize() :
+ * note : no direct equivalence in ZSTD_DCtx_setParameter,
+ * since this version sets windowSize, and the other sets windowLog */
 size_t ZSTD_DCtx_setMaxWindowSize(ZSTD_DCtx* dctx, size_t maxWindowSize)
 {
+    ZSTD_bounds const bounds = ZSTD_dParam_getBounds(ZSTD_d_windowLogMax);
+    size_t const min = (size_t)1 << bounds.lowerBound;
+    size_t const max = (size_t)1 << bounds.upperBound;
     if (dctx->streamStage != zdss_init) return ERROR(stage_wrong);
+    if (maxWindowSize < min) return ERROR(parameter_outOfBound);
+    if (maxWindowSize > max) return ERROR(parameter_outOfBound);
     dctx->maxWindowSize = maxWindowSize;
     return 0;
 }
 
 size_t ZSTD_DCtx_setFormat(ZSTD_DCtx* dctx, ZSTD_format_e format)
 {
-    DEBUGLOG(4, "ZSTD_DCtx_setFormat : %u", (unsigned)format);
+    return ZSTD_DCtx_setParameter(dctx, ZSTD_d_format, format);
+}
+
+ZSTD_bounds ZSTD_dParam_getBounds(ZSTD_dParameter dParam)
+{
+    ZSTD_bounds bounds = { 0, 0, 0 };
+    switch(dParam) {
+        case ZSTD_d_windowLogMax:
+            bounds.lowerBound = ZSTD_WINDOWLOG_ABSOLUTEMIN;
+            bounds.upperBound = ZSTD_WINDOWLOG_MAX;
+            return bounds;
+        case ZSTD_d_format:
+            bounds.lowerBound = (int)ZSTD_f_zstd1;
+            bounds.upperBound = (int)ZSTD_f_zstd1_magicless;
+            ZSTD_STATIC_ASSERT(ZSTD_f_zstd1 < ZSTD_f_zstd1_magicless);
+            return bounds;
+        default:;
+    }
+    bounds.error = ERROR(parameter_unsupported);
+    return bounds;
+}
+
+/* ZSTD_dParam_withinBounds:
+ * @return 1 if value is within dParam bounds,
+ * 0 otherwise */
+static int ZSTD_dParam_withinBounds(ZSTD_dParameter dParam, int value)
+{
+    ZSTD_bounds const bounds = ZSTD_dParam_getBounds(dParam);
+    if (ZSTD_isError(bounds.error)) return 0;
+    if (value < bounds.lowerBound) return 0;
+    if (value > bounds.upperBound) return 0;
+    return 1;
+}
+
+#define CHECK_DBOUNDS(p,v) {                \
+    if (!ZSTD_dParam_withinBounds(p, v))    \
+        return ERROR(parameter_outOfBound); \
+}
+
+size_t ZSTD_DCtx_setParameter(ZSTD_DCtx* dctx, ZSTD_dParameter dParam, int value)
+{
     if (dctx->streamStage != zdss_init) return ERROR(stage_wrong);
-    dctx->format = format;
+    switch(dParam) {
+        case ZSTD_d_windowLogMax:
+            CHECK_DBOUNDS(ZSTD_d_windowLogMax, value);
+            dctx->maxWindowSize = ((size_t)1) << value;
+            return 0;
+        case ZSTD_d_format:
+            CHECK_DBOUNDS(ZSTD_d_format, value);
+            dctx->format = (ZSTD_format_e)value;
+            return 0;
+        default:;
+    }
+    return ERROR(parameter_unsupported);
+}
+
+size_t ZSTD_DCtx_reset(ZSTD_DCtx* dctx, ZSTD_ResetDirective reset)
+{
+    if ( (reset == ZSTD_reset_session_only)
+      || (reset == ZSTD_reset_session_and_parameters) ) {
+        (void)ZSTD_initDStream(dctx);
+    }
+    if ( (reset == ZSTD_reset_parameters)
+      || (reset == ZSTD_reset_session_and_parameters) ) {
+        if (dctx->streamStage != zdss_init)
+            return ERROR(stage_wrong);
+        dctx->format = ZSTD_f_zstd1;
+        dctx->maxWindowSize = ZSTD_MAXWINDOWSIZE_DEFAULT;
+    }
     return 0;
 }
 
@@ -2799,7 +1374,7 @@ size_t ZSTD_estimateDStreamSize(size_t windowSize)
 
 size_t ZSTD_estimateDStreamSize_fromFrame(const void* src, size_t srcSize)
 {
-    U32 const windowSizeMax = 1U << ZSTD_WINDOWLOG_MAX;   /* note : should be user-selectable */
+    U32 const windowSizeMax = 1U << ZSTD_WINDOWLOG_MAX;   /* note : should be user-selectable, but requires an additional parameter (or a dctx) */
     ZSTD_frameHeader zfh;
     size_t const err = ZSTD_getFrameHeader(&zfh, src, srcSize);
     if (ZSTD_isError(err)) return err;
@@ -2868,8 +1443,8 @@ size_t ZSTD_decompressStream(ZSTD_DStream* zds, ZSTD_outBuffer* output, ZSTD_inB
 #if defined(ZSTD_LEGACY_SUPPORT) && (ZSTD_LEGACY_SUPPORT>=1)
                     U32 const legacyVersion = ZSTD_isLegacy(istart, iend-istart);
                     if (legacyVersion) {
-                        const void* const dict = zds->ddict ? zds->ddict->dictContent : NULL;
-                        size_t const dictSize = zds->ddict ? zds->ddict->dictSize : 0;
+                        const void* const dict = zds->ddict ? ZSTD_DDict_dictContent(zds->ddict) : NULL;
+                        size_t const dictSize = zds->ddict ? ZSTD_DDict_dictSize(zds->ddict) : 0;
                         DEBUGLOG(5, "ZSTD_decompressStream: detected legacy version v0.%u", legacyVersion);
                         /* legacy support is incompatible with static dctx */
                         if (zds->staticSize) return ERROR(memory_allocation);
@@ -2894,7 +1469,7 @@ size_t ZSTD_decompressStream(ZSTD_DStream* zds, ZSTD_outBuffer* output, ZSTD_inB
                             zds->lhSize += remainingInput;
                         }
                         input->pos = input->size;
-                        return (MAX(ZSTD_frameHeaderSize_min, hSize) - zds->lhSize) + ZSTD_blockHeaderSize;   /* remaining header bytes + next block header */
+                        return (MAX(ZSTD_FRAMEHEADERSIZE_MIN, hSize) - zds->lhSize) + ZSTD_blockHeaderSize;   /* remaining header bytes + next block header */
                     }
                     assert(ip != NULL);
                     memcpy(zds->headerBuffer + zds->lhSize, ip, toLoad); zds->lhSize = hSize; ip += toLoad;
@@ -2922,7 +1497,7 @@ size_t ZSTD_decompressStream(ZSTD_DStream* zds, ZSTD_outBuffer* output, ZSTD_inB
             DEBUGLOG(4, "Consume header");
             CHECK_F(ZSTD_decompressBegin_usingDDict(zds, zds->ddict));
 
-            if ((MEM_readLE32(zds->headerBuffer) & 0xFFFFFFF0U) == ZSTD_MAGIC_SKIPPABLE_START) {  /* skippable frame */
+            if ((MEM_readLE32(zds->headerBuffer) & ZSTD_MAGIC_SKIPPABLE_MASK) == ZSTD_MAGIC_SKIPPABLE_START) {  /* skippable frame */
                 zds->expected = MEM_readLE32(zds->headerBuffer + ZSTD_FRAMEIDSIZE);
                 zds->stage = ZSTDds_skipFrame;
             } else {
@@ -3038,7 +1613,9 @@ size_t ZSTD_decompressStream(ZSTD_DStream* zds, ZSTD_outBuffer* output, ZSTD_inB
             someMoreWork = 0;
             break;
 
-        default: return ERROR(GENERIC);   /* impossible */
+        default:
+            assert(0);    /* impossible */
+            return ERROR(GENERIC);   /* some compiler require default to do something */
     }   }
 
     /* result */
@@ -3080,13 +1657,7 @@ size_t ZSTD_decompressStream(ZSTD_DStream* zds, ZSTD_outBuffer* output, ZSTD_inB
     }
 }
 
-
-size_t ZSTD_decompress_generic(ZSTD_DCtx* dctx, ZSTD_outBuffer* output, ZSTD_inBuffer* input)
-{
-    return ZSTD_decompressStream(dctx, output, input);
-}
-
-size_t ZSTD_decompress_generic_simpleArgs (
+size_t ZSTD_decompressStream_simpleArgs (
                             ZSTD_DCtx* dctx,
                             void* dst, size_t dstCapacity, size_t* dstPos,
                       const void* src, size_t srcSize, size_t* srcPos)
@@ -3094,15 +1665,8 @@ size_t ZSTD_decompress_generic_simpleArgs (
     ZSTD_outBuffer output = { dst, dstCapacity, *dstPos };
     ZSTD_inBuffer  input  = { src, srcSize, *srcPos };
     /* ZSTD_compress_generic() will check validity of dstPos and srcPos */
-    size_t const cErr = ZSTD_decompress_generic(dctx, &output, &input);
+    size_t const cErr = ZSTD_decompressStream(dctx, &output, &input);
     *dstPos = output.pos;
     *srcPos = input.pos;
     return cErr;
 }
-
-void ZSTD_DCtx_reset(ZSTD_DCtx* dctx)
-{
-    (void)ZSTD_initDStream(dctx);
-    dctx->format = ZSTD_f_zstd1;
-    dctx->maxWindowSize = ZSTD_MAXWINDOWSIZE_DEFAULT;
-}
diff --git a/lib/decompress/zstd_decompress_block.c b/lib/decompress/zstd_decompress_block.c
new file mode 100644
index 000000000000..32baad9fbb5c
--- /dev/null
+++ b/lib/decompress/zstd_decompress_block.c
@@ -0,0 +1,1307 @@
+/*
+ * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
+ * You may select, at your option, one of the above-listed licenses.
+ */
+
+/* zstd_decompress_block :
+ * this module takes care of decompressing _compressed_ block */
+
+/*-*******************************************************
+*  Dependencies
+*********************************************************/
+#include       /* memcpy, memmove, memset */
+#include "compiler.h"    /* prefetch */
+#include "cpu.h"         /* bmi2 */
+#include "mem.h"         /* low level memory routines */
+#define FSE_STATIC_LINKING_ONLY
+#include "fse.h"
+#define HUF_STATIC_LINKING_ONLY
+#include "huf.h"
+#include "zstd_internal.h"
+#include "zstd_decompress_internal.h"   /* ZSTD_DCtx */
+#include "zstd_ddict.h"  /* ZSTD_DDictDictContent */
+#include "zstd_decompress_block.h"
+
+/*_*******************************************************
+*  Macros
+**********************************************************/
+
+/* These two optional macros force the use one way or another of the two
+ * ZSTD_decompressSequences implementations. You can't force in both directions
+ * at the same time.
+ */
+#if defined(ZSTD_FORCE_DECOMPRESS_SEQUENCES_SHORT) && \
+    defined(ZSTD_FORCE_DECOMPRESS_SEQUENCES_LONG)
+#error "Cannot force the use of the short and the long ZSTD_decompressSequences variants!"
+#endif
+
+
+/*_*******************************************************
+*  Memory operations
+**********************************************************/
+static void ZSTD_copy4(void* dst, const void* src) { memcpy(dst, src, 4); }
+
+
+/*-*************************************************************
+ *   Block decoding
+ ***************************************************************/
+
+/*! ZSTD_getcBlockSize() :
+ *  Provides the size of compressed block from block header `src` */
+size_t ZSTD_getcBlockSize(const void* src, size_t srcSize,
+                          blockProperties_t* bpPtr)
+{
+    if (srcSize < ZSTD_blockHeaderSize) return ERROR(srcSize_wrong);
+    {   U32 const cBlockHeader = MEM_readLE24(src);
+        U32 const cSize = cBlockHeader >> 3;
+        bpPtr->lastBlock = cBlockHeader & 1;
+        bpPtr->blockType = (blockType_e)((cBlockHeader >> 1) & 3);
+        bpPtr->origSize = cSize;   /* only useful for RLE */
+        if (bpPtr->blockType == bt_rle) return 1;
+        if (bpPtr->blockType == bt_reserved) return ERROR(corruption_detected);
+        return cSize;
+    }
+}
+
+
+/* Hidden declaration for fullbench */
+size_t ZSTD_decodeLiteralsBlock(ZSTD_DCtx* dctx,
+                          const void* src, size_t srcSize);
+/*! ZSTD_decodeLiteralsBlock() :
+ * @return : nb of bytes read from src (< srcSize )
+ *  note : symbol not declared but exposed for fullbench */
+size_t ZSTD_decodeLiteralsBlock(ZSTD_DCtx* dctx,
+                          const void* src, size_t srcSize)   /* note : srcSize < BLOCKSIZE */
+{
+    if (srcSize < MIN_CBLOCK_SIZE) return ERROR(corruption_detected);
+
+    {   const BYTE* const istart = (const BYTE*) src;
+        symbolEncodingType_e const litEncType = (symbolEncodingType_e)(istart[0] & 3);
+
+        switch(litEncType)
+        {
+        case set_repeat:
+            if (dctx->litEntropy==0) return ERROR(dictionary_corrupted);
+            /* fall-through */
+
+        case set_compressed:
+            if (srcSize < 5) return ERROR(corruption_detected);   /* srcSize >= MIN_CBLOCK_SIZE == 3; here we need up to 5 for case 3 */
+            {   size_t lhSize, litSize, litCSize;
+                U32 singleStream=0;
+                U32 const lhlCode = (istart[0] >> 2) & 3;
+                U32 const lhc = MEM_readLE32(istart);
+                size_t hufSuccess;
+                switch(lhlCode)
+                {
+                case 0: case 1: default:   /* note : default is impossible, since lhlCode into [0..3] */
+                    /* 2 - 2 - 10 - 10 */
+                    singleStream = !lhlCode;
+                    lhSize = 3;
+                    litSize  = (lhc >> 4) & 0x3FF;
+                    litCSize = (lhc >> 14) & 0x3FF;
+                    break;
+                case 2:
+                    /* 2 - 2 - 14 - 14 */
+                    lhSize = 4;
+                    litSize  = (lhc >> 4) & 0x3FFF;
+                    litCSize = lhc >> 18;
+                    break;
+                case 3:
+                    /* 2 - 2 - 18 - 18 */
+                    lhSize = 5;
+                    litSize  = (lhc >> 4) & 0x3FFFF;
+                    litCSize = (lhc >> 22) + (istart[4] << 10);
+                    break;
+                }
+                if (litSize > ZSTD_BLOCKSIZE_MAX) return ERROR(corruption_detected);
+                if (litCSize + lhSize > srcSize) return ERROR(corruption_detected);
+
+                /* prefetch huffman table if cold */
+                if (dctx->ddictIsCold && (litSize > 768 /* heuristic */)) {
+                    PREFETCH_AREA(dctx->HUFptr, sizeof(dctx->entropy.hufTable));
+                }
+
+                if (litEncType==set_repeat) {
+                    if (singleStream) {
+                        hufSuccess = HUF_decompress1X_usingDTable_bmi2(
+                            dctx->litBuffer, litSize, istart+lhSize, litCSize,
+                            dctx->HUFptr, dctx->bmi2);
+                    } else {
+                        hufSuccess = HUF_decompress4X_usingDTable_bmi2(
+                            dctx->litBuffer, litSize, istart+lhSize, litCSize,
+                            dctx->HUFptr, dctx->bmi2);
+                    }
+                } else {
+                    if (singleStream) {
+#if defined(HUF_FORCE_DECOMPRESS_X2)
+                        hufSuccess = HUF_decompress1X_DCtx_wksp(
+                            dctx->entropy.hufTable, dctx->litBuffer, litSize,
+                            istart+lhSize, litCSize, dctx->workspace,
+                            sizeof(dctx->workspace));
+#else
+                        hufSuccess = HUF_decompress1X1_DCtx_wksp_bmi2(
+                            dctx->entropy.hufTable, dctx->litBuffer, litSize,
+                            istart+lhSize, litCSize, dctx->workspace,
+                            sizeof(dctx->workspace), dctx->bmi2);
+#endif
+                    } else {
+                        hufSuccess = HUF_decompress4X_hufOnly_wksp_bmi2(
+                            dctx->entropy.hufTable, dctx->litBuffer, litSize,
+                            istart+lhSize, litCSize, dctx->workspace,
+                            sizeof(dctx->workspace), dctx->bmi2);
+                    }
+                }
+
+                if (HUF_isError(hufSuccess)) return ERROR(corruption_detected);
+
+                dctx->litPtr = dctx->litBuffer;
+                dctx->litSize = litSize;
+                dctx->litEntropy = 1;
+                if (litEncType==set_compressed) dctx->HUFptr = dctx->entropy.hufTable;
+                memset(dctx->litBuffer + dctx->litSize, 0, WILDCOPY_OVERLENGTH);
+                return litCSize + lhSize;
+            }
+
+        case set_basic:
+            {   size_t litSize, lhSize;
+                U32 const lhlCode = ((istart[0]) >> 2) & 3;
+                switch(lhlCode)
+                {
+                case 0: case 2: default:   /* note : default is impossible, since lhlCode into [0..3] */
+                    lhSize = 1;
+                    litSize = istart[0] >> 3;
+                    break;
+                case 1:
+                    lhSize = 2;
+                    litSize = MEM_readLE16(istart) >> 4;
+                    break;
+                case 3:
+                    lhSize = 3;
+                    litSize = MEM_readLE24(istart) >> 4;
+                    break;
+                }
+
+                if (lhSize+litSize+WILDCOPY_OVERLENGTH > srcSize) {  /* risk reading beyond src buffer with wildcopy */
+                    if (litSize+lhSize > srcSize) return ERROR(corruption_detected);
+                    memcpy(dctx->litBuffer, istart+lhSize, litSize);
+                    dctx->litPtr = dctx->litBuffer;
+                    dctx->litSize = litSize;
+                    memset(dctx->litBuffer + dctx->litSize, 0, WILDCOPY_OVERLENGTH);
+                    return lhSize+litSize;
+                }
+                /* direct reference into compressed stream */
+                dctx->litPtr = istart+lhSize;
+                dctx->litSize = litSize;
+                return lhSize+litSize;
+            }
+
+        case set_rle:
+            {   U32 const lhlCode = ((istart[0]) >> 2) & 3;
+                size_t litSize, lhSize;
+                switch(lhlCode)
+                {
+                case 0: case 2: default:   /* note : default is impossible, since lhlCode into [0..3] */
+                    lhSize = 1;
+                    litSize = istart[0] >> 3;
+                    break;
+                case 1:
+                    lhSize = 2;
+                    litSize = MEM_readLE16(istart) >> 4;
+                    break;
+                case 3:
+                    lhSize = 3;
+                    litSize = MEM_readLE24(istart) >> 4;
+                    if (srcSize<4) return ERROR(corruption_detected);   /* srcSize >= MIN_CBLOCK_SIZE == 3; here we need lhSize+1 = 4 */
+                    break;
+                }
+                if (litSize > ZSTD_BLOCKSIZE_MAX) return ERROR(corruption_detected);
+                memset(dctx->litBuffer, istart[lhSize], litSize + WILDCOPY_OVERLENGTH);
+                dctx->litPtr = dctx->litBuffer;
+                dctx->litSize = litSize;
+                return lhSize+1;
+            }
+        default:
+            return ERROR(corruption_detected);   /* impossible */
+        }
+    }
+}
+
+/* Default FSE distribution tables.
+ * These are pre-calculated FSE decoding tables using default distributions as defined in specification :
+ * https://github.com/facebook/zstd/blob/master/doc/zstd_compression_format.md#default-distributions
+ * They were generated programmatically with following method :
+ * - start from default distributions, present in /lib/common/zstd_internal.h
+ * - generate tables normally, using ZSTD_buildFSETable()
+ * - printout the content of tables
+ * - pretify output, report below, test with fuzzer to ensure it's correct */
+
+/* Default FSE distribution table for Literal Lengths */
+static const ZSTD_seqSymbol LL_defaultDTable[(1<tableLog = 0;
+    DTableH->fastMode = 0;
+
+    cell->nbBits = 0;
+    cell->nextState = 0;
+    assert(nbAddBits < 255);
+    cell->nbAdditionalBits = (BYTE)nbAddBits;
+    cell->baseValue = baseValue;
+}
+
+
+/* ZSTD_buildFSETable() :
+ * generate FSE decoding table for one symbol (ll, ml or off)
+ * cannot fail if input is valid =>
+ * all inputs are presumed validated at this stage */
+void
+ZSTD_buildFSETable(ZSTD_seqSymbol* dt,
+            const short* normalizedCounter, unsigned maxSymbolValue,
+            const U32* baseValue, const U32* nbAdditionalBits,
+            unsigned tableLog)
+{
+    ZSTD_seqSymbol* const tableDecode = dt+1;
+    U16 symbolNext[MaxSeq+1];
+
+    U32 const maxSV1 = maxSymbolValue + 1;
+    U32 const tableSize = 1 << tableLog;
+    U32 highThreshold = tableSize-1;
+
+    /* Sanity Checks */
+    assert(maxSymbolValue <= MaxSeq);
+    assert(tableLog <= MaxFSELog);
+
+    /* Init, lay down lowprob symbols */
+    {   ZSTD_seqSymbol_header DTableH;
+        DTableH.tableLog = tableLog;
+        DTableH.fastMode = 1;
+        {   S16 const largeLimit= (S16)(1 << (tableLog-1));
+            U32 s;
+            for (s=0; s= largeLimit) DTableH.fastMode=0;
+                    symbolNext[s] = normalizedCounter[s];
+        }   }   }
+        memcpy(dt, &DTableH, sizeof(DTableH));
+    }
+
+    /* Spread symbols */
+    {   U32 const tableMask = tableSize-1;
+        U32 const step = FSE_TABLESTEP(tableSize);
+        U32 s, position = 0;
+        for (s=0; s highThreshold) position = (position + step) & tableMask;   /* lowprob area */
+        }   }
+        assert(position == 0); /* position must reach all cells once, otherwise normalizedCounter is incorrect */
+    }
+
+    /* Build Decoding table */
+    {   U32 u;
+        for (u=0; u max) return ERROR(corruption_detected);
+        {   U32 const symbol = *(const BYTE*)src;
+            U32 const baseline = baseValue[symbol];
+            U32 const nbBits = nbAdditionalBits[symbol];
+            ZSTD_buildSeqTable_rle(DTableSpace, baseline, nbBits);
+        }
+        *DTablePtr = DTableSpace;
+        return 1;
+    case set_basic :
+        *DTablePtr = defaultTable;
+        return 0;
+    case set_repeat:
+        if (!flagRepeatTable) return ERROR(corruption_detected);
+        /* prefetch FSE table if used */
+        if (ddictIsCold && (nbSeq > 24 /* heuristic */)) {
+            const void* const pStart = *DTablePtr;
+            size_t const pSize = sizeof(ZSTD_seqSymbol) * (SEQSYMBOL_TABLE_SIZE(maxLog));
+            PREFETCH_AREA(pStart, pSize);
+        }
+        return 0;
+    case set_compressed :
+        {   unsigned tableLog;
+            S16 norm[MaxSeq+1];
+            size_t const headerSize = FSE_readNCount(norm, &max, &tableLog, src, srcSize);
+            if (FSE_isError(headerSize)) return ERROR(corruption_detected);
+            if (tableLog > maxLog) return ERROR(corruption_detected);
+            ZSTD_buildFSETable(DTableSpace, norm, max, baseValue, nbAdditionalBits, tableLog);
+            *DTablePtr = DTableSpace;
+            return headerSize;
+        }
+    default :   /* impossible */
+        assert(0);
+        return ERROR(GENERIC);
+    }
+}
+
+size_t ZSTD_decodeSeqHeaders(ZSTD_DCtx* dctx, int* nbSeqPtr,
+                             const void* src, size_t srcSize)
+{
+    const BYTE* const istart = (const BYTE* const)src;
+    const BYTE* const iend = istart + srcSize;
+    const BYTE* ip = istart;
+    int nbSeq;
+    DEBUGLOG(5, "ZSTD_decodeSeqHeaders");
+
+    /* check */
+    if (srcSize < MIN_SEQUENCES_SIZE) return ERROR(srcSize_wrong);
+
+    /* SeqHead */
+    nbSeq = *ip++;
+    if (!nbSeq) {
+        *nbSeqPtr=0;
+        if (srcSize != 1) return ERROR(srcSize_wrong);
+        return 1;
+    }
+    if (nbSeq > 0x7F) {
+        if (nbSeq == 0xFF) {
+            if (ip+2 > iend) return ERROR(srcSize_wrong);
+            nbSeq = MEM_readLE16(ip) + LONGNBSEQ, ip+=2;
+        } else {
+            if (ip >= iend) return ERROR(srcSize_wrong);
+            nbSeq = ((nbSeq-0x80)<<8) + *ip++;
+        }
+    }
+    *nbSeqPtr = nbSeq;
+
+    /* FSE table descriptors */
+    if (ip+4 > iend) return ERROR(srcSize_wrong); /* minimum possible size */
+    {   symbolEncodingType_e const LLtype = (symbolEncodingType_e)(*ip >> 6);
+        symbolEncodingType_e const OFtype = (symbolEncodingType_e)((*ip >> 4) & 3);
+        symbolEncodingType_e const MLtype = (symbolEncodingType_e)((*ip >> 2) & 3);
+        ip++;
+
+        /* Build DTables */
+        {   size_t const llhSize = ZSTD_buildSeqTable(dctx->entropy.LLTable, &dctx->LLTptr,
+                                                      LLtype, MaxLL, LLFSELog,
+                                                      ip, iend-ip,
+                                                      LL_base, LL_bits,
+                                                      LL_defaultDTable, dctx->fseEntropy,
+                                                      dctx->ddictIsCold, nbSeq);
+            if (ZSTD_isError(llhSize)) return ERROR(corruption_detected);
+            ip += llhSize;
+        }
+
+        {   size_t const ofhSize = ZSTD_buildSeqTable(dctx->entropy.OFTable, &dctx->OFTptr,
+                                                      OFtype, MaxOff, OffFSELog,
+                                                      ip, iend-ip,
+                                                      OF_base, OF_bits,
+                                                      OF_defaultDTable, dctx->fseEntropy,
+                                                      dctx->ddictIsCold, nbSeq);
+            if (ZSTD_isError(ofhSize)) return ERROR(corruption_detected);
+            ip += ofhSize;
+        }
+
+        {   size_t const mlhSize = ZSTD_buildSeqTable(dctx->entropy.MLTable, &dctx->MLTptr,
+                                                      MLtype, MaxML, MLFSELog,
+                                                      ip, iend-ip,
+                                                      ML_base, ML_bits,
+                                                      ML_defaultDTable, dctx->fseEntropy,
+                                                      dctx->ddictIsCold, nbSeq);
+            if (ZSTD_isError(mlhSize)) return ERROR(corruption_detected);
+            ip += mlhSize;
+        }
+    }
+
+    return ip-istart;
+}
+
+
+typedef struct {
+    size_t litLength;
+    size_t matchLength;
+    size_t offset;
+    const BYTE* match;
+} seq_t;
+
+typedef struct {
+    size_t state;
+    const ZSTD_seqSymbol* table;
+} ZSTD_fseState;
+
+typedef struct {
+    BIT_DStream_t DStream;
+    ZSTD_fseState stateLL;
+    ZSTD_fseState stateOffb;
+    ZSTD_fseState stateML;
+    size_t prevOffset[ZSTD_REP_NUM];
+    const BYTE* prefixStart;
+    const BYTE* dictEnd;
+    size_t pos;
+} seqState_t;
+
+
+/* ZSTD_execSequenceLast7():
+ * exceptional case : decompress a match starting within last 7 bytes of output buffer.
+ * requires more careful checks, to ensure there is no overflow.
+ * performance does not matter though.
+ * note : this case is supposed to be never generated "naturally" by reference encoder,
+ *        since in most cases it needs at least 8 bytes to look for a match.
+ *        but it's allowed by the specification. */
+FORCE_NOINLINE
+size_t ZSTD_execSequenceLast7(BYTE* op,
+                              BYTE* const oend, seq_t sequence,
+                              const BYTE** litPtr, const BYTE* const litLimit,
+                              const BYTE* const base, const BYTE* const vBase, const BYTE* const dictEnd)
+{
+    BYTE* const oLitEnd = op + sequence.litLength;
+    size_t const sequenceLength = sequence.litLength + sequence.matchLength;
+    BYTE* const oMatchEnd = op + sequenceLength;   /* risk : address space overflow (32-bits) */
+    const BYTE* const iLitEnd = *litPtr + sequence.litLength;
+    const BYTE* match = oLitEnd - sequence.offset;
+
+    /* check */
+    if (oMatchEnd>oend) return ERROR(dstSize_tooSmall);   /* last match must fit within dstBuffer */
+    if (iLitEnd > litLimit) return ERROR(corruption_detected);   /* try to read beyond literal buffer */
+
+    /* copy literals */
+    while (op < oLitEnd) *op++ = *(*litPtr)++;
+
+    /* copy Match */
+    if (sequence.offset > (size_t)(oLitEnd - base)) {
+        /* offset beyond prefix */
+        if (sequence.offset > (size_t)(oLitEnd - vBase)) return ERROR(corruption_detected);
+        match = dictEnd - (base-match);
+        if (match + sequence.matchLength <= dictEnd) {
+            memmove(oLitEnd, match, sequence.matchLength);
+            return sequenceLength;
+        }
+        /* span extDict & currentPrefixSegment */
+        {   size_t const length1 = dictEnd - match;
+            memmove(oLitEnd, match, length1);
+            op = oLitEnd + length1;
+            sequence.matchLength -= length1;
+            match = base;
+    }   }
+    while (op < oMatchEnd) *op++ = *match++;
+    return sequenceLength;
+}
+
+
+HINT_INLINE
+size_t ZSTD_execSequence(BYTE* op,
+                         BYTE* const oend, seq_t sequence,
+                         const BYTE** litPtr, const BYTE* const litLimit,
+                         const BYTE* const prefixStart, const BYTE* const virtualStart, const BYTE* const dictEnd)
+{
+    BYTE* const oLitEnd = op + sequence.litLength;
+    size_t const sequenceLength = sequence.litLength + sequence.matchLength;
+    BYTE* const oMatchEnd = op + sequenceLength;   /* risk : address space overflow (32-bits) */
+    BYTE* const oend_w = oend - WILDCOPY_OVERLENGTH;
+    const BYTE* const iLitEnd = *litPtr + sequence.litLength;
+    const BYTE* match = oLitEnd - sequence.offset;
+
+    /* check */
+    if (oMatchEnd>oend) return ERROR(dstSize_tooSmall); /* last match must start at a minimum distance of WILDCOPY_OVERLENGTH from oend */
+    if (iLitEnd > litLimit) return ERROR(corruption_detected);   /* over-read beyond lit buffer */
+    if (oLitEnd>oend_w) return ZSTD_execSequenceLast7(op, oend, sequence, litPtr, litLimit, prefixStart, virtualStart, dictEnd);
+
+    /* copy Literals */
+    ZSTD_copy8(op, *litPtr);
+    if (sequence.litLength > 8)
+        ZSTD_wildcopy(op+8, (*litPtr)+8, sequence.litLength - 8);   /* note : since oLitEnd <= oend-WILDCOPY_OVERLENGTH, no risk of overwrite beyond oend */
+    op = oLitEnd;
+    *litPtr = iLitEnd;   /* update for next sequence */
+
+    /* copy Match */
+    if (sequence.offset > (size_t)(oLitEnd - prefixStart)) {
+        /* offset beyond prefix -> go into extDict */
+        if (sequence.offset > (size_t)(oLitEnd - virtualStart))
+            return ERROR(corruption_detected);
+        match = dictEnd + (match - prefixStart);
+        if (match + sequence.matchLength <= dictEnd) {
+            memmove(oLitEnd, match, sequence.matchLength);
+            return sequenceLength;
+        }
+        /* span extDict & currentPrefixSegment */
+        {   size_t const length1 = dictEnd - match;
+            memmove(oLitEnd, match, length1);
+            op = oLitEnd + length1;
+            sequence.matchLength -= length1;
+            match = prefixStart;
+            if (op > oend_w || sequence.matchLength < MINMATCH) {
+              U32 i;
+              for (i = 0; i < sequence.matchLength; ++i) op[i] = match[i];
+              return sequenceLength;
+            }
+    }   }
+    /* Requirement: op <= oend_w && sequence.matchLength >= MINMATCH */
+
+    /* match within prefix */
+    if (sequence.offset < 8) {
+        /* close range match, overlap */
+        static const U32 dec32table[] = { 0, 1, 2, 1, 4, 4, 4, 4 };   /* added */
+        static const int dec64table[] = { 8, 8, 8, 7, 8, 9,10,11 };   /* subtracted */
+        int const sub2 = dec64table[sequence.offset];
+        op[0] = match[0];
+        op[1] = match[1];
+        op[2] = match[2];
+        op[3] = match[3];
+        match += dec32table[sequence.offset];
+        ZSTD_copy4(op+4, match);
+        match -= sub2;
+    } else {
+        ZSTD_copy8(op, match);
+    }
+    op += 8; match += 8;
+
+    if (oMatchEnd > oend-(16-MINMATCH)) {
+        if (op < oend_w) {
+            ZSTD_wildcopy(op, match, oend_w - op);
+            match += oend_w - op;
+            op = oend_w;
+        }
+        while (op < oMatchEnd) *op++ = *match++;
+    } else {
+        ZSTD_wildcopy(op, match, (ptrdiff_t)sequence.matchLength-8);   /* works even if matchLength < 8 */
+    }
+    return sequenceLength;
+}
+
+
+HINT_INLINE
+size_t ZSTD_execSequenceLong(BYTE* op,
+                             BYTE* const oend, seq_t sequence,
+                             const BYTE** litPtr, const BYTE* const litLimit,
+                             const BYTE* const prefixStart, const BYTE* const dictStart, const BYTE* const dictEnd)
+{
+    BYTE* const oLitEnd = op + sequence.litLength;
+    size_t const sequenceLength = sequence.litLength + sequence.matchLength;
+    BYTE* const oMatchEnd = op + sequenceLength;   /* risk : address space overflow (32-bits) */
+    BYTE* const oend_w = oend - WILDCOPY_OVERLENGTH;
+    const BYTE* const iLitEnd = *litPtr + sequence.litLength;
+    const BYTE* match = sequence.match;
+
+    /* check */
+    if (oMatchEnd > oend) return ERROR(dstSize_tooSmall); /* last match must start at a minimum distance of WILDCOPY_OVERLENGTH from oend */
+    if (iLitEnd > litLimit) return ERROR(corruption_detected);   /* over-read beyond lit buffer */
+    if (oLitEnd > oend_w) return ZSTD_execSequenceLast7(op, oend, sequence, litPtr, litLimit, prefixStart, dictStart, dictEnd);
+
+    /* copy Literals */
+    ZSTD_copy8(op, *litPtr);  /* note : op <= oLitEnd <= oend_w == oend - 8 */
+    if (sequence.litLength > 8)
+        ZSTD_wildcopy(op+8, (*litPtr)+8, sequence.litLength - 8);   /* note : since oLitEnd <= oend-WILDCOPY_OVERLENGTH, no risk of overwrite beyond oend */
+    op = oLitEnd;
+    *litPtr = iLitEnd;   /* update for next sequence */
+
+    /* copy Match */
+    if (sequence.offset > (size_t)(oLitEnd - prefixStart)) {
+        /* offset beyond prefix */
+        if (sequence.offset > (size_t)(oLitEnd - dictStart)) return ERROR(corruption_detected);
+        if (match + sequence.matchLength <= dictEnd) {
+            memmove(oLitEnd, match, sequence.matchLength);
+            return sequenceLength;
+        }
+        /* span extDict & currentPrefixSegment */
+        {   size_t const length1 = dictEnd - match;
+            memmove(oLitEnd, match, length1);
+            op = oLitEnd + length1;
+            sequence.matchLength -= length1;
+            match = prefixStart;
+            if (op > oend_w || sequence.matchLength < MINMATCH) {
+              U32 i;
+              for (i = 0; i < sequence.matchLength; ++i) op[i] = match[i];
+              return sequenceLength;
+            }
+    }   }
+    assert(op <= oend_w);
+    assert(sequence.matchLength >= MINMATCH);
+
+    /* match within prefix */
+    if (sequence.offset < 8) {
+        /* close range match, overlap */
+        static const U32 dec32table[] = { 0, 1, 2, 1, 4, 4, 4, 4 };   /* added */
+        static const int dec64table[] = { 8, 8, 8, 7, 8, 9,10,11 };   /* subtracted */
+        int const sub2 = dec64table[sequence.offset];
+        op[0] = match[0];
+        op[1] = match[1];
+        op[2] = match[2];
+        op[3] = match[3];
+        match += dec32table[sequence.offset];
+        ZSTD_copy4(op+4, match);
+        match -= sub2;
+    } else {
+        ZSTD_copy8(op, match);
+    }
+    op += 8; match += 8;
+
+    if (oMatchEnd > oend-(16-MINMATCH)) {
+        if (op < oend_w) {
+            ZSTD_wildcopy(op, match, oend_w - op);
+            match += oend_w - op;
+            op = oend_w;
+        }
+        while (op < oMatchEnd) *op++ = *match++;
+    } else {
+        ZSTD_wildcopy(op, match, (ptrdiff_t)sequence.matchLength-8);   /* works even if matchLength < 8 */
+    }
+    return sequenceLength;
+}
+
+static void
+ZSTD_initFseState(ZSTD_fseState* DStatePtr, BIT_DStream_t* bitD, const ZSTD_seqSymbol* dt)
+{
+    const void* ptr = dt;
+    const ZSTD_seqSymbol_header* const DTableH = (const ZSTD_seqSymbol_header*)ptr;
+    DStatePtr->state = BIT_readBits(bitD, DTableH->tableLog);
+    DEBUGLOG(6, "ZSTD_initFseState : val=%u using %u bits",
+                (U32)DStatePtr->state, DTableH->tableLog);
+    BIT_reloadDStream(bitD);
+    DStatePtr->table = dt + 1;
+}
+
+FORCE_INLINE_TEMPLATE void
+ZSTD_updateFseState(ZSTD_fseState* DStatePtr, BIT_DStream_t* bitD)
+{
+    ZSTD_seqSymbol const DInfo = DStatePtr->table[DStatePtr->state];
+    U32 const nbBits = DInfo.nbBits;
+    size_t const lowBits = BIT_readBits(bitD, nbBits);
+    DStatePtr->state = DInfo.nextState + lowBits;
+}
+
+/* We need to add at most (ZSTD_WINDOWLOG_MAX_32 - 1) bits to read the maximum
+ * offset bits. But we can only read at most (STREAM_ACCUMULATOR_MIN_32 - 1)
+ * bits before reloading. This value is the maximum number of bytes we read
+ * after reloading when we are decoding long offets.
+ */
+#define LONG_OFFSETS_MAX_EXTRA_BITS_32                       \
+    (ZSTD_WINDOWLOG_MAX_32 > STREAM_ACCUMULATOR_MIN_32       \
+        ? ZSTD_WINDOWLOG_MAX_32 - STREAM_ACCUMULATOR_MIN_32  \
+        : 0)
+
+typedef enum { ZSTD_lo_isRegularOffset, ZSTD_lo_isLongOffset=1 } ZSTD_longOffset_e;
+
+#ifndef ZSTD_FORCE_DECOMPRESS_SEQUENCES_LONG
+FORCE_INLINE_TEMPLATE seq_t
+ZSTD_decodeSequence(seqState_t* seqState, const ZSTD_longOffset_e longOffsets)
+{
+    seq_t seq;
+    U32 const llBits = seqState->stateLL.table[seqState->stateLL.state].nbAdditionalBits;
+    U32 const mlBits = seqState->stateML.table[seqState->stateML.state].nbAdditionalBits;
+    U32 const ofBits = seqState->stateOffb.table[seqState->stateOffb.state].nbAdditionalBits;
+    U32 const totalBits = llBits+mlBits+ofBits;
+    U32 const llBase = seqState->stateLL.table[seqState->stateLL.state].baseValue;
+    U32 const mlBase = seqState->stateML.table[seqState->stateML.state].baseValue;
+    U32 const ofBase = seqState->stateOffb.table[seqState->stateOffb.state].baseValue;
+
+    /* sequence */
+    {   size_t offset;
+        if (!ofBits)
+            offset = 0;
+        else {
+            ZSTD_STATIC_ASSERT(ZSTD_lo_isLongOffset == 1);
+            ZSTD_STATIC_ASSERT(LONG_OFFSETS_MAX_EXTRA_BITS_32 == 5);
+            assert(ofBits <= MaxOff);
+            if (MEM_32bits() && longOffsets && (ofBits >= STREAM_ACCUMULATOR_MIN_32)) {
+                U32 const extraBits = ofBits - MIN(ofBits, 32 - seqState->DStream.bitsConsumed);
+                offset = ofBase + (BIT_readBitsFast(&seqState->DStream, ofBits - extraBits) << extraBits);
+                BIT_reloadDStream(&seqState->DStream);
+                if (extraBits) offset += BIT_readBitsFast(&seqState->DStream, extraBits);
+                assert(extraBits <= LONG_OFFSETS_MAX_EXTRA_BITS_32);   /* to avoid another reload */
+            } else {
+                offset = ofBase + BIT_readBitsFast(&seqState->DStream, ofBits/*>0*/);   /* <=  (ZSTD_WINDOWLOG_MAX-1) bits */
+                if (MEM_32bits()) BIT_reloadDStream(&seqState->DStream);
+            }
+        }
+
+        if (ofBits <= 1) {
+            offset += (llBase==0);
+            if (offset) {
+                size_t temp = (offset==3) ? seqState->prevOffset[0] - 1 : seqState->prevOffset[offset];
+                temp += !temp;   /* 0 is not valid; input is corrupted; force offset to 1 */
+                if (offset != 1) seqState->prevOffset[2] = seqState->prevOffset[1];
+                seqState->prevOffset[1] = seqState->prevOffset[0];
+                seqState->prevOffset[0] = offset = temp;
+            } else {  /* offset == 0 */
+                offset = seqState->prevOffset[0];
+            }
+        } else {
+            seqState->prevOffset[2] = seqState->prevOffset[1];
+            seqState->prevOffset[1] = seqState->prevOffset[0];
+            seqState->prevOffset[0] = offset;
+        }
+        seq.offset = offset;
+    }
+
+    seq.matchLength = mlBase
+                    + ((mlBits>0) ? BIT_readBitsFast(&seqState->DStream, mlBits/*>0*/) : 0);  /* <=  16 bits */
+    if (MEM_32bits() && (mlBits+llBits >= STREAM_ACCUMULATOR_MIN_32-LONG_OFFSETS_MAX_EXTRA_BITS_32))
+        BIT_reloadDStream(&seqState->DStream);
+    if (MEM_64bits() && (totalBits >= STREAM_ACCUMULATOR_MIN_64-(LLFSELog+MLFSELog+OffFSELog)))
+        BIT_reloadDStream(&seqState->DStream);
+    /* Ensure there are enough bits to read the rest of data in 64-bit mode. */
+    ZSTD_STATIC_ASSERT(16+LLFSELog+MLFSELog+OffFSELog < STREAM_ACCUMULATOR_MIN_64);
+
+    seq.litLength = llBase
+                  + ((llBits>0) ? BIT_readBitsFast(&seqState->DStream, llBits/*>0*/) : 0);    /* <=  16 bits */
+    if (MEM_32bits())
+        BIT_reloadDStream(&seqState->DStream);
+
+    DEBUGLOG(6, "seq: litL=%u, matchL=%u, offset=%u",
+                (U32)seq.litLength, (U32)seq.matchLength, (U32)seq.offset);
+
+    /* ANS state update */
+    ZSTD_updateFseState(&seqState->stateLL, &seqState->DStream);    /* <=  9 bits */
+    ZSTD_updateFseState(&seqState->stateML, &seqState->DStream);    /* <=  9 bits */
+    if (MEM_32bits()) BIT_reloadDStream(&seqState->DStream);    /* <= 18 bits */
+    ZSTD_updateFseState(&seqState->stateOffb, &seqState->DStream);  /* <=  8 bits */
+
+    return seq;
+}
+
+FORCE_INLINE_TEMPLATE size_t
+ZSTD_decompressSequences_body( ZSTD_DCtx* dctx,
+                               void* dst, size_t maxDstSize,
+                         const void* seqStart, size_t seqSize, int nbSeq,
+                         const ZSTD_longOffset_e isLongOffset)
+{
+    const BYTE* ip = (const BYTE*)seqStart;
+    const BYTE* const iend = ip + seqSize;
+    BYTE* const ostart = (BYTE* const)dst;
+    BYTE* const oend = ostart + maxDstSize;
+    BYTE* op = ostart;
+    const BYTE* litPtr = dctx->litPtr;
+    const BYTE* const litEnd = litPtr + dctx->litSize;
+    const BYTE* const prefixStart = (const BYTE*) (dctx->prefixStart);
+    const BYTE* const vBase = (const BYTE*) (dctx->virtualStart);
+    const BYTE* const dictEnd = (const BYTE*) (dctx->dictEnd);
+    DEBUGLOG(5, "ZSTD_decompressSequences_body");
+
+    /* Regen sequences */
+    if (nbSeq) {
+        seqState_t seqState;
+        dctx->fseEntropy = 1;
+        { U32 i; for (i=0; ientropy.rep[i]; }
+        CHECK_E(BIT_initDStream(&seqState.DStream, ip, iend-ip), corruption_detected);
+        ZSTD_initFseState(&seqState.stateLL, &seqState.DStream, dctx->LLTptr);
+        ZSTD_initFseState(&seqState.stateOffb, &seqState.DStream, dctx->OFTptr);
+        ZSTD_initFseState(&seqState.stateML, &seqState.DStream, dctx->MLTptr);
+
+        for ( ; (BIT_reloadDStream(&(seqState.DStream)) <= BIT_DStream_completed) && nbSeq ; ) {
+            nbSeq--;
+            {   seq_t const sequence = ZSTD_decodeSequence(&seqState, isLongOffset);
+                size_t const oneSeqSize = ZSTD_execSequence(op, oend, sequence, &litPtr, litEnd, prefixStart, vBase, dictEnd);
+                DEBUGLOG(6, "regenerated sequence size : %u", (U32)oneSeqSize);
+                if (ZSTD_isError(oneSeqSize)) return oneSeqSize;
+                op += oneSeqSize;
+        }   }
+
+        /* check if reached exact end */
+        DEBUGLOG(5, "ZSTD_decompressSequences_body: after decode loop, remaining nbSeq : %i", nbSeq);
+        if (nbSeq) return ERROR(corruption_detected);
+        /* save reps for next block */
+        { U32 i; for (i=0; ientropy.rep[i] = (U32)(seqState.prevOffset[i]); }
+    }
+
+    /* last literal segment */
+    {   size_t const lastLLSize = litEnd - litPtr;
+        if (lastLLSize > (size_t)(oend-op)) return ERROR(dstSize_tooSmall);
+        memcpy(op, litPtr, lastLLSize);
+        op += lastLLSize;
+    }
+
+    return op-ostart;
+}
+
+static size_t
+ZSTD_decompressSequences_default(ZSTD_DCtx* dctx,
+                                 void* dst, size_t maxDstSize,
+                           const void* seqStart, size_t seqSize, int nbSeq,
+                           const ZSTD_longOffset_e isLongOffset)
+{
+    return ZSTD_decompressSequences_body(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset);
+}
+#endif /* ZSTD_FORCE_DECOMPRESS_SEQUENCES_LONG */
+
+
+
+#ifndef ZSTD_FORCE_DECOMPRESS_SEQUENCES_SHORT
+FORCE_INLINE_TEMPLATE seq_t
+ZSTD_decodeSequenceLong(seqState_t* seqState, ZSTD_longOffset_e const longOffsets)
+{
+    seq_t seq;
+    U32 const llBits = seqState->stateLL.table[seqState->stateLL.state].nbAdditionalBits;
+    U32 const mlBits = seqState->stateML.table[seqState->stateML.state].nbAdditionalBits;
+    U32 const ofBits = seqState->stateOffb.table[seqState->stateOffb.state].nbAdditionalBits;
+    U32 const totalBits = llBits+mlBits+ofBits;
+    U32 const llBase = seqState->stateLL.table[seqState->stateLL.state].baseValue;
+    U32 const mlBase = seqState->stateML.table[seqState->stateML.state].baseValue;
+    U32 const ofBase = seqState->stateOffb.table[seqState->stateOffb.state].baseValue;
+
+    /* sequence */
+    {   size_t offset;
+        if (!ofBits)
+            offset = 0;
+        else {
+            ZSTD_STATIC_ASSERT(ZSTD_lo_isLongOffset == 1);
+            ZSTD_STATIC_ASSERT(LONG_OFFSETS_MAX_EXTRA_BITS_32 == 5);
+            assert(ofBits <= MaxOff);
+            if (MEM_32bits() && longOffsets) {
+                U32 const extraBits = ofBits - MIN(ofBits, STREAM_ACCUMULATOR_MIN_32-1);
+                offset = ofBase + (BIT_readBitsFast(&seqState->DStream, ofBits - extraBits) << extraBits);
+                if (MEM_32bits() || extraBits) BIT_reloadDStream(&seqState->DStream);
+                if (extraBits) offset += BIT_readBitsFast(&seqState->DStream, extraBits);
+            } else {
+                offset = ofBase + BIT_readBitsFast(&seqState->DStream, ofBits);   /* <=  (ZSTD_WINDOWLOG_MAX-1) bits */
+                if (MEM_32bits()) BIT_reloadDStream(&seqState->DStream);
+            }
+        }
+
+        if (ofBits <= 1) {
+            offset += (llBase==0);
+            if (offset) {
+                size_t temp = (offset==3) ? seqState->prevOffset[0] - 1 : seqState->prevOffset[offset];
+                temp += !temp;   /* 0 is not valid; input is corrupted; force offset to 1 */
+                if (offset != 1) seqState->prevOffset[2] = seqState->prevOffset[1];
+                seqState->prevOffset[1] = seqState->prevOffset[0];
+                seqState->prevOffset[0] = offset = temp;
+            } else {
+                offset = seqState->prevOffset[0];
+            }
+        } else {
+            seqState->prevOffset[2] = seqState->prevOffset[1];
+            seqState->prevOffset[1] = seqState->prevOffset[0];
+            seqState->prevOffset[0] = offset;
+        }
+        seq.offset = offset;
+    }
+
+    seq.matchLength = mlBase + ((mlBits>0) ? BIT_readBitsFast(&seqState->DStream, mlBits) : 0);  /* <=  16 bits */
+    if (MEM_32bits() && (mlBits+llBits >= STREAM_ACCUMULATOR_MIN_32-LONG_OFFSETS_MAX_EXTRA_BITS_32))
+        BIT_reloadDStream(&seqState->DStream);
+    if (MEM_64bits() && (totalBits >= STREAM_ACCUMULATOR_MIN_64-(LLFSELog+MLFSELog+OffFSELog)))
+        BIT_reloadDStream(&seqState->DStream);
+    /* Verify that there is enough bits to read the rest of the data in 64-bit mode. */
+    ZSTD_STATIC_ASSERT(16+LLFSELog+MLFSELog+OffFSELog < STREAM_ACCUMULATOR_MIN_64);
+
+    seq.litLength = llBase + ((llBits>0) ? BIT_readBitsFast(&seqState->DStream, llBits) : 0);    /* <=  16 bits */
+    if (MEM_32bits())
+        BIT_reloadDStream(&seqState->DStream);
+
+    {   size_t const pos = seqState->pos + seq.litLength;
+        const BYTE* const matchBase = (seq.offset > pos) ? seqState->dictEnd : seqState->prefixStart;
+        seq.match = matchBase + pos - seq.offset;  /* note : this operation can overflow when seq.offset is really too large, which can only happen when input is corrupted.
+                                                    * No consequence though : no memory access will occur, overly large offset will be detected in ZSTD_execSequenceLong() */
+        seqState->pos = pos + seq.matchLength;
+    }
+
+    /* ANS state update */
+    ZSTD_updateFseState(&seqState->stateLL, &seqState->DStream);    /* <=  9 bits */
+    ZSTD_updateFseState(&seqState->stateML, &seqState->DStream);    /* <=  9 bits */
+    if (MEM_32bits()) BIT_reloadDStream(&seqState->DStream);    /* <= 18 bits */
+    ZSTD_updateFseState(&seqState->stateOffb, &seqState->DStream);  /* <=  8 bits */
+
+    return seq;
+}
+
+FORCE_INLINE_TEMPLATE size_t
+ZSTD_decompressSequencesLong_body(
+                               ZSTD_DCtx* dctx,
+                               void* dst, size_t maxDstSize,
+                         const void* seqStart, size_t seqSize, int nbSeq,
+                         const ZSTD_longOffset_e isLongOffset)
+{
+    const BYTE* ip = (const BYTE*)seqStart;
+    const BYTE* const iend = ip + seqSize;
+    BYTE* const ostart = (BYTE* const)dst;
+    BYTE* const oend = ostart + maxDstSize;
+    BYTE* op = ostart;
+    const BYTE* litPtr = dctx->litPtr;
+    const BYTE* const litEnd = litPtr + dctx->litSize;
+    const BYTE* const prefixStart = (const BYTE*) (dctx->prefixStart);
+    const BYTE* const dictStart = (const BYTE*) (dctx->virtualStart);
+    const BYTE* const dictEnd = (const BYTE*) (dctx->dictEnd);
+
+    /* Regen sequences */
+    if (nbSeq) {
+#define STORED_SEQS 4
+#define STORED_SEQS_MASK (STORED_SEQS-1)
+#define ADVANCED_SEQS 4
+        seq_t sequences[STORED_SEQS];
+        int const seqAdvance = MIN(nbSeq, ADVANCED_SEQS);
+        seqState_t seqState;
+        int seqNb;
+        dctx->fseEntropy = 1;
+        { int i; for (i=0; ientropy.rep[i]; }
+        seqState.prefixStart = prefixStart;
+        seqState.pos = (size_t)(op-prefixStart);
+        seqState.dictEnd = dictEnd;
+        assert(iend >= ip);
+        CHECK_E(BIT_initDStream(&seqState.DStream, ip, iend-ip), corruption_detected);
+        ZSTD_initFseState(&seqState.stateLL, &seqState.DStream, dctx->LLTptr);
+        ZSTD_initFseState(&seqState.stateOffb, &seqState.DStream, dctx->OFTptr);
+        ZSTD_initFseState(&seqState.stateML, &seqState.DStream, dctx->MLTptr);
+
+        /* prepare in advance */
+        for (seqNb=0; (BIT_reloadDStream(&seqState.DStream) <= BIT_DStream_completed) && (seqNbentropy.rep[i] = (U32)(seqState.prevOffset[i]); }
+    }
+
+    /* last literal segment */
+    {   size_t const lastLLSize = litEnd - litPtr;
+        if (lastLLSize > (size_t)(oend-op)) return ERROR(dstSize_tooSmall);
+        memcpy(op, litPtr, lastLLSize);
+        op += lastLLSize;
+    }
+
+    return op-ostart;
+}
+
+static size_t
+ZSTD_decompressSequencesLong_default(ZSTD_DCtx* dctx,
+                                 void* dst, size_t maxDstSize,
+                           const void* seqStart, size_t seqSize, int nbSeq,
+                           const ZSTD_longOffset_e isLongOffset)
+{
+    return ZSTD_decompressSequencesLong_body(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset);
+}
+#endif /* ZSTD_FORCE_DECOMPRESS_SEQUENCES_SHORT */
+
+
+
+#if DYNAMIC_BMI2
+
+#ifndef ZSTD_FORCE_DECOMPRESS_SEQUENCES_LONG
+static TARGET_ATTRIBUTE("bmi2") size_t
+ZSTD_decompressSequences_bmi2(ZSTD_DCtx* dctx,
+                                 void* dst, size_t maxDstSize,
+                           const void* seqStart, size_t seqSize, int nbSeq,
+                           const ZSTD_longOffset_e isLongOffset)
+{
+    return ZSTD_decompressSequences_body(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset);
+}
+#endif /* ZSTD_FORCE_DECOMPRESS_SEQUENCES_LONG */
+
+#ifndef ZSTD_FORCE_DECOMPRESS_SEQUENCES_SHORT
+static TARGET_ATTRIBUTE("bmi2") size_t
+ZSTD_decompressSequencesLong_bmi2(ZSTD_DCtx* dctx,
+                                 void* dst, size_t maxDstSize,
+                           const void* seqStart, size_t seqSize, int nbSeq,
+                           const ZSTD_longOffset_e isLongOffset)
+{
+    return ZSTD_decompressSequencesLong_body(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset);
+}
+#endif /* ZSTD_FORCE_DECOMPRESS_SEQUENCES_SHORT */
+
+#endif /* DYNAMIC_BMI2 */
+
+typedef size_t (*ZSTD_decompressSequences_t)(
+                            ZSTD_DCtx* dctx,
+                            void* dst, size_t maxDstSize,
+                            const void* seqStart, size_t seqSize, int nbSeq,
+                            const ZSTD_longOffset_e isLongOffset);
+
+#ifndef ZSTD_FORCE_DECOMPRESS_SEQUENCES_LONG
+static size_t
+ZSTD_decompressSequences(ZSTD_DCtx* dctx, void* dst, size_t maxDstSize,
+                   const void* seqStart, size_t seqSize, int nbSeq,
+                   const ZSTD_longOffset_e isLongOffset)
+{
+    DEBUGLOG(5, "ZSTD_decompressSequences");
+#if DYNAMIC_BMI2
+    if (dctx->bmi2) {
+        return ZSTD_decompressSequences_bmi2(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset);
+    }
+#endif
+  return ZSTD_decompressSequences_default(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset);
+}
+#endif /* ZSTD_FORCE_DECOMPRESS_SEQUENCES_LONG */
+
+
+#ifndef ZSTD_FORCE_DECOMPRESS_SEQUENCES_SHORT
+/* ZSTD_decompressSequencesLong() :
+ * decompression function triggered when a minimum share of offsets is considered "long",
+ * aka out of cache.
+ * note : "long" definition seems overloaded here, sometimes meaning "wider than bitstream register", and sometimes mearning "farther than memory cache distance".
+ * This function will try to mitigate main memory latency through the use of prefetching */
+static size_t
+ZSTD_decompressSequencesLong(ZSTD_DCtx* dctx,
+                             void* dst, size_t maxDstSize,
+                             const void* seqStart, size_t seqSize, int nbSeq,
+                             const ZSTD_longOffset_e isLongOffset)
+{
+    DEBUGLOG(5, "ZSTD_decompressSequencesLong");
+#if DYNAMIC_BMI2
+    if (dctx->bmi2) {
+        return ZSTD_decompressSequencesLong_bmi2(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset);
+    }
+#endif
+  return ZSTD_decompressSequencesLong_default(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset);
+}
+#endif /* ZSTD_FORCE_DECOMPRESS_SEQUENCES_SHORT */
+
+
+
+#if !defined(ZSTD_FORCE_DECOMPRESS_SEQUENCES_SHORT) && \
+    !defined(ZSTD_FORCE_DECOMPRESS_SEQUENCES_LONG)
+/* ZSTD_getLongOffsetsShare() :
+ * condition : offTable must be valid
+ * @return : "share" of long offsets (arbitrarily defined as > (1<<23))
+ *           compared to maximum possible of (1< 22) total += 1;
+    }
+
+    assert(tableLog <= OffFSELog);
+    total <<= (OffFSELog - tableLog);  /* scale to OffFSELog */
+
+    return total;
+}
+#endif
+
+
+size_t
+ZSTD_decompressBlock_internal(ZSTD_DCtx* dctx,
+                              void* dst, size_t dstCapacity,
+                        const void* src, size_t srcSize, const int frame)
+{   /* blockType == blockCompressed */
+    const BYTE* ip = (const BYTE*)src;
+    /* isLongOffset must be true if there are long offsets.
+     * Offsets are long if they are larger than 2^STREAM_ACCUMULATOR_MIN.
+     * We don't expect that to be the case in 64-bit mode.
+     * In block mode, window size is not known, so we have to be conservative.
+     * (note: but it could be evaluated from current-lowLimit)
+     */
+    ZSTD_longOffset_e const isLongOffset = (ZSTD_longOffset_e)(MEM_32bits() && (!frame || (dctx->fParams.windowSize > (1ULL << STREAM_ACCUMULATOR_MIN))));
+    DEBUGLOG(5, "ZSTD_decompressBlock_internal (size : %u)", (U32)srcSize);
+
+    if (srcSize >= ZSTD_BLOCKSIZE_MAX) return ERROR(srcSize_wrong);
+
+    /* Decode literals section */
+    {   size_t const litCSize = ZSTD_decodeLiteralsBlock(dctx, src, srcSize);
+        DEBUGLOG(5, "ZSTD_decodeLiteralsBlock : %u", (U32)litCSize);
+        if (ZSTD_isError(litCSize)) return litCSize;
+        ip += litCSize;
+        srcSize -= litCSize;
+    }
+
+    /* Build Decoding Tables */
+    {
+        /* These macros control at build-time which decompressor implementation
+         * we use. If neither is defined, we do some inspection and dispatch at
+         * runtime.
+         */
+#if !defined(ZSTD_FORCE_DECOMPRESS_SEQUENCES_SHORT) && \
+    !defined(ZSTD_FORCE_DECOMPRESS_SEQUENCES_LONG)
+        int usePrefetchDecoder = dctx->ddictIsCold;
+#endif
+        int nbSeq;
+        size_t const seqHSize = ZSTD_decodeSeqHeaders(dctx, &nbSeq, ip, srcSize);
+        if (ZSTD_isError(seqHSize)) return seqHSize;
+        ip += seqHSize;
+        srcSize -= seqHSize;
+
+#if !defined(ZSTD_FORCE_DECOMPRESS_SEQUENCES_SHORT) && \
+    !defined(ZSTD_FORCE_DECOMPRESS_SEQUENCES_LONG)
+        if ( !usePrefetchDecoder
+          && (!frame || (dctx->fParams.windowSize > (1<<24)))
+          && (nbSeq>ADVANCED_SEQS) ) {  /* could probably use a larger nbSeq limit */
+            U32 const shareLongOffsets = ZSTD_getLongOffsetsShare(dctx->OFTptr);
+            U32 const minShare = MEM_64bits() ? 7 : 20; /* heuristic values, correspond to 2.73% and 7.81% */
+            usePrefetchDecoder = (shareLongOffsets >= minShare);
+        }
+#endif
+
+        dctx->ddictIsCold = 0;
+
+#if !defined(ZSTD_FORCE_DECOMPRESS_SEQUENCES_SHORT) && \
+    !defined(ZSTD_FORCE_DECOMPRESS_SEQUENCES_LONG)
+        if (usePrefetchDecoder)
+#endif
+#ifndef ZSTD_FORCE_DECOMPRESS_SEQUENCES_SHORT
+            return ZSTD_decompressSequencesLong(dctx, dst, dstCapacity, ip, srcSize, nbSeq, isLongOffset);
+#endif
+
+#ifndef ZSTD_FORCE_DECOMPRESS_SEQUENCES_LONG
+        /* else */
+        return ZSTD_decompressSequences(dctx, dst, dstCapacity, ip, srcSize, nbSeq, isLongOffset);
+#endif
+    }
+}
+
+
+size_t ZSTD_decompressBlock(ZSTD_DCtx* dctx,
+                            void* dst, size_t dstCapacity,
+                      const void* src, size_t srcSize)
+{
+    size_t dSize;
+    ZSTD_checkContinuity(dctx, dst);
+    dSize = ZSTD_decompressBlock_internal(dctx, dst, dstCapacity, src, srcSize, /* frame */ 0);
+    dctx->previousDstEnd = (char*)dst + dSize;
+    return dSize;
+}
diff --git a/lib/decompress/zstd_decompress_block.h b/lib/decompress/zstd_decompress_block.h
new file mode 100644
index 000000000000..7e9296041023
--- /dev/null
+++ b/lib/decompress/zstd_decompress_block.h
@@ -0,0 +1,59 @@
+/*
+ * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
+ * You may select, at your option, one of the above-listed licenses.
+ */
+
+
+#ifndef ZSTD_DEC_BLOCK_H
+#define ZSTD_DEC_BLOCK_H
+
+/*-*******************************************************
+ *  Dependencies
+ *********************************************************/
+#include    /* size_t */
+#include "zstd.h"    /* DCtx, and some public functions */
+#include "zstd_internal.h"  /* blockProperties_t, and some public functions */
+#include "zstd_decompress_internal.h"  /* ZSTD_seqSymbol */
+
+
+/* ===   Prototypes   === */
+
+/* note: prototypes already published within `zstd.h` :
+ * ZSTD_decompressBlock()
+ */
+
+/* note: prototypes already published within `zstd_internal.h` :
+ * ZSTD_getcBlockSize()
+ * ZSTD_decodeSeqHeaders()
+ */
+
+
+/* ZSTD_decompressBlock_internal() :
+ * decompress block, starting at `src`,
+ * into destination buffer `dst`.
+ * @return : decompressed block size,
+ *           or an error code (which can be tested using ZSTD_isError())
+ */
+size_t ZSTD_decompressBlock_internal(ZSTD_DCtx* dctx,
+                               void* dst, size_t dstCapacity,
+                         const void* src, size_t srcSize, const int frame);
+
+/* ZSTD_buildFSETable() :
+ * generate FSE decoding table for one symbol (ll, ml or off)
+ * this function must be called with valid parameters only
+ * (dt is large enough, normalizedCounter distribution total is a power of 2, max is within range, etc.)
+ * in which case it cannot fail.
+ * Internal use only.
+ */
+void ZSTD_buildFSETable(ZSTD_seqSymbol* dt,
+             const short* normalizedCounter, unsigned maxSymbolValue,
+             const U32* baseValue, const U32* nbAdditionalBits,
+                   unsigned tableLog);
+
+
+#endif /* ZSTD_DEC_BLOCK_H */
diff --git a/lib/decompress/zstd_decompress_internal.h b/lib/decompress/zstd_decompress_internal.h
new file mode 100644
index 000000000000..abd0030519d2
--- /dev/null
+++ b/lib/decompress/zstd_decompress_internal.h
@@ -0,0 +1,168 @@
+/*
+ * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
+ * You may select, at your option, one of the above-listed licenses.
+ */
+
+
+/* zstd_decompress_internal:
+ * objects and definitions shared within lib/decompress modules */
+
+ #ifndef ZSTD_DECOMPRESS_INTERNAL_H
+ #define ZSTD_DECOMPRESS_INTERNAL_H
+
+
+/*-*******************************************************
+ *  Dependencies
+ *********************************************************/
+#include "mem.h"             /* BYTE, U16, U32 */
+#include "zstd_internal.h"   /* ZSTD_seqSymbol */
+
+
+
+/*-*******************************************************
+ *  Constants
+ *********************************************************/
+static const U32 LL_base[MaxLL+1] = {
+                 0,    1,    2,     3,     4,     5,     6,      7,
+                 8,    9,   10,    11,    12,    13,    14,     15,
+                16,   18,   20,    22,    24,    28,    32,     40,
+                48,   64, 0x80, 0x100, 0x200, 0x400, 0x800, 0x1000,
+                0x2000, 0x4000, 0x8000, 0x10000 };
+
+static const U32 OF_base[MaxOff+1] = {
+                 0,        1,       1,       5,     0xD,     0x1D,     0x3D,     0x7D,
+                 0xFD,   0x1FD,   0x3FD,   0x7FD,   0xFFD,   0x1FFD,   0x3FFD,   0x7FFD,
+                 0xFFFD, 0x1FFFD, 0x3FFFD, 0x7FFFD, 0xFFFFD, 0x1FFFFD, 0x3FFFFD, 0x7FFFFD,
+                 0xFFFFFD, 0x1FFFFFD, 0x3FFFFFD, 0x7FFFFFD, 0xFFFFFFD, 0x1FFFFFFD, 0x3FFFFFFD, 0x7FFFFFFD };
+
+static const U32 OF_bits[MaxOff+1] = {
+                     0,  1,  2,  3,  4,  5,  6,  7,
+                     8,  9, 10, 11, 12, 13, 14, 15,
+                    16, 17, 18, 19, 20, 21, 22, 23,
+                    24, 25, 26, 27, 28, 29, 30, 31 };
+
+static const U32 ML_base[MaxML+1] = {
+                     3,  4,  5,    6,     7,     8,     9,    10,
+                    11, 12, 13,   14,    15,    16,    17,    18,
+                    19, 20, 21,   22,    23,    24,    25,    26,
+                    27, 28, 29,   30,    31,    32,    33,    34,
+                    35, 37, 39,   41,    43,    47,    51,    59,
+                    67, 83, 99, 0x83, 0x103, 0x203, 0x403, 0x803,
+                    0x1003, 0x2003, 0x4003, 0x8003, 0x10003 };
+
+
+/*-*******************************************************
+ *  Decompression types
+ *********************************************************/
+ typedef struct {
+     U32 fastMode;
+     U32 tableLog;
+ } ZSTD_seqSymbol_header;
+
+ typedef struct {
+     U16  nextState;
+     BYTE nbAdditionalBits;
+     BYTE nbBits;
+     U32  baseValue;
+ } ZSTD_seqSymbol;
+
+ #define SEQSYMBOL_TABLE_SIZE(log)   (1 + (1 << (log)))
+
+typedef struct {
+    ZSTD_seqSymbol LLTable[SEQSYMBOL_TABLE_SIZE(LLFSELog)];    /* Note : Space reserved for FSE Tables */
+    ZSTD_seqSymbol OFTable[SEQSYMBOL_TABLE_SIZE(OffFSELog)];   /* is also used as temporary workspace while building hufTable during DDict creation */
+    ZSTD_seqSymbol MLTable[SEQSYMBOL_TABLE_SIZE(MLFSELog)];    /* and therefore must be at least HUF_DECOMPRESS_WORKSPACE_SIZE large */
+    HUF_DTable hufTable[HUF_DTABLE_SIZE(HufLog)];  /* can accommodate HUF_decompress4X */
+    U32 rep[ZSTD_REP_NUM];
+} ZSTD_entropyDTables_t;
+
+typedef enum { ZSTDds_getFrameHeaderSize, ZSTDds_decodeFrameHeader,
+               ZSTDds_decodeBlockHeader, ZSTDds_decompressBlock,
+               ZSTDds_decompressLastBlock, ZSTDds_checkChecksum,
+               ZSTDds_decodeSkippableHeader, ZSTDds_skipFrame } ZSTD_dStage;
+
+typedef enum { zdss_init=0, zdss_loadHeader,
+               zdss_read, zdss_load, zdss_flush } ZSTD_dStreamStage;
+
+struct ZSTD_DCtx_s
+{
+    const ZSTD_seqSymbol* LLTptr;
+    const ZSTD_seqSymbol* MLTptr;
+    const ZSTD_seqSymbol* OFTptr;
+    const HUF_DTable* HUFptr;
+    ZSTD_entropyDTables_t entropy;
+    U32 workspace[HUF_DECOMPRESS_WORKSPACE_SIZE_U32];   /* space needed when building huffman tables */
+    const void* previousDstEnd;   /* detect continuity */
+    const void* prefixStart;      /* start of current segment */
+    const void* virtualStart;     /* virtual start of previous segment if it was just before current one */
+    const void* dictEnd;          /* end of previous segment */
+    size_t expected;
+    ZSTD_frameHeader fParams;
+    U64 decodedSize;
+    blockType_e bType;            /* used in ZSTD_decompressContinue(), store blockType between block header decoding and block decompression stages */
+    ZSTD_dStage stage;
+    U32 litEntropy;
+    U32 fseEntropy;
+    XXH64_state_t xxhState;
+    size_t headerSize;
+    ZSTD_format_e format;
+    const BYTE* litPtr;
+    ZSTD_customMem customMem;
+    size_t litSize;
+    size_t rleSize;
+    size_t staticSize;
+    int bmi2;                     /* == 1 if the CPU supports BMI2 and 0 otherwise. CPU support is determined dynamically once per context lifetime. */
+
+    /* dictionary */
+    ZSTD_DDict* ddictLocal;
+    const ZSTD_DDict* ddict;     /* set by ZSTD_initDStream_usingDDict(), or ZSTD_DCtx_refDDict() */
+    U32 dictID;
+    int ddictIsCold;             /* if == 1 : dictionary is "new" for working context, and presumed "cold" (not in cpu cache) */
+
+    /* streaming */
+    ZSTD_dStreamStage streamStage;
+    char*  inBuff;
+    size_t inBuffSize;
+    size_t inPos;
+    size_t maxWindowSize;
+    char*  outBuff;
+    size_t outBuffSize;
+    size_t outStart;
+    size_t outEnd;
+    size_t lhSize;
+    void* legacyContext;
+    U32 previousLegacyVersion;
+    U32 legacyVersion;
+    U32 hostageByte;
+    int noForwardProgress;
+
+    /* workspace */
+    BYTE litBuffer[ZSTD_BLOCKSIZE_MAX + WILDCOPY_OVERLENGTH];
+    BYTE headerBuffer[ZSTD_FRAMEHEADERSIZE_MAX];
+};  /* typedef'd to ZSTD_DCtx within "zstd.h" */
+
+
+/*-*******************************************************
+ *  Shared internal functions
+ *********************************************************/
+
+/*! ZSTD_loadDEntropy() :
+ *  dict : must point at beginning of a valid zstd dictionary.
+ * @return : size of entropy tables read */
+size_t ZSTD_loadDEntropy(ZSTD_entropyDTables_t* entropy,
+                   const void* const dict, size_t const dictSize);
+
+/*! ZSTD_checkContinuity() :
+ *  check if next `dst` follows previous position, where decompression ended.
+ *  If yes, do nothing (continue on current segment).
+ *  If not, classify previous segment as "external dictionary", and start a new segment.
+ *  This function cannot fail. */
+void ZSTD_checkContinuity(ZSTD_DCtx* dctx, const void* dst);
+
+
+#endif /* ZSTD_DECOMPRESS_INTERNAL_H */
diff --git a/lib/dictBuilder/cover.c b/lib/dictBuilder/cover.c
index 6b4af69d29c5..b55bfb510b72 100644
--- a/lib/dictBuilder/cover.c
+++ b/lib/dictBuilder/cover.c
@@ -39,7 +39,7 @@
 /*-*************************************
 *  Constants
 ***************************************/
-#define COVER_MAX_SAMPLES_SIZE (sizeof(size_t) == 8 ? ((U32)-1) : ((U32)1 GB))
+#define COVER_MAX_SAMPLES_SIZE (sizeof(size_t) == 8 ? ((unsigned)-1) : ((unsigned)1 GB))
 #define DEFAULT_SPLITPOINT 1.0
 
 /*-*************************************
@@ -543,7 +543,7 @@ static int COVER_ctx_init(COVER_ctx_t *ctx, const void *samplesBuffer,
   if (totalSamplesSize < MAX(d, sizeof(U64)) ||
       totalSamplesSize >= (size_t)COVER_MAX_SAMPLES_SIZE) {
     DISPLAYLEVEL(1, "Total samples size is too large (%u MB), maximum size is %u MB\n",
-                 (U32)(totalSamplesSize>>20), (COVER_MAX_SAMPLES_SIZE >> 20));
+                 (unsigned)(totalSamplesSize>>20), (COVER_MAX_SAMPLES_SIZE >> 20));
     return 0;
   }
   /* Check if there are at least 5 training samples */
@@ -559,9 +559,9 @@ static int COVER_ctx_init(COVER_ctx_t *ctx, const void *samplesBuffer,
   /* Zero the context */
   memset(ctx, 0, sizeof(*ctx));
   DISPLAYLEVEL(2, "Training on %u samples of total size %u\n", nbTrainSamples,
-               (U32)trainingSamplesSize);
+               (unsigned)trainingSamplesSize);
   DISPLAYLEVEL(2, "Testing on %u samples of total size %u\n", nbTestSamples,
-               (U32)testSamplesSize);
+               (unsigned)testSamplesSize);
   ctx->samples = samples;
   ctx->samplesSizes = samplesSizes;
   ctx->nbSamples = nbSamples;
@@ -639,11 +639,11 @@ static size_t COVER_buildDictionary(const COVER_ctx_t *ctx, U32 *freqs,
   /* Divide the data up into epochs of equal size.
    * We will select at least one segment from each epoch.
    */
-  const U32 epochs = MAX(1, (U32)(dictBufferCapacity / parameters.k / 4));
-  const U32 epochSize = (U32)(ctx->suffixSize / epochs);
+  const unsigned epochs = MAX(1, (U32)(dictBufferCapacity / parameters.k / 4));
+  const unsigned epochSize = (U32)(ctx->suffixSize / epochs);
   size_t epoch;
-  DISPLAYLEVEL(2, "Breaking content into %u epochs of size %u\n", epochs,
-               epochSize);
+  DISPLAYLEVEL(2, "Breaking content into %u epochs of size %u\n",
+                epochs, epochSize);
   /* Loop through the epochs until there are no more segments or the dictionary
    * is full.
    */
@@ -670,7 +670,7 @@ static size_t COVER_buildDictionary(const COVER_ctx_t *ctx, U32 *freqs,
     memcpy(dict + tail, ctx->samples + segment.begin, segmentSize);
     DISPLAYUPDATE(
         2, "\r%u%%       ",
-        (U32)(((dictBufferCapacity - tail) * 100) / dictBufferCapacity));
+        (unsigned)(((dictBufferCapacity - tail) * 100) / dictBufferCapacity));
   }
   DISPLAYLEVEL(2, "\r%79s\r", "");
   return tail;
@@ -722,7 +722,7 @@ ZDICTLIB_API size_t ZDICT_trainFromBuffer_cover(
         samplesBuffer, samplesSizes, nbSamples, parameters.zParams);
     if (!ZSTD_isError(dictionarySize)) {
       DISPLAYLEVEL(2, "Constructed dictionary of size %u\n",
-                   (U32)dictionarySize);
+                   (unsigned)dictionarySize);
     }
     COVER_ctx_destroy(&ctx);
     COVER_map_destroy(&activeDmers);
@@ -868,6 +868,8 @@ void COVER_best_finish(COVER_best_t *best, size_t compressedSize,
         if (!best->dict) {
           best->compressedSize = ERROR(GENERIC);
           best->dictSize = 0;
+          ZSTD_pthread_cond_signal(&best->cond);
+          ZSTD_pthread_mutex_unlock(&best->mutex);
           return;
         }
       }
@@ -1054,7 +1056,7 @@ ZDICTLIB_API size_t ZDICT_optimizeTrainFromBuffer_cover(
       }
       /* Print status */
       LOCALDISPLAYUPDATE(displayLevel, 2, "\r%u%%       ",
-                         (U32)((iteration * 100) / kIterations));
+                         (unsigned)((iteration * 100) / kIterations));
       ++iteration;
     }
     COVER_best_wait(&best);
diff --git a/lib/dictBuilder/fastcover.c b/lib/dictBuilder/fastcover.c
index dfee45743413..c289c0690145 100644
--- a/lib/dictBuilder/fastcover.c
+++ b/lib/dictBuilder/fastcover.c
@@ -20,7 +20,7 @@
 /*-*************************************
 *  Constants
 ***************************************/
-#define FASTCOVER_MAX_SAMPLES_SIZE (sizeof(size_t) == 8 ? ((U32)-1) : ((U32)1 GB))
+#define FASTCOVER_MAX_SAMPLES_SIZE (sizeof(size_t) == 8 ? ((unsigned)-1) : ((unsigned)1 GB))
 #define FASTCOVER_MAX_F 31
 #define FASTCOVER_MAX_ACCEL 10
 #define DEFAULT_SPLITPOINT 0.75
@@ -159,15 +159,15 @@ static COVER_segment_t FASTCOVER_selectSegment(const FASTCOVER_ctx_t *ctx,
    */
   while (activeSegment.end < end) {
     /* Get hash value of current dmer */
-    const size_t index = FASTCOVER_hashPtrToIndex(ctx->samples + activeSegment.end, f, d);
+    const size_t idx = FASTCOVER_hashPtrToIndex(ctx->samples + activeSegment.end, f, d);
 
     /* Add frequency of this index to score if this is the first occurence of index in active segment */
-    if (segmentFreqs[index] == 0) {
-      activeSegment.score += freqs[index];
+    if (segmentFreqs[idx] == 0) {
+      activeSegment.score += freqs[idx];
     }
     /* Increment end of segment and segmentFreqs*/
     activeSegment.end += 1;
-    segmentFreqs[index] += 1;
+    segmentFreqs[idx] += 1;
     /* If the window is now too large, drop the first position */
     if (activeSegment.end - activeSegment.begin == dmersInK + 1) {
       /* Get hash value of the dmer to be eliminated from active segment */
@@ -309,7 +309,7 @@ FASTCOVER_ctx_init(FASTCOVER_ctx_t* ctx,
     if (totalSamplesSize < MAX(d, sizeof(U64)) ||
         totalSamplesSize >= (size_t)FASTCOVER_MAX_SAMPLES_SIZE) {
         DISPLAYLEVEL(1, "Total samples size is too large (%u MB), maximum size is %u MB\n",
-                    (U32)(totalSamplesSize >> 20), (FASTCOVER_MAX_SAMPLES_SIZE >> 20));
+                    (unsigned)(totalSamplesSize >> 20), (FASTCOVER_MAX_SAMPLES_SIZE >> 20));
         return 0;
     }
 
@@ -328,9 +328,9 @@ FASTCOVER_ctx_init(FASTCOVER_ctx_t* ctx,
     /* Zero the context */
     memset(ctx, 0, sizeof(*ctx));
     DISPLAYLEVEL(2, "Training on %u samples of total size %u\n", nbTrainSamples,
-                    (U32)trainingSamplesSize);
+                    (unsigned)trainingSamplesSize);
     DISPLAYLEVEL(2, "Testing on %u samples of total size %u\n", nbTestSamples,
-                    (U32)testSamplesSize);
+                    (unsigned)testSamplesSize);
 
     ctx->samples = samples;
     ctx->samplesSizes = samplesSizes;
@@ -389,11 +389,11 @@ FASTCOVER_buildDictionary(const FASTCOVER_ctx_t* ctx,
   /* Divide the data up into epochs of equal size.
    * We will select at least one segment from each epoch.
    */
-  const U32 epochs = MAX(1, (U32)(dictBufferCapacity / parameters.k));
-  const U32 epochSize = (U32)(ctx->nbDmers / epochs);
+  const unsigned epochs = MAX(1, (U32)(dictBufferCapacity / parameters.k));
+  const unsigned epochSize = (U32)(ctx->nbDmers / epochs);
   size_t epoch;
-  DISPLAYLEVEL(2, "Breaking content into %u epochs of size %u\n", epochs,
-               epochSize);
+  DISPLAYLEVEL(2, "Breaking content into %u epochs of size %u\n",
+                epochs, epochSize);
   /* Loop through the epochs until there are no more segments or the dictionary
    * is full.
    */
@@ -423,7 +423,7 @@ FASTCOVER_buildDictionary(const FASTCOVER_ctx_t* ctx,
     memcpy(dict + tail, ctx->samples + segment.begin, segmentSize);
     DISPLAYUPDATE(
         2, "\r%u%%       ",
-        (U32)(((dictBufferCapacity - tail) * 100) / dictBufferCapacity));
+        (unsigned)(((dictBufferCapacity - tail) * 100) / dictBufferCapacity));
   }
   DISPLAYLEVEL(2, "\r%79s\r", "");
   return tail;
@@ -577,7 +577,7 @@ ZDICT_trainFromBuffer_fastCover(void* dictBuffer, size_t dictBufferCapacity,
           samplesBuffer, samplesSizes, nbFinalizeSamples, coverParams.zParams);
       if (!ZSTD_isError(dictionarySize)) {
           DISPLAYLEVEL(2, "Constructed dictionary of size %u\n",
-                      (U32)dictionarySize);
+                      (unsigned)dictionarySize);
       }
       FASTCOVER_ctx_destroy(&ctx);
       free(segmentFreqs);
@@ -702,7 +702,7 @@ ZDICT_optimizeTrainFromBuffer_fastCover(
         }
         /* Print status */
         LOCALDISPLAYUPDATE(displayLevel, 2, "\r%u%%       ",
-                           (U32)((iteration * 100) / kIterations));
+                           (unsigned)((iteration * 100) / kIterations));
         ++iteration;
       }
       COVER_best_wait(&best);
diff --git a/lib/dictBuilder/zdict.c b/lib/dictBuilder/zdict.c
index 2964b69fff04..c753da0dbbfe 100644
--- a/lib/dictBuilder/zdict.c
+++ b/lib/dictBuilder/zdict.c
@@ -255,15 +255,15 @@ static dictItem ZDICT_analyzePos(
     }
 
     {   int i;
-        U32 searchLength;
+        U32 mml;
         U32 refinedStart = start;
         U32 refinedEnd = end;
 
         DISPLAYLEVEL(4, "\n");
-        DISPLAYLEVEL(4, "found %3u matches of length >= %i at pos %7u  ", (U32)(end-start), MINMATCHLENGTH, (U32)pos);
+        DISPLAYLEVEL(4, "found %3u matches of length >= %i at pos %7u  ", (unsigned)(end-start), MINMATCHLENGTH, (unsigned)pos);
         DISPLAYLEVEL(4, "\n");
 
-        for (searchLength = MINMATCHLENGTH ; ; searchLength++) {
+        for (mml = MINMATCHLENGTH ; ; mml++) {
             BYTE currentChar = 0;
             U32 currentCount = 0;
             U32 currentID = refinedStart;
@@ -271,13 +271,13 @@ static dictItem ZDICT_analyzePos(
             U32 selectedCount = 0;
             U32 selectedID = currentID;
             for (id =refinedStart; id < refinedEnd; id++) {
-                if (b[suffix[id] + searchLength] != currentChar) {
+                if (b[suffix[id] + mml] != currentChar) {
                     if (currentCount > selectedCount) {
                         selectedCount = currentCount;
                         selectedID = currentID;
                     }
                     currentID = id;
-                    currentChar = b[ suffix[id] + searchLength];
+                    currentChar = b[ suffix[id] + mml];
                     currentCount = 0;
                 }
                 currentCount ++;
@@ -342,7 +342,7 @@ static dictItem ZDICT_analyzePos(
             savings[i] = savings[i-1] + (lengthList[i] * (i-3));
 
         DISPLAYLEVEL(4, "Selected dict at position %u, of length %u : saves %u (ratio: %.2f)  \n",
-                     (U32)pos, (U32)maxLength, savings[maxLength], (double)savings[maxLength] / maxLength);
+                     (unsigned)pos, (unsigned)maxLength, (unsigned)savings[maxLength], (double)savings[maxLength] / maxLength);
 
         solution.pos = (U32)pos;
         solution.length = (U32)maxLength;
@@ -497,7 +497,7 @@ static U32 ZDICT_dictSize(const dictItem* dictList)
 static size_t ZDICT_trainBuffer_legacy(dictItem* dictList, U32 dictListSize,
                             const void* const buffer, size_t bufferSize,   /* buffer must end with noisy guard band */
                             const size_t* fileSizes, unsigned nbFiles,
-                            U32 minRatio, U32 notificationLevel)
+                            unsigned minRatio, U32 notificationLevel)
 {
     int* const suffix0 = (int*)malloc((bufferSize+2)*sizeof(*suffix0));
     int* const suffix = suffix0+1;
@@ -523,11 +523,11 @@ static size_t ZDICT_trainBuffer_legacy(dictItem* dictList, U32 dictListSize,
     memset(doneMarks, 0, bufferSize+16);
 
     /* limit sample set size (divsufsort limitation)*/
-    if (bufferSize > ZDICT_MAX_SAMPLES_SIZE) DISPLAYLEVEL(3, "sample set too large : reduced to %u MB ...\n", (U32)(ZDICT_MAX_SAMPLES_SIZE>>20));
+    if (bufferSize > ZDICT_MAX_SAMPLES_SIZE) DISPLAYLEVEL(3, "sample set too large : reduced to %u MB ...\n", (unsigned)(ZDICT_MAX_SAMPLES_SIZE>>20));
     while (bufferSize > ZDICT_MAX_SAMPLES_SIZE) bufferSize -= fileSizes[--nbFiles];
 
     /* sort */
-    DISPLAYLEVEL(2, "sorting %u files of total size %u MB ...\n", nbFiles, (U32)(bufferSize>>20));
+    DISPLAYLEVEL(2, "sorting %u files of total size %u MB ...\n", nbFiles, (unsigned)(bufferSize>>20));
     {   int const divSuftSortResult = divsufsort((const unsigned char*)buffer, suffix, (int)bufferSize, 0);
         if (divSuftSortResult != 0) { result = ERROR(GENERIC); goto _cleanup; }
     }
@@ -589,7 +589,7 @@ typedef struct
 #define MAXREPOFFSET 1024
 
 static void ZDICT_countEStats(EStats_ress_t esr, ZSTD_parameters params,
-                              U32* countLit, U32* offsetcodeCount, U32* matchlengthCount, U32* litlengthCount, U32* repOffsets,
+                              unsigned* countLit, unsigned* offsetcodeCount, unsigned* matchlengthCount, unsigned* litlengthCount, U32* repOffsets,
                               const void* src, size_t srcSize,
                               U32 notificationLevel)
 {
@@ -602,7 +602,7 @@ static void ZDICT_countEStats(EStats_ress_t esr, ZSTD_parameters params,
 
     }
     cSize = ZSTD_compressBlock(esr.zc, esr.workPlace, ZSTD_BLOCKSIZE_MAX, src, srcSize);
-    if (ZSTD_isError(cSize)) { DISPLAYLEVEL(3, "warning : could not compress sample size %u \n", (U32)srcSize); return; }
+    if (ZSTD_isError(cSize)) { DISPLAYLEVEL(3, "warning : could not compress sample size %u \n", (unsigned)srcSize); return; }
 
     if (cSize) {  /* if == 0; block is not compressible */
         const seqStore_t* const seqStorePtr = ZSTD_getSeqStore(esr.zc);
@@ -671,7 +671,7 @@ static void ZDICT_insertSortCount(offsetCount_t table[ZSTD_REP_NUM+1], U32 val,
  * rewrite `countLit` to contain a mostly flat but still compressible distribution of literals.
  * necessary to avoid generating a non-compressible distribution that HUF_writeCTable() cannot encode.
  */
-static void ZDICT_flatLit(U32* countLit)
+static void ZDICT_flatLit(unsigned* countLit)
 {
     int u;
     for (u=1; u<256; u++) countLit[u] = 2;
@@ -687,14 +687,14 @@ static size_t ZDICT_analyzeEntropy(void*  dstBuffer, size_t maxDstSize,
                              const void* dictBuffer, size_t  dictBufferSize,
                                    unsigned notificationLevel)
 {
-    U32 countLit[256];
+    unsigned countLit[256];
     HUF_CREATE_STATIC_CTABLE(hufTable, 255);
-    U32 offcodeCount[OFFCODE_MAX+1];
+    unsigned offcodeCount[OFFCODE_MAX+1];
     short offcodeNCount[OFFCODE_MAX+1];
     U32 offcodeMax = ZSTD_highbit32((U32)(dictBufferSize + 128 KB));
-    U32 matchLengthCount[MaxML+1];
+    unsigned matchLengthCount[MaxML+1];
     short matchLengthNCount[MaxML+1];
-    U32 litLengthCount[MaxLL+1];
+    unsigned litLengthCount[MaxLL+1];
     short litLengthNCount[MaxLL+1];
     U32 repOffset[MAXREPOFFSET];
     offsetCount_t bestRepOffset[ZSTD_REP_NUM+1];
@@ -983,33 +983,33 @@ size_t ZDICT_trainFromBuffer_unsafe_legacy(
 
     /* display best matches */
     if (params.zParams.notificationLevel>= 3) {
-        U32 const nb = MIN(25, dictList[0].pos);
-        U32 const dictContentSize = ZDICT_dictSize(dictList);
-        U32 u;
-        DISPLAYLEVEL(3, "\n %u segments found, of total size %u \n", dictList[0].pos-1, dictContentSize);
+        unsigned const nb = MIN(25, dictList[0].pos);
+        unsigned const dictContentSize = ZDICT_dictSize(dictList);
+        unsigned u;
+        DISPLAYLEVEL(3, "\n %u segments found, of total size %u \n", (unsigned)dictList[0].pos-1, dictContentSize);
         DISPLAYLEVEL(3, "list %u best segments \n", nb-1);
         for (u=1; u samplesBuffSize) || ((pos + length) > samplesBuffSize)) {
                 free(dictList);
                 return ERROR(GENERIC);   /* should never happen */
             }
             DISPLAYLEVEL(3, "%3u:%3u bytes at pos %8u, savings %7u bytes |",
-                         u, length, pos, dictList[u].savings);
+                         u, length, pos, (unsigned)dictList[u].savings);
             ZDICT_printHex((const char*)samplesBuffer+pos, printedLength);
             DISPLAYLEVEL(3, "| \n");
     }   }
 
 
     /* create dictionary */
-    {   U32 dictContentSize = ZDICT_dictSize(dictList);
+    {   unsigned dictContentSize = ZDICT_dictSize(dictList);
         if (dictContentSize < ZDICT_CONTENTSIZE_MIN) { free(dictList); return ERROR(dictionaryCreation_failed); }   /* dictionary content too small */
         if (dictContentSize < targetDictSize/4) {
-            DISPLAYLEVEL(2, "!  warning : selected content significantly smaller than requested (%u < %u) \n", dictContentSize, (U32)maxDictSize);
+            DISPLAYLEVEL(2, "!  warning : selected content significantly smaller than requested (%u < %u) \n", dictContentSize, (unsigned)maxDictSize);
             if (samplesBuffSize < 10 * targetDictSize)
-                DISPLAYLEVEL(2, "!  consider increasing the number of samples (total size : %u MB)\n", (U32)(samplesBuffSize>>20));
+                DISPLAYLEVEL(2, "!  consider increasing the number of samples (total size : %u MB)\n", (unsigned)(samplesBuffSize>>20));
             if (minRep > MINRATIO) {
                 DISPLAYLEVEL(2, "!  consider increasing selectivity to produce larger dictionary (-s%u) \n", selectivity+1);
                 DISPLAYLEVEL(2, "!  note : larger dictionaries are not necessarily better, test its efficiency on samples \n");
@@ -1017,9 +1017,9 @@ size_t ZDICT_trainFromBuffer_unsafe_legacy(
         }
 
         if ((dictContentSize > targetDictSize*3) && (nbSamples > 2*MINRATIO) && (selectivity>1)) {
-            U32 proposedSelectivity = selectivity-1;
+            unsigned proposedSelectivity = selectivity-1;
             while ((nbSamples >> proposedSelectivity) <= MINRATIO) { proposedSelectivity--; }
-            DISPLAYLEVEL(2, "!  note : calculated dictionary significantly larger than requested (%u > %u) \n", dictContentSize, (U32)maxDictSize);
+            DISPLAYLEVEL(2, "!  note : calculated dictionary significantly larger than requested (%u > %u) \n", dictContentSize, (unsigned)maxDictSize);
             DISPLAYLEVEL(2, "!  consider increasing dictionary size, or produce denser dictionary (-s%u) \n", proposedSelectivity);
             DISPLAYLEVEL(2, "!  always test dictionary efficiency on real samples \n");
         }
diff --git a/lib/legacy/zstd_v04.c b/lib/legacy/zstd_v04.c
index e852bb91116a..d7522025e7f0 100644
--- a/lib/legacy/zstd_v04.c
+++ b/lib/legacy/zstd_v04.c
@@ -240,17 +240,7 @@ MEM_STATIC size_t MEM_readLEST(const void* memPtr)
 /* *************************************
 *  Types
 ***************************************/
-#define ZSTD_WINDOWLOG_MAX 26
-#define ZSTD_WINDOWLOG_MIN 18
 #define ZSTD_WINDOWLOG_ABSOLUTEMIN 11
-#define ZSTD_CONTENTLOG_MAX (ZSTD_WINDOWLOG_MAX+1)
-#define ZSTD_CONTENTLOG_MIN 4
-#define ZSTD_HASHLOG_MAX 28
-#define ZSTD_HASHLOG_MIN 4
-#define ZSTD_SEARCHLOG_MAX (ZSTD_CONTENTLOG_MAX-1)
-#define ZSTD_SEARCHLOG_MIN 1
-#define ZSTD_SEARCHLENGTH_MAX 7
-#define ZSTD_SEARCHLENGTH_MIN 4
 
 /** from faster to stronger */
 typedef enum { ZSTD_fast, ZSTD_greedy, ZSTD_lazy, ZSTD_lazy2, ZSTD_btlazy2 } ZSTD_strategy;
diff --git a/lib/legacy/zstd_v05.c b/lib/legacy/zstd_v05.c
index a1580a271eea..96bffc44a1a4 100644
--- a/lib/legacy/zstd_v05.c
+++ b/lib/legacy/zstd_v05.c
@@ -836,7 +836,7 @@ MEM_STATIC void BITv05_skipBits(BITv05_DStream_t* bitD, U32 nbBits)
     bitD->bitsConsumed += nbBits;
 }
 
-MEM_STATIC size_t BITv05_readBits(BITv05_DStream_t* bitD, U32 nbBits)
+MEM_STATIC size_t BITv05_readBits(BITv05_DStream_t* bitD, unsigned nbBits)
 {
     size_t value = BITv05_lookBits(bitD, nbBits);
     BITv05_skipBits(bitD, nbBits);
@@ -845,7 +845,7 @@ MEM_STATIC size_t BITv05_readBits(BITv05_DStream_t* bitD, U32 nbBits)
 
 /*!BITv05_readBitsFast :
 *  unsafe version; only works only if nbBits >= 1 */
-MEM_STATIC size_t BITv05_readBitsFast(BITv05_DStream_t* bitD, U32 nbBits)
+MEM_STATIC size_t BITv05_readBitsFast(BITv05_DStream_t* bitD, unsigned nbBits)
 {
     size_t value = BITv05_lookBitsFast(bitD, nbBits);
     BITv05_skipBits(bitD, nbBits);
@@ -1162,7 +1162,7 @@ MEM_STATIC unsigned FSEv05_endOfDState(const FSEv05_DState_t* DStatePtr)
 /* **************************************************************
 *  Complex types
 ****************************************************************/
-typedef U32 DTable_max_t[FSEv05_DTABLE_SIZE_U32(FSEv05_MAX_TABLELOG)];
+typedef unsigned DTable_max_t[FSEv05_DTABLE_SIZE_U32(FSEv05_MAX_TABLELOG)];
 
 
 /* **************************************************************
@@ -2191,7 +2191,7 @@ static void HUFv05_fillDTableX4(HUFv05_DEltX4* DTable, const U32 targetLog,
     }
 }
 
-size_t HUFv05_readDTableX4 (U32* DTable, const void* src, size_t srcSize)
+size_t HUFv05_readDTableX4 (unsigned* DTable, const void* src, size_t srcSize)
 {
     BYTE weightList[HUFv05_MAX_SYMBOL_VALUE + 1];
     sortedSymbol_t sortedSymbol[HUFv05_MAX_SYMBOL_VALUE + 1];
@@ -2205,7 +2205,7 @@ size_t HUFv05_readDTableX4 (U32* DTable, const void* src, size_t srcSize)
     void* dtPtr = DTable;
     HUFv05_DEltX4* const dt = ((HUFv05_DEltX4*)dtPtr) + 1;
 
-    HUFv05_STATIC_ASSERT(sizeof(HUFv05_DEltX4) == sizeof(U32));   /* if compilation fails here, assertion is false */
+    HUFv05_STATIC_ASSERT(sizeof(HUFv05_DEltX4) == sizeof(unsigned));   /* if compilation fails here, assertion is false */
     if (memLog > HUFv05_ABSOLUTEMAX_TABLELOG) return ERROR(tableLog_tooLarge);
     //memset(weightList, 0, sizeof(weightList));   /* is not necessary, even though some analyzer complain ... */
 
@@ -2332,7 +2332,7 @@ static inline size_t HUFv05_decodeStreamX4(BYTE* p, BITv05_DStream_t* bitDPtr, B
 size_t HUFv05_decompress1X4_usingDTable(
           void* dst,  size_t dstSize,
     const void* cSrc, size_t cSrcSize,
-    const U32* DTable)
+    const unsigned* DTable)
 {
     const BYTE* const istart = (const BYTE*) cSrc;
     BYTE* const ostart = (BYTE*) dst;
@@ -2375,7 +2375,7 @@ size_t HUFv05_decompress1X4 (void* dst, size_t dstSize, const void* cSrc, size_t
 size_t HUFv05_decompress4X4_usingDTable(
           void* dst,  size_t dstSize,
     const void* cSrc, size_t cSrcSize,
-    const U32* DTable)
+    const unsigned* DTable)
 {
     if (cSrcSize < 10) return ERROR(corruption_detected);   /* strict minimum : jump table + 1 byte per stream */
 
@@ -2999,7 +2999,7 @@ static size_t ZSTDv05_decodeSeqHeaders(int* nbSeq, const BYTE** dumpsPtr, size_t
     const BYTE* ip = istart;
     const BYTE* const iend = istart + srcSize;
     U32 LLtype, Offtype, MLtype;
-    U32 LLlog, Offlog, MLlog;
+    unsigned LLlog, Offlog, MLlog;
     size_t dumpsLength;
 
     /* check */
@@ -3057,7 +3057,7 @@ static size_t ZSTDv05_decodeSeqHeaders(int* nbSeq, const BYTE** dumpsPtr, size_t
             break;
         case FSEv05_ENCODING_DYNAMIC :
         default :   /* impossible */
-            {   U32 max = MaxLL;
+            {   unsigned max = MaxLL;
                 headerSize = FSEv05_readNCount(norm, &max, &LLlog, ip, iend-ip);
                 if (FSEv05_isError(headerSize)) return ERROR(GENERIC);
                 if (LLlog > LLFSEv05Log) return ERROR(corruption_detected);
@@ -3081,7 +3081,7 @@ static size_t ZSTDv05_decodeSeqHeaders(int* nbSeq, const BYTE** dumpsPtr, size_t
             break;
         case FSEv05_ENCODING_DYNAMIC :
         default :   /* impossible */
-            {   U32 max = MaxOff;
+            {   unsigned max = MaxOff;
                 headerSize = FSEv05_readNCount(norm, &max, &Offlog, ip, iend-ip);
                 if (FSEv05_isError(headerSize)) return ERROR(GENERIC);
                 if (Offlog > OffFSEv05Log) return ERROR(corruption_detected);
@@ -3105,7 +3105,7 @@ static size_t ZSTDv05_decodeSeqHeaders(int* nbSeq, const BYTE** dumpsPtr, size_t
             break;
         case FSEv05_ENCODING_DYNAMIC :
         default :   /* impossible */
-            {   U32 max = MaxML;
+            {   unsigned max = MaxML;
                 headerSize = FSEv05_readNCount(norm, &max, &MLlog, ip, iend-ip);
                 if (FSEv05_isError(headerSize)) return ERROR(GENERIC);
                 if (MLlog > MLFSEv05Log) return ERROR(corruption_detected);
@@ -3305,9 +3305,9 @@ static size_t ZSTDv05_decompressSequences(
     const BYTE* const litEnd = litPtr + dctx->litSize;
     int nbSeq=0;
     const BYTE* dumps = NULL;
-    U32* DTableLL = dctx->LLTable;
-    U32* DTableML = dctx->MLTable;
-    U32* DTableOffb = dctx->OffTable;
+    unsigned* DTableLL = dctx->LLTable;
+    unsigned* DTableML = dctx->MLTable;
+    unsigned* DTableOffb = dctx->OffTable;
     const BYTE* const base = (const BYTE*) (dctx->base);
     const BYTE* const vBase = (const BYTE*) (dctx->vBase);
     const BYTE* const dictEnd = (const BYTE*) (dctx->dictEnd);
@@ -3633,7 +3633,7 @@ static size_t ZSTDv05_loadEntropy(ZSTDv05_DCtx* dctx, const void* dict, size_t d
 {
     size_t hSize, offcodeHeaderSize, matchlengthHeaderSize, errorCode, litlengthHeaderSize;
     short offcodeNCount[MaxOff+1];
-    U32 offcodeMaxValue=MaxOff, offcodeLog;
+    unsigned offcodeMaxValue=MaxOff, offcodeLog;
     short matchlengthNCount[MaxML+1];
     unsigned matchlengthMaxValue = MaxML, matchlengthLog;
     short litlengthNCount[MaxLL+1];
diff --git a/lib/zstd.h b/lib/zstd.h
index f2af4ac8c429..b18fc8a44bf9 100644
--- a/lib/zstd.h
+++ b/lib/zstd.h
@@ -71,16 +71,16 @@ extern "C" {
 /*------   Version   ------*/
 #define ZSTD_VERSION_MAJOR    1
 #define ZSTD_VERSION_MINOR    3
-#define ZSTD_VERSION_RELEASE  7
+#define ZSTD_VERSION_RELEASE  8
 
 #define ZSTD_VERSION_NUMBER  (ZSTD_VERSION_MAJOR *100*100 + ZSTD_VERSION_MINOR *100 + ZSTD_VERSION_RELEASE)
-ZSTDLIB_API unsigned ZSTD_versionNumber(void);   /**< useful to check dll version */
+ZSTDLIB_API unsigned ZSTD_versionNumber(void);   /**< to check runtime library version */
 
 #define ZSTD_LIB_VERSION ZSTD_VERSION_MAJOR.ZSTD_VERSION_MINOR.ZSTD_VERSION_RELEASE
 #define ZSTD_QUOTE(str) #str
 #define ZSTD_EXPAND_AND_QUOTE(str) ZSTD_QUOTE(str)
 #define ZSTD_VERSION_STRING ZSTD_EXPAND_AND_QUOTE(ZSTD_LIB_VERSION)
-ZSTDLIB_API const char* ZSTD_versionString(void);   /* v1.3.0+ */
+ZSTDLIB_API const char* ZSTD_versionString(void);   /* requires v1.3.0+ */
 
 /***************************************
 *  Default constant
@@ -110,7 +110,7 @@ ZSTDLIB_API size_t ZSTD_compress( void* dst, size_t dstCapacity,
 ZSTDLIB_API size_t ZSTD_decompress( void* dst, size_t dstCapacity,
                               const void* src, size_t compressedSize);
 
-/*! ZSTD_getFrameContentSize() : added in v1.3.0
+/*! ZSTD_getFrameContentSize() : requires v1.3.0+
  *  `src` should point to the start of a ZSTD encoded frame.
  *  `srcSize` must be at least as large as the frame header.
  *            hint : any size >= `ZSTD_frameHeaderSize_max` is large enough.
@@ -167,8 +167,10 @@ ZSTDLIB_API ZSTD_CCtx* ZSTD_createCCtx(void);
 ZSTDLIB_API size_t     ZSTD_freeCCtx(ZSTD_CCtx* cctx);
 
 /*! ZSTD_compressCCtx() :
- *  Same as ZSTD_compress(), requires an allocated ZSTD_CCtx (see ZSTD_createCCtx()). */
-ZSTDLIB_API size_t ZSTD_compressCCtx(ZSTD_CCtx* ctx,
+ *  Same as ZSTD_compress(), using an explicit ZSTD_CCtx
+ *  The function will compress at requested compression level,
+ *  ignoring any other parameter */
+ZSTDLIB_API size_t ZSTD_compressCCtx(ZSTD_CCtx* cctx,
                                      void* dst, size_t dstCapacity,
                                const void* src, size_t srcSize,
                                      int compressionLevel);
@@ -184,8 +186,11 @@ ZSTDLIB_API ZSTD_DCtx* ZSTD_createDCtx(void);
 ZSTDLIB_API size_t     ZSTD_freeDCtx(ZSTD_DCtx* dctx);
 
 /*! ZSTD_decompressDCtx() :
- *  Same as ZSTD_decompress(), requires an allocated ZSTD_DCtx (see ZSTD_createDCtx()) */
-ZSTDLIB_API size_t ZSTD_decompressDCtx(ZSTD_DCtx* ctx,
+ *  Same as ZSTD_decompress(),
+ *  requires an allocated ZSTD_DCtx.
+ *  Compatible with sticky parameters.
+ */
+ZSTDLIB_API size_t ZSTD_decompressDCtx(ZSTD_DCtx* dctx,
                                        void* dst, size_t dstCapacity,
                                  const void* src, size_t srcSize);
 
@@ -194,9 +199,12 @@ ZSTDLIB_API size_t ZSTD_decompressDCtx(ZSTD_DCtx* ctx,
 *  Simple dictionary API
 ***************************/
 /*! ZSTD_compress_usingDict() :
- *  Compression using a predefined Dictionary (see dictBuilder/zdict.h).
+ *  Compression at an explicit compression level using a Dictionary.
+ *  A dictionary can be any arbitrary data segment (also called a prefix),
+ *  or a buffer with specified information (see dictBuilder/zdict.h).
  *  Note : This function loads the dictionary, resulting in significant startup delay.
- *  Note : When `dict == NULL || dictSize < 8` no dictionary is used. */
+ *         It's intended for a dictionary used only once.
+ *  Note 2 : When `dict == NULL || dictSize < 8` no dictionary is used. */
 ZSTDLIB_API size_t ZSTD_compress_usingDict(ZSTD_CCtx* ctx,
                                            void* dst, size_t dstCapacity,
                                      const void* src, size_t srcSize,
@@ -204,9 +212,10 @@ ZSTDLIB_API size_t ZSTD_compress_usingDict(ZSTD_CCtx* ctx,
                                            int compressionLevel);
 
 /*! ZSTD_decompress_usingDict() :
- *  Decompression using a predefined Dictionary (see dictBuilder/zdict.h).
+ *  Decompression using a known Dictionary.
  *  Dictionary must be identical to the one used during compression.
  *  Note : This function loads the dictionary, resulting in significant startup delay.
+ *         It's intended for a dictionary used only once.
  *  Note : When `dict == NULL || dictSize < 8` no dictionary is used. */
 ZSTDLIB_API size_t ZSTD_decompress_usingDict(ZSTD_DCtx* dctx,
                                              void* dst, size_t dstCapacity,
@@ -214,17 +223,18 @@ ZSTDLIB_API size_t ZSTD_decompress_usingDict(ZSTD_DCtx* dctx,
                                        const void* dict,size_t dictSize);
 
 
-/**********************************
+/***********************************
  *  Bulk processing dictionary API
- *********************************/
+ **********************************/
 typedef struct ZSTD_CDict_s ZSTD_CDict;
 
 /*! ZSTD_createCDict() :
- *  When compressing multiple messages / blocks with the same dictionary, it's recommended to load it just once.
- *  ZSTD_createCDict() will create a digested dictionary, ready to start future compression operations without startup delay.
+ *  When compressing multiple messages / blocks using the same dictionary, it's recommended to load it only once.
+ *  ZSTD_createCDict() will create a digested dictionary, ready to start future compression operations without startup cost.
  *  ZSTD_CDict can be created once and shared by multiple threads concurrently, since its usage is read-only.
- *  `dictBuffer` can be released after ZSTD_CDict creation, since its content is copied within CDict
- *  Note : A ZSTD_CDict can be created with an empty dictionary, but it is inefficient for small data. */
+ * `dictBuffer` can be released after ZSTD_CDict creation, because its content is copied within CDict.
+ *  Consider experimental function `ZSTD_createCDict_byReference()` if you prefer to not duplicate `dictBuffer` content.
+ *  Note : A ZSTD_CDict can be created from an empty dictBuffer, but it is inefficient when used to compress small data. */
 ZSTDLIB_API ZSTD_CDict* ZSTD_createCDict(const void* dictBuffer, size_t dictSize,
                                          int compressionLevel);
 
@@ -234,11 +244,9 @@ ZSTDLIB_API size_t      ZSTD_freeCDict(ZSTD_CDict* CDict);
 
 /*! ZSTD_compress_usingCDict() :
  *  Compression using a digested Dictionary.
- *  Faster startup than ZSTD_compress_usingDict(), recommended when same dictionary is used multiple times.
- *  Note that compression level is decided during dictionary creation.
- *  Frame parameters are hardcoded (dictID=yes, contentSize=yes, checksum=no)
- *  Note : ZSTD_compress_usingCDict() can be used with a ZSTD_CDict created from an empty dictionary.
- *         But it is inefficient for small data, and it is recommended to use ZSTD_compressCCtx(). */
+ *  Recommended when same dictionary is used multiple times.
+ *  Note : compression level is _decided at dictionary creation time_,
+ *     and frame parameters are hardcoded (dictID=yes, contentSize=yes, checksum=no) */
 ZSTDLIB_API size_t ZSTD_compress_usingCDict(ZSTD_CCtx* cctx,
                                             void* dst, size_t dstCapacity,
                                       const void* src, size_t srcSize,
@@ -249,7 +257,7 @@ typedef struct ZSTD_DDict_s ZSTD_DDict;
 
 /*! ZSTD_createDDict() :
  *  Create a digested dictionary, ready to start decompression operation without startup delay.
- *  dictBuffer can be released after DDict creation, as its content is copied inside DDict */
+ *  dictBuffer can be released after DDict creation, as its content is copied inside DDict. */
 ZSTDLIB_API ZSTD_DDict* ZSTD_createDDict(const void* dictBuffer, size_t dictSize);
 
 /*! ZSTD_freeDDict() :
@@ -258,7 +266,7 @@ ZSTDLIB_API size_t      ZSTD_freeDDict(ZSTD_DDict* ddict);
 
 /*! ZSTD_decompress_usingDDict() :
  *  Decompression using a digested Dictionary.
- *  Faster startup than ZSTD_decompress_usingDict(), recommended when same dictionary is used multiple times. */
+ *  Recommended when same dictionary is used multiple times. */
 ZSTDLIB_API size_t ZSTD_decompress_usingDDict(ZSTD_DCtx* dctx,
                                               void* dst, size_t dstCapacity,
                                         const void* src, size_t srcSize,
@@ -289,13 +297,17 @@ typedef struct ZSTD_outBuffer_s {
 *  A ZSTD_CStream object is required to track streaming operation.
 *  Use ZSTD_createCStream() and ZSTD_freeCStream() to create/release resources.
 *  ZSTD_CStream objects can be reused multiple times on consecutive compression operations.
-*  It is recommended to re-use ZSTD_CStream in situations where many streaming operations will be achieved consecutively,
-*  since it will play nicer with system's memory, by re-using already allocated memory.
-*  Use one separate ZSTD_CStream per thread for parallel execution.
+*  It is recommended to re-use ZSTD_CStream since it will play nicer with system's memory, by re-using already allocated memory.
 *
-*  Start a new compression by initializing ZSTD_CStream context.
-*  Use ZSTD_initCStream() to start a new compression operation.
-*  Use variants ZSTD_initCStream_usingDict() or ZSTD_initCStream_usingCDict() for streaming with dictionary (experimental section)
+*  For parallel execution, use one separate ZSTD_CStream per thread.
+*
+*  note : since v1.3.0, ZSTD_CStream and ZSTD_CCtx are the same thing.
+*
+*  Parameters are sticky : when starting a new compression on the same context,
+*  it will re-use the same sticky parameters as previous compression session.
+*  When in doubt, it's recommended to fully initialize the context before usage.
+*  Use ZSTD_initCStream() to set the parameter to a selected compression level.
+*  Use advanced API (ZSTD_CCtx_setParameter(), etc.) to set more specific parameters.
 *
 *  Use ZSTD_compressStream() as many times as necessary to consume input stream.
 *  The function will automatically update both `pos` fields within `input` and `output`.
@@ -304,12 +316,11 @@ typedef struct ZSTD_outBuffer_s {
 *  in which case `input.pos < input.size`.
 *  The caller must check if input has been entirely consumed.
 *  If not, the caller must make some room to receive more compressed data,
-*  typically by emptying output buffer, or allocating a new output buffer,
 *  and then present again remaining input data.
-*  @return : a size hint, preferred nb of bytes to use as input for next function call
-*            or an error code, which can be tested using ZSTD_isError().
-*            Note 1 : it's just a hint, to help latency a little, any other value will work fine.
-*            Note 2 : size hint is guaranteed to be <= ZSTD_CStreamInSize()
+* @return : a size hint, preferred nb of bytes to use as input for next function call
+*           or an error code, which can be tested using ZSTD_isError().
+*           Note 1 : it's just a hint, to help latency a little, any value will work fine.
+*           Note 2 : size hint is guaranteed to be <= ZSTD_CStreamInSize()
 *
 *  At any moment, it's possible to flush whatever data might remain stuck within internal buffer,
 *  using ZSTD_flushStream(). `output->pos` will be updated.
@@ -353,25 +364,24 @@ ZSTDLIB_API size_t ZSTD_CStreamOutSize(void);   /**< recommended size for output
 *  Use ZSTD_createDStream() and ZSTD_freeDStream() to create/release resources.
 *  ZSTD_DStream objects can be re-used multiple times.
 *
-*  Use ZSTD_initDStream() to start a new decompression operation,
-*   or ZSTD_initDStream_usingDict() if decompression requires a dictionary.
-*   @return : recommended first input size
+*  Use ZSTD_initDStream() to start a new decompression operation.
+* @return : recommended first input size
+*  Alternatively, use advanced API to set specific properties.
 *
 *  Use ZSTD_decompressStream() repetitively to consume your input.
 *  The function will update both `pos` fields.
 *  If `input.pos < input.size`, some input has not been consumed.
 *  It's up to the caller to present again remaining data.
-*  The function tries to flush all data decoded immediately, repecting buffer sizes.
+*  The function tries to flush all data decoded immediately, respecting output buffer size.
 *  If `output.pos < output.size`, decoder has flushed everything it could.
-*  But if `output.pos == output.size`, there is no such guarantee,
-*  it's likely that some decoded data was not flushed and still remains within internal buffers.
+*  But if `output.pos == output.size`, there might be some data left within internal buffers.,
 *  In which case, call ZSTD_decompressStream() again to flush whatever remains in the buffer.
-*  When no additional input is provided, amount of data flushed is necessarily <= ZSTD_BLOCKSIZE_MAX.
+*  Note : with no additional input provided, amount of data flushed is necessarily <= ZSTD_BLOCKSIZE_MAX.
 * @return : 0 when a frame is completely decoded and fully flushed,
 *        or an error code, which can be tested using ZSTD_isError(),
 *        or any other value > 0, which means there is still some decoding or flushing to do to complete current frame :
-*                                the return value is a suggested next input size (a hint for better latency)
-*                                that will never load more than the current frame.
+*                                the return value is a suggested next input size (just a hint for better latency)
+*                                that will never request more than the remaining frame size.
 * *******************************************************************************/
 
 typedef ZSTD_DCtx ZSTD_DStream;  /**< DCtx and DStream are now effectively same object (>= v1.3.0) */
@@ -392,77 +402,602 @@ ZSTDLIB_API size_t ZSTD_DStreamOutSize(void);   /*!< recommended size for output
 
 
 
-#if defined(ZSTD_STATIC_LINKING_ONLY) && !defined(ZSTD_H_ZSTD_STATIC_LINKING_ONLY)
-#define ZSTD_H_ZSTD_STATIC_LINKING_ONLY
-
 /****************************************************************************************
  *   ADVANCED AND EXPERIMENTAL FUNCTIONS
  ****************************************************************************************
- * The definitions in this section are considered experimental.
- * They should never be used with a dynamic library, as prototypes may change in the future.
+ * The definitions in the following section are considered experimental.
  * They are provided for advanced scenarios.
+ * They should never be used with a dynamic library, as prototypes may change in the future.
  * Use them only in association with static linking.
  * ***************************************************************************************/
 
+#if defined(ZSTD_STATIC_LINKING_ONLY) && !defined(ZSTD_H_ZSTD_STATIC_LINKING_ONLY)
+#define ZSTD_H_ZSTD_STATIC_LINKING_ONLY
+
+
+/****************************************************************************************
+ *   Candidate API for promotion to stable status
+ ****************************************************************************************
+ * The following symbols and constants form the "staging area" :
+ * they are considered to join "stable API" by v1.4.0.
+ * The proposal is written so that it can be made stable "as is",
+ * though it's still possible to suggest improvements.
+ * Staging is in fact last chance for changes,
+ * the API is locked once reaching "stable" status.
+ * ***************************************************************************************/
+
+
+/* ===  Constants   === */
+
+/* all magic numbers are supposed read/written to/from files/memory using little-endian convention */
+#define ZSTD_MAGICNUMBER            0xFD2FB528    /* valid since v0.8.0 */
+#define ZSTD_MAGIC_DICTIONARY       0xEC30A437    /* valid since v0.7.0 */
+#define ZSTD_MAGIC_SKIPPABLE_START  0x184D2A50    /* all 16 values, from 0x184D2A50 to 0x184D2A5F, signal the beginning of a skippable frame */
+#define ZSTD_MAGIC_SKIPPABLE_MASK   0xFFFFFFF0
+
+#define ZSTD_BLOCKSIZELOG_MAX  17
+#define ZSTD_BLOCKSIZE_MAX     (1<= first frame size
+ * @return : the compressed size of the first frame starting at `src`,
+ *           suitable to pass as `srcSize` to `ZSTD_decompress` or similar,
+ *        or an error code if input is invalid */
+ZSTDLIB_API size_t ZSTD_findFrameCompressedSize(const void* src, size_t srcSize);
 
-#define ZSTD_FRAMEHEADERSIZE_PREFIX 5   /* minimum input size to know frame header size */
+
+/* ===   Memory management   === */
+
+/*! ZSTD_sizeof_*() :
+ *  These functions give the _current_ memory usage of selected object.
+ *  Note that object memory usage can evolve (increase or decrease) over time. */
+ZSTDLIB_API size_t ZSTD_sizeof_CCtx(const ZSTD_CCtx* cctx);
+ZSTDLIB_API size_t ZSTD_sizeof_DCtx(const ZSTD_DCtx* dctx);
+ZSTDLIB_API size_t ZSTD_sizeof_CStream(const ZSTD_CStream* zcs);
+ZSTDLIB_API size_t ZSTD_sizeof_DStream(const ZSTD_DStream* zds);
+ZSTDLIB_API size_t ZSTD_sizeof_CDict(const ZSTD_CDict* cdict);
+ZSTDLIB_API size_t ZSTD_sizeof_DDict(const ZSTD_DDict* ddict);
+
+
+/***************************************
+*  Advanced compression API
+***************************************/
+
+/* API design :
+ *   Parameters are pushed one by one into an existing context,
+ *   using ZSTD_CCtx_set*() functions.
+ *   Pushed parameters are sticky : they are valid for next compressed frame, and any subsequent frame.
+ *   "sticky" parameters are applicable to `ZSTD_compress2()` and `ZSTD_compressStream*()` !
+ *   They do not apply to "simple" one-shot variants such as ZSTD_compressCCtx()
+ *
+ *   It's possible to reset all parameters to "default" using ZSTD_CCtx_reset().
+ *
+ *   This API supercedes all other "advanced" API entry points in the experimental section.
+ *   In the future, we expect to remove from experimental API entry points which are redundant with this API.
+ */
+
+
+/* Compression strategies, listed from fastest to strongest */
+typedef enum { ZSTD_fast=1,
+               ZSTD_dfast=2,
+               ZSTD_greedy=3,
+               ZSTD_lazy=4,
+               ZSTD_lazy2=5,
+               ZSTD_btlazy2=6,
+               ZSTD_btopt=7,
+               ZSTD_btultra=8,
+               ZSTD_btultra2=9
+               /* note : new strategies _might_ be added in the future.
+                         Only the order (from fast to strong) is guaranteed */
+} ZSTD_strategy;
+
+
+typedef enum {
+
+    /* compression parameters */
+    ZSTD_c_compressionLevel=100, /* Update all compression parameters according to pre-defined cLevel table
+                              * Default level is ZSTD_CLEVEL_DEFAULT==3.
+                              * Special: value 0 means default, which is controlled by ZSTD_CLEVEL_DEFAULT.
+                              * Note 1 : it's possible to pass a negative compression level.
+                              * Note 2 : setting a level sets all default values of other compression parameters */
+    ZSTD_c_windowLog=101,    /* Maximum allowed back-reference distance, expressed as power of 2.
+                              * Must be clamped between ZSTD_WINDOWLOG_MIN and ZSTD_WINDOWLOG_MAX.
+                              * Special: value 0 means "use default windowLog".
+                              * Note: Using a windowLog greater than ZSTD_WINDOWLOG_LIMIT_DEFAULT
+                              *       requires explicitly allowing such window size at decompression stage if using streaming. */
+    ZSTD_c_hashLog=102,      /* Size of the initial probe table, as a power of 2.
+                              * Resulting memory usage is (1 << (hashLog+2)).
+                              * Must be clamped between ZSTD_HASHLOG_MIN and ZSTD_HASHLOG_MAX.
+                              * Larger tables improve compression ratio of strategies <= dFast,
+                              * and improve speed of strategies > dFast.
+                              * Special: value 0 means "use default hashLog". */
+    ZSTD_c_chainLog=103,     /* Size of the multi-probe search table, as a power of 2.
+                              * Resulting memory usage is (1 << (chainLog+2)).
+                              * Must be clamped between ZSTD_CHAINLOG_MIN and ZSTD_CHAINLOG_MAX.
+                              * Larger tables result in better and slower compression.
+                              * This parameter is useless when using "fast" strategy.
+                              * It's still useful when using "dfast" strategy,
+                              * in which case it defines a secondary probe table.
+                              * Special: value 0 means "use default chainLog". */
+    ZSTD_c_searchLog=104,    /* Number of search attempts, as a power of 2.
+                              * More attempts result in better and slower compression.
+                              * This parameter is useless when using "fast" and "dFast" strategies.
+                              * Special: value 0 means "use default searchLog". */
+    ZSTD_c_minMatch=105,     /* Minimum size of searched matches.
+                              * Note that Zstandard can still find matches of smaller size,
+                              * it just tweaks its search algorithm to look for this size and larger.
+                              * Larger values increase compression and decompression speed, but decrease ratio.
+                              * Must be clamped between ZSTD_MINMATCH_MIN and ZSTD_MINMATCH_MAX.
+                              * Note that currently, for all strategies < btopt, effective minimum is 4.
+                              *                    , for all strategies > fast, effective maximum is 6.
+                              * Special: value 0 means "use default minMatchLength". */
+    ZSTD_c_targetLength=106, /* Impact of this field depends on strategy.
+                              * For strategies btopt, btultra & btultra2:
+                              *     Length of Match considered "good enough" to stop search.
+                              *     Larger values make compression stronger, and slower.
+                              * For strategy fast:
+                              *     Distance between match sampling.
+                              *     Larger values make compression faster, and weaker.
+                              * Special: value 0 means "use default targetLength". */
+    ZSTD_c_strategy=107,     /* See ZSTD_strategy enum definition.
+                              * The higher the value of selected strategy, the more complex it is,
+                              * resulting in stronger and slower compression.
+                              * Special: value 0 means "use default strategy". */
+
+    /* LDM mode parameters */
+    ZSTD_c_enableLongDistanceMatching=160, /* Enable long distance matching.
+                                     * This parameter is designed to improve compression ratio
+                                     * for large inputs, by finding large matches at long distance.
+                                     * It increases memory usage and window size.
+                                     * Note: enabling this parameter increases default ZSTD_c_windowLog to 128 MB
+                                     * except when expressly set to a different value. */
+    ZSTD_c_ldmHashLog=161,   /* Size of the table for long distance matching, as a power of 2.
+                              * Larger values increase memory usage and compression ratio,
+                              * but decrease compression speed.
+                              * Must be clamped between ZSTD_HASHLOG_MIN and ZSTD_HASHLOG_MAX
+                              * default: windowlog - 7.
+                              * Special: value 0 means "automatically determine hashlog". */
+    ZSTD_c_ldmMinMatch=162,  /* Minimum match size for long distance matcher.
+                              * Larger/too small values usually decrease compression ratio.
+                              * Must be clamped between ZSTD_LDM_MINMATCH_MIN and ZSTD_LDM_MINMATCH_MAX.
+                              * Special: value 0 means "use default value" (default: 64). */
+    ZSTD_c_ldmBucketSizeLog=163, /* Log size of each bucket in the LDM hash table for collision resolution.
+                              * Larger values improve collision resolution but decrease compression speed.
+                              * The maximum value is ZSTD_LDM_BUCKETSIZELOG_MAX.
+                              * Special: value 0 means "use default value" (default: 3). */
+    ZSTD_c_ldmHashRateLog=164, /* Frequency of inserting/looking up entries into the LDM hash table.
+                              * Must be clamped between 0 and (ZSTD_WINDOWLOG_MAX - ZSTD_HASHLOG_MIN).
+                              * Default is MAX(0, (windowLog - ldmHashLog)), optimizing hash table usage.
+                              * Larger values improve compression speed.
+                              * Deviating far from default value will likely result in a compression ratio decrease.
+                              * Special: value 0 means "automatically determine hashRateLog". */
+
+    /* frame parameters */
+    ZSTD_c_contentSizeFlag=200, /* Content size will be written into frame header _whenever known_ (default:1)
+                              * Content size must be known at the beginning of compression.
+                              * This is automatically the case when using ZSTD_compress2(),
+                              * For streaming variants, content size must be provided with ZSTD_CCtx_setPledgedSrcSize() */
+    ZSTD_c_checksumFlag=201, /* A 32-bits checksum of content is written at end of frame (default:0) */
+    ZSTD_c_dictIDFlag=202,   /* When applicable, dictionary's ID is written into frame header (default:1) */
+
+    /* multi-threading parameters */
+    /* These parameters are only useful if multi-threading is enabled (compiled with build macro ZSTD_MULTITHREAD).
+     * They return an error otherwise. */
+    ZSTD_c_nbWorkers=400,    /* Select how many threads will be spawned to compress in parallel.
+                              * When nbWorkers >= 1, triggers asynchronous mode when used with ZSTD_compressStream*() :
+                              * ZSTD_compressStream*() consumes input and flush output if possible, but immediately gives back control to caller,
+                              * while compression work is performed in parallel, within worker threads.
+                              * (note : a strong exception to this rule is when first invocation of ZSTD_compressStream2() sets ZSTD_e_end :
+                              *  in which case, ZSTD_compressStream2() delegates to ZSTD_compress2(), which is always a blocking call).
+                              * More workers improve speed, but also increase memory usage.
+                              * Default value is `0`, aka "single-threaded mode" : no worker is spawned, compression is performed inside Caller's thread, all invocations are blocking */
+    ZSTD_c_jobSize=401,      /* Size of a compression job. This value is enforced only when nbWorkers >= 1.
+                              * Each compression job is completed in parallel, so this value can indirectly impact the nb of active threads.
+                              * 0 means default, which is dynamically determined based on compression parameters.
+                              * Job size must be a minimum of overlap size, or 1 MB, whichever is largest.
+                              * The minimum size is automatically and transparently enforced */
+    ZSTD_c_overlapLog=402,   /* Control the overlap size, as a fraction of window size.
+                              * The overlap size is an amount of data reloaded from previous job at the beginning of a new job.
+                              * It helps preserve compression ratio, while each job is compressed in parallel.
+                              * This value is enforced only when nbWorkers >= 1.
+                              * Larger values increase compression ratio, but decrease speed.
+                              * Possible values range from 0 to 9 :
+                              * - 0 means "default" : value will be determined by the library, depending on strategy
+                              * - 1 means "no overlap"
+                              * - 9 means "full overlap", using a full window size.
+                              * Each intermediate rank increases/decreases load size by a factor 2 :
+                              * 9: full window;  8: w/2;  7: w/4;  6: w/8;  5:w/16;  4: w/32;  3:w/64;  2:w/128;  1:no overlap;  0:default
+                              * default value varies between 6 and 9, depending on strategy */
+
+    /* note : additional experimental parameters are also available
+     * within the experimental section of the API.
+     * At the time of this writing, they include :
+     * ZSTD_c_rsyncable
+     * ZSTD_c_format
+     * ZSTD_c_forceMaxWindow
+     * ZSTD_c_forceAttachDict
+     * Because they are not stable, it's necessary to define ZSTD_STATIC_LINKING_ONLY to access them.
+     * note : never ever use experimentalParam? names directly;
+     *        also, the enums values themselves are unstable and can still change.
+     */
+     ZSTD_c_experimentalParam1=500,
+     ZSTD_c_experimentalParam2=10,
+     ZSTD_c_experimentalParam3=1000,
+     ZSTD_c_experimentalParam4=1001
+} ZSTD_cParameter;
+
+
+typedef struct {
+    size_t error;
+    int lowerBound;
+    int upperBound;
+} ZSTD_bounds;
+
+/*! ZSTD_cParam_getBounds() :
+ *  All parameters must belong to an interval with lower and upper bounds,
+ *  otherwise they will either trigger an error or be automatically clamped.
+ * @return : a structure, ZSTD_bounds, which contains
+ *         - an error status field, which must be tested using ZSTD_isError()
+ *         - lower and upper bounds, both inclusive
+ */
+ZSTDLIB_API ZSTD_bounds ZSTD_cParam_getBounds(ZSTD_cParameter cParam);
+
+/*! ZSTD_CCtx_setParameter() :
+ *  Set one compression parameter, selected by enum ZSTD_cParameter.
+ *  All parameters have valid bounds. Bounds can be queried using ZSTD_cParam_getBounds().
+ *  Providing a value beyond bound will either clamp it, or trigger an error (depending on parameter).
+ *  Setting a parameter is generally only possible during frame initialization (before starting compression).
+ *  Exception : when using multi-threading mode (nbWorkers >= 1),
+ *              the following parameters can be updated _during_ compression (within same frame):
+ *              => compressionLevel, hashLog, chainLog, searchLog, minMatch, targetLength and strategy.
+ *              new parameters will be active for next job only (after a flush()).
+ * @return : an error code (which can be tested using ZSTD_isError()).
+ */
+ZSTDLIB_API size_t ZSTD_CCtx_setParameter(ZSTD_CCtx* cctx, ZSTD_cParameter param, int value);
+
+/*! ZSTD_CCtx_setPledgedSrcSize() :
+ *  Total input data size to be compressed as a single frame.
+ *  Value will be written in frame header, unless if explicitly forbidden using ZSTD_c_contentSizeFlag.
+ *  This value will also be controlled at end of frame, and trigger an error if not respected.
+ * @result : 0, or an error code (which can be tested with ZSTD_isError()).
+ *  Note 1 : pledgedSrcSize==0 actually means zero, aka an empty frame.
+ *           In order to mean "unknown content size", pass constant ZSTD_CONTENTSIZE_UNKNOWN.
+ *           ZSTD_CONTENTSIZE_UNKNOWN is default value for any new frame.
+ *  Note 2 : pledgedSrcSize is only valid once, for the next frame.
+ *           It's discarded at the end of the frame, and replaced by ZSTD_CONTENTSIZE_UNKNOWN.
+ *  Note 3 : Whenever all input data is provided and consumed in a single round,
+ *           for example with ZSTD_compress2(),
+ *           or invoking immediately ZSTD_compressStream2(,,,ZSTD_e_end),
+ *           this value is automatically overriden by srcSize instead.
+ */
+ZSTDLIB_API size_t ZSTD_CCtx_setPledgedSrcSize(ZSTD_CCtx* cctx, unsigned long long pledgedSrcSize);
+
+/*! ZSTD_CCtx_loadDictionary() :
+ *  Create an internal CDict from `dict` buffer.
+ *  Decompression will have to use same dictionary.
+ * @result : 0, or an error code (which can be tested with ZSTD_isError()).
+ *  Special: Loading a NULL (or 0-size) dictionary invalidates previous dictionary,
+ *           meaning "return to no-dictionary mode".
+ *  Note 1 : Dictionary is sticky, it will be used for all future compressed frames.
+ *           To return to "no-dictionary" situation, load a NULL dictionary (or reset parameters).
+ *  Note 2 : Loading a dictionary involves building tables.
+ *           It's also a CPU consuming operation, with non-negligible impact on latency.
+ *           Tables are dependent on compression parameters, and for this reason,
+ *           compression parameters can no longer be changed after loading a dictionary.
+ *  Note 3 :`dict` content will be copied internally.
+ *           Use experimental ZSTD_CCtx_loadDictionary_byReference() to reference content instead.
+ *           In such a case, dictionary buffer must outlive its users.
+ *  Note 4 : Use ZSTD_CCtx_loadDictionary_advanced()
+ *           to precisely select how dictionary content must be interpreted. */
+ZSTDLIB_API size_t ZSTD_CCtx_loadDictionary(ZSTD_CCtx* cctx, const void* dict, size_t dictSize);
+
+/*! ZSTD_CCtx_refCDict() :
+ *  Reference a prepared dictionary, to be used for all next compressed frames.
+ *  Note that compression parameters are enforced from within CDict,
+ *  and supercede any compression parameter previously set within CCtx.
+ *  The dictionary will remain valid for future compressed frames using same CCtx.
+ * @result : 0, or an error code (which can be tested with ZSTD_isError()).
+ *  Special : Referencing a NULL CDict means "return to no-dictionary mode".
+ *  Note 1 : Currently, only one dictionary can be managed.
+ *           Referencing a new dictionary effectively "discards" any previous one.
+ *  Note 2 : CDict is just referenced, its lifetime must outlive its usage within CCtx. */
+ZSTDLIB_API size_t ZSTD_CCtx_refCDict(ZSTD_CCtx* cctx, const ZSTD_CDict* cdict);
+
+/*! ZSTD_CCtx_refPrefix() :
+ *  Reference a prefix (single-usage dictionary) for next compressed frame.
+ *  A prefix is **only used once**. Tables are discarded at end of frame (ZSTD_e_end).
+ *  Decompression will need same prefix to properly regenerate data.
+ *  Compressing with a prefix is similar in outcome as performing a diff and compressing it,
+ *  but performs much faster, especially during decompression (compression speed is tunable with compression level).
+ * @result : 0, or an error code (which can be tested with ZSTD_isError()).
+ *  Special: Adding any prefix (including NULL) invalidates any previous prefix or dictionary
+ *  Note 1 : Prefix buffer is referenced. It **must** outlive compression.
+ *           Its content must remain unmodified during compression.
+ *  Note 2 : If the intention is to diff some large src data blob with some prior version of itself,
+ *           ensure that the window size is large enough to contain the entire source.
+ *           See ZSTD_c_windowLog.
+ *  Note 3 : Referencing a prefix involves building tables, which are dependent on compression parameters.
+ *           It's a CPU consuming operation, with non-negligible impact on latency.
+ *           If there is a need to use the same prefix multiple times, consider loadDictionary instead.
+ *  Note 4 : By default, the prefix is interpreted as raw content (ZSTD_dm_rawContent).
+ *           Use experimental ZSTD_CCtx_refPrefix_advanced() to alter dictionary interpretation. */
+ZSTDLIB_API size_t ZSTD_CCtx_refPrefix(ZSTD_CCtx* cctx,
+                                 const void* prefix, size_t prefixSize);
+
+
+typedef enum {
+    ZSTD_reset_session_only = 1,
+    ZSTD_reset_parameters = 2,
+    ZSTD_reset_session_and_parameters = 3
+} ZSTD_ResetDirective;
+
+/*! ZSTD_CCtx_reset() :
+ *  There are 2 different things that can be reset, independently or jointly :
+ *  - The session : will stop compressing current frame, and make CCtx ready to start a new one.
+ *                  Useful after an error, or to interrupt any ongoing compression.
+ *                  Any internal data not yet flushed is cancelled.
+ *                  Compression parameters and dictionary remain unchanged.
+ *                  They will be used to compress next frame.
+ *                  Resetting session never fails.
+ *  - The parameters : changes all parameters back to "default".
+ *                  This removes any reference to any dictionary too.
+ *                  Parameters can only be changed between 2 sessions (i.e. no compression is currently ongoing)
+ *                  otherwise the reset fails, and function returns an error value (which can be tested using ZSTD_isError())
+ *  - Both : similar to resetting the session, followed by resetting parameters.
+ */
+ZSTDLIB_API size_t ZSTD_CCtx_reset(ZSTD_CCtx* cctx, ZSTD_ResetDirective reset);
+
+
+
+/*! ZSTD_compress2() :
+ *  Behave the same as ZSTD_compressCCtx(), but compression parameters are set using the advanced API.
+ *  ZSTD_compress2() always starts a new frame.
+ *  Should cctx hold data from a previously unfinished frame, everything about it is forgotten.
+ *  - Compression parameters are pushed into CCtx before starting compression, using ZSTD_CCtx_set*()
+ *  - The function is always blocking, returns when compression is completed.
+ *  Hint : compression runs faster if `dstCapacity` >=  `ZSTD_compressBound(srcSize)`.
+ * @return : compressed size written into `dst` (<= `dstCapacity),
+ *           or an error code if it fails (which can be tested using ZSTD_isError()).
+ */
+ZSTDLIB_API size_t ZSTD_compress2( ZSTD_CCtx* cctx,
+                                   void* dst, size_t dstCapacity,
+                             const void* src, size_t srcSize);
+
+typedef enum {
+    ZSTD_e_continue=0, /* collect more data, encoder decides when to output compressed result, for optimal compression ratio */
+    ZSTD_e_flush=1,    /* flush any data provided so far,
+                        * it creates (at least) one new block, that can be decoded immediately on reception;
+                        * frame will continue: any future data can still reference previously compressed data, improving compression. */
+    ZSTD_e_end=2       /* flush any remaining data _and_ close current frame.
+                        * note that frame is only closed after compressed data is fully flushed (return value == 0).
+                        * After that point, any additional data starts a new frame.
+                        * note : each frame is independent (does not reference any content from previous frame). */
+} ZSTD_EndDirective;
+
+/*! ZSTD_compressStream2() :
+ *  Behaves about the same as ZSTD_compressStream, with additional control on end directive.
+ *  - Compression parameters are pushed into CCtx before starting compression, using ZSTD_CCtx_set*()
+ *  - Compression parameters cannot be changed once compression is started (save a list of exceptions in multi-threading mode)
+ *  - outpot->pos must be <= dstCapacity, input->pos must be <= srcSize
+ *  - outpot->pos and input->pos will be updated. They are guaranteed to remain below their respective limit.
+ *  - When nbWorkers==0 (default), function is blocking : it completes its job before returning to caller.
+ *  - When nbWorkers>=1, function is non-blocking : it just acquires a copy of input, and distributes jobs to internal worker threads, flush whatever is available,
+ *                                                  and then immediately returns, just indicating that there is some data remaining to be flushed.
+ *                                                  The function nonetheless guarantees forward progress : it will return only after it reads or write at least 1+ byte.
+ *  - Exception : if the first call requests a ZSTD_e_end directive and provides enough dstCapacity, the function delegates to ZSTD_compress2() which is always blocking.
+ *  - @return provides a minimum amount of data remaining to be flushed from internal buffers
+ *            or an error code, which can be tested using ZSTD_isError().
+ *            if @return != 0, flush is not fully completed, there is still some data left within internal buffers.
+ *            This is useful for ZSTD_e_flush, since in this case more flushes are necessary to empty all buffers.
+ *            For ZSTD_e_end, @return == 0 when internal buffers are fully flushed and frame is completed.
+ *  - after a ZSTD_e_end directive, if internal buffer is not fully flushed (@return != 0),
+ *            only ZSTD_e_end or ZSTD_e_flush operations are allowed.
+ *            Before starting a new compression job, or changing compression parameters,
+ *            it is required to fully flush internal buffers.
+ */
+ZSTDLIB_API size_t ZSTD_compressStream2( ZSTD_CCtx* cctx,
+                                         ZSTD_outBuffer* output,
+                                         ZSTD_inBuffer* input,
+                                         ZSTD_EndDirective endOp);
+
+
+
+/* ============================== */
+/*   Advanced decompression API   */
+/* ============================== */
+
+/* The advanced API pushes parameters one by one into an existing DCtx context.
+ * Parameters are sticky, and remain valid for all following frames
+ * using the same DCtx context.
+ * It's possible to reset parameters to default values using ZSTD_DCtx_reset().
+ * Note : This API is compatible with existing ZSTD_decompressDCtx() and ZSTD_decompressStream().
+ *        Therefore, no new decompression function is necessary.
+ */
+
+
+typedef enum {
+
+    ZSTD_d_windowLogMax=100, /* Select a size limit (in power of 2) beyond which
+                              * the streaming API will refuse to allocate memory buffer
+                              * in order to protect the host from unreasonable memory requirements.
+                              * This parameter is only useful in streaming mode, since no internal buffer is allocated in single-pass mode.
+                              * By default, a decompression context accepts window sizes <= (1 << ZSTD_WINDOWLOG_LIMIT_DEFAULT) */
+
+    /* note : additional experimental parameters are also available
+     * within the experimental section of the API.
+     * At the time of this writing, they include :
+     * ZSTD_c_format
+     * Because they are not stable, it's necessary to define ZSTD_STATIC_LINKING_ONLY to access them.
+     * note : never ever use experimentalParam? names directly
+     */
+     ZSTD_d_experimentalParam1=1000
+
+} ZSTD_dParameter;
+
+
+/*! ZSTD_dParam_getBounds() :
+ *  All parameters must belong to an interval with lower and upper bounds,
+ *  otherwise they will either trigger an error or be automatically clamped.
+ * @return : a structure, ZSTD_bounds, which contains
+ *         - an error status field, which must be tested using ZSTD_isError()
+ *         - both lower and upper bounds, inclusive
+ */
+ZSTDLIB_API ZSTD_bounds ZSTD_dParam_getBounds(ZSTD_dParameter dParam);
+
+/*! ZSTD_DCtx_setParameter() :
+ *  Set one compression parameter, selected by enum ZSTD_dParameter.
+ *  All parameters have valid bounds. Bounds can be queried using ZSTD_dParam_getBounds().
+ *  Providing a value beyond bound will either clamp it, or trigger an error (depending on parameter).
+ *  Setting a parameter is only possible during frame initialization (before starting decompression).
+ * @return : 0, or an error code (which can be tested using ZSTD_isError()).
+ */
+ZSTDLIB_API size_t ZSTD_DCtx_setParameter(ZSTD_DCtx* dctx, ZSTD_dParameter param, int value);
+
+
+/*! ZSTD_DCtx_loadDictionary() :
+ *  Create an internal DDict from dict buffer,
+ *  to be used to decompress next frames.
+ *  The dictionary remains valid for all future frames, until explicitly invalidated.
+ * @result : 0, or an error code (which can be tested with ZSTD_isError()).
+ *  Special : Adding a NULL (or 0-size) dictionary invalidates any previous dictionary,
+ *            meaning "return to no-dictionary mode".
+ *  Note 1 : Loading a dictionary involves building tables,
+ *           which has a non-negligible impact on CPU usage and latency.
+ *           It's recommended to "load once, use many times", to amortize the cost
+ *  Note 2 :`dict` content will be copied internally, so `dict` can be released after loading.
+ *           Use ZSTD_DCtx_loadDictionary_byReference() to reference dictionary content instead.
+ *  Note 3 : Use ZSTD_DCtx_loadDictionary_advanced() to take control of
+ *           how dictionary content is loaded and interpreted.
+ */
+ZSTDLIB_API size_t ZSTD_DCtx_loadDictionary(ZSTD_DCtx* dctx, const void* dict, size_t dictSize);
+
+/*! ZSTD_DCtx_refDDict() :
+ *  Reference a prepared dictionary, to be used to decompress next frames.
+ *  The dictionary remains active for decompression of future frames using same DCtx.
+ * @result : 0, or an error code (which can be tested with ZSTD_isError()).
+ *  Note 1 : Currently, only one dictionary can be managed.
+ *           Referencing a new dictionary effectively "discards" any previous one.
+ *  Special: referencing a NULL DDict means "return to no-dictionary mode".
+ *  Note 2 : DDict is just referenced, its lifetime must outlive its usage from DCtx.
+ */
+ZSTDLIB_API size_t ZSTD_DCtx_refDDict(ZSTD_DCtx* dctx, const ZSTD_DDict* ddict);
+
+/*! ZSTD_DCtx_refPrefix() :
+ *  Reference a prefix (single-usage dictionary) to decompress next frame.
+ *  This is the reverse operation of ZSTD_CCtx_refPrefix(),
+ *  and must use the same prefix as the one used during compression.
+ *  Prefix is **only used once**. Reference is discarded at end of frame.
+ *  End of frame is reached when ZSTD_decompressStream() returns 0.
+ * @result : 0, or an error code (which can be tested with ZSTD_isError()).
+ *  Note 1 : Adding any prefix (including NULL) invalidates any previously set prefix or dictionary
+ *  Note 2 : Prefix buffer is referenced. It **must** outlive decompression.
+ *           Prefix buffer must remain unmodified up to the end of frame,
+ *           reached when ZSTD_decompressStream() returns 0.
+ *  Note 3 : By default, the prefix is treated as raw content (ZSTD_dm_rawContent).
+ *           Use ZSTD_CCtx_refPrefix_advanced() to alter dictMode (Experimental section)
+ *  Note 4 : Referencing a raw content prefix has almost no cpu nor memory cost.
+ *           A full dictionary is more costly, as it requires building tables.
+ */
+ZSTDLIB_API size_t ZSTD_DCtx_refPrefix(ZSTD_DCtx* dctx,
+                                 const void* prefix, size_t prefixSize);
+
+/*! ZSTD_DCtx_reset() :
+ *  Return a DCtx to clean state.
+ *  Session and parameters can be reset jointly or separately.
+ *  Parameters can only be reset when no active frame is being decompressed.
+ * @return : 0, or an error code, which can be tested with ZSTD_isError()
+ */
+ZSTDLIB_API size_t ZSTD_DCtx_reset(ZSTD_DCtx* dctx, ZSTD_ResetDirective reset);
+
+
+
+/****************************************************************************************
+ *   experimental API (static linking only)
+ ****************************************************************************************
+ * The following symbols and constants
+ * are not planned to join "stable API" status in the near future.
+ * They can still change in future versions.
+ * Some of them are planned to remain in the static_only section indefinitely.
+ * Some of them might be removed in the future (especially when redundant with existing stable functions)
+ * ***************************************************************************************/
+
+#define ZSTD_FRAMEHEADERSIZE_PREFIX 5   /* minimum input size required to query frame header size */
 #define ZSTD_FRAMEHEADERSIZE_MIN    6
-#define ZSTD_FRAMEHEADERSIZE_MAX   18   /* for static allocation */
-static const size_t ZSTD_frameHeaderSize_prefix = ZSTD_FRAMEHEADERSIZE_PREFIX;
-static const size_t ZSTD_frameHeaderSize_min = ZSTD_FRAMEHEADERSIZE_MIN;
-static const size_t ZSTD_frameHeaderSize_max = ZSTD_FRAMEHEADERSIZE_MAX;
-static const size_t ZSTD_skippableHeaderSize = 8;  /* magic number + skippable frame length */
+#define ZSTD_FRAMEHEADERSIZE_MAX   18   /* can be useful for static allocation */
+#define ZSTD_SKIPPABLEHEADERSIZE    8
 
+/* compression parameter bounds */
+#define ZSTD_WINDOWLOG_MAX_32    30
+#define ZSTD_WINDOWLOG_MAX_64    31
+#define ZSTD_WINDOWLOG_MAX     ((int)(sizeof(size_t) == 4 ? ZSTD_WINDOWLOG_MAX_32 : ZSTD_WINDOWLOG_MAX_64))
+#define ZSTD_WINDOWLOG_MIN       10
+#define ZSTD_HASHLOG_MAX       ((ZSTD_WINDOWLOG_MAX < 30) ? ZSTD_WINDOWLOG_MAX : 30)
+#define ZSTD_HASHLOG_MIN          6
+#define ZSTD_CHAINLOG_MAX_32     29
+#define ZSTD_CHAINLOG_MAX_64     30
+#define ZSTD_CHAINLOG_MAX      ((int)(sizeof(size_t) == 4 ? ZSTD_CHAINLOG_MAX_32 : ZSTD_CHAINLOG_MAX_64))
+#define ZSTD_CHAINLOG_MIN        ZSTD_HASHLOG_MIN
+#define ZSTD_SEARCHLOG_MAX      (ZSTD_WINDOWLOG_MAX-1)
+#define ZSTD_SEARCHLOG_MIN        1
+#define ZSTD_MINMATCH_MAX         7   /* only for ZSTD_fast, other strategies are limited to 6 */
+#define ZSTD_MINMATCH_MIN         3   /* only for ZSTD_btopt+, faster strategies are limited to 4 */
+#define ZSTD_TARGETLENGTH_MAX    ZSTD_BLOCKSIZE_MAX
+#define ZSTD_TARGETLENGTH_MIN     0   /* note : comparing this constant to an unsigned results in a tautological test */
+#define ZSTD_STRATEGY_MIN        ZSTD_fast
+#define ZSTD_STRATEGY_MAX        ZSTD_btultra2
+
+
+#define ZSTD_OVERLAPLOG_MIN       0
+#define ZSTD_OVERLAPLOG_MAX       9
+
+#define ZSTD_WINDOWLOG_LIMIT_DEFAULT 27   /* by default, the streaming decoder will refuse any frame
+                                           * requiring larger than (1<= first frame size
- *  @return : the compressed size of the first frame starting at `src`,
- *            suitable to pass to `ZSTD_decompress` or similar,
- *            or an error code if input is invalid */
-ZSTDLIB_API size_t ZSTD_findFrameCompressedSize(const void* src, size_t srcSize);
-
 /*! ZSTD_findDecompressedSize() :
  *  `src` should point the start of a series of ZSTD encoded and/or skippable frames
  *  `srcSize` must be the _exact_ size of this series
@@ -521,7 +1093,7 @@ ZSTDLIB_API size_t ZSTD_findFrameCompressedSize(const void* src, size_t srcSize)
 ZSTDLIB_API unsigned long long ZSTD_findDecompressedSize(const void* src, size_t srcSize);
 
 /*! ZSTD_frameHeaderSize() :
- *  srcSize must be >= ZSTD_frameHeaderSize_prefix.
+ *  srcSize must be >= ZSTD_FRAMEHEADERSIZE_PREFIX.
  * @return : size of the Frame Header,
  *           or an error code (if srcSize is too small) */
 ZSTDLIB_API size_t ZSTD_frameHeaderSize(const void* src, size_t srcSize);
@@ -531,16 +1103,6 @@ ZSTDLIB_API size_t ZSTD_frameHeaderSize(const void* src, size_t srcSize);
 *  Memory management
 ***************************************/
 
-/*! ZSTD_sizeof_*() :
- *  These functions give the current memory usage of selected object.
- *  Object memory usage can evolve when re-used. */
-ZSTDLIB_API size_t ZSTD_sizeof_CCtx(const ZSTD_CCtx* cctx);
-ZSTDLIB_API size_t ZSTD_sizeof_DCtx(const ZSTD_DCtx* dctx);
-ZSTDLIB_API size_t ZSTD_sizeof_CStream(const ZSTD_CStream* zcs);
-ZSTDLIB_API size_t ZSTD_sizeof_DStream(const ZSTD_DStream* zds);
-ZSTDLIB_API size_t ZSTD_sizeof_CDict(const ZSTD_CDict* cdict);
-ZSTDLIB_API size_t ZSTD_sizeof_DDict(const ZSTD_DDict* ddict);
-
 /*! ZSTD_estimate*() :
  *  These functions make it possible to estimate memory usage
  *  of a future {D,C}Ctx, before its creation.
@@ -548,7 +1110,7 @@ ZSTDLIB_API size_t ZSTD_sizeof_DDict(const ZSTD_DDict* ddict);
  *  It will also consider src size to be arbitrarily "large", which is worst case.
  *  If srcSize is known to always be small, ZSTD_estimateCCtxSize_usingCParams() can provide a tighter estimation.
  *  ZSTD_estimateCCtxSize_usingCParams() can be used in tandem with ZSTD_getCParams() to create cParams from compressionLevel.
- *  ZSTD_estimateCCtxSize_usingCCtxParams() can be used in tandem with ZSTD_CCtxParam_setParameter(). Only single-threaded compression is supported. This function will return an error code if ZSTD_p_nbWorkers is >= 1.
+ *  ZSTD_estimateCCtxSize_usingCCtxParams() can be used in tandem with ZSTD_CCtxParam_setParameter(). Only single-threaded compression is supported. This function will return an error code if ZSTD_c_nbWorkers is >= 1.
  *  Note : CCtx size estimation is only correct for single-threaded compression. */
 ZSTDLIB_API size_t ZSTD_estimateCCtxSize(int compressionLevel);
 ZSTDLIB_API size_t ZSTD_estimateCCtxSize_usingCParams(ZSTD_compressionParameters cParams);
@@ -560,7 +1122,7 @@ ZSTDLIB_API size_t ZSTD_estimateDCtxSize(void);
  *  It will also consider src size to be arbitrarily "large", which is worst case.
  *  If srcSize is known to always be small, ZSTD_estimateCStreamSize_usingCParams() can provide a tighter estimation.
  *  ZSTD_estimateCStreamSize_usingCParams() can be used in tandem with ZSTD_getCParams() to create cParams from compressionLevel.
- *  ZSTD_estimateCStreamSize_usingCCtxParams() can be used in tandem with ZSTD_CCtxParam_setParameter(). Only single-threaded compression is supported. This function will return an error code if ZSTD_p_nbWorkers is >= 1.
+ *  ZSTD_estimateCStreamSize_usingCCtxParams() can be used in tandem with ZSTD_CCtxParam_setParameter(). Only single-threaded compression is supported. This function will return an error code if ZSTD_c_nbWorkers is >= 1.
  *  Note : CStream size estimation is only correct for single-threaded compression.
  *  ZSTD_DStream memory budget depends on window Size.
  *  This information can be passed manually, using ZSTD_estimateDStreamSize,
@@ -623,6 +1185,7 @@ ZSTDLIB_API const ZSTD_DDict* ZSTD_initStaticDDict(
                                         ZSTD_dictLoadMethod_e dictLoadMethod,
                                         ZSTD_dictContentType_e dictContentType);
 
+
 /*! Custom memory allocation :
  *  These prototypes make it possible to pass your own allocation/free functions.
  *  ZSTD_customMem is provided at creation time, using ZSTD_create*_advanced() variants listed below.
@@ -657,8 +1220,9 @@ ZSTDLIB_API ZSTD_DDict* ZSTD_createDDict_advanced(const void* dict, size_t dictS
 
 /*! ZSTD_createCDict_byReference() :
  *  Create a digested dictionary for compression
- *  Dictionary content is simply referenced, and therefore stays in dictBuffer.
- *  It is important that dictBuffer outlives CDict, it must remain read accessible throughout the lifetime of CDict */
+ *  Dictionary content is just referenced, not duplicated.
+ *  As a consequence, `dictBuffer` **must** outlive CDict,
+ *  and its content must remain unmodified throughout the lifetime of CDict. */
 ZSTDLIB_API ZSTD_CDict* ZSTD_createCDict_byReference(const void* dictBuffer, size_t dictSize, int compressionLevel);
 
 /*! ZSTD_getCParams() :
@@ -681,22 +1245,161 @@ ZSTDLIB_API size_t ZSTD_checkCParams(ZSTD_compressionParameters params);
 ZSTDLIB_API ZSTD_compressionParameters ZSTD_adjustCParams(ZSTD_compressionParameters cPar, unsigned long long srcSize, size_t dictSize);
 
 /*! ZSTD_compress_advanced() :
-*   Same as ZSTD_compress_usingDict(), with fine-tune control over each compression parameter */
-ZSTDLIB_API size_t ZSTD_compress_advanced (ZSTD_CCtx* cctx,
-                                  void* dst, size_t dstCapacity,
-                            const void* src, size_t srcSize,
-                            const void* dict,size_t dictSize,
-                                  ZSTD_parameters params);
+ *  Same as ZSTD_compress_usingDict(), with fine-tune control over compression parameters (by structure) */
+ZSTDLIB_API size_t ZSTD_compress_advanced(ZSTD_CCtx* cctx,
+                                          void* dst, size_t dstCapacity,
+                                    const void* src, size_t srcSize,
+                                    const void* dict,size_t dictSize,
+                                          ZSTD_parameters params);
 
 /*! ZSTD_compress_usingCDict_advanced() :
-*   Same as ZSTD_compress_usingCDict(), with fine-tune control over frame parameters */
+ *  Same as ZSTD_compress_usingCDict(), with fine-tune control over frame parameters */
 ZSTDLIB_API size_t ZSTD_compress_usingCDict_advanced(ZSTD_CCtx* cctx,
-                                  void* dst, size_t dstCapacity,
-                            const void* src, size_t srcSize,
-                            const ZSTD_CDict* cdict, ZSTD_frameParameters fParams);
+                                              void* dst, size_t dstCapacity,
+                                        const void* src, size_t srcSize,
+                                        const ZSTD_CDict* cdict,
+                                              ZSTD_frameParameters fParams);
 
 
-/*--- Advanced decompression functions ---*/
+/*! ZSTD_CCtx_loadDictionary_byReference() :
+ *  Same as ZSTD_CCtx_loadDictionary(), but dictionary content is referenced, instead of being copied into CCtx.
+ *  It saves some memory, but also requires that `dict` outlives its usage within `cctx` */
+ZSTDLIB_API size_t ZSTD_CCtx_loadDictionary_byReference(ZSTD_CCtx* cctx, const void* dict, size_t dictSize);
+
+/*! ZSTD_CCtx_loadDictionary_advanced() :
+ *  Same as ZSTD_CCtx_loadDictionary(), but gives finer control over
+ *  how to load the dictionary (by copy ? by reference ?)
+ *  and how to interpret it (automatic ? force raw mode ? full mode only ?) */
+ZSTDLIB_API size_t ZSTD_CCtx_loadDictionary_advanced(ZSTD_CCtx* cctx, const void* dict, size_t dictSize, ZSTD_dictLoadMethod_e dictLoadMethod, ZSTD_dictContentType_e dictContentType);
+
+/*! ZSTD_CCtx_refPrefix_advanced() :
+ *  Same as ZSTD_CCtx_refPrefix(), but gives finer control over
+ *  how to interpret prefix content (automatic ? force raw mode (default) ? full mode only ?) */
+ZSTDLIB_API size_t ZSTD_CCtx_refPrefix_advanced(ZSTD_CCtx* cctx, const void* prefix, size_t prefixSize, ZSTD_dictContentType_e dictContentType);
+
+/* ===   experimental parameters   === */
+/* these parameters can be used with ZSTD_setParameter()
+ * they are not guaranteed to remain supported in the future */
+
+ /* Enables rsyncable mode,
+  * which makes compressed files more rsync friendly
+  * by adding periodic synchronization points to the compressed data.
+  * The target average block size is ZSTD_c_jobSize / 2.
+  * It's possible to modify the job size to increase or decrease
+  * the granularity of the synchronization point.
+  * Once the jobSize is smaller than the window size,
+  * it will result in compression ratio degradation.
+  * NOTE 1: rsyncable mode only works when multithreading is enabled.
+  * NOTE 2: rsyncable performs poorly in combination with long range mode,
+  * since it will decrease the effectiveness of synchronization points,
+  * though mileage may vary.
+  * NOTE 3: Rsyncable mode limits maximum compression speed to ~400 MB/s.
+  * If the selected compression level is already running significantly slower,
+  * the overall speed won't be significantly impacted.
+  */
+ #define ZSTD_c_rsyncable ZSTD_c_experimentalParam1
+
+/* Select a compression format.
+ * The value must be of type ZSTD_format_e.
+ * See ZSTD_format_e enum definition for details */
+#define ZSTD_c_format ZSTD_c_experimentalParam2
+
+/* Force back-reference distances to remain < windowSize,
+ * even when referencing into Dictionary content (default:0) */
+#define ZSTD_c_forceMaxWindow ZSTD_c_experimentalParam3
+
+/* Controls whether the contents of a CDict
+ * are used in place, or copied into the working context.
+ * Accepts values from the ZSTD_dictAttachPref_e enum.
+ * See the comments on that enum for an explanation of the feature. */
+#define ZSTD_c_forceAttachDict ZSTD_c_experimentalParam4
+
+/*! ZSTD_CCtx_getParameter() :
+ *  Get the requested compression parameter value, selected by enum ZSTD_cParameter,
+ *  and store it into int* value.
+ * @return : 0, or an error code (which can be tested with ZSTD_isError()).
+ */
+ZSTDLIB_API size_t ZSTD_CCtx_getParameter(ZSTD_CCtx* cctx, ZSTD_cParameter param, int* value);
+
+
+/*! ZSTD_CCtx_params :
+ *  Quick howto :
+ *  - ZSTD_createCCtxParams() : Create a ZSTD_CCtx_params structure
+ *  - ZSTD_CCtxParam_setParameter() : Push parameters one by one into
+ *                                    an existing ZSTD_CCtx_params structure.
+ *                                    This is similar to
+ *                                    ZSTD_CCtx_setParameter().
+ *  - ZSTD_CCtx_setParametersUsingCCtxParams() : Apply parameters to
+ *                                    an existing CCtx.
+ *                                    These parameters will be applied to
+ *                                    all subsequent frames.
+ *  - ZSTD_compressStream2() : Do compression using the CCtx.
+ *  - ZSTD_freeCCtxParams() : Free the memory.
+ *
+ *  This can be used with ZSTD_estimateCCtxSize_advanced_usingCCtxParams()
+ *  for static allocation of CCtx for single-threaded compression.
+ */
+ZSTDLIB_API ZSTD_CCtx_params* ZSTD_createCCtxParams(void);
+ZSTDLIB_API size_t ZSTD_freeCCtxParams(ZSTD_CCtx_params* params);
+
+/*! ZSTD_CCtxParams_reset() :
+ *  Reset params to default values.
+ */
+ZSTDLIB_API size_t ZSTD_CCtxParams_reset(ZSTD_CCtx_params* params);
+
+/*! ZSTD_CCtxParams_init() :
+ *  Initializes the compression parameters of cctxParams according to
+ *  compression level. All other parameters are reset to their default values.
+ */
+ZSTDLIB_API size_t ZSTD_CCtxParams_init(ZSTD_CCtx_params* cctxParams, int compressionLevel);
+
+/*! ZSTD_CCtxParams_init_advanced() :
+ *  Initializes the compression and frame parameters of cctxParams according to
+ *  params. All other parameters are reset to their default values.
+ */
+ZSTDLIB_API size_t ZSTD_CCtxParams_init_advanced(ZSTD_CCtx_params* cctxParams, ZSTD_parameters params);
+
+/*! ZSTD_CCtxParam_setParameter() :
+ *  Similar to ZSTD_CCtx_setParameter.
+ *  Set one compression parameter, selected by enum ZSTD_cParameter.
+ *  Parameters must be applied to a ZSTD_CCtx using ZSTD_CCtx_setParametersUsingCCtxParams().
+ * @result : 0, or an error code (which can be tested with ZSTD_isError()).
+ */
+ZSTDLIB_API size_t ZSTD_CCtxParam_setParameter(ZSTD_CCtx_params* params, ZSTD_cParameter param, int value);
+
+/*! ZSTD_CCtxParam_getParameter() :
+ * Similar to ZSTD_CCtx_getParameter.
+ * Get the requested value of one compression parameter, selected by enum ZSTD_cParameter.
+ * @result : 0, or an error code (which can be tested with ZSTD_isError()).
+ */
+ZSTDLIB_API size_t ZSTD_CCtxParam_getParameter(ZSTD_CCtx_params* params, ZSTD_cParameter param, int* value);
+
+/*! ZSTD_CCtx_setParametersUsingCCtxParams() :
+ *  Apply a set of ZSTD_CCtx_params to the compression context.
+ *  This can be done even after compression is started,
+ *    if nbWorkers==0, this will have no impact until a new compression is started.
+ *    if nbWorkers>=1, new parameters will be picked up at next job,
+ *       with a few restrictions (windowLog, pledgedSrcSize, nbWorkers, jobSize, and overlapLog are not updated).
+ */
+ZSTDLIB_API size_t ZSTD_CCtx_setParametersUsingCCtxParams(
+        ZSTD_CCtx* cctx, const ZSTD_CCtx_params* params);
+
+/*! ZSTD_compressStream2_simpleArgs() :
+ *  Same as ZSTD_compressStream2(),
+ *  but using only integral types as arguments.
+ *  This variant might be helpful for binders from dynamic languages
+ *  which have troubles handling structures containing memory pointers.
+ */
+ZSTDLIB_API size_t ZSTD_compressStream2_simpleArgs (
+                            ZSTD_CCtx* cctx,
+                            void* dst, size_t dstCapacity, size_t* dstPos,
+                      const void* src, size_t srcSize, size_t* srcPos,
+                            ZSTD_EndDirective endOp);
+
+
+/***************************************
+*  Advanced decompression functions
+***************************************/
 
 /*! ZSTD_isFrame() :
  *  Tells if the content of `buffer` starts with a valid Frame Identifier.
@@ -737,9 +1440,64 @@ ZSTDLIB_API unsigned ZSTD_getDictID_fromDDict(const ZSTD_DDict* ddict);
  *  When identifying the exact failure cause, it's possible to use ZSTD_getFrameHeader(), which will provide a more precise error code. */
 ZSTDLIB_API unsigned ZSTD_getDictID_fromFrame(const void* src, size_t srcSize);
 
+/*! ZSTD_DCtx_loadDictionary_byReference() :
+ *  Same as ZSTD_DCtx_loadDictionary(),
+ *  but references `dict` content instead of copying it into `dctx`.
+ *  This saves memory if `dict` remains around.,
+ *  However, it's imperative that `dict` remains accessible (and unmodified) while being used, so it must outlive decompression. */
+ZSTDLIB_API size_t ZSTD_DCtx_loadDictionary_byReference(ZSTD_DCtx* dctx, const void* dict, size_t dictSize);
+
+/*! ZSTD_DCtx_loadDictionary_advanced() :
+ *  Same as ZSTD_DCtx_loadDictionary(),
+ *  but gives direct control over
+ *  how to load the dictionary (by copy ? by reference ?)
+ *  and how to interpret it (automatic ? force raw mode ? full mode only ?). */
+ZSTDLIB_API size_t ZSTD_DCtx_loadDictionary_advanced(ZSTD_DCtx* dctx, const void* dict, size_t dictSize, ZSTD_dictLoadMethod_e dictLoadMethod, ZSTD_dictContentType_e dictContentType);
+
+/*! ZSTD_DCtx_refPrefix_advanced() :
+ *  Same as ZSTD_DCtx_refPrefix(), but gives finer control over
+ *  how to interpret prefix content (automatic ? force raw mode (default) ? full mode only ?) */
+ZSTDLIB_API size_t ZSTD_DCtx_refPrefix_advanced(ZSTD_DCtx* dctx, const void* prefix, size_t prefixSize, ZSTD_dictContentType_e dictContentType);
+
+/*! ZSTD_DCtx_setMaxWindowSize() :
+ *  Refuses allocating internal buffers for frames requiring a window size larger than provided limit.
+ *  This protects a decoder context from reserving too much memory for itself (potential attack scenario).
+ *  This parameter is only useful in streaming mode, since no internal buffer is allocated in single-pass mode.
+ *  By default, a decompression context accepts all window sizes <= (1 << ZSTD_WINDOWLOG_LIMIT_DEFAULT)
+ * @return : 0, or an error code (which can be tested using ZSTD_isError()).
+ */
+ZSTDLIB_API size_t ZSTD_DCtx_setMaxWindowSize(ZSTD_DCtx* dctx, size_t maxWindowSize);
+
+/* ZSTD_d_format
+ * experimental parameter,
+ * allowing selection between ZSTD_format_e input compression formats
+ */
+#define ZSTD_d_format ZSTD_d_experimentalParam1
+
+/*! ZSTD_DCtx_setFormat() :
+ *  Instruct the decoder context about what kind of data to decode next.
+ *  This instruction is mandatory to decode data without a fully-formed header,
+ *  such ZSTD_f_zstd1_magicless for example.
+ * @return : 0, or an error code (which can be tested using ZSTD_isError()). */
+ZSTDLIB_API size_t ZSTD_DCtx_setFormat(ZSTD_DCtx* dctx, ZSTD_format_e format);
+
+/*! ZSTD_decompressStream_simpleArgs() :
+ *  Same as ZSTD_decompressStream(),
+ *  but using only integral types as arguments.
+ *  This can be helpful for binders from dynamic languages
+ *  which have troubles handling structures containing memory pointers.
+ */
+ZSTDLIB_API size_t ZSTD_decompressStream_simpleArgs (
+                            ZSTD_DCtx* dctx,
+                            void* dst, size_t dstCapacity, size_t* dstPos,
+                      const void* src, size_t srcSize, size_t* srcPos);
+
 
 /********************************************************************
 *  Advanced streaming functions
+*  Warning : most of these functions are now redundant with the Advanced API.
+*  Once Advanced API reaches "stable" status,
+*  redundant functions will be deprecated, and then at some point removed.
 ********************************************************************/
 
 /*=====   Advanced Streaming compression functions  =====*/
@@ -751,7 +1509,7 @@ ZSTDLIB_API size_t ZSTD_initCStream_usingCDict(ZSTD_CStream* zcs, const ZSTD_CDi
 ZSTDLIB_API size_t ZSTD_initCStream_usingCDict_advanced(ZSTD_CStream* zcs, const ZSTD_CDict* cdict, ZSTD_frameParameters fParams, unsigned long long pledgedSrcSize);  /**< same as ZSTD_initCStream_usingCDict(), with control over frame parameters. pledgedSrcSize must be correct. If srcSize is not known at init time, use value ZSTD_CONTENTSIZE_UNKNOWN. */
 
 /*! ZSTD_resetCStream() :
- *  start a new compression job, using same parameters from previous job.
+ *  start a new frame, using same parameters from previous frame.
  *  This is typically useful to skip dictionary loading stage, since it will re-use it in-place.
  *  Note that zcs must be init at least once before using ZSTD_resetCStream().
  *  If pledgedSrcSize is not known at reset time, use macro ZSTD_CONTENTSIZE_UNKNOWN.
@@ -790,16 +1548,13 @@ ZSTDLIB_API ZSTD_frameProgression ZSTD_getFrameProgression(const ZSTD_CCtx* cctx
  *  + there is no active job (could be checked with ZSTD_frameProgression()), or
  *  + oldest job is still actively compressing data,
  *    but everything it has produced has also been flushed so far,
- *    therefore flushing speed is currently limited by production speed of oldest job
- *    irrespective of the speed of concurrent newer jobs.
+ *    therefore flush speed is limited by production speed of oldest job
+ *    irrespective of the speed of concurrent (and newer) jobs.
  */
 ZSTDLIB_API size_t ZSTD_toFlushNow(ZSTD_CCtx* cctx);
 
 
-
 /*=====   Advanced Streaming decompression functions  =====*/
-typedef enum { DStream_p_maxWindowSize } ZSTD_DStreamParameter_e;
-ZSTDLIB_API size_t ZSTD_setDStreamParameter(ZSTD_DStream* zds, ZSTD_DStreamParameter_e paramType, unsigned paramValue);   /* obsolete : this API will be removed in a future version */
 ZSTDLIB_API size_t ZSTD_initDStream_usingDict(ZSTD_DStream* zds, const void* dict, size_t dictSize); /**< note: no dictionary will be used if dict == NULL or dictSize < 8 */
 ZSTDLIB_API size_t ZSTD_initDStream_usingDDict(ZSTD_DStream* zds, const ZSTD_DDict* ddict);  /**< note : ddict is referenced, it must outlive decompression session */
 ZSTDLIB_API size_t ZSTD_resetDStream(ZSTD_DStream* zds);  /**< re-use decompression parameters from previous init; saves dictionary loading */
@@ -940,12 +1695,17 @@ typedef struct {
     unsigned dictID;
     unsigned checksumFlag;
 } ZSTD_frameHeader;
+
 /** ZSTD_getFrameHeader() :
  *  decode Frame Header, or requires larger `srcSize`.
  * @return : 0, `zfhPtr` is correctly filled,
  *          >0, `srcSize` is too small, value is wanted `srcSize` amount,
  *           or an error code, which can be tested using ZSTD_isError() */
 ZSTDLIB_API size_t ZSTD_getFrameHeader(ZSTD_frameHeader* zfhPtr, const void* src, size_t srcSize);   /**< doesn't consume input */
+/*! ZSTD_getFrameHeader_advanced() :
+ *  same as ZSTD_getFrameHeader(),
+ *  with added capability to select a format (like ZSTD_f_zstd1_magicless) */
+ZSTDLIB_API size_t ZSTD_getFrameHeader_advanced(ZSTD_frameHeader* zfhPtr, const void* src, size_t srcSize, ZSTD_format_e format);
 ZSTDLIB_API size_t ZSTD_decodingBufferSize_min(unsigned long long windowSize, unsigned long long frameContentSize);  /**< when frame content size is not known, pass in frameContentSize == ZSTD_CONTENTSIZE_UNKNOWN */
 
 ZSTDLIB_API size_t ZSTD_decompressBegin(ZSTD_DCtx* dctx);
@@ -962,526 +1722,6 @@ ZSTDLIB_API ZSTD_nextInputType_e ZSTD_nextInputType(ZSTD_DCtx* dctx);
 
 
 
-/* ============================================ */
-/**       New advanced API (experimental)       */
-/* ============================================ */
-
-/* API design :
- *   In this advanced API, parameters are pushed one by one into an existing context,
- *   using ZSTD_CCtx_set*() functions.
- *   Pushed parameters are sticky : they are applied to next job, and any subsequent job.
- *   It's possible to reset parameters to "default" using ZSTD_CCtx_reset().
- *   Important : "sticky" parameters only work with `ZSTD_compress_generic()` !
- *               For any other entry point, "sticky" parameters are ignored !
- *
- *   This API is intended to replace all others advanced / experimental API entry points.
- */
-
-/* note on enum design :
- * All enum will be pinned to explicit values before reaching "stable API" status */
-
-typedef enum {
-    /* Opened question : should we have a format ZSTD_f_auto ?
-     * Today, it would mean exactly the same as ZSTD_f_zstd1.
-     * But, in the future, should several formats become supported,
-     * on the compression side, it would mean "default format".
-     * On the decompression side, it would mean "automatic format detection",
-     * so that ZSTD_f_zstd1 would mean "accept *only* zstd frames".
-     * Since meaning is a little different, another option could be to define different enums for compression and decompression.
-     * This question could be kept for later, when there are actually multiple formats to support,
-     * but there is also the question of pinning enum values, and pinning value `0` is especially important */
-    ZSTD_f_zstd1 = 0,        /* zstd frame format, specified in zstd_compression_format.md (default) */
-    ZSTD_f_zstd1_magicless,  /* Variant of zstd frame format, without initial 4-bytes magic number.
-                              * Useful to save 4 bytes per generated frame.
-                              * Decoder cannot recognise automatically this format, requiring instructions. */
-} ZSTD_format_e;
-
-typedef enum {
-    /* compression format */
-    ZSTD_p_format = 10,      /* See ZSTD_format_e enum definition.
-                              * Cast selected format as unsigned for ZSTD_CCtx_setParameter() compatibility. */
-
-    /* compression parameters */
-    ZSTD_p_compressionLevel=100, /* Update all compression parameters according to pre-defined cLevel table
-                              * Default level is ZSTD_CLEVEL_DEFAULT==3.
-                              * Special: value 0 means default, which is controlled by ZSTD_CLEVEL_DEFAULT.
-                              * Note 1 : it's possible to pass a negative compression level by casting it to unsigned type.
-                              * Note 2 : setting a level sets all default values of other compression parameters.
-                              * Note 3 : setting compressionLevel automatically updates ZSTD_p_compressLiterals. */
-    ZSTD_p_windowLog,        /* Maximum allowed back-reference distance, expressed as power of 2.
-                              * Must be clamped between ZSTD_WINDOWLOG_MIN and ZSTD_WINDOWLOG_MAX.
-                              * Special: value 0 means "use default windowLog".
-                              * Note: Using a window size greater than ZSTD_MAXWINDOWSIZE_DEFAULT (default: 2^27)
-                              *       requires explicitly allowing such window size during decompression stage. */
-    ZSTD_p_hashLog,          /* Size of the initial probe table, as a power of 2.
-                              * Resulting table size is (1 << (hashLog+2)).
-                              * Must be clamped between ZSTD_HASHLOG_MIN and ZSTD_HASHLOG_MAX.
-                              * Larger tables improve compression ratio of strategies <= dFast,
-                              * and improve speed of strategies > dFast.
-                              * Special: value 0 means "use default hashLog". */
-    ZSTD_p_chainLog,         /* Size of the multi-probe search table, as a power of 2.
-                              * Resulting table size is (1 << (chainLog+2)).
-                              * Must be clamped between ZSTD_CHAINLOG_MIN and ZSTD_CHAINLOG_MAX.
-                              * Larger tables result in better and slower compression.
-                              * This parameter is useless when using "fast" strategy.
-                              * Note it's still useful when using "dfast" strategy,
-                              * in which case it defines a secondary probe table.
-                              * Special: value 0 means "use default chainLog". */
-    ZSTD_p_searchLog,        /* Number of search attempts, as a power of 2.
-                              * More attempts result in better and slower compression.
-                              * This parameter is useless when using "fast" and "dFast" strategies.
-                              * Special: value 0 means "use default searchLog". */
-    ZSTD_p_minMatch,         /* Minimum size of searched matches (note : repCode matches can be smaller).
-                              * Larger values make faster compression and decompression, but decrease ratio.
-                              * Must be clamped between ZSTD_SEARCHLENGTH_MIN and ZSTD_SEARCHLENGTH_MAX.
-                              * Note that currently, for all strategies < btopt, effective minimum is 4.
-                              *                    , for all strategies > fast, effective maximum is 6.
-                              * Special: value 0 means "use default minMatchLength". */
-    ZSTD_p_targetLength,     /* Impact of this field depends on strategy.
-                              * For strategies btopt & btultra:
-                              *     Length of Match considered "good enough" to stop search.
-                              *     Larger values make compression stronger, and slower.
-                              * For strategy fast:
-                              *     Distance between match sampling.
-                              *     Larger values make compression faster, and weaker.
-                              * Special: value 0 means "use default targetLength". */
-    ZSTD_p_compressionStrategy, /* See ZSTD_strategy enum definition.
-                              * Cast selected strategy as unsigned for ZSTD_CCtx_setParameter() compatibility.
-                              * The higher the value of selected strategy, the more complex it is,
-                              * resulting in stronger and slower compression.
-                              * Special: value 0 means "use default strategy". */
-
-    ZSTD_p_enableLongDistanceMatching=160, /* Enable long distance matching.
-                                         * This parameter is designed to improve compression ratio
-                                         * for large inputs, by finding large matches at long distance.
-                                         * It increases memory usage and window size.
-                                         * Note: enabling this parameter increases ZSTD_p_windowLog to 128 MB
-                                         * except when expressly set to a different value. */
-    ZSTD_p_ldmHashLog,       /* Size of the table for long distance matching, as a power of 2.
-                              * Larger values increase memory usage and compression ratio,
-                              * but decrease compression speed.
-                              * Must be clamped between ZSTD_HASHLOG_MIN and ZSTD_HASHLOG_MAX
-                              * default: windowlog - 7.
-                              * Special: value 0 means "automatically determine hashlog". */
-    ZSTD_p_ldmMinMatch,      /* Minimum match size for long distance matcher.
-                              * Larger/too small values usually decrease compression ratio.
-                              * Must be clamped between ZSTD_LDM_MINMATCH_MIN and ZSTD_LDM_MINMATCH_MAX.
-                              * Special: value 0 means "use default value" (default: 64). */
-    ZSTD_p_ldmBucketSizeLog, /* Log size of each bucket in the LDM hash table for collision resolution.
-                              * Larger values improve collision resolution but decrease compression speed.
-                              * The maximum value is ZSTD_LDM_BUCKETSIZELOG_MAX .
-                              * Special: value 0 means "use default value" (default: 3). */
-    ZSTD_p_ldmHashEveryLog,  /* Frequency of inserting/looking up entries in the LDM hash table.
-                              * Must be clamped between 0 and (ZSTD_WINDOWLOG_MAX - ZSTD_HASHLOG_MIN).
-                              * Default is MAX(0, (windowLog - ldmHashLog)), optimizing hash table usage.
-                              * Larger values improve compression speed.
-                              * Deviating far from default value will likely result in a compression ratio decrease.
-                              * Special: value 0 means "automatically determine hashEveryLog". */
-
-    /* frame parameters */
-    ZSTD_p_contentSizeFlag=200, /* Content size will be written into frame header _whenever known_ (default:1)
-                              * Content size must be known at the beginning of compression,
-                              * it is provided using ZSTD_CCtx_setPledgedSrcSize() */
-    ZSTD_p_checksumFlag,     /* A 32-bits checksum of content is written at end of frame (default:0) */
-    ZSTD_p_dictIDFlag,       /* When applicable, dictionary's ID is written into frame header (default:1) */
-
-    /* multi-threading parameters */
-    /* These parameters are only useful if multi-threading is enabled (ZSTD_MULTITHREAD).
-     * They return an error otherwise. */
-    ZSTD_p_nbWorkers=400,    /* Select how many threads will be spawned to compress in parallel.
-                              * When nbWorkers >= 1, triggers asynchronous mode :
-                              * ZSTD_compress_generic() consumes some input, flush some output if possible, and immediately gives back control to caller,
-                              * while compression work is performed in parallel, within worker threads.
-                              * (note : a strong exception to this rule is when first invocation sets ZSTD_e_end : it becomes a blocking call).
-                              * More workers improve speed, but also increase memory usage.
-                              * Default value is `0`, aka "single-threaded mode" : no worker is spawned, compression is performed inside Caller's thread, all invocations are blocking */
-    ZSTD_p_jobSize,          /* Size of a compression job. This value is enforced only in non-blocking mode.
-                              * Each compression job is completed in parallel, so this value indirectly controls the nb of active threads.
-                              * 0 means default, which is dynamically determined based on compression parameters.
-                              * Job size must be a minimum of overlapSize, or 1 MB, whichever is largest.
-                              * The minimum size is automatically and transparently enforced */
-    ZSTD_p_overlapSizeLog,   /* Size of previous input reloaded at the beginning of each job.
-                              * 0 => no overlap, 6(default) => use 1/8th of windowSize, >=9 => use full windowSize */
-
-    /* =================================================================== */
-    /* experimental parameters - no stability guaranteed                   */
-    /* =================================================================== */
-
-    ZSTD_p_forceMaxWindow=1100, /* Force back-reference distances to remain < windowSize,
-                              * even when referencing into Dictionary content (default:0) */
-    ZSTD_p_forceAttachDict,  /* ZSTD supports usage of a CDict in-place
-                              * (avoiding having to copy the compression tables
-                              * from the CDict into the working context). Using
-                              * a CDict in this way saves an initial setup step,
-                              * but comes at the cost of more work per byte of
-                              * input. ZSTD has a simple internal heuristic that
-                              * guesses which strategy will be faster. You can
-                              * use this flag to override that guess.
-                              *
-                              * Note that the by-reference, in-place strategy is
-                              * only used when reusing a compression context
-                              * with compatible compression parameters. (If
-                              * incompatible / uninitialized, the working
-                              * context needs to be cleared anyways, which is
-                              * about as expensive as overwriting it with the
-                              * dictionary context, so there's no savings in
-                              * using the CDict by-ref.)
-                              *
-                              * Values greater than 0 force attaching the dict.
-                              * Values less than 0 force copying the dict.
-                              * 0 selects the default heuristic-guided behavior.
-                              */
-
-} ZSTD_cParameter;
-
-
-/*! ZSTD_CCtx_setParameter() :
- *  Set one compression parameter, selected by enum ZSTD_cParameter.
- *  Setting a parameter is generally only possible during frame initialization (before starting compression).
- *  Exception : when using multi-threading mode (nbThreads >= 1),
- *              following parameters can be updated _during_ compression (within same frame):
- *              => compressionLevel, hashLog, chainLog, searchLog, minMatch, targetLength and strategy.
- *              new parameters will be active on next job, or after a flush().
- *  Note : when `value` type is not unsigned (int, or enum), cast it to unsigned for proper type checking.
- *  @result : informational value (typically, value being set, correctly clamped),
- *            or an error code (which can be tested with ZSTD_isError()). */
-ZSTDLIB_API size_t ZSTD_CCtx_setParameter(ZSTD_CCtx* cctx, ZSTD_cParameter param, unsigned value);
-
-/*! ZSTD_CCtx_getParameter() :
- * Get the requested value of one compression parameter, selected by enum ZSTD_cParameter.
- * @result : 0, or an error code (which can be tested with ZSTD_isError()).
- */
-ZSTDLIB_API size_t ZSTD_CCtx_getParameter(ZSTD_CCtx* cctx, ZSTD_cParameter param, unsigned* value);
-
-/*! ZSTD_CCtx_setPledgedSrcSize() :
- *  Total input data size to be compressed as a single frame.
- *  This value will be controlled at the end, and result in error if not respected.
- * @result : 0, or an error code (which can be tested with ZSTD_isError()).
- *  Note 1 : 0 means zero, empty.
- *           In order to mean "unknown content size", pass constant ZSTD_CONTENTSIZE_UNKNOWN.
- *           ZSTD_CONTENTSIZE_UNKNOWN is default value for any new compression job.
- *  Note 2 : If all data is provided and consumed in a single round,
- *           this value is overriden by srcSize instead. */
-ZSTDLIB_API size_t ZSTD_CCtx_setPledgedSrcSize(ZSTD_CCtx* cctx, unsigned long long pledgedSrcSize);
-
-/*! ZSTD_CCtx_loadDictionary() :
- *  Create an internal CDict from `dict` buffer.
- *  Decompression will have to use same dictionary.
- * @result : 0, or an error code (which can be tested with ZSTD_isError()).
- *  Special: Adding a NULL (or 0-size) dictionary invalidates previous dictionary,
- *           meaning "return to no-dictionary mode".
- *  Note 1 : Dictionary will be used for all future compression jobs.
- *           To return to "no-dictionary" situation, load a NULL dictionary
- *  Note 2 : Loading a dictionary involves building tables, which are dependent on compression parameters.
- *           For this reason, compression parameters cannot be changed anymore after loading a dictionary.
- *           It's also a CPU consuming operation, with non-negligible impact on latency.
- *  Note 3 :`dict` content will be copied internally.
- *           Use ZSTD_CCtx_loadDictionary_byReference() to reference dictionary content instead.
- *           In such a case, dictionary buffer must outlive its users.
- *  Note 4 : Use ZSTD_CCtx_loadDictionary_advanced()
- *           to precisely select how dictionary content must be interpreted. */
-ZSTDLIB_API size_t ZSTD_CCtx_loadDictionary(ZSTD_CCtx* cctx, const void* dict, size_t dictSize);
-ZSTDLIB_API size_t ZSTD_CCtx_loadDictionary_byReference(ZSTD_CCtx* cctx, const void* dict, size_t dictSize);
-ZSTDLIB_API size_t ZSTD_CCtx_loadDictionary_advanced(ZSTD_CCtx* cctx, const void* dict, size_t dictSize, ZSTD_dictLoadMethod_e dictLoadMethod, ZSTD_dictContentType_e dictContentType);
-
-
-/*! ZSTD_CCtx_refCDict() :
- *  Reference a prepared dictionary, to be used for all next compression jobs.
- *  Note that compression parameters are enforced from within CDict,
- *  and supercede any compression parameter previously set within CCtx.
- *  The dictionary will remain valid for future compression jobs using same CCtx.
- * @result : 0, or an error code (which can be tested with ZSTD_isError()).
- *  Special : adding a NULL CDict means "return to no-dictionary mode".
- *  Note 1 : Currently, only one dictionary can be managed.
- *           Adding a new dictionary effectively "discards" any previous one.
- *  Note 2 : CDict is just referenced, its lifetime must outlive CCtx. */
-ZSTDLIB_API size_t ZSTD_CCtx_refCDict(ZSTD_CCtx* cctx, const ZSTD_CDict* cdict);
-
-/*! ZSTD_CCtx_refPrefix() :
- *  Reference a prefix (single-usage dictionary) for next compression job.
- *  Decompression will need same prefix to properly regenerate data.
- *  Compressing with a prefix is similar in outcome as performing a diff and compressing it,
- *  but performs much faster, especially during decompression (compression speed is tunable with compression level).
- *  Note that prefix is **only used once**. Tables are discarded at end of compression job (ZSTD_e_end).
- * @result : 0, or an error code (which can be tested with ZSTD_isError()).
- *  Special: Adding any prefix (including NULL) invalidates any previous prefix or dictionary
- *  Note 1 : Prefix buffer is referenced. It **must** outlive compression job.
- *           Its contain must remain unmodified up to end of compression (ZSTD_e_end).
- *  Note 2 : If the intention is to diff some large src data blob with some prior version of itself,
- *           ensure that the window size is large enough to contain the entire source.
- *           See ZSTD_p_windowLog.
- *  Note 3 : Referencing a prefix involves building tables, which are dependent on compression parameters.
- *           It's a CPU consuming operation, with non-negligible impact on latency.
- *           If there is a need to use same prefix multiple times, consider loadDictionary instead.
- *  Note 4 : By default, the prefix is treated as raw content (ZSTD_dm_rawContent).
- *           Use ZSTD_CCtx_refPrefix_advanced() to alter dictMode. */
-ZSTDLIB_API size_t ZSTD_CCtx_refPrefix(ZSTD_CCtx* cctx,
-                                       const void* prefix, size_t prefixSize);
-ZSTDLIB_API size_t ZSTD_CCtx_refPrefix_advanced(ZSTD_CCtx* cctx,
-                                       const void* prefix, size_t prefixSize,
-                                       ZSTD_dictContentType_e dictContentType);
-
-/*! ZSTD_CCtx_reset() :
- *  Return a CCtx to clean state.
- *  Useful after an error, or to interrupt an ongoing compression job and start a new one.
- *  Any internal data not yet flushed is cancelled.
- *  The parameters and dictionary are kept unchanged, to reset them use ZSTD_CCtx_resetParameters().
- */
-ZSTDLIB_API void ZSTD_CCtx_reset(ZSTD_CCtx* cctx);
-
-/*! ZSTD_CCtx_resetParameters() :
- *  All parameters are back to default values (compression level is ZSTD_CLEVEL_DEFAULT).
- *  Dictionary (if any) is dropped.
- *  Resetting parameters is only possible during frame initialization (before starting compression).
- *  To reset the context use ZSTD_CCtx_reset().
- *  @return 0 or an error code (which can be checked with ZSTD_isError()).
- */
-ZSTDLIB_API size_t ZSTD_CCtx_resetParameters(ZSTD_CCtx* cctx);
-
-
-
-typedef enum {
-    ZSTD_e_continue=0, /* collect more data, encoder decides when to output compressed result, for optimal compression ratio */
-    ZSTD_e_flush,      /* flush any data provided so far,
-                        * it creates (at least) one new block, that can be decoded immediately on reception;
-                        * frame will continue: any future data can still reference previously compressed data, improving compression. */
-    ZSTD_e_end         /* flush any remaining data and close current frame.
-                        * any additional data starts a new frame.
-                        * each frame is independent (does not reference any content from previous frame). */
-} ZSTD_EndDirective;
-
-/*! ZSTD_compress_generic() :
- *  Behave about the same as ZSTD_compressStream. To note :
- *  - Compression parameters are pushed into CCtx before starting compression, using ZSTD_CCtx_setParameter()
- *  - Compression parameters cannot be changed once compression is started.
- *  - outpot->pos must be <= dstCapacity, input->pos must be <= srcSize
- *  - outpot->pos and input->pos will be updated. They are guaranteed to remain below their respective limit.
- *  - In single-thread mode (default), function is blocking : it completed its job before returning to caller.
- *  - In multi-thread mode, function is non-blocking : it just acquires a copy of input, and distribute job to internal worker threads,
- *                                                     and then immediately returns, just indicating that there is some data remaining to be flushed.
- *                                                     The function nonetheless guarantees forward progress : it will return only after it reads or write at least 1+ byte.
- *  - Exception : in multi-threading mode, if the first call requests a ZSTD_e_end directive, it is blocking : it will complete compression before giving back control to caller.
- *  - @return provides a minimum amount of data remaining to be flushed from internal buffers
- *            or an error code, which can be tested using ZSTD_isError().
- *            if @return != 0, flush is not fully completed, there is still some data left within internal buffers.
- *            This is useful for ZSTD_e_flush, since in this case more flushes are necessary to empty all buffers.
- *            For ZSTD_e_end, @return == 0 when internal buffers are fully flushed and frame is completed.
- *  - after a ZSTD_e_end directive, if internal buffer is not fully flushed (@return != 0),
- *            only ZSTD_e_end or ZSTD_e_flush operations are allowed.
- *            Before starting a new compression job, or changing compression parameters,
- *            it is required to fully flush internal buffers.
- */
-ZSTDLIB_API size_t ZSTD_compress_generic (ZSTD_CCtx* cctx,
-                                          ZSTD_outBuffer* output,
-                                          ZSTD_inBuffer* input,
-                                          ZSTD_EndDirective endOp);
-
-
-/*! ZSTD_compress_generic_simpleArgs() :
- *  Same as ZSTD_compress_generic(),
- *  but using only integral types as arguments.
- *  Argument list is larger than ZSTD_{in,out}Buffer,
- *  but can be helpful for binders from dynamic languages
- *  which have troubles handling structures containing memory pointers.
- */
-ZSTDLIB_API size_t ZSTD_compress_generic_simpleArgs (
-                            ZSTD_CCtx* cctx,
-                            void* dst, size_t dstCapacity, size_t* dstPos,
-                      const void* src, size_t srcSize, size_t* srcPos,
-                            ZSTD_EndDirective endOp);
-
-
-/*! ZSTD_CCtx_params :
- *  Quick howto :
- *  - ZSTD_createCCtxParams() : Create a ZSTD_CCtx_params structure
- *  - ZSTD_CCtxParam_setParameter() : Push parameters one by one into
- *                                    an existing ZSTD_CCtx_params structure.
- *                                    This is similar to
- *                                    ZSTD_CCtx_setParameter().
- *  - ZSTD_CCtx_setParametersUsingCCtxParams() : Apply parameters to
- *                                    an existing CCtx.
- *                                    These parameters will be applied to
- *                                    all subsequent compression jobs.
- *  - ZSTD_compress_generic() : Do compression using the CCtx.
- *  - ZSTD_freeCCtxParams() : Free the memory.
- *
- *  This can be used with ZSTD_estimateCCtxSize_advanced_usingCCtxParams()
- *  for static allocation for single-threaded compression.
- */
-ZSTDLIB_API ZSTD_CCtx_params* ZSTD_createCCtxParams(void);
-ZSTDLIB_API size_t ZSTD_freeCCtxParams(ZSTD_CCtx_params* params);
-
-
-/*! ZSTD_CCtxParams_reset() :
- *  Reset params to default values.
- */
-ZSTDLIB_API size_t ZSTD_CCtxParams_reset(ZSTD_CCtx_params* params);
-
-/*! ZSTD_CCtxParams_init() :
- *  Initializes the compression parameters of cctxParams according to
- *  compression level. All other parameters are reset to their default values.
- */
-ZSTDLIB_API size_t ZSTD_CCtxParams_init(ZSTD_CCtx_params* cctxParams, int compressionLevel);
-
-/*! ZSTD_CCtxParams_init_advanced() :
- *  Initializes the compression and frame parameters of cctxParams according to
- *  params. All other parameters are reset to their default values.
- */
-ZSTDLIB_API size_t ZSTD_CCtxParams_init_advanced(ZSTD_CCtx_params* cctxParams, ZSTD_parameters params);
-
-
-/*! ZSTD_CCtxParam_setParameter() :
- *  Similar to ZSTD_CCtx_setParameter.
- *  Set one compression parameter, selected by enum ZSTD_cParameter.
- *  Parameters must be applied to a ZSTD_CCtx using ZSTD_CCtx_setParametersUsingCCtxParams().
- *  Note : when `value` is an enum, cast it to unsigned for proper type checking.
- * @result : 0, or an error code (which can be tested with ZSTD_isError()).
- */
-ZSTDLIB_API size_t ZSTD_CCtxParam_setParameter(ZSTD_CCtx_params* params, ZSTD_cParameter param, unsigned value);
-
-/*! ZSTD_CCtxParam_getParameter() :
- * Similar to ZSTD_CCtx_getParameter.
- * Get the requested value of one compression parameter, selected by enum ZSTD_cParameter.
- * @result : 0, or an error code (which can be tested with ZSTD_isError()).
- */
-ZSTDLIB_API size_t ZSTD_CCtxParam_getParameter(ZSTD_CCtx_params* params, ZSTD_cParameter param, unsigned* value);
-
-/*! ZSTD_CCtx_setParametersUsingCCtxParams() :
- *  Apply a set of ZSTD_CCtx_params to the compression context.
- *  This can be done even after compression is started,
- *    if nbWorkers==0, this will have no impact until a new compression is started.
- *    if nbWorkers>=1, new parameters will be picked up at next job,
- *       with a few restrictions (windowLog, pledgedSrcSize, nbWorkers, jobSize, and overlapLog are not updated).
- */
-ZSTDLIB_API size_t ZSTD_CCtx_setParametersUsingCCtxParams(
-        ZSTD_CCtx* cctx, const ZSTD_CCtx_params* params);
-
-
-/* ==================================== */
-/*===   Advanced decompression API   ===*/
-/* ==================================== */
-
-/* The following API works the same way as the advanced compression API :
- * a context is created, parameters are pushed into it one by one,
- * then the context can be used to decompress data using an interface similar to the straming API.
- */
-
-/*! ZSTD_DCtx_loadDictionary() :
- *  Create an internal DDict from dict buffer,
- *  to be used to decompress next frames.
- * @result : 0, or an error code (which can be tested with ZSTD_isError()).
- *  Special : Adding a NULL (or 0-size) dictionary invalidates any previous dictionary,
- *            meaning "return to no-dictionary mode".
- *  Note 1 : `dict` content will be copied internally.
- *            Use ZSTD_DCtx_loadDictionary_byReference()
- *            to reference dictionary content instead.
- *            In which case, the dictionary buffer must outlive its users.
- *  Note 2 : Loading a dictionary involves building tables,
- *           which has a non-negligible impact on CPU usage and latency.
- *  Note 3 : Use ZSTD_DCtx_loadDictionary_advanced() to select
- *           how dictionary content will be interpreted and loaded.
- */
-ZSTDLIB_API size_t ZSTD_DCtx_loadDictionary(ZSTD_DCtx* dctx, const void* dict, size_t dictSize);
-ZSTDLIB_API size_t ZSTD_DCtx_loadDictionary_byReference(ZSTD_DCtx* dctx, const void* dict, size_t dictSize);
-ZSTDLIB_API size_t ZSTD_DCtx_loadDictionary_advanced(ZSTD_DCtx* dctx, const void* dict, size_t dictSize, ZSTD_dictLoadMethod_e dictLoadMethod, ZSTD_dictContentType_e dictContentType);
-
-
-/*! ZSTD_DCtx_refDDict() :
- *  Reference a prepared dictionary, to be used to decompress next frames.
- *  The dictionary remains active for decompression of future frames using same DCtx.
- * @result : 0, or an error code (which can be tested with ZSTD_isError()).
- *  Note 1 : Currently, only one dictionary can be managed.
- *           Referencing a new dictionary effectively "discards" any previous one.
- *  Special : adding a NULL DDict means "return to no-dictionary mode".
- *  Note 2 : DDict is just referenced, its lifetime must outlive its usage from DCtx.
- */
-ZSTDLIB_API size_t ZSTD_DCtx_refDDict(ZSTD_DCtx* dctx, const ZSTD_DDict* ddict);
-
-
-/*! ZSTD_DCtx_refPrefix() :
- *  Reference a prefix (single-usage dictionary) for next compression job.
- *  This is the reverse operation of ZSTD_CCtx_refPrefix(),
- *  and must use the same prefix as the one used during compression.
- *  Prefix is **only used once**. Reference is discarded at end of frame.
- *  End of frame is reached when ZSTD_DCtx_decompress_generic() returns 0.
- * @result : 0, or an error code (which can be tested with ZSTD_isError()).
- *  Note 1 : Adding any prefix (including NULL) invalidates any previously set prefix or dictionary
- *  Note 2 : Prefix buffer is referenced. It **must** outlive decompression job.
- *           Prefix buffer must remain unmodified up to the end of frame,
- *           reached when ZSTD_DCtx_decompress_generic() returns 0.
- *  Note 3 : By default, the prefix is treated as raw content (ZSTD_dm_rawContent).
- *           Use ZSTD_CCtx_refPrefix_advanced() to alter dictMode.
- *  Note 4 : Referencing a raw content prefix has almost no cpu nor memory cost.
- *           A fulldict prefix is more costly though.
- */
-ZSTDLIB_API size_t ZSTD_DCtx_refPrefix(ZSTD_DCtx* dctx,
-                                    const void* prefix, size_t prefixSize);
-ZSTDLIB_API size_t ZSTD_DCtx_refPrefix_advanced(ZSTD_DCtx* dctx,
-                                    const void* prefix, size_t prefixSize,
-                                    ZSTD_dictContentType_e dictContentType);
-
-
-/*! ZSTD_DCtx_setMaxWindowSize() :
- *  Refuses allocating internal buffers for frames requiring a window size larger than provided limit.
- *  This is useful to prevent a decoder context from reserving too much memory for itself (potential attack scenario).
- *  This parameter is only useful in streaming mode, since no internal buffer is allocated in direct mode.
- *  By default, a decompression context accepts all window sizes <= (1 << ZSTD_WINDOWLOG_MAX)
- * @return : 0, or an error code (which can be tested using ZSTD_isError()).
- */
-ZSTDLIB_API size_t ZSTD_DCtx_setMaxWindowSize(ZSTD_DCtx* dctx, size_t maxWindowSize);
-
-
-/*! ZSTD_DCtx_setFormat() :
- *  Instruct the decoder context about what kind of data to decode next.
- *  This instruction is mandatory to decode data without a fully-formed header,
- *  such ZSTD_f_zstd1_magicless for example.
- * @return : 0, or an error code (which can be tested using ZSTD_isError()).
- */
-ZSTDLIB_API size_t ZSTD_DCtx_setFormat(ZSTD_DCtx* dctx, ZSTD_format_e format);
-
-
-/*! ZSTD_getFrameHeader_advanced() :
- *  same as ZSTD_getFrameHeader(),
- *  with added capability to select a format (like ZSTD_f_zstd1_magicless) */
-ZSTDLIB_API size_t ZSTD_getFrameHeader_advanced(ZSTD_frameHeader* zfhPtr,
-                        const void* src, size_t srcSize, ZSTD_format_e format);
-
-
-/*! ZSTD_decompress_generic() :
- *  Behave the same as ZSTD_decompressStream.
- *  Decompression parameters cannot be changed once decompression is started.
- * @return : an error code, which can be tested using ZSTD_isError()
- *           if >0, a hint, nb of expected input bytes for next invocation.
- *           `0` means : a frame has just been fully decoded and flushed.
- */
-ZSTDLIB_API size_t ZSTD_decompress_generic(ZSTD_DCtx* dctx,
-                                           ZSTD_outBuffer* output,
-                                           ZSTD_inBuffer* input);
-
-
-/*! ZSTD_decompress_generic_simpleArgs() :
- *  Same as ZSTD_decompress_generic(),
- *  but using only integral types as arguments.
- *  Argument list is larger than ZSTD_{in,out}Buffer,
- *  but can be helpful for binders from dynamic languages
- *  which have troubles handling structures containing memory pointers.
- */
-ZSTDLIB_API size_t ZSTD_decompress_generic_simpleArgs (
-                            ZSTD_DCtx* dctx,
-                            void* dst, size_t dstCapacity, size_t* dstPos,
-                      const void* src, size_t srcSize, size_t* srcPos);
-
-
-/*! ZSTD_DCtx_reset() :
- *  Return a DCtx to clean state.
- *  If a decompression was ongoing, any internal data not yet flushed is cancelled.
- *  All parameters are back to default values, including sticky ones.
- *  Dictionary (if any) is dropped.
- *  Parameters can be modified again after a reset.
- */
-ZSTDLIB_API void ZSTD_DCtx_reset(ZSTD_DCtx* dctx);
-
-
 
 /* ============================ */
 /**       Block level API       */
@@ -1501,10 +1741,10 @@ ZSTDLIB_API void ZSTD_DCtx_reset(ZSTD_DCtx* dctx);
       + copyCCtx() and copyDCtx() can be used too
     - Block size is limited, it must be <= ZSTD_getBlockSize() <= ZSTD_BLOCKSIZE_MAX == 128 KB
       + If input is larger than a block size, it's necessary to split input data into multiple blocks
-      + For inputs larger than a single block size, consider using the regular ZSTD_compress() instead.
+      + For inputs larger than a single block, really consider using regular ZSTD_compress() instead.
         Frame metadata is not that costly, and quickly becomes negligible as source size grows larger.
     - When a block is considered not compressible enough, ZSTD_compressBlock() result will be zero.
-      In which case, nothing is produced into `dst`.
+      In which case, nothing is produced into `dst` !
       + User must test for such outcome and deal directly with uncompressed data
       + ZSTD_decompressBlock() doesn't accept uncompressed data as input !!!
       + In case of multiple successive blocks, should some of them be uncompressed,
diff --git a/programs/Makefile b/programs/Makefile
index 32dbc67eff5c..d1910fbb4b79 100644
--- a/programs/Makefile
+++ b/programs/Makefile
@@ -29,7 +29,12 @@ LIBVER  := $(shell echo $(LIBVER_SCRIPT))
 
 ZSTD_VERSION = $(LIBVER)
 
-GREP = grep --color=never
+HAVE_COLORNEVER = $(shell echo a | grep --color=never a > /dev/null 2> /dev/null && echo 1 || echo 0)
+GREP_OPTIONS ?=
+ifeq ($HAVE_COLORNEVER, 1)
+GREP_OPTIONS += --color=never
+endif
+GREP = grep $(GREP_OPTIONS)
 
 ifeq ($(shell $(CC) -v 2>&1 | $(GREP) -c "gcc version "), 1)
 ALIGN_LOOP = -falign-loops=32
@@ -48,7 +53,7 @@ DEBUGFLAGS+=-Wall -Wextra -Wcast-qual -Wcast-align -Wshadow \
             -Wstrict-aliasing=1 -Wswitch-enum -Wdeclaration-after-statement \
             -Wstrict-prototypes -Wundef -Wpointer-arith -Wformat-security \
             -Wvla -Wformat=2 -Winit-self -Wfloat-equal -Wwrite-strings \
-            -Wredundant-decls -Wmissing-prototypes
+            -Wredundant-decls -Wmissing-prototypes -Wc++-compat
 CFLAGS  += $(DEBUGFLAGS) $(MOREFLAGS)
 FLAGS    = $(CPPFLAGS) $(CFLAGS) $(LDFLAGS)
 
@@ -91,7 +96,7 @@ VOID = /dev/null
 
 # thread detection
 NO_THREAD_MSG := ==> no threads, building without multithreading support
-HAVE_PTHREAD := $(shell printf '\#include \nint main(void) { return 0; }' | $(CC) $(FLAGS) -o have_pthread$(EXT) -x c - -pthread 2> $(VOID) && rm have_pthread$(EXT) && echo 1 || echo 0)
+HAVE_PTHREAD := $(shell printf '\#include \nint main(void) { return 0; }' > have_pthread.c && $(CC) $(FLAGS) -o have_pthread$(EXT) have_pthread.c -pthread 2> $(VOID) && rm have_pthread$(EXT) && echo 1 || echo 0; rm have_pthread.c)
 HAVE_THREAD := $(shell [ "$(HAVE_PTHREAD)" -eq "1" -o -n "$(filter Windows%,$(OS))" ] && echo 1 || echo 0)
 ifeq ($(HAVE_THREAD), 1)
 THREAD_MSG := ==> building with threading support
@@ -103,7 +108,7 @@ endif
 
 # zlib detection
 NO_ZLIB_MSG := ==> no zlib, building zstd without .gz support
-HAVE_ZLIB := $(shell printf '\#include \nint main(void) { return 0; }' | $(CC) $(FLAGS) -o have_zlib$(EXT) -x c - -lz 2> $(VOID) && rm have_zlib$(EXT) && echo 1 || echo 0)
+HAVE_ZLIB := $(shell printf '\#include \nint main(void) { return 0; }' > have_zlib.c && $(CC) $(FLAGS) -o have_zlib$(EXT) have_zlib.c -lz 2> $(VOID) && rm have_zlib$(EXT) && echo 1 || echo 0; rm have_zlib.c)
 ifeq ($(HAVE_ZLIB), 1)
 ZLIB_MSG := ==> building zstd with .gz compression support
 ZLIBCPP = -DZSTD_GZCOMPRESS -DZSTD_GZDECOMPRESS
@@ -114,7 +119,7 @@ endif
 
 # lzma detection
 NO_LZMA_MSG := ==> no liblzma, building zstd without .xz/.lzma support
-HAVE_LZMA := $(shell printf '\#include \nint main(void) { return 0; }' | $(CC) $(FLAGS) -o have_lzma$(EXT) -x c - -llzma 2> $(VOID) && rm have_lzma$(EXT) && echo 1 || echo 0)
+HAVE_LZMA := $(shell printf '\#include \nint main(void) { return 0; }' > have_lzma.c && $(CC) $(FLAGS) -o have_lzma$(EXT) have_lzma.c -llzma 2> $(VOID) && rm have_lzma$(EXT) && echo 1 || echo 0; rm have_lzma.c)
 ifeq ($(HAVE_LZMA), 1)
 LZMA_MSG := ==> building zstd with .xz/.lzma compression support
 LZMACPP = -DZSTD_LZMACOMPRESS -DZSTD_LZMADECOMPRESS
@@ -125,7 +130,7 @@ endif
 
 # lz4 detection
 NO_LZ4_MSG := ==> no liblz4, building zstd without .lz4 support
-HAVE_LZ4 := $(shell printf '\#include \n\#include \nint main(void) { return 0; }' | $(CC) $(FLAGS) -o have_lz4$(EXT) -x c - -llz4 2> $(VOID) && rm have_lz4$(EXT) && echo 1 || echo 0)
+HAVE_LZ4 := $(shell printf '\#include \n\#include \nint main(void) { return 0; }' > have_lz4.c && $(CC) $(FLAGS) -o have_lz4$(EXT) have_lz4.c -llz4 2> $(VOID) && rm have_lz4$(EXT) && echo 1 || echo 0; rm have_lz4.c)
 ifeq ($(HAVE_LZ4), 1)
 LZ4_MSG := ==> building zstd with .lz4 compression support
 LZ4CPP = -DZSTD_LZ4COMPRESS -DZSTD_LZ4DECOMPRESS
@@ -160,7 +165,7 @@ $(ZSTDDECOMP_O): CFLAGS += $(ALIGN_LOOP)
 zstd : CPPFLAGS += $(THREAD_CPP) $(ZLIBCPP) $(LZMACPP) $(LZ4CPP)
 zstd : LDFLAGS += $(THREAD_LD) $(ZLIBLD) $(LZMALD) $(LZ4LD) $(DEBUGFLAGS_LD)
 zstd : CPPFLAGS += -DZSTD_LEGACY_SUPPORT=$(ZSTD_LEGACY_SUPPORT)
-zstd : $(ZSTDLIB_FILES) zstdcli.o fileio.o bench.o datagen.o dibio.o
+zstd : $(ZSTDLIB_FILES) zstdcli.o util.o fileio.o benchfn.o benchzstd.o datagen.o dibio.o
 	@echo "$(THREAD_MSG)"
 	@echo "$(ZLIB_MSG)"
 	@echo "$(LZMA_MSG)"
@@ -178,13 +183,13 @@ zstd-release: zstd
 zstd32 : CPPFLAGS += $(THREAD_CPP)
 zstd32 : LDFLAGS  += $(THREAD_LD)
 zstd32 : CPPFLAGS += -DZSTD_LEGACY_SUPPORT=$(ZSTD_LEGACY_SUPPORT)
-zstd32 : $(ZSTDLIB_FILES) zstdcli.c fileio.c bench.c datagen.c dibio.c
+zstd32 : $(ZSTDLIB_FILES) zstdcli.c util.c fileio.c benchfn.c benchzstd.c datagen.c dibio.c
 ifneq (,$(filter Windows%,$(OS)))
 	windres/generate_res.bat
 endif
 	$(CC) -m32 $(FLAGS) $^ $(RES32_FILE) -o $@$(EXT)
 
-zstd-nolegacy : $(ZSTD_FILES) $(ZDICT_FILES) zstdcli.o fileio.c bench.o datagen.o dibio.o
+zstd-nolegacy : $(ZSTD_FILES) $(ZDICT_FILES) zstdcli.o util.o fileio.c benchfn.o benchzstd.o datagen.o dibio.o
 	$(CC) $(FLAGS) $^ -o $@$(EXT) $(LDFLAGS)
 
 zstd-nomt : THREAD_CPP :=
@@ -203,27 +208,27 @@ zstd-noxz : LZMA_MSG := - xz/lzma support is disabled
 zstd-noxz : zstd
 
 
-zstd-pgo : MOREFLAGS = -fprofile-generate
-zstd-pgo : clean zstd
+zstd-pgo :
+	$(MAKE) clean
+	$(MAKE) zstd MOREFLAGS=-fprofile-generate
 	./zstd -b19i1 $(PROFILE_WITH)
 	./zstd -b16i1 $(PROFILE_WITH)
 	./zstd -b9i2 $(PROFILE_WITH)
 	./zstd -b $(PROFILE_WITH)
 	./zstd -b7i2 $(PROFILE_WITH)
 	./zstd -b5 $(PROFILE_WITH)
-	$(RM) zstd
-	$(RM) $(ZSTDDECOMP_O)
+	$(RM) zstd *.o $(ZSTDDECOMP_O) $(ZSTDDIR)/compress/*.o
 	$(MAKE) zstd MOREFLAGS=-fprofile-use
 
 # minimal target, with only zstd compression and decompression. no bench. no legacy.
 zstd-small: CFLAGS = -Os -s
-zstd-frugal zstd-small: $(ZSTD_FILES) zstdcli.c fileio.c
+zstd-frugal zstd-small: $(ZSTD_FILES) zstdcli.c util.c fileio.c
 	$(CC) $(FLAGS) -DZSTD_NOBENCH -DZSTD_NODICT $^ -o $@$(EXT)
 
-zstd-decompress: $(ZSTDCOMMON_FILES) $(ZSTDDECOMP_FILES) zstdcli.c fileio.c
+zstd-decompress: $(ZSTDCOMMON_FILES) $(ZSTDDECOMP_FILES) zstdcli.c util.c fileio.c
 	$(CC) $(FLAGS) -DZSTD_NOBENCH -DZSTD_NODICT -DZSTD_NOCOMPRESS $^ -o $@$(EXT)
 
-zstd-compress: $(ZSTDCOMMON_FILES) $(ZSTDCOMP_FILES) zstdcli.c fileio.c
+zstd-compress: $(ZSTDCOMMON_FILES) $(ZSTDCOMP_FILES) zstdcli.c util.c fileio.c
 	$(CC) $(FLAGS) -DZSTD_NOBENCH -DZSTD_NODICT -DZSTD_NODECOMPRESS $^ -o $@$(EXT)
 
 zstdmt: zstd
@@ -275,7 +280,12 @@ preview-man: clean-man man
 #-----------------------------------------------------------------------------
 ifneq (,$(filter $(shell uname),Linux Darwin GNU/kFreeBSD GNU OpenBSD FreeBSD NetBSD DragonFly SunOS Haiku))
 
-EGREP = egrep --color=never
+HAVE_COLORNEVER = $(shell echo a | egrep --color=never a > /dev/null 2> /dev/null && echo 1 || echo 0)
+EGREP_OPTIONS ?=
+ifeq ($HAVE_COLORNEVER, 1)
+EGREP_OPTIONS += --color=never
+endif
+EGREP = egrep $(EGREP_OPTIONS)
 
 # Print a two column output of targets and their description. To add a target description, put a
 # comment in the Makefile with the format "## : ".  For example:
@@ -352,6 +362,7 @@ uninstall:
 	@$(RM) $(DESTDIR)$(BINDIR)/zstdless
 	@$(RM) $(DESTDIR)$(BINDIR)/zstdcat
 	@$(RM) $(DESTDIR)$(BINDIR)/unzstd
+	@$(RM) $(DESTDIR)$(BINDIR)/zstdmt
 	@$(RM) $(DESTDIR)$(BINDIR)/zstd
 	@$(RM) $(DESTDIR)$(MAN1DIR)/zstdless.1
 	@$(RM) $(DESTDIR)$(MAN1DIR)/zstdgrep.1
diff --git a/programs/README.md b/programs/README.md
index ca9056eaaa43..afbebaa16fdf 100644
--- a/programs/README.md
+++ b/programs/README.md
@@ -172,6 +172,11 @@ Benchmark arguments :
 --priority=rt : set process priority to real-time
 ```
 
+#### Restricted usage of Environment Variables
+Using environment variables to set compression/decompression parameters has security implications. Therefore,
+we intentionally restrict its usage. Currently, only `ZSTD_CLEVEL` is supported for setting compression level.
+If the value of `ZSTD_CLEVEL` is not a valid integer, it will be ignored with a warning message.
+Note that command line options will override corresponding environment variable settings.
 
 #### Long distance matching mode
 The long distance matching mode, enabled with `--long`, is designed to improve
diff --git a/programs/benchfn.c b/programs/benchfn.c
new file mode 100644
index 000000000000..f5118d0871f1
--- /dev/null
+++ b/programs/benchfn.c
@@ -0,0 +1,263 @@
+/*
+ * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
+ * You may select, at your option, one of the above-listed licenses.
+ */
+
+
+
+/* *************************************
+*  Includes
+***************************************/
+#include "platform.h"    /* Large Files support */
+#include "util.h"        /* UTIL_getFileSize, UTIL_sleep */
+#include       /* malloc, free */
+#include       /* memset */
+#include        /* fprintf, fopen */
+#undef NDEBUG            /* assert must not be disabled */
+#include       /* assert */
+
+#include "mem.h"
+#include "benchfn.h"
+
+
+/* *************************************
+*  Constants
+***************************************/
+#define TIMELOOP_MICROSEC     (1*1000000ULL) /* 1 second */
+#define TIMELOOP_NANOSEC      (1*1000000000ULL) /* 1 second */
+#define ACTIVEPERIOD_MICROSEC (70*TIMELOOP_MICROSEC) /* 70 seconds */
+#define COOLPERIOD_SEC        10
+
+#define KB *(1 <<10)
+#define MB *(1 <<20)
+#define GB *(1U<<30)
+
+
+/* *************************************
+*  Errors
+***************************************/
+#ifndef DEBUG
+#  define DEBUG 0
+#endif
+
+#define DISPLAY(...)       fprintf(stderr, __VA_ARGS__)
+#define DEBUGOUTPUT(...) { if (DEBUG) DISPLAY(__VA_ARGS__); }
+
+/* error without displaying */
+#define RETURN_QUIET_ERROR(retValue, ...) {           \
+    DEBUGOUTPUT("%s: %i: \n", __FILE__, __LINE__);    \
+    DEBUGOUTPUT("Error : ");                          \
+    DEBUGOUTPUT(__VA_ARGS__);                         \
+    DEBUGOUTPUT(" \n");                               \
+    return retValue;                                  \
+}
+
+
+/* *************************************
+*  Benchmarking an arbitrary function
+***************************************/
+
+int BMK_isSuccessful_runOutcome(BMK_runOutcome_t outcome)
+{
+    return outcome.error_tag_never_ever_use_directly == 0;
+}
+
+/* warning : this function will stop program execution if outcome is invalid !
+ *           check outcome validity first, using BMK_isValid_runResult() */
+BMK_runTime_t BMK_extract_runTime(BMK_runOutcome_t outcome)
+{
+    assert(outcome.error_tag_never_ever_use_directly == 0);
+    return outcome.internal_never_ever_use_directly;
+}
+
+size_t BMK_extract_errorResult(BMK_runOutcome_t outcome)
+{
+    assert(outcome.error_tag_never_ever_use_directly != 0);
+    return outcome.error_result_never_ever_use_directly;
+}
+
+static BMK_runOutcome_t BMK_runOutcome_error(size_t errorResult)
+{
+    BMK_runOutcome_t b;
+    memset(&b, 0, sizeof(b));
+    b.error_tag_never_ever_use_directly = 1;
+    b.error_result_never_ever_use_directly = errorResult;
+    return b;
+}
+
+static BMK_runOutcome_t BMK_setValid_runTime(BMK_runTime_t runTime)
+{
+    BMK_runOutcome_t outcome;
+    outcome.error_tag_never_ever_use_directly = 0;
+    outcome.internal_never_ever_use_directly = runTime;
+    return outcome;
+}
+
+
+/* initFn will be measured once, benchFn will be measured `nbLoops` times */
+/* initFn is optional, provide NULL if none */
+/* benchFn must return a size_t value that errorFn can interpret */
+/* takes # of blocks and list of size & stuff for each. */
+/* can report result of benchFn for each block into blockResult. */
+/* blockResult is optional, provide NULL if this information is not required */
+/* note : time per loop can be reported as zero if run time < timer resolution */
+BMK_runOutcome_t BMK_benchFunction(BMK_benchParams_t p,
+                                   unsigned nbLoops)
+{
+    size_t dstSize = 0;
+    nbLoops += !nbLoops;   /* minimum nbLoops is 1 */
+
+    /* init */
+    {   size_t i;
+        for(i = 0; i < p.blockCount; i++) {
+            memset(p.dstBuffers[i], 0xE5, p.dstCapacities[i]);  /* warm up and erase result buffer */
+        }
+#if 0
+        /* based on testing these seem to lower accuracy of multiple calls of 1 nbLoops vs 1 call of multiple nbLoops
+         * (Makes former slower)
+         */
+        UTIL_sleepMilli(5);  /* give processor time to other processes */
+        UTIL_waitForNextTick();
+#endif
+    }
+
+    /* benchmark */
+    {   UTIL_time_t const clockStart = UTIL_getTime();
+        unsigned loopNb, blockNb;
+        if (p.initFn != NULL) p.initFn(p.initPayload);
+        for (loopNb = 0; loopNb < nbLoops; loopNb++) {
+            for (blockNb = 0; blockNb < p.blockCount; blockNb++) {
+                size_t const res = p.benchFn(p.srcBuffers[blockNb], p.srcSizes[blockNb],
+                                   p.dstBuffers[blockNb], p.dstCapacities[blockNb],
+                                   p.benchPayload);
+                if (loopNb == 0) {
+                    if (p.blockResults != NULL) p.blockResults[blockNb] = res;
+                    if ((p.errorFn != NULL) && (p.errorFn(res))) {
+                        RETURN_QUIET_ERROR(BMK_runOutcome_error(res),
+                            "Function benchmark failed on block %u (of size %u) with error %i",
+                            blockNb, (unsigned)p.srcSizes[blockNb], (int)res);
+                    }
+                    dstSize += res;
+            }   }
+        }  /* for (loopNb = 0; loopNb < nbLoops; loopNb++) */
+
+        {   U64 const totalTime = UTIL_clockSpanNano(clockStart);
+            BMK_runTime_t rt;
+            rt.nanoSecPerRun = totalTime / nbLoops;
+            rt.sumOfReturn = dstSize;
+            return BMK_setValid_runTime(rt);
+    }   }
+}
+
+
+/* ====  Benchmarking any function, providing intermediate results  ==== */
+
+struct BMK_timedFnState_s {
+    U64 timeSpent_ns;
+    U64 timeBudget_ns;
+    U64 runBudget_ns;
+    BMK_runTime_t fastestRun;
+    unsigned nbLoops;
+    UTIL_time_t coolTime;
+};  /* typedef'd to BMK_timedFnState_t within bench.h */
+
+BMK_timedFnState_t* BMK_createTimedFnState(unsigned total_ms, unsigned run_ms)
+{
+    BMK_timedFnState_t* const r = (BMK_timedFnState_t*)malloc(sizeof(*r));
+    if (r == NULL) return NULL;   /* malloc() error */
+    BMK_resetTimedFnState(r, total_ms, run_ms);
+    return r;
+}
+
+void BMK_freeTimedFnState(BMK_timedFnState_t* state) {
+    free(state);
+}
+
+void BMK_resetTimedFnState(BMK_timedFnState_t* timedFnState, unsigned total_ms, unsigned run_ms)
+{
+    if (!total_ms) total_ms = 1 ;
+    if (!run_ms) run_ms = 1;
+    if (run_ms > total_ms) run_ms = total_ms;
+    timedFnState->timeSpent_ns = 0;
+    timedFnState->timeBudget_ns = (U64)total_ms * TIMELOOP_NANOSEC / 1000;
+    timedFnState->runBudget_ns = (U64)run_ms * TIMELOOP_NANOSEC / 1000;
+    timedFnState->fastestRun.nanoSecPerRun = (U64)(-1LL);
+    timedFnState->fastestRun.sumOfReturn = (size_t)(-1LL);
+    timedFnState->nbLoops = 1;
+    timedFnState->coolTime = UTIL_getTime();
+}
+
+/* Tells if nb of seconds set in timedFnState for all runs is spent.
+ * note : this function will return 1 if BMK_benchFunctionTimed() has actually errored. */
+int BMK_isCompleted_TimedFn(const BMK_timedFnState_t* timedFnState)
+{
+    return (timedFnState->timeSpent_ns >= timedFnState->timeBudget_ns);
+}
+
+
+#undef MIN
+#define MIN(a,b)   ( (a) < (b) ? (a) : (b) )
+
+#define MINUSABLETIME  (TIMELOOP_NANOSEC / 2)  /* 0.5 seconds */
+
+BMK_runOutcome_t BMK_benchTimedFn(BMK_timedFnState_t* cont,
+                                  BMK_benchParams_t p)
+{
+    U64 const runBudget_ns = cont->runBudget_ns;
+    U64 const runTimeMin_ns = runBudget_ns / 2;
+    int completed = 0;
+    BMK_runTime_t bestRunTime = cont->fastestRun;
+
+    while (!completed) {
+        BMK_runOutcome_t runResult;
+
+        /* Overheat protection */
+        if (UTIL_clockSpanMicro(cont->coolTime) > ACTIVEPERIOD_MICROSEC) {
+            DEBUGOUTPUT("\rcooling down ...    \r");
+            UTIL_sleep(COOLPERIOD_SEC);
+            cont->coolTime = UTIL_getTime();
+        }
+
+        /* reinitialize capacity */
+        runResult = BMK_benchFunction(p, cont->nbLoops);
+
+        if(!BMK_isSuccessful_runOutcome(runResult)) { /* error : move out */
+            return runResult;
+        }
+
+        {   BMK_runTime_t const newRunTime = BMK_extract_runTime(runResult);
+            U64 const loopDuration_ns = newRunTime.nanoSecPerRun * cont->nbLoops;
+
+            cont->timeSpent_ns += loopDuration_ns;
+
+            /* estimate nbLoops for next run to last approximately 1 second */
+            if (loopDuration_ns > (runBudget_ns / 50)) {
+                U64 const fastestRun_ns = MIN(bestRunTime.nanoSecPerRun, newRunTime.nanoSecPerRun);
+                cont->nbLoops = (U32)(runBudget_ns / fastestRun_ns) + 1;
+            } else {
+                /* previous run was too short : blindly increase workload by x multiplier */
+                const unsigned multiplier = 10;
+                assert(cont->nbLoops < ((unsigned)-1) / multiplier);  /* avoid overflow */
+                cont->nbLoops *= multiplier;
+            }
+
+            if(loopDuration_ns < runTimeMin_ns) {
+                /* don't report results for which benchmark run time was too small : increased risks of rounding errors */
+                assert(completed == 0);
+                continue;
+            } else {
+                if(newRunTime.nanoSecPerRun < bestRunTime.nanoSecPerRun) {
+                    bestRunTime = newRunTime;
+                }
+                completed = 1;
+            }
+        }
+    }   /* while (!completed) */
+
+    return BMK_setValid_runTime(bestRunTime);
+}
diff --git a/programs/benchfn.h b/programs/benchfn.h
new file mode 100644
index 000000000000..3ca36e3623da
--- /dev/null
+++ b/programs/benchfn.h
@@ -0,0 +1,167 @@
+/*
+ * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
+ * You may select, at your option, one of the above-listed licenses.
+ */
+
+
+/* benchfn :
+ * benchmark any function on a set of input
+ * providing result in nanoSecPerRun
+ * or detecting and returning an error
+ */
+
+#if defined (__cplusplus)
+extern "C" {
+#endif
+
+#ifndef BENCH_FN_H_23876
+#define BENCH_FN_H_23876
+
+/* ===  Dependencies  === */
+#include    /* size_t */
+
+
+/* ====  Benchmark any function, iterated on a set of blocks  ==== */
+
+/* BMK_runTime_t: valid result return type */
+
+typedef struct {
+    unsigned long long nanoSecPerRun;  /* time per iteration (over all blocks) */
+    size_t sumOfReturn;         /* sum of return values */
+} BMK_runTime_t;
+
+
+/* BMK_runOutcome_t:
+ * type expressing the outcome of a benchmark run by BMK_benchFunction(),
+ * which can be either valid or invalid.
+ * benchmark outcome can be invalid if errorFn is provided.
+ * BMK_runOutcome_t must be considered "opaque" : never access its members directly.
+ * Instead, use its assigned methods :
+ * BMK_isSuccessful_runOutcome, BMK_extract_runTime, BMK_extract_errorResult.
+ * The structure is only described here to allow its allocation on stack. */
+
+typedef struct {
+    BMK_runTime_t internal_never_ever_use_directly;
+    size_t error_result_never_ever_use_directly;
+    int error_tag_never_ever_use_directly;
+} BMK_runOutcome_t;
+
+
+/* prototypes for benchmarked functions */
+typedef size_t (*BMK_benchFn_t)(const void* src, size_t srcSize, void* dst, size_t dstCapacity, void* customPayload);
+typedef size_t (*BMK_initFn_t)(void* initPayload);
+typedef unsigned (*BMK_errorFn_t)(size_t);
+
+
+/* BMK_benchFunction() parameters are provided through following structure.
+ * This is preferable for readability,
+ * as the number of parameters required is pretty large.
+ * No initializer is provided, because it doesn't make sense to provide some "default" :
+ * all parameters should be specified by the caller */
+typedef struct {
+    BMK_benchFn_t benchFn;   /* the function to benchmark, over the set of blocks */
+    void* benchPayload;      /* pass custom parameters to benchFn  :
+                              * (*benchFn)(srcBuffers[i], srcSizes[i], dstBuffers[i], dstCapacities[i], benchPayload) */
+    BMK_initFn_t initFn;     /* (*initFn)(initPayload) is run once per run, at the beginning. */
+    void* initPayload;       /* Both arguments can be NULL, in which case nothing is run. */
+    BMK_errorFn_t errorFn;   /* errorFn will check each return value of benchFn over each block, to determine if it failed or not.
+                              * errorFn can be NULL, in which case no check is performed.
+                              * errorFn must return 0 when benchFn was successful, and >= 1 if it detects an error.
+                              * Execution is stopped as soon as an error is detected.
+                              * the triggering return value can be retrieved using BMK_extract_errorResult(). */
+    size_t blockCount;       /* number of blocks to operate benchFn on.
+                              * It's also the size of all array parameters :
+                              * srcBuffers, srcSizes, dstBuffers, dstCapacities, blockResults */
+    const void *const * srcBuffers; /* array of buffers to be operated on by benchFn */
+    const size_t* srcSizes;  /* array of the sizes of srcBuffers buffers */
+    void *const * dstBuffers;/* array of buffers to be written into by benchFn */
+    const size_t* dstCapacities; /* array of the capacities of dstBuffers buffers */
+    size_t* blockResults;    /* Optional: store the return value of benchFn for each block. Use NULL if this result is not requested. */
+} BMK_benchParams_t;
+
+
+/* BMK_benchFunction() :
+ * This function benchmarks benchFn and initFn, providing a result.
+ *
+ * params : see description of BMK_benchParams_t above.
+ * nbLoops: defines number of times benchFn is run over the full set of blocks.
+ *          Minimum value is 1. A 0 is interpreted as a 1.
+ *
+ * @return: can express either an error or a successful result.
+ *          Use BMK_isSuccessful_runOutcome() to check if benchmark was successful.
+ *          If yes, extract the result with BMK_extract_runTime(),
+ *          it will contain :
+ *              .sumOfReturn : the sum of all return values of benchFn through all of blocks
+ *              .nanoSecPerRun : time per run of benchFn + (time for initFn / nbLoops)
+ *          .sumOfReturn is generally intended for functions which return a # of bytes written into dstBuffer,
+ *              in which case, this value will be the total amount of bytes written into dstBuffer.
+ *
+ * blockResults : when provided (!= NULL), and when benchmark is successful,
+ *                params.blockResults contains all return values of `benchFn` over all blocks.
+ *                when provided (!= NULL), and when benchmark failed,
+ *                params.blockResults contains return values of `benchFn` over all blocks preceding and including the failed block.
+ */
+BMK_runOutcome_t BMK_benchFunction(BMK_benchParams_t params, unsigned nbLoops);
+
+
+
+/* check first if the benchmark was successful or not */
+int BMK_isSuccessful_runOutcome(BMK_runOutcome_t outcome);
+
+/* If the benchmark was successful, extract the result.
+ * note : this function will abort() program execution if benchmark failed !
+ *        always check if benchmark was successful first !
+ */
+BMK_runTime_t BMK_extract_runTime(BMK_runOutcome_t outcome);
+
+/* when benchmark failed, it means one invocation of `benchFn` failed.
+ * The failure was detected by `errorFn`, operating on return values of `benchFn`.
+ * Returns the faulty return value.
+ * note : this function will abort() program execution if benchmark did not failed.
+ *        always check if benchmark failed first !
+ */
+size_t BMK_extract_errorResult(BMK_runOutcome_t outcome);
+
+
+
+/* ====  Benchmark any function, returning intermediate results  ==== */
+
+/* state information tracking benchmark session */
+typedef struct BMK_timedFnState_s BMK_timedFnState_t;
+
+/* BMK_benchTimedFn() :
+ * Similar to BMK_benchFunction(), most arguments being identical.
+ * Automatically determines `nbLoops` so that each result is regularly produced at interval of about run_ms.
+ * Note : minimum `nbLoops` is 1, therefore a run may last more than run_ms, and possibly even more than total_ms.
+ * Usage - initialize timedFnState, select benchmark duration (total_ms) and each measurement duration (run_ms)
+ *         call BMK_benchTimedFn() repetitively, each measurement is supposed to last about run_ms
+ *         Check if total time budget is spent or exceeded, using BMK_isCompleted_TimedFn()
+ */
+BMK_runOutcome_t BMK_benchTimedFn(BMK_timedFnState_t* timedFnState,
+                                  BMK_benchParams_t params);
+
+/* Tells if duration of all benchmark runs has exceeded total_ms
+ */
+int BMK_isCompleted_TimedFn(const BMK_timedFnState_t* timedFnState);
+
+/* BMK_createTimedFnState() and BMK_resetTimedFnState() :
+ * Create/Set BMK_timedFnState_t for next benchmark session,
+ * which shall last a minimum of total_ms milliseconds,
+ * producing intermediate results, paced at interval of (approximately) run_ms.
+ */
+BMK_timedFnState_t* BMK_createTimedFnState(unsigned total_ms, unsigned run_ms);
+void BMK_resetTimedFnState(BMK_timedFnState_t* timedFnState, unsigned total_ms, unsigned run_ms);
+void BMK_freeTimedFnState(BMK_timedFnState_t* state);
+
+
+
+#endif   /* BENCH_FN_H_23876 */
+
+#if defined (__cplusplus)
+}
+#endif
diff --git a/programs/bench.c b/programs/benchzstd.c
similarity index 69%
rename from programs/bench.c
rename to programs/benchzstd.c
index 326c1c1c56e5..0be74a965047 100644
--- a/programs/bench.c
+++ b/programs/benchzstd.c
@@ -9,7 +9,6 @@
  */
 
 
-
 /* **************************************
 *  Tuning parameters
 ****************************************/
@@ -18,30 +17,24 @@
 #endif
 
 
-/* **************************************
-*  Compiler Warnings
-****************************************/
-#ifdef _MSC_VER
-#  pragma warning(disable : 4127)   /* disable: C4127: conditional expression is constant */
-#endif
-
-
 /* *************************************
 *  Includes
 ***************************************/
 #include "platform.h"    /* Large Files support */
 #include "util.h"        /* UTIL_getFileSize, UTIL_sleep */
 #include       /* malloc, free */
-#include       /* memset */
+#include       /* memset, strerror */
 #include        /* fprintf, fopen */
+#include 
 #include       /* assert */
 
+#include "benchfn.h"
 #include "mem.h"
 #define ZSTD_STATIC_LINKING_ONLY
 #include "zstd.h"
 #include "datagen.h"     /* RDG_genBuffer */
 #include "xxhash.h"
-#include "bench.h"
+#include "benchzstd.h"
 #include "zstd_errors.h"
 
 
@@ -102,6 +95,18 @@ static UTIL_time_t g_displayClock = UTIL_TIME_INITIALIZER;
     return errorNum;                                  \
 }
 
+#define CHECK_Z(zf) {              \
+    size_t const zerr = zf;        \
+    if (ZSTD_isError(zerr)) {      \
+        DEBUGOUTPUT("%s: %i: \n", __FILE__, __LINE__);  \
+        DISPLAY("Error : ");       \
+        DISPLAY("%s failed : %s",  \
+                #zf, ZSTD_getErrorName(zerr));   \
+        DISPLAY(" \n");            \
+        exit(1);                   \
+    }                              \
+}
+
 #define RETURN_ERROR(errorNum, retType, ...)  {       \
     retType r;                                        \
     memset(&r, 0, sizeof(retType));                   \
@@ -113,17 +118,6 @@ static UTIL_time_t g_displayClock = UTIL_TIME_INITIALIZER;
     return r;                                         \
 }
 
-/* error without displaying */
-#define RETURN_QUIET_ERROR(errorNum, retType, ...)  { \
-    retType r;                                        \
-    memset(&r, 0, sizeof(retType));                   \
-    DEBUGOUTPUT("%s: %i: \n", __FILE__, __LINE__);    \
-    DEBUGOUTPUT("Error %i : ", errorNum);             \
-    DEBUGOUTPUT(__VA_ARGS__);                         \
-    DEBUGOUTPUT(" \n");                               \
-    r.tag = errorNum;                                 \
-    return r;                                         \
-}
 
 /* *************************************
 *  Benchmark Parameters
@@ -141,7 +135,7 @@ BMK_advancedParams_t BMK_initAdvancedParams(void) {
         0, /* ldmMinMatch */
         0, /* ldmHashLog */
         0, /* ldmBuckSizeLog */
-        0  /* ldmHashEveryLog */
+        0  /* ldmHashRateLog */
     };
     return res;
 }
@@ -168,33 +162,32 @@ typedef struct {
 static void BMK_initCCtx(ZSTD_CCtx* ctx,
     const void* dictBuffer, size_t dictBufferSize, int cLevel,
     const ZSTD_compressionParameters* comprParams, const BMK_advancedParams_t* adv) {
-    ZSTD_CCtx_reset(ctx);
-    ZSTD_CCtx_resetParameters(ctx);
+    ZSTD_CCtx_reset(ctx, ZSTD_reset_session_and_parameters);
     if (adv->nbWorkers==1) {
-        ZSTD_CCtx_setParameter(ctx, ZSTD_p_nbWorkers, 0);
+        CHECK_Z(ZSTD_CCtx_setParameter(ctx, ZSTD_c_nbWorkers, 0));
     } else {
-        ZSTD_CCtx_setParameter(ctx, ZSTD_p_nbWorkers, adv->nbWorkers);
+        CHECK_Z(ZSTD_CCtx_setParameter(ctx, ZSTD_c_nbWorkers, adv->nbWorkers));
     }
-    ZSTD_CCtx_setParameter(ctx, ZSTD_p_compressionLevel, cLevel);
-    ZSTD_CCtx_setParameter(ctx, ZSTD_p_enableLongDistanceMatching, adv->ldmFlag);
-    ZSTD_CCtx_setParameter(ctx, ZSTD_p_ldmMinMatch, adv->ldmMinMatch);
-    ZSTD_CCtx_setParameter(ctx, ZSTD_p_ldmHashLog, adv->ldmHashLog);
-    ZSTD_CCtx_setParameter(ctx, ZSTD_p_ldmBucketSizeLog, adv->ldmBucketSizeLog);
-    ZSTD_CCtx_setParameter(ctx, ZSTD_p_ldmHashEveryLog, adv->ldmHashEveryLog);
-    ZSTD_CCtx_setParameter(ctx, ZSTD_p_windowLog, comprParams->windowLog);
-    ZSTD_CCtx_setParameter(ctx, ZSTD_p_hashLog, comprParams->hashLog);
-    ZSTD_CCtx_setParameter(ctx, ZSTD_p_chainLog, comprParams->chainLog);
-    ZSTD_CCtx_setParameter(ctx, ZSTD_p_searchLog, comprParams->searchLog);
-    ZSTD_CCtx_setParameter(ctx, ZSTD_p_minMatch, comprParams->searchLength);
-    ZSTD_CCtx_setParameter(ctx, ZSTD_p_targetLength, comprParams->targetLength);
-    ZSTD_CCtx_setParameter(ctx, ZSTD_p_compressionStrategy, comprParams->strategy);
-    ZSTD_CCtx_loadDictionary(ctx, dictBuffer, dictBufferSize);
+    CHECK_Z(ZSTD_CCtx_setParameter(ctx, ZSTD_c_compressionLevel, cLevel));
+    CHECK_Z(ZSTD_CCtx_setParameter(ctx, ZSTD_c_enableLongDistanceMatching, adv->ldmFlag));
+    CHECK_Z(ZSTD_CCtx_setParameter(ctx, ZSTD_c_ldmMinMatch, adv->ldmMinMatch));
+    CHECK_Z(ZSTD_CCtx_setParameter(ctx, ZSTD_c_ldmHashLog, adv->ldmHashLog));
+    CHECK_Z(ZSTD_CCtx_setParameter(ctx, ZSTD_c_ldmBucketSizeLog, adv->ldmBucketSizeLog));
+    CHECK_Z(ZSTD_CCtx_setParameter(ctx, ZSTD_c_ldmHashRateLog, adv->ldmHashRateLog));
+    CHECK_Z(ZSTD_CCtx_setParameter(ctx, ZSTD_c_windowLog, comprParams->windowLog));
+    CHECK_Z(ZSTD_CCtx_setParameter(ctx, ZSTD_c_hashLog, comprParams->hashLog));
+    CHECK_Z(ZSTD_CCtx_setParameter(ctx, ZSTD_c_chainLog, comprParams->chainLog));
+    CHECK_Z(ZSTD_CCtx_setParameter(ctx, ZSTD_c_searchLog, comprParams->searchLog));
+    CHECK_Z(ZSTD_CCtx_setParameter(ctx, ZSTD_c_minMatch, comprParams->minMatch));
+    CHECK_Z(ZSTD_CCtx_setParameter(ctx, ZSTD_c_targetLength, comprParams->targetLength));
+    CHECK_Z(ZSTD_CCtx_setParameter(ctx, ZSTD_c_strategy, comprParams->strategy));
+    CHECK_Z(ZSTD_CCtx_loadDictionary(ctx, dictBuffer, dictBufferSize));
 }
 
 static void BMK_initDCtx(ZSTD_DCtx* dctx,
     const void* dictBuffer, size_t dictBufferSize) {
-    ZSTD_DCtx_reset(dctx);
-    ZSTD_DCtx_loadDictionary(dctx, dictBuffer, dictBufferSize);
+    CHECK_Z(ZSTD_DCtx_reset(dctx, ZSTD_reset_session_and_parameters));
+    CHECK_Z(ZSTD_DCtx_loadDictionary(dctx, dictBuffer, dictBufferSize));
 }
 
 
@@ -232,22 +225,8 @@ static size_t local_defaultCompress(
                     void* dstBuffer, size_t dstSize,
                     void* addArgs)
 {
-    size_t moreToFlush = 1;
     ZSTD_CCtx* const cctx = (ZSTD_CCtx*)addArgs;
-    ZSTD_inBuffer in;
-    ZSTD_outBuffer out;
-    in.src = srcBuffer; in.size = srcSize; in.pos = 0;
-    out.dst = dstBuffer; out.size = dstSize; out.pos = 0;
-    while (moreToFlush) {
-        if(out.pos == out.size) {
-            return (size_t)-ZSTD_error_dstSize_tooSmall;
-        }
-        moreToFlush = ZSTD_compress_generic(cctx, &out, &in, ZSTD_e_end);
-        if (ZSTD_isError(moreToFlush)) {
-            return moreToFlush;
-        }
-    }
-    return out.pos;
+    return ZSTD_compress2(cctx, dstBuffer, dstSize, srcBuffer, srcSize);
 }
 
 /* `addArgs` is the context */
@@ -266,7 +245,7 @@ static size_t local_defaultDecompress(
         if(out.pos == out.size) {
             return (size_t)-ZSTD_error_dstSize_tooSmall;
         }
-        moreToFlush = ZSTD_decompress_generic(dctx, &out, &in);
+        moreToFlush = ZSTD_decompressStream(dctx, &out, &in);
         if (ZSTD_isError(moreToFlush)) {
             return moreToFlush;
         }
@@ -276,219 +255,6 @@ static size_t local_defaultDecompress(
 }
 
 
-/*===  Benchmarking an arbitrary function  ===*/
-
-int BMK_isSuccessful_runOutcome(BMK_runOutcome_t outcome)
-{
-    return outcome.tag == 0;
-}
-
-/* warning : this function will stop program execution if outcome is invalid !
- *           check outcome validity first, using BMK_isValid_runResult() */
-BMK_runTime_t BMK_extract_runTime(BMK_runOutcome_t outcome)
-{
-    assert(outcome.tag == 0);
-    return outcome.internal_never_use_directly;
-}
-
-static BMK_runOutcome_t BMK_runOutcome_error(void)
-{
-    BMK_runOutcome_t b;
-    memset(&b, 0, sizeof(b));
-    b.tag = 1;
-    return b;
-}
-
-static BMK_runOutcome_t BMK_setValid_runTime(BMK_runTime_t runTime)
-{
-    BMK_runOutcome_t outcome;
-    outcome.tag = 0;
-    outcome.internal_never_use_directly = runTime;
-    return outcome;
-}
-
-
-/* initFn will be measured once, benchFn will be measured `nbLoops` times */
-/* initFn is optional, provide NULL if none */
-/* benchFn must return size_t field compliant with ZSTD_isError for error valuee */
-/* takes # of blocks and list of size & stuff for each. */
-/* can report result of benchFn for each block into blockResult. */
-/* blockResult is optional, provide NULL if this information is not required */
-/* note : time per loop could be zero if run time < timer resolution */
-BMK_runOutcome_t BMK_benchFunction(
-            BMK_benchFn_t benchFn, void* benchPayload,
-            BMK_initFn_t initFn, void* initPayload,
-            size_t blockCount,
-            const void* const * srcBlockBuffers, const size_t* srcBlockSizes,
-            void* const * dstBlockBuffers, const size_t* dstBlockCapacities,
-            size_t* blockResults,
-            unsigned nbLoops)
-{
-    size_t dstSize = 0;
-
-    if(!nbLoops) {
-        RETURN_QUIET_ERROR(2, BMK_runOutcome_t, "nbLoops must be nonzero ");
-    }
-
-    /* init */
-    {   size_t i;
-        for(i = 0; i < blockCount; i++) {
-            memset(dstBlockBuffers[i], 0xE5, dstBlockCapacities[i]);  /* warm up and erase result buffer */
-        }
-#if 0
-        /* based on testing these seem to lower accuracy of multiple calls of 1 nbLoops vs 1 call of multiple nbLoops
-         * (Makes former slower)
-         */
-        UTIL_sleepMilli(5);  /* give processor time to other processes */
-        UTIL_waitForNextTick();
-#endif
-    }
-
-    /* benchmark */
-    {   UTIL_time_t const clockStart = UTIL_getTime();
-        unsigned loopNb, blockNb;
-        if (initFn != NULL) initFn(initPayload);
-        for (loopNb = 0; loopNb < nbLoops; loopNb++) {
-            for (blockNb = 0; blockNb < blockCount; blockNb++) {
-                size_t const res = benchFn(srcBlockBuffers[blockNb], srcBlockSizes[blockNb],
-                                    dstBlockBuffers[blockNb], dstBlockCapacities[blockNb],
-                                    benchPayload);
-                if(ZSTD_isError(res)) {
-                    RETURN_QUIET_ERROR(2, BMK_runOutcome_t,
-                        "Function benchmark failed on block %u of size %u : %s",
-                        blockNb, (U32)dstBlockCapacities[blockNb], ZSTD_getErrorName(res));
-                } else if (loopNb == 0) {
-                    dstSize += res;
-                    if (blockResults != NULL) blockResults[blockNb] = res;
-            }   }
-        }  /* for (loopNb = 0; loopNb < nbLoops; loopNb++) */
-
-        {   U64 const totalTime = UTIL_clockSpanNano(clockStart);
-            BMK_runTime_t rt;
-            rt.nanoSecPerRun = totalTime / nbLoops;
-            rt.sumOfReturn = dstSize;
-            return BMK_setValid_runTime(rt);
-    }   }
-}
-
-
-/* ====  Benchmarking any function, providing intermediate results  ==== */
-
-struct BMK_timedFnState_s {
-    U64 timeSpent_ns;
-    U64 timeBudget_ns;
-    U64 runBudget_ns;
-    BMK_runTime_t fastestRun;
-    unsigned nbLoops;
-    UTIL_time_t coolTime;
-};  /* typedef'd to BMK_timedFnState_t within bench.h */
-
-BMK_timedFnState_t* BMK_createTimedFnState(unsigned total_ms, unsigned run_ms)
-{
-    BMK_timedFnState_t* const r = (BMK_timedFnState_t*)malloc(sizeof(*r));
-    if (r == NULL) return NULL;   /* malloc() error */
-    BMK_resetTimedFnState(r, total_ms, run_ms);
-    return r;
-}
-
-void BMK_freeTimedFnState(BMK_timedFnState_t* state) {
-    free(state);
-}
-
-void BMK_resetTimedFnState(BMK_timedFnState_t* timedFnState, unsigned total_ms, unsigned run_ms)
-{
-    if (!total_ms) total_ms = 1 ;
-    if (!run_ms) run_ms = 1;
-    if (run_ms > total_ms) run_ms = total_ms;
-    timedFnState->timeSpent_ns = 0;
-    timedFnState->timeBudget_ns = (U64)total_ms * TIMELOOP_NANOSEC / 1000;
-    timedFnState->runBudget_ns = (U64)run_ms * TIMELOOP_NANOSEC / 1000;
-    timedFnState->fastestRun.nanoSecPerRun = (U64)(-1LL);
-    timedFnState->fastestRun.sumOfReturn = (size_t)(-1LL);
-    timedFnState->nbLoops = 1;
-    timedFnState->coolTime = UTIL_getTime();
-}
-
-/* Tells if nb of seconds set in timedFnState for all runs is spent.
- * note : this function will return 1 if BMK_benchFunctionTimed() has actually errored. */
-int BMK_isCompleted_TimedFn(const BMK_timedFnState_t* timedFnState)
-{
-    return (timedFnState->timeSpent_ns >= timedFnState->timeBudget_ns);
-}
-
-
-#define MINUSABLETIME  (TIMELOOP_NANOSEC / 2)  /* 0.5 seconds */
-
-BMK_runOutcome_t BMK_benchTimedFn(
-            BMK_timedFnState_t* cont,
-            BMK_benchFn_t benchFn, void* benchPayload,
-            BMK_initFn_t initFn, void* initPayload,
-            size_t blockCount,
-            const void* const* srcBlockBuffers, const size_t* srcBlockSizes,
-            void * const * dstBlockBuffers, const size_t * dstBlockCapacities,
-            size_t* blockResults)
-{
-    U64 const runBudget_ns = cont->runBudget_ns;
-    U64 const runTimeMin_ns = runBudget_ns / 2;
-    int completed = 0;
-    BMK_runTime_t bestRunTime = cont->fastestRun;
-
-    while (!completed) {
-        BMK_runOutcome_t runResult;
-
-        /* Overheat protection */
-        if (UTIL_clockSpanMicro(cont->coolTime) > ACTIVEPERIOD_MICROSEC) {
-            DEBUGOUTPUT("\rcooling down ...    \r");
-            UTIL_sleep(COOLPERIOD_SEC);
-            cont->coolTime = UTIL_getTime();
-        }
-
-        /* reinitialize capacity */
-        runResult = BMK_benchFunction(benchFn, benchPayload,
-                                    initFn, initPayload,
-                                    blockCount,
-                                    srcBlockBuffers, srcBlockSizes,
-                                    dstBlockBuffers, dstBlockCapacities,
-                                    blockResults,
-                                    cont->nbLoops);
-
-        if(!BMK_isSuccessful_runOutcome(runResult)) { /* error : move out */
-            return BMK_runOutcome_error();
-        }
-
-        {   BMK_runTime_t const newRunTime = BMK_extract_runTime(runResult);
-            U64 const loopDuration_ns = newRunTime.nanoSecPerRun * cont->nbLoops;
-
-            cont->timeSpent_ns += loopDuration_ns;
-
-            /* estimate nbLoops for next run to last approximately 1 second */
-            if (loopDuration_ns > (runBudget_ns / 50)) {
-                U64 const fastestRun_ns = MIN(bestRunTime.nanoSecPerRun, newRunTime.nanoSecPerRun);
-                cont->nbLoops = (U32)(runBudget_ns / fastestRun_ns) + 1;
-            } else {
-                /* previous run was too short : blindly increase workload by x multiplier */
-                const unsigned multiplier = 10;
-                assert(cont->nbLoops < ((unsigned)-1) / multiplier);  /* avoid overflow */
-                cont->nbLoops *= multiplier;
-            }
-
-            if(loopDuration_ns < runTimeMin_ns) {
-                /* don't report results for which benchmark run time was too small : increased risks of rounding errors */
-                assert(completed == 0);
-                continue;
-            } else {
-                if(newRunTime.nanoSecPerRun < bestRunTime.nanoSecPerRun) {
-                    bestRunTime = newRunTime;
-                }
-                completed = 1;
-            }
-        }
-    }   /* while (!completed) */
-
-    return BMK_setValid_runTime(bestRunTime);
-}
-
-
 /* ================================================================= */
 /*      Benchmark Zstandard, mem-to-mem scenarios                    */
 /* ================================================================= */
@@ -522,22 +288,24 @@ static BMK_benchOutcome_t BMK_benchOutcome_setValidResult(BMK_benchResult_t resu
 
 
 /* benchMem with no allocation */
-static BMK_benchOutcome_t BMK_benchMemAdvancedNoAlloc(
-            const void** srcPtrs, size_t* srcSizes,
-            void** cPtrs, size_t* cCapacities, size_t* cSizes,
-            void** resPtrs, size_t* resSizes,
-            void** resultBufferPtr, void* compressedBuffer,
-            size_t maxCompressedSize,
-            BMK_timedFnState_t* timeStateCompress,
-            BMK_timedFnState_t* timeStateDecompress,
+static BMK_benchOutcome_t
+BMK_benchMemAdvancedNoAlloc(
+                    const void** srcPtrs, size_t* srcSizes,
+                    void** cPtrs, size_t* cCapacities, size_t* cSizes,
+                    void** resPtrs, size_t* resSizes,
+                    void** resultBufferPtr, void* compressedBuffer,
+                    size_t maxCompressedSize,
+                    BMK_timedFnState_t* timeStateCompress,
+                    BMK_timedFnState_t* timeStateDecompress,
 
-            const void* srcBuffer, size_t srcSize,
-            const size_t* fileSizes, unsigned nbFiles,
-            const int cLevel, const ZSTD_compressionParameters* comprParams,
-            const void* dictBuffer, size_t dictBufferSize,
-            ZSTD_CCtx* cctx, ZSTD_DCtx* dctx,
-            int displayLevel, const char* displayName,
-            const BMK_advancedParams_t* adv)
+                    const void* srcBuffer, size_t srcSize,
+                    const size_t* fileSizes, unsigned nbFiles,
+                    const int cLevel,
+                    const ZSTD_compressionParameters* comprParams,
+                    const void* dictBuffer, size_t dictBufferSize,
+                    ZSTD_CCtx* cctx, ZSTD_DCtx* dctx,
+                    int displayLevel, const char* displayName,
+                    const BMK_advancedParams_t* adv)
 {
     size_t const blockSize = ((adv->blockSize>=32 && (adv->mode != BMK_decodeOnly)) ? adv->blockSize : srcSize) + (!srcSize);  /* avoid div by 0 */
     BMK_benchResult_t benchResult;
@@ -599,6 +367,11 @@ static BMK_benchOutcome_t BMK_benchMemAdvancedNoAlloc(
                 cPtr += cCapacities[nbBlocks];
                 resPtr += thisBlockSize;
                 remaining -= thisBlockSize;
+                if (adv->mode == BMK_decodeOnly) {
+                    assert(nbBlocks==0);
+                    cSizes[nbBlocks] = thisBlockSize;
+                    benchResult.cSize = thisBlockSize;
+                }
             }
         }
     }
@@ -617,32 +390,51 @@ static BMK_benchOutcome_t BMK_benchMemAdvancedNoAlloc(
         U32 markNb = 0;
         int compressionCompleted = (adv->mode == BMK_decodeOnly);
         int decompressionCompleted = (adv->mode == BMK_compressOnly);
+        BMK_benchParams_t cbp, dbp;
         BMK_initCCtxArgs cctxprep;
         BMK_initDCtxArgs dctxprep;
+
+        cbp.benchFn = local_defaultCompress;
+        cbp.benchPayload = cctx;
+        cbp.initFn = local_initCCtx;
+        cbp.initPayload = &cctxprep;
+        cbp.errorFn = ZSTD_isError;
+        cbp.blockCount = nbBlocks;
+        cbp.srcBuffers = srcPtrs;
+        cbp.srcSizes = srcSizes;
+        cbp.dstBuffers = cPtrs;
+        cbp.dstCapacities = cCapacities;
+        cbp.blockResults = cSizes;
+
         cctxprep.cctx = cctx;
         cctxprep.dictBuffer = dictBuffer;
         cctxprep.dictBufferSize = dictBufferSize;
         cctxprep.cLevel = cLevel;
         cctxprep.comprParams = comprParams;
         cctxprep.adv = adv;
+
+        dbp.benchFn = local_defaultDecompress;
+        dbp.benchPayload = dctx;
+        dbp.initFn = local_initDCtx;
+        dbp.initPayload = &dctxprep;
+        dbp.errorFn = ZSTD_isError;
+        dbp.blockCount = nbBlocks;
+        dbp.srcBuffers = (const void* const *) cPtrs;
+        dbp.srcSizes = cSizes;
+        dbp.dstBuffers = resPtrs;
+        dbp.dstCapacities = resSizes;
+        dbp.blockResults = NULL;
+
         dctxprep.dctx = dctx;
         dctxprep.dictBuffer = dictBuffer;
         dctxprep.dictBufferSize = dictBufferSize;
 
         DISPLAYLEVEL(2, "\r%70s\r", "");   /* blank line */
-        DISPLAYLEVEL(2, "%2s-%-17.17s :%10u ->\r", marks[markNb], displayName, (U32)srcSize);
+        DISPLAYLEVEL(2, "%2s-%-17.17s :%10u ->\r", marks[markNb], displayName, (unsigned)srcSize);
 
         while (!(compressionCompleted && decompressionCompleted)) {
-
             if (!compressionCompleted) {
-                BMK_runOutcome_t const cOutcome =
-                        BMK_benchTimedFn( timeStateCompress,
-                                        &local_defaultCompress, cctx,
-                                        &local_initCCtx, &cctxprep,
-                                        nbBlocks,
-                                        srcPtrs, srcSizes,
-                                        cPtrs, cCapacities,
-                                        cSizes);
+                BMK_runOutcome_t const cOutcome = BMK_benchTimedFn( timeStateCompress, cbp);
 
                 if (!BMK_isSuccessful_runOutcome(cOutcome)) {
                     return BMK_benchOutcome_error();
@@ -659,10 +451,9 @@ static BMK_benchOutcome_t BMK_benchMemAdvancedNoAlloc(
                 }   }
 
                 {   int const ratioAccuracy = (ratio < 10.) ? 3 : 2;
-                    markNb = (markNb+1) % NB_MARKS;
                     DISPLAYLEVEL(2, "%2s-%-17.17s :%10u ->%10u (%5.*f),%6.*f MB/s\r",
                             marks[markNb], displayName,
-                            (U32)srcSize, (U32)cSize,
+                            (unsigned)srcSize, (unsigned)cSize,
                             ratioAccuracy, ratio,
                             benchResult.cSpeed < (10 MB) ? 2 : 1, (double)benchResult.cSpeed / MB_UNIT);
                 }
@@ -670,14 +461,7 @@ static BMK_benchOutcome_t BMK_benchMemAdvancedNoAlloc(
             }
 
             if(!decompressionCompleted) {
-                BMK_runOutcome_t const dOutcome =
-                        BMK_benchTimedFn(timeStateDecompress,
-                                        &local_defaultDecompress, dctx,
-                                        &local_initDCtx, &dctxprep,
-                                        nbBlocks,
-                                        (const void *const *)cPtrs, cSizes,
-                                        resPtrs, resSizes,
-                                        NULL);
+                BMK_runOutcome_t const dOutcome = BMK_benchTimedFn(timeStateDecompress, dbp);
 
                 if(!BMK_isSuccessful_runOutcome(dOutcome)) {
                     return BMK_benchOutcome_error();
@@ -690,16 +474,16 @@ static BMK_benchOutcome_t BMK_benchMemAdvancedNoAlloc(
                 }
 
                 {   int const ratioAccuracy = (ratio < 10.) ? 3 : 2;
-                    markNb = (markNb+1) % NB_MARKS;
                     DISPLAYLEVEL(2, "%2s-%-17.17s :%10u ->%10u (%5.*f),%6.*f MB/s ,%6.1f MB/s \r",
                             marks[markNb], displayName,
-                            (U32)srcSize, (U32)benchResult.cSize,
+                            (unsigned)srcSize, (unsigned)benchResult.cSize,
                             ratioAccuracy, ratio,
                             benchResult.cSpeed < (10 MB) ? 2 : 1, (double)benchResult.cSpeed / MB_UNIT,
                             (double)benchResult.dSpeed / MB_UNIT);
                 }
                 decompressionCompleted = BMK_isCompleted_TimedFn(timeStateDecompress);
             }
+            markNb = (markNb+1) % NB_MARKS;
         }   /* while (!(compressionCompleted && decompressionCompleted)) */
 
         /* CRC Checking */
@@ -707,12 +491,13 @@ static BMK_benchOutcome_t BMK_benchMemAdvancedNoAlloc(
             U64 const crcCheck = XXH64(resultBuffer, srcSize, 0);
             if ((adv->mode == BMK_both) && (crcOrig!=crcCheck)) {
                 size_t u;
-                DISPLAY("!!! WARNING !!! %14s : Invalid Checksum : %x != %x   \n", displayName, (unsigned)crcOrig, (unsigned)crcCheck);
+                DISPLAY("!!! WARNING !!! %14s : Invalid Checksum : %x != %x   \n",
+                        displayName, (unsigned)crcOrig, (unsigned)crcCheck);
                 for (u=0; u u) break;
                             bacc += srcSizes[segNb];
@@ -883,7 +668,7 @@ static BMK_benchOutcome_t BMK_benchCLevel(const void* srcBuffer, size_t benchedS
     if (displayLevel == 1 && !adv->additionalParam)   /* --quiet mode */
         DISPLAY("bench %s %s: input %u bytes, %u seconds, %u KB blocks\n",
                 ZSTD_VERSION_STRING, ZSTD_GIT_COMMIT_STRING,
-                (U32)benchedSize, adv->nbSeconds, (U32)(adv->blockSize>>10));
+                (unsigned)benchedSize, adv->nbSeconds, (unsigned)(adv->blockSize>>10));
 
     return BMK_benchMemAdvanced(srcBuffer, benchedSize,
                                 NULL, 0,
@@ -1015,6 +800,11 @@ BMK_benchOutcome_t BMK_benchFilesAdvanced(
     /* Load dictionary */
     if (dictFileName != NULL) {
         U64 const dictFileSize = UTIL_getFileSize(dictFileName);
+        if (dictFileSize == UTIL_FILESIZE_UNKNOWN) {
+            DISPLAYLEVEL(1, "error loading %s : %s \n", dictFileName, strerror(errno));
+            free(fileSizes);
+            RETURN_ERROR(9, BMK_benchOutcome_t, "benchmark aborted");
+        }
         if (dictFileSize > 64 MB) {
             free(fileSizes);
             RETURN_ERROR(10, BMK_benchOutcome_t, "dictionary file %s too large", dictFileName);
@@ -1024,7 +814,7 @@ BMK_benchOutcome_t BMK_benchFilesAdvanced(
         if (dictBuffer==NULL) {
             free(fileSizes);
             RETURN_ERROR(11, BMK_benchOutcome_t, "not enough memory for dictionary (%u bytes)",
-                            (U32)dictBufferSize);
+                            (unsigned)dictBufferSize);
         }
 
         {   int const errorCode = BMK_loadFiles(dictBuffer, dictBufferSize,
@@ -1040,7 +830,7 @@ BMK_benchOutcome_t BMK_benchFilesAdvanced(
     benchedSize = BMK_findMaxMem(totalSizeToLoad * 3) / 3;
     if ((U64)benchedSize > totalSizeToLoad) benchedSize = (size_t)totalSizeToLoad;
     if (benchedSize < totalSizeToLoad)
-        DISPLAY("Not enough memory; testing %u MB only...\n", (U32)(benchedSize >> 20));
+        DISPLAY("Not enough memory; testing %u MB only...\n", (unsigned)(benchedSize >> 20));
 
     srcBuffer = benchedSize ? malloc(benchedSize) : NULL;
     if (!srcBuffer) {
diff --git a/programs/bench.h b/programs/benchzstd.h
similarity index 61%
rename from programs/bench.h
rename to programs/benchzstd.h
index 13ca5b50b461..3a8b893e7ee6 100644
--- a/programs/bench.h
+++ b/programs/benchzstd.h
@@ -8,12 +8,18 @@
  * You may select, at your option, one of the above-listed licenses.
  */
 
+ /* benchzstd :
+  * benchmark Zstandard compression / decompression
+  * over a set of files or buffers
+  * and display progress result and final summary
+  */
+
 #if defined (__cplusplus)
 extern "C" {
 #endif
 
-#ifndef BENCH_H_121279284357
-#define BENCH_H_121279284357
+#ifndef BENCH_ZSTD_H_3242387
+#define BENCH_ZSTD_H_3242387
 
 /* ===  Dependencies  === */
 #include    /* size_t */
@@ -109,7 +115,7 @@ typedef struct {
     unsigned ldmMinMatch;       /* below: parameters for long distance matching, see zstd.1.md */
     unsigned ldmHashLog;
     unsigned ldmBucketSizeLog;
-    unsigned ldmHashEveryLog;
+    unsigned ldmHashRateLog;
 } BMK_advancedParams_t;
 
 /* returns default parameters used by nonAdvanced functions */
@@ -142,9 +148,9 @@ BMK_benchOutcome_t BMK_benchFilesAdvanced(
  *          .cMem  : memory budget required for the compression context
  */
 BMK_benchOutcome_t BMK_syntheticTest(
-                              int cLevel, double compressibility,
-                              const ZSTD_compressionParameters* compressionParams,
-                              int displayLevel, const BMK_advancedParams_t* adv);
+                          int cLevel, double compressibility,
+                          const ZSTD_compressionParameters* compressionParams,
+                          int displayLevel, const BMK_advancedParams_t* adv);
 
 
 
@@ -181,6 +187,7 @@ BMK_benchOutcome_t BMK_benchMem(const void* srcBuffer, size_t srcSize,
                         const void* dictBuffer, size_t dictBufferSize,
                         int displayLevel, const char* displayName);
 
+
 /* BMK_benchMemAdvanced() : same as BMK_benchMem()
  * with following additional options :
  * dstBuffer - destination buffer to write compressed output in, NULL if none provided.
@@ -197,106 +204,8 @@ BMK_benchOutcome_t BMK_benchMemAdvanced(const void* srcBuffer, size_t srcSize,
 
 
 
-/* ====  Benchmarking any function, iterated on a set of blocks  ==== */
 
-typedef struct {
-    unsigned long long nanoSecPerRun;  /* time per iteration */
-    size_t sumOfReturn;       /* sum of return values */
-} BMK_runTime_t;
-
-VARIANT_ERROR_RESULT(BMK_runTime_t, BMK_runOutcome_t);
-
-/* check first if the return structure represents an error or a valid result */
-int BMK_isSuccessful_runOutcome(BMK_runOutcome_t outcome);
-
-/* extract result from variant type.
- * note : this function will abort() program execution if result is not valid
- *        check result validity first, by using BMK_isSuccessful_runOutcome()
- */
-BMK_runTime_t BMK_extract_runTime(BMK_runOutcome_t outcome);
-
-
-
-typedef size_t (*BMK_benchFn_t)(const void* src, size_t srcSize, void* dst, size_t dstCapacity, void* customPayload);
-typedef size_t (*BMK_initFn_t)(void* initPayload);
-
-
-/* BMK_benchFunction() :
- * This function times the execution of 2 argument functions, benchFn and initFn  */
-
-/* benchFn - (*benchFn)(srcBuffers[i], srcSizes[i], dstBuffers[i], dstCapacities[i], benchPayload)
- *      is run nbLoops times
- * initFn - (*initFn)(initPayload) is run once per benchmark, at the beginning.
- *      This argument can be NULL, in which case nothing is run.
- * blockCount - number of blocks. Size of all array parameters : srcBuffers, srcSizes, dstBuffers, dstCapacities, blockResults
- * srcBuffers - an array of buffers to be operated on by benchFn
- * srcSizes - an array of the sizes of above buffers
- * dstBuffers - an array of buffers to be written into by benchFn
- * dstCapacities - an array of the capacities of above buffers
- * blockResults - Optional: store the return value of benchFn for each block. Use NULL if this result is not requested.
- * nbLoops - defines number of times benchFn is run.
- * @return: a variant, which express either an error, or can generate a valid BMK_runTime_t result.
- *          Use BMK_isSuccessful_runOutcome() to check if function was successful.
- *          If yes, extract the result with BMK_extract_runTime(),
- *          it will contain :
- *              .sumOfReturn : the sum of all return values of benchFn through all of blocks
- *              .nanoSecPerRun : time per run of benchFn + (time for initFn / nbLoops)
- *          .sumOfReturn is generally intended for functions which return a # of bytes written into dstBuffer,
- *              in which case, this value will be the total amount of bytes written into dstBuffer.
- */
-BMK_runOutcome_t BMK_benchFunction(
-                        BMK_benchFn_t benchFn, void* benchPayload,
-                        BMK_initFn_t initFn, void* initPayload,
-                        size_t blockCount,
-                        const void *const * srcBuffers, const size_t* srcSizes,
-                        void *const * dstBuffers, const size_t* dstCapacities,
-                        size_t* blockResults,
-                        unsigned nbLoops);
-
-
-
-/* ====  Benchmark any function, providing intermediate results  ==== */
-
-/* state information tracking benchmark session */
-typedef struct BMK_timedFnState_s BMK_timedFnState_t;
-
-/* BMK_createTimedFnState() and BMK_resetTimedFnState() :
- * Create/Set BMK_timedFnState_t for next benchmark session,
- * which shall last a minimum of total_ms milliseconds,
- * producing intermediate results, paced at interval of (approximately) run_ms.
- */
-BMK_timedFnState_t* BMK_createTimedFnState(unsigned total_ms, unsigned run_ms);
-void BMK_resetTimedFnState(BMK_timedFnState_t* timedFnState, unsigned total_ms, unsigned run_ms);
-void BMK_freeTimedFnState(BMK_timedFnState_t* state);
-
-
-/* Tells if duration of all benchmark runs has exceeded total_ms
- */
-int BMK_isCompleted_TimedFn(const BMK_timedFnState_t* timedFnState);
-
-
-/* BMK_benchTimedFn() :
- * Similar to BMK_benchFunction(), most arguments being identical.
- * Automatically determines `nbLoops` so that each result is regularly produced at interval of about run_ms.
- * Note : minimum `nbLoops` is 1, therefore a run may last more than run_ms, and possibly even more than total_ms.
- * Usage - initialize timedFnState, select benchmark duration (total_ms) and each measurement duration (run_ms)
- *         call BMK_benchTimedFn() repetitively, each measurement is supposed to last about run_ms
- *         Check if total time budget is spent or exceeded, using BMK_isCompleted_TimedFn()
- */
-BMK_runOutcome_t BMK_benchTimedFn(
-                    BMK_timedFnState_t* timedFnState,
-                    BMK_benchFn_t benchFn, void* benchPayload,
-                    BMK_initFn_t initFn, void* initPayload,
-                    size_t blockCount,
-                    const void *const * srcBlockBuffers, const size_t* srcBlockSizes,
-                    void *const * dstBlockBuffers, const size_t* dstBlockCapacities,
-                    size_t* blockResults);
-
-
-
-
-
-#endif   /* BENCH_H_121279284357 */
+#endif   /* BENCH_ZSTD_H_3242387 */
 
 #if defined (__cplusplus)
 }
diff --git a/programs/datagen.c b/programs/datagen.c
index c8383658488b..026b9a1855da 100644
--- a/programs/datagen.c
+++ b/programs/datagen.c
@@ -81,18 +81,18 @@ static BYTE RDG_genChar(U32* seed, const BYTE* ldt)
 }
 
 
-static U32 RDG_rand15Bits (unsigned* seedPtr)
+static U32 RDG_rand15Bits (U32* seedPtr)
 {
     return RDG_rand(seedPtr) & 0x7FFF;
 }
 
-static U32 RDG_randLength(unsigned* seedPtr)
+static U32 RDG_randLength(U32* seedPtr)
 {
     if (RDG_rand(seedPtr) & 7) return (RDG_rand(seedPtr) & 0xF);   /* small length */
     return (RDG_rand(seedPtr) & 0x1FF) + 0xF;
 }
 
-static void RDG_genBlock(void* buffer, size_t buffSize, size_t prefixSize, double matchProba, const BYTE* ldt, unsigned* seedPtr)
+static void RDG_genBlock(void* buffer, size_t buffSize, size_t prefixSize, double matchProba, const BYTE* ldt, U32* seedPtr)
 {
     BYTE* const buffPtr = (BYTE*)buffer;
     U32 const matchProba32 = (U32)(32768 * matchProba);
@@ -141,16 +141,18 @@ static void RDG_genBlock(void* buffer, size_t buffSize, size_t prefixSize, doubl
 
 void RDG_genBuffer(void* buffer, size_t size, double matchProba, double litProba, unsigned seed)
 {
+    U32 seed32 = seed;
     BYTE ldt[LTSIZE];
     memset(ldt, '0', sizeof(ldt));  /* yes, character '0', this is intentional */
     if (litProba<=0.0) litProba = matchProba / 4.5;
     RDG_fillLiteralDistrib(ldt, litProba);
-    RDG_genBlock(buffer, size, 0, matchProba, ldt, &seed);
+    RDG_genBlock(buffer, size, 0, matchProba, ldt, &seed32);
 }
 
 
 void RDG_genStdout(unsigned long long size, double matchProba, double litProba, unsigned seed)
 {
+    U32 seed32 = seed;
     size_t const stdBlockSize = 128 KB;
     size_t const stdDictSize = 32 KB;
     BYTE* const buff = (BYTE*)malloc(stdDictSize + stdBlockSize);
@@ -165,12 +167,12 @@ void RDG_genStdout(unsigned long long size, double matchProba, double litProba,
     SET_BINARY_MODE(stdout);
 
     /* Generate initial dict */
-    RDG_genBlock(buff, stdDictSize, 0, matchProba, ldt, &seed);
+    RDG_genBlock(buff, stdDictSize, 0, matchProba, ldt, &seed32);
 
     /* Generate compressible data */
     while (total < size) {
         size_t const genBlockSize = (size_t) (MIN (stdBlockSize, size-total));
-        RDG_genBlock(buff, stdDictSize+stdBlockSize, stdDictSize, matchProba, ldt, &seed);
+        RDG_genBlock(buff, stdDictSize+stdBlockSize, stdDictSize, matchProba, ldt, &seed32);
         total += genBlockSize;
         { size_t const unused = fwrite(buff, 1, genBlockSize, stdout); (void)unused; }
         /* update dict */
diff --git a/programs/dibio.c b/programs/dibio.c
index d3fd8cc053de..c9d214e754e6 100644
--- a/programs/dibio.c
+++ b/programs/dibio.c
@@ -139,7 +139,7 @@ static unsigned DiB_loadFiles(void* buffer, size_t* bufferSizePtr,
     }
     DISPLAYLEVEL(2, "\r%79s\r", "");
     *bufferSizePtr = pos;
-    DISPLAYLEVEL(4, "loaded : %u KB \n", (U32)(pos >> 10))
+    DISPLAYLEVEL(4, "loaded : %u KB \n", (unsigned)(pos >> 10))
     return nbLoadedChunks;
 }
 
@@ -249,7 +249,7 @@ static fileStats DiB_fileStats(const char** fileNamesTable, unsigned nbFiles, si
         fs.oneSampleTooLarge |= (chunkSize > 2*SAMPLESIZE_MAX);
         fs.nbSamples += nbSamples;
     }
-    DISPLAYLEVEL(4, "Preparing to load : %u KB \n", (U32)(fs.totalSizeToLoad >> 10));
+    DISPLAYLEVEL(4, "Preparing to load : %u KB \n", (unsigned)(fs.totalSizeToLoad >> 10));
     return fs;
 }
 
@@ -358,7 +358,7 @@ int DiB_trainFromFiles(const char* dictFileName, unsigned maxDictSize,
             goto _cleanup;
         }
         /* save dict */
-        DISPLAYLEVEL(2, "Save dictionary of size %u into file %s \n", (U32)dictSize, dictFileName);
+        DISPLAYLEVEL(2, "Save dictionary of size %u into file %s \n", (unsigned)dictSize, dictFileName);
         DiB_saveDict(dictFileName, dictBuffer, dictSize);
     }
 
diff --git a/programs/fileio.c b/programs/fileio.c
index c24f4defbb9a..9fb795ed3f27 100644
--- a/programs/fileio.c
+++ b/programs/fileio.c
@@ -88,10 +88,10 @@ void FIO_setNotificationLevel(unsigned level) { g_displayLevel=level; }
 static const U64 g_refreshRate = SEC_TO_MICRO / 6;
 static UTIL_time_t g_displayClock = UTIL_TIME_INITIALIZER;
 
-#define READY_FOR_UPDATE() (UTIL_clockSpanMicro(g_displayClock) > g_refreshRate)
+#define READY_FOR_UPDATE() (!g_noProgress && UTIL_clockSpanMicro(g_displayClock) > g_refreshRate)
 #define DELAY_NEXT_UPDATE() { g_displayClock = UTIL_getTime(); }
 #define DISPLAYUPDATE(l, ...) {                              \
-        if (g_displayLevel>=l) {                             \
+        if (g_displayLevel>=l && !g_noProgress) {            \
             if (READY_FOR_UPDATE() || (g_displayLevel>=4)) { \
                 DELAY_NEXT_UPDATE();                         \
                 DISPLAY(__VA_ARGS__);                        \
@@ -270,18 +270,18 @@ void FIO_addAbortHandler()
 static FIO_compressionType_t g_compressionType = FIO_zstdCompression;
 void FIO_setCompressionType(FIO_compressionType_t compressionType) { g_compressionType = compressionType; }
 static U32 g_overwrite = 0;
-void FIO_overwriteMode(void) { g_overwrite=1; }
-static U32 g_sparseFileSupport = 1;   /* 0: no sparse allowed; 1: auto (file yes, stdout no); 2: force sparse */
-void FIO_setSparseWrite(unsigned sparse) { g_sparseFileSupport=sparse; }
+void FIO_overwriteMode(void) { g_overwrite = 1; }
+static U32 g_sparseFileSupport = ZSTD_SPARSE_DEFAULT;   /* 0: no sparse allowed; 1: auto (file yes, stdout no); 2: force sparse */
+void FIO_setSparseWrite(unsigned sparse) { g_sparseFileSupport = sparse; }
 static U32 g_dictIDFlag = 1;
 void FIO_setDictIDFlag(unsigned dictIDFlag) { g_dictIDFlag = dictIDFlag; }
 static U32 g_checksumFlag = 1;
 void FIO_setChecksumFlag(unsigned checksumFlag) { g_checksumFlag = checksumFlag; }
 static U32 g_removeSrcFile = 0;
 void FIO_setRemoveSrcFile(unsigned flag) { g_removeSrcFile = (flag>0); }
-static U32 g_memLimit = 0;
+static unsigned g_memLimit = 0;
 void FIO_setMemLimit(unsigned memLimit) { g_memLimit = memLimit; }
-static U32 g_nbWorkers = 1;
+static unsigned g_nbWorkers = 1;
 void FIO_setNbWorkers(unsigned nbWorkers) {
 #ifndef ZSTD_MULTITHREAD
     if (nbWorkers > 0) DISPLAYLEVEL(2, "Note : multi-threading is disabled \n");
@@ -295,7 +295,7 @@ void FIO_setBlockSize(unsigned blockSize) {
     g_blockSize = blockSize;
 }
 #define FIO_OVERLAP_LOG_NOTSET 9999
-static U32 g_overlapLog = FIO_OVERLAP_LOG_NOTSET;
+static unsigned g_overlapLog = FIO_OVERLAP_LOG_NOTSET;
 void FIO_setOverlapLog(unsigned overlapLog){
     if (overlapLog && g_nbWorkers==0)
         DISPLAYLEVEL(2, "Setting overlapLog is useless in single-thread mode \n");
@@ -307,6 +307,12 @@ void FIO_setAdaptiveMode(unsigned adapt) {
         EXM_THROW(1, "Adaptive mode is not compatible with single thread mode \n");
     g_adaptiveMode = adapt;
 }
+static U32 g_rsyncable = 0;
+void FIO_setRsyncable(unsigned rsyncable) {
+    if ((rsyncable>0) && (g_nbWorkers==0))
+        EXM_THROW(1, "Rsyncable mode is not compatible with single thread mode \n");
+    g_rsyncable = rsyncable;
+}
 static int g_minAdaptLevel = -50;   /* initializing this value requires a constant, so ZSTD_minCLevel() doesn't work */
 void FIO_setAdaptMin(int minCLevel)
 {
@@ -340,9 +346,13 @@ void FIO_setLdmBucketSizeLog(unsigned ldmBucketSizeLog) {
     g_ldmBucketSizeLog = ldmBucketSizeLog;
 }
 
-static U32 g_ldmHashEveryLog = FIO_LDM_PARAM_NOTSET;
-void FIO_setLdmHashEveryLog(unsigned ldmHashEveryLog) {
-    g_ldmHashEveryLog = ldmHashEveryLog;
+static U32 g_ldmHashRateLog = FIO_LDM_PARAM_NOTSET;
+void FIO_setLdmHashRateLog(unsigned ldmHashRateLog) {
+    g_ldmHashRateLog = ldmHashRateLog;
+}
+static U32 g_noProgress = 0;
+void FIO_setNoProgress(unsigned noProgress) {
+    g_noProgress = noProgress;
 }
 
 
@@ -355,7 +365,7 @@ void FIO_setLdmHashEveryLog(unsigned ldmHashEveryLog) {
 static int FIO_remove(const char* path)
 {
     if (!UTIL_isRegularFile(path)) {
-        DISPLAYLEVEL(2, "zstd: Refusing to remove non-regular file %s\n", path);
+        DISPLAYLEVEL(2, "zstd: Refusing to remove non-regular file %s \n", path);
         return 0;
     }
 #if defined(_WIN32) || defined(WIN32)
@@ -373,11 +383,17 @@ static FILE* FIO_openSrcFile(const char* srcFileName)
 {
     assert(srcFileName != NULL);
     if (!strcmp (srcFileName, stdinmark)) {
-        DISPLAYLEVEL(4,"Using stdin for input\n");
+        DISPLAYLEVEL(4,"Using stdin for input \n");
         SET_BINARY_MODE(stdin);
         return stdin;
     }
 
+    if (!UTIL_fileExist(srcFileName)) {
+        DISPLAYLEVEL(1, "zstd: can't stat %s : %s -- ignored \n",
+                        srcFileName, strerror(errno));
+        return NULL;
+    }
+
     if (!UTIL_isRegularFile(srcFileName)) {
         DISPLAYLEVEL(1, "zstd: %s is not a regular file -- ignored \n",
                         srcFileName);
@@ -394,30 +410,54 @@ static FILE* FIO_openSrcFile(const char* srcFileName)
 /** FIO_openDstFile() :
  *  condition : `dstFileName` must be non-NULL.
  * @result : FILE* to `dstFileName`, or NULL if it fails */
-static FILE* FIO_openDstFile(const char* dstFileName)
+static FILE* FIO_openDstFile(const char* srcFileName, const char* dstFileName)
 {
     assert(dstFileName != NULL);
     if (!strcmp (dstFileName, stdoutmark)) {
-        DISPLAYLEVEL(4,"Using stdout for output\n");
+        DISPLAYLEVEL(4,"Using stdout for output \n");
         SET_BINARY_MODE(stdout);
-        if (g_sparseFileSupport==1) {
+        if (g_sparseFileSupport == 1) {
             g_sparseFileSupport = 0;
             DISPLAYLEVEL(4, "Sparse File Support is automatically disabled on stdout ; try --sparse \n");
         }
         return stdout;
     }
 
+    /* ensure dst is not the same file as src */
+    if (srcFileName != NULL) {
+#ifdef _MSC_VER
+        /* note : Visual does not support file identification by inode.
+         *        The following work-around is limited to detecting exact name repetition only,
+         *        aka `filename` is considered different from `subdir/../filename` */
+        if (!strcmp(srcFileName, dstFileName)) {
+            DISPLAYLEVEL(1, "zstd: Refusing to open a output file which will overwrite the input file \n");
+            return NULL;
+        }
+#else
+        stat_t srcStat;
+        stat_t dstStat;
+        if (UTIL_getFileStat(srcFileName, &srcStat)
+            && UTIL_getFileStat(dstFileName, &dstStat)) {
+            if (srcStat.st_dev == dstStat.st_dev
+                && srcStat.st_ino == dstStat.st_ino) {
+                DISPLAYLEVEL(1, "zstd: Refusing to open a output file which will overwrite the input file \n");
+                return NULL;
+            }
+        }
+#endif
+    }
+
     if (g_sparseFileSupport == 1) {
         g_sparseFileSupport = ZSTD_SPARSE_DEFAULT;
     }
 
     if (UTIL_isRegularFile(dstFileName)) {
-        FILE* fCheck;
-        if (!strcmp(dstFileName, nulmark)) {
-            EXM_THROW(40, "%s is unexpectedly a regular file", dstFileName);
-        }
         /* Check if destination file already exists */
-        fCheck = fopen( dstFileName, "rb" );
+        FILE* const fCheck = fopen( dstFileName, "rb" );
+        if (!strcmp(dstFileName, nulmark)) {
+            EXM_THROW(40, "%s is unexpectedly categorized as a regular file",
+                        dstFileName);
+        }
         if (fCheck != NULL) {  /* dst file exists, authorization prompt */
             fclose(fCheck);
             if (!g_overwrite) {
@@ -467,6 +507,7 @@ static size_t FIO_createDictBuffer(void** bufferPtr, const char* fileName)
     DISPLAYLEVEL(4,"Loading %s as dictionary \n", fileName);
     fileHandle = fopen(fileName, "rb");
     if (fileHandle==NULL) EXM_THROW(31, "%s: %s", fileName, strerror(errno));
+
     fileSize = UTIL_getFileSize(fileName);
     if (fileSize > DICTSIZE_MAX) {
         EXM_THROW(32, "Dictionary file %s is too large (> %u MB)",
@@ -475,8 +516,9 @@ static size_t FIO_createDictBuffer(void** bufferPtr, const char* fileName)
     *bufferPtr = malloc((size_t)fileSize);
     if (*bufferPtr==NULL) EXM_THROW(34, "%s", strerror(errno));
     {   size_t const readSize = fread(*bufferPtr, 1, (size_t)fileSize, fileHandle);
-        if (readSize!=fileSize)
-            EXM_THROW(35, "Error reading dictionary file %s", fileName);
+        if (readSize != fileSize)
+            EXM_THROW(35, "Error reading dictionary file %s : %s",
+                    fileName, strerror(errno));
     }
     fclose(fileHandle);
     return (size_t)fileSize;
@@ -506,7 +548,8 @@ static cRess_t FIO_createCResources(const char* dictFileName, int cLevel,
     DISPLAYLEVEL(6, "FIO_createCResources \n");
     ress.cctx = ZSTD_createCCtx();
     if (ress.cctx == NULL)
-        EXM_THROW(30, "allocation error : can't create ZSTD_CCtx");
+        EXM_THROW(30, "allocation error (%s): can't create ZSTD_CCtx",
+                    strerror(errno));
     ress.srcBufferSize = ZSTD_CStreamInSize();
     ress.srcBuffer = malloc(ress.srcBufferSize);
     ress.dstBufferSize = ZSTD_CStreamOutSize();
@@ -523,40 +566,39 @@ static cRess_t FIO_createCResources(const char* dictFileName, int cLevel,
         if (g_adaptiveMode && !g_ldmFlag && !comprParams.windowLog)
             comprParams.windowLog = ADAPT_WINDOWLOG_DEFAULT;
 
-        CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_p_contentSizeFlag, 1) );  /* always enable content size when available (note: supposed to be default) */
-        CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_p_dictIDFlag, g_dictIDFlag) );
-        CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_p_checksumFlag, g_checksumFlag) );
+        CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_contentSizeFlag, 1) );  /* always enable content size when available (note: supposed to be default) */
+        CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_dictIDFlag, g_dictIDFlag) );
+        CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_checksumFlag, g_checksumFlag) );
         /* compression level */
-        CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_p_compressionLevel, (unsigned)cLevel) );
+        CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_compressionLevel, cLevel) );
         /* long distance matching */
-        CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_p_enableLongDistanceMatching, g_ldmFlag) );
-        CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_p_ldmHashLog, g_ldmHashLog) );
-        CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_p_ldmMinMatch, g_ldmMinMatch) );
+        CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_enableLongDistanceMatching, g_ldmFlag) );
+        CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_ldmHashLog, g_ldmHashLog) );
+        CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_ldmMinMatch, g_ldmMinMatch) );
         if (g_ldmBucketSizeLog != FIO_LDM_PARAM_NOTSET) {
-            CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_p_ldmBucketSizeLog, g_ldmBucketSizeLog) );
+            CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_ldmBucketSizeLog, g_ldmBucketSizeLog) );
         }
-        if (g_ldmHashEveryLog != FIO_LDM_PARAM_NOTSET) {
-            CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_p_ldmHashEveryLog, g_ldmHashEveryLog) );
+        if (g_ldmHashRateLog != FIO_LDM_PARAM_NOTSET) {
+            CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_ldmHashRateLog, g_ldmHashRateLog) );
         }
         /* compression parameters */
-        CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_p_windowLog, comprParams.windowLog) );
-        CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_p_chainLog, comprParams.chainLog) );
-        CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_p_hashLog, comprParams.hashLog) );
-        CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_p_searchLog, comprParams.searchLog) );
-        CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_p_minMatch, comprParams.searchLength) );
-        CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_p_targetLength, comprParams.targetLength) );
-        CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_p_compressionStrategy, (U32)comprParams.strategy) );
+        CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_windowLog, comprParams.windowLog) );
+        CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_chainLog, comprParams.chainLog) );
+        CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_hashLog, comprParams.hashLog) );
+        CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_searchLog, comprParams.searchLog) );
+        CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_minMatch, comprParams.minMatch) );
+        CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_targetLength, comprParams.targetLength) );
+        CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_strategy, comprParams.strategy) );
         /* multi-threading */
 #ifdef ZSTD_MULTITHREAD
         DISPLAYLEVEL(5,"set nb workers = %u \n", g_nbWorkers);
-        CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_p_nbWorkers, g_nbWorkers) );
-        if ( (g_overlapLog == FIO_OVERLAP_LOG_NOTSET)
-          && (cLevel == ZSTD_maxCLevel()) )
-            g_overlapLog = 9;   /* full overlap */
+        CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_nbWorkers, g_nbWorkers) );
+        CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_jobSize, g_blockSize) );
         if (g_overlapLog != FIO_OVERLAP_LOG_NOTSET) {
             DISPLAYLEVEL(3,"set overlapLog = %u \n", g_overlapLog);
-            CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_p_overlapSizeLog, g_overlapLog) );
+            CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_overlapLog, g_overlapLog) );
         }
+        CHECK( ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_rsyncable, g_rsyncable) );
 #endif
         /* dictionary */
         CHECK( ZSTD_CCtx_setPledgedSrcSize(ress.cctx, srcSize) );  /* set the value temporarily for dictionary loading, to adapt compression parameters */
@@ -627,11 +669,11 @@ FIO_compressGzFrame(cRess_t* ress,
         }
         if (srcFileSize == UTIL_FILESIZE_UNKNOWN)
             DISPLAYUPDATE(2, "\rRead : %u MB ==> %.2f%%",
-                            (U32)(inFileSize>>20),
+                            (unsigned)(inFileSize>>20),
                             (double)outFileSize/inFileSize*100)
         else
             DISPLAYUPDATE(2, "\rRead : %u / %u MB ==> %.2f%%",
-                            (U32)(inFileSize>>20), (U32)(srcFileSize>>20),
+                            (unsigned)(inFileSize>>20), (unsigned)(srcFileSize>>20),
                             (double)outFileSize/inFileSize*100);
     }
 
@@ -640,7 +682,7 @@ FIO_compressGzFrame(cRess_t* ress,
         {   size_t const decompBytes = ress->dstBufferSize - strm.avail_out;
             if (decompBytes) {
                 if (fwrite(ress->dstBuffer, 1, decompBytes, ress->dstFile) != decompBytes)
-                    EXM_THROW(75, "Write error : cannot write to output file");
+                    EXM_THROW(75, "Write error : %s", strerror(errno));
                 outFileSize += decompBytes;
                 strm.next_out = (Bytef*)ress->dstBuffer;
                 strm.avail_out = (uInt)ress->dstBufferSize;
@@ -708,18 +750,18 @@ FIO_compressLzmaFrame(cRess_t* ress,
         {   size_t const compBytes = ress->dstBufferSize - strm.avail_out;
             if (compBytes) {
                 if (fwrite(ress->dstBuffer, 1, compBytes, ress->dstFile) != compBytes)
-                    EXM_THROW(73, "Write error : cannot write to output file");
+                    EXM_THROW(73, "Write error : %s", strerror(errno));
                 outFileSize += compBytes;
                 strm.next_out = (BYTE*)ress->dstBuffer;
                 strm.avail_out = ress->dstBufferSize;
         }   }
         if (srcFileSize == UTIL_FILESIZE_UNKNOWN)
             DISPLAYUPDATE(2, "\rRead : %u MB ==> %.2f%%",
-                            (U32)(inFileSize>>20),
+                            (unsigned)(inFileSize>>20),
                             (double)outFileSize/inFileSize*100)
         else
             DISPLAYUPDATE(2, "\rRead : %u / %u MB ==> %.2f%%",
-                            (U32)(inFileSize>>20), (U32)(srcFileSize>>20),
+                            (unsigned)(inFileSize>>20), (unsigned)(srcFileSize>>20),
                             (double)outFileSize/inFileSize*100);
         if (ret == LZMA_STREAM_END) break;
     }
@@ -773,7 +815,7 @@ FIO_compressLz4Frame(cRess_t* ress,
             EXM_THROW(33, "File header generation failed : %s",
                             LZ4F_getErrorName(headerSize));
         if (fwrite(ress->dstBuffer, 1, headerSize, ress->dstFile) != headerSize)
-            EXM_THROW(34, "Write error : cannot write header");
+            EXM_THROW(34, "Write error : %s (cannot write header)", strerror(errno));
         outFileSize += headerSize;
 
         /* Read first block */
@@ -782,26 +824,28 @@ FIO_compressLz4Frame(cRess_t* ress,
 
         /* Main Loop */
         while (readSize>0) {
-            size_t outSize;
-
-            /* Compress Block */
-            outSize = LZ4F_compressUpdate(ctx, ress->dstBuffer, ress->dstBufferSize, ress->srcBuffer, readSize, NULL);
+            size_t const outSize = LZ4F_compressUpdate(ctx,
+                                        ress->dstBuffer, ress->dstBufferSize,
+                                        ress->srcBuffer, readSize, NULL);
             if (LZ4F_isError(outSize))
                 EXM_THROW(35, "zstd: %s: lz4 compression failed : %s",
                             srcFileName, LZ4F_getErrorName(outSize));
             outFileSize += outSize;
-            if (srcFileSize == UTIL_FILESIZE_UNKNOWN)
+            if (srcFileSize == UTIL_FILESIZE_UNKNOWN) {
                 DISPLAYUPDATE(2, "\rRead : %u MB ==> %.2f%%",
-                                (U32)(inFileSize>>20),
+                                (unsigned)(inFileSize>>20),
                                 (double)outFileSize/inFileSize*100)
-            else
+            } else {
                 DISPLAYUPDATE(2, "\rRead : %u / %u MB ==> %.2f%%",
-                                (U32)(inFileSize>>20), (U32)(srcFileSize>>20),
+                                (unsigned)(inFileSize>>20), (unsigned)(srcFileSize>>20),
                                 (double)outFileSize/inFileSize*100);
+            }
 
             /* Write Block */
-            { size_t const sizeCheck = fwrite(ress->dstBuffer, 1, outSize, ress->dstFile);
-              if (sizeCheck!=outSize) EXM_THROW(36, "Write error : cannot write compressed block"); }
+            {   size_t const sizeCheck = fwrite(ress->dstBuffer, 1, outSize, ress->dstFile);
+                if (sizeCheck != outSize)
+                    EXM_THROW(36, "Write error : %s", strerror(errno));
+            }
 
             /* Read next block */
             readSize  = fread(ress->srcBuffer, (size_t)1, (size_t)blockSize, ress->srcFile);
@@ -815,8 +859,11 @@ FIO_compressLz4Frame(cRess_t* ress,
             EXM_THROW(38, "zstd: %s: lz4 end of file generation failed : %s",
                         srcFileName, LZ4F_getErrorName(headerSize));
 
-        { size_t const sizeCheck = fwrite(ress->dstBuffer, 1, headerSize, ress->dstFile);
-          if (sizeCheck!=headerSize) EXM_THROW(39, "Write error : cannot write end of stream"); }
+        {   size_t const sizeCheck = fwrite(ress->dstBuffer, 1, headerSize, ress->dstFile);
+            if (sizeCheck != headerSize)
+                EXM_THROW(39, "Write error : %s (cannot write end of stream)",
+                            strerror(errno));
+        }
         outFileSize += headerSize;
     }
 
@@ -863,7 +910,7 @@ FIO_compressZstdFrame(const cRess_t* ressPtr,
         /* Fill input Buffer */
         size_t const inSize = fread(ress.srcBuffer, (size_t)1, ress.srcBufferSize, srcFile);
         ZSTD_inBuffer inBuff = { ress.srcBuffer, inSize, 0 };
-        DISPLAYLEVEL(6, "fread %u bytes from source \n", (U32)inSize);
+        DISPLAYLEVEL(6, "fread %u bytes from source \n", (unsigned)inSize);
         *readsize += inSize;
 
         if ((inSize == 0) || (*readsize == fileSize))
@@ -876,7 +923,7 @@ FIO_compressZstdFrame(const cRess_t* ressPtr,
             size_t const oldIPos = inBuff.pos;
             ZSTD_outBuffer outBuff = { ress.dstBuffer, ress.dstBufferSize, 0 };
             size_t const toFlushNow = ZSTD_toFlushNow(ress.cctx);
-            CHECK_V(stillToFlush, ZSTD_compress_generic(ress.cctx, &outBuff, &inBuff, directive));
+            CHECK_V(stillToFlush, ZSTD_compressStream2(ress.cctx, &outBuff, &inBuff, directive));
 
             /* count stats */
             inputPresented++;
@@ -885,11 +932,12 @@ FIO_compressZstdFrame(const cRess_t* ressPtr,
 
             /* Write compressed stream */
             DISPLAYLEVEL(6, "ZSTD_compress_generic(end:%u) => input pos(%u)<=(%u)size ; output generated %u bytes \n",
-                            (U32)directive, (U32)inBuff.pos, (U32)inBuff.size, (U32)outBuff.pos);
+                            (unsigned)directive, (unsigned)inBuff.pos, (unsigned)inBuff.size, (unsigned)outBuff.pos);
             if (outBuff.pos) {
                 size_t const sizeCheck = fwrite(ress.dstBuffer, 1, outBuff.pos, dstFile);
                 if (sizeCheck != outBuff.pos)
-                    EXM_THROW(25, "Write error : cannot write compressed block");
+                    EXM_THROW(25, "Write error : %s (cannot write compressed block)",
+                                    strerror(errno));
                 compressedfilesize += outBuff.pos;
             }
 
@@ -902,14 +950,14 @@ FIO_compressZstdFrame(const cRess_t* ressPtr,
                 if (g_displayLevel >= 3) {
                     DISPLAYUPDATE(3, "\r(L%i) Buffered :%4u MB - Consumed :%4u MB - Compressed :%4u MB => %.2f%% ",
                                 compressionLevel,
-                                (U32)((zfp.ingested - zfp.consumed) >> 20),
-                                (U32)(zfp.consumed >> 20),
-                                (U32)(zfp.produced >> 20),
+                                (unsigned)((zfp.ingested - zfp.consumed) >> 20),
+                                (unsigned)(zfp.consumed >> 20),
+                                (unsigned)(zfp.produced >> 20),
                                 cShare );
                 } else {   /* summarized notifications if == 2; */
-                    DISPLAYLEVEL(2, "\rRead : %u ", (U32)(zfp.consumed >> 20));
+                    DISPLAYLEVEL(2, "\rRead : %u ", (unsigned)(zfp.consumed >> 20));
                     if (fileSize != UTIL_FILESIZE_UNKNOWN)
-                        DISPLAYLEVEL(2, "/ %u ", (U32)(fileSize >> 20));
+                        DISPLAYLEVEL(2, "/ %u ", (unsigned)(fileSize >> 20));
                     DISPLAYLEVEL(2, "MB ==> %2.f%% ", cShare);
                     DELAY_NEXT_UPDATE();
                 }
@@ -965,8 +1013,8 @@ FIO_compressZstdFrame(const cRess_t* ressPtr,
                                 assert(inputPresented > 0);
                                 DISPLAYLEVEL(6, "input blocked %u/%u(%.2f) - ingested:%u vs %u:consumed - flushed:%u vs %u:produced \n",
                                                 inputBlocked, inputPresented, (double)inputBlocked/inputPresented*100,
-                                                (U32)newlyIngested, (U32)newlyConsumed,
-                                                (U32)newlyFlushed, (U32)newlyProduced);
+                                                (unsigned)newlyIngested, (unsigned)newlyConsumed,
+                                                (unsigned)newlyFlushed, (unsigned)newlyProduced);
                                 if ( (inputBlocked > inputPresented / 8)     /* input is waiting often, because input buffers is full : compression or output too slow */
                                   && (newlyFlushed * 33 / 32 > newlyProduced)  /* flush everything that is produced */
                                   && (newlyIngested * 33 / 32 > newlyConsumed) /* input speed as fast or faster than compression speed */
@@ -986,14 +1034,14 @@ FIO_compressZstdFrame(const cRess_t* ressPtr,
                             if (compressionLevel > ZSTD_maxCLevel()) compressionLevel = ZSTD_maxCLevel();
                             if (compressionLevel > g_maxAdaptLevel) compressionLevel = g_maxAdaptLevel;
                             compressionLevel += (compressionLevel == 0);   /* skip 0 */
-                            ZSTD_CCtx_setParameter(ress.cctx, ZSTD_p_compressionLevel, (unsigned)compressionLevel);
+                            ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_compressionLevel, compressionLevel);
                         }
                         if (speedChange == faster) {
                             DISPLAYLEVEL(6, "faster speed , lighter compression \n")
                             compressionLevel --;
                             if (compressionLevel < g_minAdaptLevel) compressionLevel = g_minAdaptLevel;
                             compressionLevel -= (compressionLevel == 0);   /* skip 0 */
-                            ZSTD_CCtx_setParameter(ress.cctx, ZSTD_p_compressionLevel, (unsigned)compressionLevel);
+                            ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_compressionLevel, compressionLevel);
                         }
                         speedChange = noChange;
 
@@ -1028,7 +1076,7 @@ FIO_compressFilename_internal(cRess_t ress,
     U64 readsize = 0;
     U64 compressedfilesize = 0;
     U64 const fileSize = UTIL_getFileSize(srcFileName);
-    DISPLAYLEVEL(5, "%s: %u bytes \n", srcFileName, (U32)fileSize);
+    DISPLAYLEVEL(5, "%s: %u bytes \n", srcFileName, (unsigned)fileSize);
 
     /* compression format selection */
     switch (g_compressionType) {
@@ -1105,7 +1153,7 @@ static int FIO_compressFilename_dstFile(cRess_t ress,
     if (ress.dstFile == NULL) {
         closeDstFile = 1;
         DISPLAYLEVEL(6, "FIO_compressFilename_dstFile: opening dst: %s", dstFileName);
-        ress.dstFile = FIO_openDstFile(dstFileName);
+        ress.dstFile = FIO_openDstFile(srcFileName, dstFileName);
         if (ress.dstFile==NULL) return 1;  /* could not open dstFileName */
         /* Must only be added after FIO_openDstFile() succeeds.
          * Otherwise we may delete the destination file if it already exists,
@@ -1255,7 +1303,7 @@ int FIO_compressMultipleFilenames(const char** inFileNamesTable, unsigned nbFile
     assert(outFileName != NULL || suffix != NULL);
 
     if (outFileName != NULL) {   /* output into a single destination (stdout typically) */
-        ress.dstFile = FIO_openDstFile(outFileName);
+        ress.dstFile = FIO_openDstFile(NULL, outFileName);
         if (ress.dstFile == NULL) {  /* could not open outFileName */
             error = 1;
         } else {
@@ -1263,7 +1311,8 @@ int FIO_compressMultipleFilenames(const char** inFileNamesTable, unsigned nbFile
             for (u=0; u 1 GB) {
         int const seekResult = LONG_SEEK(file, 1 GB, SEEK_CUR);
-        if (seekResult != 0) EXM_THROW(71, "1 GB skip error (sparse file support)");
+        if (seekResult != 0)
+            EXM_THROW(71, "1 GB skip error (sparse file support)");
         storedSkips -= 1 GB;
     }
 
@@ -1401,12 +1454,14 @@ static unsigned FIO_fwriteSparse(FILE* file, const void* buffer, size_t bufferSi
 
 static void FIO_fwriteSparseEnd(FILE* file, unsigned storedSkips)
 {
-    if (storedSkips-->0) {   /* implies g_sparseFileSupport>0 */
-        int const seekResult = LONG_SEEK(file, storedSkips, SEEK_CUR);
-        if (seekResult != 0) EXM_THROW(69, "Final skip error (sparse file)");
+    if (storedSkips>0) {
+        assert(g_sparseFileSupport > 0);  /* storedSkips>0 implies sparse support is enabled */
+        if (LONG_SEEK(file, storedSkips-1, SEEK_CUR) != 0)
+            EXM_THROW(69, "Final skip error (sparse file support)");
+        /* last zero must be explicitly written,
+         * so that skipped ones get implicitly translated as zero by FS */
         {   const char lastZeroByte[1] = { 0 };
-            size_t const sizeCheck = fwrite(lastZeroByte, 1, 1, file);
-            if (sizeCheck != 1)
+            if (fwrite(lastZeroByte, 1, 1, file) != 1)
                 EXM_THROW(69, "Write error : cannot write last zero");
     }   }
 }
@@ -1463,12 +1518,12 @@ static void FIO_zstdErrorHelp(dRess_t* ress, size_t err, char const* srcFileName
     err = ZSTD_getFrameHeader(&header, ress->srcBuffer, ress->srcBufferLoaded);
     if (err == 0) {
         unsigned long long const windowSize = header.windowSize;
-        U32 const windowLog = FIO_highbit64(windowSize) + ((windowSize & (windowSize - 1)) != 0);
+        unsigned const windowLog = FIO_highbit64(windowSize) + ((windowSize & (windowSize - 1)) != 0);
         assert(g_memLimit > 0);
         DISPLAYLEVEL(1, "%s : Window size larger than maximum : %llu > %u\n",
                         srcFileName, windowSize, g_memLimit);
         if (windowLog <= ZSTD_WINDOWLOG_MAX) {
-            U32 const windowMB = (U32)((windowSize >> 20) + ((windowSize & ((1 MB) - 1)) != 0));
+            unsigned const windowMB = (unsigned)((windowSize >> 20) + ((windowSize & ((1 MB) - 1)) != 0));
             assert(windowSize < (U64)(1ULL << 52));   /* ensure now overflow for windowMB */
             DISPLAYLEVEL(1, "%s : Use --long=%u or --memory=%uMB\n",
                             srcFileName, windowLog, windowMB);
@@ -1520,7 +1575,7 @@ static unsigned long long FIO_decompressZstdFrame(dRess_t* ress,
         storedSkips = FIO_fwriteSparse(ress->dstFile, ress->dstBuffer, outBuff.pos, storedSkips);
         frameSize += outBuff.pos;
         DISPLAYUPDATE(2, "\r%-20.20s : %u MB...     ",
-                         srcFileName, (U32)((alreadyDecoded+frameSize)>>20) );
+                         srcFileName, (unsigned)((alreadyDecoded+frameSize)>>20) );
 
         if (inBuff.pos > 0) {
             memmove(ress->srcBuffer, (char*)ress->srcBuffer + inBuff.pos, inBuff.size - inBuff.pos);
@@ -1871,7 +1926,7 @@ static int FIO_decompressDstFile(dRess_t ress, FILE* srcFile,
     if (ress.dstFile == NULL) {
         releaseDstFile = 1;
 
-        ress.dstFile = FIO_openDstFile(dstFileName);
+        ress.dstFile = FIO_openDstFile(srcFileName, dstFileName);
         if (ress.dstFile==0) return 1;
 
         /* Must only be added after FIO_openDstFile() succeeds.
@@ -2025,7 +2080,7 @@ FIO_determineDstName(const char* srcFileName)
         dfnbCapacity = sfnSize + 20;
         dstFileNameBuffer = (char*)malloc(dfnbCapacity);
         if (dstFileNameBuffer==NULL)
-            EXM_THROW(74, "not enough memory for dstFileName");
+            EXM_THROW(74, "%s : not enough memory for dstFileName", strerror(errno));
     }
 
     /* return dst name == src name truncated from suffix */
@@ -2048,12 +2103,13 @@ FIO_decompressMultipleFilenames(const char* srcNamesTable[], unsigned nbFiles,
 
     if (outFileName) {
         unsigned u;
-        ress.dstFile = FIO_openDstFile(outFileName);
+        ress.dstFile = FIO_openDstFile(NULL, outFileName);
         if (ress.dstFile == 0) EXM_THROW(71, "cannot open %s", outFileName);
         for (u=0; ucompressedSize > 0)
@@ -2164,7 +2220,7 @@ FIO_analyzeFrames(fileInfo_t* info, FILE* const srcFile)
                 info->numActualFrames++;
             }
             /* Skippable frame */
-            else if ((magicNumber & 0xFFFFFFF0U) == ZSTD_MAGIC_SKIPPABLE_START) {
+            else if ((magicNumber & ZSTD_MAGIC_SKIPPABLE_MASK) == ZSTD_MAGIC_SKIPPABLE_START) {
                 U32 const frameSize = MEM_readLE32(headerBuffer + 4);
                 long const seek = (long)(8 + frameSize - numBytesRead);
                 ERROR_IF(LONG_SEEK(srcFile, seek, SEEK_CUR) != 0,
@@ -2292,7 +2348,7 @@ FIO_listFile(fileInfo_t* total, const char* inFileName, int displayLevel)
         }
         displayInfo(inFileName, &info, displayLevel);
         *total = FIO_addFInfo(*total, info);
-        assert(error>=0 || error<=1);
+        assert(error == info_success || error == info_frame_error);
         return error;
     }
 }
@@ -2339,13 +2395,13 @@ int FIO_listMultipleFiles(unsigned numFiles, const char** filenameTable, int dis
                         total.numSkippableFrames + total.numActualFrames,
                         total.numSkippableFrames,
                         compressedSizeUnit, unitStr,
-                        checkString, total.nbFiles);
+                        checkString, (unsigned)total.nbFiles);
             } else {
                 DISPLAYOUT("%6d  %5d  %7.2f %2s  %9.2f %2s  %5.3f  %5s  %u files\n",
                         total.numSkippableFrames + total.numActualFrames,
                         total.numSkippableFrames,
                         compressedSizeUnit, unitStr, decompressedSizeUnit, unitStr,
-                        ratio, checkString, total.nbFiles);
+                        ratio, checkString, (unsigned)total.nbFiles);
         }   }
         return error;
     }
diff --git a/programs/fileio.h b/programs/fileio.h
index 4c7049cb7167..97f270632071 100644
--- a/programs/fileio.h
+++ b/programs/fileio.h
@@ -56,7 +56,7 @@ void FIO_setChecksumFlag(unsigned checksumFlag);
 void FIO_setDictIDFlag(unsigned dictIDFlag);
 void FIO_setLdmBucketSizeLog(unsigned ldmBucketSizeLog);
 void FIO_setLdmFlag(unsigned ldmFlag);
-void FIO_setLdmHashEveryLog(unsigned ldmHashEveryLog);
+void FIO_setLdmHashRateLog(unsigned ldmHashRateLog);
 void FIO_setLdmHashLog(unsigned ldmHashLog);
 void FIO_setLdmMinMatch(unsigned ldmMinMatch);
 void FIO_setMemLimit(unsigned memLimit);
@@ -65,6 +65,8 @@ void FIO_setNotificationLevel(unsigned level);
 void FIO_setOverlapLog(unsigned overlapLog);
 void FIO_setRemoveSrcFile(unsigned flag);
 void FIO_setSparseWrite(unsigned sparse);  /**< 0: no sparse; 1: disable on stdout; 2: always enabled */
+void FIO_setRsyncable(unsigned rsyncable);
+void FIO_setNoProgress(unsigned noProgress);
 
 
 /*-*************************************
diff --git a/programs/platform.h b/programs/platform.h
index 155ebcd1eb9c..1a8f97bc4dff 100644
--- a/programs/platform.h
+++ b/programs/platform.h
@@ -26,6 +26,7 @@ extern "C" {
 #    define _CRT_SECURE_NO_DEPRECATE /* VS2005 - must be declared before  and  */
 #    define snprintf sprintf_s       /* snprintf unsupported by Visual <= 2013 */
 #  endif
+#  pragma warning(disable : 4127)    /* disable: C4127: conditional expression is constant */
 #endif
 
 
diff --git a/programs/util.c b/programs/util.c
new file mode 100644
index 000000000000..34634318c638
--- /dev/null
+++ b/programs/util.c
@@ -0,0 +1,673 @@
+/*
+ * Copyright (c) 2016-present, Przemyslaw Skibinski, Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
+ * You may select, at your option, one of the above-listed licenses.
+ */
+
+#if defined (__cplusplus)
+extern "C" {
+#endif
+
+
+/*-****************************************
+*  Dependencies
+******************************************/
+#include "util.h"       /* note : ensure that platform.h is included first ! */
+#include 
+#include 
+
+
+int UTIL_fileExist(const char* filename)
+{
+    stat_t statbuf;
+#if defined(_MSC_VER)
+    int const stat_error = _stat64(filename, &statbuf);
+#else
+    int const stat_error = stat(filename, &statbuf);
+#endif
+    return !stat_error;
+}
+
+int UTIL_isRegularFile(const char* infilename)
+{
+    stat_t statbuf;
+    return UTIL_getFileStat(infilename, &statbuf); /* Only need to know whether it is a regular file */
+}
+
+int UTIL_getFileStat(const char* infilename, stat_t *statbuf)
+{
+    int r;
+#if defined(_MSC_VER)
+    r = _stat64(infilename, statbuf);
+    if (r || !(statbuf->st_mode & S_IFREG)) return 0;   /* No good... */
+#else
+    r = stat(infilename, statbuf);
+    if (r || !S_ISREG(statbuf->st_mode)) return 0;   /* No good... */
+#endif
+    return 1;
+}
+
+int UTIL_setFileStat(const char *filename, stat_t *statbuf)
+{
+    int res = 0;
+    struct utimbuf timebuf;
+
+    if (!UTIL_isRegularFile(filename))
+        return -1;
+
+    timebuf.actime = time(NULL);
+    timebuf.modtime = statbuf->st_mtime;
+    res += utime(filename, &timebuf);  /* set access and modification times */
+
+#if !defined(_WIN32)
+    res += chown(filename, statbuf->st_uid, statbuf->st_gid);  /* Copy ownership */
+#endif
+
+    res += chmod(filename, statbuf->st_mode & 07777);  /* Copy file permissions */
+
+    errno = 0;
+    return -res; /* number of errors is returned */
+}
+
+U32 UTIL_isDirectory(const char* infilename)
+{
+    int r;
+    stat_t statbuf;
+#if defined(_MSC_VER)
+    r = _stat64(infilename, &statbuf);
+    if (!r && (statbuf.st_mode & _S_IFDIR)) return 1;
+#else
+    r = stat(infilename, &statbuf);
+    if (!r && S_ISDIR(statbuf.st_mode)) return 1;
+#endif
+    return 0;
+}
+
+U32 UTIL_isLink(const char* infilename)
+{
+/* macro guards, as defined in : https://linux.die.net/man/2/lstat */
+#ifndef __STRICT_ANSI__
+#if defined(_BSD_SOURCE) \
+    || (defined(_XOPEN_SOURCE) && (_XOPEN_SOURCE >= 500)) \
+    || (defined(_XOPEN_SOURCE) && defined(_XOPEN_SOURCE_EXTENDED)) \
+    || (defined(_POSIX_C_SOURCE) && (_POSIX_C_SOURCE >= 200112L)) \
+    || (defined(__APPLE__) && defined(__MACH__))
+    int r;
+    stat_t statbuf;
+    r = lstat(infilename, &statbuf);
+    if (!r && S_ISLNK(statbuf.st_mode)) return 1;
+#endif
+#endif
+    (void)infilename;
+    return 0;
+}
+
+U64 UTIL_getFileSize(const char* infilename)
+{
+    if (!UTIL_isRegularFile(infilename)) return UTIL_FILESIZE_UNKNOWN;
+    {   int r;
+#if defined(_MSC_VER)
+        struct __stat64 statbuf;
+        r = _stat64(infilename, &statbuf);
+        if (r || !(statbuf.st_mode & S_IFREG)) return UTIL_FILESIZE_UNKNOWN;
+#elif defined(__MINGW32__) && defined (__MSVCRT__)
+        struct _stati64 statbuf;
+        r = _stati64(infilename, &statbuf);
+        if (r || !(statbuf.st_mode & S_IFREG)) return UTIL_FILESIZE_UNKNOWN;
+#else
+        struct stat statbuf;
+        r = stat(infilename, &statbuf);
+        if (r || !S_ISREG(statbuf.st_mode)) return UTIL_FILESIZE_UNKNOWN;
+#endif
+        return (U64)statbuf.st_size;
+    }
+}
+
+
+U64 UTIL_getTotalFileSize(const char* const * const fileNamesTable, unsigned nbFiles)
+{
+    U64 total = 0;
+    int error = 0;
+    unsigned n;
+    for (n=0; n= *bufEnd) {
+                ptrdiff_t const newListSize = (*bufEnd - *bufStart) + LIST_SIZE_INCREASE;
+                *bufStart = (char*)UTIL_realloc(*bufStart, newListSize);
+                if (*bufStart == NULL) { free(path); FindClose(hFile); return 0; }
+                *bufEnd = *bufStart + newListSize;
+            }
+            if (*bufStart + *pos + pathLength < *bufEnd) {
+                memcpy(*bufStart + *pos, path, pathLength+1 /* include final \0 */);
+                *pos += pathLength + 1;
+                nbFiles++;
+            }
+        }
+        free(path);
+    } while (FindNextFileA(hFile, &cFile));
+
+    FindClose(hFile);
+    return nbFiles;
+}
+
+#elif defined(__linux__) || (PLATFORM_POSIX_VERSION >= 200112L)  /* opendir, readdir require POSIX.1-2001 */
+
+int UTIL_prepareFileList(const char *dirName, char** bufStart, size_t* pos, char** bufEnd, int followLinks)
+{
+    DIR *dir;
+    struct dirent *entry;
+    char* path;
+    int dirLength, fnameLength, pathLength, nbFiles = 0;
+
+    if (!(dir = opendir(dirName))) {
+        UTIL_DISPLAYLEVEL(1, "Cannot open directory '%s': %s\n", dirName, strerror(errno));
+        return 0;
+    }
+
+    dirLength = (int)strlen(dirName);
+    errno = 0;
+    while ((entry = readdir(dir)) != NULL) {
+        if (strcmp (entry->d_name, "..") == 0 ||
+            strcmp (entry->d_name, ".") == 0) continue;
+        fnameLength = (int)strlen(entry->d_name);
+        path = (char*) malloc(dirLength + fnameLength + 2);
+        if (!path) { closedir(dir); return 0; }
+        memcpy(path, dirName, dirLength);
+
+        path[dirLength] = '/';
+        memcpy(path+dirLength+1, entry->d_name, fnameLength);
+        pathLength = dirLength+1+fnameLength;
+        path[pathLength] = 0;
+
+        if (!followLinks && UTIL_isLink(path)) {
+            UTIL_DISPLAYLEVEL(2, "Warning : %s is a symbolic link, ignoring\n", path);
+            continue;
+        }
+
+        if (UTIL_isDirectory(path)) {
+            nbFiles += UTIL_prepareFileList(path, bufStart, pos, bufEnd, followLinks);  /* Recursively call "UTIL_prepareFileList" with the new path. */
+            if (*bufStart == NULL) { free(path); closedir(dir); return 0; }
+        } else {
+            if (*bufStart + *pos + pathLength >= *bufEnd) {
+                ptrdiff_t newListSize = (*bufEnd - *bufStart) + LIST_SIZE_INCREASE;
+                *bufStart = (char*)UTIL_realloc(*bufStart, newListSize);
+                *bufEnd = *bufStart + newListSize;
+                if (*bufStart == NULL) { free(path); closedir(dir); return 0; }
+            }
+            if (*bufStart + *pos + pathLength < *bufEnd) {
+                memcpy(*bufStart + *pos, path, pathLength + 1);  /* with final \0 */
+                *pos += pathLength + 1;
+                nbFiles++;
+            }
+        }
+        free(path);
+        errno = 0; /* clear errno after UTIL_isDirectory, UTIL_prepareFileList */
+    }
+
+    if (errno != 0) {
+        UTIL_DISPLAYLEVEL(1, "readdir(%s) error: %s\n", dirName, strerror(errno));
+        free(*bufStart);
+        *bufStart = NULL;
+    }
+    closedir(dir);
+    return nbFiles;
+}
+
+#else
+
+int UTIL_prepareFileList(const char *dirName, char** bufStart, size_t* pos, char** bufEnd, int followLinks)
+{
+    (void)bufStart; (void)bufEnd; (void)pos; (void)followLinks;
+    UTIL_DISPLAYLEVEL(1, "Directory %s ignored (compiled without _WIN32 or _POSIX_C_SOURCE)\n", dirName);
+    return 0;
+}
+
+#endif /* #ifdef _WIN32 */
+
+/*
+ * UTIL_createFileList - takes a list of files and directories (params: inputNames, inputNamesNb), scans directories,
+ *                       and returns a new list of files (params: return value, allocatedBuffer, allocatedNamesNb).
+ * After finishing usage of the list the structures should be freed with UTIL_freeFileList(params: return value, allocatedBuffer)
+ * In case of error UTIL_createFileList returns NULL and UTIL_freeFileList should not be called.
+ */
+const char**
+UTIL_createFileList(const char **inputNames, unsigned inputNamesNb,
+                    char** allocatedBuffer, unsigned* allocatedNamesNb,
+                    int followLinks)
+{
+    size_t pos;
+    unsigned i, nbFiles;
+    char* buf = (char*)malloc(LIST_SIZE_INCREASE);
+    char* bufend = buf + LIST_SIZE_INCREASE;
+    const char** fileTable;
+
+    if (!buf) return NULL;
+
+    for (i=0, pos=0, nbFiles=0; i= bufend) {
+                ptrdiff_t newListSize = (bufend - buf) + LIST_SIZE_INCREASE;
+                buf = (char*)UTIL_realloc(buf, newListSize);
+                bufend = buf + newListSize;
+                if (!buf) return NULL;
+            }
+            if (buf + pos + len < bufend) {
+                memcpy(buf+pos, inputNames[i], len+1);  /* with final \0 */
+                pos += len + 1;
+                nbFiles++;
+            }
+        } else {
+            nbFiles += UTIL_prepareFileList(inputNames[i], &buf, &pos, &bufend, followLinks);
+            if (buf == NULL) return NULL;
+    }   }
+
+    if (nbFiles == 0) { free(buf); return NULL; }
+
+    fileTable = (const char**)malloc((nbFiles+1) * sizeof(const char*));
+    if (!fileTable) { free(buf); return NULL; }
+
+    for (i=0, pos=0; i bufend) { free(buf); free((void*)fileTable); return NULL; }
+
+    *allocatedBuffer = buf;
+    *allocatedNamesNb = nbFiles;
+
+    return fileTable;
+}
+
+/*-****************************************
+*  Console log
+******************************************/
+int g_utilDisplayLevel;
+
+
+/*-****************************************
+*  Time functions
+******************************************/
+#if defined(_WIN32)   /* Windows */
+
+UTIL_time_t UTIL_getTime(void) { UTIL_time_t x; QueryPerformanceCounter(&x); return x; }
+
+U64 UTIL_getSpanTimeMicro(UTIL_time_t clockStart, UTIL_time_t clockEnd)
+{
+    static LARGE_INTEGER ticksPerSecond;
+    static int init = 0;
+    if (!init) {
+        if (!QueryPerformanceFrequency(&ticksPerSecond))
+            UTIL_DISPLAYLEVEL(1, "ERROR: QueryPerformanceFrequency() failure\n");
+        init = 1;
+    }
+    return 1000000ULL*(clockEnd.QuadPart - clockStart.QuadPart)/ticksPerSecond.QuadPart;
+}
+
+U64 UTIL_getSpanTimeNano(UTIL_time_t clockStart, UTIL_time_t clockEnd)
+{
+    static LARGE_INTEGER ticksPerSecond;
+    static int init = 0;
+    if (!init) {
+        if (!QueryPerformanceFrequency(&ticksPerSecond))
+            UTIL_DISPLAYLEVEL(1, "ERROR: QueryPerformanceFrequency() failure\n");
+        init = 1;
+    }
+    return 1000000000ULL*(clockEnd.QuadPart - clockStart.QuadPart)/ticksPerSecond.QuadPart;
+}
+
+#elif defined(__APPLE__) && defined(__MACH__)
+
+UTIL_time_t UTIL_getTime(void) { return mach_absolute_time(); }
+
+U64 UTIL_getSpanTimeMicro(UTIL_time_t clockStart, UTIL_time_t clockEnd)
+{
+    static mach_timebase_info_data_t rate;
+    static int init = 0;
+    if (!init) {
+        mach_timebase_info(&rate);
+        init = 1;
+    }
+    return (((clockEnd - clockStart) * (U64)rate.numer) / ((U64)rate.denom))/1000ULL;
+}
+
+U64 UTIL_getSpanTimeNano(UTIL_time_t clockStart, UTIL_time_t clockEnd)
+{
+    static mach_timebase_info_data_t rate;
+    static int init = 0;
+    if (!init) {
+        mach_timebase_info(&rate);
+        init = 1;
+    }
+    return ((clockEnd - clockStart) * (U64)rate.numer) / ((U64)rate.denom);
+}
+
+#elif (PLATFORM_POSIX_VERSION >= 200112L) \
+   && (defined(__UCLIBC__)                \
+      || (defined(__GLIBC__)              \
+          && ((__GLIBC__ == 2 && __GLIBC_MINOR__ >= 17) \
+             || (__GLIBC__ > 2))))
+
+UTIL_time_t UTIL_getTime(void)
+{
+    UTIL_time_t time;
+    if (clock_gettime(CLOCK_MONOTONIC, &time))
+        UTIL_DISPLAYLEVEL(1, "ERROR: Failed to get time\n");   /* we could also exit() */
+    return time;
+}
+
+UTIL_time_t UTIL_getSpanTime(UTIL_time_t begin, UTIL_time_t end)
+{
+    UTIL_time_t diff;
+    if (end.tv_nsec < begin.tv_nsec) {
+        diff.tv_sec = (end.tv_sec - 1) - begin.tv_sec;
+        diff.tv_nsec = (end.tv_nsec + 1000000000ULL) - begin.tv_nsec;
+    } else {
+        diff.tv_sec = end.tv_sec - begin.tv_sec;
+        diff.tv_nsec = end.tv_nsec - begin.tv_nsec;
+    }
+    return diff;
+}
+
+U64 UTIL_getSpanTimeMicro(UTIL_time_t begin, UTIL_time_t end)
+{
+    UTIL_time_t const diff = UTIL_getSpanTime(begin, end);
+    U64 micro = 0;
+    micro += 1000000ULL * diff.tv_sec;
+    micro += diff.tv_nsec / 1000ULL;
+    return micro;
+}
+
+U64 UTIL_getSpanTimeNano(UTIL_time_t begin, UTIL_time_t end)
+{
+    UTIL_time_t const diff = UTIL_getSpanTime(begin, end);
+    U64 nano = 0;
+    nano += 1000000000ULL * diff.tv_sec;
+    nano += diff.tv_nsec;
+    return nano;
+}
+
+#else   /* relies on standard C (note : clock_t measurements can be wrong when using multi-threading) */
+
+UTIL_time_t UTIL_getTime(void) { return clock(); }
+U64 UTIL_getSpanTimeMicro(UTIL_time_t clockStart, UTIL_time_t clockEnd) { return 1000000ULL * (clockEnd - clockStart) / CLOCKS_PER_SEC; }
+U64 UTIL_getSpanTimeNano(UTIL_time_t clockStart, UTIL_time_t clockEnd) { return 1000000000ULL * (clockEnd - clockStart) / CLOCKS_PER_SEC; }
+
+#endif
+
+/* returns time span in microseconds */
+U64 UTIL_clockSpanMicro(UTIL_time_t clockStart )
+{
+    UTIL_time_t const clockEnd = UTIL_getTime();
+    return UTIL_getSpanTimeMicro(clockStart, clockEnd);
+}
+
+/* returns time span in microseconds */
+U64 UTIL_clockSpanNano(UTIL_time_t clockStart )
+{
+    UTIL_time_t const clockEnd = UTIL_getTime();
+    return UTIL_getSpanTimeNano(clockStart, clockEnd);
+}
+
+void UTIL_waitForNextTick(void)
+{
+    UTIL_time_t const clockStart = UTIL_getTime();
+    UTIL_time_t clockEnd;
+    do {
+        clockEnd = UTIL_getTime();
+    } while (UTIL_getSpanTimeNano(clockStart, clockEnd) == 0);
+}
+
+/* count the number of physical cores */
+#if defined(_WIN32) || defined(WIN32)
+
+#include 
+
+typedef BOOL(WINAPI* LPFN_GLPI)(PSYSTEM_LOGICAL_PROCESSOR_INFORMATION, PDWORD);
+
+int UTIL_countPhysicalCores(void)
+{
+    static int numPhysicalCores = 0;
+    if (numPhysicalCores != 0) return numPhysicalCores;
+
+    {   LPFN_GLPI glpi;
+        BOOL done = FALSE;
+        PSYSTEM_LOGICAL_PROCESSOR_INFORMATION buffer = NULL;
+        PSYSTEM_LOGICAL_PROCESSOR_INFORMATION ptr = NULL;
+        DWORD returnLength = 0;
+        size_t byteOffset = 0;
+
+        glpi = (LPFN_GLPI)GetProcAddress(GetModuleHandle(TEXT("kernel32")),
+                                         "GetLogicalProcessorInformation");
+
+        if (glpi == NULL) {
+            goto failed;
+        }
+
+        while(!done) {
+            DWORD rc = glpi(buffer, &returnLength);
+            if (FALSE == rc) {
+                if (GetLastError() == ERROR_INSUFFICIENT_BUFFER) {
+                    if (buffer)
+                        free(buffer);
+                    buffer = (PSYSTEM_LOGICAL_PROCESSOR_INFORMATION)malloc(returnLength);
+
+                    if (buffer == NULL) {
+                        perror("zstd");
+                        exit(1);
+                    }
+                } else {
+                    /* some other error */
+                    goto failed;
+                }
+            } else {
+                done = TRUE;
+            }
+        }
+
+        ptr = buffer;
+
+        while (byteOffset + sizeof(SYSTEM_LOGICAL_PROCESSOR_INFORMATION) <= returnLength) {
+
+            if (ptr->Relationship == RelationProcessorCore) {
+                numPhysicalCores++;
+            }
+
+            ptr++;
+            byteOffset += sizeof(SYSTEM_LOGICAL_PROCESSOR_INFORMATION);
+        }
+
+        free(buffer);
+
+        return numPhysicalCores;
+    }
+
+failed:
+    /* try to fall back on GetSystemInfo */
+    {   SYSTEM_INFO sysinfo;
+        GetSystemInfo(&sysinfo);
+        numPhysicalCores = sysinfo.dwNumberOfProcessors;
+        if (numPhysicalCores == 0) numPhysicalCores = 1; /* just in case */
+    }
+    return numPhysicalCores;
+}
+
+#elif defined(__APPLE__)
+
+#include 
+
+/* Use apple-provided syscall
+ * see: man 3 sysctl */
+int UTIL_countPhysicalCores(void)
+{
+    static S32 numPhysicalCores = 0; /* apple specifies int32_t */
+    if (numPhysicalCores != 0) return numPhysicalCores;
+
+    {   size_t size = sizeof(S32);
+        int const ret = sysctlbyname("hw.physicalcpu", &numPhysicalCores, &size, NULL, 0);
+        if (ret != 0) {
+            if (errno == ENOENT) {
+                /* entry not present, fall back on 1 */
+                numPhysicalCores = 1;
+            } else {
+                perror("zstd: can't get number of physical cpus");
+                exit(1);
+            }
+        }
+
+        return numPhysicalCores;
+    }
+}
+
+#elif defined(__linux__)
+
+/* parse /proc/cpuinfo
+ * siblings / cpu cores should give hyperthreading ratio
+ * otherwise fall back on sysconf */
+int UTIL_countPhysicalCores(void)
+{
+    static int numPhysicalCores = 0;
+
+    if (numPhysicalCores != 0) return numPhysicalCores;
+
+    numPhysicalCores = (int)sysconf(_SC_NPROCESSORS_ONLN);
+    if (numPhysicalCores == -1) {
+        /* value not queryable, fall back on 1 */
+        return numPhysicalCores = 1;
+    }
+
+    /* try to determine if there's hyperthreading */
+    {   FILE* const cpuinfo = fopen("/proc/cpuinfo", "r");
+#define BUF_SIZE 80
+        char buff[BUF_SIZE];
+
+        int siblings = 0;
+        int cpu_cores = 0;
+        int ratio = 1;
+
+        if (cpuinfo == NULL) {
+            /* fall back on the sysconf value */
+            return numPhysicalCores;
+        }
+
+        /* assume the cpu cores/siblings values will be constant across all
+         * present processors */
+        while (!feof(cpuinfo)) {
+            if (fgets(buff, BUF_SIZE, cpuinfo) != NULL) {
+                if (strncmp(buff, "siblings", 8) == 0) {
+                    const char* const sep = strchr(buff, ':');
+                    if (*sep == '\0') {
+                        /* formatting was broken? */
+                        goto failed;
+                    }
+
+                    siblings = atoi(sep + 1);
+                }
+                if (strncmp(buff, "cpu cores", 9) == 0) {
+                    const char* const sep = strchr(buff, ':');
+                    if (*sep == '\0') {
+                        /* formatting was broken? */
+                        goto failed;
+                    }
+
+                    cpu_cores = atoi(sep + 1);
+                }
+            } else if (ferror(cpuinfo)) {
+                /* fall back on the sysconf value */
+                goto failed;
+            }
+        }
+        if (siblings && cpu_cores) {
+            ratio = siblings / cpu_cores;
+        }
+failed:
+        fclose(cpuinfo);
+        return numPhysicalCores = numPhysicalCores / ratio;
+    }
+}
+
+#elif defined(__FreeBSD__) || defined(__NetBSD__) || defined(__OpenBSD__) || defined(__DragonFly__)
+
+/* Use apple-provided syscall
+ * see: man 3 sysctl */
+int UTIL_countPhysicalCores(void)
+{
+    static int numPhysicalCores = 0;
+
+    if (numPhysicalCores != 0) return numPhysicalCores;
+
+    numPhysicalCores = (int)sysconf(_SC_NPROCESSORS_ONLN);
+    if (numPhysicalCores == -1) {
+        /* value not queryable, fall back on 1 */
+        return numPhysicalCores = 1;
+    }
+    return numPhysicalCores;
+}
+
+#else
+
+int UTIL_countPhysicalCores(void)
+{
+    /* assume 1 */
+    return 1;
+}
+
+#endif
+
+#if defined (__cplusplus)
+}
+#endif
diff --git a/programs/util.h b/programs/util.h
index 67aa7a56b967..f78bcbe1b3f8 100644
--- a/programs/util.h
+++ b/programs/util.h
@@ -16,15 +16,13 @@ extern "C" {
 #endif
 
 
-
 /*-****************************************
 *  Dependencies
 ******************************************/
 #include "platform.h"     /* PLATFORM_POSIX_VERSION, ZSTD_NANOSLEEP_SUPPORT, ZSTD_SETPRIORITY_SUPPORT */
-#include        /* malloc */
+#include        /* malloc, realloc, free */
 #include        /* size_t, ptrdiff_t */
 #include         /* fprintf */
-#include        /* strncmp */
 #include     /* stat, utime */
 #include      /* stat, chmod */
 #if defined(_MSC_VER)
@@ -35,11 +33,10 @@ extern "C" {
 #  include       /* utime */
 #endif
 #include          /* clock_t, clock, CLOCKS_PER_SEC, nanosleep */
-#include 
 #include "mem.h"          /* U32, U64 */
 
 
-/* ************************************************************
+/*-************************************************************
 * Avoid fseek()'s 2GiB barrier with MSVC, macOS, *BSD, MinGW
 ***************************************************************/
 #if defined(_MSC_VER) && (_MSC_VER >= 1400)
@@ -84,7 +81,7 @@ extern "C" {
 #endif
 
 
-/* *************************************
+/*-*************************************
 *  Constants
 ***************************************/
 #define LIST_SIZE_INCREASE   (8*1024)
@@ -110,7 +107,7 @@ extern "C" {
 /*-****************************************
 *  Console log
 ******************************************/
-static int g_utilDisplayLevel;
+extern int g_utilDisplayLevel;
 #define UTIL_DISPLAY(...)         fprintf(stderr, __VA_ARGS__)
 #define UTIL_DISPLAYLEVEL(l, ...) { if (g_utilDisplayLevel>=l) { UTIL_DISPLAY(__VA_ARGS__); } }
 
@@ -119,61 +116,16 @@ static int g_utilDisplayLevel;
 *  Time functions
 ******************************************/
 #if defined(_WIN32)   /* Windows */
+
     #define UTIL_TIME_INITIALIZER { { 0, 0 } }
     typedef LARGE_INTEGER UTIL_time_t;
 
-    UTIL_STATIC UTIL_time_t UTIL_getTime(void) { UTIL_time_t x; QueryPerformanceCounter(&x); return x; }
-    UTIL_STATIC U64 UTIL_getSpanTimeMicro(UTIL_time_t clockStart, UTIL_time_t clockEnd)
-    {
-        static LARGE_INTEGER ticksPerSecond;
-        static int init = 0;
-        if (!init) {
-            if (!QueryPerformanceFrequency(&ticksPerSecond))
-                UTIL_DISPLAYLEVEL(1, "ERROR: QueryPerformanceFrequency() failure\n");
-            init = 1;
-        }
-        return 1000000ULL*(clockEnd.QuadPart - clockStart.QuadPart)/ticksPerSecond.QuadPart;
-    }
-    UTIL_STATIC U64 UTIL_getSpanTimeNano(UTIL_time_t clockStart, UTIL_time_t clockEnd)
-    {
-        static LARGE_INTEGER ticksPerSecond;
-        static int init = 0;
-        if (!init) {
-            if (!QueryPerformanceFrequency(&ticksPerSecond))
-                UTIL_DISPLAYLEVEL(1, "ERROR: QueryPerformanceFrequency() failure\n");
-            init = 1;
-        }
-        return 1000000000ULL*(clockEnd.QuadPart - clockStart.QuadPart)/ticksPerSecond.QuadPart;
-    }
-
 #elif defined(__APPLE__) && defined(__MACH__)
 
     #include 
     #define UTIL_TIME_INITIALIZER 0
     typedef U64 UTIL_time_t;
 
-    UTIL_STATIC UTIL_time_t UTIL_getTime(void) { return mach_absolute_time(); }
-    UTIL_STATIC U64 UTIL_getSpanTimeMicro(UTIL_time_t clockStart, UTIL_time_t clockEnd)
-    {
-        static mach_timebase_info_data_t rate;
-        static int init = 0;
-        if (!init) {
-            mach_timebase_info(&rate);
-            init = 1;
-        }
-        return (((clockEnd - clockStart) * (U64)rate.numer) / ((U64)rate.denom))/1000ULL;
-    }
-    UTIL_STATIC U64 UTIL_getSpanTimeNano(UTIL_time_t clockStart, UTIL_time_t clockEnd)
-    {
-        static mach_timebase_info_data_t rate;
-        static int init = 0;
-        if (!init) {
-            mach_timebase_info(&rate);
-            init = 1;
-        }
-        return ((clockEnd - clockStart) * (U64)rate.numer) / ((U64)rate.denom);
-    }
-
 #elif (PLATFORM_POSIX_VERSION >= 200112L) \
    && (defined(__UCLIBC__)                \
       || (defined(__GLIBC__)              \
@@ -184,79 +136,27 @@ static int g_utilDisplayLevel;
     typedef struct timespec UTIL_freq_t;
     typedef struct timespec UTIL_time_t;
 
-    UTIL_STATIC UTIL_time_t UTIL_getTime(void)
-    {
-        UTIL_time_t time;
-        if (clock_gettime(CLOCK_MONOTONIC, &time))
-            UTIL_DISPLAYLEVEL(1, "ERROR: Failed to get time\n");   /* we could also exit() */
-        return time;
-    }
-
-    UTIL_STATIC UTIL_time_t UTIL_getSpanTime(UTIL_time_t begin, UTIL_time_t end)
-    {
-        UTIL_time_t diff;
-        if (end.tv_nsec < begin.tv_nsec) {
-            diff.tv_sec = (end.tv_sec - 1) - begin.tv_sec;
-            diff.tv_nsec = (end.tv_nsec + 1000000000ULL) - begin.tv_nsec;
-        } else {
-            diff.tv_sec = end.tv_sec - begin.tv_sec;
-            diff.tv_nsec = end.tv_nsec - begin.tv_nsec;
-        }
-        return diff;
-    }
-
-    UTIL_STATIC U64 UTIL_getSpanTimeMicro(UTIL_time_t begin, UTIL_time_t end)
-    {
-        UTIL_time_t const diff = UTIL_getSpanTime(begin, end);
-        U64 micro = 0;
-        micro += 1000000ULL * diff.tv_sec;
-        micro += diff.tv_nsec / 1000ULL;
-        return micro;
-    }
-
-    UTIL_STATIC U64 UTIL_getSpanTimeNano(UTIL_time_t begin, UTIL_time_t end)
-    {
-        UTIL_time_t const diff = UTIL_getSpanTime(begin, end);
-        U64 nano = 0;
-        nano += 1000000000ULL * diff.tv_sec;
-        nano += diff.tv_nsec;
-        return nano;
-    }
+    UTIL_time_t UTIL_getSpanTime(UTIL_time_t begin, UTIL_time_t end);
 
 #else   /* relies on standard C (note : clock_t measurements can be wrong when using multi-threading) */
+
     typedef clock_t UTIL_time_t;
     #define UTIL_TIME_INITIALIZER 0
-    UTIL_STATIC UTIL_time_t UTIL_getTime(void) { return clock(); }
-    UTIL_STATIC U64 UTIL_getSpanTimeMicro(UTIL_time_t clockStart, UTIL_time_t clockEnd) { return 1000000ULL * (clockEnd - clockStart) / CLOCKS_PER_SEC; }
-    UTIL_STATIC U64 UTIL_getSpanTimeNano(UTIL_time_t clockStart, UTIL_time_t clockEnd) { return 1000000000ULL * (clockEnd - clockStart) / CLOCKS_PER_SEC; }
+
 #endif
 
+UTIL_time_t UTIL_getTime(void);
+U64 UTIL_getSpanTimeMicro(UTIL_time_t clockStart, UTIL_time_t clockEnd);
+U64 UTIL_getSpanTimeNano(UTIL_time_t clockStart, UTIL_time_t clockEnd);
+
 #define SEC_TO_MICRO 1000000
 
 /* returns time span in microseconds */
-UTIL_STATIC U64 UTIL_clockSpanMicro(UTIL_time_t clockStart )
-{
-    UTIL_time_t const clockEnd = UTIL_getTime();
-    return UTIL_getSpanTimeMicro(clockStart, clockEnd);
-}
+U64 UTIL_clockSpanMicro(UTIL_time_t clockStart);
 
 /* returns time span in microseconds */
-UTIL_STATIC U64 UTIL_clockSpanNano(UTIL_time_t clockStart )
-{
-    UTIL_time_t const clockEnd = UTIL_getTime();
-    return UTIL_getSpanTimeNano(clockStart, clockEnd);
-}
-
-UTIL_STATIC void UTIL_waitForNextTick(void)
-{
-    UTIL_time_t const clockStart = UTIL_getTime();
-    UTIL_time_t clockEnd;
-    do {
-        clockEnd = UTIL_getTime();
-    } while (UTIL_getSpanTimeNano(clockStart, clockEnd) == 0);
-}
-
-
+U64 UTIL_clockSpanNano(UTIL_time_t clockStart);
+void UTIL_waitForNextTick(void);
 
 /*-****************************************
 *  File functions
@@ -269,129 +169,23 @@ UTIL_STATIC void UTIL_waitForNextTick(void)
 #endif
 
 
-UTIL_STATIC int UTIL_isRegularFile(const char* infilename);
-
-
-UTIL_STATIC int UTIL_setFileStat(const char *filename, stat_t *statbuf)
-{
-    int res = 0;
-    struct utimbuf timebuf;
-
-    if (!UTIL_isRegularFile(filename))
-        return -1;
-
-    timebuf.actime = time(NULL);
-    timebuf.modtime = statbuf->st_mtime;
-    res += utime(filename, &timebuf);  /* set access and modification times */
-
-#if !defined(_WIN32)
-    res += chown(filename, statbuf->st_uid, statbuf->st_gid);  /* Copy ownership */
-#endif
-
-    res += chmod(filename, statbuf->st_mode & 07777);  /* Copy file permissions */
-
-    errno = 0;
-    return -res; /* number of errors is returned */
-}
-
-
-UTIL_STATIC int UTIL_getFileStat(const char* infilename, stat_t *statbuf)
-{
-    int r;
-#if defined(_MSC_VER)
-    r = _stat64(infilename, statbuf);
-    if (r || !(statbuf->st_mode & S_IFREG)) return 0;   /* No good... */
-#else
-    r = stat(infilename, statbuf);
-    if (r || !S_ISREG(statbuf->st_mode)) return 0;   /* No good... */
-#endif
-    return 1;
-}
-
-
-UTIL_STATIC int UTIL_isRegularFile(const char* infilename)
-{
-    stat_t statbuf;
-    return UTIL_getFileStat(infilename, &statbuf); /* Only need to know whether it is a regular file */
-}
-
-
-UTIL_STATIC U32 UTIL_isDirectory(const char* infilename)
-{
-    int r;
-    stat_t statbuf;
-#if defined(_MSC_VER)
-    r = _stat64(infilename, &statbuf);
-    if (!r && (statbuf.st_mode & _S_IFDIR)) return 1;
-#else
-    r = stat(infilename, &statbuf);
-    if (!r && S_ISDIR(statbuf.st_mode)) return 1;
-#endif
-    return 0;
-}
-
-UTIL_STATIC U32 UTIL_isLink(const char* infilename)
-{
-/* macro guards, as defined in : https://linux.die.net/man/2/lstat */
-#ifndef __STRICT_ANSI__
-#if defined(_BSD_SOURCE) \
-    || (defined(_XOPEN_SOURCE) && (_XOPEN_SOURCE >= 500)) \
-    || (defined(_XOPEN_SOURCE) && defined(_XOPEN_SOURCE_EXTENDED)) \
-    || (defined(_POSIX_C_SOURCE) && (_POSIX_C_SOURCE >= 200112L)) \
-    || (defined(__APPLE__) && defined(__MACH__))
-    int r;
-    stat_t statbuf;
-    r = lstat(infilename, &statbuf);
-    if (!r && S_ISLNK(statbuf.st_mode)) return 1;
-#endif
-#endif
-    (void)infilename;
-    return 0;
-}
-
+int UTIL_fileExist(const char* filename);
+int UTIL_isRegularFile(const char* infilename);
+int UTIL_setFileStat(const char* filename, stat_t* statbuf);
+U32 UTIL_isDirectory(const char* infilename);
+int UTIL_getFileStat(const char* infilename, stat_t* statbuf);
 
+U32 UTIL_isLink(const char* infilename);
 #define UTIL_FILESIZE_UNKNOWN  ((U64)(-1))
-UTIL_STATIC U64 UTIL_getFileSize(const char* infilename)
-{
-    if (!UTIL_isRegularFile(infilename)) return UTIL_FILESIZE_UNKNOWN;
-    {   int r;
-#if defined(_MSC_VER)
-        struct __stat64 statbuf;
-        r = _stat64(infilename, &statbuf);
-        if (r || !(statbuf.st_mode & S_IFREG)) return UTIL_FILESIZE_UNKNOWN;
-#elif defined(__MINGW32__) && defined (__MSVCRT__)
-        struct _stati64 statbuf;
-        r = _stati64(infilename, &statbuf);
-        if (r || !(statbuf.st_mode & S_IFREG)) return UTIL_FILESIZE_UNKNOWN;
-#else
-        struct stat statbuf;
-        r = stat(infilename, &statbuf);
-        if (r || !S_ISREG(statbuf.st_mode)) return UTIL_FILESIZE_UNKNOWN;
-#endif
-        return (U64)statbuf.st_size;
-    }
-}
-
-
-UTIL_STATIC U64 UTIL_getTotalFileSize(const char* const * const fileNamesTable, unsigned nbFiles)
-{
-    U64 total = 0;
-    int error = 0;
-    unsigned n;
-    for (n=0; n= *bufEnd) {
-                ptrdiff_t newListSize = (*bufEnd - *bufStart) + LIST_SIZE_INCREASE;
-                *bufStart = (char*)UTIL_realloc(*bufStart, newListSize);
-                *bufEnd = *bufStart + newListSize;
-                if (*bufStart == NULL) { free(path); FindClose(hFile); return 0; }
-            }
-            if (*bufStart + *pos + pathLength < *bufEnd) {
-                strncpy(*bufStart + *pos, path, *bufEnd - (*bufStart + *pos));
-                *pos += pathLength + 1;
-                nbFiles++;
-            }
-        }
-        free(path);
-    } while (FindNextFileA(hFile, &cFile));
-
-    FindClose(hFile);
-    return nbFiles;
-}
-
 #elif defined(__linux__) || (PLATFORM_POSIX_VERSION >= 200112L)  /* opendir, readdir require POSIX.1-2001 */
 #  define UTIL_HAS_CREATEFILELIST
 #  include        /* opendir, readdir */
 #  include        /* strerror, memcpy */
-
-UTIL_STATIC int UTIL_prepareFileList(const char *dirName, char** bufStart, size_t* pos, char** bufEnd, int followLinks)
-{
-    DIR *dir;
-    struct dirent *entry;
-    char* path;
-    int dirLength, fnameLength, pathLength, nbFiles = 0;
-
-    if (!(dir = opendir(dirName))) {
-        UTIL_DISPLAYLEVEL(1, "Cannot open directory '%s': %s\n", dirName, strerror(errno));
-        return 0;
-    }
-
-    dirLength = (int)strlen(dirName);
-    errno = 0;
-    while ((entry = readdir(dir)) != NULL) {
-        if (strcmp (entry->d_name, "..") == 0 ||
-            strcmp (entry->d_name, ".") == 0) continue;
-        fnameLength = (int)strlen(entry->d_name);
-        path = (char*) malloc(dirLength + fnameLength + 2);
-        if (!path) { closedir(dir); return 0; }
-        memcpy(path, dirName, dirLength);
-
-        path[dirLength] = '/';
-        memcpy(path+dirLength+1, entry->d_name, fnameLength);
-        pathLength = dirLength+1+fnameLength;
-        path[pathLength] = 0;
-
-        if (!followLinks && UTIL_isLink(path)) {
-            UTIL_DISPLAYLEVEL(2, "Warning : %s is a symbolic link, ignoring\n", path);
-            continue;
-        }
-
-        if (UTIL_isDirectory(path)) {
-            nbFiles += UTIL_prepareFileList(path, bufStart, pos, bufEnd, followLinks);  /* Recursively call "UTIL_prepareFileList" with the new path. */
-            if (*bufStart == NULL) { free(path); closedir(dir); return 0; }
-        } else {
-            if (*bufStart + *pos + pathLength >= *bufEnd) {
-                ptrdiff_t newListSize = (*bufEnd - *bufStart) + LIST_SIZE_INCREASE;
-                *bufStart = (char*)UTIL_realloc(*bufStart, newListSize);
-                *bufEnd = *bufStart + newListSize;
-                if (*bufStart == NULL) { free(path); closedir(dir); return 0; }
-            }
-            if (*bufStart + *pos + pathLength < *bufEnd) {
-                strncpy(*bufStart + *pos, path, *bufEnd - (*bufStart + *pos));
-                *pos += pathLength + 1;
-                nbFiles++;
-            }
-        }
-        free(path);
-        errno = 0; /* clear errno after UTIL_isDirectory, UTIL_prepareFileList */
-    }
-
-    if (errno != 0) {
-        UTIL_DISPLAYLEVEL(1, "readdir(%s) error: %s\n", dirName, strerror(errno));
-        free(*bufStart);
-        *bufStart = NULL;
-    }
-    closedir(dir);
-    return nbFiles;
-}
-
 #else
-
-UTIL_STATIC int UTIL_prepareFileList(const char *dirName, char** bufStart, size_t* pos, char** bufEnd, int followLinks)
-{
-    (void)bufStart; (void)bufEnd; (void)pos; (void)followLinks;
-    UTIL_DISPLAYLEVEL(1, "Directory %s ignored (compiled without _WIN32 or _POSIX_C_SOURCE)\n", dirName);
-    return 0;
-}
-
 #endif /* #ifdef _WIN32 */
 
 /*
@@ -544,56 +209,10 @@ UTIL_STATIC int UTIL_prepareFileList(const char *dirName, char** bufStart, size_
  * After finishing usage of the list the structures should be freed with UTIL_freeFileList(params: return value, allocatedBuffer)
  * In case of error UTIL_createFileList returns NULL and UTIL_freeFileList should not be called.
  */
-UTIL_STATIC const char**
+const char**
 UTIL_createFileList(const char **inputNames, unsigned inputNamesNb,
                     char** allocatedBuffer, unsigned* allocatedNamesNb,
-                    int followLinks)
-{
-    size_t pos;
-    unsigned i, nbFiles;
-    char* buf = (char*)malloc(LIST_SIZE_INCREASE);
-    char* bufend = buf + LIST_SIZE_INCREASE;
-    const char** fileTable;
-
-    if (!buf) return NULL;
-
-    for (i=0, pos=0, nbFiles=0; i= bufend) {
-                ptrdiff_t newListSize = (bufend - buf) + LIST_SIZE_INCREASE;
-                buf = (char*)UTIL_realloc(buf, newListSize);
-                bufend = buf + newListSize;
-                if (!buf) return NULL;
-            }
-            if (buf + pos + len < bufend) {
-                strncpy(buf + pos, inputNames[i], bufend - (buf + pos));
-                pos += len + 1;
-                nbFiles++;
-            }
-        } else {
-            nbFiles += UTIL_prepareFileList(inputNames[i], &buf, &pos, &bufend, followLinks);
-            if (buf == NULL) return NULL;
-    }   }
-
-    if (nbFiles == 0) { free(buf); return NULL; }
-
-    fileTable = (const char**)malloc((nbFiles+1) * sizeof(const char*));
-    if (!fileTable) { free(buf); return NULL; }
-
-    for (i=0, pos=0; i bufend) { free(buf); free((void*)fileTable); return NULL; }
-
-    *allocatedBuffer = buf;
-    *allocatedNamesNb = nbFiles;
-
-    return fileTable;
-}
-
+                    int followLinks);
 
 UTIL_STATIC void UTIL_freeFileList(const char** filenameTable, char* allocatedBuffer)
 {
@@ -601,201 +220,7 @@ UTIL_STATIC void UTIL_freeFileList(const char** filenameTable, char* allocatedBu
     if (filenameTable) free((void*)filenameTable);
 }
 
-/* count the number of physical cores */
-#if defined(_WIN32) || defined(WIN32)
-
-#include 
-
-typedef BOOL(WINAPI* LPFN_GLPI)(PSYSTEM_LOGICAL_PROCESSOR_INFORMATION, PDWORD);
-
-UTIL_STATIC int UTIL_countPhysicalCores(void)
-{
-    static int numPhysicalCores = 0;
-    if (numPhysicalCores != 0) return numPhysicalCores;
-
-    {   LPFN_GLPI glpi;
-        BOOL done = FALSE;
-        PSYSTEM_LOGICAL_PROCESSOR_INFORMATION buffer = NULL;
-        PSYSTEM_LOGICAL_PROCESSOR_INFORMATION ptr = NULL;
-        DWORD returnLength = 0;
-        size_t byteOffset = 0;
-
-        glpi = (LPFN_GLPI)GetProcAddress(GetModuleHandle(TEXT("kernel32")),
-                                         "GetLogicalProcessorInformation");
-
-        if (glpi == NULL) {
-            goto failed;
-        }
-
-        while(!done) {
-            DWORD rc = glpi(buffer, &returnLength);
-            if (FALSE == rc) {
-                if (GetLastError() == ERROR_INSUFFICIENT_BUFFER) {
-                    if (buffer)
-                        free(buffer);
-                    buffer = (PSYSTEM_LOGICAL_PROCESSOR_INFORMATION)malloc(returnLength);
-
-                    if (buffer == NULL) {
-                        perror("zstd");
-                        exit(1);
-                    }
-                } else {
-                    /* some other error */
-                    goto failed;
-                }
-            } else {
-                done = TRUE;
-            }
-        }
-
-        ptr = buffer;
-
-        while (byteOffset + sizeof(SYSTEM_LOGICAL_PROCESSOR_INFORMATION) <= returnLength) {
-
-            if (ptr->Relationship == RelationProcessorCore) {
-                numPhysicalCores++;
-            }
-
-            ptr++;
-            byteOffset += sizeof(SYSTEM_LOGICAL_PROCESSOR_INFORMATION);
-        }
-
-        free(buffer);
-
-        return numPhysicalCores;
-    }
-
-failed:
-    /* try to fall back on GetSystemInfo */
-    {   SYSTEM_INFO sysinfo;
-        GetSystemInfo(&sysinfo);
-        numPhysicalCores = sysinfo.dwNumberOfProcessors;
-        if (numPhysicalCores == 0) numPhysicalCores = 1; /* just in case */
-    }
-    return numPhysicalCores;
-}
-
-#elif defined(__APPLE__)
-
-#include 
-
-/* Use apple-provided syscall
- * see: man 3 sysctl */
-UTIL_STATIC int UTIL_countPhysicalCores(void)
-{
-    static S32 numPhysicalCores = 0; /* apple specifies int32_t */
-    if (numPhysicalCores != 0) return numPhysicalCores;
-
-    {   size_t size = sizeof(S32);
-        int const ret = sysctlbyname("hw.physicalcpu", &numPhysicalCores, &size, NULL, 0);
-        if (ret != 0) {
-            if (errno == ENOENT) {
-                /* entry not present, fall back on 1 */
-                numPhysicalCores = 1;
-            } else {
-                perror("zstd: can't get number of physical cpus");
-                exit(1);
-            }
-        }
-
-        return numPhysicalCores;
-    }
-}
-
-#elif defined(__linux__)
-
-/* parse /proc/cpuinfo
- * siblings / cpu cores should give hyperthreading ratio
- * otherwise fall back on sysconf */
-UTIL_STATIC int UTIL_countPhysicalCores(void)
-{
-    static int numPhysicalCores = 0;
-
-    if (numPhysicalCores != 0) return numPhysicalCores;
-
-    numPhysicalCores = (int)sysconf(_SC_NPROCESSORS_ONLN);
-    if (numPhysicalCores == -1) {
-        /* value not queryable, fall back on 1 */
-        return numPhysicalCores = 1;
-    }
-
-    /* try to determine if there's hyperthreading */
-    {   FILE* const cpuinfo = fopen("/proc/cpuinfo", "r");
-#define BUF_SIZE 80
-        char buff[BUF_SIZE];
-
-        int siblings = 0;
-        int cpu_cores = 0;
-        int ratio = 1;
-
-        if (cpuinfo == NULL) {
-            /* fall back on the sysconf value */
-            return numPhysicalCores;
-        }
-
-        /* assume the cpu cores/siblings values will be constant across all
-         * present processors */
-        while (!feof(cpuinfo)) {
-            if (fgets(buff, BUF_SIZE, cpuinfo) != NULL) {
-                if (strncmp(buff, "siblings", 8) == 0) {
-                    const char* const sep = strchr(buff, ':');
-                    if (*sep == '\0') {
-                        /* formatting was broken? */
-                        goto failed;
-                    }
-
-                    siblings = atoi(sep + 1);
-                }
-                if (strncmp(buff, "cpu cores", 9) == 0) {
-                    const char* const sep = strchr(buff, ':');
-                    if (*sep == '\0') {
-                        /* formatting was broken? */
-                        goto failed;
-                    }
-
-                    cpu_cores = atoi(sep + 1);
-                }
-            } else if (ferror(cpuinfo)) {
-                /* fall back on the sysconf value */
-                goto failed;
-            }
-        }
-        if (siblings && cpu_cores) {
-            ratio = siblings / cpu_cores;
-        }
-failed:
-        fclose(cpuinfo);
-        return numPhysicalCores = numPhysicalCores / ratio;
-    }
-}
-
-#elif defined(__FreeBSD__) || defined(__NetBSD__) || defined(__OpenBSD__) || defined(__DragonFly__)
-
-/* Use apple-provided syscall
- * see: man 3 sysctl */
-UTIL_STATIC int UTIL_countPhysicalCores(void)
-{
-    static int numPhysicalCores = 0;
-
-    if (numPhysicalCores != 0) return numPhysicalCores;
-
-    numPhysicalCores = (int)sysconf(_SC_NPROCESSORS_ONLN);
-    if (numPhysicalCores == -1) {
-        /* value not queryable, fall back on 1 */
-        return numPhysicalCores = 1;
-    }
-    return numPhysicalCores;
-}
-
-#else
-
-UTIL_STATIC int UTIL_countPhysicalCores(void)
-{
-    /* assume 1 */
-    return 1;
-}
-
-#endif
+int UTIL_countPhysicalCores(void);
 
 #if defined (__cplusplus)
 }
diff --git a/programs/zstd.1 b/programs/zstd.1
index 674f89841ce8..c93755f877ed 100644
--- a/programs/zstd.1
+++ b/programs/zstd.1
@@ -1,5 +1,5 @@
 .
-.TH "ZSTD" "1" "October 2018" "zstd 1.3.7" "User Commands"
+.TH "ZSTD" "1" "December 2018" "zstd 1.3.8" "User Commands"
 .
 .SH "NAME"
 \fBzstd\fR \- zstd, zstdmt, unzstd, zstdcat \- Compress or decompress \.zst files
@@ -127,6 +127,10 @@ Does not spawn a thread for compression, use a single thread for both I/O and co
 \fBzstd\fR will dynamically adapt compression level to perceived I/O conditions\. Compression level adaptation can be observed live by using command \fB\-v\fR\. Adaptation can be constrained between supplied \fBmin\fR and \fBmax\fR levels\. The feature works when combined with multi\-threading and \fB\-\-long\fR mode\. It does not work with \fB\-\-single\-thread\fR\. It sets window size to 8 MB by default (can be changed manually, see \fBwlog\fR)\. Due to the chaotic nature of dynamic adaptation, compressed result is not reproducible\. \fInote\fR : at the time of this writing, \fB\-\-adapt\fR can remain stuck at low speed when combined with multiple worker threads (>=2)\.
 .
 .TP
+\fB\-\-rsyncable\fR
+\fBzstd\fR will periodically synchronize the compression state to make the compressed file more rsync\-friendly\. There is a negligible impact to compression ratio, and the faster compression levels will see a small compression speed hit\. This feature does not work with \fB\-\-single\-thread\fR\. You probably don\'t want to use it with long range mode, since it will decrease the effectiveness of the synchronization points, but your milage may vary\.
+.
+.TP
 \fB\-D file\fR
 use \fBfile\fR as Dictionary to compress or decompress FILE(s)
 .
@@ -312,7 +316,7 @@ set process priority to real\-time
 Specify a strategy used by a match finder\.
 .
 .IP
-There are 8 strategies numbered from 1 to 8, from faster to stronger: 1=ZSTD_fast, 2=ZSTD_dfast, 3=ZSTD_greedy, 4=ZSTD_lazy, 5=ZSTD_lazy2, 6=ZSTD_btlazy2, 7=ZSTD_btopt, 8=ZSTD_btultra\.
+There are 9 strategies numbered from 1 to 9, from faster to stronger: 1=ZSTD_fast, 2=ZSTD_dfast, 3=ZSTD_greedy, 4=ZSTD_lazy, 5=ZSTD_lazy2, 6=ZSTD_btlazy2, 7=ZSTD_btopt, 8=ZSTD_btultra, 9=ZSTD_btultra2\.
 .
 .TP
 \fBwindowLog\fR=\fIwlog\fR, \fBwlog\fR=\fIwlog\fR
@@ -355,21 +359,21 @@ More searches increases the chance to find a match which usually increases compr
 The minimum \fIslog\fR is 1 and the maximum is 26\.
 .
 .TP
-\fBsearchLength\fR=\fIslen\fR, \fBslen\fR=\fIslen\fR
+\fBminMatch\fR=\fImml\fR, \fBmml\fR=\fImml\fR
 Specify the minimum searched length of a match in a hash table\.
 .
 .IP
 Larger search lengths usually decrease compression ratio but improve decompression speed\.
 .
 .IP
-The minimum \fIslen\fR is 3 and the maximum is 7\.
+The minimum \fImml\fR is 3 and the maximum is 7\.
 .
 .TP
 \fBtargetLen\fR=\fItlen\fR, \fBtlen\fR=\fItlen\fR
 The impact of this field vary depending on selected strategy\.
 .
 .IP
-For ZSTD_btopt and ZSTD_btultra, it specifies the minimum match length that causes match finder to stop searching for better matches\. A larger \fBtargetLen\fR usually improves compression ratio but decreases compression speed\.
+For ZSTD_btopt, ZSTD_btultra and ZSTD_btultra2, it specifies the minimum match length that causes match finder to stop searching\. A larger \fBtargetLen\fR usually improves compression ratio but decreases compression speed\.
 .
 .IP
 For ZSTD_fast, it triggers ultra\-fast mode when > 0\. The value represents the amount of data skipped between match sampling\. Impact is reversed : a larger \fBtargetLen\fR increases compression speed but decreases compression ratio\.
@@ -385,10 +389,10 @@ The minimum \fItlen\fR is 0 and the maximum is 999\.
 Determine \fBoverlapSize\fR, amount of data reloaded from previous job\. This parameter is only available when multithreading is enabled\. Reloading more data improves compression ratio, but decreases speed\.
 .
 .IP
-The minimum \fIovlog\fR is 0, and the maximum is 9\. 0 means "no overlap", hence completely independent jobs\. 9 means "full overlap", meaning up to \fBwindowSize\fR is reloaded from previous job\. Reducing \fIovlog\fR by 1 reduces the amount of reload by a factor 2\. Default \fIovlog\fR is 6, which means "reload \fBwindowSize / 8\fR"\. Exception : the maximum compression level (22) has a default \fIovlog\fR of 9\.
+The minimum \fIovlog\fR is 0, and the maximum is 9\. 1 means "no overlap", hence completely independent jobs\. 9 means "full overlap", meaning up to \fBwindowSize\fR is reloaded from previous job\. Reducing \fIovlog\fR by 1 reduces the reloaded amount by a factor 2\. For example, 8 means "windowSize/2", and 6 means "windowSize/8"\. Value 0 is special and means "default" : \fIovlog\fR is automatically determined by \fBzstd\fR\. In which case, \fIovlog\fR will range from 6 to 9, depending on selected \fIstrat\fR\.
 .
 .TP
-\fBldmHashLog\fR=\fIldmhlog\fR, \fBldmhlog\fR=\fIldmhlog\fR
+\fBldmHashLog\fR=\fIlhlog\fR, \fBlhlog\fR=\fIlhlog\fR
 Specify the maximum size for a hash table used for long distance matching\.
 .
 .IP
@@ -398,10 +402,10 @@ This option is ignored unless long distance matching is enabled\.
 Bigger hash tables usually improve compression ratio at the expense of more memory during compression and a decrease in compression speed\.
 .
 .IP
-The minimum \fIldmhlog\fR is 6 and the maximum is 26 (default: 20)\.
+The minimum \fIlhlog\fR is 6 and the maximum is 26 (default: 20)\.
 .
 .TP
-\fBldmSearchLength\fR=\fIldmslen\fR, \fBldmslen\fR=\fIldmslen\fR
+\fBldmMinMatch\fR=\fIlmml\fR, \fBlmml\fR=\fIlmml\fR
 Specify the minimum searched length of a match for long distance matching\.
 .
 .IP
@@ -411,10 +415,10 @@ This option is ignored unless long distance matching is enabled\.
 Larger/very small values usually decrease compression ratio\.
 .
 .IP
-The minimum \fIldmslen\fR is 4 and the maximum is 4096 (default: 64)\.
+The minimum \fIlmml\fR is 4 and the maximum is 4096 (default: 64)\.
 .
 .TP
-\fBldmBucketSizeLog\fR=\fIldmblog\fR, \fBldmblog\fR=\fIldmblog\fR
+\fBldmBucketSizeLog\fR=\fIlblog\fR, \fBlblog\fR=\fIlblog\fR
 Specify the size of each bucket for the hash table used for long distance matching\.
 .
 .IP
@@ -424,10 +428,10 @@ This option is ignored unless long distance matching is enabled\.
 Larger bucket sizes improve collision resolution but decrease compression speed\.
 .
 .IP
-The minimum \fIldmblog\fR is 0 and the maximum is 8 (default: 3)\.
+The minimum \fIlblog\fR is 0 and the maximum is 8 (default: 3)\.
 .
 .TP
-\fBldmHashEveryLog\fR=\fIldmhevery\fR, \fBldmhevery\fR=\fIldmhevery\fR
+\fBldmHashRateLog\fR=\fIlhrlog\fR, \fBlhrlog\fR=\fIlhrlog\fR
 Specify the frequency of inserting entries into the long distance matching hash table\.
 .
 .IP
@@ -437,13 +441,13 @@ This option is ignored unless long distance matching is enabled\.
 Larger values will improve compression speed\. Deviating far from the default value will likely result in a decrease in compression ratio\.
 .
 .IP
-The default value is \fBwlog \- ldmhlog\fR\.
+The default value is \fBwlog \- lhlog\fR\.
 .
 .SS "Example"
 The following parameters sets advanced compression options to something similar to predefined level 19 for files bigger than 256 KB:
 .
 .P
-\fB\-\-zstd\fR=wlog=23,clog=23,hlog=22,slog=6,slen=3,tlen=48,strat=6
+\fB\-\-zstd\fR=wlog=23,clog=23,hlog=22,slog=6,mml=3,tlen=48,strat=6
 .
 .SS "\-B#:"
 Select the size of each compression job\. This parameter is available only when multi\-threading is enabled\. Default value is \fB4 * windowSize\fR, which means it varies depending on compression level\. \fB\-B#\fR makes it possible to select a custom value\. Note that job size must respect a minimum value which is enforced transparently\. This minimum is either 1 MB, or \fBoverlapSize\fR, whichever is largest\.
diff --git a/programs/zstd.1.md b/programs/zstd.1.md
index c0c04698ddc2..a029af5ff3ce 100644
--- a/programs/zstd.1.md
+++ b/programs/zstd.1.md
@@ -144,6 +144,14 @@ the last one takes effect.
     Due to the chaotic nature of dynamic adaptation, compressed result is not reproducible.
     _note_ : at the time of this writing, `--adapt` can remain stuck at low speed
     when combined with multiple worker threads (>=2).
+* `--rsyncable` :
+    `zstd` will periodically synchronize the compression state to make the
+    compressed file more rsync-friendly. There is a negligible impact to
+    compression ratio, and the faster compression levels will see a small
+    compression speed hit.
+    This feature does not work with `--single-thread`. You probably don't want
+    to use it with long range mode, since it will decrease the effectiveness of
+    the synchronization points, but your milage may vary.
 * `-D file`:
     use `file` as Dictionary to compress or decompress FILE(s)
 * `--no-dictID`:
@@ -187,6 +195,8 @@ the last one takes effect.
 * `-q`, `--quiet`:
     suppress warnings, interactivity, and notifications.
     specify twice to suppress errors too.
+* `--no-progress`:
+    do not display the progress bar, but keep all other messages.
 * `-C`, `--[no-]check`:
     add integrity check computed from uncompressed data (default: enabled)
 * `--`:
@@ -331,9 +341,10 @@ The list of available _options_:
 - `strategy`=_strat_, `strat`=_strat_:
     Specify a strategy used by a match finder.
 
-    There are 8 strategies numbered from 1 to 8, from faster to stronger:
-    1=ZSTD\_fast, 2=ZSTD\_dfast, 3=ZSTD\_greedy, 4=ZSTD\_lazy,
-    5=ZSTD\_lazy2, 6=ZSTD\_btlazy2, 7=ZSTD\_btopt, 8=ZSTD\_btultra.
+    There are 9 strategies numbered from 1 to 9, from faster to stronger:
+    1=ZSTD\_fast, 2=ZSTD\_dfast, 3=ZSTD\_greedy,
+    4=ZSTD\_lazy, 5=ZSTD\_lazy2, 6=ZSTD\_btlazy2,
+    7=ZSTD\_btopt, 8=ZSTD\_btultra, 9=ZSTD\_btultra2.
 
 - `windowLog`=_wlog_, `wlog`=_wlog_:
     Specify the maximum number of bits for a match distance.
@@ -375,19 +386,19 @@ The list of available _options_:
 
     The minimum _slog_ is 1 and the maximum is 26.
 
-- `searchLength`=_slen_, `slen`=_slen_:
+- `minMatch`=_mml_, `mml`=_mml_:
     Specify the minimum searched length of a match in a hash table.
 
     Larger search lengths usually decrease compression ratio but improve
     decompression speed.
 
-    The minimum _slen_ is 3 and the maximum is 7.
+    The minimum _mml_ is 3 and the maximum is 7.
 
 - `targetLen`=_tlen_, `tlen`=_tlen_:
     The impact of this field vary depending on selected strategy.
 
-    For ZSTD\_btopt and ZSTD\_btultra, it specifies the minimum match length
-    that causes match finder to stop searching for better matches.
+    For ZSTD\_btopt, ZSTD\_btultra and ZSTD\_btultra2, it specifies
+    the minimum match length that causes match finder to stop searching.
     A larger `targetLen` usually improves compression ratio
     but decreases compression speed.
 
@@ -406,13 +417,14 @@ The list of available _options_:
     Reloading more data improves compression ratio, but decreases speed.
 
     The minimum _ovlog_ is 0, and the maximum is 9.
-    0 means "no overlap", hence completely independent jobs.
+    1 means "no overlap", hence completely independent jobs.
     9 means "full overlap", meaning up to `windowSize` is reloaded from previous job.
-    Reducing _ovlog_ by 1 reduces the amount of reload by a factor 2.
-    Default _ovlog_ is 6, which means "reload `windowSize / 8`".
-    Exception : the maximum compression level (22) has a default _ovlog_ of 9.
+    Reducing _ovlog_ by 1 reduces the reloaded amount by a factor 2.
+    For example, 8 means "windowSize/2", and 6 means "windowSize/8".
+    Value 0 is special and means "default" : _ovlog_ is automatically determined by `zstd`.
+    In which case, _ovlog_ will range from 6 to 9, depending on selected _strat_.
 
-- `ldmHashLog`=_ldmhlog_, `ldmhlog`=_ldmhlog_:
+- `ldmHashLog`=_lhlog_, `lhlog`=_lhlog_:
     Specify the maximum size for a hash table used for long distance matching.
 
     This option is ignored unless long distance matching is enabled.
@@ -420,18 +432,18 @@ The list of available _options_:
     Bigger hash tables usually improve compression ratio at the expense of more
     memory during compression and a decrease in compression speed.
 
-    The minimum _ldmhlog_ is 6 and the maximum is 26 (default: 20).
+    The minimum _lhlog_ is 6 and the maximum is 26 (default: 20).
 
-- `ldmSearchLength`=_ldmslen_, `ldmslen`=_ldmslen_:
+- `ldmMinMatch`=_lmml_, `lmml`=_lmml_:
     Specify the minimum searched length of a match for long distance matching.
 
     This option is ignored unless long distance matching is enabled.
 
     Larger/very small values usually decrease compression ratio.
 
-    The minimum _ldmslen_ is 4 and the maximum is 4096 (default: 64).
+    The minimum _lmml_ is 4 and the maximum is 4096 (default: 64).
 
-- `ldmBucketSizeLog`=_ldmblog_, `ldmblog`=_ldmblog_:
+- `ldmBucketSizeLog`=_lblog_, `lblog`=_lblog_:
     Specify the size of each bucket for the hash table used for long distance
     matching.
 
@@ -440,9 +452,9 @@ The list of available _options_:
     Larger bucket sizes improve collision resolution but decrease compression
     speed.
 
-    The minimum _ldmblog_ is 0 and the maximum is 8 (default: 3).
+    The minimum _lblog_ is 0 and the maximum is 8 (default: 3).
 
-- `ldmHashEveryLog`=_ldmhevery_, `ldmhevery`=_ldmhevery_:
+- `ldmHashRateLog`=_lhrlog_, `lhrlog`=_lhrlog_:
     Specify the frequency of inserting entries into the long distance matching
     hash table.
 
@@ -451,13 +463,13 @@ The list of available _options_:
     Larger values will improve compression speed. Deviating far from the
     default value will likely result in a decrease in compression ratio.
 
-    The default value is `wlog - ldmhlog`.
+    The default value is `wlog - lhlog`.
 
 ### Example
 The following parameters sets advanced compression options to something
 similar to predefined level 19 for files bigger than 256 KB:
 
-`--zstd`=wlog=23,clog=23,hlog=22,slog=6,slen=3,tlen=48,strat=6
+`--zstd`=wlog=23,clog=23,hlog=22,slog=6,mml=3,tlen=48,strat=6
 
 ### -B#:
 Select the size of each compression job.
diff --git a/programs/zstdcli.c b/programs/zstdcli.c
index 1545d1cac579..64dad110c7af 100644
--- a/programs/zstdcli.c
+++ b/programs/zstdcli.c
@@ -28,11 +28,12 @@
 #include "platform.h" /* IS_CONSOLE, PLATFORM_POSIX_VERSION */
 #include "util.h"     /* UTIL_HAS_CREATEFILELIST, UTIL_createFileList */
 #include     /* fprintf(), stdin, stdout, stderr */
+#include    /* getenv */
 #include    /* strcmp, strlen */
 #include     /* errno */
 #include "fileio.h"   /* stdinmark, stdoutmark, ZSTD_EXTENSION */
 #ifndef ZSTD_NOBENCH
-#  include "bench.h"  /* BMK_benchFiles */
+#  include "benchzstd.h"  /* BMK_benchFiles */
 #endif
 #ifndef ZSTD_NODICT
 #  include "dibio.h"  /* ZDICT_cover_params_t, DiB_trainFromFiles() */
@@ -81,7 +82,7 @@ static const unsigned g_defaultMaxWindowLog = 27;
 static U32 g_overlapLog = OVERLAP_LOG_DEFAULT;
 static U32 g_ldmHashLog = 0;
 static U32 g_ldmMinMatch = 0;
-static U32 g_ldmHashEveryLog = LDM_PARAM_DEFAULT;
+static U32 g_ldmHashRateLog = LDM_PARAM_DEFAULT;
 static U32 g_ldmBucketSizeLog = LDM_PARAM_DEFAULT;
 
 
@@ -143,6 +144,7 @@ static int usage_advanced(const char* programName)
 #ifdef ZSTD_MULTITHREAD
     DISPLAY( " -T#    : spawns # compression threads (default: 1, 0==# cores) \n");
     DISPLAY( " -B#    : select size of each job (default: 0==automatic) \n");
+    DISPLAY( " --rsyncable : compress using a rsync-friendly method (-B sets block size) \n");
 #endif
     DISPLAY( "--no-dictID : don't write dictID into header (dictionary compression)\n");
     DISPLAY( "--[no-]check : integrity check (default: enabled) \n");
@@ -150,7 +152,7 @@ static int usage_advanced(const char* programName)
 #ifdef UTIL_HAS_CREATEFILELIST
     DISPLAY( " -r     : operate recursively on directories \n");
 #endif
-    DISPLAY( "--format=zstd : compress files to the .zstd format (default) \n");
+    DISPLAY( "--format=zstd : compress files to the .zst format (default) \n");
 #ifdef ZSTD_GZCOMPRESS
     DISPLAY( "--format=gzip : compress files to the .gz format \n");
 #endif
@@ -170,6 +172,7 @@ static int usage_advanced(const char* programName)
 #endif
 #endif
     DISPLAY( " -M#    : Set a memory usage limit for decompression \n");
+    DISPLAY( "--no-progress : do not display the progress bar \n");
     DISPLAY( "--      : All arguments after \"--\" are treated as files \n");
 #ifndef ZSTD_NODICT
     DISPLAY( "\n");
@@ -231,32 +234,44 @@ static void errorOut(const char* msg)
     DISPLAY("%s \n", msg); exit(1);
 }
 
-/*! readU32FromChar() :
- * @return : unsigned integer value read from input in `char` format.
+/*! readU32FromCharChecked() :
+ * @return 0 if success, and store the result in *value.
  *  allows and interprets K, KB, KiB, M, MB and MiB suffix.
  *  Will also modify `*stringPtr`, advancing it to position where it stopped reading.
- *  Note : function will exit() program if digit sequence overflows */
-static unsigned readU32FromChar(const char** stringPtr)
+ * @return 1 if an overflow error occurs */
+static int readU32FromCharChecked(const char** stringPtr, unsigned* value)
 {
-    const char errorMsg[] = "error: numeric value too large";
+    static unsigned const max = (((unsigned)(-1)) / 10) - 1;
     unsigned result = 0;
     while ((**stringPtr >='0') && (**stringPtr <='9')) {
-        unsigned const max = (((unsigned)(-1)) / 10) - 1;
-        if (result > max) errorOut(errorMsg);
+        if (result > max) return 1; // overflow error
         result *= 10, result += **stringPtr - '0', (*stringPtr)++ ;
     }
     if ((**stringPtr=='K') || (**stringPtr=='M')) {
         unsigned const maxK = ((unsigned)(-1)) >> 10;
-        if (result > maxK) errorOut(errorMsg);
+        if (result > maxK) return 1; // overflow error
         result <<= 10;
         if (**stringPtr=='M') {
-            if (result > maxK) errorOut(errorMsg);
+            if (result > maxK) return 1; // overflow error
             result <<= 10;
         }
         (*stringPtr)++;  /* skip `K` or `M` */
         if (**stringPtr=='i') (*stringPtr)++;
         if (**stringPtr=='B') (*stringPtr)++;
     }
+    *value = result;
+    return 0;
+}
+
+/*! readU32FromChar() :
+ * @return : unsigned integer value read from input in `char` format.
+ *  allows and interprets K, KB, KiB, M, MB and MiB suffix.
+ *  Will also modify `*stringPtr`, advancing it to position where it stopped reading.
+ *  Note : function will exit() program if digit sequence overflows */
+static unsigned readU32FromChar(const char** stringPtr) {
+    static const char errorMsg[] = "error: numeric value too large";
+    unsigned result;
+    if (readU32FromCharChecked(stringPtr, &result)) { errorOut(errorMsg); }
     return result;
 }
 
@@ -391,7 +406,7 @@ static unsigned parseAdaptParameters(const char* stringPtr, int* adaptMinPtr, in
 
 
 /** parseCompressionParameters() :
- *  reads compression parameters from *stringPtr (e.g. "--zstd=wlog=23,clog=23,hlog=22,slog=6,slen=3,tlen=48,strat=6") into *params
+ *  reads compression parameters from *stringPtr (e.g. "--zstd=wlog=23,clog=23,hlog=22,slog=6,mml=3,tlen=48,strat=6") into *params
  *  @return 1 means that compression parameters were correct
  *  @return 0 in case of malformed parameters
  */
@@ -402,20 +417,20 @@ static unsigned parseCompressionParameters(const char* stringPtr, ZSTD_compressi
         if (longCommandWArg(&stringPtr, "chainLog=") || longCommandWArg(&stringPtr, "clog=")) { params->chainLog = readU32FromChar(&stringPtr); if (stringPtr[0]==',') { stringPtr++; continue; } else break; }
         if (longCommandWArg(&stringPtr, "hashLog=") || longCommandWArg(&stringPtr, "hlog=")) { params->hashLog = readU32FromChar(&stringPtr); if (stringPtr[0]==',') { stringPtr++; continue; } else break; }
         if (longCommandWArg(&stringPtr, "searchLog=") || longCommandWArg(&stringPtr, "slog=")) { params->searchLog = readU32FromChar(&stringPtr); if (stringPtr[0]==',') { stringPtr++; continue; } else break; }
-        if (longCommandWArg(&stringPtr, "searchLength=") || longCommandWArg(&stringPtr, "slen=")) { params->searchLength = readU32FromChar(&stringPtr); if (stringPtr[0]==',') { stringPtr++; continue; } else break; }
+        if (longCommandWArg(&stringPtr, "minMatch=") || longCommandWArg(&stringPtr, "mml=")) { params->minMatch = readU32FromChar(&stringPtr); if (stringPtr[0]==',') { stringPtr++; continue; } else break; }
         if (longCommandWArg(&stringPtr, "targetLength=") || longCommandWArg(&stringPtr, "tlen=")) { params->targetLength = readU32FromChar(&stringPtr); if (stringPtr[0]==',') { stringPtr++; continue; } else break; }
         if (longCommandWArg(&stringPtr, "strategy=") || longCommandWArg(&stringPtr, "strat=")) { params->strategy = (ZSTD_strategy)(readU32FromChar(&stringPtr)); if (stringPtr[0]==',') { stringPtr++; continue; } else break; }
         if (longCommandWArg(&stringPtr, "overlapLog=") || longCommandWArg(&stringPtr, "ovlog=")) { g_overlapLog = readU32FromChar(&stringPtr); if (stringPtr[0]==',') { stringPtr++; continue; } else break; }
-        if (longCommandWArg(&stringPtr, "ldmHashLog=") || longCommandWArg(&stringPtr, "ldmhlog=")) { g_ldmHashLog = readU32FromChar(&stringPtr); if (stringPtr[0]==',') { stringPtr++; continue; } else break; }
-        if (longCommandWArg(&stringPtr, "ldmSearchLength=") || longCommandWArg(&stringPtr, "ldmslen=")) { g_ldmMinMatch = readU32FromChar(&stringPtr); if (stringPtr[0]==',') { stringPtr++; continue; } else break; }
-        if (longCommandWArg(&stringPtr, "ldmBucketSizeLog=") || longCommandWArg(&stringPtr, "ldmblog=")) { g_ldmBucketSizeLog = readU32FromChar(&stringPtr); if (stringPtr[0]==',') { stringPtr++; continue; } else break; }
-        if (longCommandWArg(&stringPtr, "ldmHashEveryLog=") || longCommandWArg(&stringPtr, "ldmhevery=")) { g_ldmHashEveryLog = readU32FromChar(&stringPtr); if (stringPtr[0]==',') { stringPtr++; continue; } else break; }
+        if (longCommandWArg(&stringPtr, "ldmHashLog=") || longCommandWArg(&stringPtr, "lhlog=")) { g_ldmHashLog = readU32FromChar(&stringPtr); if (stringPtr[0]==',') { stringPtr++; continue; } else break; }
+        if (longCommandWArg(&stringPtr, "ldmMinMatch=") || longCommandWArg(&stringPtr, "lmml=")) { g_ldmMinMatch = readU32FromChar(&stringPtr); if (stringPtr[0]==',') { stringPtr++; continue; } else break; }
+        if (longCommandWArg(&stringPtr, "ldmBucketSizeLog=") || longCommandWArg(&stringPtr, "lblog=")) { g_ldmBucketSizeLog = readU32FromChar(&stringPtr); if (stringPtr[0]==',') { stringPtr++; continue; } else break; }
+        if (longCommandWArg(&stringPtr, "ldmHashRateLog=") || longCommandWArg(&stringPtr, "lhrlog=")) { g_ldmHashRateLog = readU32FromChar(&stringPtr); if (stringPtr[0]==',') { stringPtr++; continue; } else break; }
         DISPLAYLEVEL(4, "invalid compression parameter \n");
         return 0;
     }
 
     DISPLAYLEVEL(4, "windowLog=%d, chainLog=%d, hashLog=%d, searchLog=%d \n", params->windowLog, params->chainLog, params->hashLog, params->searchLog);
-    DISPLAYLEVEL(4, "searchLength=%d, targetLength=%d, strategy=%d \n", params->searchLength, params->targetLength, params->strategy);
+    DISPLAYLEVEL(4, "minMatch=%d, targetLength=%d, strategy=%d \n", params->minMatch, params->targetLength, params->strategy);
     if (stringPtr[0] != 0) return 0; /* check the end of string */
     return 1;
 }
@@ -450,6 +465,38 @@ static void printVersion(void)
 #endif
 }
 
+/* Environment variables for parameter setting */
+#define ENV_CLEVEL "ZSTD_CLEVEL"
+
+/* functions that pick up environment variables */
+static int init_cLevel(void) {
+    const char* const env = getenv(ENV_CLEVEL);
+    if (env) {
+        const char *ptr = env;
+        int sign = 1;
+        if (*ptr == '-') {
+            sign = -1;
+            ptr++;
+        } else if (*ptr == '+') {
+            ptr++;
+        }
+
+        if ((*ptr>='0') && (*ptr<='9')) {
+            unsigned absLevel;
+            if (readU32FromCharChecked(&ptr, &absLevel)) { 
+                DISPLAYLEVEL(2, "Ignore environment variable setting %s=%s: numeric value too large\n", ENV_CLEVEL, env);
+                return ZSTDCLI_CLEVEL_DEFAULT;
+            } else if (*ptr == 0) {
+                return sign * absLevel;
+            }
+        }
+
+        DISPLAYLEVEL(2, "Ignore environment variable setting %s=%s: not a valid integer value\n", ENV_CLEVEL, env);
+    }
+
+    return ZSTDCLI_CLEVEL_DEFAULT;
+}
+
 typedef enum { zom_compress, zom_decompress, zom_test, zom_bench, zom_train, zom_list } zstd_operation_mode;
 
 #define CLEAN_RETURN(i) { operationResult = (i); goto _end; }
@@ -475,6 +522,7 @@ int main(int argCount, const char* argv[])
         adapt = 0,
         adaptMin = MINCLEVEL,
         adaptMax = MAXCLEVEL,
+        rsyncable = 0,
         nextArgumentIsOutFileName = 0,
         nextArgumentIsMaxDict = 0,
         nextArgumentIsDictID = 0,
@@ -490,7 +538,7 @@ int main(int argCount, const char* argv[])
     size_t blockSize = 0;
     zstd_operation_mode operation = zom_compress;
     ZSTD_compressionParameters compressionParams;
-    int cLevel = ZSTDCLI_CLEVEL_DEFAULT;
+    int cLevel;
     int cLevelLast = -1000000000;
     unsigned recursive = 0;
     unsigned memLimit = 0;
@@ -525,6 +573,7 @@ int main(int argCount, const char* argv[])
     if (filenameTable==NULL) { DISPLAY("zstd: %s \n", strerror(errno)); exit(1); }
     filenameTable[0] = stdinmark;
     g_displayOut = stderr;
+    cLevel = init_cLevel();
     programName = lastNameFromPath(programName);
 #ifdef ZSTD_MULTITHREAD
     nbWorkers = 1;
@@ -607,6 +656,8 @@ int main(int argCount, const char* argv[])
 #ifdef ZSTD_LZ4COMPRESS
                     if (!strcmp(argument, "--format=lz4")) { suffix = LZ4_EXTENSION; FIO_setCompressionType(FIO_lz4Compression);  continue; }
 #endif
+                    if (!strcmp(argument, "--rsyncable")) { rsyncable = 1; continue; }
+                    if (!strcmp(argument, "--no-progress")) { FIO_setNoProgress(1); continue; }
 
                     /* long commands with arguments */
 #ifndef ZSTD_NODICT
@@ -937,8 +988,8 @@ int main(int argCount, const char* argv[])
         if (g_ldmBucketSizeLog != LDM_PARAM_DEFAULT) {
             benchParams.ldmBucketSizeLog = g_ldmBucketSizeLog;
         }
-        if (g_ldmHashEveryLog != LDM_PARAM_DEFAULT) {
-            benchParams.ldmHashEveryLog = g_ldmHashEveryLog;
+        if (g_ldmHashRateLog != LDM_PARAM_DEFAULT) {
+            benchParams.ldmHashRateLog = g_ldmHashRateLog;
         }
 
         if (cLevel > ZSTD_maxCLevel()) cLevel = ZSTD_maxCLevel();
@@ -1048,10 +1099,11 @@ int main(int argCount, const char* argv[])
         FIO_setLdmHashLog(g_ldmHashLog);
         FIO_setLdmMinMatch(g_ldmMinMatch);
         if (g_ldmBucketSizeLog != LDM_PARAM_DEFAULT) FIO_setLdmBucketSizeLog(g_ldmBucketSizeLog);
-        if (g_ldmHashEveryLog != LDM_PARAM_DEFAULT) FIO_setLdmHashEveryLog(g_ldmHashEveryLog);
+        if (g_ldmHashRateLog != LDM_PARAM_DEFAULT) FIO_setLdmHashRateLog(g_ldmHashRateLog);
         FIO_setAdaptiveMode(adapt);
         FIO_setAdaptMin(adaptMin);
         FIO_setAdaptMax(adaptMax);
+        FIO_setRsyncable(rsyncable);
         if (adaptMin > cLevel) cLevel = adaptMin;
         if (adaptMax < cLevel) cLevel = adaptMax;
 
@@ -1060,7 +1112,7 @@ int main(int argCount, const char* argv[])
         else
           operationResult = FIO_compressMultipleFilenames(filenameTable, filenameIdx, outFileName, suffix, dictFileName, cLevel, compressionParams);
 #else
-        (void)suffix; (void)adapt; (void)ultra; (void)cLevel; (void)ldmFlag; /* not used when ZSTD_NOCOMPRESS set */
+        (void)suffix; (void)adapt; (void)rsyncable; (void)ultra; (void)cLevel; (void)ldmFlag; /* not used when ZSTD_NOCOMPRESS set */
         DISPLAY("Compression not supported \n");
 #endif
     } else {  /* decompression or test */
diff --git a/programs/zstdgrep b/programs/zstdgrep
index 9f871c03f1aa..a10e0710a296 100755
--- a/programs/zstdgrep
+++ b/programs/zstdgrep
@@ -31,94 +31,101 @@ grep_args=""
 hyphen=0
 silent=0
 
-prg=$(basename $0)
+prg=$(basename "$0")
 
 # handle being called 'zegrep' or 'zfgrep'
-case ${prg} in
-    *zegrep)
-	grep_args="-E";;
-    *zfgrep)
-	grep_args="-F";;
+case "${prg}" in
+    *zegrep) grep_args="-E";;
+    *zfgrep) grep_args="-F";;
 esac
 
 # skip all options and pass them on to grep taking care of options
 # with arguments, and if -e was supplied
 
-while [ $# -gt 0 -a ${endofopts} -eq 0 ]
-do
-    case $1 in
+while [ "$#" -gt 0 ] && [ "${endofopts}" -eq 0 ]; do
+    case "$1" in
     # from GNU grep-2.5.1 -- keep in sync!
-	-[ABCDXdefm])
-	    if [ $# -lt 2 ]
-		then
-		echo "${prg}: missing argument for $1 flag" >&2
-		exit 1
-	    fi
-	    case $1 in
-		-e)
-		    pattern="$2"
-		    pattern_found=1
-		    shift 2
-		    break
-		    ;;
-		*)
-		    ;;
-	    esac
-	    grep_args="${grep_args} $1 $2"
-	    shift 2
-	    ;;
-	--)
-	    shift
-	    endofopts=1
-	    ;;
-	-)
-	    hyphen=1
-	    shift
-	    ;;
-	-h)
-	    silent=1
-	    shift
-	    ;;
-	-*)
-	    grep_args="${grep_args} $1"
-	    shift
-	    ;;
-	*)
-	    # pattern to grep for
-	    endofopts=1
-	    ;;
+        -[ABCDXdefm])
+            if [ "$#" -lt 2 ]; then
+                printf '%s: missing argument for %s flag\n' "${prg}" "$1" >&2
+                exit 1
+            fi
+            case "$1" in
+                -e)
+                    pattern="$2"
+                    pattern_found=1
+                    shift 2
+                    break
+                    ;;
+                *)
+                    ;;
+            esac
+            grep_args="${grep_args} $1 $2"
+            shift 2
+            ;;
+        --)
+            shift
+            endofopts=1
+            ;;
+        -)
+            hyphen=1
+            shift
+            ;;
+        -h)
+            silent=1
+            shift
+            ;;
+        -*)
+            grep_args="${grep_args} $1"
+            shift
+            ;;
+        *)
+            # pattern to grep for
+            endofopts=1
+            ;;
     esac
 done
 
 # if no -e option was found, take next argument as grep-pattern
-if [ ${pattern_found} -lt 1 ]
-then
-    if [ $# -ge 1 ]; then
-	pattern="$1"
-	shift
-    elif [ ${hyphen} -gt 0 ]; then
-	pattern="-"
+if [ "${pattern_found}" -lt 1 ]; then
+    if [ "$#" -ge 1 ]; then
+        pattern="$1"
+        shift
+    elif [ "${hyphen}" -gt 0 ]; then
+        pattern="-"
     else
-	echo "${prg}: missing pattern" >&2
-	exit 1
+        printf '%s: missing pattern\n' "${prg}" >&2
+        exit 1
     fi
 fi
 
+EXIT_CODE=0
 # call grep ...
-if [ $# -lt 1 ]
-then
+if [ "$#" -lt 1 ]; then
     # ... on stdin
-    ${zcat} -fq - | ${grep} ${grep_args} -- "${pattern}" -
+    set -f # Disable file name generation (globbing).
+    # shellcheck disable=SC2086
+    "${zcat}" -fq - | "${grep}" ${grep_args} -- "${pattern}" -
+    EXIT_CODE=$?
+    set +f
 else
     # ... on all files given on the command line
-    if [ ${silent} -lt 1 -a $# -gt 1 ]; then
-	grep_args="-H ${grep_args}"
+    if [ "${silent}" -lt 1 ] && [ "$#" -gt 1 ]; then
+        grep_args="-H ${grep_args}"
     fi
-    while [ $# -gt 0 ]
-    do
-	${zcat} -fq -- "$1" | ${grep} --label="${1}" ${grep_args} -- "${pattern}" -
-	shift
+    CUR_EXIT_CODE=0
+    EXIT_CODE=1
+    set -f
+    while [ "$#" -gt 0 ]; do
+        # shellcheck disable=SC2086
+        "${zcat}" -fq -- "$1" | "${grep}" --label="${1}" ${grep_args} -- "${pattern}" -
+        CUR_EXIT_CODE=$?
+        if [ "${CUR_EXIT_CODE}" -eq 0 ] && [ "${EXIT_CODE}" -ne 1 ]; then
+            EXIT_CODE=0
+        fi
+        shift
     done
+    set +f
 fi
 
-exit 0
+exit "${EXIT_CODE}"
diff --git a/programs/zstdgrep.1 b/programs/zstdgrep.1
index 716d28fc8e7e..57bc14de3587 100644
--- a/programs/zstdgrep.1
+++ b/programs/zstdgrep.1
@@ -1,5 +1,5 @@
 .
-.TH "ZSTDGREP" "1" "October 2018" "zstd 1.3.7" "User Commands"
+.TH "ZSTDGREP" "1" "December 2018" "zstd 1.3.8" "User Commands"
 .
 .SH "NAME"
 \fBzstdgrep\fR \- print lines matching a pattern in zstandard\-compressed files
diff --git a/programs/zstdless.1 b/programs/zstdless.1
index bf4965e7dfa6..ff39742bf8a6 100644
--- a/programs/zstdless.1
+++ b/programs/zstdless.1
@@ -1,5 +1,5 @@
 .
-.TH "ZSTDLESS" "1" "October 2018" "zstd 1.3.7" "User Commands"
+.TH "ZSTDLESS" "1" "December 2018" "zstd 1.3.8" "User Commands"
 .
 .SH "NAME"
 \fBzstdless\fR \- view zstandard\-compressed files
diff --git a/tests/Makefile b/tests/Makefile
index 2a96829f61c5..25bd5c84ef90 100644
--- a/tests/Makefile
+++ b/tests/Makefile
@@ -132,18 +132,18 @@ fullbench fullbench32 : CPPFLAGS += $(MULTITHREAD_CPP)
 fullbench fullbench32 : LDFLAGS += $(MULTITHREAD_LD)
 fullbench fullbench32 : DEBUGFLAGS = -DNDEBUG  # turn off assert() for speed measurements
 fullbench fullbench32 : $(ZSTD_FILES)
-fullbench fullbench32 : $(PRGDIR)/datagen.c $(PRGDIR)/bench.c fullbench.c
+fullbench fullbench32 : $(PRGDIR)/datagen.c $(PRGDIR)/util.c $(PRGDIR)/benchfn.c fullbench.c
 	$(CC) $(FLAGS) $^ -o $@$(EXT)
 
 fullbench-lib : CPPFLAGS += -DXXH_NAMESPACE=ZSTD_
 fullbench-lib : zstd-staticLib
-fullbench-lib : $(PRGDIR)/datagen.c $(PRGDIR)/bench.c fullbench.c
+fullbench-lib : $(PRGDIR)/datagen.c $(PRGDIR)/util.c $(PRGDIR)/benchfn.c fullbench.c
 	$(CC) $(FLAGS) $(filter %.c,$^) -o $@$(EXT) $(ZSTDDIR)/libzstd.a
 
 # note : broken : requires unavailable symbols
 fullbench-dll : zstd-dll
 fullbench-dll : LDFLAGS+= -L$(ZSTDDIR) -lzstd
-fullbench-dll: $(PRGDIR)/datagen.c fullbench.c
+fullbench-dll: $(PRGDIR)/datagen.c $(PRGDIR)/util.c $(PRGDIR)/benchfn.c fullbench.c
 #	$(CC) $(FLAGS) $(filter %.c,$^) -o $@$(EXT) -DZSTD_DLL_IMPORT=1 $(ZSTDDIR)/dll/libzstd.dll
 	$(CC) $(FLAGS) $(filter %.c,$^) -o $@$(EXT)
 
@@ -152,32 +152,32 @@ fuzzer  : LDFLAGS += $(MULTITHREAD_LD)
 fuzzer32: CFLAGS += -m32
 fuzzer  : $(ZSTDMT_OBJECTS)
 fuzzer32: $(ZSTD_FILES)
-fuzzer fuzzer32 : $(ZDICT_FILES) $(PRGDIR)/datagen.c fuzzer.c
+fuzzer fuzzer32 : $(ZDICT_FILES) $(PRGDIR)/util.c $(PRGDIR)/datagen.c fuzzer.c
 	$(CC) $(FLAGS) $^ -o $@$(EXT)
 
 fuzzer-dll : zstd-dll
 fuzzer-dll : LDFLAGS+= -L$(ZSTDDIR) -lzstd
-fuzzer-dll : $(ZSTDDIR)/common/xxhash.c $(PRGDIR)/datagen.c fuzzer.c
+fuzzer-dll : $(ZSTDDIR)/common/xxhash.c $(PRGDIR)/util.c $(PRGDIR)/datagen.c fuzzer.c
 	$(CC) $(CPPFLAGS) $(CFLAGS) $(filter %.c,$^) $(LDFLAGS) -o $@$(EXT)
 
 zbufftest : CPPFLAGS += -I$(ZSTDDIR)/deprecated
 zbufftest : CFLAGS += -Wno-deprecated-declarations   # required to silence deprecation warnings
-zbufftest : $(ZSTD_OBJECTS) $(ZBUFF_FILES) $(PRGDIR)/datagen.c zbufftest.c
+zbufftest : $(ZSTD_OBJECTS) $(ZBUFF_FILES) $(PRGDIR)/util.c $(PRGDIR)/datagen.c zbufftest.c
 	$(CC) $(FLAGS) $^ -o $@$(EXT)
 
 zbufftest32 : CPPFLAGS += -I$(ZSTDDIR)/deprecated
 zbufftest32 : CFLAGS += -Wno-deprecated-declarations -m32
-zbufftest32 : $(ZSTD_FILES) $(ZBUFF_FILES) $(PRGDIR)/datagen.c zbufftest.c
+zbufftest32 : $(ZSTD_FILES) $(ZBUFF_FILES) $(PRGDIR)/util.c $(PRGDIR)/datagen.c zbufftest.c
 	$(CC) $(FLAGS) $^ -o $@$(EXT)
 
 zbufftest-dll : zstd-dll
 zbufftest-dll : CPPFLAGS += -I$(ZSTDDIR)/deprecated
 zbufftest-dll : CFLAGS += -Wno-deprecated-declarations   # required to silence deprecation warnings
 zbufftest-dll : LDFLAGS+= -L$(ZSTDDIR) -lzstd
-zbufftest-dll : $(ZSTDDIR)/common/xxhash.c $(PRGDIR)/datagen.c zbufftest.c
+zbufftest-dll : $(ZSTDDIR)/common/xxhash.c $(PRGDIR)/util.c $(PRGDIR)/datagen.c zbufftest.c
 	$(CC) $(CPPFLAGS) $(CFLAGS) $(filter %.c,$^) $(LDFLAGS) -o $@$(EXT)
 
-ZSTREAM_LOCAL_FILES := $(PRGDIR)/datagen.c seqgen.c zstreamtest.c
+ZSTREAM_LOCAL_FILES := $(PRGDIR)/datagen.c $(PRGDIR)/util.c seqgen.c zstreamtest.c
 ZSTREAM_PROPER_FILES := $(ZDICT_FILES) $(ZSTREAM_LOCAL_FILES)
 ZSTREAMFILES := $(ZSTD_FILES) $(ZSTREAM_PROPER_FILES)
 zstreamtest32 : CFLAGS += -m32
@@ -203,7 +203,7 @@ zstreamtest-dll : $(ZSTREAM_LOCAL_FILES)
 	$(CC) $(CPPFLAGS) $(CFLAGS) $(filter %.c,$^) $(LDFLAGS) -o $@$(EXT)
 
 paramgrill : DEBUGFLAGS =  # turn off assert() by default for speed measurements
-paramgrill : $(ZSTD_FILES) $(PRGDIR)/bench.c $(PRGDIR)/datagen.c paramgrill.c
+paramgrill : $(ZSTD_FILES) $(PRGDIR)/util.c $(PRGDIR)/benchfn.c $(PRGDIR)/benchzstd.c $(PRGDIR)/datagen.c paramgrill.c
 	$(CC) $(FLAGS) $^ -lm -o $@$(EXT)
 
 datagen : $(PRGDIR)/datagen.c datagencli.c
@@ -222,7 +222,7 @@ legacy : CPPFLAGS += -I$(ZSTDDIR)/legacy -DZSTD_LEGACY_SUPPORT=4
 legacy : $(ZSTD_FILES) $(wildcard $(ZSTDDIR)/legacy/*.c) legacy.c
 	$(CC) $(FLAGS) $^ -o $@$(EXT)
 
-decodecorpus : $(filter-out zstdc_zstd_compress.o, $(ZSTD_OBJECTS)) $(ZDICT_FILES) decodecorpus.c
+decodecorpus : $(filter-out zstdc_zstd_compress.o, $(ZSTD_OBJECTS)) $(ZDICT_FILES) $(PRGDIR)/util.c decodecorpus.c
 	$(CC) $(FLAGS) $^ -o $@$(EXT) -lm
 
 symbols  : symbols.c zstd-dll
@@ -233,7 +233,7 @@ else
 	$(CC) $(FLAGS) $< -o $@$(EXT) -Wl,-rpath=$(ZSTDDIR) $(ZSTDDIR)/libzstd.so   # broken on Mac
 endif
 
-poolTests : poolTests.c $(ZSTDDIR)/common/pool.c $(ZSTDDIR)/common/threading.c $(ZSTDDIR)/common/zstd_common.c $(ZSTDDIR)/common/error_private.c
+poolTests : $(PRGDIR)/util.c poolTests.c $(ZSTDDIR)/common/pool.c $(ZSTDDIR)/common/threading.c $(ZSTDDIR)/common/zstd_common.c $(ZSTDDIR)/common/error_private.c
 	$(CC) $(FLAGS) $(MULTITHREAD) $^ -o $@$(EXT)
 
 .PHONY: versionsTest
@@ -320,7 +320,7 @@ test32: test-zstd32 test-fullbench32 test-fuzzer32 test-zstream32
 test-all: test test32 valgrindTest test-decodecorpus-cli
 
 
-.PHONY: test-zstd test-zstd32 test-zstd-nolegacy
+.PHONY: test-zstd test-zstd32 test-zstd-nolegacy test-zstdgrep
 test-zstd: ZSTD = $(PRGDIR)/zstd
 test-zstd: zstd
 
@@ -352,6 +352,10 @@ test-gzstd: gzstd
 	$(PRGDIR)/zstd -dcf - > (32 - r)))
     static const U32 prime1 = 2654435761U;
@@ -350,7 +350,7 @@ static void writeFrameHeader(U32* seed, frame_t* frame, dictInfo info)
         }
     }
 
-    DISPLAYLEVEL(3, " frame content size:\t%u\n", (U32)fh.contentSize);
+    DISPLAYLEVEL(3, " frame content size:\t%u\n", (unsigned)fh.contentSize);
     DISPLAYLEVEL(3, " frame window size:\t%u\n", fh.windowSize);
     DISPLAYLEVEL(3, " content size flag:\t%d\n", contentSizeFlag);
     DISPLAYLEVEL(3, " single segment flag:\t%d\n", singleSegment);
@@ -412,7 +412,7 @@ static size_t writeLiteralsBlockSimple(U32* seed, frame_t* frame, size_t content
         /* RLE literals */
         BYTE const symb = (BYTE) (RAND(seed) % 256);
 
-        DISPLAYLEVEL(4, "   rle literals: 0x%02x\n", (U32)symb);
+        DISPLAYLEVEL(4, "   rle literals: 0x%02x\n", (unsigned)symb);
 
         memset(LITERAL_BUFFER, symb, litSize);
         op[0] = symb;
@@ -432,12 +432,12 @@ static size_t writeHufHeader(U32* seed, HUF_CElt* hufTable, void* dst, size_t ds
     BYTE* op = ostart;
 
     unsigned huffLog = 11;
-    U32 maxSymbolValue = 255;
+    unsigned maxSymbolValue = 255;
 
-    U32 count[HUF_SYMBOLVALUE_MAX+1];
+    unsigned count[HUF_SYMBOLVALUE_MAX+1];
 
     /* Scan input and build symbol stats */
-    {   size_t const largest = HIST_count_wksp (count, &maxSymbolValue, (const BYTE*)src, srcSize, WKSP);
+    {   size_t const largest = HIST_count_wksp (count, &maxSymbolValue, (const BYTE*)src, srcSize, WKSP, sizeof(WKSP));
         assert(!HIST_isError(largest));
         if (largest == srcSize) { *ostart = ((const BYTE*)src)[0]; return 0; }   /* single symbol, rle */
         if (largest <= (srcSize >> 7)+1) return 0;   /* Fast heuristic : not compressible enough */
@@ -568,8 +568,8 @@ static size_t writeLiteralsBlockCompressed(U32* seed, frame_t* frame, size_t con
         op += compressedSize;
 
         compressedSize += hufHeaderSize;
-        DISPLAYLEVEL(5, "    regenerated size: %u\n", (U32)litSize);
-        DISPLAYLEVEL(5, "    compressed size: %u\n", (U32)compressedSize);
+        DISPLAYLEVEL(5, "    regenerated size: %u\n", (unsigned)litSize);
+        DISPLAYLEVEL(5, "    compressed size: %u\n", (unsigned)compressedSize);
         if (compressedSize >= litSize) {
             DISPLAYLEVEL(5, "     trying again\n");
             /* if we have to try again, reset the stats so we don't accidentally
@@ -656,7 +656,7 @@ static U32 generateSequences(U32* seed, frame_t* frame, seqStore_t* seqStore,
         excessMatch = remainingMatch - numSequences * MIN_SEQ_LEN;
     }
 
-    DISPLAYLEVEL(5, "    total match lengths: %u\n", (U32)remainingMatch);
+    DISPLAYLEVEL(5, "    total match lengths: %u\n", (unsigned)remainingMatch);
     for (i = 0; i < numSequences; i++) {
         /* Generate match and literal lengths by exponential distribution to
          * ensure nice numbers */
@@ -748,12 +748,13 @@ static U32 generateSequences(U32* seed, frame_t* frame, seqStore_t* seqStore,
             frame->stats.rep[0] = offset;
         }
 
-        DISPLAYLEVEL(6, "      LL: %5u OF: %5u ML: %5u", literalLen, offset, matchLen);
+        DISPLAYLEVEL(6, "      LL: %5u OF: %5u ML: %5u",
+                    (unsigned)literalLen, (unsigned)offset, (unsigned)matchLen);
         DISPLAYLEVEL(7, " srcPos: %8u seqNb: %3u",
-                     (U32)((BYTE*)srcPtr - (BYTE*)frame->srcStart), i);
+                     (unsigned)((BYTE*)srcPtr - (BYTE*)frame->srcStart), (unsigned)i);
         DISPLAYLEVEL(6, "\n");
         if (offsetCode < 3) {
-            DISPLAYLEVEL(7, "        repeat offset: %d\n", repIndex);
+            DISPLAYLEVEL(7, "        repeat offset: %d\n", (int)repIndex);
         }
         /* use libzstd sequence handling */
         ZSTD_storeSeq(seqStore, literalLen, literals, offsetCode,
@@ -766,8 +767,8 @@ static U32 generateSequences(U32* seed, frame_t* frame, seqStore_t* seqStore,
 
     memcpy(srcPtr, literals, literalsSize);
     srcPtr += literalsSize;
-    DISPLAYLEVEL(6, "      excess literals: %5u", (U32)literalsSize);
-    DISPLAYLEVEL(7, " srcPos: %8u", (U32)((BYTE*)srcPtr - (BYTE*)frame->srcStart));
+    DISPLAYLEVEL(6, "      excess literals: %5u", (unsigned)literalsSize);
+    DISPLAYLEVEL(7, " srcPos: %8u", (unsigned)((BYTE*)srcPtr - (BYTE*)frame->srcStart));
     DISPLAYLEVEL(6, "\n");
 
     return numSequences;
@@ -800,7 +801,7 @@ static size_t writeSequences(U32* seed, frame_t* frame, seqStore_t* seqStorePtr,
                              size_t nbSeq)
 {
     /* This code is mostly copied from ZSTD_compressSequences in zstd_compress.c */
-    U32 count[MaxSeq+1];
+    unsigned count[MaxSeq+1];
     S16 norm[MaxSeq+1];
     FSE_CTable* CTable_LitLength = frame->stats.litlengthCTable;
     FSE_CTable* CTable_OffsetBits = frame->stats.offcodeCTable;
@@ -823,21 +824,20 @@ static size_t writeSequences(U32* seed, frame_t* frame, seqStore_t* seqStorePtr,
     else if (nbSeq < LONGNBSEQ) op[0] = (BYTE)((nbSeq>>8) + 0x80), op[1] = (BYTE)nbSeq, op+=2;
     else op[0]=0xFF, MEM_writeLE16(op+1, (U16)(nbSeq - LONGNBSEQ)), op+=3;
 
-    /* seqHead : flags for FSE encoding type */
-    seqHead = op++;
-
     if (nbSeq==0) {
         frame->data = op;
-
         return 0;
     }
 
+    /* seqHead : flags for FSE encoding type */
+    seqHead = op++;
+
     /* convert length/distances into codes */
     ZSTD_seqToCodes(seqStorePtr);
 
     /* CTable for Literal Lengths */
-    {   U32 max = MaxLL;
-        size_t const mostFrequent = HIST_countFast_wksp(count, &max, llCodeTable, nbSeq, WKSP);   /* cannot fail */
+    {   unsigned max = MaxLL;
+        size_t const mostFrequent = HIST_countFast_wksp(count, &max, llCodeTable, nbSeq, WKSP, sizeof(WKSP));   /* cannot fail */
         assert(!HIST_isError(mostFrequent));
         if (mostFrequent == nbSeq) {
             /* do RLE if we have the chance */
@@ -868,8 +868,8 @@ static size_t writeSequences(U32* seed, frame_t* frame, seqStore_t* seqStorePtr,
 
     /* CTable for Offsets */
     /* see Literal Lengths for descriptions of mode choices */
-    {   U32 max = MaxOff;
-        size_t const mostFrequent = HIST_countFast_wksp(count, &max, ofCodeTable, nbSeq, WKSP);   /* cannot fail */
+    {   unsigned max = MaxOff;
+        size_t const mostFrequent = HIST_countFast_wksp(count, &max, ofCodeTable, nbSeq, WKSP, sizeof(WKSP));   /* cannot fail */
         assert(!HIST_isError(mostFrequent));
         if (mostFrequent == nbSeq) {
             *op++ = ofCodeTable[0];
@@ -896,8 +896,8 @@ static size_t writeSequences(U32* seed, frame_t* frame, seqStore_t* seqStorePtr,
 
     /* CTable for MatchLengths */
     /* see Literal Lengths for descriptions of mode choices */
-    {   U32 max = MaxML;
-        size_t const mostFrequent = HIST_countFast_wksp(count, &max, mlCodeTable, nbSeq, WKSP);   /* cannot fail */
+    {   unsigned max = MaxML;
+        size_t const mostFrequent = HIST_countFast_wksp(count, &max, mlCodeTable, nbSeq, WKSP, sizeof(WKSP));   /* cannot fail */
         assert(!HIST_isError(mostFrequent));
         if (mostFrequent == nbSeq) {
             *op++ = *mlCodeTable;
@@ -928,7 +928,7 @@ static size_t writeSequences(U32* seed, frame_t* frame, seqStore_t* seqStorePtr,
     initSymbolSet(ofCodeTable, nbSeq, frame->stats.offsetSymbolSet, 28);
     initSymbolSet(mlCodeTable, nbSeq, frame->stats.matchlengthSymbolSet, 52);
 
-    DISPLAYLEVEL(5, "    LL type: %d OF type: %d ML type: %d\n", LLtype, Offtype, MLtype);
+    DISPLAYLEVEL(5, "    LL type: %d OF type: %d ML type: %d\n", (unsigned)LLtype, (unsigned)Offtype, (unsigned)MLtype);
 
     *seqHead = (BYTE)((LLtype<<6) + (Offtype<<4) + (MLtype<<2));
 
@@ -1015,11 +1015,11 @@ static size_t writeCompressedBlock(U32* seed, frame_t* frame, size_t contentSize
 
     literalsSize = writeLiteralsBlock(seed, frame, contentSize);
 
-    DISPLAYLEVEL(4, "   literals size: %u\n", (U32)literalsSize);
+    DISPLAYLEVEL(4, "   literals size: %u\n", (unsigned)literalsSize);
 
     nbSeq = writeSequencesBlock(seed, frame, contentSize, literalsSize, info);
 
-    DISPLAYLEVEL(4, "   number of sequences: %u\n", (U32)nbSeq);
+    DISPLAYLEVEL(4, "   number of sequences: %u\n", (unsigned)nbSeq);
 
     return (BYTE*)frame->data - blockStart;
 }
@@ -1035,7 +1035,7 @@ static void writeBlock(U32* seed, frame_t* frame, size_t contentSize,
     BYTE *op = header + 3;
 
     DISPLAYLEVEL(4, " block:\n");
-    DISPLAYLEVEL(4, "  block content size: %u\n", (U32)contentSize);
+    DISPLAYLEVEL(4, "  block content size: %u\n", (unsigned)contentSize);
     DISPLAYLEVEL(4, "  last block: %s\n", lastBlock ? "yes" : "no");
 
     if (blockTypeDesc == 0) {
@@ -1083,7 +1083,7 @@ static void writeBlock(U32* seed, frame_t* frame, size_t contentSize,
     frame->src = (BYTE*)frame->src + contentSize;
 
     DISPLAYLEVEL(4, "  block type: %s\n", BLOCK_TYPES[blockType]);
-    DISPLAYLEVEL(4, "  block size field: %u\n", (U32)blockSize);
+    DISPLAYLEVEL(4, "  block size field: %u\n", (unsigned)blockSize);
 
     header[0] = (BYTE) ((lastBlock | (blockType << 1) | (blockSize << 3)) & 0xff);
     MEM_writeLE16(header + 1, (U16) (blockSize >> 5));
@@ -1125,7 +1125,7 @@ static void writeChecksum(frame_t* frame)
 {
     /* write checksum so implementations can verify their output */
     U64 digest = XXH64(frame->srcStart, (BYTE*)frame->src-(BYTE*)frame->srcStart, 0);
-    DISPLAYLEVEL(3, "  checksum: %08x\n", (U32)digest);
+    DISPLAYLEVEL(3, "  checksum: %08x\n", (unsigned)digest);
     MEM_writeLE32(frame->data, (U32)digest);
     frame->data = (BYTE*)frame->data + 4;
 }
@@ -1186,7 +1186,7 @@ static U32 generateCompressedBlock(U32 seed, frame_t* frame, dictInfo info)
     size_t blockContentSize;
     int blockWritten = 0;
     BYTE* op;
-    DISPLAYLEVEL(4, "block seed: %u\n", seed);
+    DISPLAYLEVEL(4, "block seed: %u\n", (unsigned)seed);
     initFrame(frame);
     op = (BYTE*)frame->data;
 
@@ -1223,7 +1223,7 @@ static U32 generateCompressedBlock(U32 seed, frame_t* frame, dictInfo info)
             DISPLAYLEVEL(5, "   can't compress block : try again \n");
         } else {
             blockWritten = 1;
-            DISPLAYLEVEL(4, "   block size: %u \n", (U32)cSize);
+            DISPLAYLEVEL(4, "   block size: %u \n", (unsigned)cSize);
             frame->src = (BYTE*)frame->src + blockContentSize;
         }
     }
@@ -1234,7 +1234,7 @@ static U32 generateCompressedBlock(U32 seed, frame_t* frame, dictInfo info)
 static U32 generateFrame(U32 seed, frame_t* fr, dictInfo info)
 {
     /* generate a complete frame */
-    DISPLAYLEVEL(3, "frame seed: %u\n", seed);
+    DISPLAYLEVEL(3, "frame seed: %u\n", (unsigned)seed);
     initFrame(fr);
 
     writeFrameHeader(&seed, fr, info);
@@ -1480,8 +1480,8 @@ static int runBlockTest(U32* seed)
 
     {   size_t const r = testDecodeRawBlock(&fr);
         if (ZSTD_isError(r)) {
-            DISPLAY("Error in block mode on test seed %u: %s\n", seedCopy,
-                    ZSTD_getErrorName(r));
+            DISPLAY("Error in block mode on test seed %u: %s\n",
+                    (unsigned)seedCopy, ZSTD_getErrorName(r));
             return 1;
         }
     }
@@ -1489,7 +1489,7 @@ static int runBlockTest(U32* seed)
     {   size_t const r = testDecodeWithDict(*seed, gt_block);
         if (ZSTD_isError(r)) {
             DISPLAY("Error in block mode with dictionary on test seed %u: %s\n",
-                    seedCopy, ZSTD_getErrorName(r));
+                    (unsigned)seedCopy, ZSTD_getErrorName(r));
             return 1;
         }
     }
@@ -1507,21 +1507,21 @@ static int runFrameTest(U32* seed)
     {   size_t const r = testDecodeSimple(&fr);
         if (ZSTD_isError(r)) {
             DISPLAY("Error in simple mode on test seed %u: %s\n",
-                    seedCopy, ZSTD_getErrorName(r));
+                    (unsigned)seedCopy, ZSTD_getErrorName(r));
             return 1;
         }
     }
     {   size_t const r = testDecodeStreaming(&fr);
         if (ZSTD_isError(r)) {
             DISPLAY("Error in streaming mode on test seed %u: %s\n",
-                    seedCopy, ZSTD_getErrorName(r));
+                    (unsigned)seedCopy, ZSTD_getErrorName(r));
             return 1;
         }
     }
     {   size_t const r = testDecodeWithDict(*seed, gt_frame);  /* avoid big dictionaries */
         if (ZSTD_isError(r)) {
             DISPLAY("Error in dictionary mode on test seed %u: %s\n",
-                    seedCopy, ZSTD_getErrorName(r));
+                    (unsigned)seedCopy, ZSTD_getErrorName(r));
             return 1;
         }
     }
@@ -1538,7 +1538,7 @@ static int runTestMode(U32 seed, unsigned numFiles, unsigned const testDurationS
 
     if (numFiles == 0 && !testDurationS) numFiles = 1;
 
-    DISPLAY("seed: %u\n", seed);
+    DISPLAY("seed: %u\n", (unsigned)seed);
 
     for (fnum = 0; fnum < numFiles || UTIL_clockSpanMicro(startClock) < maxClockSpan; fnum++) {
         if (fnum < numFiles)
@@ -1568,7 +1568,7 @@ static int generateFile(U32 seed, const char* const path,
 {
     frame_t fr;
 
-    DISPLAY("seed: %u\n", seed);
+    DISPLAY("seed: %u\n", (unsigned)seed);
 
     {   dictInfo const info = initDictInfo(0, 0, NULL, 0);
         if (genType == gt_frame) {
@@ -1590,7 +1590,7 @@ static int generateCorpus(U32 seed, unsigned numFiles, const char* const path,
     char outPath[MAX_PATH];
     unsigned fnum;
 
-    DISPLAY("seed: %u\n", seed);
+    DISPLAY("seed: %u\n", (unsigned)seed);
 
     for (fnum = 0; fnum < numFiles; fnum++) {
         frame_t fr;
diff --git a/tests/fullbench.c b/tests/fullbench.c
index b05f1537cd70..8644a2e3a136 100644
--- a/tests/fullbench.c
+++ b/tests/fullbench.c
@@ -19,7 +19,7 @@
 
 #include "mem.h"         /* U32 */
 #ifndef ZSTD_DLL_IMPORT
-    #include "zstd_internal.h"   /* ZSTD_blockHeaderSize, blockType_e, KB, MB */
+    #include "zstd_internal.h"   /* ZSTD_decodeSeqHeaders, ZSTD_blockHeaderSize, blockType_e, KB, MB */
 #else
     #define KB *(1 <<10)
     #define MB *(1 <<20)
@@ -30,7 +30,8 @@
 #include "zstd.h"        /* ZSTD_versionString */
 #include "util.h"        /* time functions */
 #include "datagen.h"
-#include "bench.h"       /* CustomBench*/
+#include "benchfn.h"       /* CustomBench*/
+#include "benchzstd.h"     /* MB_UNIT */
 
 
 /*_************************************
@@ -63,10 +64,10 @@ static const size_t g_sampleSize = 10000000;
 /*_************************************
 *  Benchmark Parameters
 **************************************/
-static U32 g_nbIterations = NBLOOPS;
+static unsigned g_nbIterations = NBLOOPS;
 static double g_compressibility = COMPRESSIBILITY_DEFAULT;
 
-static void BMK_SetNbIterations(U32 nbLoops)
+static void BMK_SetNbIterations(unsigned nbLoops)
 {
     g_nbIterations = nbLoops;
     DISPLAY("- %i iterations -\n", g_nbIterations);
@@ -133,7 +134,6 @@ static size_t local_ZSTD_decodeLiteralsBlock(const void* src, size_t srcSize, vo
     return ZSTD_decodeLiteralsBlock((ZSTD_DCtx*)g_zdc, buff2, g_cSize);
 }
 
-extern size_t ZSTD_decodeSeqHeaders(ZSTD_DCtx* dctx, int* nbSeq, const void* src, size_t srcSize);
 static size_t local_ZSTD_decodeSeqHeaders(const void* src, size_t srcSize, void* dst, size_t dstSize, void* buff2)
 {
     int nbSeq;
@@ -171,17 +171,8 @@ local_ZSTD_compress_generic_end(const void* src, size_t srcSize,
                                 void* dst, size_t dstCapacity,
                                 void* buff2)
 {
-    ZSTD_outBuffer buffOut;
-    ZSTD_inBuffer buffIn;
     (void)buff2;
-    buffOut.dst = dst;
-    buffOut.size = dstCapacity;
-    buffOut.pos = 0;
-    buffIn.src = src;
-    buffIn.size = srcSize;
-    buffIn.pos = 0;
-    ZSTD_compress_generic(g_cstream, &buffOut, &buffIn, ZSTD_e_end);
-    return buffOut.pos;
+    return ZSTD_compress2(g_cstream, dst, dstCapacity, src, srcSize);
 }
 
 static size_t
@@ -198,8 +189,8 @@ local_ZSTD_compress_generic_continue(const void* src, size_t srcSize,
     buffIn.src = src;
     buffIn.size = srcSize;
     buffIn.pos = 0;
-    ZSTD_compress_generic(g_cstream, &buffOut, &buffIn, ZSTD_e_continue);
-    ZSTD_compress_generic(g_cstream, &buffOut, &buffIn, ZSTD_e_end);
+    ZSTD_compressStream2(g_cstream, &buffOut, &buffIn, ZSTD_e_continue);
+    ZSTD_compressStream2(g_cstream, &buffOut, &buffIn, ZSTD_e_end);
     return buffOut.pos;
 }
 
@@ -208,18 +199,9 @@ local_ZSTD_compress_generic_T2_end(const void* src, size_t srcSize,
                                    void* dst, size_t dstCapacity,
                                    void* buff2)
 {
-    ZSTD_outBuffer buffOut;
-    ZSTD_inBuffer buffIn;
     (void)buff2;
-    ZSTD_CCtx_setParameter(g_cstream, ZSTD_p_nbWorkers, 2);
-    buffOut.dst = dst;
-    buffOut.size = dstCapacity;
-    buffOut.pos = 0;
-    buffIn.src = src;
-    buffIn.size = srcSize;
-    buffIn.pos = 0;
-    while (ZSTD_compress_generic(g_cstream, &buffOut, &buffIn, ZSTD_e_end)) {}
-    return buffOut.pos;
+    ZSTD_CCtx_setParameter(g_cstream, ZSTD_c_nbWorkers, 2);
+    return ZSTD_compress2(g_cstream, dst, dstCapacity, src, srcSize);
 }
 
 static size_t
@@ -230,15 +212,15 @@ local_ZSTD_compress_generic_T2_continue(const void* src, size_t srcSize,
     ZSTD_outBuffer buffOut;
     ZSTD_inBuffer buffIn;
     (void)buff2;
-    ZSTD_CCtx_setParameter(g_cstream, ZSTD_p_nbWorkers, 2);
+    ZSTD_CCtx_setParameter(g_cstream, ZSTD_c_nbWorkers, 2);
     buffOut.dst = dst;
     buffOut.size = dstCapacity;
     buffOut.pos = 0;
     buffIn.src = src;
     buffIn.size = srcSize;
     buffIn.pos = 0;
-    ZSTD_compress_generic(g_cstream, &buffOut, &buffIn, ZSTD_e_continue);
-    while(ZSTD_compress_generic(g_cstream, &buffOut, &buffIn, ZSTD_e_end)) {}
+    ZSTD_compressStream2(g_cstream, &buffOut, &buffIn, ZSTD_e_continue);
+    while(ZSTD_compressStream2(g_cstream, &buffOut, &buffIn, ZSTD_e_end)) {}
     return buffOut.pos;
 }
 
@@ -334,7 +316,7 @@ static size_t local_ZSTD_decompressContinue(const void* src, size_t srcSize,
 /*_*******************************************************
 *  Bench functions
 *********************************************************/
-static size_t benchMem(U32 benchNb,
+static size_t benchMem(unsigned benchNb,
                        const void* src, size_t srcSize,
                        int cLevel, ZSTD_compressionParameters cparams)
 {
@@ -408,28 +390,28 @@ static size_t benchMem(U32 benchNb,
     if (g_cstream==NULL) g_cstream = ZSTD_createCStream();
     if (g_dstream==NULL) g_dstream = ZSTD_createDStream();
 
-    /* DISPLAY("params: cLevel %d, wlog %d hlog %d clog %d slog %d slen %d tlen %d strat %d \n",
+    /* DISPLAY("params: cLevel %d, wlog %d hlog %d clog %d slog %d mml %d tlen %d strat %d \n",
           cLevel, cparams->windowLog, cparams->hashLog, cparams->chainLog, cparams->searchLog,
-          cparams->searchLength, cparams->targetLength, cparams->strategy); */
+          cparams->minMatch, cparams->targetLength, cparams->strategy); */
 
-    ZSTD_CCtx_setParameter(g_zcc, ZSTD_p_compressionLevel, cLevel);
-    ZSTD_CCtx_setParameter(g_zcc, ZSTD_p_windowLog, cparams.windowLog);
-    ZSTD_CCtx_setParameter(g_zcc, ZSTD_p_hashLog, cparams.hashLog);
-    ZSTD_CCtx_setParameter(g_zcc, ZSTD_p_chainLog, cparams.chainLog);
-    ZSTD_CCtx_setParameter(g_zcc, ZSTD_p_searchLog, cparams.searchLog);
-    ZSTD_CCtx_setParameter(g_zcc, ZSTD_p_minMatch, cparams.searchLength);
-    ZSTD_CCtx_setParameter(g_zcc, ZSTD_p_targetLength, cparams.targetLength);
-    ZSTD_CCtx_setParameter(g_zcc, ZSTD_p_compressionStrategy, cparams.strategy);
+    ZSTD_CCtx_setParameter(g_zcc, ZSTD_c_compressionLevel, cLevel);
+    ZSTD_CCtx_setParameter(g_zcc, ZSTD_c_windowLog, cparams.windowLog);
+    ZSTD_CCtx_setParameter(g_zcc, ZSTD_c_hashLog, cparams.hashLog);
+    ZSTD_CCtx_setParameter(g_zcc, ZSTD_c_chainLog, cparams.chainLog);
+    ZSTD_CCtx_setParameter(g_zcc, ZSTD_c_searchLog, cparams.searchLog);
+    ZSTD_CCtx_setParameter(g_zcc, ZSTD_c_minMatch, cparams.minMatch);
+    ZSTD_CCtx_setParameter(g_zcc, ZSTD_c_targetLength, cparams.targetLength);
+    ZSTD_CCtx_setParameter(g_zcc, ZSTD_c_strategy, cparams.strategy);
 
 
-    ZSTD_CCtx_setParameter(g_cstream, ZSTD_p_compressionLevel, cLevel);
-    ZSTD_CCtx_setParameter(g_cstream, ZSTD_p_windowLog, cparams.windowLog);
-    ZSTD_CCtx_setParameter(g_cstream, ZSTD_p_hashLog, cparams.hashLog);
-    ZSTD_CCtx_setParameter(g_cstream, ZSTD_p_chainLog, cparams.chainLog);
-    ZSTD_CCtx_setParameter(g_cstream, ZSTD_p_searchLog, cparams.searchLog);
-    ZSTD_CCtx_setParameter(g_cstream, ZSTD_p_minMatch, cparams.searchLength);
-    ZSTD_CCtx_setParameter(g_cstream, ZSTD_p_targetLength, cparams.targetLength);
-    ZSTD_CCtx_setParameter(g_cstream, ZSTD_p_compressionStrategy, cparams.strategy);
+    ZSTD_CCtx_setParameter(g_cstream, ZSTD_c_compressionLevel, cLevel);
+    ZSTD_CCtx_setParameter(g_cstream, ZSTD_c_windowLog, cparams.windowLog);
+    ZSTD_CCtx_setParameter(g_cstream, ZSTD_c_hashLog, cparams.hashLog);
+    ZSTD_CCtx_setParameter(g_cstream, ZSTD_c_chainLog, cparams.chainLog);
+    ZSTD_CCtx_setParameter(g_cstream, ZSTD_c_searchLog, cparams.searchLog);
+    ZSTD_CCtx_setParameter(g_cstream, ZSTD_c_minMatch, cparams.minMatch);
+    ZSTD_CCtx_setParameter(g_cstream, ZSTD_c_targetLength, cparams.targetLength);
+    ZSTD_CCtx_setParameter(g_cstream, ZSTD_c_strategy, cparams.strategy);
 
     /* Preparation */
     switch(benchNb)
@@ -455,8 +437,8 @@ static size_t benchMem(U32 benchNb,
             ZSTD_frameHeader zfp;
             size_t frameHeaderSize, skippedSize;
             g_cSize = ZSTD_compress(dstBuff, dstBuffSize, src, srcSize, cLevel);
-            frameHeaderSize = ZSTD_getFrameHeader(&zfp, dstBuff, ZSTD_frameHeaderSize_min);
-            if (frameHeaderSize==0) frameHeaderSize = ZSTD_frameHeaderSize_min;
+            frameHeaderSize = ZSTD_getFrameHeader(&zfp, dstBuff, ZSTD_FRAMEHEADERSIZE_MIN);
+            if (frameHeaderSize==0) frameHeaderSize = ZSTD_FRAMEHEADERSIZE_MIN;
             ZSTD_getcBlockSize(dstBuff+frameHeaderSize, dstBuffSize, &bp);  /* Get 1st block type */
             if (bp.blockType != bt_compressed) {
                 DISPLAY("ZSTD_decodeLiteralsBlock : impossible to test on this sample (not compressible)\n");
@@ -476,8 +458,8 @@ static size_t benchMem(U32 benchNb,
             size_t frameHeaderSize, cBlockSize;
             ZSTD_compress(dstBuff, dstBuffSize, src, srcSize, cLevel);   /* it would be better to use direct block compression here */
             g_cSize = ZSTD_compress(dstBuff, dstBuffSize, src, srcSize, cLevel);
-            frameHeaderSize = ZSTD_getFrameHeader(&zfp, dstBuff, ZSTD_frameHeaderSize_min);
-            if (frameHeaderSize==0) frameHeaderSize = ZSTD_frameHeaderSize_min;
+            frameHeaderSize = ZSTD_getFrameHeader(&zfp, dstBuff, ZSTD_FRAMEHEADERSIZE_MIN);
+            if (frameHeaderSize==0) frameHeaderSize = ZSTD_FRAMEHEADERSIZE_MIN;
             ip += frameHeaderSize;   /* Skip frame Header */
             cBlockSize = ZSTD_getcBlockSize(ip, dstBuffSize, &bp);   /* Get 1st block type */
             if (bp.blockType != bt_compressed) {
@@ -515,20 +497,28 @@ static size_t benchMem(U32 benchNb,
 
     /* benchmark loop */
     {   BMK_timedFnState_t* const tfs = BMK_createTimedFnState(g_nbIterations * 1000, 1000);
+        void* const avoidStrictAliasingPtr = &dstBuff;
+        BMK_benchParams_t bp;
         BMK_runTime_t bestResult;
         bestResult.sumOfReturn = 0;
         bestResult.nanoSecPerRun = (unsigned long long)(-1LL);
         assert(tfs != NULL);
+
+        bp.benchFn = benchFunction;
+        bp.benchPayload = buff2;
+        bp.initFn = NULL;
+        bp.initPayload = NULL;
+        bp.errorFn = ZSTD_isError;
+        bp.blockCount = 1;
+        bp.srcBuffers = &src;
+        bp.srcSizes = &srcSize;
+        bp.dstBuffers = (void* const*) avoidStrictAliasingPtr;  /* circumvent strict aliasing warning on gcc-8,
+                                                                 * because gcc considers that `void* const *`  and `void**` are 2 different types */
+        bp.dstCapacities = &dstBuffSize;
+        bp.blockResults = NULL;
+
         for (;;) {
-            void* const dstBuffv = dstBuff;
-            BMK_runOutcome_t const bOutcome =
-                    BMK_benchTimedFn( tfs,
-                            benchFunction, buff2,
-                            NULL, NULL,   /* initFn */
-                            1,  /* blockCount */
-                            &src, &srcSize,
-                            &dstBuffv, &dstBuffSize,
-                            NULL);
+            BMK_runOutcome_t const bOutcome = BMK_benchTimedFn(tfs, bp);
 
             if (!BMK_isSuccessful_runOutcome(bOutcome)) {
                 DISPLAY("ERROR benchmarking function ! ! \n");
@@ -616,7 +606,7 @@ static int benchFiles(U32 benchNb,
                 benchedSize = (size_t)inFileSize;
             if ((U64)benchedSize < inFileSize) {
                 DISPLAY("Not enough memory for '%s' full size; testing %u MB only... \n",
-                        inFileName, (U32)(benchedSize>>20));
+                        inFileName, (unsigned)(benchedSize>>20));
         }   }
 
         /* Alloc */
@@ -744,7 +734,7 @@ int main(int argc, const char** argv)
                 if (longCommandWArg(&argument, "chainLog=") || longCommandWArg(&argument, "clog=")) { cparams.chainLog = readU32FromChar(&argument); if (argument[0]==',') { argument++; continue; } else break; }
                 if (longCommandWArg(&argument, "hashLog=") || longCommandWArg(&argument, "hlog=")) { cparams.hashLog = readU32FromChar(&argument); if (argument[0]==',') { argument++; continue; } else break; }
                 if (longCommandWArg(&argument, "searchLog=") || longCommandWArg(&argument, "slog=")) { cparams.searchLog = readU32FromChar(&argument); if (argument[0]==',') { argument++; continue; } else break; }
-                if (longCommandWArg(&argument, "searchLength=") || longCommandWArg(&argument, "slen=")) { cparams.searchLength = readU32FromChar(&argument); if (argument[0]==',') { argument++; continue; } else break; }
+                if (longCommandWArg(&argument, "minMatch=") || longCommandWArg(&argument, "mml=")) { cparams.minMatch = readU32FromChar(&argument); if (argument[0]==',') { argument++; continue; } else break; }
                 if (longCommandWArg(&argument, "targetLength=") || longCommandWArg(&argument, "tlen=")) { cparams.targetLength = readU32FromChar(&argument); if (argument[0]==',') { argument++; continue; } else break; }
                 if (longCommandWArg(&argument, "strategy=") || longCommandWArg(&argument, "strat=")) { cparams.strategy = (ZSTD_strategy)(readU32FromChar(&argument)); if (argument[0]==',') { argument++; continue; } else break; }
                 if (longCommandWArg(&argument, "level=") || longCommandWArg(&argument, "lvl=")) { cLevel = (int)readU32FromChar(&argument); cparams = ZSTD_getCParams(cLevel, 0, 0); if (argument[0]==',') { argument++; continue; } else break; }
diff --git a/tests/fuzz/Makefile b/tests/fuzz/Makefile
index d9b00fd2af79..4130f18e330c 100644
--- a/tests/fuzz/Makefile
+++ b/tests/fuzz/Makefile
@@ -41,7 +41,7 @@ FUZZ_ARFLAGS := $(ARFLAGS)
 FUZZ_TARGET_FLAGS = $(FUZZ_CPPFLAGS) $(FUZZ_CXXFLAGS) $(FUZZ_LDFLAGS)
 
 FUZZ_HEADERS := fuzz_helpers.h fuzz.h zstd_helpers.h
-FUZZ_SRC := zstd_helpers.c
+FUZZ_SRC := $(PRGDIR)/util.c zstd_helpers.c
 
 ZSTDCOMMON_SRC := $(ZSTDDIR)/common/*.c
 ZSTDCOMP_SRC   := $(ZSTDDIR)/compress/*.c
@@ -90,7 +90,7 @@ stream_decompress: $(FUZZ_HEADERS) $(FUZZ_OBJ) stream_decompress.o
 block_decompress: $(FUZZ_HEADERS) $(FUZZ_OBJ) block_decompress.o
 	$(CXX) $(FUZZ_TARGET_FLAGS) $(FUZZ_OBJ) block_decompress.o $(LIB_FUZZING_ENGINE) -o $@
 
-libregression.a: $(FUZZ_HEADERS) $(PRGDIR)/util.h regression_driver.o
+libregression.a: $(FUZZ_HEADERS) $(PRGDIR)/util.h $(PRGDIR)/util.c regression_driver.o
 	$(AR) $(FUZZ_ARFLAGS) $@ regression_driver.o
 
 # Install libfuzzer (not usable for MSAN testing)
@@ -99,7 +99,7 @@ libregression.a: $(FUZZ_HEADERS) $(PRGDIR)/util.h regression_driver.o
 .PHONY: libFuzzer
 libFuzzer:
 	@$(RM) -rf Fuzzer
-	@git clone https://chromium.googlesource.com/chromium/llvm-project/llvm/lib/Fuzzer
+	@git clone https://chromium.googlesource.com/chromium/llvm-project/compiler-rt/lib/fuzzer Fuzzer
 	@cd Fuzzer && ./build.sh
 
 corpora/%_seed_corpus.zip:
diff --git a/tests/fuzz/simple_round_trip.c b/tests/fuzz/simple_round_trip.c
index 0921106dea8e..83608b6e79ea 100644
--- a/tests/fuzz/simple_round_trip.c
+++ b/tests/fuzz/simple_round_trip.c
@@ -40,9 +40,9 @@ static size_t roundTripTest(void *result, size_t resultCapacity,
         ZSTD_outBuffer out = {compressed, compressedCapacity, 0};
         size_t err;
 
-        ZSTD_CCtx_reset(cctx);
+        ZSTD_CCtx_reset(cctx, ZSTD_reset_session_only);
         FUZZ_setRandomParameters(cctx, srcSize, &seed);
-        err = ZSTD_compress_generic(cctx, &out, &in, ZSTD_e_end);
+        err = ZSTD_compressStream2(cctx, &out, &in, ZSTD_e_end);
         FUZZ_ZASSERT(err);
         FUZZ_ASSERT(err == 0);
         cSize = out.pos;
diff --git a/tests/fuzz/stream_round_trip.c b/tests/fuzz/stream_round_trip.c
index 72d70495f836..d903bcb29b11 100644
--- a/tests/fuzz/stream_round_trip.c
+++ b/tests/fuzz/stream_round_trip.c
@@ -56,7 +56,7 @@ static size_t compress(uint8_t *dst, size_t capacity,
                        const uint8_t *src, size_t srcSize)
 {
     size_t dstSize = 0;
-    ZSTD_CCtx_reset(cctx);
+    ZSTD_CCtx_reset(cctx, ZSTD_reset_session_only);
     FUZZ_setRandomParameters(cctx, srcSize, &seed);
 
     while (srcSize > 0) {
@@ -72,7 +72,7 @@ static size_t compress(uint8_t *dst, size_t capacity,
                 case 1: /* fall-though */
                 case 2: {
                     size_t const ret =
-                        ZSTD_compress_generic(cctx, &out, &in, ZSTD_e_flush);
+                        ZSTD_compressStream2(cctx, &out, &in, ZSTD_e_flush);
                     FUZZ_ZASSERT(ret);
                     if (ret == 0)
                         mode = -1;
@@ -80,11 +80,11 @@ static size_t compress(uint8_t *dst, size_t capacity,
                 }
                 case 3: {
                     size_t ret =
-                        ZSTD_compress_generic(cctx, &out, &in, ZSTD_e_end);
+                        ZSTD_compressStream2(cctx, &out, &in, ZSTD_e_end);
                     FUZZ_ZASSERT(ret);
                     /* Reset the compressor when the frame is finished */
                     if (ret == 0) {
-                        ZSTD_CCtx_reset(cctx);
+                        ZSTD_CCtx_reset(cctx, ZSTD_reset_session_only);
                         if ((FUZZ_rand(&seed) & 7) == 0) {
                             size_t const remaining = in.size - in.pos;
                             FUZZ_setRandomParameters(cctx, remaining, &seed);
@@ -95,7 +95,7 @@ static size_t compress(uint8_t *dst, size_t capacity,
                 }
                 default: {
                     size_t const ret =
-                        ZSTD_compress_generic(cctx, &out, &in, ZSTD_e_continue);
+                        ZSTD_compressStream2(cctx, &out, &in, ZSTD_e_continue);
                     FUZZ_ZASSERT(ret);
                     mode = -1;
                 }
@@ -108,7 +108,7 @@ static size_t compress(uint8_t *dst, size_t capacity,
     for (;;) {
         ZSTD_inBuffer in = {NULL, 0, 0};
         ZSTD_outBuffer out = makeOutBuffer(dst, capacity);
-        size_t const ret = ZSTD_compress_generic(cctx, &out, &in, ZSTD_e_end);
+        size_t const ret = ZSTD_compressStream2(cctx, &out, &in, ZSTD_e_end);
         FUZZ_ZASSERT(ret);
 
         dst += out.pos;
diff --git a/tests/fuzz/zstd_helpers.c b/tests/fuzz/zstd_helpers.c
index bf5eccff83c7..10163e1512b0 100644
--- a/tests/fuzz/zstd_helpers.c
+++ b/tests/fuzz/zstd_helpers.c
@@ -13,7 +13,7 @@
 #include "fuzz_helpers.h"
 #include "zstd.h"
 
-static void set(ZSTD_CCtx *cctx, ZSTD_cParameter param, unsigned value)
+static void set(ZSTD_CCtx *cctx, ZSTD_cParameter param, int value)
 {
     FUZZ_ZASSERT(ZSTD_CCtx_setParameter(cctx, param, value));
 }
@@ -32,10 +32,10 @@ ZSTD_compressionParameters FUZZ_randomCParams(size_t srcSize, uint32_t *state)
     cParams.hashLog = FUZZ_rand32(state, ZSTD_HASHLOG_MIN, 15);
     cParams.chainLog = FUZZ_rand32(state, ZSTD_CHAINLOG_MIN, 16);
     cParams.searchLog = FUZZ_rand32(state, ZSTD_SEARCHLOG_MIN, 9);
-    cParams.searchLength = FUZZ_rand32(state, ZSTD_SEARCHLENGTH_MIN,
-                                              ZSTD_SEARCHLENGTH_MAX);
+    cParams.minMatch = FUZZ_rand32(state, ZSTD_MINMATCH_MIN,
+                                          ZSTD_MINMATCH_MAX);
     cParams.targetLength = FUZZ_rand32(state, 0, 512);
-    cParams.strategy = FUZZ_rand32(state, ZSTD_fast, ZSTD_btultra);
+    cParams.strategy = FUZZ_rand32(state, ZSTD_STRATEGY_MIN, ZSTD_STRATEGY_MAX);
     return ZSTD_adjustCParams(cParams, srcSize, 0);
 }
 
@@ -60,25 +60,25 @@ ZSTD_parameters FUZZ_randomParams(size_t srcSize, uint32_t *state)
 void FUZZ_setRandomParameters(ZSTD_CCtx *cctx, size_t srcSize, uint32_t *state)
 {
     ZSTD_compressionParameters cParams = FUZZ_randomCParams(srcSize, state);
-    set(cctx, ZSTD_p_windowLog, cParams.windowLog);
-    set(cctx, ZSTD_p_hashLog, cParams.hashLog);
-    set(cctx, ZSTD_p_chainLog, cParams.chainLog);
-    set(cctx, ZSTD_p_searchLog, cParams.searchLog);
-    set(cctx, ZSTD_p_minMatch, cParams.searchLength);
-    set(cctx, ZSTD_p_targetLength, cParams.targetLength);
-    set(cctx, ZSTD_p_compressionStrategy, cParams.strategy);
+    set(cctx, ZSTD_c_windowLog, cParams.windowLog);
+    set(cctx, ZSTD_c_hashLog, cParams.hashLog);
+    set(cctx, ZSTD_c_chainLog, cParams.chainLog);
+    set(cctx, ZSTD_c_searchLog, cParams.searchLog);
+    set(cctx, ZSTD_c_minMatch, cParams.minMatch);
+    set(cctx, ZSTD_c_targetLength, cParams.targetLength);
+    set(cctx, ZSTD_c_strategy, cParams.strategy);
     /* Select frame parameters */
-    setRand(cctx, ZSTD_p_contentSizeFlag, 0, 1, state);
-    setRand(cctx, ZSTD_p_checksumFlag, 0, 1, state);
-    setRand(cctx, ZSTD_p_dictIDFlag, 0, 1, state);
-    setRand(cctx, ZSTD_p_forceAttachDict, -2, 2, state);
+    setRand(cctx, ZSTD_c_contentSizeFlag, 0, 1, state);
+    setRand(cctx, ZSTD_c_checksumFlag, 0, 1, state);
+    setRand(cctx, ZSTD_c_dictIDFlag, 0, 1, state);
+    setRand(cctx, ZSTD_c_forceAttachDict, 0, 2, state);
     /* Select long distance matchig parameters */
-    setRand(cctx, ZSTD_p_enableLongDistanceMatching, 0, 1, state);
-    setRand(cctx, ZSTD_p_ldmHashLog, ZSTD_HASHLOG_MIN, 16, state);
-    setRand(cctx, ZSTD_p_ldmMinMatch, ZSTD_LDM_MINMATCH_MIN,
+    setRand(cctx, ZSTD_c_enableLongDistanceMatching, 0, 1, state);
+    setRand(cctx, ZSTD_c_ldmHashLog, ZSTD_HASHLOG_MIN, 16, state);
+    setRand(cctx, ZSTD_c_ldmMinMatch, ZSTD_LDM_MINMATCH_MIN,
             ZSTD_LDM_MINMATCH_MAX, state);
-    setRand(cctx, ZSTD_p_ldmBucketSizeLog, 0, ZSTD_LDM_BUCKETSIZELOG_MAX,
+    setRand(cctx, ZSTD_c_ldmBucketSizeLog, 0, ZSTD_LDM_BUCKETSIZELOG_MAX,
             state);
-    setRand(cctx, ZSTD_p_ldmHashEveryLog, 0,
-            ZSTD_WINDOWLOG_MAX - ZSTD_HASHLOG_MIN, state);
+    setRand(cctx, ZSTD_c_ldmHashRateLog, ZSTD_LDM_HASHRATELOG_MIN,
+            ZSTD_LDM_HASHRATELOG_MAX, state);
 }
diff --git a/tests/fuzzer.c b/tests/fuzzer.c
index 5616285b9ed7..946844e1614a 100644
--- a/tests/fuzzer.c
+++ b/tests/fuzzer.c
@@ -47,8 +47,8 @@
 #define MB *(1U<<20)
 #define GB *(1U<<30)
 
-static const U32 FUZ_compressibility_default = 50;
-static const U32 nbTestsDefault = 30000;
+static const int FUZ_compressibility_default = 50;
+static const int nbTestsDefault = 30000;
 
 
 /*-************************************
@@ -88,7 +88,7 @@ void FUZ_bug976(void)
 #define MAX(a,b) ((a)>(b)?(a):(b))
 
 #define FUZ_rotl32(x,r) ((x << r) | (x >> (32 - r)))
-static unsigned FUZ_rand(unsigned* src)
+static U32 FUZ_rand(U32* src)
 {
     static const U32 prime1 = 2654435761U;
     static const U32 prime2 = 2246822519U;
@@ -100,7 +100,7 @@ static unsigned FUZ_rand(unsigned* src)
     return rand32 >> 5;
 }
 
-static unsigned FUZ_highbit32(U32 v32)
+static U32 FUZ_highbit32(U32 v32)
 {
     unsigned nbBits = 0;
     if (v32==0) return 0;
@@ -155,7 +155,7 @@ static void* FUZ_mallocDebug(void* counter, size_t size)
     void* const ptr = malloc(size);
     if (ptr==NULL) return NULL;
     DISPLAYLEVEL(4, "allocating %u KB => effectively %u KB \n",
-        (U32)(size >> 10), (U32)(malloc_size(ptr) >> 10));  /* OS-X specific */
+        (unsigned)(size >> 10), (unsigned)(malloc_size(ptr) >> 10));  /* OS-X specific */
     mcPtr->totalMalloc += size;
     mcPtr->currentMalloc += size;
     if (mcPtr->currentMalloc > mcPtr->peakMalloc)
@@ -167,7 +167,7 @@ static void* FUZ_mallocDebug(void* counter, size_t size)
 static void FUZ_freeDebug(void* counter, void* address)
 {
     mallocCounter_t* const mcPtr = (mallocCounter_t*)counter;
-    DISPLAYLEVEL(4, "freeing %u KB \n", (U32)(malloc_size(address) >> 10));
+    DISPLAYLEVEL(4, "freeing %u KB \n", (unsigned)(malloc_size(address) >> 10));
     mcPtr->nbFree += 1;
     mcPtr->currentMalloc -= malloc_size(address);  /* OS-X specific */
     free(address);
@@ -176,9 +176,9 @@ static void FUZ_freeDebug(void* counter, void* address)
 static void FUZ_displayMallocStats(mallocCounter_t count)
 {
     DISPLAYLEVEL(3, "peak:%6u KB,  nbMallocs:%2u, total:%6u KB \n",
-        (U32)(count.peakMalloc >> 10),
+        (unsigned)(count.peakMalloc >> 10),
         count.nbMalloc,
-        (U32)(count.totalMalloc >> 10));
+        (unsigned)(count.totalMalloc >> 10));
 }
 
 static int FUZ_mallocTests_internal(unsigned seed, double compressibility, unsigned part,
@@ -226,18 +226,16 @@ static int FUZ_mallocTests_internal(unsigned seed, double compressibility, unsig
 
     /* advanced MT API test */
     if (part <= 3)
-    {   U32 nbThreads;
+    {   unsigned nbThreads;
         for (nbThreads=1; nbThreads<=4; nbThreads++) {
             int compressionLevel;
             for (compressionLevel=1; compressionLevel<=6; compressionLevel++) {
                 mallocCounter_t malcount = INIT_MALLOC_COUNTER;
                 ZSTD_customMem const cMem = { FUZ_mallocDebug, FUZ_freeDebug, &malcount };
                 ZSTD_CCtx* const cctx = ZSTD_createCCtx_advanced(cMem);
-                ZSTD_outBuffer out = { outBuffer, outSize, 0 };
-                ZSTD_inBuffer in = { inBuffer, inSize, 0 };
-                CHECK_Z( ZSTD_CCtx_setParameter(cctx, ZSTD_p_compressionLevel, (U32)compressionLevel) );
-                CHECK_Z( ZSTD_CCtx_setParameter(cctx, ZSTD_p_nbWorkers, nbThreads) );
-                while ( ZSTD_compress_generic(cctx, &out, &in, ZSTD_e_end) ) {}
+                CHECK_Z( ZSTD_CCtx_setParameter(cctx, ZSTD_c_compressionLevel, compressionLevel) );
+                CHECK_Z( ZSTD_CCtx_setParameter(cctx, ZSTD_c_nbWorkers, nbThreads) );
+                CHECK_Z( ZSTD_compress2(cctx, outBuffer, outSize, inBuffer, inSize) );
                 ZSTD_freeCCtx(cctx);
                 DISPLAYLEVEL(3, "compress_generic,-T%u,end level %i : ",
                                 nbThreads, compressionLevel);
@@ -246,7 +244,7 @@ static int FUZ_mallocTests_internal(unsigned seed, double compressibility, unsig
 
     /* advanced MT streaming API test */
     if (part <= 4)
-    {   U32 nbThreads;
+    {   unsigned nbThreads;
         for (nbThreads=1; nbThreads<=4; nbThreads++) {
             int compressionLevel;
             for (compressionLevel=1; compressionLevel<=6; compressionLevel++) {
@@ -255,10 +253,10 @@ static int FUZ_mallocTests_internal(unsigned seed, double compressibility, unsig
                 ZSTD_CCtx* const cctx = ZSTD_createCCtx_advanced(cMem);
                 ZSTD_outBuffer out = { outBuffer, outSize, 0 };
                 ZSTD_inBuffer in = { inBuffer, inSize, 0 };
-                CHECK_Z( ZSTD_CCtx_setParameter(cctx, ZSTD_p_compressionLevel, (U32)compressionLevel) );
-                CHECK_Z( ZSTD_CCtx_setParameter(cctx, ZSTD_p_nbWorkers, nbThreads) );
-                CHECK_Z( ZSTD_compress_generic(cctx, &out, &in, ZSTD_e_continue) );
-                while ( ZSTD_compress_generic(cctx, &out, &in, ZSTD_e_end) ) {}
+                CHECK_Z( ZSTD_CCtx_setParameter(cctx, ZSTD_c_compressionLevel, compressionLevel) );
+                CHECK_Z( ZSTD_CCtx_setParameter(cctx, ZSTD_c_nbWorkers, nbThreads) );
+                CHECK_Z( ZSTD_compressStream2(cctx, &out, &in, ZSTD_e_continue) );
+                while ( ZSTD_compressStream2(cctx, &out, &in, ZSTD_e_end) ) {}
                 ZSTD_freeCCtx(cctx);
                 DISPLAYLEVEL(3, "compress_generic,-T%u,continue level %i : ",
                                 nbThreads, compressionLevel);
@@ -311,9 +309,8 @@ static int basicUnitTests(U32 seed, double compressibility)
     size_t const compressedBufferSize = ZSTD_compressBound(CNBuffSize);
     void* const compressedBuffer = malloc(compressedBufferSize);
     void* const decodedBuffer = malloc(CNBuffSize);
-    ZSTD_DCtx* dctx = ZSTD_createDCtx();
     int testResult = 0;
-    U32 testNb=0;
+    unsigned testNb=0;
     size_t cSize;
 
     /* Create compressible noise */
@@ -325,37 +322,45 @@ static int basicUnitTests(U32 seed, double compressibility)
     RDG_genBuffer(CNBuffer, CNBuffSize, compressibility, 0., seed);
 
     /* Basic tests */
-    DISPLAYLEVEL(3, "test%3i : ZSTD_getErrorName : ", testNb++);
+    DISPLAYLEVEL(3, "test%3u : ZSTD_getErrorName : ", testNb++);
     {   const char* errorString = ZSTD_getErrorName(0);
         DISPLAYLEVEL(3, "OK : %s \n", errorString);
     }
 
-    DISPLAYLEVEL(3, "test%3i : ZSTD_getErrorName with wrong value : ", testNb++);
+    DISPLAYLEVEL(3, "test%3u : ZSTD_getErrorName with wrong value : ", testNb++);
     {   const char* errorString = ZSTD_getErrorName(499);
         DISPLAYLEVEL(3, "OK : %s \n", errorString);
     }
 
-    DISPLAYLEVEL(3, "test%3i : min compression level : ", testNb++);
+    DISPLAYLEVEL(3, "test%3u : min compression level : ", testNb++);
     {   int const mcl = ZSTD_minCLevel();
         DISPLAYLEVEL(3, "%i (OK) \n", mcl);
     }
 
-    DISPLAYLEVEL(3, "test%3i : compress %u bytes : ", testNb++, (U32)CNBuffSize);
+    DISPLAYLEVEL(3, "test%3u : compress %u bytes : ", testNb++, (unsigned)CNBuffSize);
     {   ZSTD_CCtx* const cctx = ZSTD_createCCtx();
         if (cctx==NULL) goto _output_error;
         CHECKPLUS(r, ZSTD_compressCCtx(cctx,
                             compressedBuffer, compressedBufferSize,
                             CNBuffer, CNBuffSize, 1),
                   cSize=r );
-        DISPLAYLEVEL(3, "OK (%u bytes : %.2f%%)\n", (U32)cSize, (double)cSize/CNBuffSize*100);
+        DISPLAYLEVEL(3, "OK (%u bytes : %.2f%%)\n", (unsigned)cSize, (double)cSize/CNBuffSize*100);
 
         DISPLAYLEVEL(3, "test%3i : size of cctx for level 1 : ", testNb++);
         {   size_t const cctxSize = ZSTD_sizeof_CCtx(cctx);
-            DISPLAYLEVEL(3, "%u bytes \n", (U32)cctxSize);
+            DISPLAYLEVEL(3, "%u bytes \n", (unsigned)cctxSize);
         }
         ZSTD_freeCCtx(cctx);
     }
 
+    DISPLAYLEVEL(3, "test%3i : decompress skippable frame -8 size : ", testNb++);
+    {
+       char const skippable8[] = "\x50\x2a\x4d\x18\xf8\xff\xff\xff";
+       size_t const size = ZSTD_decompress(NULL, 0, skippable8, 8);
+       if (!ZSTD_isError(size)) goto _output_error;
+    }
+    DISPLAYLEVEL(3, "OK \n");
+
 
     DISPLAYLEVEL(3, "test%3i : ZSTD_getFrameContentSize test : ", testNb++);
     {   unsigned long long const rSize = ZSTD_getFrameContentSize(compressedBuffer, cSize);
@@ -369,7 +374,7 @@ static int basicUnitTests(U32 seed, double compressibility)
     }
     DISPLAYLEVEL(3, "OK \n");
 
-    DISPLAYLEVEL(3, "test%3i : decompress %u bytes : ", testNb++, (U32)CNBuffSize);
+    DISPLAYLEVEL(3, "test%3i : decompress %u bytes : ", testNb++, (unsigned)CNBuffSize);
     { size_t const r = ZSTD_decompress(decodedBuffer, CNBuffSize, compressedBuffer, cSize);
       if (r != CNBuffSize) goto _output_error; }
     DISPLAYLEVEL(3, "OK \n");
@@ -383,13 +388,27 @@ static int basicUnitTests(U32 seed, double compressibility)
 
 
     DISPLAYLEVEL(3, "test%3i : decompress with null dict : ", testNb++);
-    { size_t const r = ZSTD_decompress_usingDict(dctx, decodedBuffer, CNBuffSize, compressedBuffer, cSize, NULL, 0);
-      if (r != CNBuffSize) goto _output_error; }
+    {   ZSTD_DCtx* const dctx = ZSTD_createDCtx(); assert(dctx != NULL);
+        {   size_t const r = ZSTD_decompress_usingDict(dctx,
+                                                    decodedBuffer, CNBuffSize,
+                                                    compressedBuffer, cSize,
+                                                    NULL, 0);
+            if (r != CNBuffSize) goto _output_error;
+        }
+        ZSTD_freeDCtx(dctx);
+    }
     DISPLAYLEVEL(3, "OK \n");
 
     DISPLAYLEVEL(3, "test%3i : decompress with null DDict : ", testNb++);
-    { size_t const r = ZSTD_decompress_usingDDict(dctx, decodedBuffer, CNBuffSize, compressedBuffer, cSize, NULL);
-      if (r != CNBuffSize) goto _output_error; }
+    {   ZSTD_DCtx* const dctx = ZSTD_createDCtx(); assert(dctx != NULL);
+        {   size_t const r = ZSTD_decompress_usingDDict(dctx,
+                                                    decodedBuffer, CNBuffSize,
+                                                    compressedBuffer, cSize,
+                                                    NULL);
+            if (r != CNBuffSize) goto _output_error;
+        }
+        ZSTD_freeDCtx(dctx);
+    }
     DISPLAYLEVEL(3, "OK \n");
 
     DISPLAYLEVEL(3, "test%3i : decompress with 1 missing byte : ", testNb++);
@@ -411,11 +430,55 @@ static int basicUnitTests(U32 seed, double compressibility)
     DISPLAYLEVEL(3, "OK \n");
 
     DISPLAYLEVEL(3, "test%3d : check CCtx size after compressing empty input : ", testNb++);
-    {   ZSTD_CCtx* cctx = ZSTD_createCCtx();
+    {   ZSTD_CCtx* const cctx = ZSTD_createCCtx();
         size_t const r = ZSTD_compressCCtx(cctx, compressedBuffer, compressedBufferSize, NULL, 0, 19);
         if (ZSTD_isError(r)) goto _output_error;
         if (ZSTD_sizeof_CCtx(cctx) > (1U << 20)) goto _output_error;
         ZSTD_freeCCtx(cctx);
+        cSize = r;
+    }
+    DISPLAYLEVEL(3, "OK \n");
+
+    DISPLAYLEVEL(3, "test%3d : decompress empty frame into NULL : ", testNb++);
+    {   size_t const r = ZSTD_decompress(NULL, 0, compressedBuffer, cSize);
+        if (ZSTD_isError(r)) goto _output_error;
+        if (r != 0) goto _output_error;
+    }
+    {   ZSTD_CCtx* const cctx = ZSTD_createCCtx();
+        ZSTD_outBuffer output;
+        if (cctx==NULL) goto _output_error;
+        output.dst = compressedBuffer;
+        output.size = compressedBufferSize;
+        output.pos = 0;
+        CHECK_Z( ZSTD_initCStream(cctx, 1) );    /* content size unknown */
+        CHECK_Z( ZSTD_flushStream(cctx, &output) );   /* ensure no possibility to "concatenate" and determine the content size */
+        CHECK_Z( ZSTD_endStream(cctx, &output) );
+        ZSTD_freeCCtx(cctx);
+        /* single scan decompression */
+        {   size_t const r = ZSTD_decompress(NULL, 0, compressedBuffer, output.pos);
+            if (ZSTD_isError(r)) goto _output_error;
+            if (r != 0) goto _output_error;
+        }
+        /* streaming decompression */
+        {   ZSTD_DCtx* const dstream = ZSTD_createDStream();
+            ZSTD_inBuffer dinput;
+            ZSTD_outBuffer doutput;
+            size_t ipos;
+            if (dstream==NULL) goto _output_error;
+            dinput.src = compressedBuffer;
+            dinput.size = 0;
+            dinput.pos = 0;
+            doutput.dst = NULL;
+            doutput.size = 0;
+            doutput.pos = 0;
+            CHECK_Z ( ZSTD_initDStream(dstream) );
+            for (ipos=1; ipos<=output.pos; ipos++) {
+                dinput.size = ipos;
+                CHECK_Z ( ZSTD_decompressStream(dstream, &doutput, &dinput) );
+            }
+            if (doutput.pos != 0) goto _output_error;
+            ZSTD_freeDStream(dstream);
+        }
     }
     DISPLAYLEVEL(3, "OK \n");
 
@@ -437,7 +500,8 @@ static int basicUnitTests(U32 seed, double compressibility)
     DISPLAYLEVEL(3, "OK \n");
 
     DISPLAYLEVEL(3, "test%3d : re-using a CCtx should compress the same : ", testNb++);
-    {   int i;
+    {   size_t const sampleSize = 30;
+        int i;
         for (i=0; i<20; i++)
             ((char*)CNBuffer)[i] = (char)i;   /* ensure no match during initial section */
         memcpy((char*)CNBuffer + 20, CNBuffer, 10);   /* create one match, starting from beginning of sample, which is the difficult case (see #1241) */
@@ -445,54 +509,117 @@ static int basicUnitTests(U32 seed, double compressibility)
             ZSTD_CCtx* const cctx = ZSTD_createCCtx();
             size_t size1, size2;
             DISPLAYLEVEL(5, "l%i ", i);
-            size1 = ZSTD_compressCCtx(cctx, compressedBuffer, compressedBufferSize, CNBuffer, 30, i);
+            size1 = ZSTD_compressCCtx(cctx, compressedBuffer, compressedBufferSize, CNBuffer, sampleSize, i);
             CHECK_Z(size1);
-            size2 = ZSTD_compressCCtx(cctx, compressedBuffer, compressedBufferSize, CNBuffer, 30, i);
+
+            size2 = ZSTD_compressCCtx(cctx, compressedBuffer, compressedBufferSize, CNBuffer, sampleSize, i);
             CHECK_Z(size2);
             CHECK_EQ(size1, size2);
 
+            CHECK_Z( ZSTD_CCtx_setParameter(cctx, ZSTD_c_compressionLevel, i) );
+            size2 = ZSTD_compress2(cctx, compressedBuffer, compressedBufferSize, CNBuffer, sampleSize);
+            CHECK_Z(size2);
+            CHECK_EQ(size1, size2);
+
+            size2 = ZSTD_compress2(cctx, compressedBuffer, ZSTD_compressBound(sampleSize) - 1, CNBuffer, sampleSize);  /* force streaming, as output buffer is not large enough to guarantee success */
+            CHECK_Z(size2);
+            CHECK_EQ(size1, size2);
+
+            {   ZSTD_inBuffer inb;
+                ZSTD_outBuffer outb;
+                inb.src = CNBuffer;
+                inb.pos = 0;
+                inb.size = sampleSize;
+                outb.dst = compressedBuffer;
+                outb.pos = 0;
+                outb.size = ZSTD_compressBound(sampleSize) - 1;  /* force streaming, as output buffer is not large enough to guarantee success */
+                CHECK_Z( ZSTD_compressStream2(cctx, &outb, &inb, ZSTD_e_end) );
+                assert(inb.pos == inb.size);
+                CHECK_EQ(size1, outb.pos);
+            }
+
             ZSTD_freeCCtx(cctx);
         }
     }
     DISPLAYLEVEL(3, "OK \n");
 
+    DISPLAYLEVEL(3, "test%3d : btultra2 & 1st block : ", testNb++);
+    {   size_t const sampleSize = 1024;
+        ZSTD_CCtx* const cctx = ZSTD_createCCtx();
+        ZSTD_inBuffer inb;
+        ZSTD_outBuffer outb;
+        inb.src = CNBuffer;
+        inb.pos = 0;
+        inb.size = 0;
+        outb.dst = compressedBuffer;
+        outb.pos = 0;
+        outb.size = compressedBufferSize;
+        CHECK_Z( ZSTD_CCtx_setParameter(cctx, ZSTD_c_compressionLevel, ZSTD_maxCLevel()) );
+
+        inb.size = sampleSize;   /* start with something, so that context is already used */
+        CHECK_Z( ZSTD_compressStream2(cctx, &outb, &inb, ZSTD_e_end) );   /* will break internal assert if stats_init is not disabled */
+        assert(inb.pos == inb.size);
+        outb.pos = 0;     /* cancel output */
+
+        CHECK_Z( ZSTD_CCtx_setPledgedSrcSize(cctx, sampleSize) );
+        inb.size = 4;   /* too small size : compression will be skipped */
+        inb.pos = 0;
+        CHECK_Z( ZSTD_compressStream2(cctx, &outb, &inb, ZSTD_e_flush) );
+        assert(inb.pos == inb.size);
+
+        inb.size += 5;   /* too small size : compression will be skipped */
+        CHECK_Z( ZSTD_compressStream2(cctx, &outb, &inb, ZSTD_e_flush) );
+        assert(inb.pos == inb.size);
+
+        inb.size += 11;   /* small enough to attempt compression */
+        CHECK_Z( ZSTD_compressStream2(cctx, &outb, &inb, ZSTD_e_flush) );
+        assert(inb.pos == inb.size);
+
+        assert(inb.pos < sampleSize);
+        inb.size = sampleSize;   /* large enough to trigger stats_init, but no longer at beginning */
+        CHECK_Z( ZSTD_compressStream2(cctx, &outb, &inb, ZSTD_e_end) );   /* will break internal assert if stats_init is not disabled */
+        assert(inb.pos == inb.size);
+        ZSTD_freeCCtx(cctx);
+    }
+    DISPLAYLEVEL(3, "OK \n");
+
     DISPLAYLEVEL(3, "test%3d : ZSTD_CCtx_getParameter() : ", testNb++);
     {   ZSTD_CCtx* const cctx = ZSTD_createCCtx();
         ZSTD_outBuffer out = {NULL, 0, 0};
         ZSTD_inBuffer in = {NULL, 0, 0};
-        unsigned value;
+        int value;
 
-        CHECK_Z(ZSTD_CCtx_getParameter(cctx, ZSTD_p_compressionLevel, &value));
+        CHECK_Z(ZSTD_CCtx_getParameter(cctx, ZSTD_c_compressionLevel, &value));
         CHECK_EQ(value, 3);
-        CHECK_Z(ZSTD_CCtx_getParameter(cctx, ZSTD_p_hashLog, &value));
+        CHECK_Z(ZSTD_CCtx_getParameter(cctx, ZSTD_c_hashLog, &value));
         CHECK_EQ(value, 0);
-        CHECK_Z(ZSTD_CCtx_setParameter(cctx, ZSTD_p_hashLog, ZSTD_HASHLOG_MIN));
-        CHECK_Z(ZSTD_CCtx_getParameter(cctx, ZSTD_p_compressionLevel, &value));
+        CHECK_Z(ZSTD_CCtx_setParameter(cctx, ZSTD_c_hashLog, ZSTD_HASHLOG_MIN));
+        CHECK_Z(ZSTD_CCtx_getParameter(cctx, ZSTD_c_compressionLevel, &value));
         CHECK_EQ(value, 3);
-        CHECK_Z(ZSTD_CCtx_getParameter(cctx, ZSTD_p_hashLog, &value));
+        CHECK_Z(ZSTD_CCtx_getParameter(cctx, ZSTD_c_hashLog, &value));
         CHECK_EQ(value, ZSTD_HASHLOG_MIN);
-        CHECK_Z(ZSTD_CCtx_setParameter(cctx, ZSTD_p_compressionLevel, 7));
-        CHECK_Z(ZSTD_CCtx_getParameter(cctx, ZSTD_p_compressionLevel, &value));
+        CHECK_Z(ZSTD_CCtx_setParameter(cctx, ZSTD_c_compressionLevel, 7));
+        CHECK_Z(ZSTD_CCtx_getParameter(cctx, ZSTD_c_compressionLevel, &value));
         CHECK_EQ(value, 7);
-        CHECK_Z(ZSTD_CCtx_getParameter(cctx, ZSTD_p_hashLog, &value));
+        CHECK_Z(ZSTD_CCtx_getParameter(cctx, ZSTD_c_hashLog, &value));
         CHECK_EQ(value, ZSTD_HASHLOG_MIN);
         /* Start a compression job */
-        ZSTD_compress_generic(cctx, &out, &in, ZSTD_e_continue);
-        CHECK_Z(ZSTD_CCtx_getParameter(cctx, ZSTD_p_compressionLevel, &value));
+        ZSTD_compressStream2(cctx, &out, &in, ZSTD_e_continue);
+        CHECK_Z(ZSTD_CCtx_getParameter(cctx, ZSTD_c_compressionLevel, &value));
         CHECK_EQ(value, 7);
-        CHECK_Z(ZSTD_CCtx_getParameter(cctx, ZSTD_p_hashLog, &value));
+        CHECK_Z(ZSTD_CCtx_getParameter(cctx, ZSTD_c_hashLog, &value));
         CHECK_EQ(value, ZSTD_HASHLOG_MIN);
         /* Reset the CCtx */
-        ZSTD_CCtx_reset(cctx);
-        CHECK_Z(ZSTD_CCtx_getParameter(cctx, ZSTD_p_compressionLevel, &value));
+        ZSTD_CCtx_reset(cctx, ZSTD_reset_session_only);
+        CHECK_Z(ZSTD_CCtx_getParameter(cctx, ZSTD_c_compressionLevel, &value));
         CHECK_EQ(value, 7);
-        CHECK_Z(ZSTD_CCtx_getParameter(cctx, ZSTD_p_hashLog, &value));
+        CHECK_Z(ZSTD_CCtx_getParameter(cctx, ZSTD_c_hashLog, &value));
         CHECK_EQ(value, ZSTD_HASHLOG_MIN);
         /* Reset the parameters */
-        ZSTD_CCtx_resetParameters(cctx);
-        CHECK_Z(ZSTD_CCtx_getParameter(cctx, ZSTD_p_compressionLevel, &value));
+        ZSTD_CCtx_reset(cctx, ZSTD_reset_parameters);
+        CHECK_Z(ZSTD_CCtx_getParameter(cctx, ZSTD_c_compressionLevel, &value));
         CHECK_EQ(value, 3);
-        CHECK_Z(ZSTD_CCtx_getParameter(cctx, ZSTD_p_hashLog, &value));
+        CHECK_Z(ZSTD_CCtx_getParameter(cctx, ZSTD_c_hashLog, &value));
         CHECK_EQ(value, 0);
 
         ZSTD_freeCCtx(cctx);
@@ -502,8 +629,8 @@ static int basicUnitTests(U32 seed, double compressibility)
     /* this test is really too long, and should be made faster */
     DISPLAYLEVEL(3, "test%3d : overflow protection with large windowLog : ", testNb++);
     {   ZSTD_CCtx* const cctx = ZSTD_createCCtx();
-        ZSTD_parameters params = ZSTD_getParams(-9, ZSTD_CONTENTSIZE_UNKNOWN, 0);
-        size_t const nbCompressions = ((1U << 31) / CNBuffSize) + 1;   /* ensure U32 overflow protection is triggered */
+        ZSTD_parameters params = ZSTD_getParams(-999, ZSTD_CONTENTSIZE_UNKNOWN, 0);
+        size_t const nbCompressions = ((1U << 31) / CNBuffSize) + 2;   /* ensure U32 overflow protection is triggered */
         size_t cnb;
         assert(cctx != NULL);
         params.fParams.contentSizeFlag = 0;
@@ -537,7 +664,7 @@ static int basicUnitTests(U32 seed, double compressibility)
             {   U32 const maxNbAttempts = 1100;   /* nb of usages before triggering size down is handled within zstd_compress.c.
                                                    * currently defined as 128x, but could be adjusted in the future.
                                                    * make this test long enough so that it's not too much tied to the current definition within zstd_compress.c */
-                U32 u;
+                unsigned u;
                 for (u=0; u same size */
         }
-        DISPLAYLEVEL(3, "OK (%u bytes : %.2f%%)\n", (U32)cSize, (double)cSize/CNBuffSize*100);
+        DISPLAYLEVEL(3, "OK (%u bytes : %.2f%%)\n", (unsigned)cSize, (double)cSize/CNBuffSize*100);
 
         DISPLAYLEVEL(3, "test%3i : frame built with duplicated context should be decompressible : ", testNb++);
         CHECKPLUS(r, ZSTD_decompress_usingDict(dctx,
@@ -782,7 +922,7 @@ static int basicUnitTests(U32 seed, double compressibility)
         {   ZSTD_DDict* const ddict = ZSTD_createDDict(CNBuffer, dictSize);
             size_t const r = ZSTD_decompress_usingDDict(dctx, decodedBuffer, CNBuffSize, compressedBuffer, cSize, ddict);
             if (r != CNBuffSize - dictSize) goto _output_error;
-            DISPLAYLEVEL(3, "OK (size of DDict : %u) \n", (U32)ZSTD_sizeof_DDict(ddict));
+            DISPLAYLEVEL(3, "OK (size of DDict : %u) \n", (unsigned)ZSTD_sizeof_DDict(ddict));
             ZSTD_freeDDict(ddict);
         }
 
@@ -795,7 +935,7 @@ static int basicUnitTests(U32 seed, double compressibility)
                 if (r != CNBuffSize - dictSize) goto _output_error;
             }
             free(ddictBuffer);
-            DISPLAYLEVEL(3, "OK (size of static DDict : %u) \n", (U32)ddictBufferSize);
+            DISPLAYLEVEL(3, "OK (size of static DDict : %u) \n", (unsigned)ddictBufferSize);
         }
 
         DISPLAYLEVEL(3, "test%3i : check content size on duplicated context : ", testNb++);
@@ -817,6 +957,7 @@ static int basicUnitTests(U32 seed, double compressibility)
 
         ZSTD_freeCCtx(ctxOrig);
         ZSTD_freeCCtx(ctxDuplicated);
+        ZSTD_freeDCtx(dctx);
     }
 
     /* Dictionary and dictBuilder tests */
@@ -843,7 +984,7 @@ static int basicUnitTests(U32 seed, double compressibility)
         {   size_t const sDictSize = ZDICT_trainFromBuffer(dictBuffer, dictBufferCapacity,
                                          decodedBuffer, samplesSizes, nbSamples);
             if (ZDICT_isError(sDictSize)) goto _output_error;
-            DISPLAYLEVEL(3, "OK, created dictionary of size %u \n", (U32)sDictSize);
+            DISPLAYLEVEL(3, "OK, created dictionary of size %u \n", (unsigned)sDictSize);
         }
 
         DISPLAYLEVEL(3, "test%3i : dictBuilder : ", testNb++);
@@ -851,19 +992,47 @@ static int basicUnitTests(U32 seed, double compressibility)
         dictSize = ZDICT_trainFromBuffer(dictBuffer, dictBufferCapacity,
                                          CNBuffer, samplesSizes, nbSamples);
         if (ZDICT_isError(dictSize)) goto _output_error;
-        DISPLAYLEVEL(3, "OK, created dictionary of size %u \n", (U32)dictSize);
+        DISPLAYLEVEL(3, "OK, created dictionary of size %u \n", (unsigned)dictSize);
+
+        DISPLAYLEVEL(3, "test%3i : Multithreaded COVER dictBuilder : ", testNb++);
+        { U32 u; for (u=0; u %u bytes)\n", (U32)inputSize, (U32)cSize);
+            DISPLAYLEVEL(3, "OK (compress : %u -> %u bytes)\n", (unsigned)inputSize, (unsigned)cSize);
             ZSTD_freeCCtx(cctx);
         }
 
         {   ZSTD_CCtx* cctx = ZSTD_createCCtx();
             DISPLAYLEVEL(3, "test%3i : parameters disordered : ", testNb++);
-            CHECK( ZSTD_CCtx_setParameter(cctx, ZSTD_p_windowLog, 18) );
-            CHECK( ZSTD_CCtx_setParameter(cctx, ZSTD_p_enableLongDistanceMatching, 1) );
-            CHECK( ZSTD_CCtx_setParameter(cctx, ZSTD_p_compressionLevel, 2) );
-            {   ZSTD_inBuffer in = { CNBuffer, inputSize, 0 };
-                ZSTD_outBuffer out = { compressedBuffer, ZSTD_compressBound(inputSize), 0 };
-                size_t const result = ZSTD_compress_generic(cctx, &out, &in, ZSTD_e_end);
-                if (result != 0) goto _output_error;
-                if (in.pos != in.size) goto _output_error;
-                if (out.pos != cSize) goto _output_error;   /* must result in same compressed result, hence same size */
-                if (XXH64(out.dst, out.pos, 0) != xxh64) goto _output_error;  /* must result in exactly same content, hence same hash */
-                DISPLAYLEVEL(3, "OK (compress : %u -> %u bytes)\n", (U32)inputSize, (U32)out.pos);
+            CHECK( ZSTD_CCtx_setParameter(cctx, ZSTD_c_windowLog, 18) );
+            CHECK( ZSTD_CCtx_setParameter(cctx, ZSTD_c_enableLongDistanceMatching, 1) );
+            CHECK( ZSTD_CCtx_setParameter(cctx, ZSTD_c_compressionLevel, 2) );
+            {   size_t const result = ZSTD_compress2(cctx,
+                                compressedBuffer, ZSTD_compressBound(inputSize),
+                                CNBuffer, inputSize);
+                CHECK(result);
+                if (result != cSize) goto _output_error;   /* must result in same compressed result, hence same size */
+                if (XXH64(compressedBuffer, result, 0) != xxh64) goto _output_error;  /* must result in exactly same content, hence same hash */
+                DISPLAYLEVEL(3, "OK (compress : %u -> %u bytes)\n", (unsigned)inputSize, (unsigned)result);
             }
             ZSTD_freeCCtx(cctx);
         }
     }
 
+    /* advanced parameters for decompression */
+    {   ZSTD_DCtx* const dctx = ZSTD_createDCtx();
+        assert(dctx != NULL);
+
+        DISPLAYLEVEL(3, "test%3i : get dParameter bounds ", testNb++);
+        {   ZSTD_bounds const bounds = ZSTD_dParam_getBounds(ZSTD_d_windowLogMax);
+            CHECK(bounds.error);
+        }
+        DISPLAYLEVEL(3, "OK \n");
+
+        DISPLAYLEVEL(3, "test%3i : wrong dParameter : ", testNb++);
+        {   size_t const sr = ZSTD_DCtx_setParameter(dctx, (ZSTD_dParameter)999999, 0);
+            if (!ZSTD_isError(sr)) goto _output_error;
+        }
+        {   ZSTD_bounds const bounds = ZSTD_dParam_getBounds((ZSTD_dParameter)999998);
+            if (!ZSTD_isError(bounds.error)) goto _output_error;
+        }
+        DISPLAYLEVEL(3, "OK \n");
+
+        DISPLAYLEVEL(3, "test%3i : out of bound dParameter : ", testNb++);
+        {   size_t const sr = ZSTD_DCtx_setParameter(dctx, ZSTD_d_windowLogMax, 9999);
+            if (!ZSTD_isError(sr)) goto _output_error;
+        }
+        {   size_t const sr = ZSTD_DCtx_setParameter(dctx, ZSTD_d_format, (ZSTD_format_e)888);
+            if (!ZSTD_isError(sr)) goto _output_error;
+        }
+        DISPLAYLEVEL(3, "OK \n");
+
+        ZSTD_freeDCtx(dctx);
+    }
+
+
     /* custom formats tests */
     {   ZSTD_CCtx* const cctx = ZSTD_createCCtx();
+        ZSTD_DCtx* const dctx = ZSTD_createDCtx();
         size_t const inputSize = CNBuffSize / 2;   /* won't cause pb with small dict size */
+        assert(dctx != NULL); assert(cctx != NULL);
 
         /* basic block compression */
         DISPLAYLEVEL(3, "test%3i : magic-less format test : ", testNb++);
-        CHECK( ZSTD_CCtx_setParameter(cctx, ZSTD_p_format, ZSTD_f_zstd1_magicless) );
+        CHECK( ZSTD_CCtx_setParameter(cctx, ZSTD_c_format, ZSTD_f_zstd1_magicless) );
         {   ZSTD_inBuffer in = { CNBuffer, inputSize, 0 };
             ZSTD_outBuffer out = { compressedBuffer, ZSTD_compressBound(inputSize), 0 };
-            size_t const result = ZSTD_compress_generic(cctx, &out, &in, ZSTD_e_end);
+            size_t const result = ZSTD_compressStream2(cctx, &out, &in, ZSTD_e_end);
             if (result != 0) goto _output_error;
             if (in.pos != in.size) goto _output_error;
             cSize = out.pos;
         }
-        DISPLAYLEVEL(3, "OK (compress : %u -> %u bytes)\n", (U32)inputSize, (U32)cSize);
+        DISPLAYLEVEL(3, "OK (compress : %u -> %u bytes)\n", (unsigned)inputSize, (unsigned)cSize);
 
         DISPLAYLEVEL(3, "test%3i : decompress normally (should fail) : ", testNb++);
         {   size_t const decodeResult = ZSTD_decompressDCtx(dctx, decodedBuffer, CNBuffSize, compressedBuffer, cSize);
@@ -1255,29 +1475,38 @@ static int basicUnitTests(U32 seed, double compressibility)
         }
 
         DISPLAYLEVEL(3, "test%3i : decompress of magic-less frame : ", testNb++);
-        ZSTD_DCtx_reset(dctx);
-        CHECK( ZSTD_DCtx_setFormat(dctx, ZSTD_f_zstd1_magicless) );
+        ZSTD_DCtx_reset(dctx, ZSTD_reset_session_and_parameters);
+        CHECK( ZSTD_DCtx_setParameter(dctx, ZSTD_d_format, ZSTD_f_zstd1_magicless) );
         {   ZSTD_frameHeader zfh;
             size_t const zfhrt = ZSTD_getFrameHeader_advanced(&zfh, compressedBuffer, cSize, ZSTD_f_zstd1_magicless);
             if (zfhrt != 0) goto _output_error;
         }
+        /* one shot */
+        {   size_t const result = ZSTD_decompressDCtx(dctx, decodedBuffer, CNBuffSize, compressedBuffer, cSize);
+            if (result != inputSize) goto _output_error;
+            DISPLAYLEVEL(3, "one-shot OK, ");
+        }
+        /* streaming */
         {   ZSTD_inBuffer in = { compressedBuffer, cSize, 0 };
             ZSTD_outBuffer out = { decodedBuffer, CNBuffSize, 0 };
-            size_t const result = ZSTD_decompress_generic(dctx, &out, &in);
+            size_t const result = ZSTD_decompressStream(dctx, &out, &in);
             if (result != 0) goto _output_error;
             if (in.pos != in.size) goto _output_error;
             if (out.pos != inputSize) goto _output_error;
-            DISPLAYLEVEL(3, "OK : regenerated %u bytes \n", (U32)out.pos);
+            DISPLAYLEVEL(3, "streaming OK : regenerated %u bytes \n", (unsigned)out.pos);
         }
 
         ZSTD_freeCCtx(cctx);
+        ZSTD_freeDCtx(dctx);
     }
 
     /* block API tests */
     {   ZSTD_CCtx* const cctx = ZSTD_createCCtx();
+        ZSTD_DCtx* const dctx = ZSTD_createDCtx();
         static const size_t dictSize = 65 KB;
         static const size_t blockSize = 100 KB;   /* won't cause pb with small dict size */
         size_t cSize2;
+        assert(cctx != NULL); assert(dctx != NULL);
 
         /* basic block compression */
         DISPLAYLEVEL(3, "test%3i : Block compression test : ", testNb++);
@@ -1295,13 +1524,14 @@ static int basicUnitTests(U32 seed, double compressibility)
 
         /* very long stream of block compression */
         DISPLAYLEVEL(3, "test%3i : Huge block streaming compression test : ", testNb++);
-        CHECK( ZSTD_compressBegin(cctx, -99) );  /* we just want to quickly overflow internal U32 index */
+        CHECK( ZSTD_compressBegin(cctx, -199) );  /* we just want to quickly overflow internal U32 index */
         CHECK( ZSTD_getBlockSize(cctx) >= blockSize);
         {   U64 const toCompress = 5000000000ULL;   /* > 4 GB */
             U64 compressed = 0;
             while (compressed < toCompress) {
                 size_t const blockCSize = ZSTD_compressBlock(cctx, compressedBuffer, ZSTD_compressBound(blockSize), CNBuffer, blockSize);
-                if (ZSTD_isError(cSize)) goto _output_error;
+                assert(blockCSize != 0);
+                if (ZSTD_isError(blockCSize)) goto _output_error;
                 compressed += blockCSize;
             }
         }
@@ -1339,8 +1569,8 @@ static int basicUnitTests(U32 seed, double compressibility)
         DISPLAYLEVEL(3, "OK \n");
 
         ZSTD_freeCCtx(cctx);
+        ZSTD_freeDCtx(dctx);
     }
-    ZSTD_freeDCtx(dctx);
 
     /* long rle test */
     {   size_t sampleSize = 0;
@@ -1363,7 +1593,7 @@ static int basicUnitTests(U32 seed, double compressibility)
     memset(CNBuffer, 0, ZEROESLENGTH);
     { CHECK_V(r, ZSTD_compress(compressedBuffer, ZSTD_compressBound(ZEROESLENGTH), CNBuffer, ZEROESLENGTH, 1) );
       cSize = r; }
-    DISPLAYLEVEL(3, "OK (%u bytes : %.2f%%)\n", (U32)cSize, (double)cSize/ZEROESLENGTH*100);
+    DISPLAYLEVEL(3, "OK (%u bytes : %.2f%%)\n", (unsigned)cSize, (double)cSize/ZEROESLENGTH*100);
 
     DISPLAYLEVEL(3, "test%3i : decompress %u zeroes : ", testNb++, ZEROESLENGTH);
     { CHECK_V(r, ZSTD_decompress(decodedBuffer, ZEROESLENGTH, compressedBuffer, cSize) );
@@ -1417,7 +1647,7 @@ static int basicUnitTests(U32 seed, double compressibility)
     { CHECK_V(r, ZSTD_compress(compressedBuffer, ZSTD_compressBound(_3BYTESTESTLENGTH),
                                  CNBuffer, _3BYTESTESTLENGTH, 19) );
       cSize = r; }
-    DISPLAYLEVEL(3, "OK (%u bytes : %.2f%%)\n", (U32)cSize, (double)cSize/_3BYTESTESTLENGTH*100);
+    DISPLAYLEVEL(3, "OK (%u bytes : %.2f%%)\n", (unsigned)cSize, (double)cSize/_3BYTESTESTLENGTH*100);
 
     DISPLAYLEVEL(3, "test%3i : decompress lots 3-bytes sequence : ", testNb++);
     { CHECK_V(r, ZSTD_decompress(decodedBuffer, _3BYTESTESTLENGTH, compressedBuffer, cSize) );
@@ -1586,7 +1816,7 @@ static size_t FUZ_randomLength(U32* seed, U32 maxLog)
     if (cond) {                                               \
         DISPLAY("Error => ");                                 \
         DISPLAY(__VA_ARGS__);                                 \
-        DISPLAY(" (seed %u, test nb %u)  \n", seed, testNb);  \
+        DISPLAY(" (seed %u, test nb %u)  \n", (unsigned)seed, testNb);  \
         goto _output_error;                                   \
 }   }
 
@@ -1596,12 +1826,12 @@ static size_t FUZ_randomLength(U32* seed, U32 maxLog)
     if (ZSTD_isError(err)) {                                  \
         DISPLAY("Error => %s : %s ",                          \
                 #f, ZSTD_getErrorName(err));                  \
-        DISPLAY(" (seed %u, test nb %u)  \n", seed, testNb);  \
+        DISPLAY(" (seed %u, test nb %u)  \n", (unsigned)seed, testNb);  \
         goto _output_error;                                   \
 }   }
 
 
-static int fuzzerTests(U32 seed, U32 nbTests, unsigned startTest, U32 const maxDurationS, double compressibility, int bigTests)
+static int fuzzerTests(U32 seed, unsigned nbTests, unsigned startTest, U32 const maxDurationS, double compressibility, int bigTests)
 {
     static const U32 maxSrcLog = 23;
     static const U32 maxSampleLog = 22;
@@ -1616,7 +1846,7 @@ static int fuzzerTests(U32 seed, U32 nbTests, unsigned startTest, U32 const maxD
     ZSTD_CCtx* const ctx = ZSTD_createCCtx();
     ZSTD_DCtx* const dctx = ZSTD_createDCtx();
     U32 result = 0;
-    U32 testNb = 0;
+    unsigned testNb = 0;
     U32 coreSeed = seed;
     UTIL_time_t const startClock = UTIL_getTime();
     U64 const maxClockSpan = maxDurationS * SEC_TO_MICRO;
@@ -1698,15 +1928,17 @@ static int fuzzerTests(U32 seed, U32 nbTests, unsigned startTest, U32 const maxD
             CHECK(ZSTD_isError(cSize), "ZSTD_compressCCtx failed : %s", ZSTD_getErrorName(cSize));
 
             /* compression failure test : too small dest buffer */
-            if (cSize > 3) {
-                const size_t missing = (FUZ_rand(&lseed) % (cSize-2)) + 1;   /* no problem, as cSize > 4 (frameHeaderSizer) */
+            assert(cSize > 3);
+            {   const size_t missing = (FUZ_rand(&lseed) % (cSize-2)) + 1;
                 const size_t tooSmallSize = cSize - missing;
-                const U32 endMark = 0x4DC2B1A9;
-                memcpy(dstBuffer+tooSmallSize, &endMark, 4);
+                const unsigned endMark = 0x4DC2B1A9;
+                memcpy(dstBuffer+tooSmallSize, &endMark, sizeof(endMark));
+                DISPLAYLEVEL(5, "fuzzer t%u: compress into too small buffer of size %u (missing %u bytes) \n",
+                            testNb, (unsigned)tooSmallSize, (unsigned)missing);
                 { size_t const errorCode = ZSTD_compressCCtx(ctx, dstBuffer, tooSmallSize, sampleBuffer, sampleSize, cLevel);
-                  CHECK(!ZSTD_isError(errorCode), "ZSTD_compressCCtx should have failed ! (buffer too small : %u < %u)", (U32)tooSmallSize, (U32)cSize); }
-                { U32 endCheck; memcpy(&endCheck, dstBuffer+tooSmallSize, 4);
-                  CHECK(endCheck != endMark, "ZSTD_compressCCtx : dst buffer overflow"); }
+                  CHECK(!ZSTD_isError(errorCode), "ZSTD_compressCCtx should have failed ! (buffer too small : %u < %u)", (unsigned)tooSmallSize, (unsigned)cSize); }
+                { unsigned endCheck; memcpy(&endCheck, dstBuffer+tooSmallSize, sizeof(endCheck));
+                  CHECK(endCheck != endMark, "ZSTD_compressCCtx : dst buffer overflow  (check.%08X != %08X.mark)", endCheck, endMark); }
         }   }
 
         /* frame header decompression test */
@@ -1724,9 +1956,9 @@ static int fuzzerTests(U32 seed, U32 nbTests, unsigned startTest, U32 const maxD
         DISPLAYLEVEL(5, "fuzzer t%u: simple decompression test \n", testNb);
         {   size_t const margin = (FUZ_rand(&lseed) & 1) ? 0 : (FUZ_rand(&lseed) & 31) + 1;
             size_t const dSize = ZSTD_decompress(dstBuffer, sampleSize + margin, cBuffer, cSize);
-            CHECK(dSize != sampleSize, "ZSTD_decompress failed (%s) (srcSize : %u ; cSize : %u)", ZSTD_getErrorName(dSize), (U32)sampleSize, (U32)cSize);
+            CHECK(dSize != sampleSize, "ZSTD_decompress failed (%s) (srcSize : %u ; cSize : %u)", ZSTD_getErrorName(dSize), (unsigned)sampleSize, (unsigned)cSize);
             {   U64 const crcDest = XXH64(dstBuffer, sampleSize, 0);
-                CHECK(crcOrig != crcDest, "decompression result corrupted (pos %u / %u)", (U32)findDiff(sampleBuffer, dstBuffer, sampleSize), (U32)sampleSize);
+                CHECK(crcOrig != crcDest, "decompression result corrupted (pos %u / %u)", (unsigned)findDiff(sampleBuffer, dstBuffer, sampleSize), (unsigned)sampleSize);
         }   }
 
         free(sampleBuffer);   /* no longer useful after this point */
@@ -1751,7 +1983,7 @@ static int fuzzerTests(U32 seed, U32 nbTests, unsigned startTest, U32 const maxD
             static const BYTE token = 0xA9;
             dstBuffer[tooSmallSize] = token;
             { size_t const errorCode = ZSTD_decompress(dstBuffer, tooSmallSize, cBuffer, cSize);
-              CHECK(!ZSTD_isError(errorCode), "ZSTD_decompress should have failed : %u > %u (dst buffer too small)", (U32)errorCode, (U32)tooSmallSize); }
+              CHECK(!ZSTD_isError(errorCode), "ZSTD_decompress should have failed : %u > %u (dst buffer too small)", (unsigned)errorCode, (unsigned)tooSmallSize); }
             CHECK(dstBuffer[tooSmallSize] != token, "ZSTD_decompress : dst buffer overflow");
         }
 
@@ -1786,7 +2018,7 @@ static int fuzzerTests(U32 seed, U32 nbTests, unsigned startTest, U32 const maxD
                 {   size_t const decompressResult = ZSTD_decompress(dstBuffer, sampleSize, cBuffer, cSize);
                     /* result *may* be an unlikely success, but even then, it must strictly respect dst buffer boundaries */
                     CHECK((!ZSTD_isError(decompressResult)) && (decompressResult>sampleSize),
-                          "ZSTD_decompress on noisy src : result is too large : %u > %u (dst buffer)", (U32)decompressResult, (U32)sampleSize);
+                          "ZSTD_decompress on noisy src : result is too large : %u > %u (dst buffer)", (unsigned)decompressResult, (unsigned)sampleSize);
                 }
                 {   U32 endCheck; memcpy(&endCheck, dstBuffer+sampleSize, 4);
                     CHECK(endMark!=endCheck, "ZSTD_decompress on noisy src : dst buffer overflow");
@@ -1807,7 +2039,7 @@ static int fuzzerTests(U32 seed, U32 nbTests, unsigned startTest, U32 const maxD
             dict = srcBuffer + (FUZ_rand(&lseed) % (srcBufferSize - dictSize));
 
             DISPLAYLEVEL(6, "fuzzer t%u: Compressing up to <=%u bytes at level %i with dictionary size %u \n",
-                            testNb, (U32)maxTestSize, cLevel, (U32)dictSize);
+                            testNb, (unsigned)maxTestSize, cLevel, (unsigned)dictSize);
 
             if (FUZ_rand(&lseed) & 0xF) {
                 CHECK_Z ( ZSTD_compressBegin_usingDict(refCtx, dict, dictSize, cLevel) );
@@ -1853,13 +2085,13 @@ static int fuzzerTests(U32 seed, U32 nbTests, unsigned startTest, U32 const maxD
         DISPLAYLEVEL(5, "fuzzer t%u: Bufferless streaming decompression test \n", testNb);
         /* ensure memory requirement is good enough (should always be true) */
         {   ZSTD_frameHeader zfh;
-            CHECK( ZSTD_getFrameHeader(&zfh, cBuffer, ZSTD_frameHeaderSize_max),
+            CHECK( ZSTD_getFrameHeader(&zfh, cBuffer, ZSTD_FRAMEHEADERSIZE_MAX),
                   "ZSTD_getFrameHeader(): error retrieving frame information");
             {   size_t const roundBuffSize = ZSTD_decodingBufferSize_min(zfh.windowSize, zfh.frameContentSize);
                 CHECK_Z(roundBuffSize);
                 CHECK((roundBuffSize > totalTestSize) && (zfh.frameContentSize!=ZSTD_CONTENTSIZE_UNKNOWN),
                       "ZSTD_decodingBufferSize_min() requires more memory (%u) than necessary (%u)",
-                      (U32)roundBuffSize, (U32)totalTestSize );
+                      (unsigned)roundBuffSize, (unsigned)totalTestSize );
         }   }
         if (dictSize<8) dictSize=0, dict=NULL;   /* disable dictionary */
         CHECK_Z( ZSTD_decompressBegin_usingDict(dctx, dict, dictSize) );
@@ -1877,7 +2109,7 @@ static int fuzzerTests(U32 seed, U32 nbTests, unsigned startTest, U32 const maxD
         CHECK (totalCSize != cSize, "compressed data should be fully read")
         {   U64 const crcDest = XXH64(dstBuffer, totalTestSize, 0);
             CHECK(crcOrig != crcDest, "streaming decompressed data corrupted (pos %u / %u)",
-                (U32)findDiff(mirrorBuffer, dstBuffer, totalTestSize), (U32)totalTestSize);
+                (unsigned)findDiff(mirrorBuffer, dstBuffer, totalTestSize), (unsigned)totalTestSize);
         }
     }   /* for ( ; (testNb <= nbTests) */
     DISPLAY("\r%u fuzzer tests completed   \n", testNb-1);
@@ -1911,10 +2143,10 @@ static int FUZ_usage(const char* programName)
     DISPLAY( "      %s [args]\n", programName);
     DISPLAY( "\n");
     DISPLAY( "Arguments :\n");
-    DISPLAY( " -i#    : Nb of tests (default:%u) \n", nbTestsDefault);
+    DISPLAY( " -i#    : Nb of tests (default:%i) \n", nbTestsDefault);
     DISPLAY( " -s#    : Select seed (default:prompt user)\n");
     DISPLAY( " -t#    : Select starting test number (default:0)\n");
-    DISPLAY( " -P#    : Select compressibility in %% (default:%u%%)\n", FUZ_compressibility_default);
+    DISPLAY( " -P#    : Select compressibility in %% (default:%i%%)\n", FUZ_compressibility_default);
     DISPLAY( " -v     : verbose\n");
     DISPLAY( " -p     : pause at the end\n");
     DISPLAY( " -h     : display help and exit\n");
@@ -1961,7 +2193,7 @@ int main(int argc, const char** argv)
     int argNb;
     int nbTests = nbTestsDefault;
     int testNb = 0;
-    U32 proba = FUZ_compressibility_default;
+    int proba = FUZ_compressibility_default;
     int result = 0;
     U32 mainPause = 0;
     U32 maxDuration = 0;
@@ -2048,8 +2280,8 @@ int main(int argc, const char** argv)
         seed = h % 10000;
     }
 
-    DISPLAY("Seed = %u\n", seed);
-    if (proba!=FUZ_compressibility_default) DISPLAY("Compressibility : %u%%\n", proba);
+    DISPLAY("Seed = %u\n", (unsigned)seed);
+    if (proba!=FUZ_compressibility_default) DISPLAY("Compressibility : %i%%\n", proba);
 
     if (memTestsOnly) {
         g_displayLevel = MAX(3, g_displayLevel);
diff --git a/tests/libzstd_partial_builds.sh b/tests/libzstd_partial_builds.sh
index 34d8ea55231e..b1c1e3b1a7e3 100755
--- a/tests/libzstd_partial_builds.sh
+++ b/tests/libzstd_partial_builds.sh
@@ -1,36 +1,89 @@
 #!/bin/sh -e
 
+DIR="$( cd "$( dirname "${BASH_SOURCE[0]}" )" && pwd )"
+
+ECHO=echo
+RM="rm -f"
+GREP="grep"
+INTOVOID="/dev/null"
+
 die() {
     $ECHO "$@" 1>&2
     exit 1
 }
 
-DIR="$( cd "$( dirname "${BASH_SOURCE[0]}" )" && pwd )"
+isPresent() {
+    $GREP $@ tmplog || die "$@" "should be present"
+}
 
-INTOVOID="/dev/null"
-case "$OS" in
-  Windows*)
-    INTOVOID="NUL"
-    ;;
-esac
+mustBeAbsent() {
+    $GREP $@ tmplog && die "$@ should not be there !!"
+    $ECHO "$@ correctly not present"  # for some reason, this $ECHO must exist, otherwise mustBeAbsent() always fails (??)
+}
 
+# default compilation : all features enabled
+make clean > /dev/null
+$ECHO "testing default library compilation"
+CFLAGS= make -C $DIR/../lib libzstd.a > $INTOVOID
+nm $DIR/../lib/libzstd.a | $GREP "\.o" > tmplog
+isPresent "zstd_compress.o"
+isPresent "zstd_decompress.o"
+isPresent "zdict.o"
+isPresent "zstd_v07.o"
+isPresent "zbuff_compress.o"
+$RM $DIR/../lib/libzstd.a tmplog
+
+# compression disabled => also disable zdict and zbuff
+$ECHO "testing with compression disabled"
 ZSTD_LIB_COMPRESSION=0 CFLAGS= make -C $DIR/../lib libzstd.a > $INTOVOID
-nm $DIR/../lib/libzstd.a | grep ".*\.o:" > tmplog
-! grep -q "zstd_compress" tmplog && grep -q "zstd_decompress" tmplog && ! grep -q "dict" tmplog && grep -q "zstd_v" tmplog && ! grep -q "zbuff" tmplog && make clean && rm -f tmplog || die "Compression macro failed"
-
+nm $DIR/../lib/libzstd.a | $GREP "\.o" > tmplog
+mustBeAbsent "zstd_compress.o"
+isPresent "zstd_decompress.o"
+mustBeAbsent "zdict.o"
+isPresent "zstd_v07.o"
+mustBeAbsent "zbuff_compress.o"
+$RM $DIR/../lib/libzstd.a tmplog
 
+# decompression disabled => also disable legacy and zbuff
+$ECHO "testing with decompression disabled"
 ZSTD_LIB_DECOMPRESSION=0 CFLAGS= make -C $DIR/../lib libzstd.a > $INTOVOID
-nm $DIR/../lib/libzstd.a | grep ".*\.o:" > tmplog
-grep -q "zstd_compress" tmplog && ! grep -q "zstd_decompress" tmplog && grep -q "dict" tmplog && ! grep -q "zstd_v" tmplog && ! grep -q "zbuff" tmplog && make clean && rm -f tmplog || die "Decompression macro failed"
+nm $DIR/../lib/libzstd.a | $GREP "\.o" > tmplog
+isPresent "zstd_compress.o"
+mustBeAbsent "zstd_decompress.o"
+isPresent "zdict.o"
+mustBeAbsent "zstd_v07.o"
+mustBeAbsent "zbuff_compress.o"
+$RM $DIR/../lib/libzstd.a tmplog
 
+# deprecated function disabled => only remove zbuff
+$ECHO "testing with deprecated functions disabled"
 ZSTD_LIB_DEPRECATED=0 CFLAGS= make -C $DIR/../lib libzstd.a > $INTOVOID
-nm $DIR/../lib/libzstd.a | grep ".*\.o:" > tmplog
-grep -q "zstd_compress" tmplog && grep -q "zstd_decompress" tmplog && grep -q "dict" tmplog && grep -q "zstd_v" tmplog && ! grep -q "zbuff" tmplog && make clean && rm -f tmplog || die "Deprecated macro failed"
+nm $DIR/../lib/libzstd.a | $GREP "\.o" > tmplog
+isPresent "zstd_compress.o"
+isPresent "zstd_decompress.o"
+isPresent "zdict.o"
+isPresent "zstd_v07.o"
+mustBeAbsent "zbuff_compress.o"
+$RM $DIR/../lib/libzstd.a tmplog
 
+# dictionary builder disabled => only remove zdict
+$ECHO "testing with dictionary builder disabled"
 ZSTD_LIB_DICTBUILDER=0 CFLAGS= make -C $DIR/../lib libzstd.a > $INTOVOID
-nm $DIR/../lib/libzstd.a | grep ".*\.o:" > tmplog
-grep -q "zstd_compress" tmplog && grep -q "zstd_decompress" tmplog && ! grep -q "dict" tmplog && grep -q "zstd_v" tmplog && grep -q "zbuff" tmplog && make clean && rm -f tmplog || die "Dictbuilder macro failed"
+nm $DIR/../lib/libzstd.a | $GREP "\.o" > tmplog
+isPresent "zstd_compress.o"
+isPresent "zstd_decompress.o"
+mustBeAbsent "zdict.o"
+isPresent "zstd_v07.o"
+isPresent "zbuff_compress.o"
+$RM $DIR/../lib/libzstd.a tmplog
 
+# both decompression and dictionary builder disabled => only compression remains
+$ECHO "testing with both decompression and dictionary builder disabled (only compression remains)"
 ZSTD_LIB_DECOMPRESSION=0 ZSTD_LIB_DICTBUILDER=0 CFLAGS= make -C $DIR/../lib libzstd.a > $INTOVOID
-nm $DIR/../lib/libzstd.a | grep ".*\.o:" > tmplog
-grep -q "zstd_compress" tmplog && ! grep -q "zstd_decompress" tmplog && ! grep -q "dict" tmplog && ! grep -q "zstd_v" tmplog && ! grep -q "zbuff" tmplog && make clean && rm -f tmplog || die "Multi-macro failed"
\ No newline at end of file
+nm $DIR/../lib/libzstd.a | $GREP "\.o" > tmplog
+isPresent "zstd_compress.o"
+mustBeAbsent "zstd_decompress.o"
+mustBeAbsent "zdict.o"
+mustBeAbsent "zstd_v07.o"
+mustBeAbsent "zbuff_compress.o"
+$RM $DIR/../lib/libzstd.a tmplog
diff --git a/tests/longmatch.c b/tests/longmatch.c
index 1271e9ab1039..b673baa60140 100644
--- a/tests/longmatch.c
+++ b/tests/longmatch.c
@@ -50,7 +50,7 @@ int main(int argc, const char** argv)
   params.cParams.chainLog = 13;
   params.cParams.hashLog = 14;
   params.cParams.searchLog = 1;
-  params.cParams.searchLength = 7;
+  params.cParams.minMatch = 7;
   params.cParams.targetLength = 16;
   params.cParams.strategy = ZSTD_fast;
   windowLog = params.cParams.windowLog;
diff --git a/tests/paramgrill.c b/tests/paramgrill.c
index 7a4be854a46b..415551b9578d 100644
--- a/tests/paramgrill.c
+++ b/tests/paramgrill.c
@@ -12,7 +12,7 @@
 /*-************************************
 *  Dependencies
 **************************************/
-#include "util.h"      /* Compiler options, UTIL_GetFileSize */
+#include "util.h"      /* Ensure platform.h is compiled first; also : compiler options, UTIL_GetFileSize */
 #include     /* malloc */
 #include      /* fprintf, fopen, ftello64 */
 #include     /* strcmp */
@@ -24,8 +24,8 @@
 #include "zstd.h"
 #include "datagen.h"
 #include "xxhash.h"
-#include "util.h"
-#include "bench.h"
+#include "benchfn.h"
+#include "benchzstd.h"
 #include "zstd_errors.h"
 #include "zstd_internal.h"     /* should not be needed */
 
@@ -74,20 +74,21 @@ static const int g_maxNbVariations = 64;
 #define CLOG_RANGE (ZSTD_CHAINLOG_MAX - ZSTD_CHAINLOG_MIN + 1)
 #define HLOG_RANGE (ZSTD_HASHLOG_MAX - ZSTD_HASHLOG_MIN + 1)
 #define SLOG_RANGE (ZSTD_SEARCHLOG_MAX - ZSTD_SEARCHLOG_MIN + 1)
-#define SLEN_RANGE (ZSTD_SEARCHLENGTH_MAX - ZSTD_SEARCHLENGTH_MIN + 1)
-#define TLEN_RANGE 17
-#define STRT_RANGE (ZSTD_btultra - ZSTD_fast + 1)
-#define FADT_RANGE 3
+#define MML_RANGE  (ZSTD_MINMATCH_MAX - ZSTD_MINMATCH_MIN + 1)
+#define TLEN_RANGE  17
+#define STRT_RANGE (ZSTD_STRATEGY_MAX - ZSTD_STRATEGY_MIN + 1)
+#define FADT_RANGE   3
 
-#define CHECKTIME(r) { if(BMK_timeSpan(g_time) > g_timeLimit_s) { DEBUGOUTPUT("Time Limit Reached\n"); return r; } }
-#define CHECKTIMEGT(ret, val, _gototag) {if(BMK_timeSpan(g_time) > g_timeLimit_s) { DEBUGOUTPUT("Time Limit Reached\n"); ret = val; goto _gototag; } }
+#define CHECKTIME(r) { if(BMK_timeSpan_s(g_time) > g_timeLimit_s) { DEBUGOUTPUT("Time Limit Reached\n"); return r; } }
+#define CHECKTIMEGT(ret, val, _gototag) { if(BMK_timeSpan_s(g_time) > g_timeLimit_s) { DEBUGOUTPUT("Time Limit Reached\n"); ret = val; goto _gototag; } }
 
 #define PARAM_UNSET ((U32)-2) /* can't be -1 b/c fadt uses -1 */
 
-static const char* g_stratName[ZSTD_btultra+1] = {
+static const char* g_stratName[ZSTD_STRATEGY_MAX+1] = {
                 "(none)       ", "ZSTD_fast    ", "ZSTD_dfast   ",
                 "ZSTD_greedy  ", "ZSTD_lazy    ", "ZSTD_lazy2   ",
-                "ZSTD_btlazy2 ", "ZSTD_btopt   ", "ZSTD_btultra "};
+                "ZSTD_btlazy2 ", "ZSTD_btopt   ", "ZSTD_btultra ",
+                "ZSTD_btultra2"};
 
 static const U32 tlen_table[TLEN_RANGE] = { 0, 1, 2, 4, 6, 8, 12, 16, 24, 32, 48, 64, 96, 128, 256, 512, 999 };
 
@@ -102,7 +103,7 @@ typedef enum {
     clog_ind = 1,
     hlog_ind = 2,
     slog_ind = 3,
-    slen_ind = 4,
+    mml_ind  = 4,
     tlen_ind = 5,
     strt_ind = 6,
     fadt_ind = 7, /* forceAttachDict */
@@ -113,56 +114,66 @@ typedef struct {
     U32 vals[NUM_PARAMS];
 } paramValues_t;
 
-/* maximum value of parameters */
-static const U32 mintable[NUM_PARAMS] =
-        { ZSTD_WINDOWLOG_MIN, ZSTD_CHAINLOG_MIN, ZSTD_HASHLOG_MIN, ZSTD_SEARCHLOG_MIN, ZSTD_SEARCHLENGTH_MIN, ZSTD_TARGETLENGTH_MIN, ZSTD_fast, FADT_MIN };
-
 /* minimum value of parameters */
+static const U32 mintable[NUM_PARAMS] =
+        { ZSTD_WINDOWLOG_MIN, ZSTD_CHAINLOG_MIN, ZSTD_HASHLOG_MIN, ZSTD_SEARCHLOG_MIN, ZSTD_MINMATCH_MIN, ZSTD_TARGETLENGTH_MIN, ZSTD_STRATEGY_MIN, FADT_MIN };
+
+/* maximum value of parameters */
 static const U32 maxtable[NUM_PARAMS] =
-        { ZSTD_WINDOWLOG_MAX, ZSTD_CHAINLOG_MAX, ZSTD_HASHLOG_MAX, ZSTD_SEARCHLOG_MAX, ZSTD_SEARCHLENGTH_MAX, ZSTD_TARGETLENGTH_MAX, ZSTD_btultra, FADT_MAX };
+        { ZSTD_WINDOWLOG_MAX, ZSTD_CHAINLOG_MAX, ZSTD_HASHLOG_MAX, ZSTD_SEARCHLOG_MAX, ZSTD_MINMATCH_MAX, ZSTD_TARGETLENGTH_MAX, ZSTD_STRATEGY_MAX, FADT_MAX };
 
 /* # of values parameters can take on */
 static const U32 rangetable[NUM_PARAMS] =
-        { WLOG_RANGE, CLOG_RANGE, HLOG_RANGE, SLOG_RANGE, SLEN_RANGE, TLEN_RANGE, STRT_RANGE, FADT_RANGE };
+        { WLOG_RANGE, CLOG_RANGE, HLOG_RANGE, SLOG_RANGE, MML_RANGE, TLEN_RANGE, STRT_RANGE, FADT_RANGE };
 
 /* ZSTD_cctxSetParameter() index to set */
 static const ZSTD_cParameter cctxSetParamTable[NUM_PARAMS] =
-        { ZSTD_p_windowLog, ZSTD_p_chainLog, ZSTD_p_hashLog, ZSTD_p_searchLog, ZSTD_p_minMatch, ZSTD_p_targetLength, ZSTD_p_compressionStrategy, ZSTD_p_forceAttachDict };
+        { ZSTD_c_windowLog, ZSTD_c_chainLog, ZSTD_c_hashLog, ZSTD_c_searchLog, ZSTD_c_minMatch, ZSTD_c_targetLength, ZSTD_c_strategy, ZSTD_c_forceAttachDict };
 
 /* names of parameters */
 static const char* g_paramNames[NUM_PARAMS] =
-        { "windowLog", "chainLog", "hashLog","searchLog", "searchLength", "targetLength", "strategy", "forceAttachDict" };
+        { "windowLog", "chainLog", "hashLog","searchLog", "minMatch", "targetLength", "strategy", "forceAttachDict" };
 
 /* shortened names of parameters */
 static const char* g_shortParamNames[NUM_PARAMS] =
-        { "wlog", "clog", "hlog","slog", "slen", "tlen", "strt", "fadt" };
+        { "wlog", "clog", "hlog", "slog", "mml", "tlen", "strat", "fadt" };
 
 /* maps value from { 0 to rangetable[param] - 1 } to valid paramvalues */
-static U32 rangeMap(varInds_t param, int ind) {
+static U32 rangeMap(varInds_t param, int ind)
+{
     ind = MAX(MIN(ind, (int)rangetable[param] - 1), 0);
     switch(param) {
-        case tlen_ind:
-            return tlen_table[ind];
-        case fadt_ind: /* 0, 1, 2 -> -1, 0, 1 */
-            return ind - 1;
         case wlog_ind: /* using default: triggers -Wswitch-enum */
         case clog_ind:
         case hlog_ind:
         case slog_ind:
-        case slen_ind:
+        case mml_ind:
         case strt_ind:
             return mintable[param] + ind;
+        case tlen_ind:
+            return tlen_table[ind];
+        case fadt_ind: /* 0, 1, 2 -> -1, 0, 1 */
+            return ind - 1;
         case NUM_PARAMS:
-            DISPLAY("Error, not a valid param\n ");
-            return (U32)-1;
+        default:;
     }
-    return 0; /* should never happen, stop compiler warnings */
+    DISPLAY("Error, not a valid param\n ");
+    assert(0);
+    return (U32)-1;
 }
 
 /* inverse of rangeMap */
-static int invRangeMap(varInds_t param, U32 value) {
+static int invRangeMap(varInds_t param, U32 value)
+{
     value = MIN(MAX(mintable[param], value), maxtable[param]);
     switch(param) {
+        case wlog_ind:
+        case clog_ind:
+        case hlog_ind:
+        case slog_ind:
+        case mml_ind:
+        case strt_ind:
+            return value - mintable[param];
         case tlen_ind: /* bin search */
         {
             int lo = 0;
@@ -181,33 +192,49 @@ static int invRangeMap(varInds_t param, U32 value) {
         }
         case fadt_ind:
             return (int)value + 1;
-        case wlog_ind:
-        case clog_ind:
-        case hlog_ind:
-        case slog_ind:
-        case slen_ind:
-        case strt_ind:
-            return value - mintable[param];
         case NUM_PARAMS:
-            DISPLAY("Error, not a valid param\n ");
-            return -2;
+        default:;
     }
-    return 0; /* should never happen, stop compiler warnings */
+    DISPLAY("Error, not a valid param\n ");
+    assert(0);
+    return -2;
 }
 
 /* display of params */
-static void displayParamVal(FILE* f, varInds_t param, U32 value, int width) {
+static void displayParamVal(FILE* f, varInds_t param, unsigned value, int width)
+{
     switch(param) {
-        case fadt_ind: if(width) { fprintf(f, "%*d", width, (int)value); } else { fprintf(f, "%d", (int)value); } break;
-        case strt_ind: if(width) { fprintf(f, "%*s", width, g_stratName[value]); } else { fprintf(f, "%s", g_stratName[value]); } break;
         case wlog_ind:
         case clog_ind:
         case hlog_ind:
         case slog_ind:
-        case slen_ind:
-        case tlen_ind: if(width) { fprintf(f, "%*u", width, value); } else { fprintf(f, "%u", value); } break;
+        case mml_ind:
+        case tlen_ind:
+            if(width) {
+                fprintf(f, "%*u", width, value);
+            } else {
+                fprintf(f, "%u", value);
+            }
+            break;
+        case strt_ind:
+            if(width) {
+                fprintf(f, "%*s", width, g_stratName[value]);
+            } else {
+                fprintf(f, "%s", g_stratName[value]);
+            }
+            break;
+        case fadt_ind:   /* force attach dict */
+            if(width) {
+                fprintf(f, "%*d", width, (int)value);
+            } else {
+                fprintf(f, "%d", (int)value);
+            }
+            break;
         case NUM_PARAMS:
-            DISPLAY("Error, not a valid param\n "); break;
+        default:
+            DISPLAY("Error, not a valid param\n ");
+            assert(0);
+            break;
     }
 }
 
@@ -216,8 +243,6 @@ static void displayParamVal(FILE* f, varInds_t param, U32 value, int width) {
 *  Benchmark Parameters/Global Variables
 **************************************/
 
-typedef BYTE U8;
-
 /* General Utility */
 static U32 g_timeLimit_s = 99999;   /* about 27 hours */
 static UTIL_time_t g_time; /* to be used to compare solution finding speeds to compare to original */
@@ -226,7 +251,7 @@ static U32 g_rand = 1;
 
 /* Display */
 static int g_displayLevel = 3;
-static BYTE g_silenceParams[NUM_PARAMS];
+static BYTE g_silenceParams[NUM_PARAMS];   /* can selectively silence some params when displaying them */
 
 /* Mode Selection */
 static U32 g_singleRun = 0;
@@ -297,50 +322,54 @@ static paramValues_t sanitizeParams(paramValues_t params)
         params.vals[clog_ind] = 0, params.vals[slog_ind] = 0;
     if (params.vals[strt_ind] == ZSTD_dfast)
         params.vals[slog_ind] = 0;
-    if (params.vals[strt_ind] != ZSTD_btopt && params.vals[strt_ind] != ZSTD_btultra && params.vals[strt_ind] != ZSTD_fast)
+    if ( (params.vals[strt_ind] < ZSTD_btopt) && (params.vals[strt_ind] != ZSTD_fast) )
         params.vals[tlen_ind] = 0;
 
     return params;
 }
 
-static ZSTD_compressionParameters pvalsToCParams(paramValues_t p) {
+static ZSTD_compressionParameters pvalsToCParams(paramValues_t p)
+{
     ZSTD_compressionParameters c;
     memset(&c, 0, sizeof(ZSTD_compressionParameters));
     c.windowLog = p.vals[wlog_ind];
     c.chainLog = p.vals[clog_ind];
     c.hashLog = p.vals[hlog_ind];
     c.searchLog = p.vals[slog_ind];
-    c.searchLength = p.vals[slen_ind];
+    c.minMatch = p.vals[mml_ind];
     c.targetLength = p.vals[tlen_ind];
     c.strategy = p.vals[strt_ind];
     /* no forceAttachDict */
     return c;
 }
 
-static paramValues_t cParamsToPVals(ZSTD_compressionParameters c) {
+static paramValues_t cParamsToPVals(ZSTD_compressionParameters c)
+{
     paramValues_t p;
     varInds_t i;
     p.vals[wlog_ind] = c.windowLog;
     p.vals[clog_ind] = c.chainLog;
     p.vals[hlog_ind] = c.hashLog;
     p.vals[slog_ind] = c.searchLog;
-    p.vals[slen_ind] = c.searchLength;
+    p.vals[mml_ind]  = c.minMatch;
     p.vals[tlen_ind] = c.targetLength;
     p.vals[strt_ind] = c.strategy;
 
     /* set all other params to their minimum value */
-    for(i = strt_ind + 1; i < NUM_PARAMS; i++) {
+    for (i = strt_ind + 1; i < NUM_PARAMS; i++) {
         p.vals[i] = mintable[i];
     }
     return p;
 }
 
 /* equivalent of ZSTD_adjustCParams for paramValues_t */
-static paramValues_t adjustParams(paramValues_t p, const size_t maxBlockSize, const size_t dictSize) {
+static paramValues_t
+adjustParams(paramValues_t p, const size_t maxBlockSize, const size_t dictSize)
+{
     paramValues_t ot = p;
     varInds_t i;
     p = cParamsToPVals(ZSTD_adjustCParams(pvalsToCParams(p), maxBlockSize, dictSize));
-    if(!dictSize) { p.vals[fadt_ind] = 0; }
+    if (!dictSize) { p.vals[fadt_ind] = 0; }
     /* retain value of all other parameters */
     for(i = strt_ind + 1; i < NUM_PARAMS; i++) {
         p.vals[i] = ot.vals[i];
@@ -367,7 +396,10 @@ static size_t BMK_findMaxMem(U64 requiredMem)
 }
 
 /* accuracy in seconds only, span can be multiple years */
-static U32 BMK_timeSpan(const UTIL_time_t tStart) { return (U32)(UTIL_clockSpanMicro(tStart) / 1000000ULL); }
+static U32 BMK_timeSpan_s(const UTIL_time_t tStart)
+{
+    return (U32)(UTIL_clockSpanMicro(tStart) / 1000000ULL);
+}
 
 static U32 FUZ_rotl32(U32 x, U32 r)
 {
@@ -386,32 +418,38 @@ static U32 FUZ_rand(U32* src)
     return rand32 >> 5;
 }
 
-/* allows zeros */
-#define CLAMPCHECK(val,min,max) {                     \
+#define BOUNDCHECK(val,min,max) {                     \
     if (((val)<(min)) | ((val)>(max))) {              \
         DISPLAY("INVALID PARAMETER CONSTRAINTS\n");   \
         return 0;                                     \
 }   }
 
-static int paramValid(const paramValues_t paramTarget) {
+static int paramValid(const paramValues_t paramTarget)
+{
     U32 i;
     for(i = 0; i < NUM_PARAMS; i++) {
-        CLAMPCHECK(paramTarget.vals[i], mintable[i], maxtable[i]);
+        BOUNDCHECK(paramTarget.vals[i], mintable[i], maxtable[i]);
     }
     return 1;
 }
 
-static paramValues_t cParamUnsetMin(paramValues_t paramTarget) {
-    varInds_t i;
-    for(i = 0; i < NUM_PARAMS; i++) {
-        if(paramTarget.vals[i] == PARAM_UNSET) {
-            paramTarget.vals[i] = mintable[i];
+/* cParamUnsetMin() :
+ * if any parameter in paramTarget is not yet set,
+ * it will receive its corresponding minimal value.
+ * This function never fails */
+static paramValues_t cParamUnsetMin(paramValues_t paramTarget)
+{
+    varInds_t vi;
+    for (vi = 0; vi < NUM_PARAMS; vi++) {
+        if (paramTarget.vals[vi] == PARAM_UNSET) {
+            paramTarget.vals[vi] = mintable[vi];
         }
     }
     return paramTarget;
 }
 
-static paramValues_t emptyParams(void) {
+static paramValues_t emptyParams(void)
+{
     U32 i;
     paramValues_t p;
     for(i = 0; i < NUM_PARAMS; i++) {
@@ -420,7 +458,8 @@ static paramValues_t emptyParams(void) {
     return p;
 }
 
-static winnerInfo_t initWinnerInfo(const paramValues_t p) {
+static winnerInfo_t initWinnerInfo(const paramValues_t p)
+{
     winnerInfo_t w1;
     w1.result.cSpeed = 0.;
     w1.result.dSpeed = 0.;
@@ -430,7 +469,9 @@ static winnerInfo_t initWinnerInfo(const paramValues_t p) {
     return w1;
 }
 
-static paramValues_t overwriteParams(paramValues_t base, const paramValues_t mask) {
+static paramValues_t
+overwriteParams(paramValues_t base, const paramValues_t mask)
+{
     U32 i;
     for(i = 0; i < NUM_PARAMS; i++) {
         if(mask.vals[i] != PARAM_UNSET) {
@@ -440,12 +481,16 @@ static paramValues_t overwriteParams(paramValues_t base, const paramValues_t mas
     return base;
 }
 
-static void paramVaryOnce(const varInds_t paramIndex, const int amt, paramValues_t* ptr) {
-    ptr->vals[paramIndex] = rangeMap(paramIndex, invRangeMap(paramIndex, ptr->vals[paramIndex]) + amt);
+static void
+paramVaryOnce(const varInds_t paramIndex, const int amt, paramValues_t* ptr)
+{
+    ptr->vals[paramIndex] = rangeMap(paramIndex,
+                                     invRangeMap(paramIndex, ptr->vals[paramIndex]) + amt);
 }
 
 /* varies ptr by nbChanges respecting varyParams*/
-static void paramVariation(paramValues_t* ptr, memoTable_t* mtAll, const U32 nbChanges)
+static void
+paramVariation(paramValues_t* ptr, memoTable_t* mtAll, const U32 nbChanges)
 {
     paramValues_t p;
     U32 validated = 0;
@@ -473,8 +518,9 @@ static paramValues_t randomParams(void)
 
 static U64 g_clockGranularity = 100000000ULL;
 
-static void findClockGranularity(void) {
-    UTIL_time_t clockStart = UTIL_getTime();
+static void init_clockGranularity(void)
+{
+    UTIL_time_t const clockStart = UTIL_getTime();
     U64 el1 = 0, el2 = 0;
     int i = 0;
     do {
@@ -511,7 +557,9 @@ static int feasible(const BMK_benchResult_t results, const constraint_t target)
  * bonus to exceeding the constraint value. We also give linear ratio for compression ratio.
  * The constant factors are experimental.
  */
-static double resultScore(const BMK_benchResult_t res, const size_t srcSize, const constraint_t target) {
+static double
+resultScore(const BMK_benchResult_t res, const size_t srcSize, const constraint_t target)
+{
     double cs = 0., ds = 0., rt, cm = 0.;
     const double r1 = 1, r2 = 0.1, rtr = 0.5;
     double ret;
@@ -527,7 +575,9 @@ static double resultScore(const BMK_benchResult_t res, const size_t srcSize, con
 }
 
 /* calculates normalized squared euclidean distance of result1 if it is in the first quadrant relative to lvlRes */
-static double resultDistLvl(const BMK_benchResult_t result1, const BMK_benchResult_t lvlRes) {
+static double
+resultDistLvl(const BMK_benchResult_t result1, const BMK_benchResult_t lvlRes)
+{
     double normalizedCSpeedGain1 = (result1.cSpeed / lvlRes.cSpeed) - 1;
     double normalizedRatioGain1 = ((double)lvlRes.cSize / result1.cSize) - 1;
     if(normalizedRatioGain1 < 0 || normalizedCSpeedGain1 < 0) {
@@ -537,16 +587,21 @@ static double resultDistLvl(const BMK_benchResult_t result1, const BMK_benchResu
 }
 
 /* return true if r2 strictly better than r1 */
-static int compareResultLT(const BMK_benchResult_t result1, const BMK_benchResult_t result2, const constraint_t target, size_t srcSize) {
+static int
+compareResultLT(const BMK_benchResult_t result1, const BMK_benchResult_t result2, const constraint_t target, size_t srcSize)
+{
     if(feasible(result1, target) && feasible(result2, target)) {
         if(g_optmode) {
             return resultDistLvl(result1, g_lvltarget) < resultDistLvl(result2, g_lvltarget);
         } else {
-            return (result1.cSize > result2.cSize) || (result1.cSize == result2.cSize && result2.cSpeed > result1.cSpeed)
-            || (result1.cSize == result2.cSize && result2.cSpeed == result1.cSpeed && result2.dSpeed > result1.dSpeed);
+            return (result1.cSize > result2.cSize)
+                || (result1.cSize == result2.cSize && result2.cSpeed > result1.cSpeed)
+                || (result1.cSize == result2.cSize && result2.cSpeed == result1.cSpeed && result2.dSpeed > result1.dSpeed);
         }
     }
-    return feasible(result2, target) || (!feasible(result1, target) && (resultScore(result1, srcSize, target) < resultScore(result2, srcSize, target)));
+    return feasible(result2, target)
+        || (!feasible(result1, target)
+            && (resultScore(result1, srcSize, target) < resultScore(result2, srcSize, target)));
 }
 
 static constraint_t relaxTarget(constraint_t target) {
@@ -556,14 +611,17 @@ static constraint_t relaxTarget(constraint_t target) {
     return target;
 }
 
-static void optimizerAdjustInput(paramValues_t* pc, const size_t maxBlockSize) {
+static void optimizerAdjustInput(paramValues_t* pc, const size_t maxBlockSize)
+{
     varInds_t v;
     for(v = 0; v < NUM_PARAMS; v++) {
         if(pc->vals[v] != PARAM_UNSET) {
             U32 newval = MIN(MAX(pc->vals[v], mintable[v]), maxtable[v]);
             if(newval != pc->vals[v]) {
                 pc->vals[v] = newval;
-                DISPLAY("Warning: parameter %s not in valid range, adjusting to ", g_paramNames[v]); displayParamVal(stderr, v, newval, 0); DISPLAY("\n");
+                DISPLAY("Warning: parameter %s not in valid range, adjusting to ",
+                        g_paramNames[v]);
+                displayParamVal(stderr, v, newval, 0); DISPLAY("\n");
             }
         }
     }
@@ -577,7 +635,8 @@ static void optimizerAdjustInput(paramValues_t* pc, const size_t maxBlockSize) {
             U32 adjust = MAX(mintable[wlog_ind], sshb);
             if(adjust != pc->vals[wlog_ind]) {
                 pc->vals[wlog_ind] = adjust;
-                DISPLAY("Warning: windowLog larger than src/block size, adjusted to %u\n", pc->vals[wlog_ind]);
+                DISPLAY("Warning: windowLog larger than src/block size, adjusted to %u\n",
+                        (unsigned)pc->vals[wlog_ind]);
             }
         }
     }
@@ -592,40 +651,52 @@ static void optimizerAdjustInput(paramValues_t* pc, const size_t maxBlockSize) {
 
         if(pc->vals[clog_ind] > maxclog) {
             pc->vals[clog_ind] = maxclog;
-            DISPLAY("Warning: chainlog too much larger than windowLog size, adjusted to %u\n", pc->vals[clog_ind]);
+            DISPLAY("Warning: chainlog too much larger than windowLog size, adjusted to %u\n",
+                    (unsigned)pc->vals[clog_ind]);
         }
     }
 
     if(pc->vals[wlog_ind] != PARAM_UNSET && pc->vals[hlog_ind] != PARAM_UNSET) {
         if(pc->vals[wlog_ind] + 1 < pc->vals[hlog_ind]) {
             pc->vals[hlog_ind] = pc->vals[wlog_ind] + 1;
-            DISPLAY("Warning: hashlog too much larger than windowLog size, adjusted to %u\n", pc->vals[hlog_ind]);
+            DISPLAY("Warning: hashlog too much larger than windowLog size, adjusted to %u\n",
+                    (unsigned)pc->vals[hlog_ind]);
         }
     }
 
     if(pc->vals[slog_ind] != PARAM_UNSET && pc->vals[clog_ind] != PARAM_UNSET) {
         if(pc->vals[slog_ind] > pc->vals[clog_ind]) {
             pc->vals[clog_ind] = pc->vals[slog_ind];
-            DISPLAY("Warning: searchLog larger than chainLog, adjusted to %u\n", pc->vals[slog_ind]);
+            DISPLAY("Warning: searchLog larger than chainLog, adjusted to %u\n",
+                    (unsigned)pc->vals[slog_ind]);
         }
     }
 }
 
-static int redundantParams(const paramValues_t paramValues, const constraint_t target, const size_t maxBlockSize) {
+static int
+redundantParams(const paramValues_t paramValues, const constraint_t target, const size_t maxBlockSize)
+{
     return
        (ZSTD_estimateCStreamSize_usingCParams(pvalsToCParams(paramValues)) > (size_t)target.cMem) /* Uses too much memory */
     || ((1ULL << (paramValues.vals[wlog_ind] - 1)) >= maxBlockSize && paramValues.vals[wlog_ind] != mintable[wlog_ind]) /* wlog too much bigger than src size */
     || (paramValues.vals[clog_ind] > (paramValues.vals[wlog_ind] + (paramValues.vals[strt_ind] > ZSTD_btlazy2))) /* chainLog larger than windowLog*/
     || (paramValues.vals[slog_ind] > paramValues.vals[clog_ind]) /* searchLog larger than chainLog */
     || (paramValues.vals[hlog_ind] > paramValues.vals[wlog_ind] + 1); /* hashLog larger than windowLog + 1 */
-
 }
 
+
 /*-************************************
 *  Display Functions
 **************************************/
 
-static void BMK_translateAdvancedParams(FILE* f, const paramValues_t params) {
+/* BMK_paramValues_into_commandLine() :
+ * transform a set of parameters paramValues_t
+ * into a command line compatible with `zstd` syntax
+ * and writes it into FILE* f.
+ * f must be already opened and writable */
+static void
+BMK_paramValues_into_commandLine(FILE* f, const paramValues_t params)
+{
     varInds_t v;
     int first = 1;
     fprintf(f,"--zstd=");
@@ -634,60 +705,13 @@ static void BMK_translateAdvancedParams(FILE* f, const paramValues_t params) {
         if (!first) { fprintf(f, ","); }
         fprintf(f,"%s=", g_paramNames[v]);
 
-        if (v == strt_ind) { fprintf(f,"%u", params.vals[v]); }
+        if (v == strt_ind) { fprintf(f,"%u", (unsigned)params.vals[v]); }
         else { displayParamVal(f, v, params.vals[v], 0); }
         first = 0;
     }
     fprintf(f, "\n");
 }
 
-static void BMK_displayOneResult(FILE* f, winnerInfo_t res, const size_t srcSize)
-{
-    varInds_t v;
-    int first = 1;
-    res.params = cParamUnsetMin(res.params);
-    fprintf(f, "    {");
-    for (v = 0; v < NUM_PARAMS; v++) {
-        if (g_silenceParams[v]) { continue; }
-        if (!first) { fprintf(f, ","); }
-        displayParamVal(f, v, res.params.vals[v], 3);
-        first = 0;
-    }
-
-    {   double const ratio = res.result.cSize ?
-                            (double)srcSize / res.result.cSize : 0;
-        double const cSpeedMBps = (double)res.result.cSpeed / MB_UNIT;
-        double const dSpeedMBps = (double)res.result.dSpeed / MB_UNIT;
-
-        fprintf(f, " },     /* R:%5.3f at %5.1f MB/s - %5.1f MB/s */\n",
-                            ratio, cSpeedMBps, dSpeedMBps);
-    }
-}
-
-/* Writes to f the results of a parameter benchmark */
-/* when used with --optimize, will only print results better than previously discovered */
-static void BMK_printWinner(FILE* f, const int cLevel, const BMK_benchResult_t result, const paramValues_t params, const size_t srcSize)
-{
-    char lvlstr[15] = "Custom Level";
-    winnerInfo_t w;
-    w.params = params;
-    w.result = result;
-
-    fprintf(f, "\r%79s\r", "");
-
-    if(cLevel != CUSTOM_LEVEL) {
-        snprintf(lvlstr, 15, "  Level %2d  ", cLevel);
-    }
-
-    if(TIMED) {
-        const U64 time = UTIL_clockSpanNano(g_time);
-        const U64 minutes = time / (60ULL * TIMELOOP_NANOSEC);
-        fprintf(f, "%1lu:%2lu:%05.2f - ", (unsigned long) minutes / 60,(unsigned long) minutes % 60, (double)(time - minutes * TIMELOOP_NANOSEC * 60ULL)/TIMELOOP_NANOSEC);
-    }
-
-    fprintf(f, "/* %s */   ", lvlstr);
-    BMK_displayOneResult(f, w, srcSize);
-}
 
 /* comparison function: */
 /* strictly better, strictly worse, equal, speed-side adv, size-side adv */
@@ -698,7 +722,9 @@ static void BMK_printWinner(FILE* f, const int cLevel, const BMK_benchResult_t r
 #define SPEED_RESULT 4
 #define SIZE_RESULT 5
 /* maybe have epsilon-eq to limit table size? */
-static int speedSizeCompare(const BMK_benchResult_t r1, const BMK_benchResult_t r2) {
+static int
+speedSizeCompare(const BMK_benchResult_t r1, const BMK_benchResult_t r2)
+{
     if(r1.cSpeed < r2.cSpeed) {
         if(r1.cSize >= r2.cSize) {
             return BETTER_RESULT;
@@ -714,7 +740,9 @@ static int speedSizeCompare(const BMK_benchResult_t r1, const BMK_benchResult_t
 
 /* 0 for insertion, 1 for no insert */
 /* maintain invariant speedSizeCompare(n, n->next) = SPEED_RESULT */
-static int insertWinner(const winnerInfo_t w, const constraint_t targetConstraints) {
+static int
+insertWinner(const winnerInfo_t w, const constraint_t targetConstraints)
+{
     BMK_benchResult_t r = w.result;
     winner_ll_node* cur_node = g_winners;
     /* first node to insert */
@@ -808,20 +836,82 @@ static int insertWinner(const winnerInfo_t w, const constraint_t targetConstrain
     }
 }
 
-static void BMK_printWinnerOpt(FILE* f, const U32 cLevel, const BMK_benchResult_t result, const paramValues_t params, const constraint_t targetConstraints, const size_t srcSize)
+static void
+BMK_displayOneResult(FILE* f, winnerInfo_t res, const size_t srcSize)
+{
+    varInds_t v;
+    int first = 1;
+    res.params = cParamUnsetMin(res.params);
+    fprintf(f, "    {");
+    for (v = 0; v < NUM_PARAMS; v++) {
+        if (g_silenceParams[v]) { continue; }
+        if (!first) { fprintf(f, ","); }
+        displayParamVal(f, v, res.params.vals[v], 3);
+        first = 0;
+    }
+
+    {   double const ratio = res.result.cSize ?
+                            (double)srcSize / res.result.cSize : 0;
+        double const cSpeedMBps = (double)res.result.cSpeed / MB_UNIT;
+        double const dSpeedMBps = (double)res.result.dSpeed / MB_UNIT;
+
+        fprintf(f, " },     /* R:%5.3f at %5.1f MB/s - %5.1f MB/s */\n",
+                            ratio, cSpeedMBps, dSpeedMBps);
+    }
+}
+
+/* Writes to f the results of a parameter benchmark */
+/* when used with --optimize, will only print results better than previously discovered */
+static void
+BMK_printWinner(FILE* f, const int cLevel, const BMK_benchResult_t result, const paramValues_t params, const size_t srcSize)
+{
+    char lvlstr[15] = "Custom Level";
+    winnerInfo_t w;
+    w.params = params;
+    w.result = result;
+
+    fprintf(f, "\r%79s\r", "");
+
+    if(cLevel != CUSTOM_LEVEL) {
+        snprintf(lvlstr, 15, "  Level %2d  ", cLevel);
+    }
+
+    if(TIMED) {
+        const U64 mn_in_ns = 60ULL * TIMELOOP_NANOSEC;
+        const U64 time_ns = UTIL_clockSpanNano(g_time);
+        const U64 minutes = time_ns / mn_in_ns;
+        fprintf(f, "%1lu:%2lu:%05.2f - ",
+                (unsigned long) minutes / 60,
+                (unsigned long) minutes % 60,
+                (double)(time_ns - (minutes * mn_in_ns)) / TIMELOOP_NANOSEC );
+    }
+
+    fprintf(f, "/* %s */   ", lvlstr);
+    BMK_displayOneResult(f, w, srcSize);
+}
+
+static void
+BMK_printWinnerOpt(FILE* f, const U32 cLevel, const BMK_benchResult_t result, const paramValues_t params, const constraint_t targetConstraints, const size_t srcSize)
 {
     /* global winner used for constraints */
                                     /* cSize, cSpeed, dSpeed, cMem */
-    static winnerInfo_t g_winner = { { (size_t)-1LL, 0, 0, (size_t)-1LL }, { { PARAM_UNSET, PARAM_UNSET, PARAM_UNSET, PARAM_UNSET, PARAM_UNSET, PARAM_UNSET, PARAM_UNSET, PARAM_UNSET } } };
-    if(DEBUG || compareResultLT(g_winner.result, result, targetConstraints, srcSize) || g_displayLevel >= 4) {
-        if(DEBUG && compareResultLT(g_winner.result, result, targetConstraints, srcSize)) {
+    static winnerInfo_t g_winner = { { (size_t)-1LL, 0, 0, (size_t)-1LL },
+                                     { { PARAM_UNSET, PARAM_UNSET, PARAM_UNSET, PARAM_UNSET, PARAM_UNSET, PARAM_UNSET, PARAM_UNSET, PARAM_UNSET } }
+                                   };
+    if ( DEBUG
+      || compareResultLT(g_winner.result, result, targetConstraints, srcSize)
+      || g_displayLevel >= 4) {
+        if ( DEBUG
+          && compareResultLT(g_winner.result, result, targetConstraints, srcSize)) {
             DISPLAY("New Winner: \n");
         }
 
-        if(g_displayLevel >= 2) { BMK_printWinner(f, cLevel, result, params, srcSize); }
+        if(g_displayLevel >= 2) {
+            BMK_printWinner(f, cLevel, result, params, srcSize);
+        }
 
         if(compareResultLT(g_winner.result, result, targetConstraints, srcSize)) {
-            if(g_displayLevel >= 1) { BMK_translateAdvancedParams(f, params); }
+            if(g_displayLevel >= 1) { BMK_paramValues_into_commandLine(f, params); }
             g_winner.result = result;
             g_winner.params = params;
         }
@@ -849,31 +939,88 @@ static void BMK_printWinnerOpt(FILE* f, const U32 cLevel, const BMK_benchResult_
 
         fprintf(f, "Overall Winner: \n");
         BMK_displayOneResult(f, g_winner, srcSize);
-        BMK_translateAdvancedParams(f, g_winner.params);
+        BMK_paramValues_into_commandLine(f, g_winner.params);
 
         fprintf(f, "Latest BMK: \n");\
         BMK_displayOneResult(f, w, srcSize);
     }
 }
 
-static void BMK_printWinners2(FILE* f, const winnerInfo_t* winners, const size_t srcSize)
+
+/* BMK_print_cLevelEntry() :
+ * Writes one cLevelTable entry, for one level.
+ * f must exist, be already opened, and be seekable.
+ * this function cannot error.
+ */
+static void
+BMK_print_cLevelEntry(FILE* f, const int cLevel,
+                      paramValues_t params,
+                      const BMK_benchResult_t result, const size_t srcSize)
+{
+    varInds_t v;
+    int first = 1;
+
+    assert(cLevel >= 0);
+    assert(cLevel <= NB_LEVELS_TRACKED);
+    params = cParamUnsetMin(params);
+
+    fprintf(f, "   {");
+    /* print cParams.
+     * assumption : all cParams are present and in order in the following range */
+    for (v = 0; v <= strt_ind; v++) {
+        if (!first) { fprintf(f, ","); }
+        displayParamVal(f, v, params.vals[v], 3);
+        first = 0;
+    }
+    /* print comment */
+    {   double const ratio = result.cSize ?
+                            (double)srcSize / result.cSize : 0;
+        double const cSpeedMBps = (double)result.cSpeed / MB_UNIT;
+        double const dSpeedMBps = (double)result.dSpeed / MB_UNIT;
+
+        fprintf(f, " },   /* level %2i:  R=%5.3f at %5.1f MB/s - %5.1f MB/s */\n",
+                             cLevel, ratio, cSpeedMBps, dSpeedMBps);
+    }
+}
+
+
+/* BMK_print_cLevelTable() :
+ * print candidate compression table into proposed FILE* f.
+ * f must exist, be already opened, and be seekable.
+ * winners must be a table of NB_LEVELS_TRACKED+1 elements winnerInfo_t, all entries presumed initialized
+ * this function cannot error.
+ */
+static void
+BMK_print_cLevelTable(FILE* f, const winnerInfo_t* winners, const size_t srcSize)
 {
     int cLevel;
 
     fprintf(f, "\n /* Proposed configurations : */ \n");
-    fprintf(f, "    /* W,  C,  H,  S,  L,  T, strat */ \n");
+    fprintf(f, "   /* W,  C,  H,  S,  L,  T, strat */ \n");
 
     for (cLevel=0; cLevel <= NB_LEVELS_TRACKED; cLevel++)
-        BMK_printWinner(f, cLevel, winners[cLevel].result, winners[cLevel].params, srcSize);
+        BMK_print_cLevelEntry(f,
+                              cLevel, winners[cLevel].params,
+                              winners[cLevel].result, srcSize);
 }
 
 
-static void BMK_printWinners(FILE* f, const winnerInfo_t* winners, const size_t srcSize)
+/* BMK_saveAndPrint_cLevelTable() :
+ * save candidate compression table into FILE* f,
+ * and then to stdout.
+ * f must exist, be already opened, and be seekable.
+ * winners must be a table of NB_LEVELS_TRACKED+1 elements winnerInfo_t, all entries presumed initialized
+ * this function cannot error.
+ */
+static void
+BMK_saveAndPrint_cLevelTable(FILE* const f,
+                       const winnerInfo_t* winners,
+                       const size_t srcSize)
 {
     fseek(f, 0, SEEK_SET);
-    BMK_printWinners2(f, winners, srcSize);
+    BMK_print_cLevelTable(f, winners, srcSize);
     fflush(f);
-    BMK_printWinners2(stdout, winners, srcSize);
+    BMK_print_cLevelTable(stdout, winners, srcSize);
 }
 
 
@@ -892,9 +1039,8 @@ typedef struct {
 static size_t local_initCCtx(void* payload) {
     const BMK_initCCtxArgs* ag = (const BMK_initCCtxArgs*)payload;
     varInds_t i;
-    ZSTD_CCtx_reset(ag->cctx);
-    ZSTD_CCtx_resetParameters(ag->cctx);
-    ZSTD_CCtx_setParameter(ag->cctx, ZSTD_p_compressionLevel, ag->cLevel);
+    ZSTD_CCtx_reset(ag->cctx, ZSTD_reset_session_and_parameters);
+    ZSTD_CCtx_setParameter(ag->cctx, ZSTD_c_compressionLevel, ag->cLevel);
 
     for(i = 0; i < NUM_PARAMS; i++) {
         if(ag->comprParams->vals[i] != PARAM_UNSET)
@@ -913,37 +1059,20 @@ typedef struct {
 
 static size_t local_initDCtx(void* payload) {
     const BMK_initDCtxArgs* ag = (const BMK_initDCtxArgs*)payload;
-    ZSTD_DCtx_reset(ag->dctx);
+    ZSTD_DCtx_reset(ag->dctx, ZSTD_reset_session_and_parameters);
     ZSTD_DCtx_loadDictionary(ag->dctx, ag->dictBuffer, ag->dictBufferSize);
     return 0;
 }
 
 /* additional argument is just the context */
 static size_t local_defaultCompress(
-    const void* srcBuffer, size_t srcSize,
-    void* dstBuffer, size_t dstSize,
-    void* addArgs) {
-    size_t moreToFlush = 1;
-    ZSTD_CCtx* ctx = (ZSTD_CCtx*)addArgs;
-    ZSTD_inBuffer in;
-    ZSTD_outBuffer out;
-    in.src = srcBuffer;
-    in.size = srcSize;
-    in.pos = 0;
-    out.dst = dstBuffer;
-    out.size = dstSize;
-    out.pos = 0;
+                            const void* srcBuffer, size_t srcSize,
+                            void* dstBuffer, size_t dstSize,
+                            void* addArgs)
+{
+    ZSTD_CCtx* cctx = (ZSTD_CCtx*)addArgs;
     assert(dstSize == ZSTD_compressBound(srcSize)); /* specific to this version, which is only used in paramgrill */
-    while (moreToFlush) {
-        if(out.pos == out.size) {
-            return (size_t)-ZSTD_error_dstSize_tooSmall;
-        }
-        moreToFlush = ZSTD_compress_generic(ctx, &out, &in, ZSTD_e_end);
-        if (ZSTD_isError(moreToFlush)) {
-            return moreToFlush;
-        }
-    }
-    return out.pos;
+    return ZSTD_compress2(cctx, dstBuffer, dstSize, srcBuffer, srcSize);
 }
 
 /* additional argument is just the context */
@@ -965,7 +1094,7 @@ static size_t local_defaultDecompress(
         if(out.pos == out.size) {
             return (size_t)-ZSTD_error_dstSize_tooSmall;
         }
-        moreToFlush = ZSTD_decompress_generic(dctx,
+        moreToFlush = ZSTD_decompressStream(dctx,
                             &out, &in);
         if (ZSTD_isError(moreToFlush)) {
             return moreToFlush;
@@ -1235,7 +1364,7 @@ static size_t sanitizeVarArray(varInds_t* varNew, const size_t varLength, const
         if( !((varArray[i] == clog_ind && strat == ZSTD_fast)
             || (varArray[i] == slog_ind && strat == ZSTD_fast)
             || (varArray[i] == slog_ind && strat == ZSTD_dfast)
-            || (varArray[i] == tlen_ind && strat != ZSTD_btopt && strat != ZSTD_btultra && strat != ZSTD_fast))) {
+            || (varArray[i] == tlen_ind && strat < ZSTD_btopt && strat != ZSTD_fast))) {
             varNew[j] = varArray[i];
             j++;
         }
@@ -1307,10 +1436,12 @@ static void memoTableSet(const memoTable_t* memoTableArray, const paramValues_t
 }
 
 /* frees all allocated memotables */
+/* secret contract :
+ * mtAll is a table of (ZSTD_STRATEGY_MAX+1) memoTable_t */
 static void freeMemoTableArray(memoTable_t* const mtAll) {
     int i;
     if(mtAll == NULL) { return; }
-    for(i = 1; i <= (int)ZSTD_btultra; i++) {
+    for(i = 1; i <= (int)ZSTD_STRATEGY_MAX; i++) {
         free(mtAll[i].table);
     }
     free(mtAll);
@@ -1318,21 +1449,26 @@ static void freeMemoTableArray(memoTable_t* const mtAll) {
 
 /* inits memotables for all (including mallocs), all strategies */
 /* takes unsanitized varyParams */
-static memoTable_t* createMemoTableArray(const paramValues_t p, const varInds_t* const varyParams, const size_t varyLen, const U32 memoTableLog) {
-    memoTable_t* mtAll = (memoTable_t*)calloc(sizeof(memoTable_t),(ZSTD_btultra + 1));
-    ZSTD_strategy i, stratMin = ZSTD_fast, stratMax = ZSTD_btultra;
+static memoTable_t*
+createMemoTableArray(const paramValues_t p,
+                     const varInds_t* const varyParams,
+                     const size_t varyLen,
+                     const U32 memoTableLog)
+{
+    memoTable_t* const mtAll = (memoTable_t*)calloc(sizeof(memoTable_t),(ZSTD_STRATEGY_MAX + 1));
+    ZSTD_strategy i, stratMin = ZSTD_STRATEGY_MIN, stratMax = ZSTD_STRATEGY_MAX;
 
     if(mtAll == NULL) {
         return NULL;
     }
 
-    for(i = 1; i <= (int)ZSTD_btultra; i++) {
+    for(i = 1; i <= (int)ZSTD_STRATEGY_MAX; i++) {
         mtAll[i].varLen = sanitizeVarArray(mtAll[i].varArray, varyLen, varyParams, i);
     }
 
     /* no memoization */
     if(memoTableLog == 0) {
-        for(i = 1; i <= (int)ZSTD_btultra; i++) {
+        for(i = 1; i <= (int)ZSTD_STRATEGY_MAX; i++) {
             mtAll[i].tableType = noMemo;
             mtAll[i].table = NULL;
             mtAll[i].tableLen = 0;
@@ -1391,13 +1527,24 @@ static void randomConstrainedParams(paramValues_t* pc, const memoTable_t* memoTa
 *  Benchmarking Functions
 **************************************/
 
+static void display_params_tested(paramValues_t cParams)
+{
+    varInds_t vi;
+    DISPLAYLEVEL(3, "\r testing :");
+    for (vi=0; vi < NUM_PARAMS; vi++) {
+        DISPLAYLEVEL(3, "%3u,", (unsigned)cParams.vals[vi]);
+    }
+    DISPLAYLEVEL(3, "\b    \r");
+}
+
 /* Replicate functionality of benchMemAdvanced, but with pre-split src / dst buffers */
 /* The purpose is so that sufficient information is returned so that a decompression call to benchMemInvertible is possible */
 /* BMK_benchMemAdvanced(srcBuffer,srcSize, dstBuffer, dstSize, fileSizes, nbFiles, 0, &cParams, dictBuffer, dictSize, ctx, dctx, 0, "File", &adv); */
 /* nbSeconds used in same way as in BMK_advancedParams_t */
 /* if in decodeOnly, then srcPtr's will be compressed blocks, and uncompressedBlocks will be written to dstPtrs */
 /* dictionary nullable, nothing else though. */
-/* note : it would be better if this function was in bench.c, sharing code with benchMemAdvanced(), since it's technically a part of it */
+/* note : it would be a lot better if this function was present in benchzstd.c,
+ * sharing code with benchMemAdvanced(), since it's technically a part of it */
 static BMK_benchOutcome_t
 BMK_benchMemInvertible( buffers_t buf, contexts_t ctx,
                         int cLevel, const paramValues_t* comprParams,
@@ -1420,6 +1567,7 @@ BMK_benchMemInvertible( buffers_t buf, contexts_t ctx,
     ZSTD_DCtx* dctx = ctx.dctx;
 
     /* init */
+    display_params_tested(*comprParams);
     memset(&bResult, 0, sizeof(bResult));
 
     /* warmimg up memory */
@@ -1438,13 +1586,40 @@ BMK_benchMemInvertible( buffers_t buf, contexts_t ctx,
         int decompressionCompleted = (mode == BMK_compressOnly);
         BMK_timedFnState_t* timeStateCompress = BMK_createTimedFnState(nbSeconds * 1000, 1000);
         BMK_timedFnState_t* timeStateDecompress = BMK_createTimedFnState(nbSeconds * 1000, 1000);
+        BMK_benchParams_t cbp, dbp;
         BMK_initCCtxArgs cctxprep;
         BMK_initDCtxArgs dctxprep;
+
+        cbp.benchFn = local_defaultCompress;
+        cbp.benchPayload = cctx;
+        cbp.initFn = local_initCCtx;
+        cbp.initPayload = &cctxprep;
+        cbp.errorFn = ZSTD_isError;
+        cbp.blockCount = nbBlocks;
+        cbp.srcBuffers = srcPtrs;
+        cbp.srcSizes = srcSizes;
+        cbp.dstBuffers = dstPtrs;
+        cbp.dstCapacities = dstCapacities;
+        cbp.blockResults = dstSizes;
+
         cctxprep.cctx = cctx;
         cctxprep.dictBuffer = dictBuffer;
         cctxprep.dictBufferSize = dictBufferSize;
         cctxprep.cLevel = cLevel;
         cctxprep.comprParams = comprParams;
+
+        dbp.benchFn = local_defaultDecompress;
+        dbp.benchPayload = dctx;
+        dbp.initFn = local_initDCtx;
+        dbp.initPayload = &dctxprep;
+        dbp.errorFn = ZSTD_isError;
+        dbp.blockCount = nbBlocks;
+        dbp.srcBuffers = (const void* const *) dstPtrs;
+        dbp.srcSizes = dstCapacities;
+        dbp.dstBuffers = resPtrs;
+        dbp.dstCapacities = resSizes;
+        dbp.blockResults = NULL;
+
         dctxprep.dctx = dctx;
         dctxprep.dictBuffer = dictBuffer;
         dctxprep.dictBufferSize = dictBufferSize;
@@ -1452,13 +1627,7 @@ BMK_benchMemInvertible( buffers_t buf, contexts_t ctx,
         assert(timeStateCompress != NULL);
         assert(timeStateDecompress != NULL);
         while(!compressionCompleted) {
-            BMK_runOutcome_t const cOutcome = BMK_benchTimedFn(timeStateCompress,
-                                            &local_defaultCompress, cctx,
-                                            &local_initCCtx, &cctxprep,
-                                            nbBlocks,
-                                            srcPtrs, srcSizes,
-                                            dstPtrs, dstCapacities,
-                                            dstSizes);
+            BMK_runOutcome_t const cOutcome = BMK_benchTimedFn(timeStateCompress, cbp);
 
             if (!BMK_isSuccessful_runOutcome(cOutcome)) {
                 BMK_benchOutcome_t bOut;
@@ -1476,13 +1645,7 @@ BMK_benchMemInvertible( buffers_t buf, contexts_t ctx,
         }
 
         while (!decompressionCompleted) {
-            BMK_runOutcome_t const dOutcome = BMK_benchTimedFn(timeStateDecompress,
-                                        &local_defaultDecompress, dctx,
-                                        &local_initDCtx, &dctxprep,
-                                        nbBlocks,
-                                        (const void* const*)dstPtrs, dstSizes,
-                                        resPtrs, resSizes,
-                                        NULL);
+            BMK_runOutcome_t const dOutcome = BMK_benchTimedFn(timeStateDecompress, dbp);
 
             if (!BMK_isSuccessful_runOutcome(dOutcome)) {
                 BMK_benchOutcome_t bOut;
@@ -1512,6 +1675,10 @@ BMK_benchMemInvertible( buffers_t buf, contexts_t ctx,
     }
 }
 
+/* BMK_benchParam() :
+ * benchmark a set of `cParams` over sample `buf`,
+ * store the result in `resultPtr`.
+ * @return : 0 if success, 1 if error */
 static int BMK_benchParam ( BMK_benchResult_t* resultPtr,
                             buffers_t buf, contexts_t ctx,
                             paramValues_t cParams)
@@ -1519,30 +1686,12 @@ static int BMK_benchParam ( BMK_benchResult_t* resultPtr,
     BMK_benchOutcome_t const outcome = BMK_benchMemInvertible(buf, ctx,
                                                         BASE_CLEVEL, &cParams,
                                                         BMK_both, 3);
-    int const success = BMK_isSuccessful_benchOutcome(outcome);
-    if (!success) return 1;
+    if (!BMK_isSuccessful_benchOutcome(outcome)) return 1;
     *resultPtr = BMK_extract_benchResult(outcome);
     return 0;
 }
 
 
-#define CBENCHMARK(conditional, resultvar, tmpret, mode, sec) {                                                 \
-    if(conditional) {                                                                                           \
-        BMK_benchOutcome_t const outcome = BMK_benchMemInvertible(buf, ctx, BASE_CLEVEL, &cParams, mode, sec);  \
-        if (!BMK_isSuccessful_benchOutcome(outcome)) {                                                          \
-            DEBUGOUTPUT("Benchmarking failed\n");                                                               \
-            return ERROR_RESULT;                                                                                \
-        }                                                                                                       \
-        {   BMK_benchResult_t const tmpResult = BMK_extract_benchResult(outcome);                               \
-            if (mode != BMK_decodeOnly)  {                                                                      \
-                resultvar.cSpeed = tmpResult.cSpeed;                                                            \
-                resultvar.cSize = tmpResult.cSize;                                                              \
-                resultvar.cMem = tmpResult.cMem;                                                                \
-            }                                                                                                   \
-            if (mode != BMK_compressOnly) { resultvar.dSpeed = tmpResult.dSpeed; }                              \
-    }   }                                                                                                       \
-}
-
 /* Benchmarking which stops when we are sufficiently sure the solution is infeasible / worse than the winner */
 #define VARIANCE 1.2
 static int allBench(BMK_benchResult_t* resultPtr,
@@ -1662,12 +1811,14 @@ static void BMK_init_level_constraints(int bytePerSec_level1)
             g_level_constraint[l].cSpeed_min = (g_level_constraint[l-1].cSpeed_min * 49) / 64;
             g_level_constraint[l].dSpeed_min = 0.;
             g_level_constraint[l].windowLog_max = (l<20) ? 23 : l+5;   /* only --ultra levels >= 20 can use windowlog > 23 */
-            g_level_constraint[l].strategy_max = (l<19) ? ZSTD_btopt : ZSTD_btultra;   /* level 19 is allowed to use btultra */
+            g_level_constraint[l].strategy_max = ZSTD_STRATEGY_MAX;
     }   }
 }
 
-static int BMK_seed(winnerInfo_t* winners, const paramValues_t params,
-                    const buffers_t buf, const contexts_t ctx)
+static int BMK_seed(winnerInfo_t* winners,
+                    const paramValues_t params,
+                    const buffers_t buf,
+                    const contexts_t ctx)
 {
     BMK_benchResult_t testResult;
     int better = 0;
@@ -1675,8 +1826,8 @@ static int BMK_seed(winnerInfo_t* winners, const paramValues_t params,
 
     BMK_benchParam(&testResult, buf, ctx, params);
 
-
     for (cLevel = 1; cLevel <= NB_LEVELS_TRACKED; cLevel++) {
+
         if (testResult.cSpeed < g_level_constraint[cLevel].cSpeed_min)
             continue;   /* not fast enough for this level */
         if (testResult.dSpeed < g_level_constraint[cLevel].dSpeed_min)
@@ -1689,7 +1840,7 @@ static int BMK_seed(winnerInfo_t* winners, const paramValues_t params,
             /* first solution for this cLevel */
             winners[cLevel].result = testResult;
             winners[cLevel].params = params;
-            BMK_printWinner(stdout, cLevel, testResult, params, buf.srcSize);
+            BMK_print_cLevelEntry(stdout, cLevel, params, testResult, buf.srcSize);
             better = 1;
             continue;
         }
@@ -1719,7 +1870,7 @@ static int BMK_seed(winnerInfo_t* winners, const paramValues_t params,
             if (W_DMemUsed_note < O_DMemUsed_note) {
                 /* uses too much Decompression memory for too little benefit */
                 if (W_ratio > O_ratio)
-                DISPLAY ("Decompression Memory : %5.3f @ %4.1f MB  vs  %5.3f @ %4.1f MB   : not enough for level %i\n",
+                DISPLAYLEVEL(3, "Decompression Memory : %5.3f @ %4.1f MB  vs  %5.3f @ %4.1f MB   : not enough for level %i\n",
                          W_ratio, (double)(W_DMemUsed) / 1024 / 1024,
                          O_ratio, (double)(O_DMemUsed) / 1024 / 1024,   cLevel);
                 continue;
@@ -1727,34 +1878,38 @@ static int BMK_seed(winnerInfo_t* winners, const paramValues_t params,
             if (W_CMemUsed_note < O_CMemUsed_note) {
                 /* uses too much memory for compression for too little benefit */
                 if (W_ratio > O_ratio)
-                DISPLAY ("Compression Memory : %5.3f @ %4.1f MB  vs  %5.3f @ %4.1f MB   : not enough for level %i\n",
+                DISPLAYLEVEL(3, "Compression Memory : %5.3f @ %4.1f MB  vs  %5.3f @ %4.1f MB   : not enough for level %i\n",
                          W_ratio, (double)(W_CMemUsed) / 1024 / 1024,
-                         O_ratio, (double)(O_CMemUsed) / 1024 / 1024,   cLevel);
+                         O_ratio, (double)(O_CMemUsed) / 1024 / 1024,
+                         cLevel);
                 continue;
             }
             if (W_CSpeed_note   < O_CSpeed_note  ) {
                 /* too large compression speed difference for the compression benefit */
                 if (W_ratio > O_ratio)
-                DISPLAY ("Compression Speed : %5.3f @ %4.1f MB/s  vs  %5.3f @ %4.1f MB/s   : not enough for level %i\n",
+                DISPLAYLEVEL(3, "Compression Speed : %5.3f @ %4.1f MB/s  vs  %5.3f @ %4.1f MB/s   : not enough for level %i\n",
                          W_ratio, (double)testResult.cSpeed / MB_UNIT,
-                         O_ratio, (double)winners[cLevel].result.cSpeed / MB_UNIT,   cLevel);
+                         O_ratio, (double)winners[cLevel].result.cSpeed / MB_UNIT,
+                         cLevel);
                 continue;
             }
             if (W_DSpeed_note   < O_DSpeed_note  ) {
                 /* too large decompression speed difference for the compression benefit */
                 if (W_ratio > O_ratio)
-                DISPLAY ("Decompression Speed : %5.3f @ %4.1f MB/s  vs  %5.3f @ %4.1f MB/s   : not enough for level %i\n",
+                DISPLAYLEVEL(3, "Decompression Speed : %5.3f @ %4.1f MB/s  vs  %5.3f @ %4.1f MB/s   : not enough for level %i\n",
                          W_ratio, (double)testResult.dSpeed / MB_UNIT,
-                         O_ratio, (double)winners[cLevel].result.dSpeed / MB_UNIT,   cLevel);
+                         O_ratio, (double)winners[cLevel].result.dSpeed / MB_UNIT,
+                         cLevel);
                 continue;
             }
 
             if (W_ratio < O_ratio)
-                DISPLAY("Solution %4.3f selected over %4.3f at level %i, due to better secondary statistics \n", W_ratio, O_ratio, cLevel);
+                DISPLAYLEVEL(3, "Solution %4.3f selected over %4.3f at level %i, due to better secondary statistics \n",
+                                W_ratio, O_ratio, cLevel);
 
             winners[cLevel].result = testResult;
             winners[cLevel].params = params;
-            BMK_printWinner(stdout, cLevel, testResult, params, buf.srcSize);
+            BMK_print_cLevelEntry(stdout, cLevel, params, testResult, buf.srcSize);
 
             better = 1;
     }   }
@@ -1771,58 +1926,75 @@ static int BMK_seed(winnerInfo_t* winners, const paramValues_t params,
 #define PARAMTABLEMASK (PARAMTABLESIZE-1)
 static BYTE g_alreadyTested[PARAMTABLESIZE] = {0};   /* init to zero */
 
-static BYTE* NB_TESTS_PLAYED(paramValues_t p) {
-    ZSTD_compressionParameters p2 = pvalsToCParams(sanitizeParams(p));
-    return &g_alreadyTested[(XXH64((void*)&p2, sizeof(p2), 0) >> 3) & PARAMTABLEMASK];
+static BYTE* NB_TESTS_PLAYED(paramValues_t p)
+{
+    ZSTD_compressionParameters const cParams = pvalsToCParams(sanitizeParams(p));
+    unsigned long long const h64 = XXH64(&cParams, sizeof(cParams), 0);
+    return &g_alreadyTested[(h64 >> 3) & PARAMTABLEMASK];
 }
 
-static void playAround(FILE* f, winnerInfo_t* winners,
+static void playAround(FILE* f,
+                       winnerInfo_t* winners,
                        paramValues_t p,
                        const buffers_t buf, const contexts_t ctx)
 {
-    int nbVariations = 0, i;
+    int nbVariations = 0;
     UTIL_time_t const clockStart = UTIL_getTime();
 
     while (UTIL_clockSpanMicro(clockStart) < g_maxVariationTime) {
-        BYTE* b;
-
         if (nbVariations++ > g_maxNbVariations) break;
 
-        do { for(i = 0; i < 4; i++) { paramVaryOnce(FUZ_rand(&g_rand) % (strt_ind + 1), ((FUZ_rand(&g_rand) & 1) << 1) - 1, &p); } }
-        while(!paramValid(p));
+        do {
+            int i;
+            for(i = 0; i < 4; i++) {
+                paramVaryOnce(FUZ_rand(&g_rand) % (strt_ind + 1),
+                              ((FUZ_rand(&g_rand) & 1) << 1) - 1,
+                              &p);
+            }
+        } while (!paramValid(p));
 
         /* exclude faster if already played params */
         if (FUZ_rand(&g_rand) & ((1 << *NB_TESTS_PLAYED(p))-1))
             continue;
 
         /* test */
-        b = NB_TESTS_PLAYED(p);
-        (*b)++;
+        {   BYTE* const b = NB_TESTS_PLAYED(p);
+            (*b)++;
+        }
         if (!BMK_seed(winners, p, buf, ctx)) continue;
 
         /* improvement found => search more */
-        BMK_printWinners(f, winners, buf.srcSize);
+        BMK_saveAndPrint_cLevelTable(f, winners, buf.srcSize);
         playAround(f, winners, p, buf, ctx);
     }
 
 }
 
-static void BMK_selectRandomStart(
-                       FILE* f, winnerInfo_t* winners,
+static void
+BMK_selectRandomStart( FILE* f,
+                       winnerInfo_t* winners,
                        const buffers_t buf, const contexts_t ctx)
 {
     U32 const id = FUZ_rand(&g_rand) % (NB_LEVELS_TRACKED+1);
     if ((id==0) || (winners[id].params.vals[wlog_ind]==0)) {
         /* use some random entry */
         paramValues_t const p = adjustParams(cParamsToPVals(pvalsToCParams(randomParams())), /* defaults nonCompression parameters */
-            buf.srcSize, 0);
+                                             buf.srcSize, 0);
         playAround(f, winners, p, buf, ctx);
     } else {
         playAround(f, winners, winners[id].params, buf, ctx);
     }
 }
 
-static void BMK_benchFullTable(const buffers_t buf, const contexts_t ctx)
+
+/* BMK_generate_cLevelTable() :
+ * test a large number of configurations
+ * and distribute them accross compression levels according to speed conditions.
+ * display and save all intermediate results into rfName = "grillResults.txt".
+ * the function automatically stops after g_timeLimit_s.
+ * this function cannot error, it directly exit() in case of problem.
+ */
+static void BMK_generate_cLevelTable(const buffers_t buf, const contexts_t ctx)
 {
     paramValues_t params;
     winnerInfo_t winners[NB_LEVELS_TRACKED+1];
@@ -1851,17 +2023,17 @@ static void BMK_benchFullTable(const buffers_t buf, const contexts_t ctx)
             params = cParamsToPVals(ZSTD_getCParams(i, buf.maxBlockSize, 0));
             BMK_seed(winners, params, buf, ctx);
     }   }
-    BMK_printWinners(f, winners, buf.srcSize);
+    BMK_saveAndPrint_cLevelTable(f, winners, buf.srcSize);
 
     /* start tests */
     {   const UTIL_time_t grillStart = UTIL_getTime();
         do {
             BMK_selectRandomStart(f, winners, buf, ctx);
-        } while (BMK_timeSpan(grillStart) < g_timeLimit_s);
+        } while (BMK_timeSpan_s(grillStart) < g_timeLimit_s);
     }
 
     /* end summary */
-    BMK_printWinners(f, winners, buf.srcSize);
+    BMK_saveAndPrint_cLevelTable(f, winners, buf.srcSize);
     DISPLAY("grillParams operations completed \n");
 
     /* clean up*/
@@ -1873,7 +2045,9 @@ static void BMK_benchFullTable(const buffers_t buf, const contexts_t ctx)
 *  Single Benchmark Functions
 **************************************/
 
-static int benchOnce(const buffers_t buf, const contexts_t ctx, const int cLevel) {
+static int
+benchOnce(const buffers_t buf, const contexts_t ctx, const int cLevel)
+{
     BMK_benchResult_t testResult;
     g_params = adjustParams(overwriteParams(cParamsToPVals(ZSTD_getCParams(cLevel, buf.maxBlockSize, ctx.dictSize)), g_params), buf.maxBlockSize, ctx.dictSize);
 
@@ -1923,7 +2097,7 @@ static int benchSample(double compressibility, int cLevel)
     if(g_singleRun) {
         ret = benchOnce(buf, ctx, cLevel);
     } else {
-        BMK_benchFullTable(buf, ctx);
+        BMK_generate_cLevelTable(buf, ctx);
     }
 
     freeBuffers(buf);
@@ -1963,7 +2137,7 @@ static int benchFiles(const char** fileNamesTable, int nbFiles,
     if (g_singleRun) {
         ret = benchOnce(buf, ctx, cLevel);
     } else {
-        BMK_benchFullTable(buf, ctx);
+        BMK_generate_cLevelTable(buf, ctx);
     }
 
     freeBuffers(buf);
@@ -2040,15 +2214,15 @@ static winnerInfo_t climbOnce(const constraint_t target,
                             }
                         }
                     }
-                }
+                }  /* for (offset = -1; offset <= 1; offset += 2) */
             }   /* for (i = 0; i < varLen; i++) */
 
             if(better) {
                 continue;
             }
 
-            for(dist = 2; dist < varLen + 2; dist++) { /* varLen is # dimensions */
-                for(i = 0; i < (1 << varLen) / varLen + 2; i++) {
+            for (dist = 2; dist < varLen + 2; dist++) { /* varLen is # dimensions */
+                for (i = 0; i < (1 << varLen) / varLen + 2; i++) {
                     int res;
                     CHECKTIME(winnerInfo);
                     candidateInfo.params = cparam;
@@ -2095,11 +2269,12 @@ static winnerInfo_t climbOnce(const constraint_t target,
    weight more on visit for bad results, less on good results/more on later results / ones with more failures.
    allocate memoTable here.
  */
-static winnerInfo_t optimizeFixedStrategy(
-    const buffers_t buf, const contexts_t ctx,
-    const constraint_t target, paramValues_t paramTarget,
-    const ZSTD_strategy strat,
-    memoTable_t* memoTableArray, const int tries) {
+static winnerInfo_t
+optimizeFixedStrategy(const buffers_t buf, const contexts_t ctx,
+                      const constraint_t target, paramValues_t paramTarget,
+                      const ZSTD_strategy strat,
+                      memoTable_t* memoTableArray, const int tries)
+{
     int i = 0;
 
     paramValues_t init;
@@ -2114,9 +2289,11 @@ static winnerInfo_t optimizeFixedStrategy(
 
     for(i = 0; i < tries; i++) {
         DEBUGOUTPUT("Restart\n");
-        do { randomConstrainedParams(&init, memoTableArray, strat); } while(redundantParams(init, target, buf.maxBlockSize));
+        do {
+            randomConstrainedParams(&init, memoTableArray, strat);
+        } while(redundantParams(init, target, buf.maxBlockSize));
         candidateInfo = climbOnce(target, memoTableArray, buf, ctx, init);
-        if(compareResultLT(winnerInfo.result, candidateInfo.result, target, buf.srcSize)) {
+        if (compareResultLT(winnerInfo.result, candidateInfo.result, target, buf.srcSize)) {
             winnerInfo = candidateInfo;
             BMK_printWinnerOpt(stdout, CUSTOM_LEVEL, winnerInfo.result, winnerInfo.params, target, buf.srcSize);
             i = 0;
@@ -2130,12 +2307,13 @@ static winnerInfo_t optimizeFixedStrategy(
 
 /* goes best, best-1, best+1, best-2, ... */
 /* return 0 if nothing remaining */
-static int nextStrategy(const int currentStrategy, const int bestStrategy) {
+static int nextStrategy(const int currentStrategy, const int bestStrategy)
+{
     if(bestStrategy <= currentStrategy) {
         int candidate = 2 * bestStrategy - currentStrategy - 1;
         if(candidate < 1) {
             candidate = currentStrategy + 1;
-            if(candidate > (int)ZSTD_btultra) {
+            if(candidate > (int)ZSTD_STRATEGY_MAX) {
                 return 0;
             } else {
                 return candidate;
@@ -2145,7 +2323,7 @@ static int nextStrategy(const int currentStrategy, const int bestStrategy) {
         }
     } else { /* bestStrategy >= currentStrategy */
         int candidate = 2 * bestStrategy - currentStrategy;
-        if(candidate > (int)ZSTD_btultra) {
+        if(candidate > (int)ZSTD_STRATEGY_MAX) {
             candidate = currentStrategy - 1;
             if(candidate < 1) {
                 return 0;
@@ -2176,8 +2354,12 @@ static int nextStrategy(const int currentStrategy, const int bestStrategy) {
 static int g_maxTries = 5;
 #define TRY_DECAY 1
 
-static int optimizeForSize(const char* const * const fileNamesTable, const size_t nbFiles, const char* dictFileName, constraint_t target, paramValues_t paramTarget,
-    const int cLevelOpt, const int cLevelRun, const U32 memoTableLog)
+static int
+optimizeForSize(const char* const * const fileNamesTable, const size_t nbFiles,
+                const char* dictFileName,
+                constraint_t target, paramValues_t paramTarget,
+                const int cLevelOpt, const int cLevelRun,
+                const U32 memoTableLog)
 {
     varInds_t varArray [NUM_PARAMS];
     int ret = 0;
@@ -2189,18 +2371,18 @@ static int optimizeForSize(const char* const * const fileNamesTable, const size_
     buffers_t buf;
     g_time = UTIL_getTime();
 
-    if(createBuffers(&buf, fileNamesTable, nbFiles)) {
+    if (createBuffers(&buf, fileNamesTable, nbFiles)) {
         DISPLAY("unable to load files\n");
         return 1;
     }
 
-    if(createContexts(&ctx, dictFileName)) {
+    if (createContexts(&ctx, dictFileName)) {
         DISPLAY("unable to load dictionary\n");
         freeBuffers(buf);
         return 2;
     }
 
-    if(nbFiles == 1) {
+    if (nbFiles == 1) {
         DISPLAYLEVEL(2, "Loading %s...       \r", fileNamesTable[0]);
     } else {
         DISPLAYLEVEL(2, "Loading %lu Files...       \r", (unsigned long)nbFiles);
@@ -2274,21 +2456,19 @@ static int optimizeForSize(const char* const * const fileNamesTable, const size_
         DISPLAYLEVEL(2, "optimizing for %lu Files", (unsigned long)nbFiles);
     }
 
-    if(target.cSpeed != 0) { DISPLAYLEVEL(2," - limit compression speed %u MB/s", target.cSpeed >> 20); }
-    if(target.dSpeed != 0) { DISPLAYLEVEL(2, " - limit decompression speed %u MB/s", target.dSpeed >> 20); }
-    if(target.cMem != (U32)-1) { DISPLAYLEVEL(2, " - limit memory %u MB", target.cMem >> 20); }
+    if(target.cSpeed != 0) { DISPLAYLEVEL(2," - limit compression speed %u MB/s", (unsigned)(target.cSpeed >> 20)); }
+    if(target.dSpeed != 0) { DISPLAYLEVEL(2, " - limit decompression speed %u MB/s", (unsigned)(target.dSpeed >> 20)); }
+    if(target.cMem != (U32)-1) { DISPLAYLEVEL(2, " - limit memory %u MB", (unsigned)(target.cMem >> 20)); }
 
     DISPLAYLEVEL(2, "\n");
-    findClockGranularity();
+    init_clockGranularity();
 
     {   paramValues_t CParams;
 
         /* find best solution from default params */
-        {
-            /* strategy selection */
-            const int maxSeeds = g_noSeed ? 1 : ZSTD_maxCLevel();
+        {   const int maxSeeds = g_noSeed ? 1 : ZSTD_maxCLevel();
             DEBUGOUTPUT("Strategy Selection\n");
-            if(paramTarget.vals[strt_ind] == PARAM_UNSET) {
+            if (paramTarget.vals[strt_ind] == PARAM_UNSET) {
                 BMK_benchResult_t candidate;
                 int i;
                 for (i=1; i<=maxSeeds; i++) {
@@ -2313,16 +2493,14 @@ static int optimizeForSize(const char* const * const fileNamesTable, const size_
 
         DEBUGOUTPUT("Real Opt\n");
         /* start 'real' optimization */
-        {
-            int bestStrategy = (int)winner.params.vals[strt_ind];
-            if(paramTarget.vals[strt_ind] == PARAM_UNSET) {
+        {   int bestStrategy = (int)winner.params.vals[strt_ind];
+            if (paramTarget.vals[strt_ind] == PARAM_UNSET) {
                 int st = bestStrategy;
                 int tries = g_maxTries;
 
-                {
-                    /* one iterations of hill climbing with the level-defined parameters. */
-                    winnerInfo_t w1 = climbOnce(target, allMT, buf, ctx, winner.params);
-                    if(compareResultLT(winner.result, w1.result, target, buf.srcSize)) {
+                /* one iterations of hill climbing with the level-defined parameters. */
+                {   winnerInfo_t const w1 = climbOnce(target, allMT, buf, ctx, winner.params);
+                    if (compareResultLT(winner.result, w1.result, target, buf.srcSize)) {
                         winner = w1;
                     }
                     CHECKTIMEGT(ret, 0, _displayCleanUp);
@@ -2356,13 +2534,16 @@ static int optimizeForSize(const char* const * const fileNamesTable, const size_
             DISPLAY("No feasible solution found\n");
             goto _cleanUp;
         }
+
         /* end summary */
 _displayCleanUp:
-        if(g_displayLevel >= 0) { BMK_displayOneResult(stdout, winner, buf.srcSize); }
-        BMK_translateAdvancedParams(stdout, winner.params);
+        if (g_displayLevel >= 0) {
+            BMK_displayOneResult(stdout, winner, buf.srcSize);
+        }
+        BMK_paramValues_into_commandLine(stdout, winner.params);
         DISPLAYLEVEL(1, "grillParams size - optimizer completed \n");
-
     }
+
 _cleanUp:
     freeContexts(ctx);
     freeBuffers(buf);
@@ -2459,7 +2640,8 @@ static int usage_advanced(void)
     DISPLAY( " -S           : Single run \n");
     DISPLAY( " --zstd       : Single run, parameter selection same as zstdcli \n");
     DISPLAY( " -P#          : generated sample compressibility (default : %.1f%%) \n", COMPRESSIBILITY_DEFAULT * 100);
-    DISPLAY( " -t#          : Caps runtime of operation in seconds (default : %u seconds (%.1f hours)) \n", g_timeLimit_s, (double)g_timeLimit_s / 3600);
+    DISPLAY( " -t#          : Caps runtime of operation in seconds (default : %u seconds (%.1f hours)) \n",
+                                (unsigned)g_timeLimit_s, (double)g_timeLimit_s / 3600);
     DISPLAY( " -v           : Prints Benchmarking output\n");
     DISPLAY( " -D           : Next argument dictionary file\n");
     DISPLAY( " -s           : Seperate Files\n");
@@ -2473,14 +2655,23 @@ static int badusage(const char* exename)
     return 1;
 }
 
-#define PARSE_SUB_ARGS(stringLong, stringShort, variable) { if (longCommandWArg(&argument, stringLong) || longCommandWArg(&argument, stringShort)) { variable = readU32FromChar(&argument); if (argument[0]==',') { argument++; continue; } else break; } }
+#define PARSE_SUB_ARGS(stringLong, stringShort, variable) { \
+    if ( longCommandWArg(&argument, stringLong)             \
+      || longCommandWArg(&argument, stringShort) ) {        \
+          variable = readU32FromChar(&argument);            \
+          if (argument[0]==',') {                           \
+              argument++; continue;                         \
+          } else break;                                     \
+}   }
+
 /* 1 if successful parse, 0 otherwise */
 static int parse_params(const char** argptr, paramValues_t* pv) {
     int matched = 0;
     const char* argOrig = *argptr;
     varInds_t v;
     for(v = 0; v < NUM_PARAMS; v++) {
-        if(longCommandWArg(argptr,g_shortParamNames[v]) || longCommandWArg(argptr, g_paramNames[v])) {
+        if ( longCommandWArg(argptr,g_shortParamNames[v])
+          || longCommandWArg(argptr, g_paramNames[v]) ) {
             if(**argptr == '=') {
                 (*argptr)++;
                 pv->vals[v] = readU32FromChar(argptr);
@@ -2648,7 +2839,7 @@ int main(int argc, const char** argv)
                             continue;
                         case 'l':  /* search length */
                             argument++;
-                            g_params.vals[slen_ind] = readU32FromChar(&argument);
+                            g_params.vals[mml_ind] = readU32FromChar(&argument);
                             continue;
                         case 't':  /* target length */
                             argument++;
@@ -2685,7 +2876,7 @@ int main(int argc, const char** argv)
                 case 'B':
                     argument++;
                     g_blockSize = readU32FromChar(&argument);
-                    DISPLAY("using %u KB block size \n", g_blockSize>>10);
+                    DISPLAY("using %u KB block size \n", (unsigned)(g_blockSize>>10));
                     break;
 
                     /* caps runtime (in seconds) */
diff --git a/tests/playTests.sh b/tests/playTests.sh
index b86a0dc40cb2..ef4861b52917 100755
--- a/tests/playTests.sh
+++ b/tests/playTests.sh
@@ -122,6 +122,18 @@ $ZSTD --fast=5000000000 -f tmp && die "too large numeric value : must fail"
 $ZSTD -c --fast=0 tmp > $INTOVOID && die "--fast must not accept value 0"
 $ECHO "test : too large numeric argument"
 $ZSTD --fast=9999999999 -f tmp  && die "should have refused numeric value"
+$ECHO "test : set compression level with environment variable ZSTD_CLEVEL"
+ZSTD_CLEVEL=12  $ZSTD -f tmp # positive compression level
+ZSTD_CLEVEL=-12 $ZSTD -f tmp # negative compression level
+ZSTD_CLEVEL=+12 $ZSTD -f tmp # valid: verbose '+' sign
+ZSTD_CLEVEL=    $ZSTD -f tmp # empty env var, warn and revert to default setting
+ZSTD_CLEVEL=-   $ZSTD -f tmp # malformed env var, warn and revert to default setting
+ZSTD_CLEVEL=a   $ZSTD -f tmp # malformed env var, warn and revert to default setting
+ZSTD_CLEVEL=+a  $ZSTD -f tmp # malformed env var, warn and revert to default setting
+ZSTD_CLEVEL=3a7 $ZSTD -f tmp # malformed env var, warn and revert to default setting
+ZSTD_CLEVEL=50000000000  $ZSTD -f tmp # numeric value too large, warn and revert to default setting
+$ECHO "test : override ZSTD_CLEVEL with command line option"
+ZSTD_CLEVEL=12  $ZSTD --fast=3 -f tmp # overridden by command line option
 $ECHO "test : compress to stdout"
 $ZSTD tmp -c > tmpCompressed
 $ZSTD tmp --stdout > tmpCompressed       # long command format
@@ -179,8 +191,15 @@ $ECHO foo > tmpro
 chmod 400 tmpro.zst
 $ZSTD -q tmpro && die "should have refused to overwrite read-only file"
 $ZSTD -q -f tmpro
+$ECHO "test: --no-progress flag"
+$ZSTD tmpro -c --no-progress | $ZSTD -d -f -o "$INTOVOID" --no-progress
+$ZSTD tmpro -cv --no-progress | $ZSTD -dv -f -o "$INTOVOID" --no-progress
 rm -f tmpro tmpro.zst
-
+$ECHO "test: overwrite input file (must fail)"
+$ZSTD tmp -fo tmp && die "zstd compression overwrote the input file"
+$ZSTD tmp.zst -dfo tmp.zst && die "zstd decompression overwrote the input file"
+$ECHO "test: detect that input file does not exist"
+$ZSTD nothere && die "zstd hasn't detected that input file does not exist"
 
 $ECHO "test : file removal"
 $ZSTD -f --rm tmp
@@ -213,7 +232,7 @@ rm tmp*
 $ECHO "test : compress multiple files"
 $ECHO hello > tmp1
 $ECHO world > tmp2
-$ZSTD tmp1 tmp2 -o "$INTOVOID"
+$ZSTD tmp1 tmp2 -o "$INTOVOID" -f
 $ZSTD tmp1 tmp2 -c | $ZSTD -t
 $ZSTD tmp1 tmp2 -o tmp.zst
 test ! -f tmp1.zst
@@ -221,7 +240,7 @@ test ! -f tmp2.zst
 $ZSTD tmp1 tmp2
 $ZSTD -t tmp1.zst tmp2.zst
 $ZSTD -dc tmp1.zst tmp2.zst
-$ZSTD tmp1.zst tmp2.zst -o "$INTOVOID"
+$ZSTD tmp1.zst tmp2.zst -o "$INTOVOID" -f
 $ZSTD -d tmp1.zst tmp2.zst -o tmp
 touch tmpexists
 $ZSTD tmp1 tmp2 -f -o tmpexists
@@ -237,14 +256,15 @@ $ECHO "\n===>  Advanced compression parameters "
 $ECHO "Hello world!" | $ZSTD --zstd=windowLog=21,      - -o tmp.zst && die "wrong parameters not detected!"
 $ECHO "Hello world!" | $ZSTD --zstd=windowLo=21        - -o tmp.zst && die "wrong parameters not detected!"
 $ECHO "Hello world!" | $ZSTD --zstd=windowLog=21,slog  - -o tmp.zst && die "wrong parameters not detected!"
+$ECHO "Hello world!" | $ZSTD --zstd=strategy=10        - -o tmp.zst && die "parameter out of bound not detected!"  # > btultra2 : does not exist
 test ! -f tmp.zst  # tmp.zst should not be created
 roundTripTest -g512K
-roundTripTest -g512K " --zstd=slen=3,tlen=48,strat=6"
+roundTripTest -g512K " --zstd=mml=3,tlen=48,strat=6"
 roundTripTest -g512K " --zstd=strat=6,wlog=23,clog=23,hlog=22,slog=6"
-roundTripTest -g512K " --zstd=windowLog=23,chainLog=23,hashLog=22,searchLog=6,searchLength=3,targetLength=48,strategy=6"
-roundTripTest -g512K " --single-thread --long --zstd=ldmHashLog=20,ldmSearchLength=64,ldmBucketSizeLog=1,ldmHashEveryLog=7"
-roundTripTest -g512K " --single-thread --long --zstd=ldmhlog=20,ldmslen=64,ldmblog=1,ldmhevery=7"
-roundTripTest -g512K 19
+roundTripTest -g512K " --zstd=windowLog=23,chainLog=23,hashLog=22,searchLog=6,minMatch=3,targetLength=48,strategy=6"
+roundTripTest -g512K " --single-thread --long --zstd=ldmHashLog=20,ldmMinMatch=64,ldmBucketSizeLog=1,ldmHashRateLog=7"
+roundTripTest -g512K " --single-thread --long --zstd=lhlog=20,lmml=64,lblog=1,lhrlog=7"
+roundTripTest -g64K  "19 --zstd=strat=9"   # btultra2
 
 
 $ECHO "\n===>  Pass-Through mode "
@@ -541,6 +561,9 @@ $ECHO "bench negative level"
 $ZSTD -bi0 --fast tmp1
 $ECHO "with recursive and quiet modes"
 $ZSTD -rqi1b1e2 tmp1
+$ECHO "benchmark decompression only"
+$ZSTD -f tmp1
+$ZSTD -b -d -i1 tmp1.zst
 
 $ECHO "\n===>  zstd compatibility tests "
 
@@ -744,17 +767,17 @@ then
     ./datagen -g2MB > tmp
     refSize=$($ZSTD tmp -6 -c --zstd=wlog=18         | wc -c)
     ov9Size=$($ZSTD tmp -6 -c --zstd=wlog=18,ovlog=9 | wc -c)
-    ov0Size=$($ZSTD tmp -6 -c --zstd=wlog=18,ovlog=0 | wc -c)
+    ov1Size=$($ZSTD tmp -6 -c --zstd=wlog=18,ovlog=1 | wc -c)
     if [ $refSize -eq $ov9Size ]; then
         echo ov9Size should be different from refSize
         exit 1
     fi
-    if [ $refSize -eq $ov0Size ]; then
-        echo ov0Size should be different from refSize
+    if [ $refSize -eq $ov1Size ]; then
+        echo ov1Size should be different from refSize
         exit 1
     fi
-    if [ $ov9Size -ge $ov0Size ]; then
-        echo ov9Size=$ov9Size should be smaller than ov0Size=$ov0Size
+    if [ $ov9Size -ge $ov1Size ]; then
+        echo ov9Size=$ov9Size should be smaller than ov1Size=$ov1Size
         exit 1
     fi
 
@@ -833,6 +856,12 @@ $ECHO "===>   test: --adapt must fail on incoherent bounds "
 ./datagen > tmp
 $ZSTD -f -vv --adapt=min=10,max=9 tmp && die "--adapt must fail on incoherent bounds"
 
+$ECHO "\n===>   rsyncable mode "
+roundTripTest -g10M " --rsyncable"
+roundTripTest -g10M " --rsyncable -B100K"
+$ECHO "===>   test: --rsyncable must fail with --single-thread"
+$ZSTD -f -vv --rsyncable --single-thread tmp && die "--rsyncable must fail with --single-thread"
+
 
 if [ "$1" != "--test-large-data" ]; then
     $ECHO "Skipping large data tests"
diff --git a/tests/poolTests.c b/tests/poolTests.c
index 9661b5299e52..8b9a4700229d 100644
--- a/tests/poolTests.c
+++ b/tests/poolTests.c
@@ -30,7 +30,7 @@ struct data {
   size_t i;
 };
 
-void fn(void *opaque) {
+static void fn(void *opaque) {
   struct data *data = (struct data *)opaque;
   ZSTD_pthread_mutex_lock(&data->mutex);
   data->data[data->i] = data->i;
@@ -38,7 +38,7 @@ void fn(void *opaque) {
   ZSTD_pthread_mutex_unlock(&data->mutex);
 }
 
-int testOrder(size_t numThreads, size_t queueSize) {
+static int testOrder(size_t numThreads, size_t queueSize) {
   struct data data;
   POOL_ctx *ctx = POOL_create(numThreads, queueSize);
   ASSERT_TRUE(ctx);
@@ -63,13 +63,13 @@ int testOrder(size_t numThreads, size_t queueSize) {
 
 /* --- test deadlocks --- */
 
-void waitFn(void *opaque) {
+static void waitFn(void *opaque) {
   (void)opaque;
   UTIL_sleepMilli(1);
 }
 
 /* Tests for deadlock */
-int testWait(size_t numThreads, size_t queueSize) {
+static int testWait(size_t numThreads, size_t queueSize) {
   struct data data;
   POOL_ctx *ctx = POOL_create(numThreads, queueSize);
   ASSERT_TRUE(ctx);
@@ -92,7 +92,7 @@ typedef struct {
     ZSTD_pthread_cond_t cond;
 } poolTest_t;
 
-void waitLongFn(void *opaque) {
+static void waitLongFn(void *opaque) {
   poolTest_t* test = (poolTest_t*) opaque;
   UTIL_sleepMilli(10);
   ZSTD_pthread_mutex_lock(&test->mut);
@@ -167,7 +167,7 @@ typedef struct {
     int val;
 } abruptEndCanary_t;
 
-void waitIncFn(void *opaque) {
+static void waitIncFn(void *opaque) {
   abruptEndCanary_t* test = (abruptEndCanary_t*) opaque;
   UTIL_sleepMilli(10);
   ZSTD_pthread_mutex_lock(&test->mut);
diff --git a/tests/regression/Makefile b/tests/regression/Makefile
new file mode 100644
index 000000000000..03e5b0af1d70
--- /dev/null
+++ b/tests/regression/Makefile
@@ -0,0 +1,58 @@
+# ################################################################
+# Copyright (c) 2015-present, Facebook, Inc.
+# All rights reserved.
+#
+# This source code is licensed under both the BSD-style license (found in the
+# LICENSE file in the root directory of this source tree) and the GPLv2 (found
+# in the COPYING file in the root directory of this source tree).
+# ################################################################
+
+CFLAGS ?= -O3
+
+CURL_CFLAGS := $(shell curl-config --cflags)
+CURL_LDFLAGS := $(shell curl-config --libs) -pthread
+
+PROGDIR := ../../programs
+LIBDIR := ../../lib
+ZSTD_CPPFLAGS := -I$(PROGDIR) -I$(LIBDIR) -I$(LIBDIR)/common
+
+REGRESSION_CFLAGS = $(CFLAGS) $(CURL_CFLAGS)
+REGRESSION_CPPFLAGS = $(CPPFLAGS) $(ZSTD_CPPFLAGS)
+REGRESSION_LDFLAGS = $(LDFLAGS) $(CURL_LDFLAGS)
+
+all: test
+
+xxhash.o: $(LIBDIR)/common/xxhash.c $(LIBDIR)/common/xxhash.h
+	$(CC) $(REGRESSION_CFLAGS) $(REGRESSION_CPPFLAGS) $< -c -o $@
+
+util.o: $(PROGDIR)/util.c $(PROGDIR)/util.h
+	$(CC) $(REGRESSION_CFLAGS) $(REGRESSION_CPPFLAGS) $< -c -o $@
+
+data.o: data.c data.h $(PROGDIR)/util.h $(LIBDIR)/common/xxhash.h
+	$(CC) $(REGRESSION_CFLAGS) $(REGRESSION_CPPFLAGS) $< -c -o $@
+
+config.o: config.c config.h levels.h
+	$(CC) $(REGRESSION_CFLAGS) $(REGRESSION_CPPFLAGS) $< -c -o $@
+
+method.h: data.h config.h result.h
+
+method.o: method.c method.h
+	$(CC) $(REGRESSION_CFLAGS) $(REGRESSION_CPPFLAGS) $< -c -o $@
+
+result.o: result.c result.h
+	$(CC) $(REGRESSION_CFLAGS) $(REGRESSION_CPPFLAGS) $< -c -o $@
+
+test.o: test.c data.h config.h method.h
+	$(CC) $(REGRESSION_CFLAGS) $(REGRESSION_CPPFLAGS) $< -c -o $@
+
+libzstd.a:
+	$(MAKE) -C $(LIBDIR) libzstd.a-mt
+	cp $(LIBDIR)/libzstd.a .
+
+test: test.o data.o config.o util.o method.o result.o xxhash.o libzstd.a
+	$(CC) $^ $(REGRESSION_LDFLAGS) -o $@
+
+.PHONY: clean
+clean:
+	$(MAKE) -C $(LIBDIR) clean
+	$(RM) *.o *.a test
diff --git a/tests/regression/config.c b/tests/regression/config.c
new file mode 100644
index 000000000000..8a6103b2929e
--- /dev/null
+++ b/tests/regression/config.c
@@ -0,0 +1,230 @@
+/*
+ * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
+ * You may select, at your option, one of the above-listed licenses.
+ */
+
+#include "config.h"
+
+/* Define a config for each fast level we want to test with. */
+#define FAST_LEVEL(x)                                               \
+    param_value_t const level_fast##x##_param_values[] = {          \
+        {.param = ZSTD_c_compressionLevel, .value = -x},            \
+    };                                                              \
+    config_t const level_fast##x = {                                \
+        .name = "level -" #x,                                       \
+        .cli_args = "--fast=" #x,                                   \
+        .param_values = PARAM_VALUES(level_fast##x##_param_values), \
+    };                                                              \
+    config_t const level_fast##x##_dict = {                         \
+        .name = "level -" #x " with dict",                          \
+        .cli_args = "--fast=" #x,                                   \
+        .param_values = PARAM_VALUES(level_fast##x##_param_values), \
+        .use_dictionary = 1,                                        \
+    };
+
+/* Define a config for each level we want to test with. */
+#define LEVEL(x)                                                \
+    param_value_t const level_##x##_param_values[] = {          \
+        {.param = ZSTD_c_compressionLevel, .value = x},         \
+    };                                                          \
+    config_t const level_##x = {                                \
+        .name = "level " #x,                                    \
+        .cli_args = "-" #x,                                     \
+        .param_values = PARAM_VALUES(level_##x##_param_values), \
+    };                                                          \
+    config_t const level_##x##_dict = {                         \
+        .name = "level " #x " with dict",                       \
+        .cli_args = "-" #x,                                     \
+        .param_values = PARAM_VALUES(level_##x##_param_values), \
+        .use_dictionary = 1,                                    \
+    };
+
+#define PARAM_VALUES(pv) \
+    { .data = pv, .size = sizeof(pv) / sizeof((pv)[0]) }
+
+#include "levels.h"
+
+#undef LEVEL
+#undef FAST_LEVEL
+
+static config_t no_pledged_src_size = {
+    .name = "no source size",
+    .cli_args = "",
+    .param_values = PARAM_VALUES(level_0_param_values),
+    .no_pledged_src_size = 1,
+};
+
+static param_value_t const ldm_param_values[] = {
+    {.param = ZSTD_c_enableLongDistanceMatching, .value = 1},
+};
+
+static config_t ldm = {
+    .name = "long distance mode",
+    .cli_args = "--long",
+    .param_values = PARAM_VALUES(ldm_param_values),
+};
+
+static param_value_t const mt_param_values[] = {
+    {.param = ZSTD_c_nbWorkers, .value = 2},
+};
+
+static config_t mt = {
+    .name = "multithreaded",
+    .cli_args = "-T2",
+    .param_values = PARAM_VALUES(mt_param_values),
+};
+
+static param_value_t const mt_ldm_param_values[] = {
+    {.param = ZSTD_c_nbWorkers, .value = 2},
+    {.param = ZSTD_c_enableLongDistanceMatching, .value = 1},
+};
+
+static config_t mt_ldm = {
+    .name = "multithreaded long distance mode",
+    .cli_args = "-T2 --long",
+    .param_values = PARAM_VALUES(mt_ldm_param_values),
+};
+
+static param_value_t const small_wlog_param_values[] = {
+    {.param = ZSTD_c_windowLog, .value = 10},
+};
+
+static config_t small_wlog = {
+    .name = "small window log",
+    .cli_args = "--zstd=wlog=10",
+    .param_values = PARAM_VALUES(small_wlog_param_values),
+};
+
+static param_value_t const small_hlog_param_values[] = {
+    {.param = ZSTD_c_hashLog, .value = 6},
+    {.param = ZSTD_c_strategy, .value = (int)ZSTD_btopt},
+};
+
+static config_t small_hlog = {
+    .name = "small hash log",
+    .cli_args = "--zstd=hlog=6,strat=7",
+    .param_values = PARAM_VALUES(small_hlog_param_values),
+};
+
+static param_value_t const small_clog_param_values[] = {
+    {.param = ZSTD_c_chainLog, .value = 6},
+    {.param = ZSTD_c_strategy, .value = (int)ZSTD_btopt},
+};
+
+static config_t small_clog = {
+    .name = "small chain log",
+    .cli_args = "--zstd=clog=6,strat=7",
+    .param_values = PARAM_VALUES(small_clog_param_values),
+};
+
+static param_value_t const explicit_params_param_values[] = {
+    {.param = ZSTD_c_checksumFlag, .value = 1},
+    {.param = ZSTD_c_contentSizeFlag, .value = 0},
+    {.param = ZSTD_c_dictIDFlag, .value = 0},
+    {.param = ZSTD_c_strategy, .value = (int)ZSTD_greedy},
+    {.param = ZSTD_c_windowLog, .value = 18},
+    {.param = ZSTD_c_hashLog, .value = 21},
+    {.param = ZSTD_c_chainLog, .value = 21},
+    {.param = ZSTD_c_targetLength, .value = 100},
+};
+
+static config_t explicit_params = {
+    .name = "explicit params",
+    .cli_args = "--no-check --no-dictID --zstd=strategy=3,wlog=18,hlog=21,clog=21,tlen=100",
+    .param_values = PARAM_VALUES(explicit_params_param_values),
+};
+
+static config_t const* g_configs[] = {
+
+#define FAST_LEVEL(x) &level_fast##x, &level_fast##x##_dict,
+#define LEVEL(x) &level_##x, &level_##x##_dict,
+#include "levels.h"
+#undef LEVEL
+#undef FAST_LEVEL
+
+    &no_pledged_src_size,
+    &ldm,
+    &mt,
+    &mt_ldm,
+    &small_wlog,
+    &small_hlog,
+    &small_clog,
+    &explicit_params,
+    NULL,
+};
+
+config_t const* const* configs = g_configs;
+
+int config_skip_data(config_t const* config, data_t const* data) {
+    return config->use_dictionary && !data_has_dict(data);
+}
+
+int config_get_level(config_t const* config)
+{
+    param_values_t const params = config->param_values;
+    size_t i;
+    for (i = 0; i < params.size; ++i) {
+        if (params.data[i].param == ZSTD_c_compressionLevel)
+            return (int)params.data[i].value;
+    }
+    return CONFIG_NO_LEVEL;
+}
+
+ZSTD_parameters config_get_zstd_params(
+    config_t const* config,
+    uint64_t srcSize,
+    size_t dictSize)
+{
+    ZSTD_parameters zparams = {};
+    param_values_t const params = config->param_values;
+    int level = config_get_level(config);
+    if (level == CONFIG_NO_LEVEL)
+        level = 3;
+    zparams = ZSTD_getParams(
+        level,
+        config->no_pledged_src_size ? ZSTD_CONTENTSIZE_UNKNOWN : srcSize,
+        dictSize);
+    for (size_t i = 0; i < params.size; ++i) {
+        unsigned const value = params.data[i].value;
+        switch (params.data[i].param) {
+            case ZSTD_c_contentSizeFlag:
+                zparams.fParams.contentSizeFlag = value;
+                break;
+            case ZSTD_c_checksumFlag:
+                zparams.fParams.checksumFlag = value;
+                break;
+            case ZSTD_c_dictIDFlag:
+                zparams.fParams.noDictIDFlag = !value;
+                break;
+            case ZSTD_c_windowLog:
+                zparams.cParams.windowLog = value;
+                break;
+            case ZSTD_c_chainLog:
+                zparams.cParams.chainLog = value;
+                break;
+            case ZSTD_c_hashLog:
+                zparams.cParams.hashLog = value;
+                break;
+            case ZSTD_c_searchLog:
+                zparams.cParams.searchLog = value;
+                break;
+            case ZSTD_c_minMatch:
+                zparams.cParams.minMatch = value;
+                break;
+            case ZSTD_c_targetLength:
+                zparams.cParams.targetLength = value;
+                break;
+            case ZSTD_c_strategy:
+                zparams.cParams.strategy = (ZSTD_strategy)value;
+                break;
+            default:
+                break;
+        }
+    }
+    return zparams;
+}
diff --git a/tests/regression/config.h b/tests/regression/config.h
new file mode 100644
index 000000000000..3cd0308a0987
--- /dev/null
+++ b/tests/regression/config.h
@@ -0,0 +1,86 @@
+/*
+ * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
+ * You may select, at your option, one of the above-listed licenses.
+ */
+
+#ifndef CONFIG_H
+#define CONFIG_H
+
+#include 
+
+#define ZSTD_STATIC_LINKING_ONLY
+#include 
+
+#include "data.h"
+
+typedef struct {
+    ZSTD_cParameter param;
+    int value;
+} param_value_t;
+
+typedef struct {
+    size_t size;
+    param_value_t const* data;
+} param_values_t;
+
+/**
+ * The config tells the compression method what options to use.
+ */
+typedef struct {
+    const char* name;  /**< Identifies the config in the results table */
+    /**
+     * Optional arguments to pass to the CLI. If not set, CLI-based methods
+     * will skip this config.
+     */
+    char const* cli_args;
+    /**
+     * Parameters to pass to the advanced API. If the advanced API isn't used,
+     * the parameters will be derived from these.
+     */
+    param_values_t param_values;
+    /**
+     * Boolean parameter that says if we should use a dictionary. If the data
+     * doesn't have a dictionary, this config is skipped. Defaults to no.
+     */
+    int use_dictionary;
+    /**
+     * Boolean parameter that says if we should pass the pledged source size
+     * when the method allows it. Defaults to yes.
+     */
+    int no_pledged_src_size;
+} config_t;
+
+/**
+ * Returns true if the config should skip this data.
+ * For instance, if the config requires a dictionary but the data doesn't have
+ * one.
+ */
+int config_skip_data(config_t const* config, data_t const* data);
+
+#define CONFIG_NO_LEVEL (-ZSTD_TARGETLENGTH_MAX - 1)
+/**
+ * Returns the compression level specified by the config, or CONFIG_NO_LEVEL if
+ * no level is specified. Note that 0 is a valid compression level, meaning
+ * default.
+ */
+int config_get_level(config_t const* config);
+
+/**
+ * Returns the compression parameters specified by the config.
+ */
+ZSTD_parameters config_get_zstd_params(
+    config_t const* config,
+    uint64_t srcSize,
+    size_t dictSize);
+
+/**
+ * The NULL-terminated list of configs.
+ */
+extern config_t const* const* configs;
+
+#endif
diff --git a/tests/regression/data.c b/tests/regression/data.c
new file mode 100644
index 000000000000..86e7687de5ee
--- /dev/null
+++ b/tests/regression/data.c
@@ -0,0 +1,617 @@
+/*
+ * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
+ * You may select, at your option, one of the above-listed licenses.
+ */
+
+#include "data.h"
+
+#include 
+#include 
+#include 
+#include 
+
+#include 
+
+#include 
+
+#include "mem.h"
+#include "util.h"
+#define XXH_STATIC_LINKING_ONLY
+#include "xxhash.h"
+
+/**
+ * Data objects
+ */
+
+#define REGRESSION_RELEASE(x) \
+    "https://github.com/facebook/zstd/releases/download/regression-data/" x
+
+data_t silesia = {
+    .name = "silesia",
+    .type = data_type_dir,
+    .data =
+        {
+            .url = REGRESSION_RELEASE("silesia.tar.zst"),
+            .xxhash64 = 0x48a199f92f93e977LL,
+        },
+};
+
+data_t silesia_tar = {
+    .name = "silesia.tar",
+    .type = data_type_file,
+    .data =
+        {
+            .url = REGRESSION_RELEASE("silesia.tar.zst"),
+            .xxhash64 = 0x48a199f92f93e977LL,
+        },
+};
+
+data_t github = {
+    .name = "github",
+    .type = data_type_dir,
+    .data =
+        {
+            .url = REGRESSION_RELEASE("github.tar.zst"),
+            .xxhash64 = 0xa9b1b44b020df292LL,
+        },
+    .dict =
+        {
+            .url = REGRESSION_RELEASE("github.dict.zst"),
+            .xxhash64 = 0x1eddc6f737d3cb53LL,
+
+        },
+};
+
+static data_t* g_data[] = {
+    &silesia,
+    &silesia_tar,
+    &github,
+    NULL,
+};
+
+data_t const* const* data = (data_t const* const*)g_data;
+
+/**
+ * data helpers.
+ */
+
+int data_has_dict(data_t const* data) {
+    return data->dict.url != NULL;
+}
+
+/**
+ * data buffer helper functions (documented in header).
+ */
+
+data_buffer_t data_buffer_create(size_t const capacity) {
+    data_buffer_t buffer = {};
+
+    buffer.data = (uint8_t*)malloc(capacity);
+    if (buffer.data == NULL)
+        return buffer;
+    buffer.capacity = capacity;
+    return buffer;
+}
+
+data_buffer_t data_buffer_read(char const* filename) {
+    data_buffer_t buffer = {};
+
+    uint64_t const size = UTIL_getFileSize(filename);
+    if (size == UTIL_FILESIZE_UNKNOWN) {
+        fprintf(stderr, "unknown size for %s\n", filename);
+        return buffer;
+    }
+
+    buffer.data = (uint8_t*)malloc(size);
+    if (buffer.data == NULL) {
+        fprintf(stderr, "malloc failed\n");
+        return buffer;
+    }
+    buffer.capacity = size;
+
+    FILE* file = fopen(filename, "rb");
+    if (file == NULL) {
+        fprintf(stderr, "file null\n");
+        goto err;
+    }
+    buffer.size = fread(buffer.data, 1, buffer.capacity, file);
+    fclose(file);
+    if (buffer.size != buffer.capacity) {
+        fprintf(stderr, "read %zu != %zu\n", buffer.size, buffer.capacity);
+        goto err;
+    }
+
+    return buffer;
+err:
+    free(buffer.data);
+    memset(&buffer, 0, sizeof(buffer));
+    return buffer;
+}
+
+data_buffer_t data_buffer_get_data(data_t const* data) {
+    data_buffer_t const kEmptyBuffer = {};
+
+    if (data->type != data_type_file)
+        return kEmptyBuffer;
+
+    return data_buffer_read(data->data.path);
+}
+
+data_buffer_t data_buffer_get_dict(data_t const* data) {
+    data_buffer_t const kEmptyBuffer = {};
+
+    if (!data_has_dict(data))
+        return kEmptyBuffer;
+
+    return data_buffer_read(data->dict.path);
+}
+
+int data_buffer_compare(data_buffer_t buffer1, data_buffer_t buffer2) {
+    size_t const size =
+        buffer1.size < buffer2.size ? buffer1.size : buffer2.size;
+    int const cmp = memcmp(buffer1.data, buffer2.data, size);
+    if (cmp != 0)
+        return cmp;
+    if (buffer1.size < buffer2.size)
+        return -1;
+    if (buffer1.size == buffer2.size)
+        return 0;
+    assert(buffer1.size > buffer2.size);
+    return 1;
+}
+
+void data_buffer_free(data_buffer_t buffer) {
+    free(buffer.data);
+}
+
+/**
+ * data filenames helpers.
+ */
+
+data_filenames_t data_filenames_get(data_t const* data) {
+    data_filenames_t filenames = {.buffer = NULL, .size = 0};
+    char const* path = data->data.path;
+
+    filenames.filenames = UTIL_createFileList(
+        &path,
+        1,
+        &filenames.buffer,
+        &filenames.size,
+        /* followLinks */ 0);
+    return filenames;
+}
+
+void data_filenames_free(data_filenames_t filenames) {
+    UTIL_freeFileList(filenames.filenames, filenames.buffer);
+}
+
+/**
+ * data buffers helpers.
+ */
+
+data_buffers_t data_buffers_get(data_t const* data) {
+    data_buffers_t buffers = {.size = 0};
+    data_filenames_t filenames = data_filenames_get(data);
+    if (filenames.size == 0)
+        return buffers;
+
+    data_buffer_t* buffersPtr =
+        (data_buffer_t*)malloc(filenames.size * sizeof(data_buffer_t));
+    if (buffersPtr == NULL)
+        return buffers;
+    buffers.buffers = (data_buffer_t const*)buffersPtr;
+    buffers.size = filenames.size;
+
+    for (size_t i = 0; i < filenames.size; ++i) {
+        buffersPtr[i] = data_buffer_read(filenames.filenames[i]);
+        if (buffersPtr[i].data == NULL) {
+            data_buffers_t const kEmptyBuffer = {};
+            data_buffers_free(buffers);
+            return kEmptyBuffer;
+        }
+    }
+
+    return buffers;
+}
+
+/**
+ * Frees the data buffers.
+ */
+void data_buffers_free(data_buffers_t buffers) {
+    free((data_buffer_t*)buffers.buffers);
+}
+
+/**
+ * Initialization and download functions.
+ */
+
+static char* g_data_dir = NULL;
+
+/* mkdir -p */
+static int ensure_directory_exists(char const* indir) {
+    char* const dir = strdup(indir);
+    char* end = dir;
+    int ret = 0;
+    if (dir == NULL) {
+        ret = EINVAL;
+        goto out;
+    }
+    do {
+        /* Find the next directory level. */
+        for (++end; *end != '\0' && *end != '/'; ++end)
+            ;
+        /* End the string there, make the directory, and restore the string. */
+        char const save = *end;
+        *end = '\0';
+        int const isdir = UTIL_isDirectory(dir);
+        ret = mkdir(dir, S_IRWXU);
+        *end = save;
+        /* Its okay if the directory already exists. */
+        if (ret == 0 || (errno == EEXIST && isdir))
+            continue;
+        ret = errno;
+        fprintf(stderr, "mkdir() failed\n");
+        goto out;
+    } while (*end != '\0');
+
+    ret = 0;
+out:
+    free(dir);
+    return ret;
+}
+
+/** Concatenate 3 strings into a new buffer. */
+static char* cat3(char const* str1, char const* str2, char const* str3) {
+    size_t const size1 = strlen(str1);
+    size_t const size2 = strlen(str2);
+    size_t const size3 = str3 == NULL ? 0 : strlen(str3);
+    size_t const size = size1 + size2 + size3 + 1;
+    char* const dst = (char*)malloc(size);
+    if (dst == NULL)
+        return NULL;
+    strcpy(dst, str1);
+    strcpy(dst + size1, str2);
+    if (str3 != NULL)
+        strcpy(dst + size1 + size2, str3);
+    assert(strlen(dst) == size1 + size2 + size3);
+    return dst;
+}
+
+static char* cat2(char const* str1, char const* str2) {
+    return cat3(str1, str2, NULL);
+}
+
+/**
+ * State needed by the curl callback.
+ * It takes data from curl, hashes it, and writes it to the file.
+ */
+typedef struct {
+    FILE* file;
+    XXH64_state_t xxhash64;
+    int error;
+} curl_data_t;
+
+/** Create the curl state. */
+static curl_data_t curl_data_create(
+    data_resource_t const* resource,
+    data_type_t type) {
+    curl_data_t cdata = {};
+
+    XXH64_reset(&cdata.xxhash64, 0);
+
+    assert(UTIL_isDirectory(g_data_dir));
+
+    if (type == data_type_file) {
+        /* Decompress the resource and store to the path. */
+        char* cmd = cat3("zstd -dqfo '", resource->path, "'");
+        if (cmd == NULL) {
+            cdata.error = ENOMEM;
+            return cdata;
+        }
+        cdata.file = popen(cmd, "w");
+        free(cmd);
+    } else {
+        /* Decompress and extract the resource to the cache directory. */
+        char* cmd = cat3("zstd -dc | tar -x -C '", g_data_dir, "'");
+        if (cmd == NULL) {
+            cdata.error = ENOMEM;
+            return cdata;
+        }
+        cdata.file = popen(cmd, "w");
+        free(cmd);
+    }
+    if (cdata.file == NULL) {
+        cdata.error = errno;
+    }
+
+    return cdata;
+}
+
+/** Free the curl state. */
+static int curl_data_free(curl_data_t cdata) {
+    return pclose(cdata.file);
+}
+
+/** curl callback. Updates the hash, and writes to the file. */
+static size_t curl_write(void* data, size_t size, size_t count, void* ptr) {
+    curl_data_t* cdata = (curl_data_t*)ptr;
+    size_t const written = fwrite(data, size, count, cdata->file);
+    XXH64_update(&cdata->xxhash64, data, written * size);
+    return written;
+}
+
+static int curl_download_resource(
+    CURL* curl,
+    data_resource_t const* resource,
+    data_type_t type) {
+    curl_data_t cdata;
+    /* Download the data. */
+    if (curl_easy_setopt(curl, CURLOPT_URL, resource->url) != 0)
+        return EINVAL;
+    if (curl_easy_setopt(curl, CURLOPT_WRITEDATA, &cdata) != 0)
+        return EINVAL;
+    cdata = curl_data_create(resource, type);
+    if (cdata.error != 0)
+        return cdata.error;
+    int const curl_err = curl_easy_perform(curl);
+    int const close_err = curl_data_free(cdata);
+    if (curl_err) {
+        fprintf(
+            stderr,
+            "downloading '%s' for '%s' failed\n",
+            resource->url,
+            resource->path);
+        return EIO;
+    }
+    if (close_err) {
+        fprintf(stderr, "writing data to '%s' failed\n", resource->path);
+        return EIO;
+    }
+    /* check that the file exists. */
+    if (type == data_type_file && !UTIL_isRegularFile(resource->path)) {
+        fprintf(stderr, "output file '%s' does not exist\n", resource->path);
+        return EIO;
+    }
+    if (type == data_type_dir && !UTIL_isDirectory(resource->path)) {
+        fprintf(
+            stderr, "output directory '%s' does not exist\n", resource->path);
+        return EIO;
+    }
+    /* Check that the hash matches. */
+    if (XXH64_digest(&cdata.xxhash64) != resource->xxhash64) {
+        fprintf(
+            stderr,
+            "checksum does not match: 0x%llxLL != 0x%llxLL\n",
+            (unsigned long long)XXH64_digest(&cdata.xxhash64),
+            (unsigned long long)resource->xxhash64);
+        return EINVAL;
+    }
+
+    return 0;
+}
+
+/** Download a single data object. */
+static int curl_download_datum(CURL* curl, data_t const* data) {
+    int ret;
+    ret = curl_download_resource(curl, &data->data, data->type);
+    if (ret != 0)
+        return ret;
+    if (data_has_dict(data)) {
+        ret = curl_download_resource(curl, &data->dict, data_type_file);
+        if (ret != 0)
+            return ret;
+    }
+    return ret;
+}
+
+/** Download all the data. */
+static int curl_download_data(data_t const* const* data) {
+    if (curl_global_init(CURL_GLOBAL_ALL) != 0)
+        return EFAULT;
+
+    curl_data_t cdata = {};
+    CURL* curl = curl_easy_init();
+    int err = EFAULT;
+
+    if (curl == NULL)
+        return EFAULT;
+
+    if (curl_easy_setopt(curl, CURLOPT_NOPROGRESS, 1L) != 0)
+        goto out;
+    if (curl_easy_setopt(curl, CURLOPT_FOLLOWLOCATION, 1L) != 0)
+        goto out;
+    if (curl_easy_setopt(curl, CURLOPT_WRITEFUNCTION, curl_write) != 0)
+        goto out;
+
+    assert(data != NULL);
+    for (; *data != NULL; ++data) {
+        if (curl_download_datum(curl, *data) != 0)
+            goto out;
+    }
+
+    err = 0;
+out:
+    curl_easy_cleanup(curl);
+    curl_global_cleanup();
+    return err;
+}
+
+/** Fill the path member variable of the data objects. */
+static int data_create_paths(data_t* const* data, char const* dir) {
+    size_t const dirlen = strlen(dir);
+    assert(data != NULL);
+    for (; *data != NULL; ++data) {
+        data_t* const datum = *data;
+        datum->data.path = cat3(dir, "/", datum->name);
+        if (datum->data.path == NULL)
+            return ENOMEM;
+        if (data_has_dict(datum)) {
+            datum->dict.path = cat2(datum->data.path, ".dict");
+            if (datum->dict.path == NULL)
+                return ENOMEM;
+        }
+    }
+    return 0;
+}
+
+/** Free the path member variable of the data objects. */
+static void data_free_paths(data_t* const* data) {
+    assert(data != NULL);
+    for (; *data != NULL; ++data) {
+        data_t* datum = *data;
+        free((void*)datum->data.path);
+        free((void*)datum->dict.path);
+        datum->data.path = NULL;
+        datum->dict.path = NULL;
+    }
+}
+
+static char const kStampName[] = "STAMP";
+
+static void xxh_update_le(XXH64_state_t* state, uint64_t data) {
+    if (!MEM_isLittleEndian())
+        data = MEM_swap64(data);
+    XXH64_update(state, &data, sizeof(data));
+}
+
+/** Hash the data to create the stamp. */
+static uint64_t stamp_hash(data_t const* const* data) {
+    XXH64_state_t state;
+
+    XXH64_reset(&state, 0);
+    assert(data != NULL);
+    for (; *data != NULL; ++data) {
+        data_t const* datum = *data;
+        /* We don't care about the URL that we fetch from. */
+        /* The path is derived from the name. */
+        XXH64_update(&state, datum->name, strlen(datum->name));
+        xxh_update_le(&state, datum->data.xxhash64);
+        xxh_update_le(&state, datum->dict.xxhash64);
+        xxh_update_le(&state, datum->type);
+    }
+    return XXH64_digest(&state);
+}
+
+/** Check if the stamp matches the stamp in the cache directory. */
+static int stamp_check(char const* dir, data_t const* const* data) {
+    char* stamp = cat3(dir, "/", kStampName);
+    uint64_t const expected = stamp_hash(data);
+    XXH64_canonical_t actual;
+    FILE* stampfile = NULL;
+    int matches = 0;
+
+    if (stamp == NULL)
+        goto out;
+    if (!UTIL_isRegularFile(stamp)) {
+        fprintf(stderr, "stamp does not exist: recreating the data cache\n");
+        goto out;
+    }
+
+    stampfile = fopen(stamp, "rb");
+    if (stampfile == NULL) {
+        fprintf(stderr, "could not open stamp: recreating the data cache\n");
+        goto out;
+    }
+
+    size_t b;
+    if ((b = fread(&actual, sizeof(actual), 1, stampfile)) != 1) {
+        fprintf(stderr, "invalid stamp: recreating the data cache\n");
+        goto out;
+    }
+
+    matches = (expected == XXH64_hashFromCanonical(&actual));
+    if (matches)
+        fprintf(stderr, "stamp matches: reusing the cached data\n");
+    else
+        fprintf(stderr, "stamp does not match: recreating the data cache\n");
+
+out:
+    free(stamp);
+    if (stampfile != NULL)
+        fclose(stampfile);
+    return matches;
+}
+
+/** On success write a new stamp, on failure delete the old stamp. */
+static int
+stamp_write(char const* dir, data_t const* const* data, int const data_err) {
+    char* stamp = cat3(dir, "/", kStampName);
+    FILE* stampfile = NULL;
+    int err = EIO;
+
+    if (stamp == NULL)
+        return ENOMEM;
+
+    if (data_err != 0) {
+        err = data_err;
+        goto out;
+    }
+    XXH64_canonical_t hash;
+
+    XXH64_canonicalFromHash(&hash, stamp_hash(data));
+
+    stampfile = fopen(stamp, "wb");
+    if (stampfile == NULL)
+        goto out;
+    if (fwrite(&hash, sizeof(hash), 1, stampfile) != 1)
+        goto out;
+    err = 0;
+    fprintf(stderr, "stamped new data cache\n");
+out:
+    if (err != 0)
+        /* Ignore errors. */
+        unlink(stamp);
+    free(stamp);
+    if (stampfile != NULL)
+        fclose(stampfile);
+    return err;
+}
+
+int data_init(char const* dir) {
+    int err;
+
+    if (dir == NULL)
+        return EINVAL;
+
+    /* This must be first to simplify logic. */
+    err = ensure_directory_exists(dir);
+    if (err != 0)
+        return err;
+
+    /* Save the cache directory. */
+    g_data_dir = strdup(dir);
+    if (g_data_dir == NULL)
+        return ENOMEM;
+
+    err = data_create_paths(g_data, dir);
+    if (err != 0)
+        return err;
+
+    /* If the stamp matches then we are good to go.
+     * This must be called before any modifications to the data cache.
+     * After this point, we MUST call stamp_write() to update the STAMP,
+     * since we've updated the data cache.
+     */
+    if (stamp_check(dir, data))
+        return 0;
+
+    err = curl_download_data(data);
+    if (err != 0)
+        goto out;
+
+out:
+    /* This must be last, since it must know if data_init() succeeded. */
+    stamp_write(dir, data, err);
+    return err;
+}
+
+void data_finish(void) {
+    data_free_paths(g_data);
+    free(g_data_dir);
+    g_data_dir = NULL;
+}
diff --git a/tests/regression/data.h b/tests/regression/data.h
new file mode 100644
index 000000000000..717fe1294c14
--- /dev/null
+++ b/tests/regression/data.h
@@ -0,0 +1,140 @@
+/*
+ * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
+ * You may select, at your option, one of the above-listed licenses.
+ */
+
+#ifndef DATA_H
+#define DATA_H
+
+#include 
+#include 
+
+typedef enum {
+    data_type_file = 1,  /**< This data is a file. *.zst */
+    data_type_dir = 2,   /**< This data is a directory. *.tar.zst */
+} data_type_t;
+
+typedef struct {
+    char const* url;   /**< Where to get this resource. */
+    uint64_t xxhash64; /**< Hash of the url contents. */
+    char const* path;  /**< The path of the unpacked resource (derived). */
+} data_resource_t;
+
+typedef struct {
+    data_resource_t data;
+    data_resource_t dict;
+    data_type_t type;  /**< The type of the data. */
+    char const* name;  /**< The logical name of the data (no extension). */
+} data_t;
+
+/**
+ * The NULL-terminated list of data objects.
+ */
+extern data_t const* const* data;
+
+
+int data_has_dict(data_t const* data);
+
+/**
+ * Initializes the data module and downloads the data necessary.
+ * Caches the downloads in dir. We add a stamp file in the directory after
+ * a successful download. If a stamp file already exists, and matches our
+ * current data stamp, we will use the cached data without downloading.
+ *
+ * @param dir The directory to cache the downloaded data into.
+ *
+ * @returns 0 on success.
+ */
+int data_init(char const* dir);
+
+/**
+ * Must be called at exit to free resources allocated by data_init().
+ */
+void data_finish(void);
+
+typedef struct {
+    uint8_t* data;
+    size_t size;
+    size_t capacity;
+} data_buffer_t;
+
+/**
+ * Read the file that data points to into a buffer.
+ * NOTE: data must be a file, not a directory.
+ *
+ * @returns The buffer, which is NULL on failure.
+ */
+data_buffer_t data_buffer_get_data(data_t const* data);
+
+/**
+ * Read the dictionary that the data points to into a buffer.
+ *
+ * @returns The buffer, which is NULL on failure.
+ */
+data_buffer_t data_buffer_get_dict(data_t const* data);
+
+/**
+ * Read the contents of filename into a buffer.
+ *
+ * @returns The buffer, which is NULL on failure.
+ */
+data_buffer_t data_buffer_read(char const* filename);
+
+/**
+ * Create a buffer with the specified capacity.
+ *
+ * @returns The buffer, which is NULL on failure.
+ */
+data_buffer_t data_buffer_create(size_t capacity);
+
+/**
+ * Calls memcmp() on the contents [0, size) of both buffers.
+ */
+int data_buffer_compare(data_buffer_t buffer1, data_buffer_t buffer2);
+
+/**
+ * Frees an allocated buffer.
+ */
+void data_buffer_free(data_buffer_t buffer);
+
+typedef struct {
+    char* buffer;
+    char const** filenames;
+    unsigned size;
+} data_filenames_t;
+
+/**
+ * Get a recursive list of filenames in the data object. If it is a file, it
+ * will only contain one entry. If it is a directory, it will recursively walk
+ * the directory.
+ *
+ * @returns The list of filenames, which has size 0 and NULL pointers on error.
+ */
+data_filenames_t data_filenames_get(data_t const* data);
+
+/**
+ * Frees the filenames table.
+ */
+void data_filenames_free(data_filenames_t filenames);
+
+typedef struct {
+    data_buffer_t const* buffers;
+    size_t size;
+} data_buffers_t;
+
+/**
+ * @returns a list of buffers for every file in data. It is zero sized on error.
+ */
+data_buffers_t data_buffers_get(data_t const* data);
+
+/**
+ * Frees the data buffers.
+ */
+void data_buffers_free(data_buffers_t buffers);
+
+#endif
diff --git a/tests/regression/levels.h b/tests/regression/levels.h
new file mode 100644
index 000000000000..f96689075126
--- /dev/null
+++ b/tests/regression/levels.h
@@ -0,0 +1,44 @@
+/*
+ * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
+ * You may select, at your option, one of the above-listed licenses.
+ */
+
+#ifndef LEVEL
+# error LEVEL(x) must be defined
+#endif
+#ifndef FAST_LEVEL
+# error FAST_LEVEL(x) must be defined
+#endif
+
+/**
+ * The levels are chosen to trigger every strategy in every source size,
+ * as well as some fast levels and the default level.
+ * If you change the compression levels, you should probably update these.
+ */
+
+FAST_LEVEL(5)
+
+FAST_LEVEL(3)
+
+FAST_LEVEL(1)
+LEVEL(0)
+LEVEL(1)
+
+LEVEL(3)
+LEVEL(4)
+LEVEL(5)
+LEVEL(6)
+LEVEL(7)
+
+LEVEL(9)
+
+LEVEL(13)
+
+LEVEL(16)
+
+LEVEL(19)
diff --git a/tests/regression/method.c b/tests/regression/method.c
new file mode 100644
index 000000000000..51a0fb9d14f0
--- /dev/null
+++ b/tests/regression/method.c
@@ -0,0 +1,566 @@
+/*
+ * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
+ * You may select, at your option, one of the above-listed licenses.
+ */
+
+#include "method.h"
+
+#include 
+#include 
+
+#define ZSTD_STATIC_LINKING_ONLY
+#include 
+
+#define MIN(x, y) ((x) < (y) ? (x) : (y))
+
+static char const* g_zstdcli = NULL;
+
+void method_set_zstdcli(char const* zstdcli) {
+    g_zstdcli = zstdcli;
+}
+
+/**
+ * Macro to get a pointer of type, given ptr, which is a member variable with
+ * the given name, member.
+ *
+ *     method_state_t* base = ...;
+ *     buffer_state_t* state = container_of(base, buffer_state_t, base);
+ */
+#define container_of(ptr, type, member) \
+    ((type*)(ptr == NULL ? NULL : (char*)(ptr)-offsetof(type, member)))
+
+/** State to reuse the same buffers between compression calls. */
+typedef struct {
+    method_state_t base;
+    data_buffers_t inputs; /**< The input buffer for each file. */
+    data_buffer_t dictionary; /**< The dictionary. */
+    data_buffer_t compressed; /**< The compressed data buffer. */
+    data_buffer_t decompressed; /**< The decompressed data buffer. */
+} buffer_state_t;
+
+static size_t buffers_max_size(data_buffers_t buffers) {
+    size_t max = 0;
+    for (size_t i = 0; i < buffers.size; ++i) {
+        if (buffers.buffers[i].size > max)
+            max = buffers.buffers[i].size;
+    }
+    return max;
+}
+
+static method_state_t* buffer_state_create(data_t const* data) {
+    buffer_state_t* state = (buffer_state_t*)calloc(1, sizeof(buffer_state_t));
+    if (state == NULL)
+        return NULL;
+    state->base.data = data;
+    state->inputs = data_buffers_get(data);
+    state->dictionary = data_buffer_get_dict(data);
+    size_t const max_size = buffers_max_size(state->inputs);
+    state->compressed = data_buffer_create(ZSTD_compressBound(max_size));
+    state->decompressed = data_buffer_create(max_size);
+    return &state->base;
+}
+
+static void buffer_state_destroy(method_state_t* base) {
+    if (base == NULL)
+        return;
+    buffer_state_t* state = container_of(base, buffer_state_t, base);
+    free(state);
+}
+
+static int buffer_state_bad(
+    buffer_state_t const* state,
+    config_t const* config) {
+    if (state == NULL) {
+        fprintf(stderr, "buffer_state_t is NULL\n");
+        return 1;
+    }
+    if (state->inputs.size == 0 || state->compressed.data == NULL ||
+        state->decompressed.data == NULL) {
+        fprintf(stderr, "buffer state allocation failure\n");
+        return 1;
+    }
+    if (config->use_dictionary && state->dictionary.data == NULL) {
+        fprintf(stderr, "dictionary loading failed\n");
+        return 1;
+    }
+    return 0;
+}
+
+static result_t simple_compress(method_state_t* base, config_t const* config) {
+    buffer_state_t* state = container_of(base, buffer_state_t, base);
+
+    if (buffer_state_bad(state, config))
+        return result_error(result_error_system_error);
+
+    /* Keep the tests short by skipping directories, since behavior shouldn't
+     * change.
+     */
+    if (base->data->type != data_type_file)
+        return result_error(result_error_skip);
+
+    if (config->use_dictionary || config->no_pledged_src_size)
+        return result_error(result_error_skip);
+
+    /* If the config doesn't specify a level, skip. */
+    int const level = config_get_level(config);
+    if (level == CONFIG_NO_LEVEL)
+        return result_error(result_error_skip);
+
+    data_buffer_t const input = state->inputs.buffers[0];
+
+    /* Compress, decompress, and check the result. */
+    state->compressed.size = ZSTD_compress(
+        state->compressed.data,
+        state->compressed.capacity,
+        input.data,
+        input.size,
+        level);
+    if (ZSTD_isError(state->compressed.size))
+        return result_error(result_error_compression_error);
+
+    state->decompressed.size = ZSTD_decompress(
+        state->decompressed.data,
+        state->decompressed.capacity,
+        state->compressed.data,
+        state->compressed.size);
+    if (ZSTD_isError(state->decompressed.size))
+        return result_error(result_error_decompression_error);
+    if (data_buffer_compare(input, state->decompressed))
+        return result_error(result_error_round_trip_error);
+
+    result_data_t data;
+    data.total_size = state->compressed.size;
+    return result_data(data);
+}
+
+static result_t compress_cctx_compress(
+    method_state_t* base,
+    config_t const* config) {
+    buffer_state_t* state = container_of(base, buffer_state_t, base);
+
+    if (buffer_state_bad(state, config))
+        return result_error(result_error_system_error);
+
+    if (config->no_pledged_src_size)
+        return result_error(result_error_skip);
+
+    if (base->data->type != data_type_dir)
+        return result_error(result_error_skip);
+
+    int const level = config_get_level(config);
+
+    ZSTD_CCtx* cctx = ZSTD_createCCtx();
+    ZSTD_DCtx* dctx = ZSTD_createDCtx();
+    if (cctx == NULL || dctx == NULL) {
+        fprintf(stderr, "context creation failed\n");
+        return result_error(result_error_system_error);
+    }
+
+    result_t result;
+    result_data_t data = {.total_size = 0};
+    for (size_t i = 0; i < state->inputs.size; ++i) {
+        data_buffer_t const input = state->inputs.buffers[i];
+        ZSTD_parameters const params =
+            config_get_zstd_params(config, input.size, state->dictionary.size);
+
+        if (level == CONFIG_NO_LEVEL)
+            state->compressed.size = ZSTD_compress_advanced(
+                cctx,
+                state->compressed.data,
+                state->compressed.capacity,
+                input.data,
+                input.size,
+                state->dictionary.data,
+                state->dictionary.size,
+                params);
+        else if (config->use_dictionary)
+            state->compressed.size = ZSTD_compress_usingDict(
+                cctx,
+                state->compressed.data,
+                state->compressed.capacity,
+                input.data,
+                input.size,
+                state->dictionary.data,
+                state->dictionary.size,
+                level);
+        else
+            state->compressed.size = ZSTD_compressCCtx(
+                cctx,
+                state->compressed.data,
+                state->compressed.capacity,
+                input.data,
+                input.size,
+                level);
+
+        if (ZSTD_isError(state->compressed.size)) {
+            result = result_error(result_error_compression_error);
+            goto out;
+        }
+
+        if (config->use_dictionary)
+            state->decompressed.size = ZSTD_decompress_usingDict(
+                dctx,
+                state->decompressed.data,
+                state->decompressed.capacity,
+                state->compressed.data,
+                state->compressed.size,
+                state->dictionary.data,
+                state->dictionary.size);
+        else
+            state->decompressed.size = ZSTD_decompressDCtx(
+                dctx,
+                state->decompressed.data,
+                state->decompressed.capacity,
+                state->compressed.data,
+                state->compressed.size);
+        if (ZSTD_isError(state->decompressed.size)) {
+            result = result_error(result_error_decompression_error);
+            goto out;
+        }
+        if (data_buffer_compare(input, state->decompressed)) {
+            result = result_error(result_error_round_trip_error);
+            goto out;
+        }
+
+        data.total_size += state->compressed.size;
+    }
+
+    result = result_data(data);
+out:
+    ZSTD_freeCCtx(cctx);
+    ZSTD_freeDCtx(dctx);
+    return result;
+}
+
+/** Generic state creation function. */
+static method_state_t* method_state_create(data_t const* data) {
+    method_state_t* state = (method_state_t*)malloc(sizeof(method_state_t));
+    if (state == NULL)
+        return NULL;
+    state->data = data;
+    return state;
+}
+
+static void method_state_destroy(method_state_t* state) {
+    free(state);
+}
+
+static result_t cli_compress(method_state_t* state, config_t const* config) {
+    if (config->cli_args == NULL)
+        return result_error(result_error_skip);
+
+    /* We don't support no pledged source size with directories. Too slow. */
+    if (state->data->type == data_type_dir && config->no_pledged_src_size)
+        return result_error(result_error_skip);
+
+    if (g_zstdcli == NULL)
+        return result_error(result_error_system_error);
+
+    /* '' -cqr  [-D ''] '' */
+    char cmd[1024];
+    size_t const cmd_size = snprintf(
+        cmd,
+        sizeof(cmd),
+        "'%s' -cqr %s %s%s%s %s '%s'",
+        g_zstdcli,
+        config->cli_args,
+        config->use_dictionary ? "-D '" : "",
+        config->use_dictionary ? state->data->dict.path : "",
+        config->use_dictionary ? "'" : "",
+        config->no_pledged_src_size ? "<" : "",
+        state->data->data.path);
+    if (cmd_size >= sizeof(cmd)) {
+        fprintf(stderr, "command too large: %s\n", cmd);
+        return result_error(result_error_system_error);
+    }
+    FILE* zstd = popen(cmd, "r");
+    if (zstd == NULL) {
+        fprintf(stderr, "failed to popen command: %s\n", cmd);
+        return result_error(result_error_system_error);
+    }
+
+    char out[4096];
+    size_t total_size = 0;
+    while (1) {
+        size_t const size = fread(out, 1, sizeof(out), zstd);
+        total_size += size;
+        if (size != sizeof(out))
+            break;
+    }
+    if (ferror(zstd) || pclose(zstd) != 0) {
+        fprintf(stderr, "zstd failed with command: %s\n", cmd);
+        return result_error(result_error_compression_error);
+    }
+
+    result_data_t const data = {.total_size = total_size};
+    return result_data(data);
+}
+
+static int advanced_config(
+    ZSTD_CCtx* cctx,
+    buffer_state_t* state,
+    config_t const* config) {
+    ZSTD_CCtx_reset(cctx, ZSTD_reset_session_and_parameters);
+    for (size_t p = 0; p < config->param_values.size; ++p) {
+        param_value_t const pv = config->param_values.data[p];
+        if (ZSTD_isError(ZSTD_CCtx_setParameter(cctx, pv.param, pv.value))) {
+            return 1;
+        }
+    }
+    if (config->use_dictionary) {
+        if (ZSTD_isError(ZSTD_CCtx_loadDictionary(
+                cctx, state->dictionary.data, state->dictionary.size))) {
+            return 1;
+        }
+    }
+    return 0;
+}
+
+static result_t advanced_one_pass_compress_output_adjustment(
+    method_state_t* base,
+    config_t const* config,
+    size_t const subtract) {
+    buffer_state_t* state = container_of(base, buffer_state_t, base);
+
+    if (buffer_state_bad(state, config))
+        return result_error(result_error_system_error);
+
+    ZSTD_CCtx* cctx = ZSTD_createCCtx();
+    result_t result;
+
+    if (!cctx || advanced_config(cctx, state, config)) {
+        result = result_error(result_error_compression_error);
+        goto out;
+    }
+
+    result_data_t data = {.total_size = 0};
+    for (size_t i = 0; i < state->inputs.size; ++i) {
+        data_buffer_t const input = state->inputs.buffers[i];
+
+        if (!config->no_pledged_src_size) {
+            if (ZSTD_isError(ZSTD_CCtx_setPledgedSrcSize(cctx, input.size))) {
+                result = result_error(result_error_compression_error);
+                goto out;
+            }
+        }
+        size_t const size = ZSTD_compress2(
+            cctx,
+            state->compressed.data,
+            ZSTD_compressBound(input.size) - subtract,
+            input.data,
+            input.size);
+        if (ZSTD_isError(size)) {
+            result = result_error(result_error_compression_error);
+            goto out;
+        }
+        data.total_size += size;
+    }
+
+    result = result_data(data);
+out:
+    ZSTD_freeCCtx(cctx);
+    return result;
+}
+
+static result_t advanced_one_pass_compress(
+    method_state_t* base,
+    config_t const* config) {
+  return advanced_one_pass_compress_output_adjustment(base, config, 0);
+}
+
+static result_t advanced_one_pass_compress_small_output(
+    method_state_t* base,
+    config_t const* config) {
+  return advanced_one_pass_compress_output_adjustment(base, config, 1);
+}
+
+static result_t advanced_streaming_compress(
+    method_state_t* base,
+    config_t const* config) {
+    buffer_state_t* state = container_of(base, buffer_state_t, base);
+
+    if (buffer_state_bad(state, config))
+        return result_error(result_error_system_error);
+
+    ZSTD_CCtx* cctx = ZSTD_createCCtx();
+    result_t result;
+
+    if (!cctx || advanced_config(cctx, state, config)) {
+        result = result_error(result_error_compression_error);
+        goto out;
+    }
+
+    result_data_t data = {.total_size = 0};
+    for (size_t i = 0; i < state->inputs.size; ++i) {
+        data_buffer_t input = state->inputs.buffers[i];
+
+        if (!config->no_pledged_src_size) {
+            if (ZSTD_isError(ZSTD_CCtx_setPledgedSrcSize(cctx, input.size))) {
+                result = result_error(result_error_compression_error);
+                goto out;
+            }
+        }
+
+        while (input.size > 0) {
+            ZSTD_inBuffer in = {input.data, MIN(input.size, 4096)};
+            input.data += in.size;
+            input.size -= in.size;
+            ZSTD_EndDirective const op =
+                input.size > 0 ? ZSTD_e_continue : ZSTD_e_end;
+            size_t ret = 0;
+            while (in.pos < in.size || (op == ZSTD_e_end && ret != 0)) {
+                ZSTD_outBuffer out = {state->compressed.data,
+                                      MIN(state->compressed.capacity, 1024)};
+                ret = ZSTD_compressStream2(cctx, &out, &in, op);
+                if (ZSTD_isError(ret)) {
+                    result = result_error(result_error_compression_error);
+                    goto out;
+                }
+                data.total_size += out.pos;
+            }
+        }
+    }
+
+    result = result_data(data);
+out:
+    ZSTD_freeCCtx(cctx);
+    return result;
+}
+
+static result_t old_streaming_compress(
+    method_state_t* base,
+    config_t const* config) {
+    buffer_state_t* state = container_of(base, buffer_state_t, base);
+
+    if (buffer_state_bad(state, config))
+        return result_error(result_error_system_error);
+
+    int const level = config_get_level(config);
+    if (level == CONFIG_NO_LEVEL)
+        return result_error(result_error_skip);
+
+    ZSTD_CStream* zcs = ZSTD_createCStream();
+    result_t result;
+    if (zcs == NULL) {
+        result = result_error(result_error_compression_error);
+        goto out;
+    }
+    size_t zret;
+    if (config->use_dictionary) {
+        zret = ZSTD_initCStream_usingDict(
+            zcs, state->dictionary.data, state->dictionary.size, level);
+    } else {
+        zret = ZSTD_initCStream(zcs, level);
+    }
+    if (ZSTD_isError(zret)) {
+        result = result_error(result_error_compression_error);
+        goto out;
+    }
+
+    result_data_t data = {.total_size = 0};
+    for (size_t i = 0; i < state->inputs.size; ++i) {
+        data_buffer_t input = state->inputs.buffers[i];
+        zret = ZSTD_resetCStream(
+            zcs,
+            config->no_pledged_src_size ? ZSTD_CONTENTSIZE_UNKNOWN
+                                        : input.size);
+        if (ZSTD_isError(zret)) {
+            result = result_error(result_error_compression_error);
+            goto out;
+        }
+
+        while (input.size > 0) {
+            ZSTD_inBuffer in = {input.data, MIN(input.size, 4096)};
+            input.data += in.size;
+            input.size -= in.size;
+            ZSTD_EndDirective const op =
+                input.size > 0 ? ZSTD_e_continue : ZSTD_e_end;
+            zret = 0;
+            while (in.pos < in.size || (op == ZSTD_e_end && zret != 0)) {
+                ZSTD_outBuffer out = {state->compressed.data,
+                                      MIN(state->compressed.capacity, 1024)};
+                if (op == ZSTD_e_continue || in.pos < in.size)
+                    zret = ZSTD_compressStream(zcs, &out, &in);
+                else
+                    zret = ZSTD_endStream(zcs, &out);
+                if (ZSTD_isError(zret)) {
+                    result = result_error(result_error_compression_error);
+                    goto out;
+                }
+                data.total_size += out.pos;
+            }
+        }
+    }
+
+    result = result_data(data);
+out:
+    ZSTD_freeCStream(zcs);
+    return result;
+}
+
+method_t const simple = {
+    .name = "compress simple",
+    .create = buffer_state_create,
+    .compress = simple_compress,
+    .destroy = buffer_state_destroy,
+};
+
+method_t const compress_cctx = {
+    .name = "compress cctx",
+    .create = buffer_state_create,
+    .compress = compress_cctx_compress,
+    .destroy = buffer_state_destroy,
+};
+
+method_t const advanced_one_pass = {
+    .name = "advanced one pass",
+    .create = buffer_state_create,
+    .compress = advanced_one_pass_compress,
+    .destroy = buffer_state_destroy,
+};
+
+method_t const advanced_one_pass_small_out = {
+    .name = "advanced one pass small out",
+    .create = buffer_state_create,
+    .compress = advanced_one_pass_compress,
+    .destroy = buffer_state_destroy,
+};
+
+method_t const advanced_streaming = {
+    .name = "advanced streaming",
+    .create = buffer_state_create,
+    .compress = advanced_streaming_compress,
+    .destroy = buffer_state_destroy,
+};
+
+method_t const old_streaming = {
+    .name = "old streaming",
+    .create = buffer_state_create,
+    .compress = old_streaming_compress,
+    .destroy = buffer_state_destroy,
+};
+
+method_t const cli = {
+    .name = "zstdcli",
+    .create = method_state_create,
+    .compress = cli_compress,
+    .destroy = method_state_destroy,
+};
+
+static method_t const* g_methods[] = {
+    &simple,
+    &compress_cctx,
+    &cli,
+    &advanced_one_pass,
+    &advanced_one_pass_small_out,
+    &advanced_streaming,
+    &old_streaming,
+    NULL,
+};
+
+method_t const* const* methods = g_methods;
diff --git a/tests/regression/method.h b/tests/regression/method.h
new file mode 100644
index 000000000000..d70b776b1b90
--- /dev/null
+++ b/tests/regression/method.h
@@ -0,0 +1,65 @@
+/*
+ * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
+ * You may select, at your option, one of the above-listed licenses.
+ */
+
+#ifndef METHOD_H
+#define METHOD_H
+
+#include 
+
+#include "data.h"
+#include "config.h"
+#include "result.h"
+
+/**
+ * The base class for state that methods keep.
+ * All derived method state classes must have a member of this type.
+ */
+typedef struct {
+    data_t const* data;
+} method_state_t;
+
+/**
+ * A method that compresses the data using config.
+ */
+typedef struct {
+    char const* name;  /**< The identifier for this method in the results. */
+    /**
+     * Creates a state that must contain a member variable of method_state_t,
+     * and returns a pointer to that member variable.
+     *
+     * This method can be used to do expensive work that only depends on the
+     * data, like loading the data file into a buffer.
+     */
+    method_state_t* (*create)(data_t const* data);
+    /**
+     * Compresses the data in the state using the given config.
+     *
+     * @param state A pointer to the state returned by create().
+     *
+     * @returns The total compressed size on success, or an error code.
+     */
+    result_t (*compress)(method_state_t* state, config_t const* config);
+    /**
+     * Frees the state.
+     */
+    void (*destroy)(method_state_t* state);
+} method_t;
+
+/**
+ * Set the zstd cli path. Must be called before any methods are used.
+ */
+void method_set_zstdcli(char const* zstdcli);
+
+/**
+ * A NULL-terminated list of methods.
+ */
+extern method_t const* const* methods;
+
+#endif
diff --git a/tests/regression/result.c b/tests/regression/result.c
new file mode 100644
index 000000000000..31439b08cde7
--- /dev/null
+++ b/tests/regression/result.c
@@ -0,0 +1,28 @@
+/*
+ * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
+ * You may select, at your option, one of the above-listed licenses.
+ */
+
+#include "result.h"
+
+char const* result_get_error_string(result_t result) {
+    switch (result_get_error(result)) {
+        case result_error_ok:
+            return "okay";
+        case result_error_skip:
+            return "skip";
+        case result_error_system_error:
+            return "system error";
+        case result_error_compression_error:
+            return "compression error";
+        case result_error_decompression_error:
+            return "decompression error";
+        case result_error_round_trip_error:
+            return "round trip error";
+    }
+}
diff --git a/tests/regression/result.h b/tests/regression/result.h
new file mode 100644
index 000000000000..8c80cf85a966
--- /dev/null
+++ b/tests/regression/result.h
@@ -0,0 +1,103 @@
+/*
+ * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
+ * You may select, at your option, one of the above-listed licenses.
+ */
+
+#ifndef RESULT_H
+#define RESULT_H
+
+#include 
+
+/**
+ * The error type enum.
+ */
+typedef enum {
+    result_error_ok,                   /**< No error. */
+    result_error_skip,                 /**< This method was skipped. */
+    result_error_system_error,         /**< Some internal error happened. */
+    result_error_compression_error,    /**< Compression failed. */
+    result_error_decompression_error,  /**< Decompression failed. */
+    result_error_round_trip_error,     /**< Data failed to round trip. */
+} result_error_t;
+
+/**
+ * The success type.
+ */
+typedef struct {
+    size_t total_size;  /**< The total compressed size. */
+} result_data_t;
+
+/**
+ * The result type.
+ * Do not access the member variables directory, use the helper functions.
+ */
+typedef struct {
+    result_error_t internal_error;
+    result_data_t internal_data;
+} result_t;
+
+/**
+ * Create a result of the error type.
+ */
+static result_t result_error(result_error_t error);
+/**
+ * Create a result of the success type.
+ */
+static result_t result_data(result_data_t data);
+
+/**
+ * Check if the result is an error or skip.
+ */
+static int result_is_error(result_t result);
+/**
+ * Check if the result error is skip.
+ */
+static int result_is_skip(result_t result);
+/**
+ * Get the result error or okay.
+ */
+static result_error_t result_get_error(result_t result);
+/**
+ * Get the result data. The result MUST be checked with result_is_error() first.
+ */
+static result_data_t result_get_data(result_t result);
+
+static result_t result_error(result_error_t error) {
+    result_t result = {
+        .internal_error = error,
+    };
+    return result;
+}
+
+static result_t result_data(result_data_t data) {
+    result_t result = {
+        .internal_error = result_error_ok,
+        .internal_data = data,
+    };
+    return result;
+}
+
+static int result_is_error(result_t result) {
+    return result_get_error(result) != result_error_ok;
+}
+
+static int result_is_skip(result_t result) {
+    return result_get_error(result) == result_error_skip;
+}
+
+static result_error_t result_get_error(result_t result) {
+    return result.internal_error;
+}
+
+char const* result_get_error_string(result_t result);
+
+static result_data_t result_get_data(result_t result) {
+    return result.internal_data;
+}
+
+#endif
diff --git a/tests/regression/results.csv b/tests/regression/results.csv
new file mode 100644
index 000000000000..a62178d6ed0c
--- /dev/null
+++ b/tests/regression/results.csv
@@ -0,0 +1,448 @@
+Data,                             Config,                           Method,                           Total compressed size
+silesia.tar,                      level -5,                         compress simple,                  7160438
+silesia.tar,                      level -3,                         compress simple,                  6789024
+silesia.tar,                      level -1,                         compress simple,                  6195462
+silesia.tar,                      level 0,                          compress simple,                  4875008
+silesia.tar,                      level 1,                          compress simple,                  5339697
+silesia.tar,                      level 3,                          compress simple,                  4875008
+silesia.tar,                      level 4,                          compress simple,                  4813507
+silesia.tar,                      level 5,                          compress simple,                  4722235
+silesia.tar,                      level 6,                          compress simple,                  4672194
+silesia.tar,                      level 7,                          compress simple,                  4606658
+silesia.tar,                      level 9,                          compress simple,                  4554098
+silesia.tar,                      level 13,                         compress simple,                  4491702
+silesia.tar,                      level 16,                         compress simple,                  4381277
+silesia.tar,                      level 19,                         compress simple,                  4281514
+silesia,                          level -5,                         compress cctx,                    7152294
+silesia,                          level -3,                         compress cctx,                    6789969
+silesia,                          level -1,                         compress cctx,                    6191548
+silesia,                          level 0,                          compress cctx,                    4862377
+silesia,                          level 1,                          compress cctx,                    5318036
+silesia,                          level 3,                          compress cctx,                    4862377
+silesia,                          level 4,                          compress cctx,                    4800629
+silesia,                          level 5,                          compress cctx,                    4710178
+silesia,                          level 6,                          compress cctx,                    4659996
+silesia,                          level 7,                          compress cctx,                    4596234
+silesia,                          level 9,                          compress cctx,                    4543862
+silesia,                          level 13,                         compress cctx,                    4482073
+silesia,                          level 16,                         compress cctx,                    4377391
+silesia,                          level 19,                         compress cctx,                    4293262
+silesia,                          long distance mode,               compress cctx,                    4862377
+silesia,                          multithreaded,                    compress cctx,                    4862377
+silesia,                          multithreaded long distance mode, compress cctx,                    4862377
+silesia,                          small window log,                 compress cctx,                    7115734
+silesia,                          small hash log,                   compress cctx,                    6554898
+silesia,                          small chain log,                  compress cctx,                    4931093
+silesia,                          explicit params,                  compress cctx,                    4813352
+github,                           level -5,                         compress cctx,                    232744
+github,                           level -5 with dict,               compress cctx,                    47294
+github,                           level -3,                         compress cctx,                    220611
+github,                           level -3 with dict,               compress cctx,                    48047
+github,                           level -1,                         compress cctx,                    176575
+github,                           level -1 with dict,               compress cctx,                    43527
+github,                           level 0,                          compress cctx,                    136397
+github,                           level 0 with dict,                compress cctx,                    41536
+github,                           level 1,                          compress cctx,                    143457
+github,                           level 1 with dict,                compress cctx,                    42157
+github,                           level 3,                          compress cctx,                    136397
+github,                           level 3 with dict,                compress cctx,                    41536
+github,                           level 4,                          compress cctx,                    136144
+github,                           level 4 with dict,                compress cctx,                    41721
+github,                           level 5,                          compress cctx,                    135106
+github,                           level 5 with dict,                compress cctx,                    38934
+github,                           level 6,                          compress cctx,                    135108
+github,                           level 6 with dict,                compress cctx,                    38628
+github,                           level 7,                          compress cctx,                    135108
+github,                           level 7 with dict,                compress cctx,                    38741
+github,                           level 9,                          compress cctx,                    135108
+github,                           level 9 with dict,                compress cctx,                    39335
+github,                           level 13,                         compress cctx,                    133717
+github,                           level 13 with dict,               compress cctx,                    39923
+github,                           level 16,                         compress cctx,                    133717
+github,                           level 16 with dict,               compress cctx,                    37568
+github,                           level 19,                         compress cctx,                    133717
+github,                           level 19 with dict,               compress cctx,                    37567
+github,                           long distance mode,               compress cctx,                    decompression error
+github,                           multithreaded,                    compress cctx,                    decompression error
+github,                           multithreaded long distance mode, compress cctx,                    decompression error
+github,                           small window log,                 compress cctx,                    decompression error
+github,                           small hash log,                   compress cctx,                    decompression error
+github,                           small chain log,                  compress cctx,                    decompression error
+github,                           explicit params,                  compress cctx,                    decompression error
+silesia,                          level -5,                         zstdcli,                          7152342
+silesia,                          level -3,                         zstdcli,                          6790021
+silesia,                          level -1,                         zstdcli,                          6191597
+silesia,                          level 0,                          zstdcli,                          4862425
+silesia,                          level 1,                          zstdcli,                          5318084
+silesia,                          level 3,                          zstdcli,                          4862425
+silesia,                          level 4,                          zstdcli,                          4800677
+silesia,                          level 5,                          zstdcli,                          4710226
+silesia,                          level 6,                          zstdcli,                          4660044
+silesia,                          level 7,                          zstdcli,                          4596282
+silesia,                          level 9,                          zstdcli,                          4543910
+silesia,                          level 13,                         zstdcli,                          4482121
+silesia,                          level 16,                         zstdcli,                          4377439
+silesia,                          level 19,                         zstdcli,                          4293310
+silesia,                          long distance mode,               zstdcli,                          4853437
+silesia,                          multithreaded,                    zstdcli,                          4862425
+silesia,                          multithreaded long distance mode, zstdcli,                          4853437
+silesia,                          small window log,                 zstdcli,                          7126434
+silesia,                          small hash log,                   zstdcli,                          6554946
+silesia,                          small chain log,                  zstdcli,                          4931141
+silesia,                          explicit params,                  zstdcli,                          4815380
+silesia.tar,                      level -5,                         zstdcli,                          7161160
+silesia.tar,                      level -3,                         zstdcli,                          6789865
+silesia.tar,                      level -1,                         zstdcli,                          6196433
+silesia.tar,                      level 0,                          zstdcli,                          4875136
+silesia.tar,                      level 1,                          zstdcli,                          5340573
+silesia.tar,                      level 3,                          zstdcli,                          4875136
+silesia.tar,                      level 4,                          zstdcli,                          4814531
+silesia.tar,                      level 5,                          zstdcli,                          4723284
+silesia.tar,                      level 6,                          zstdcli,                          4673591
+silesia.tar,                      level 7,                          zstdcli,                          4608342
+silesia.tar,                      level 9,                          zstdcli,                          4554700
+silesia.tar,                      level 13,                         zstdcli,                          4491706
+silesia.tar,                      level 16,                         zstdcli,                          4381281
+silesia.tar,                      level 19,                         zstdcli,                          4281518
+silesia.tar,                      no source size,                   zstdcli,                          4875132
+silesia.tar,                      long distance mode,               zstdcli,                          4866975
+silesia.tar,                      multithreaded,                    zstdcli,                          4875136
+silesia.tar,                      multithreaded long distance mode, zstdcli,                          4866975
+silesia.tar,                      small window log,                 zstdcli,                          7130434
+silesia.tar,                      small hash log,                   zstdcli,                          6587841
+silesia.tar,                      small chain log,                  zstdcli,                          4943259
+silesia.tar,                      explicit params,                  zstdcli,                          4839202
+github,                           level -5,                         zstdcli,                          234744
+github,                           level -5 with dict,               zstdcli,                          48718
+github,                           level -3,                         zstdcli,                          222611
+github,                           level -3 with dict,               zstdcli,                          47395
+github,                           level -1,                         zstdcli,                          178575
+github,                           level -1 with dict,               zstdcli,                          45170
+github,                           level 0,                          zstdcli,                          138397
+github,                           level 0 with dict,                zstdcli,                          43170
+github,                           level 1,                          zstdcli,                          145457
+github,                           level 1 with dict,                zstdcli,                          43682
+github,                           level 3,                          zstdcli,                          138397
+github,                           level 3 with dict,                zstdcli,                          43170
+github,                           level 4,                          zstdcli,                          138144
+github,                           level 4 with dict,                zstdcli,                          43306
+github,                           level 5,                          zstdcli,                          137106
+github,                           level 5 with dict,                zstdcli,                          40938
+github,                           level 6,                          zstdcli,                          137108
+github,                           level 6 with dict,                zstdcli,                          40632
+github,                           level 7,                          zstdcli,                          137108
+github,                           level 7 with dict,                zstdcli,                          40766
+github,                           level 9,                          zstdcli,                          137108
+github,                           level 9 with dict,                zstdcli,                          41326
+github,                           level 13,                         zstdcli,                          135717
+github,                           level 13 with dict,               zstdcli,                          41716
+github,                           level 16,                         zstdcli,                          135717
+github,                           level 16 with dict,               zstdcli,                          39577
+github,                           level 19,                         zstdcli,                          135717
+github,                           level 19 with dict,               zstdcli,                          39576
+github,                           long distance mode,               zstdcli,                          138397
+github,                           multithreaded,                    zstdcli,                          138397
+github,                           multithreaded long distance mode, zstdcli,                          138397
+github,                           small window log,                 zstdcli,                          138397
+github,                           small hash log,                   zstdcli,                          137467
+github,                           small chain log,                  zstdcli,                          138314
+github,                           explicit params,                  zstdcli,                          136140
+silesia,                          level -5,                         advanced one pass,                7152294
+silesia,                          level -3,                         advanced one pass,                6789969
+silesia,                          level -1,                         advanced one pass,                6191548
+silesia,                          level 0,                          advanced one pass,                4862377
+silesia,                          level 1,                          advanced one pass,                5318036
+silesia,                          level 3,                          advanced one pass,                4862377
+silesia,                          level 4,                          advanced one pass,                4800629
+silesia,                          level 5,                          advanced one pass,                4710178
+silesia,                          level 6,                          advanced one pass,                4659996
+silesia,                          level 7,                          advanced one pass,                4596234
+silesia,                          level 9,                          advanced one pass,                4543862
+silesia,                          level 13,                         advanced one pass,                4482073
+silesia,                          level 16,                         advanced one pass,                4377391
+silesia,                          level 19,                         advanced one pass,                4293262
+silesia,                          no source size,                   advanced one pass,                4862377
+silesia,                          long distance mode,               advanced one pass,                4853389
+silesia,                          multithreaded,                    advanced one pass,                4862377
+silesia,                          multithreaded long distance mode, advanced one pass,                4853389
+silesia,                          small window log,                 advanced one pass,                7126386
+silesia,                          small hash log,                   advanced one pass,                6554898
+silesia,                          small chain log,                  advanced one pass,                4931093
+silesia,                          explicit params,                  advanced one pass,                4815369
+silesia.tar,                      level -5,                         advanced one pass,                7160438
+silesia.tar,                      level -3,                         advanced one pass,                6789024
+silesia.tar,                      level -1,                         advanced one pass,                6195462
+silesia.tar,                      level 0,                          advanced one pass,                4875008
+silesia.tar,                      level 1,                          advanced one pass,                5339697
+silesia.tar,                      level 3,                          advanced one pass,                4875008
+silesia.tar,                      level 4,                          advanced one pass,                4813507
+silesia.tar,                      level 5,                          advanced one pass,                4722235
+silesia.tar,                      level 6,                          advanced one pass,                4672194
+silesia.tar,                      level 7,                          advanced one pass,                4606658
+silesia.tar,                      level 9,                          advanced one pass,                4554098
+silesia.tar,                      level 13,                         advanced one pass,                4491702
+silesia.tar,                      level 16,                         advanced one pass,                4381277
+silesia.tar,                      level 19,                         advanced one pass,                4281514
+silesia.tar,                      no source size,                   advanced one pass,                4875008
+silesia.tar,                      long distance mode,               advanced one pass,                4861218
+silesia.tar,                      multithreaded,                    advanced one pass,                4874631
+silesia.tar,                      multithreaded long distance mode, advanced one pass,                4860683
+silesia.tar,                      small window log,                 advanced one pass,                7130394
+silesia.tar,                      small hash log,                   advanced one pass,                6587833
+silesia.tar,                      small chain log,                  advanced one pass,                4943255
+silesia.tar,                      explicit params,                  advanced one pass,                4829974
+github,                           level -5,                         advanced one pass,                232744
+github,                           level -5 with dict,               advanced one pass,                46718
+github,                           level -3,                         advanced one pass,                220611
+github,                           level -3 with dict,               advanced one pass,                45395
+github,                           level -1,                         advanced one pass,                176575
+github,                           level -1 with dict,               advanced one pass,                43170
+github,                           level 0,                          advanced one pass,                136397
+github,                           level 0 with dict,                advanced one pass,                41170
+github,                           level 1,                          advanced one pass,                143457
+github,                           level 1 with dict,                advanced one pass,                41682
+github,                           level 3,                          advanced one pass,                136397
+github,                           level 3 with dict,                advanced one pass,                41170
+github,                           level 4,                          advanced one pass,                136144
+github,                           level 4 with dict,                advanced one pass,                41306
+github,                           level 5,                          advanced one pass,                135106
+github,                           level 5 with dict,                advanced one pass,                38938
+github,                           level 6,                          advanced one pass,                135108
+github,                           level 6 with dict,                advanced one pass,                38632
+github,                           level 7,                          advanced one pass,                135108
+github,                           level 7 with dict,                advanced one pass,                38766
+github,                           level 9,                          advanced one pass,                135108
+github,                           level 9 with dict,                advanced one pass,                39326
+github,                           level 13,                         advanced one pass,                133717
+github,                           level 13 with dict,               advanced one pass,                39716
+github,                           level 16,                         advanced one pass,                133717
+github,                           level 16 with dict,               advanced one pass,                37577
+github,                           level 19,                         advanced one pass,                133717
+github,                           level 19 with dict,               advanced one pass,                37576
+github,                           no source size,                   advanced one pass,                136397
+github,                           long distance mode,               advanced one pass,                136397
+github,                           multithreaded,                    advanced one pass,                136397
+github,                           multithreaded long distance mode, advanced one pass,                136397
+github,                           small window log,                 advanced one pass,                136397
+github,                           small hash log,                   advanced one pass,                135467
+github,                           small chain log,                  advanced one pass,                136314
+github,                           explicit params,                  advanced one pass,                137670
+silesia,                          level -5,                         advanced one pass small out,      7152294
+silesia,                          level -3,                         advanced one pass small out,      6789969
+silesia,                          level -1,                         advanced one pass small out,      6191548
+silesia,                          level 0,                          advanced one pass small out,      4862377
+silesia,                          level 1,                          advanced one pass small out,      5318036
+silesia,                          level 3,                          advanced one pass small out,      4862377
+silesia,                          level 4,                          advanced one pass small out,      4800629
+silesia,                          level 5,                          advanced one pass small out,      4710178
+silesia,                          level 6,                          advanced one pass small out,      4659996
+silesia,                          level 7,                          advanced one pass small out,      4596234
+silesia,                          level 9,                          advanced one pass small out,      4543862
+silesia,                          level 13,                         advanced one pass small out,      4482073
+silesia,                          level 16,                         advanced one pass small out,      4377391
+silesia,                          level 19,                         advanced one pass small out,      4293262
+silesia,                          no source size,                   advanced one pass small out,      4862377
+silesia,                          long distance mode,               advanced one pass small out,      4853389
+silesia,                          multithreaded,                    advanced one pass small out,      4862377
+silesia,                          multithreaded long distance mode, advanced one pass small out,      4853389
+silesia,                          small window log,                 advanced one pass small out,      7126386
+silesia,                          small hash log,                   advanced one pass small out,      6554898
+silesia,                          small chain log,                  advanced one pass small out,      4931093
+silesia,                          explicit params,                  advanced one pass small out,      4815369
+silesia.tar,                      level -5,                         advanced one pass small out,      7160438
+silesia.tar,                      level -3,                         advanced one pass small out,      6789024
+silesia.tar,                      level -1,                         advanced one pass small out,      6195462
+silesia.tar,                      level 0,                          advanced one pass small out,      4875008
+silesia.tar,                      level 1,                          advanced one pass small out,      5339697
+silesia.tar,                      level 3,                          advanced one pass small out,      4875008
+silesia.tar,                      level 4,                          advanced one pass small out,      4813507
+silesia.tar,                      level 5,                          advanced one pass small out,      4722235
+silesia.tar,                      level 6,                          advanced one pass small out,      4672194
+silesia.tar,                      level 7,                          advanced one pass small out,      4606658
+silesia.tar,                      level 9,                          advanced one pass small out,      4554098
+silesia.tar,                      level 13,                         advanced one pass small out,      4491702
+silesia.tar,                      level 16,                         advanced one pass small out,      4381277
+silesia.tar,                      level 19,                         advanced one pass small out,      4281514
+silesia.tar,                      no source size,                   advanced one pass small out,      4875008
+silesia.tar,                      long distance mode,               advanced one pass small out,      4861218
+silesia.tar,                      multithreaded,                    advanced one pass small out,      4874631
+silesia.tar,                      multithreaded long distance mode, advanced one pass small out,      4860683
+silesia.tar,                      small window log,                 advanced one pass small out,      7130394
+silesia.tar,                      small hash log,                   advanced one pass small out,      6587833
+silesia.tar,                      small chain log,                  advanced one pass small out,      4943255
+silesia.tar,                      explicit params,                  advanced one pass small out,      4829974
+github,                           level -5,                         advanced one pass small out,      232744
+github,                           level -5 with dict,               advanced one pass small out,      46718
+github,                           level -3,                         advanced one pass small out,      220611
+github,                           level -3 with dict,               advanced one pass small out,      45395
+github,                           level -1,                         advanced one pass small out,      176575
+github,                           level -1 with dict,               advanced one pass small out,      43170
+github,                           level 0,                          advanced one pass small out,      136397
+github,                           level 0 with dict,                advanced one pass small out,      41170
+github,                           level 1,                          advanced one pass small out,      143457
+github,                           level 1 with dict,                advanced one pass small out,      41682
+github,                           level 3,                          advanced one pass small out,      136397
+github,                           level 3 with dict,                advanced one pass small out,      41170
+github,                           level 4,                          advanced one pass small out,      136144
+github,                           level 4 with dict,                advanced one pass small out,      41306
+github,                           level 5,                          advanced one pass small out,      135106
+github,                           level 5 with dict,                advanced one pass small out,      38938
+github,                           level 6,                          advanced one pass small out,      135108
+github,                           level 6 with dict,                advanced one pass small out,      38632
+github,                           level 7,                          advanced one pass small out,      135108
+github,                           level 7 with dict,                advanced one pass small out,      38766
+github,                           level 9,                          advanced one pass small out,      135108
+github,                           level 9 with dict,                advanced one pass small out,      39326
+github,                           level 13,                         advanced one pass small out,      133717
+github,                           level 13 with dict,               advanced one pass small out,      39716
+github,                           level 16,                         advanced one pass small out,      133717
+github,                           level 16 with dict,               advanced one pass small out,      37577
+github,                           level 19,                         advanced one pass small out,      133717
+github,                           level 19 with dict,               advanced one pass small out,      37576
+github,                           no source size,                   advanced one pass small out,      136397
+github,                           long distance mode,               advanced one pass small out,      136397
+github,                           multithreaded,                    advanced one pass small out,      136397
+github,                           multithreaded long distance mode, advanced one pass small out,      136397
+github,                           small window log,                 advanced one pass small out,      136397
+github,                           small hash log,                   advanced one pass small out,      135467
+github,                           small chain log,                  advanced one pass small out,      136314
+github,                           explicit params,                  advanced one pass small out,      137670
+silesia,                          level -5,                         advanced streaming,               7152294
+silesia,                          level -3,                         advanced streaming,               6789973
+silesia,                          level -1,                         advanced streaming,               6191549
+silesia,                          level 0,                          advanced streaming,               4862377
+silesia,                          level 1,                          advanced streaming,               5318036
+silesia,                          level 3,                          advanced streaming,               4862377
+silesia,                          level 4,                          advanced streaming,               4800629
+silesia,                          level 5,                          advanced streaming,               4710178
+silesia,                          level 6,                          advanced streaming,               4659996
+silesia,                          level 7,                          advanced streaming,               4596234
+silesia,                          level 9,                          advanced streaming,               4543862
+silesia,                          level 13,                         advanced streaming,               4482073
+silesia,                          level 16,                         advanced streaming,               4377391
+silesia,                          level 19,                         advanced streaming,               4293262
+silesia,                          no source size,                   advanced streaming,               4862341
+silesia,                          long distance mode,               advanced streaming,               4853389
+silesia,                          multithreaded,                    advanced streaming,               4862377
+silesia,                          multithreaded long distance mode, advanced streaming,               4853389
+silesia,                          small window log,                 advanced streaming,               7126389
+silesia,                          small hash log,                   advanced streaming,               6554898
+silesia,                          small chain log,                  advanced streaming,               4931093
+silesia,                          explicit params,                  advanced streaming,               4815380
+silesia.tar,                      level -5,                         advanced streaming,               7160440
+silesia.tar,                      level -3,                         advanced streaming,               6789026
+silesia.tar,                      level -1,                         advanced streaming,               6195465
+silesia.tar,                      level 0,                          advanced streaming,               4875010
+silesia.tar,                      level 1,                          advanced streaming,               5339701
+silesia.tar,                      level 3,                          advanced streaming,               4875010
+silesia.tar,                      level 4,                          advanced streaming,               4813507
+silesia.tar,                      level 5,                          advanced streaming,               4722240
+silesia.tar,                      level 6,                          advanced streaming,               4672203
+silesia.tar,                      level 7,                          advanced streaming,               4606658
+silesia.tar,                      level 9,                          advanced streaming,               4554105
+silesia.tar,                      level 13,                         advanced streaming,               4491703
+silesia.tar,                      level 16,                         advanced streaming,               4381277
+silesia.tar,                      level 19,                         advanced streaming,               4281514
+silesia.tar,                      no source size,                   advanced streaming,               4875006
+silesia.tar,                      long distance mode,               advanced streaming,               4861218
+silesia.tar,                      multithreaded,                    advanced streaming,               4875132
+silesia.tar,                      multithreaded long distance mode, advanced streaming,               4866971
+silesia.tar,                      small window log,                 advanced streaming,               7130394
+silesia.tar,                      small hash log,                   advanced streaming,               6587834
+silesia.tar,                      small chain log,                  advanced streaming,               4943260
+silesia.tar,                      explicit params,                  advanced streaming,               4830002
+github,                           level -5,                         advanced streaming,               232744
+github,                           level -5 with dict,               advanced streaming,               46718
+github,                           level -3,                         advanced streaming,               220611
+github,                           level -3 with dict,               advanced streaming,               45395
+github,                           level -1,                         advanced streaming,               176575
+github,                           level -1 with dict,               advanced streaming,               43170
+github,                           level 0,                          advanced streaming,               136397
+github,                           level 0 with dict,                advanced streaming,               41170
+github,                           level 1,                          advanced streaming,               143457
+github,                           level 1 with dict,                advanced streaming,               41682
+github,                           level 3,                          advanced streaming,               136397
+github,                           level 3 with dict,                advanced streaming,               41170
+github,                           level 4,                          advanced streaming,               136144
+github,                           level 4 with dict,                advanced streaming,               41306
+github,                           level 5,                          advanced streaming,               135106
+github,                           level 5 with dict,                advanced streaming,               38938
+github,                           level 6,                          advanced streaming,               135108
+github,                           level 6 with dict,                advanced streaming,               38632
+github,                           level 7,                          advanced streaming,               135108
+github,                           level 7 with dict,                advanced streaming,               38766
+github,                           level 9,                          advanced streaming,               135108
+github,                           level 9 with dict,                advanced streaming,               39326
+github,                           level 13,                         advanced streaming,               133717
+github,                           level 13 with dict,               advanced streaming,               39716
+github,                           level 16,                         advanced streaming,               133717
+github,                           level 16 with dict,               advanced streaming,               37577
+github,                           level 19,                         advanced streaming,               133717
+github,                           level 19 with dict,               advanced streaming,               37576
+github,                           no source size,                   advanced streaming,               136397
+github,                           long distance mode,               advanced streaming,               136397
+github,                           multithreaded,                    advanced streaming,               136397
+github,                           multithreaded long distance mode, advanced streaming,               136397
+github,                           small window log,                 advanced streaming,               136397
+github,                           small hash log,                   advanced streaming,               135467
+github,                           small chain log,                  advanced streaming,               136314
+github,                           explicit params,                  advanced streaming,               137670
+silesia,                          level -5,                         old streaming,                    7152294
+silesia,                          level -3,                         old streaming,                    6789973
+silesia,                          level -1,                         old streaming,                    6191549
+silesia,                          level 0,                          old streaming,                    4862377
+silesia,                          level 1,                          old streaming,                    5318036
+silesia,                          level 3,                          old streaming,                    4862377
+silesia,                          level 4,                          old streaming,                    4800629
+silesia,                          level 5,                          old streaming,                    4710178
+silesia,                          level 6,                          old streaming,                    4659996
+silesia,                          level 7,                          old streaming,                    4596234
+silesia,                          level 9,                          old streaming,                    4543862
+silesia,                          level 13,                         old streaming,                    4482073
+silesia,                          level 16,                         old streaming,                    4377391
+silesia,                          level 19,                         old streaming,                    4293262
+silesia,                          no source size,                   old streaming,                    4862341
+silesia.tar,                      level -5,                         old streaming,                    7160440
+silesia.tar,                      level -3,                         old streaming,                    6789026
+silesia.tar,                      level -1,                         old streaming,                    6195465
+silesia.tar,                      level 0,                          old streaming,                    4875010
+silesia.tar,                      level 1,                          old streaming,                    5339701
+silesia.tar,                      level 3,                          old streaming,                    4875010
+silesia.tar,                      level 4,                          old streaming,                    4813507
+silesia.tar,                      level 5,                          old streaming,                    4722240
+silesia.tar,                      level 6,                          old streaming,                    4672203
+silesia.tar,                      level 7,                          old streaming,                    4606658
+silesia.tar,                      level 9,                          old streaming,                    4554105
+silesia.tar,                      level 13,                         old streaming,                    4491703
+silesia.tar,                      level 16,                         old streaming,                    4381277
+silesia.tar,                      level 19,                         old streaming,                    4281514
+silesia.tar,                      no source size,                   old streaming,                    4875006
+github,                           level -5,                         old streaming,                    232744
+github,                           level -5 with dict,               old streaming,                    46718
+github,                           level -3,                         old streaming,                    220611
+github,                           level -3 with dict,               old streaming,                    45395
+github,                           level -1,                         old streaming,                    176575
+github,                           level -1 with dict,               old streaming,                    43170
+github,                           level 0,                          old streaming,                    136397
+github,                           level 0 with dict,                old streaming,                    41170
+github,                           level 1,                          old streaming,                    143457
+github,                           level 1 with dict,                old streaming,                    41682
+github,                           level 3,                          old streaming,                    136397
+github,                           level 3 with dict,                old streaming,                    41170
+github,                           level 4,                          old streaming,                    136144
+github,                           level 4 with dict,                old streaming,                    41306
+github,                           level 5,                          old streaming,                    135106
+github,                           level 5 with dict,                old streaming,                    38938
+github,                           level 6,                          old streaming,                    135108
+github,                           level 6 with dict,                old streaming,                    38632
+github,                           level 7,                          old streaming,                    135108
+github,                           level 7 with dict,                old streaming,                    38766
+github,                           level 9,                          old streaming,                    135108
+github,                           level 9 with dict,                old streaming,                    39326
+github,                           level 13,                         old streaming,                    133717
+github,                           level 13 with dict,               old streaming,                    39716
+github,                           level 16,                         old streaming,                    133717
+github,                           level 16 with dict,               old streaming,                    37577
+github,                           level 19,                         old streaming,                    133717
+github,                           level 19 with dict,               old streaming,                    37576
+github,                           no source size,                   old streaming,                    141003
diff --git a/tests/regression/test.c b/tests/regression/test.c
new file mode 100644
index 000000000000..9e7b83c19daf
--- /dev/null
+++ b/tests/regression/test.c
@@ -0,0 +1,362 @@
+/*
+ * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
+ * You may select, at your option, one of the above-listed licenses.
+ */
+
+#include 
+#include 
+#include 
+#include 
+
+#include "config.h"
+#include "data.h"
+#include "method.h"
+
+static int g_max_name_len = 0;
+
+/** Check if a name contains a comma or is too long. */
+static int is_name_bad(char const* name) {
+    if (name == NULL)
+        return 1;
+    int const len = strlen(name);
+    if (len > g_max_name_len)
+        g_max_name_len = len;
+    for (; *name != '\0'; ++name)
+        if (*name == ',')
+            return 1;
+    return 0;
+}
+
+/** Check if any of the names contain a comma. */
+static int are_names_bad() {
+    for (size_t method = 0; methods[method] != NULL; ++method)
+        if (is_name_bad(methods[method]->name)) {
+            fprintf(stderr, "method name %s is bad\n", methods[method]->name);
+            return 1;
+        }
+    for (size_t datum = 0; data[datum] != NULL; ++datum)
+        if (is_name_bad(data[datum]->name)) {
+            fprintf(stderr, "data name %s is bad\n", data[datum]->name);
+            return 1;
+        }
+    for (size_t config = 0; configs[config] != NULL; ++config)
+        if (is_name_bad(configs[config]->name)) {
+            fprintf(stderr, "config name %s is bad\n", configs[config]->name);
+            return 1;
+        }
+    return 0;
+}
+
+/**
+ * Option parsing using getopt.
+ * When you add a new option update: long_options, long_extras, and
+ * short_options.
+ */
+
+/** Option variables filled by parse_args. */
+static char const* g_output = NULL;
+static char const* g_diff = NULL;
+static char const* g_cache = NULL;
+static char const* g_zstdcli = NULL;
+static char const* g_config = NULL;
+static char const* g_data = NULL;
+static char const* g_method = NULL;
+
+typedef enum {
+    required_option,
+    optional_option,
+    help_option,
+} option_type;
+
+/**
+ * Extra state that we need to keep per-option that we can't store in getopt.
+ */
+struct option_extra {
+    int id; /**< The short option name, used as an id. */
+    char const* help; /**< The help message. */
+    option_type opt_type; /**< The option type: required, optional, or help. */
+    char const** value; /**< The value to set or NULL if no_argument. */
+};
+
+/** The options. */
+static struct option long_options[] = {
+    {"cache", required_argument, NULL, 'c'},
+    {"output", required_argument, NULL, 'o'},
+    {"zstd", required_argument, NULL, 'z'},
+    {"config", required_argument, NULL, 128},
+    {"data", required_argument, NULL, 129},
+    {"method", required_argument, NULL, 130},
+    {"diff", required_argument, NULL, 'd'},
+    {"help", no_argument, NULL, 'h'},
+};
+
+static size_t const nargs = sizeof(long_options) / sizeof(long_options[0]);
+
+/** The extra info for the options. Must be in the same order as the options. */
+static struct option_extra long_extras[] = {
+    {'c', "the cache directory", required_option, &g_cache},
+    {'o', "write the results here", required_option, &g_output},
+    {'z', "zstd cli tool", required_option, &g_zstdcli},
+    {128, "use this config", optional_option, &g_config},
+    {129, "use this data", optional_option, &g_data},
+    {130, "use this method", optional_option, &g_method},
+    {'d', "compare the results to this file", optional_option, &g_diff},
+    {'h', "display this message", help_option, NULL},
+};
+
+/** The short options. Must correspond to the options. */
+static char const short_options[] = "c:d:ho:z:";
+
+/** Return the help string for the option type. */
+static char const* required_message(option_type opt_type) {
+    switch (opt_type) {
+        case required_option:
+            return "[required]";
+        case optional_option:
+            return "[optional]";
+        case help_option:
+            return "";
+        default:
+            assert(0);
+            return NULL;
+    }
+}
+
+/** Print the help for the program. */
+static void print_help(void) {
+    fprintf(stderr, "regression test runner\n");
+    size_t const nargs = sizeof(long_options) / sizeof(long_options[0]);
+    for (size_t i = 0; i < nargs; ++i) {
+        if (long_options[i].val < 128) {
+            /* Long / short  - help [option type] */
+            fprintf(
+                stderr,
+                "--%s / -%c \t- %s %s\n",
+                long_options[i].name,
+                long_options[i].val,
+                long_extras[i].help,
+                required_message(long_extras[i].opt_type));
+        } else {
+            /* Short / long  - help [option type] */
+            fprintf(
+                stderr,
+                "--%s      \t- %s %s\n",
+                long_options[i].name,
+                long_extras[i].help,
+                required_message(long_extras[i].opt_type));
+        }
+    }
+}
+
+/** Parse the arguments. Teturn 0 on success. Print help on failure. */
+static int parse_args(int argc, char** argv) {
+    int option_index = 0;
+    int c;
+
+    while (1) {
+        c = getopt_long(argc, argv, short_options, long_options, &option_index);
+        if (c == -1)
+            break;
+
+        int found = 0;
+        for (size_t i = 0; i < nargs; ++i) {
+            if (c == long_extras[i].id && long_extras[i].value != NULL) {
+                *long_extras[i].value = optarg;
+                found = 1;
+                break;
+            }
+        }
+        if (found)
+            continue;
+
+        switch (c) {
+            case 'h':
+            case '?':
+            default:
+                print_help();
+                return 1;
+        }
+    }
+
+    int bad = 0;
+    for (size_t i = 0; i < nargs; ++i) {
+        if (long_extras[i].opt_type != required_option)
+            continue;
+        if (long_extras[i].value == NULL)
+            continue;
+        if (*long_extras[i].value != NULL)
+            continue;
+        fprintf(
+            stderr,
+            "--%s is a required argument but is not set\n",
+            long_options[i].name);
+        bad = 1;
+    }
+    if (bad) {
+        fprintf(stderr, "\n");
+        print_help();
+        return 1;
+    }
+
+    return 0;
+}
+
+/** Helper macro to print to stderr and a file. */
+#define tprintf(file, ...)            \
+    do {                              \
+        fprintf(file, __VA_ARGS__);   \
+        fprintf(stderr, __VA_ARGS__); \
+    } while (0)
+/** Helper macro to flush stderr and a file. */
+#define tflush(file)    \
+    do {                \
+        fflush(file);   \
+        fflush(stderr); \
+    } while (0)
+
+void tprint_names(
+    FILE* results,
+    char const* data_name,
+    char const* config_name,
+    char const* method_name) {
+    int const data_padding = g_max_name_len - strlen(data_name);
+    int const config_padding = g_max_name_len - strlen(config_name);
+    int const method_padding = g_max_name_len - strlen(method_name);
+
+    tprintf(
+        results,
+        "%s, %*s%s, %*s%s, %*s",
+        data_name,
+        data_padding,
+        "",
+        config_name,
+        config_padding,
+        "",
+        method_name,
+        method_padding,
+        "");
+}
+
+/**
+ * Run all the regression tests and record the results table to results and
+ * stderr progressively.
+ */
+static int run_all(FILE* results) {
+    tprint_names(results, "Data", "Config", "Method");
+    tprintf(results, "Total compressed size\n");
+    for (size_t method = 0; methods[method] != NULL; ++method) {
+        if (g_method != NULL && strcmp(methods[method]->name, g_method))
+            continue;
+        for (size_t datum = 0; data[datum] != NULL; ++datum) {
+            if (g_data != NULL && strcmp(data[datum]->name, g_data))
+                continue;
+            /* Create the state common to all configs */
+            method_state_t* state = methods[method]->create(data[datum]);
+            for (size_t config = 0; configs[config] != NULL; ++config) {
+                if (g_config != NULL && strcmp(configs[config]->name, g_config))
+                    continue;
+                if (config_skip_data(configs[config], data[datum]))
+                    continue;
+                /* Print the result for the (method, data, config) tuple. */
+                result_t const result =
+                    methods[method]->compress(state, configs[config]);
+                if (result_is_skip(result))
+                    continue;
+                tprint_names(
+                    results,
+                    data[datum]->name,
+                    configs[config]->name,
+                    methods[method]->name);
+                if (result_is_error(result)) {
+                    tprintf(results, "%s\n", result_get_error_string(result));
+                } else {
+                    tprintf(
+                        results,
+                        "%llu\n",
+                        (unsigned long long)result_get_data(result).total_size);
+                }
+                tflush(results);
+            }
+            methods[method]->destroy(state);
+        }
+    }
+    return 0;
+}
+
+/** memcmp() the old results file and the new results file. */
+static int diff_results(char const* actual_file, char const* expected_file) {
+    data_buffer_t const actual = data_buffer_read(actual_file);
+    data_buffer_t const expected = data_buffer_read(expected_file);
+    int ret = 1;
+
+    if (actual.data == NULL) {
+        fprintf(stderr, "failed to open results '%s' for diff\n", actual_file);
+        goto out;
+    }
+    if (expected.data == NULL) {
+        fprintf(
+            stderr,
+            "failed to open previous results '%s' for diff\n",
+            expected_file);
+        goto out;
+    }
+
+    ret = data_buffer_compare(actual, expected);
+    if (ret != 0) {
+        fprintf(
+            stderr,
+            "actual results '%s' does not match expected results '%s'\n",
+            actual_file,
+            expected_file);
+    } else {
+        fprintf(stderr, "actual results match expected results\n");
+    }
+out:
+    data_buffer_free(actual);
+    data_buffer_free(expected);
+    return ret;
+}
+
+int main(int argc, char** argv) {
+    /* Parse args and validate modules. */
+    int ret = parse_args(argc, argv);
+    if (ret != 0)
+        return ret;
+
+    if (are_names_bad())
+        return 1;
+
+    /* Initialize modules. */
+    method_set_zstdcli(g_zstdcli);
+    ret = data_init(g_cache);
+    if (ret != 0) {
+        fprintf(stderr, "data_init() failed with error=%s\n", strerror(ret));
+        return 1;
+    }
+
+    /* Run the regression tests. */
+    ret = 1;
+    FILE* results = fopen(g_output, "w");
+    if (results == NULL) {
+        fprintf(stderr, "Failed to open the output file\n");
+        goto out;
+    }
+    ret = run_all(results);
+    fclose(results);
+
+    if (ret != 0)
+        goto out;
+
+    if (g_diff)
+        /* Diff the new results with the previous results. */
+        ret = diff_results(g_output, g_diff);
+
+out:
+    data_finish();
+    return ret;
+}
diff --git a/tests/roundTripCrash.c b/tests/roundTripCrash.c
index 90afcd4b2a8b..3f4ace8c9800 100644
--- a/tests/roundTripCrash.c
+++ b/tests/roundTripCrash.c
@@ -85,7 +85,7 @@ static size_t cctxParamRoundTripTest(void* resultBuff, size_t resultBuffCapacity
     ZSTD_CCtx* const cctx = ZSTD_createCCtx();
     ZSTD_CCtx_params* const cctxParams = ZSTD_createCCtxParams();
     ZSTD_inBuffer inBuffer = { srcBuff, srcBuffSize, 0 };
-    ZSTD_outBuffer outBuffer = {compressedBuff, compressedBuffCapacity, 0 };
+    ZSTD_outBuffer outBuffer = { compressedBuff, compressedBuffCapacity, 0 };
 
     static const int maxClevel = 19;
     size_t const hashLength = MIN(128, srcBuffSize);
@@ -93,15 +93,15 @@ static size_t cctxParamRoundTripTest(void* resultBuff, size_t resultBuffCapacity
     int const cLevel = h32 % maxClevel;
 
     /* Set parameters */
-    CHECK_Z( ZSTD_CCtxParam_setParameter(cctxParams, ZSTD_p_compressionLevel, cLevel) );
-    CHECK_Z( ZSTD_CCtxParam_setParameter(cctxParams, ZSTD_p_nbWorkers, 2) );
-    CHECK_Z( ZSTD_CCtxParam_setParameter(cctxParams, ZSTD_p_overlapSizeLog, 5) );
+    CHECK_Z( ZSTD_CCtxParam_setParameter(cctxParams, ZSTD_c_compressionLevel, cLevel) );
+    CHECK_Z( ZSTD_CCtxParam_setParameter(cctxParams, ZSTD_c_nbWorkers, 2) );
+    CHECK_Z( ZSTD_CCtxParam_setParameter(cctxParams, ZSTD_c_overlapLog, 5) );
 
 
     /* Apply parameters */
     CHECK_Z( ZSTD_CCtx_setParametersUsingCCtxParams(cctx, cctxParams) );
 
-    CHECK_Z (ZSTD_compress_generic(cctx, &outBuffer, &inBuffer, ZSTD_e_end) );
+    CHECK_Z (ZSTD_compressStream2(cctx, &outBuffer, &inBuffer, ZSTD_e_end) );
 
     ZSTD_freeCCtxParams(cctxParams);
     ZSTD_freeCCtx(cctx);
diff --git a/tests/seqgen.c b/tests/seqgen.c
index 8233ecee1cfd..06e3ed5ba0b0 100644
--- a/tests/seqgen.c
+++ b/tests/seqgen.c
@@ -17,7 +17,7 @@
 static const size_t kMatchBytes = 128;
 
 #define SEQ_rotl32(x,r) ((x << r) | (x >> (32 - r)))
-static BYTE SEQ_randByte(U32* src)
+static BYTE SEQ_randByte(unsigned* src)
 {
     static const U32 prime1 = 2654435761U;
     static const U32 prime2 = 2246822519U;
diff --git a/tests/symbols.c b/tests/symbols.c
index b37082131464..600d81670a55 100644
--- a/tests/symbols.c
+++ b/tests/symbols.c
@@ -89,7 +89,6 @@ static const void *symbols[] = {
   &ZSTD_sizeof_CStream,
   &ZSTD_createDStream_advanced,
   &ZSTD_initDStream_usingDict,
-  &ZSTD_setDStreamParameter,
   &ZSTD_initDStream_usingDDict,
   &ZSTD_resetDStream,
   &ZSTD_sizeof_DStream,
diff --git a/tests/zstreamtest.c b/tests/zstreamtest.c
index f47451a3c3d5..ac300890415f 100644
--- a/tests/zstreamtest.c
+++ b/tests/zstreamtest.c
@@ -46,7 +46,7 @@
 #define MB *(1U<<20)
 #define GB *(1U<<30)
 
-static const U32 nbTestsDefault = 10000;
+static const int nbTestsDefault = 10000;
 static const U32 g_cLevelMax_smallTests = 10;
 #define COMPRESSIBLE_NOISE_LENGTH (10 MB)
 #define FUZ_COMPRESSIBILITY_DEFAULT 50
@@ -84,7 +84,7 @@ static U64 g_clockTime = 0;
     @return : a 27 bits random value, from a 32-bits `seed`.
     `seed` is also modified */
 #define FUZ_rotl32(x,r) ((x << r) | (x >> (32 - r)))
-static unsigned int FUZ_rand(unsigned int* seedPtr)
+static U32 FUZ_rand(U32* seedPtr)
 {
     static const U32 prime2 = 2246822519U;
     U32 rand32 = *seedPtr;
@@ -100,7 +100,7 @@ static unsigned int FUZ_rand(unsigned int* seedPtr)
         DISPLAY("Error => ");                                \
         DISPLAY(__VA_ARGS__);                                \
         DISPLAY(" (seed %u, test nb %u, line %u) \n",        \
-                seed, testNb, __LINE__);                     \
+                (unsigned)seed, testNb, __LINE__);           \
         goto _output_error;                                  \
 }   }
 
@@ -175,11 +175,11 @@ static size_t SEQ_roundTrip(ZSTD_CCtx* cctx, ZSTD_DCtx* dctx,
     size_t cret;
 
     do {
-        ZSTD_outBuffer cout = {compressed, sizeof(compressed), 0};
-        ZSTD_inBuffer din = {compressed, 0, 0};
-        ZSTD_outBuffer dout = {uncompressed, 0, 0};
+        ZSTD_outBuffer cout = { compressed, sizeof(compressed), 0 };
+        ZSTD_inBuffer din   = { compressed, 0, 0 };
+        ZSTD_outBuffer dout = { uncompressed, 0, 0 };
 
-        cret = ZSTD_compress_generic(cctx, &cout, &cin, endOp);
+        cret = ZSTD_compressStream2(cctx, &cout, &cin, endOp);
         if (ZSTD_isError(cret))
             return cret;
 
@@ -223,19 +223,19 @@ static size_t SEQ_generateRoundTrip(ZSTD_CCtx* cctx, ZSTD_DCtx* dctx,
 
 static size_t getCCtxParams(ZSTD_CCtx* zc, ZSTD_parameters* savedParams)
 {
-    unsigned value;
-    CHECK_RET_Z(ZSTD_CCtx_getParameter(zc, ZSTD_p_windowLog, &savedParams->cParams.windowLog));
-    CHECK_RET_Z(ZSTD_CCtx_getParameter(zc, ZSTD_p_hashLog, &savedParams->cParams.hashLog));
-    CHECK_RET_Z(ZSTD_CCtx_getParameter(zc, ZSTD_p_chainLog, &savedParams->cParams.chainLog));
-    CHECK_RET_Z(ZSTD_CCtx_getParameter(zc, ZSTD_p_searchLog, &savedParams->cParams.searchLog));
-    CHECK_RET_Z(ZSTD_CCtx_getParameter(zc, ZSTD_p_minMatch, &savedParams->cParams.searchLength));
-    CHECK_RET_Z(ZSTD_CCtx_getParameter(zc, ZSTD_p_targetLength, &savedParams->cParams.targetLength));
-    CHECK_RET_Z(ZSTD_CCtx_getParameter(zc, ZSTD_p_compressionStrategy, &value));
+    int value;
+    CHECK_RET_Z(ZSTD_CCtx_getParameter(zc, ZSTD_c_windowLog, (int*)&savedParams->cParams.windowLog));
+    CHECK_RET_Z(ZSTD_CCtx_getParameter(zc, ZSTD_c_hashLog, (int*)&savedParams->cParams.hashLog));
+    CHECK_RET_Z(ZSTD_CCtx_getParameter(zc, ZSTD_c_chainLog, (int*)&savedParams->cParams.chainLog));
+    CHECK_RET_Z(ZSTD_CCtx_getParameter(zc, ZSTD_c_searchLog, (int*)&savedParams->cParams.searchLog));
+    CHECK_RET_Z(ZSTD_CCtx_getParameter(zc, ZSTD_c_minMatch, (int*)&savedParams->cParams.minMatch));
+    CHECK_RET_Z(ZSTD_CCtx_getParameter(zc, ZSTD_c_targetLength, (int*)&savedParams->cParams.targetLength));
+    CHECK_RET_Z(ZSTD_CCtx_getParameter(zc, ZSTD_c_strategy, &value));
     savedParams->cParams.strategy = value;
 
-    CHECK_RET_Z(ZSTD_CCtx_getParameter(zc, ZSTD_p_checksumFlag, &savedParams->fParams.checksumFlag));
-    CHECK_RET_Z(ZSTD_CCtx_getParameter(zc, ZSTD_p_contentSizeFlag, &savedParams->fParams.contentSizeFlag));
-    CHECK_RET_Z(ZSTD_CCtx_getParameter(zc, ZSTD_p_dictIDFlag, &value));
+    CHECK_RET_Z(ZSTD_CCtx_getParameter(zc, ZSTD_c_checksumFlag, &savedParams->fParams.checksumFlag));
+    CHECK_RET_Z(ZSTD_CCtx_getParameter(zc, ZSTD_c_contentSizeFlag, &savedParams->fParams.contentSizeFlag));
+    CHECK_RET_Z(ZSTD_CCtx_getParameter(zc, ZSTD_c_dictIDFlag, &value));
     savedParams->fParams.noDictIDFlag = !value;
     return 0;
 }
@@ -248,7 +248,7 @@ static U32 badParameters(ZSTD_CCtx* zc, ZSTD_parameters const savedParams)
     CHECK_RET(2, params.cParams.hashLog != savedParams.cParams.hashLog, "hashLog");
     CHECK_RET(3, params.cParams.chainLog != savedParams.cParams.chainLog, "chainLog");
     CHECK_RET(4, params.cParams.searchLog != savedParams.cParams.searchLog, "searchLog");
-    CHECK_RET(5, params.cParams.searchLength != savedParams.cParams.searchLength, "searchLength");
+    CHECK_RET(5, params.cParams.minMatch != savedParams.cParams.minMatch, "minMatch");
     CHECK_RET(6, params.cParams.targetLength != savedParams.cParams.targetLength, "targetLength");
 
     CHECK_RET(7, params.fParams.checksumFlag != savedParams.fParams.checksumFlag, "checksumFlag");
@@ -268,7 +268,7 @@ static int basicUnitTests(U32 seed, double compressibility)
     void* decodedBuffer = malloc(decodedBufferSize);
     size_t cSize;
     int testResult = 0;
-    U32 testNb = 1;
+    int testNb = 1;
     U32 coreSeed = 0;  /* this name to conform with CHECK_Z macro display */
     ZSTD_CStream* zc = ZSTD_createCStream();
     ZSTD_DStream* zd = ZSTD_createDStream();
@@ -309,7 +309,7 @@ static int basicUnitTests(U32 seed, double compressibility)
     if (inBuff.pos != inBuff.size) goto _output_error;   /* entire input should be consumed */
     { size_t const r = ZSTD_endStream(zc, &outBuff);
       if (r != 0) goto _output_error; }  /* error, or some data not flushed */
-    DISPLAYLEVEL(3, "OK (%u bytes)\n", (U32)outBuff.pos);
+    DISPLAYLEVEL(3, "OK (%u bytes)\n", (unsigned)outBuff.pos);
 
     /* generate skippable frame */
     MEM_writeLE32(compressedBuffer, ZSTD_MAGIC_SKIPPABLE_START);
@@ -331,7 +331,8 @@ static int basicUnitTests(U32 seed, double compressibility)
     { size_t const r = ZSTD_endStream(zc, &outBuff);
       if (r != 0) goto _output_error; }  /* error, or some data not flushed */
     cSize += outBuff.pos;
-    DISPLAYLEVEL(3, "OK (%u bytes : %.2f%%)\n", (U32)cSize, (double)cSize/COMPRESSIBLE_NOISE_LENGTH*100);
+    DISPLAYLEVEL(3, "OK (%u bytes : %.2f%%)\n",
+                    (unsigned)cSize, (double)cSize/COMPRESSIBLE_NOISE_LENGTH*100);
 
     /* context size functions */
     DISPLAYLEVEL(3, "test%3i : estimate CStream size : ", testNb++);
@@ -340,20 +341,20 @@ static int basicUnitTests(U32 seed, double compressibility)
         size_t const cdictSize = ZSTD_estimateCDictSize_advanced(dictSize, cParams, ZSTD_dlm_byCopy); /* uses ZSTD_initCStream_usingDict() */
         if (ZSTD_isError(cstreamSize)) goto _output_error;
         if (ZSTD_isError(cdictSize)) goto _output_error;
-        DISPLAYLEVEL(3, "OK (%u bytes) \n", (U32)(cstreamSize + cdictSize));
+        DISPLAYLEVEL(3, "OK (%u bytes) \n", (unsigned)(cstreamSize + cdictSize));
     }
 
     DISPLAYLEVEL(3, "test%3i : check actual CStream size : ", testNb++);
     {   size_t const s = ZSTD_sizeof_CStream(zc);
         if (ZSTD_isError(s)) goto _output_error;
-        DISPLAYLEVEL(3, "OK (%u bytes) \n", (U32)s);
+        DISPLAYLEVEL(3, "OK (%u bytes) \n", (unsigned)s);
     }
 
     /* Attempt bad compression parameters */
     DISPLAYLEVEL(3, "test%3i : use bad compression parameters : ", testNb++);
     {   size_t r;
         ZSTD_parameters params = ZSTD_getParams(1, 0, 0);
-        params.cParams.searchLength = 2;
+        params.cParams.minMatch = 2;
         r = ZSTD_initCStream_advanced(zc, NULL, 0, params, 0);
         if (!ZSTD_isError(r)) goto _output_error;
         DISPLAYLEVEL(3, "init error : %s \n", ZSTD_getErrorName(r));
@@ -369,7 +370,7 @@ static int basicUnitTests(U32 seed, double compressibility)
     outBuff.size = CNBufferSize;
     outBuff.pos = 0;
     {   size_t const r = ZSTD_decompressStream(zd, &outBuff, &inBuff);
-        DISPLAYLEVEL(5, " ( ZSTD_decompressStream => %u ) ", (U32)r);
+        DISPLAYLEVEL(5, " ( ZSTD_decompressStream => %u ) ", (unsigned)r);
         if (r != 0) goto _output_error;
     }
     if (outBuff.pos != 0) goto _output_error;   /* skippable frame output len is 0 */
@@ -379,7 +380,7 @@ static int basicUnitTests(U32 seed, double compressibility)
     inBuff2 = inBuff;
     DISPLAYLEVEL(3, "test%3i : decompress %u bytes : ", testNb++, COMPRESSIBLE_NOISE_LENGTH);
     ZSTD_initDStream_usingDict(zd, CNBuffer, dictSize);
-    CHECK_Z( ZSTD_setDStreamParameter(zd, DStream_p_maxWindowSize, 1000000000) );  /* large limit */
+    CHECK_Z( ZSTD_DCtx_setParameter(zd, ZSTD_d_windowLogMax, ZSTD_WINDOWLOG_LIMIT_DEFAULT+1) );  /* large limit */
     { size_t const remaining = ZSTD_decompressStream(zd, &outBuff, &inBuff);
       if (remaining != 0) goto _output_error; }  /* should reach end of frame == 0; otherwise, some data left, or an error */
     if (outBuff.pos != CNBufferSize) goto _output_error;   /* should regenerate the same amount */
@@ -409,18 +410,18 @@ static int basicUnitTests(U32 seed, double compressibility)
         const void* cStart = (char*)compressedBuffer + (skippableFrameSize + 8);
         size_t const gfhError = ZSTD_getFrameHeader(&fhi, cStart, cSize);
         if (gfhError!=0) goto _output_error;
-        DISPLAYLEVEL(5, " (windowSize : %u) ", (U32)fhi.windowSize);
+        DISPLAYLEVEL(5, " (windowSize : %u) ", (unsigned)fhi.windowSize);
         {   size_t const s = ZSTD_estimateDStreamSize(fhi.windowSize)
                             /* uses ZSTD_initDStream_usingDict() */
                            + ZSTD_estimateDDictSize(dictSize, ZSTD_dlm_byCopy);
             if (ZSTD_isError(s)) goto _output_error;
-            DISPLAYLEVEL(3, "OK (%u bytes) \n", (U32)s);
+            DISPLAYLEVEL(3, "OK (%u bytes) \n", (unsigned)s);
     }   }
 
     DISPLAYLEVEL(3, "test%3i : check actual DStream size : ", testNb++);
     { size_t const s = ZSTD_sizeof_DStream(zd);
       if (ZSTD_isError(s)) goto _output_error;
-      DISPLAYLEVEL(3, "OK (%u bytes) \n", (U32)s);
+      DISPLAYLEVEL(3, "OK (%u bytes) \n", (unsigned)s);
     }
 
     /* Decompression by small increment */
@@ -507,7 +508,7 @@ static int basicUnitTests(U32 seed, double compressibility)
       if (r != 0) goto _output_error; }  /* error, or some data not flushed */
     { unsigned long long origSize = ZSTD_findDecompressedSize(outBuff.dst, outBuff.pos);
       if ((size_t)origSize != CNBufferSize) goto _output_error; }  /* exact original size must be present */
-    DISPLAYLEVEL(3, "OK (%u bytes : %.2f%%)\n", (U32)cSize, (double)cSize/COMPRESSIBLE_NOISE_LENGTH*100);
+    DISPLAYLEVEL(3, "OK (%u bytes : %.2f%%)\n", (unsigned)cSize, (double)cSize/COMPRESSIBLE_NOISE_LENGTH*100);
 
     /* wrong _srcSize compression test */
     DISPLAYLEVEL(3, "test%3i : too large srcSize : %u bytes : ", testNb++, COMPRESSIBLE_NOISE_LENGTH-1);
@@ -598,7 +599,7 @@ static int basicUnitTests(U32 seed, double compressibility)
     DISPLAYLEVEL(3, "test%3i : digested dictionary : ", testNb++);
     {   ZSTD_CDict* const cdict = ZSTD_createCDict(dictionary.start, dictionary.filled, 1 /*byRef*/ );
         size_t const initError = ZSTD_initCStream_usingCDict(zc, cdict);
-        DISPLAYLEVEL(5, "ZSTD_initCStream_usingCDict result : %u ", (U32)initError);
+        DISPLAYLEVEL(5, "ZSTD_initCStream_usingCDict result : %u ", (unsigned)initError);
         if (ZSTD_isError(initError)) goto _output_error;
         outBuff.dst = compressedBuffer;
         outBuff.size = compressedBufferSize;
@@ -610,18 +611,18 @@ static int basicUnitTests(U32 seed, double compressibility)
         CHECK_Z( ZSTD_compressStream(zc, &outBuff, &inBuff) );
         if (inBuff.pos != inBuff.size) goto _output_error;   /* entire input should be consumed */
         {   size_t const r = ZSTD_endStream(zc, &outBuff);
-            DISPLAYLEVEL(5, "- ZSTD_endStream result : %u ", (U32)r);
+            DISPLAYLEVEL(5, "- ZSTD_endStream result : %u ", (unsigned)r);
             if (r != 0) goto _output_error;  /* error, or some data not flushed */
         }
         cSize = outBuff.pos;
         ZSTD_freeCDict(cdict);
-        DISPLAYLEVEL(3, "OK (%u bytes : %.2f%%)\n", (U32)cSize, (double)cSize/CNBufferSize*100);
+        DISPLAYLEVEL(3, "OK (%u bytes : %.2f%%)\n", (unsigned)cSize, (double)cSize/CNBufferSize*100);
     }
 
     DISPLAYLEVEL(3, "test%3i : check CStream size : ", testNb++);
     { size_t const s = ZSTD_sizeof_CStream(zc);
       if (ZSTD_isError(s)) goto _output_error;
-      DISPLAYLEVEL(3, "OK (%u bytes) \n", (U32)s);
+      DISPLAYLEVEL(3, "OK (%u bytes) \n", (unsigned)s);
     }
 
     DISPLAYLEVEL(4, "test%3i : check Dictionary ID : ", testNb++);
@@ -631,7 +632,7 @@ static int basicUnitTests(U32 seed, double compressibility)
     }
 
     /* DDict scenario */
-    DISPLAYLEVEL(3, "test%3i : decompress %u bytes with digested dictionary : ", testNb++, (U32)CNBufferSize);
+    DISPLAYLEVEL(3, "test%3i : decompress %u bytes with digested dictionary : ", testNb++, (unsigned)CNBufferSize);
     {   ZSTD_DDict* const ddict = ZSTD_createDDict(dictionary.start, dictionary.filled);
         size_t const initError = ZSTD_initDStream_usingDDict(zd, ddict);
         if (ZSTD_isError(initError)) goto _output_error;
@@ -649,16 +650,10 @@ static int basicUnitTests(U32 seed, double compressibility)
         DISPLAYLEVEL(3, "OK \n");
     }
 
-    /* test ZSTD_setDStreamParameter() resilience */
-    DISPLAYLEVEL(3, "test%3i : wrong parameter for ZSTD_setDStreamParameter(): ", testNb++);
-    { size_t const r = ZSTD_setDStreamParameter(zd, (ZSTD_DStreamParameter_e)999, 1);  /* large limit */
-      if (!ZSTD_isError(r)) goto _output_error; }
-    DISPLAYLEVEL(3, "OK \n");
-
     /* Memory restriction */
     DISPLAYLEVEL(3, "test%3i : maxWindowSize < frame requirement : ", testNb++);
     ZSTD_initDStream_usingDict(zd, CNBuffer, dictSize);
-    CHECK_Z( ZSTD_setDStreamParameter(zd, DStream_p_maxWindowSize, 1000) );  /* too small limit */
+    CHECK_Z( ZSTD_DCtx_setParameter(zd, ZSTD_d_windowLogMax, 10) );  /* too small limit */
     outBuff.dst = decodedBuffer;
     outBuff.size = CNBufferSize;
     outBuff.pos = 0;
@@ -668,7 +663,7 @@ static int basicUnitTests(U32 seed, double compressibility)
     { size_t const r = ZSTD_decompressStream(zd, &outBuff, &inBuff);
       if (!ZSTD_isError(r)) goto _output_error;  /* must fail : frame requires > 100 bytes */
       DISPLAYLEVEL(3, "OK (%s)\n", ZSTD_getErrorName(r)); }
-    ZSTD_DCtx_reset(zd);   /* leave zd in good shape for next tests */
+    ZSTD_DCtx_reset(zd, ZSTD_reset_session_and_parameters);   /* leave zd in good shape for next tests */
 
     DISPLAYLEVEL(3, "test%3i : dictionary source size and level : ", testNb++);
     {   ZSTD_DCtx* const dctx = ZSTD_createDCtx();
@@ -692,12 +687,12 @@ static int basicUnitTests(U32 seed, double compressibility)
                 inBuff.size = size;
                 inBuff.pos = 0;
                 CHECK_Z(ZSTD_CCtx_refCDict(cctx, cdict));
-                CHECK_Z(ZSTD_compress_generic(cctx, &outBuff, &inBuff, ZSTD_e_end));
+                CHECK_Z(ZSTD_compressStream2(cctx, &outBuff, &inBuff, ZSTD_e_end));
                 CHECK(badParameters(cctx, savedParams), "Bad CCtx params");
                 if (inBuff.pos != inBuff.size) goto _output_error;
                 {   ZSTD_outBuffer decOut = {decodedBuffer, size, 0};
                     ZSTD_inBuffer decIn = {outBuff.dst, outBuff.pos, 0};
-                    CHECK_Z( ZSTD_decompress_generic(dctx, &decOut, &decIn) );
+                    CHECK_Z( ZSTD_decompressStream(dctx, &decOut, &decIn) );
                     if (decIn.pos != decIn.size) goto _output_error;
                     if (decOut.pos != size) goto _output_error;
                     {   U64 const crcDec = XXH64(decOut.dst, decOut.pos, 0);
@@ -729,7 +724,7 @@ static int basicUnitTests(U32 seed, double compressibility)
           if (r != 0) goto _output_error; }  /* error, or some data not flushed */
         cSize = outBuff.pos;
         ZSTD_freeCDict(cdict);
-        DISPLAYLEVEL(3, "OK (%u bytes : %.2f%%)\n", (U32)cSize, (double)cSize/CNBufferSize*100);
+        DISPLAYLEVEL(3, "OK (%u bytes : %.2f%%)\n", (unsigned)cSize, (double)cSize/CNBufferSize*100);
     }
 
     DISPLAYLEVEL(3, "test%3i : try retrieving dictID from frame : ", testNb++);
@@ -752,10 +747,10 @@ static int basicUnitTests(U32 seed, double compressibility)
     inBuff.src = CNBuffer;
     inBuff.size = CNBufferSize;
     inBuff.pos = 0;
-    CHECK_Z( ZSTD_compress_generic(zc, &outBuff, &inBuff, ZSTD_e_end) );
+    CHECK_Z( ZSTD_compressStream2(zc, &outBuff, &inBuff, ZSTD_e_end) );
     if (inBuff.pos != inBuff.size) goto _output_error;  /* entire input should be consumed */
     cSize = outBuff.pos;
-    DISPLAYLEVEL(3, "OK (%u bytes : %.2f%%)\n", (U32)cSize, (double)cSize/CNBufferSize*100);
+    DISPLAYLEVEL(3, "OK (%u bytes : %.2f%%)\n", (unsigned)cSize, (double)cSize/CNBufferSize*100);
 
     DISPLAYLEVEL(3, "test%3i : decompress with ZSTD_DCtx_refPrefix : ", testNb++);
     CHECK_Z( ZSTD_DCtx_refPrefix(zd, dictionary.start, dictionary.filled) );
@@ -765,7 +760,7 @@ static int basicUnitTests(U32 seed, double compressibility)
     inBuff.src = compressedBuffer;
     inBuff.size = cSize;
     inBuff.pos = 0;
-    CHECK_Z( ZSTD_decompress_generic(zd, &outBuff, &inBuff) );
+    CHECK_Z( ZSTD_decompressStream(zd, &outBuff, &inBuff) );
     if (inBuff.pos != inBuff.size) goto _output_error;  /* entire input should be consumed */
     if (outBuff.pos != CNBufferSize) goto _output_error;  /* must regenerate whole input */
     DISPLAYLEVEL(3, "OK \n");
@@ -776,17 +771,17 @@ static int basicUnitTests(U32 seed, double compressibility)
         DISPLAYLEVEL(3, "OK (%s)\n", ZSTD_getErrorName(r));
     }
 
-    DISPLAYLEVEL(3, "test%3i : compress again with ZSTD_compress_generic : ", testNb++);
+    DISPLAYLEVEL(3, "test%3i : compress again with ZSTD_compressStream2 : ", testNb++);
     outBuff.dst = compressedBuffer;
     outBuff.size = compressedBufferSize;
     outBuff.pos = 0;
     inBuff.src = CNBuffer;
     inBuff.size = CNBufferSize;
     inBuff.pos = 0;
-    CHECK_Z( ZSTD_compress_generic(zc, &outBuff, &inBuff, ZSTD_e_end) );
+    CHECK_Z( ZSTD_compressStream2(zc, &outBuff, &inBuff, ZSTD_e_end) );
     if (inBuff.pos != inBuff.size) goto _output_error;  /* entire input should be consumed */
     cSize = outBuff.pos;
-    DISPLAYLEVEL(3, "OK (%u bytes : %.2f%%)\n", (U32)cSize, (double)cSize/CNBufferSize*100);
+    DISPLAYLEVEL(3, "OK (%u bytes : %.2f%%)\n", (unsigned)cSize, (double)cSize/CNBufferSize*100);
 
     DISPLAYLEVEL(3, "test%3i : decompress without dictionary (should work): ", testNb++);
     CHECK_Z( ZSTD_decompress(decodedBuffer, CNBufferSize, compressedBuffer, cSize) );
@@ -842,7 +837,7 @@ static int basicUnitTests(U32 seed, double compressibility)
     /* Basic multithreading compression test */
     DISPLAYLEVEL(3, "test%3i : compress %u bytes with multiple threads : ", testNb++, COMPRESSIBLE_NOISE_LENGTH);
     {   ZSTD_parameters const params = ZSTD_getParams(1, 0, 0);
-        unsigned jobSize;
+        int jobSize;
         CHECK_Z( ZSTDMT_getMTCtxParameter(mtctx, ZSTDMT_p_jobSize, &jobSize));
         CHECK(jobSize != 0, "job size non-zero");
         CHECK_Z( ZSTDMT_initCStream_advanced(mtctx, CNBuffer, dictSize, params, CNBufferSize) );
@@ -873,10 +868,10 @@ static int basicUnitTests(U32 seed, double compressibility)
         size_t const start = jobSize + (offset-1);
         const BYTE* const srcToCopy = (const BYTE*)CNBuffer + start;
         BYTE* const dst = (BYTE*)CNBuffer + start - offset;
-        DISPLAYLEVEL(3, "test%3i : compress %u bytes with multiple threads + dictionary : ", testNb++, (U32)srcSize);
-        CHECK_Z( ZSTD_CCtx_setParameter(zc, ZSTD_p_compressionLevel, 3) );
-        CHECK_Z( ZSTD_CCtx_setParameter(zc, ZSTD_p_nbWorkers, nbWorkers) );
-        CHECK_Z( ZSTD_CCtx_setParameter(zc, ZSTD_p_jobSize, jobSize) );
+        DISPLAYLEVEL(3, "test%3i : compress %u bytes with multiple threads + dictionary : ", testNb++, (unsigned)srcSize);
+        CHECK_Z( ZSTD_CCtx_setParameter(zc, ZSTD_c_compressionLevel, 3) );
+        CHECK_Z( ZSTD_CCtx_setParameter(zc, ZSTD_c_nbWorkers, nbWorkers) );
+        CHECK_Z( ZSTD_CCtx_setParameter(zc, ZSTD_c_jobSize, jobSize) );
         assert(start > offset);
         assert(start + segLength < COMPRESSIBLE_NOISE_LENGTH);
         memcpy(dst, srcToCopy, segLength);   /* create a long repetition at long distance for job 2 */
@@ -891,7 +886,7 @@ static int basicUnitTests(U32 seed, double compressibility)
         ZSTD_CDict* const cdict = ZSTD_createCDict_advanced(dictionary.start, dictionary.filled, ZSTD_dlm_byRef, ZSTD_dct_fullDict, cParams, ZSTD_defaultCMem);
         DISPLAYLEVEL(5, "cParams.windowLog = %u : ", cParams.windowLog);
         CHECK_Z( ZSTD_CCtx_refCDict(zc, cdict) );
-        CHECK_Z( ZSTD_compress_generic(zc, &outBuff, &inBuff, ZSTD_e_end) );
+        CHECK_Z( ZSTD_compressStream2(zc, &outBuff, &inBuff, ZSTD_e_end) );
         CHECK_Z( ZSTD_CCtx_refCDict(zc, NULL) );  /* do not keep a reference to cdict, as its lifetime ends */
         ZSTD_freeCDict(cdict);
     }
@@ -903,7 +898,7 @@ static int basicUnitTests(U32 seed, double compressibility)
     {   ZSTD_DStream* const dstream = ZSTD_createDCtx();
         ZSTD_frameHeader zfh;
         ZSTD_getFrameHeader(&zfh, compressedBuffer, cSize);
-        DISPLAYLEVEL(5, "frame windowsize = %u : ", (U32)zfh.windowSize);
+        DISPLAYLEVEL(5, "frame windowsize = %u : ", (unsigned)zfh.windowSize);
         outBuff.dst = decodedBuffer;
         outBuff.size = CNBufferSize;
         outBuff.pos = 0;
@@ -936,11 +931,11 @@ static int basicUnitTests(U32 seed, double compressibility)
 
         if (!cdict || !ddict) goto _output_error;
 
-        ZSTD_CCtx_reset(zc);
+        ZSTD_CCtx_reset(zc, ZSTD_reset_session_only);
         ZSTD_resetDStream(zd);
         CHECK_Z(ZSTD_CCtx_refCDict(zc, cdict));
         CHECK_Z(ZSTD_initDStream_usingDDict(zd, ddict));
-        CHECK_Z(ZSTD_setDStreamParameter(zd, DStream_p_maxWindowSize, 1U << kMaxWindowLog));
+        CHECK_Z(ZSTD_DCtx_setParameter(zd, ZSTD_d_windowLogMax, kMaxWindowLog));
         /* Test all values < 300 */
         for (value = 0; value < 300; ++value) {
             for (type = (SEQ_gen_type)0; type < SEQ_gen_max; ++type) {
@@ -968,12 +963,12 @@ static int basicUnitTests(U32 seed, double compressibility)
     DISPLAYLEVEL(3, "OK \n");
 
     DISPLAYLEVEL(3, "test%3i : ZSTD_initCStream_srcSize sets requestedParams : ", testNb++);
-    {   unsigned level;
+    {   int level;
         CHECK_Z(ZSTD_initCStream_srcSize(zc, 11, ZSTD_CONTENTSIZE_UNKNOWN));
-        CHECK_Z(ZSTD_CCtx_getParameter(zc, ZSTD_p_compressionLevel, &level));
+        CHECK_Z(ZSTD_CCtx_getParameter(zc, ZSTD_c_compressionLevel, &level));
         CHECK(level != 11, "Compression level does not match");
         ZSTD_resetCStream(zc, ZSTD_CONTENTSIZE_UNKNOWN);
-        CHECK_Z(ZSTD_CCtx_getParameter(zc, ZSTD_p_compressionLevel, &level));
+        CHECK_Z(ZSTD_CCtx_getParameter(zc, ZSTD_c_compressionLevel, &level));
         CHECK(level != 11, "Compression level does not match");
     }
     DISPLAYLEVEL(3, "OK \n");
@@ -1083,17 +1078,16 @@ static int basicUnitTests(U32 seed, double compressibility)
         int remainingInput = 256 * 1024;
         int offset;
 
-        ZSTD_CCtx_reset(zc);
-        CHECK_Z(ZSTD_CCtx_resetParameters(zc));
+        CHECK_Z(ZSTD_CCtx_reset(zc, ZSTD_reset_session_and_parameters));
         CHECK_Z(ZSTD_CCtx_refCDict(zc, cdict));
-        CHECK_Z(ZSTD_CCtx_setParameter(zc, ZSTD_p_checksumFlag, 1));
+        CHECK_Z(ZSTD_CCtx_setParameter(zc, ZSTD_c_checksumFlag, 1));
         /* Write a bunch of 6 byte blocks */
         while (remainingInput > 0) {
           char testBuffer[6] = "\xAA\xAA\xAA\xAA\xAA\xAA";
           const size_t kSmallBlockSize = sizeof(testBuffer);
           ZSTD_inBuffer in = {testBuffer, kSmallBlockSize, 0};
 
-          CHECK_Z(ZSTD_compress_generic(zc, &out, &in, ZSTD_e_flush));
+          CHECK_Z(ZSTD_compressStream2(zc, &out, &in, ZSTD_e_flush));
           CHECK(in.pos != in.size, "input not fully consumed");
           remainingInput -= kSmallBlockSize;
         }
@@ -1101,7 +1095,7 @@ static int basicUnitTests(U32 seed, double compressibility)
         for (offset = 1024; offset >= 0; offset -= 128) {
           ZSTD_inBuffer in = {dictionary.start + offset, 128, 0};
           ZSTD_EndDirective flush = offset > 0 ? ZSTD_e_continue : ZSTD_e_end;
-          CHECK_Z(ZSTD_compress_generic(zc, &out, &in, flush));
+          CHECK_Z(ZSTD_compressStream2(zc, &out, &in, flush));
           CHECK(in.pos != in.size, "input not fully consumed");
         }
         /* Ensure decompression works */
@@ -1139,10 +1133,10 @@ static size_t findDiff(const void* buf1, const void* buf2, size_t max)
         if (b1[u] != b2[u]) break;
     }
     if (u==max) {
-        DISPLAY("=> No difference detected within %u bytes \n", (U32)max);
+        DISPLAY("=> No difference detected within %u bytes \n", (unsigned)max);
         return u;
     }
-    DISPLAY("Error at position %u / %u \n", (U32)u, (U32)max);
+    DISPLAY("Error at position %u / %u \n", (unsigned)u, (unsigned)max);
     if (u>=3)
         DISPLAY(" %02X %02X %02X ",
                 b1[u-3], b1[u-2], b1[u-1]);
@@ -1175,7 +1169,7 @@ static U32 FUZ_randomClampedLength(U32* seed, U32 minVal, U32 maxVal)
     return (U32)((FUZ_rand(seed) % mod) + minVal);
 }
 
-static int fuzzerTests(U32 seed, U32 nbTests, unsigned startTest, double compressibility, int bigTests)
+static int fuzzerTests(U32 seed, unsigned nbTests, unsigned startTest, double compressibility, int bigTests)
 {
     U32 const maxSrcLog = bigTests ? 24 : 22;
     static const U32 maxSampleLog = 19;
@@ -1188,7 +1182,7 @@ static int fuzzerTests(U32 seed, U32 nbTests, unsigned startTest, double compres
     size_t const dstBufferSize = srcBufferSize;
     BYTE*  const dstBuffer = (BYTE*)malloc (dstBufferSize);
     U32 result = 0;
-    U32 testNb = 0;
+    unsigned testNb = 0;
     U32 coreSeed = seed;
     ZSTD_CStream* zc = ZSTD_createCStream();   /* will be re-created sometimes */
     ZSTD_DStream* zd = ZSTD_createDStream();   /* will be re-created sometimes */
@@ -1369,7 +1363,7 @@ static int fuzzerTests(U32 seed, U32 nbTests, unsigned startTest, double compres
             }
             CHECK (decompressionResult != 0, "frame not fully decoded");
             CHECK (outBuff.pos != totalTestSize, "decompressed data : wrong size (%u != %u)",
-                    (U32)outBuff.pos, (U32)totalTestSize);
+                    (unsigned)outBuff.pos, (unsigned)totalTestSize);
             CHECK (inBuff.pos != cSize, "compressed data should be fully read")
             {   U64 const crcDest = XXH64(dstBuffer, totalTestSize, 0);
                 if (crcDest!=crcOrig) findDiff(copyBuffer, dstBuffer, totalTestSize);
@@ -1427,7 +1421,7 @@ static int fuzzerTests(U32 seed, U32 nbTests, unsigned startTest, double compres
 
 
 /* fuzzing ZSTDMT_* interface */
-static int fuzzerTests_MT(U32 seed, U32 nbTests, unsigned startTest,
+static int fuzzerTests_MT(U32 seed, int nbTests, int startTest,
                           double compressibility, int bigTests)
 {
     const U32 maxSrcLog = bigTests ? 24 : 22;
@@ -1441,9 +1435,9 @@ static int fuzzerTests_MT(U32 seed, U32 nbTests, unsigned startTest,
     size_t const dstBufferSize = srcBufferSize;
     BYTE*  const dstBuffer = (BYTE*)malloc (dstBufferSize);
     U32 result = 0;
-    U32 testNb = 0;
+    int testNb = 0;
     U32 coreSeed = seed;
-    U32 nbThreads = 2;
+    int nbThreads = 2;
     ZSTDMT_CCtx* zc = ZSTDMT_createCCtx(nbThreads);   /* will be reset sometimes */
     ZSTD_DStream* zd = ZSTD_createDStream();   /* will be reset sometimes */
     ZSTD_DStream* const zd_noise = ZSTD_createDStream();
@@ -1471,7 +1465,7 @@ static int fuzzerTests_MT(U32 seed, U32 nbTests, unsigned startTest,
     RDG_genBuffer(cNoiseBuffer[4], srcBufferSize, 1.00, 0., coreSeed);    /* sparse content */
     memset(copyBuffer, 0x65, copyBufferSize);                             /* make copyBuffer considered initialized */
     ZSTD_initDStream_usingDict(zd, NULL, 0);  /* ensure at least one init */
-    DISPLAYLEVEL(6, "Creating initial context with %u threads \n", nbThreads);
+    DISPLAYLEVEL(6, "Creating initial context with %i threads \n", nbThreads);
 
     /* catch up testNb */
     for (testNb=1; testNb < startTest; testNb++)
@@ -1549,13 +1543,13 @@ static int fuzzerTests_MT(U32 seed, U32 nbTests, unsigned startTest,
                 {   U64 const pledgedSrcSize = (FUZ_rand(&lseed) & 3) ? ZSTD_CONTENTSIZE_UNKNOWN : maxTestSize;
                     ZSTD_parameters params = ZSTD_getParams(cLevel, pledgedSrcSize, dictSize);
                     DISPLAYLEVEL(5, "Init with windowLog = %u, pledgedSrcSize = %u, dictSize = %u \n",
-                        params.cParams.windowLog, (U32)pledgedSrcSize, (U32)dictSize);
+                        params.cParams.windowLog, (unsigned)pledgedSrcSize, (unsigned)dictSize);
                     params.fParams.checksumFlag = FUZ_rand(&lseed) & 1;
                     params.fParams.noDictIDFlag = FUZ_rand(&lseed) & 1;
                     params.fParams.contentSizeFlag = FUZ_rand(&lseed) & 1;
                     DISPLAYLEVEL(5, "checksumFlag : %u \n", params.fParams.checksumFlag);
-                    CHECK_Z( ZSTDMT_setMTCtxParameter(zc, ZSTDMT_p_overlapSectionLog, FUZ_rand(&lseed) % 12) );
-                    CHECK_Z( ZSTDMT_setMTCtxParameter(zc, ZSTDMT_p_jobSize, FUZ_rand(&lseed) % (2*maxTestSize+1)) );   /* custome job size */
+                    CHECK_Z( ZSTDMT_setMTCtxParameter(zc, ZSTDMT_p_overlapLog, FUZ_rand(&lseed) % 12) );
+                    CHECK_Z( ZSTDMT_setMTCtxParameter(zc, ZSTDMT_p_jobSize, FUZ_rand(&lseed) % (2*maxTestSize+1)) );   /* custom job size */
                     CHECK_Z( ZSTDMT_initCStream_advanced(zc, dict, dictSize, params, pledgedSrcSize) );
         }   }   }
 
@@ -1573,9 +1567,9 @@ static int fuzzerTests_MT(U32 seed, U32 nbTests, unsigned startTest,
                     ZSTD_inBuffer inBuff = { srcBuffer+srcStart, srcSize, 0 };
                     outBuff.size = outBuff.pos + dstBuffSize;
 
-                    DISPLAYLEVEL(6, "Sending %u bytes to compress \n", (U32)srcSize);
+                    DISPLAYLEVEL(6, "Sending %u bytes to compress \n", (unsigned)srcSize);
                     CHECK_Z( ZSTDMT_compressStream(zc, &outBuff, &inBuff) );
-                    DISPLAYLEVEL(6, "%u bytes read by ZSTDMT_compressStream \n", (U32)inBuff.pos);
+                    DISPLAYLEVEL(6, "%u bytes read by ZSTDMT_compressStream \n", (unsigned)inBuff.pos);
 
                     XXH64_update(&xxhState, srcBuffer+srcStart, inBuff.pos);
                     memcpy(copyBuffer+totalTestSize, srcBuffer+srcStart, inBuff.pos);
@@ -1588,10 +1582,10 @@ static int fuzzerTests_MT(U32 seed, U32 nbTests, unsigned startTest,
                     size_t const adjustedDstSize = MIN(cBufferSize - cSize, randomDstSize);
                     size_t const previousPos = outBuff.pos;
                     outBuff.size = outBuff.pos + adjustedDstSize;
-                    DISPLAYLEVEL(5, "Flushing into dst buffer of size %u \n", (U32)adjustedDstSize);
+                    DISPLAYLEVEL(5, "Flushing into dst buffer of size %u \n", (unsigned)adjustedDstSize);
                     CHECK_Z( ZSTDMT_flushStream(zc, &outBuff) );
                     assert(outBuff.pos >= previousPos);
-                    DISPLAYLEVEL(6, "%u bytes flushed by ZSTDMT_flushStream \n", (U32)(outBuff.pos-previousPos));
+                    DISPLAYLEVEL(6, "%u bytes flushed by ZSTDMT_flushStream \n", (unsigned)(outBuff.pos-previousPos));
             }   }
 
             /* final frame epilogue */
@@ -1601,17 +1595,17 @@ static int fuzzerTests_MT(U32 seed, U32 nbTests, unsigned startTest,
                     size_t const adjustedDstSize = MIN(cBufferSize - cSize, randomDstSize);
                     size_t const previousPos = outBuff.pos;
                     outBuff.size = outBuff.pos + adjustedDstSize;
-                    DISPLAYLEVEL(5, "Ending into dst buffer of size %u \n", (U32)adjustedDstSize);
+                    DISPLAYLEVEL(5, "Ending into dst buffer of size %u \n", (unsigned)adjustedDstSize);
                     remainingToFlush = ZSTDMT_endStream(zc, &outBuff);
                     CHECK (ZSTD_isError(remainingToFlush), "ZSTDMT_endStream error : %s", ZSTD_getErrorName(remainingToFlush));
                     assert(outBuff.pos >= previousPos);
-                    DISPLAYLEVEL(6, "%u bytes flushed by ZSTDMT_endStream \n", (U32)(outBuff.pos-previousPos));
-                    DISPLAYLEVEL(5, "endStream : remainingToFlush : %u \n", (U32)remainingToFlush);
+                    DISPLAYLEVEL(6, "%u bytes flushed by ZSTDMT_endStream \n", (unsigned)(outBuff.pos-previousPos));
+                    DISPLAYLEVEL(5, "endStream : remainingToFlush : %u \n", (unsigned)remainingToFlush);
             }   }
             crcOrig = XXH64_digest(&xxhState);
             cSize = outBuff.pos;
             DISPLAYLEVEL(5, "Frame completed : %u bytes compressed into %u bytes \n",
-                            (U32)totalTestSize, (U32)cSize);
+                            (unsigned)totalTestSize, (unsigned)cSize);
         }
 
         /* multi - fragments decompression test */
@@ -1632,7 +1626,7 @@ static int fuzzerTests_MT(U32 seed, U32 nbTests, unsigned startTest,
                 inBuff.size = inBuff.pos + readCSrcSize;
                 outBuff.size = outBuff.pos + dstBuffSize;
                 DISPLAYLEVEL(6, "ZSTD_decompressStream input %u bytes into outBuff %u bytes \n",
-                                (U32)readCSrcSize, (U32)dstBuffSize);
+                                (unsigned)readCSrcSize, (unsigned)dstBuffSize);
                 decompressionResult = ZSTD_decompressStream(zd, &outBuff, &inBuff);
                 if (ZSTD_isError(decompressionResult)) {
                     DISPLAY("ZSTD_decompressStream error : %s \n", ZSTD_getErrorName(decompressionResult));
@@ -1640,10 +1634,14 @@ static int fuzzerTests_MT(U32 seed, U32 nbTests, unsigned startTest,
                 }
                 CHECK (ZSTD_isError(decompressionResult), "decompression error : %s", ZSTD_getErrorName(decompressionResult));
                 DISPLAYLEVEL(6, "total ingested (inBuff.pos) = %u and produced (outBuff.pos) = %u \n",
-                                (U32)inBuff.pos, (U32)outBuff.pos);
+                                (unsigned)inBuff.pos, (unsigned)outBuff.pos);
             }
-            CHECK (outBuff.pos != totalTestSize, "decompressed data : wrong size (%u != %u)", (U32)outBuff.pos, (U32)totalTestSize);
-            CHECK (inBuff.pos != cSize, "compressed data should be fully read (%u != %u)", (U32)inBuff.pos, (U32)cSize);
+            CHECK (outBuff.pos != totalTestSize,
+                    "decompressed data : wrong size (%u != %u)",
+                    (unsigned)outBuff.pos, (unsigned)totalTestSize );
+            CHECK (inBuff.pos != cSize,
+                    "compressed data should be fully read (%u != %u)",
+                    (unsigned)inBuff.pos, (unsigned)cSize );
             {   U64 const crcDest = XXH64(dstBuffer, totalTestSize, 0);
                 if (crcDest!=crcOrig) findDiff(copyBuffer, dstBuffer, totalTestSize);
                 CHECK (crcDest!=crcOrig, "decompressed data corrupted");
@@ -1712,7 +1710,7 @@ static size_t setCCtxParameter(ZSTD_CCtx* zc, ZSTD_CCtx_params* cctxParams,
 }
 
 /* Tests for ZSTD_compress_generic() API */
-static int fuzzerTests_newAPI(U32 seed, U32 nbTests, unsigned startTest,
+static int fuzzerTests_newAPI(U32 seed, int nbTests, int startTest,
                               double compressibility, int bigTests)
 {
     U32 const maxSrcLog = bigTests ? 24 : 22;
@@ -1726,7 +1724,7 @@ static int fuzzerTests_newAPI(U32 seed, U32 nbTests, unsigned startTest,
     size_t const dstBufferSize = srcBufferSize;
     BYTE*  const dstBuffer = (BYTE*)malloc (dstBufferSize);
     U32 result = 0;
-    U32 testNb = 0;
+    int testNb = 0;
     U32 coreSeed = seed;
     ZSTD_CCtx* zc = ZSTD_createCCtx();   /* will be reset sometimes */
     ZSTD_DStream* zd = ZSTD_createDStream();   /* will be reset sometimes */
@@ -1824,7 +1822,7 @@ static int fuzzerTests_newAPI(U32 seed, U32 nbTests, unsigned startTest,
             if (maxTestSize >= srcBufferSize) maxTestSize = srcBufferSize-1;
             {   int const compressionLevel = (FUZ_rand(&lseed) % 5) + 1;
                 DISPLAYLEVEL(5, "t%u : compression level : %i \n", testNb, compressionLevel);
-                CHECK_Z (setCCtxParameter(zc, cctxParams, ZSTD_p_compressionLevel, compressionLevel, opaqueAPI) );
+                CHECK_Z (setCCtxParameter(zc, cctxParams, ZSTD_c_compressionLevel, compressionLevel, opaqueAPI) );
             }
         } else {
             U32 const testLog = FUZ_rand(&lseed) % maxSrcLog;
@@ -1833,10 +1831,10 @@ static int fuzzerTests_newAPI(U32 seed, U32 nbTests, unsigned startTest,
                                (ZSTD_maxCLevel() -
                                (MAX(testLog, dictLog) / 2))) +
                                1;
-            U32 const cLevel = MIN(cLevelCandidate, cLevelMax);
-            DISPLAYLEVEL(5, "t%u: base cLevel : %u \n", testNb, cLevel);
+            int const cLevel = MIN(cLevelCandidate, cLevelMax);
+            DISPLAYLEVEL(5, "t%i: base cLevel : %u \n", testNb, cLevel);
             maxTestSize = FUZ_rLogLength(&lseed, testLog);
-            DISPLAYLEVEL(5, "t%u: maxTestSize : %u \n", testNb, (U32)maxTestSize);
+            DISPLAYLEVEL(5, "t%i: maxTestSize : %u \n", testNb, (unsigned)maxTestSize);
             oldTestLog = testLog;
             /* random dictionary selection */
             dictSize  = ((FUZ_rand(&lseed)&63)==1) ? FUZ_rLogLength(&lseed, dictLog) : 0;
@@ -1858,47 +1856,47 @@ static int fuzzerTests_newAPI(U32 seed, U32 nbTests, unsigned startTest,
                 cParams.chainLog += (FUZ_rand(&lseed) & 3) - 1;
                 cParams.searchLog += (FUZ_rand(&lseed) & 3) - 1;
                 cParams.searchLog = MIN(searchLogMax, cParams.searchLog);
-                cParams.searchLength += (FUZ_rand(&lseed) & 3) - 1;
+                cParams.minMatch += (FUZ_rand(&lseed) & 3) - 1;
                 cParams.targetLength = (U32)((cParams.targetLength + 1 ) * (0.5 + ((double)(FUZ_rand(&lseed) & 127) / 128)));
                 cParams = ZSTD_adjustCParams(cParams, pledgedSrcSize, dictSize);
 
                 if (FUZ_rand(&lseed) & 1) {
                     DISPLAYLEVEL(5, "t%u: windowLog : %u \n", testNb, cParams.windowLog);
-                    CHECK_Z( setCCtxParameter(zc, cctxParams, ZSTD_p_windowLog, cParams.windowLog, opaqueAPI) );
+                    CHECK_Z( setCCtxParameter(zc, cctxParams, ZSTD_c_windowLog, cParams.windowLog, opaqueAPI) );
                     assert(cParams.windowLog >= ZSTD_WINDOWLOG_MIN);   /* guaranteed by ZSTD_adjustCParams() */
                     windowLogMalus = (cParams.windowLog - ZSTD_WINDOWLOG_MIN) / 5;
                 }
                 if (FUZ_rand(&lseed) & 1) {
                     DISPLAYLEVEL(5, "t%u: hashLog : %u \n", testNb, cParams.hashLog);
-                    CHECK_Z( setCCtxParameter(zc, cctxParams, ZSTD_p_hashLog, cParams.hashLog, opaqueAPI) );
+                    CHECK_Z( setCCtxParameter(zc, cctxParams, ZSTD_c_hashLog, cParams.hashLog, opaqueAPI) );
                 }
                 if (FUZ_rand(&lseed) & 1) {
                     DISPLAYLEVEL(5, "t%u: chainLog : %u \n", testNb, cParams.chainLog);
-                    CHECK_Z( setCCtxParameter(zc, cctxParams, ZSTD_p_chainLog, cParams.chainLog, opaqueAPI) );
+                    CHECK_Z( setCCtxParameter(zc, cctxParams, ZSTD_c_chainLog, cParams.chainLog, opaqueAPI) );
                 }
-                if (FUZ_rand(&lseed) & 1) CHECK_Z( setCCtxParameter(zc, cctxParams, ZSTD_p_searchLog, cParams.searchLog, opaqueAPI) );
-                if (FUZ_rand(&lseed) & 1) CHECK_Z( setCCtxParameter(zc, cctxParams, ZSTD_p_minMatch, cParams.searchLength, opaqueAPI) );
-                if (FUZ_rand(&lseed) & 1) CHECK_Z( setCCtxParameter(zc, cctxParams, ZSTD_p_targetLength, cParams.targetLength, opaqueAPI) );
+                if (FUZ_rand(&lseed) & 1) CHECK_Z( setCCtxParameter(zc, cctxParams, ZSTD_c_searchLog, cParams.searchLog, opaqueAPI) );
+                if (FUZ_rand(&lseed) & 1) CHECK_Z( setCCtxParameter(zc, cctxParams, ZSTD_c_minMatch, cParams.minMatch, opaqueAPI) );
+                if (FUZ_rand(&lseed) & 1) CHECK_Z( setCCtxParameter(zc, cctxParams, ZSTD_c_targetLength, cParams.targetLength, opaqueAPI) );
 
                 /* mess with long distance matching parameters */
                 if (bigTests) {
-                    if (FUZ_rand(&lseed) & 1) CHECK_Z( setCCtxParameter(zc, cctxParams, ZSTD_p_enableLongDistanceMatching, FUZ_rand(&lseed) & 63, opaqueAPI) );
-                    if (FUZ_rand(&lseed) & 3) CHECK_Z( setCCtxParameter(zc, cctxParams, ZSTD_p_ldmHashLog, FUZ_randomClampedLength(&lseed, ZSTD_HASHLOG_MIN, 23), opaqueAPI) );
-                    if (FUZ_rand(&lseed) & 3) CHECK_Z( setCCtxParameter(zc, cctxParams, ZSTD_p_ldmMinMatch, FUZ_randomClampedLength(&lseed, ZSTD_LDM_MINMATCH_MIN, ZSTD_LDM_MINMATCH_MAX), opaqueAPI) );
-                    if (FUZ_rand(&lseed) & 3) CHECK_Z( setCCtxParameter(zc, cctxParams, ZSTD_p_ldmBucketSizeLog, FUZ_randomClampedLength(&lseed, 0, ZSTD_LDM_BUCKETSIZELOG_MAX), opaqueAPI) );
-                    if (FUZ_rand(&lseed) & 3) CHECK_Z( setCCtxParameter(zc, cctxParams, ZSTD_p_ldmHashEveryLog, FUZ_randomClampedLength(&lseed, 0, ZSTD_WINDOWLOG_MAX - ZSTD_HASHLOG_MIN), opaqueAPI) );
+                    if (FUZ_rand(&lseed) & 1) CHECK_Z( setCCtxParameter(zc, cctxParams, ZSTD_c_enableLongDistanceMatching, FUZ_rand(&lseed) & 63, opaqueAPI) );
+                    if (FUZ_rand(&lseed) & 3) CHECK_Z( setCCtxParameter(zc, cctxParams, ZSTD_c_ldmHashLog, FUZ_randomClampedLength(&lseed, ZSTD_HASHLOG_MIN, 23), opaqueAPI) );
+                    if (FUZ_rand(&lseed) & 3) CHECK_Z( setCCtxParameter(zc, cctxParams, ZSTD_c_ldmMinMatch, FUZ_randomClampedLength(&lseed, ZSTD_LDM_MINMATCH_MIN, ZSTD_LDM_MINMATCH_MAX), opaqueAPI) );
+                    if (FUZ_rand(&lseed) & 3) CHECK_Z( setCCtxParameter(zc, cctxParams, ZSTD_c_ldmBucketSizeLog, FUZ_randomClampedLength(&lseed, ZSTD_LDM_BUCKETSIZELOG_MIN, ZSTD_LDM_BUCKETSIZELOG_MAX), opaqueAPI) );
+                    if (FUZ_rand(&lseed) & 3) CHECK_Z( setCCtxParameter(zc, cctxParams, ZSTD_c_ldmHashRateLog, FUZ_randomClampedLength(&lseed, ZSTD_LDM_HASHRATELOG_MIN, ZSTD_LDM_HASHRATELOG_MAX), opaqueAPI) );
                 }
 
                 /* mess with frame parameters */
                 if (FUZ_rand(&lseed) & 1) {
-                    U32 const checksumFlag = FUZ_rand(&lseed) & 1;
+                    int const checksumFlag = FUZ_rand(&lseed) & 1;
                     DISPLAYLEVEL(5, "t%u: frame checksum : %u \n", testNb, checksumFlag);
-                    CHECK_Z( setCCtxParameter(zc, cctxParams, ZSTD_p_checksumFlag, checksumFlag, opaqueAPI) );
+                    CHECK_Z( setCCtxParameter(zc, cctxParams, ZSTD_c_checksumFlag, checksumFlag, opaqueAPI) );
                 }
-                if (FUZ_rand(&lseed) & 1) CHECK_Z( setCCtxParameter(zc, cctxParams, ZSTD_p_dictIDFlag, FUZ_rand(&lseed) & 1, opaqueAPI) );
-                if (FUZ_rand(&lseed) & 1) CHECK_Z( setCCtxParameter(zc, cctxParams, ZSTD_p_contentSizeFlag, FUZ_rand(&lseed) & 1, opaqueAPI) );
+                if (FUZ_rand(&lseed) & 1) CHECK_Z( setCCtxParameter(zc, cctxParams, ZSTD_c_dictIDFlag, FUZ_rand(&lseed) & 1, opaqueAPI) );
+                if (FUZ_rand(&lseed) & 1) CHECK_Z( setCCtxParameter(zc, cctxParams, ZSTD_c_contentSizeFlag, FUZ_rand(&lseed) & 1, opaqueAPI) );
                 if (FUZ_rand(&lseed) & 1) {
-                    DISPLAYLEVEL(5, "t%u: pledgedSrcSize : %u \n", testNb, (U32)pledgedSrcSize);
+                    DISPLAYLEVEL(5, "t%u: pledgedSrcSize : %u \n", testNb, (unsigned)pledgedSrcSize);
                     CHECK_Z( ZSTD_CCtx_setPledgedSrcSize(zc, pledgedSrcSize) );
                 }
 
@@ -1906,17 +1904,19 @@ static int fuzzerTests_newAPI(U32 seed, U32 nbTests, unsigned startTest,
                 if (bigTests || (FUZ_rand(&lseed) & 0xF) == 0xF) {
                     U32 const nbThreadsCandidate = (FUZ_rand(&lseed) & 4) + 1;
                     U32 const nbThreadsAdjusted = (windowLogMalus < nbThreadsCandidate) ? nbThreadsCandidate - windowLogMalus : 1;
-                    U32 const nbThreads = MIN(nbThreadsAdjusted, nbThreadsMax);
-                    DISPLAYLEVEL(5, "t%u: nbThreads : %u \n", testNb, nbThreads);
-                    CHECK_Z( setCCtxParameter(zc, cctxParams, ZSTD_p_nbWorkers, nbThreads, opaqueAPI) );
+                    int const nbThreads = MIN(nbThreadsAdjusted, nbThreadsMax);
+                    DISPLAYLEVEL(5, "t%i: nbThreads : %u \n", testNb, nbThreads);
+                    CHECK_Z( setCCtxParameter(zc, cctxParams, ZSTD_c_nbWorkers, nbThreads, opaqueAPI) );
                     if (nbThreads > 1) {
                         U32 const jobLog = FUZ_rand(&lseed) % (testLog+1);
-                        CHECK_Z( setCCtxParameter(zc, cctxParams, ZSTD_p_overlapSizeLog, FUZ_rand(&lseed) % 10, opaqueAPI) );
-                        CHECK_Z( setCCtxParameter(zc, cctxParams, ZSTD_p_jobSize, (U32)FUZ_rLogLength(&lseed, jobLog), opaqueAPI) );
+                        CHECK_Z( setCCtxParameter(zc, cctxParams, ZSTD_c_overlapLog, FUZ_rand(&lseed) % 10, opaqueAPI) );
+                        CHECK_Z( setCCtxParameter(zc, cctxParams, ZSTD_c_jobSize, (U32)FUZ_rLogLength(&lseed, jobLog), opaqueAPI) );
                     }
                 }
+                /* Enable rsyncable mode 1 in 4 times. */
+                setCCtxParameter(zc, cctxParams, ZSTD_c_rsyncable, (FUZ_rand(&lseed) % 4 == 0), opaqueAPI);
 
-                if (FUZ_rand(&lseed) & 1) CHECK_Z( setCCtxParameter(zc, cctxParams, ZSTD_p_forceMaxWindow, FUZ_rand(&lseed) & 1, opaqueAPI) );
+                if (FUZ_rand(&lseed) & 1) CHECK_Z( setCCtxParameter(zc, cctxParams, ZSTD_c_forceMaxWindow, FUZ_rand(&lseed) & 1, opaqueAPI) );
 
                 /* Apply parameters */
                 if (opaqueAPI) {
@@ -1936,7 +1936,7 @@ static int fuzzerTests_newAPI(U32 seed, U32 nbTests, unsigned startTest,
                             size_t const setError = ZSTD_CCtx_setParametersUsingCCtxParams(zc, cctxParams);
                             CHECK(!ZSTD_isError(setError), "ZSTD_CCtx_setParametersUsingCCtxParams should have failed");
                         } else {
-                            size_t const setError = ZSTD_CCtx_setParameter(zc, ZSTD_p_windowLog, cParams.windowLog-1);
+                            size_t const setError = ZSTD_CCtx_setParameter(zc, ZSTD_c_windowLog, cParams.windowLog-1);
                             CHECK(!ZSTD_isError(setError), "ZSTD_CCtx_setParameter should have failed");
                         }
                     }
@@ -1961,9 +1961,9 @@ static int fuzzerTests_newAPI(U32 seed, U32 nbTests, unsigned startTest,
                 ZSTD_inBuffer inBuff = { srcBuffer+srcStart, srcSize, 0 };
                 outBuff.size = outBuff.pos + dstBuffSize;
 
-                CHECK_Z( ZSTD_compress_generic(zc, &outBuff, &inBuff, flush) );
+                CHECK_Z( ZSTD_compressStream2(zc, &outBuff, &inBuff, flush) );
                 DISPLAYLEVEL(6, "t%u: compress consumed %u bytes (total : %u) ; flush: %u (total : %u) \n",
-                    testNb, (U32)inBuff.pos, (U32)(totalTestSize + inBuff.pos), (U32)flush, (U32)outBuff.pos);
+                    testNb, (unsigned)inBuff.pos, (unsigned)(totalTestSize + inBuff.pos), (unsigned)flush, (unsigned)outBuff.pos);
 
                 XXH64_update(&xxhState, srcBuffer+srcStart, inBuff.pos);
                 memcpy(copyBuffer+totalTestSize, srcBuffer+srcStart, inBuff.pos);
@@ -1977,11 +1977,11 @@ static int fuzzerTests_newAPI(U32 seed, U32 nbTests, unsigned startTest,
                     size_t const randomDstSize = FUZ_randomLength(&lseed, maxSampleLog+1);
                     size_t const adjustedDstSize = MIN(cBufferSize - cSize, randomDstSize);
                     outBuff.size = outBuff.pos + adjustedDstSize;
-                    DISPLAYLEVEL(6, "t%u: End-flush into dst buffer of size %u \n", testNb, (U32)adjustedDstSize);
-                    remainingToFlush = ZSTD_compress_generic(zc, &outBuff, &inBuff, ZSTD_e_end);
-                    DISPLAYLEVEL(6, "t%u: Total flushed so far : %u bytes \n", testNb, (U32)outBuff.pos);
+                    DISPLAYLEVEL(6, "t%u: End-flush into dst buffer of size %u \n", testNb, (unsigned)adjustedDstSize);
+                    remainingToFlush = ZSTD_compressStream2(zc, &outBuff, &inBuff, ZSTD_e_end);
+                    DISPLAYLEVEL(6, "t%u: Total flushed so far : %u bytes \n", testNb, (unsigned)outBuff.pos);
                     CHECK( ZSTD_isError(remainingToFlush),
-                          "ZSTD_compress_generic w/ ZSTD_e_end error : %s",
+                          "ZSTD_compressStream2 w/ ZSTD_e_end error : %s",
                            ZSTD_getErrorName(remainingToFlush) );
             }   }
             crcOrig = XXH64_digest(&xxhState);
@@ -1993,7 +1993,7 @@ static int fuzzerTests_newAPI(U32 seed, U32 nbTests, unsigned startTest,
 
         /* multi - fragments decompression test */
         if (!dictSize /* don't reset if dictionary : could be different */ && (FUZ_rand(&lseed) & 1)) {
-            DISPLAYLEVEL(5, "resetting DCtx (dict:%08X) \n", (U32)(size_t)dict);
+            DISPLAYLEVEL(5, "resetting DCtx (dict:%p) \n", dict);
             CHECK_Z( ZSTD_resetDStream(zd) );
         } else {
             if (dictSize)
@@ -2010,19 +2010,19 @@ static int fuzzerTests_newAPI(U32 seed, U32 nbTests, unsigned startTest,
                 inBuff.size = inBuff.pos + readCSrcSize;
                 outBuff.size = outBuff.pos + dstBuffSize;
                 DISPLAYLEVEL(6, "decompression presented %u new bytes (pos:%u/%u)\n",
-                                (U32)readCSrcSize, (U32)inBuff.pos, (U32)cSize);
+                                (unsigned)readCSrcSize, (unsigned)inBuff.pos, (unsigned)cSize);
                 decompressionResult = ZSTD_decompressStream(zd, &outBuff, &inBuff);
                 DISPLAYLEVEL(6, "so far: consumed = %u, produced = %u \n",
-                                (U32)inBuff.pos, (U32)outBuff.pos);
+                                (unsigned)inBuff.pos, (unsigned)outBuff.pos);
                 if (ZSTD_isError(decompressionResult)) {
                     DISPLAY("ZSTD_decompressStream error : %s \n", ZSTD_getErrorName(decompressionResult));
                     findDiff(copyBuffer, dstBuffer, totalTestSize);
                 }
                 CHECK (ZSTD_isError(decompressionResult), "decompression error : %s", ZSTD_getErrorName(decompressionResult));
-                CHECK (inBuff.pos > cSize, "ZSTD_decompressStream consumes too much input : %u > %u ", (U32)inBuff.pos, (U32)cSize);
+                CHECK (inBuff.pos > cSize, "ZSTD_decompressStream consumes too much input : %u > %u ", (unsigned)inBuff.pos, (unsigned)cSize);
             }
-            CHECK (inBuff.pos != cSize, "compressed data should be fully read (%u != %u)", (U32)inBuff.pos, (U32)cSize);
-            CHECK (outBuff.pos != totalTestSize, "decompressed data : wrong size (%u != %u)", (U32)outBuff.pos, (U32)totalTestSize);
+            CHECK (inBuff.pos != cSize, "compressed data should be fully read (%u != %u)", (unsigned)inBuff.pos, (unsigned)cSize);
+            CHECK (outBuff.pos != totalTestSize, "decompressed data : wrong size (%u != %u)", (unsigned)outBuff.pos, (unsigned)totalTestSize);
             {   U64 const crcDest = XXH64(dstBuffer, totalTestSize, 0);
                 if (crcDest!=crcOrig) findDiff(copyBuffer, dstBuffer, totalTestSize);
                 CHECK (crcDest!=crcOrig, "decompressed data corrupted");
@@ -2219,7 +2219,7 @@ int main(int argc, const char** argv)
         seed = h % 10000;
     }
 
-    DISPLAY("Seed = %u\n", seed);
+    DISPLAY("Seed = %u\n", (unsigned)seed);
     if (proba!=FUZ_COMPRESSIBILITY_DEFAULT) DISPLAY("Compressibility : %i%%\n", proba);
 
     if (nbTests<=0) nbTests=1;
diff --git a/zlibWrapper/Makefile b/zlibWrapper/Makefile
index c1896f8b80a3..0c19107b3e6f 100644
--- a/zlibWrapper/Makefile
+++ b/zlibWrapper/Makefile
@@ -71,25 +71,25 @@ valgrindTest: clean example fitblk example_zstd fitblk_zstd zwrapbench
 #	$(CC) $(CFLAGS) $(CPPFLAGS) -c -o $@ $<
 
 minigzip: $(EXAMPLE_PATH)/minigzip.o $(ZLIBWRAPPER_PATH)/zstd_zlibwrapper.o $(GZFILES) $(ZSTDLIBRARY)
-	$(CC) $(LDFLAGS) $^ $(ZSTDLIBRARY) $(ZLIB_LIBRARY) -o $@
+	$(CC) $(CFLAGS) $(CPPFLAGS) $(LDFLAGS) $^ $(ZSTDLIBRARY) $(ZLIB_LIBRARY) -o $@
 
 minigzip_zstd: $(EXAMPLE_PATH)/minigzip.o $(ZLIBWRAPPER_PATH)/zstdTurnedOn_zlibwrapper.o $(GZFILES) $(ZSTDLIBRARY)
-	$(CC) $(LDFLAGS) $^ $(ZSTDLIBRARY) $(ZLIB_LIBRARY) -o $@
+	$(CC) $(CFLAGS) $(CPPFLAGS) $(LDFLAGS) $^ $(ZSTDLIBRARY) $(ZLIB_LIBRARY) -o $@
 
 example: $(EXAMPLE_PATH)/example.o $(ZLIBWRAPPER_PATH)/zstd_zlibwrapper.o $(GZFILES) $(ZSTDLIBRARY)
-	$(CC) $(LDFLAGS) $^ $(ZLIB_LIBRARY) -o $@
+	$(CC) $(CFLAGS) $(CPPFLAGS) $(LDFLAGS) $^ $(ZLIB_LIBRARY) -o $@
 
 example_zstd: $(EXAMPLE_PATH)/example.o $(ZLIBWRAPPER_PATH)/zstdTurnedOn_zlibwrapper.o $(GZFILES) $(ZSTDLIBRARY)
-	$(CC) $(LDFLAGS) $^ $(ZLIB_LIBRARY) -o $@
+	$(CC) $(CFLAGS) $(CPPFLAGS) $(LDFLAGS) $^ $(ZLIB_LIBRARY) -o $@
 
 fitblk: $(EXAMPLE_PATH)/fitblk.o $(ZLIBWRAPPER_PATH)/zstd_zlibwrapper.o $(ZSTDLIBRARY)
-	$(CC) $(LDFLAGS) $^ $(ZLIB_LIBRARY) -o $@
+	$(CC) $(CFLAGS) $(CPPFLAGS) $(LDFLAGS) $^ $(ZLIB_LIBRARY) -o $@
 
 fitblk_zstd: $(EXAMPLE_PATH)/fitblk.o $(ZLIBWRAPPER_PATH)/zstdTurnedOn_zlibwrapper.o $(ZSTDLIBRARY)
-	$(CC) $(LDFLAGS) $^ $(ZLIB_LIBRARY) -o $@
+	$(CC) $(CFLAGS) $(CPPFLAGS) $(LDFLAGS) $^ $(ZLIB_LIBRARY) -o $@
 
-zwrapbench: $(EXAMPLE_PATH)/zwrapbench.o $(ZLIBWRAPPER_PATH)/zstd_zlibwrapper.o $(PROGRAMS_PATH)/datagen.o $(ZSTDLIBRARY)
-	$(CC) $(LDFLAGS) $^ $(ZLIB_LIBRARY) -o $@
+zwrapbench: $(EXAMPLE_PATH)/zwrapbench.o $(ZLIBWRAPPER_PATH)/zstd_zlibwrapper.o $(PROGRAMS_PATH)/util.o $(PROGRAMS_PATH)/datagen.o $(ZSTDLIBRARY)
+	$(CC) $(CFLAGS) $(CPPFLAGS) $(LDFLAGS) $^ $(ZLIB_LIBRARY) -o $@
 
 
 $(ZLIBWRAPPER_PATH)/zstd_zlibwrapper.o: $(ZLIBWRAPPER_PATH)/zstd_zlibwrapper.c $(ZLIBWRAPPER_PATH)/zstd_zlibwrapper.h
diff --git a/zlibWrapper/examples/zwrapbench.c b/zlibWrapper/examples/zwrapbench.c
index d2d6073f9c0e..e071c3098525 100644
--- a/zlibWrapper/examples/zwrapbench.c
+++ b/zlibWrapper/examples/zwrapbench.c
@@ -17,6 +17,7 @@
 #include        /* fprintf, fopen, ftello64 */
 #include         /* clock_t, clock, CLOCKS_PER_SEC */
 #include        /* toupper */
+#include        /* errno */
 
 #include "mem.h"
 #define ZSTD_STATIC_LINKING_ONLY
@@ -103,7 +104,7 @@ static clock_t g_time = 0;
 /* *************************************
 *  Benchmark Parameters
 ***************************************/
-static U32 g_nbIterations = NBLOOPS;
+static unsigned g_nbIterations = NBLOOPS;
 static size_t g_blockSize = 0;
 int g_additionalParam = 0;
 
@@ -120,7 +121,7 @@ void BMK_SetNbIterations(unsigned nbLoops)
 void BMK_SetBlockSize(size_t blockSize)
 {
     g_blockSize = blockSize;
-    DISPLAYLEVEL(2, "using blocks of size %u KB \n", (U32)(blockSize>>10));
+    DISPLAYLEVEL(2, "using blocks of size %u KB \n", (unsigned)(blockSize>>10));
 }
 
 
@@ -221,7 +222,7 @@ static int BMK_benchMem(z_const void* srcBuffer, size_t srcSize,
             }
 
             /* Compression */
-            DISPLAYLEVEL(2, "%2s-%-17.17s :%10u ->\r", marks[markNb], displayName, (U32)srcSize);
+            DISPLAYLEVEL(2, "%2s-%-17.17s :%10u ->\r", marks[markNb], displayName, (unsigned)srcSize);
             if (!cCompleted) memset(compressedBuffer, 0xE5, maxCompressedSize);  /* warm up and erase result buffer */
 
             UTIL_sleepMilli(1);  /* give processor time to other processes */
@@ -370,7 +371,7 @@ static int BMK_benchMem(z_const void* srcBuffer, size_t srcSize,
             ratio = (double)srcSize / (double)cSize;
             markNb = (markNb+1) % NB_MARKS;
             DISPLAYLEVEL(2, "%2s-%-17.17s :%10u ->%10u (%5.3f),%6.1f MB/s\r",
-                    marks[markNb], displayName, (U32)srcSize, (U32)cSize, ratio,
+                    marks[markNb], displayName, (unsigned)srcSize, (unsigned)cSize, ratio,
                     (double)srcSize / fastestC );
 
             (void)fastestD; (void)crcOrig;   /*  unused when decompression disabled */
@@ -388,7 +389,7 @@ static int BMK_benchMem(z_const void* srcBuffer, size_t srcSize,
                     ZSTD_DDict* ddict = ZSTD_createDDict(dictBuffer, dictBufferSize);
                     if (!ddict) EXM_THROW(2, "ZSTD_createDDict() allocation failure");
                     do {
-                        U32 blockNb;
+                        unsigned blockNb;
                         for (blockNb=0; blockNb%10u (%5.3f),%6.1f MB/s ,%6.1f MB/s\r",
-                    marks[markNb], displayName, (U32)srcSize, (U32)cSize, ratio,
+                    marks[markNb], displayName, (unsigned)srcSize, (unsigned)cSize, ratio,
                     (double)srcSize / fastestC,
                     (double)srcSize / fastestD );
 
@@ -520,9 +521,9 @@ static int BMK_benchMem(z_const void* srcBuffer, size_t srcSize,
                     DISPLAY("!!! WARNING !!! %14s : Invalid Checksum : %x != %x   \n", displayName, (unsigned)crcOrig, (unsigned)crcCheck);
                     for (u=0; u u) break;
                                 bacc += blockTable[segNb].srcSize;
@@ -593,7 +594,9 @@ static void BMK_benchCLevel(void* srcBuffer, size_t benchedSize,
     SET_REALTIME_PRIORITY;
 
     if (g_displayLevel == 1 && !g_additionalParam)
-        DISPLAY("bench %s %s: input %u bytes, %u seconds, %u KB blocks\n", ZSTD_VERSION_STRING, ZSTD_GIT_COMMIT_STRING, (U32)benchedSize, g_nbIterations, (U32)(g_blockSize>>10));
+        DISPLAY("bench %s %s: input %u bytes, %u seconds, %u KB blocks\n",
+                ZSTD_VERSION_STRING, ZSTD_GIT_COMMIT_STRING,
+                (unsigned)benchedSize, g_nbIterations, (unsigned)(g_blockSize>>10));
 
     if (cLevelLast < cLevel) cLevelLast = cLevel;
 
@@ -725,7 +728,7 @@ static void BMK_benchFileTable(const char** fileNamesTable, unsigned nbFiles,
         dictBufferSize = (size_t)dictFileSize;
         dictBuffer = malloc(dictBufferSize);
         if (dictBuffer==NULL)
-            EXM_THROW(11, "not enough memory for dictionary (%u bytes)", (U32)dictBufferSize);
+            EXM_THROW(11, "not enough memory for dictionary (%u bytes)", (unsigned)dictBufferSize);
         BMK_loadFiles(dictBuffer, dictBufferSize, fileSizes, &dictFileName, 1);
     }
 
@@ -733,7 +736,7 @@ static void BMK_benchFileTable(const char** fileNamesTable, unsigned nbFiles,
     benchedSize = BMK_findMaxMem(totalSizeToLoad * 3) / 3;
     if ((U64)benchedSize > totalSizeToLoad) benchedSize = (size_t)totalSizeToLoad;
     if (benchedSize < totalSizeToLoad)
-        DISPLAY("Not enough memory; testing %u MB only...\n", (U32)(benchedSize >> 20));
+        DISPLAY("Not enough memory; testing %u MB only...\n", (unsigned)(benchedSize >> 20));
     srcBuffer = malloc(benchedSize + !benchedSize);
     if (!srcBuffer) EXM_THROW(12, "not enough memory");
 
diff --git a/zlibWrapper/zstd_zlibwrapper.c b/zlibWrapper/zstd_zlibwrapper.c
index b3e5f36725a1..0ee5a310873b 100644
--- a/zlibWrapper/zstd_zlibwrapper.c
+++ b/zlibWrapper/zstd_zlibwrapper.c
@@ -31,7 +31,7 @@
 /* ===   Constants   === */
 #define Z_INFLATE_SYNC              8
 #define ZLIB_HEADERSIZE             4
-#define ZSTD_HEADERSIZE             ZSTD_frameHeaderSize_min
+#define ZSTD_HEADERSIZE             ZSTD_FRAMEHEADERSIZE_MIN
 #define ZWRAP_DEFAULT_CLEVEL        3   /* Z_DEFAULT_COMPRESSION is translated to ZWRAP_DEFAULT_CLEVEL for zstd */
 
 
@@ -140,8 +140,8 @@ static int ZWRAP_initializeCStream(ZWRAP_CCtx* zwc, const void* dict, size_t dic
     if (!pledgedSrcSize) pledgedSrcSize = zwc->pledgedSrcSize;
     {   ZSTD_parameters const params = ZSTD_getParams(zwc->compressionLevel, pledgedSrcSize, dictSize);
         size_t initErr;
-        LOG_WRAPPERC("pledgedSrcSize=%d windowLog=%d chainLog=%d hashLog=%d searchLog=%d searchLength=%d strategy=%d\n",
-                    (int)pledgedSrcSize, params.cParams.windowLog, params.cParams.chainLog, params.cParams.hashLog, params.cParams.searchLog, params.cParams.searchLength, params.cParams.strategy);
+        LOG_WRAPPERC("pledgedSrcSize=%d windowLog=%d chainLog=%d hashLog=%d searchLog=%d minMatch=%d strategy=%d\n",
+                    (int)pledgedSrcSize, params.cParams.windowLog, params.cParams.chainLog, params.cParams.hashLog, params.cParams.searchLog, params.cParams.minMatch, params.cParams.strategy);
         initErr = ZSTD_initCStream_advanced(zwc->zbc, dict, dictSize, params, pledgedSrcSize);
         if (ZSTD_isError(initErr)) return Z_STREAM_ERROR;
     }