d/freebsd-skq
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Increase loop unrolling for skein hashes

Browse Source 
This patch was inspired by an opposite change made to shrink the code
for the boot loader.

On my i7-4770, it increases the skein1024 speed from 470 to 550 MB/s

Reviewed by:	sbruno
MFC after:	1 month
Sponsored by:	ScaleEngine Inc.
Differential Revision:	https://reviews.freebsd.org/D7824
This commit is contained in:
allanjude 2017-07-01 21:18:06 +00:00
parent 57c61ab190
commit 162ef49ab1
2 changed files with 8 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
lib/libmd/Makefile
View File

@ -88,6 +88,8 @@ sys/md5.h: ${SRCTOP}/sys/${.TARGET} .NOMETA
CFLAGS+= -I${.CURDIR} -I${SRCTOP}/sys/crypto/sha2 CFLAGS+= -I${.CURDIR} -I${SRCTOP}/sys/crypto/sha2
CFLAGS+= -I${SRCTOP}/sys/crypto/skein CFLAGS+= -I${SRCTOP}/sys/crypto/skein
CFLAGS+= -DWEAK_REFS CFLAGS+= -DWEAK_REFS
# unroll the 256 and 512 loops, half unroll the 1024
CFLAGS+= -DSKEIN_LOOP=995
.PATH: ${.CURDIR}/${MACHINE_ARCH} ${SRCTOP}/sys/crypto/sha2 .PATH: ${.CURDIR}/${MACHINE_ARCH} ${SRCTOP}/sys/crypto/sha2
.PATH: ${SRCTOP}/sys/crypto/skein ${SRCTOP}/sys/crypto/skein/${MACHINE_ARCH} .PATH: ${SRCTOP}/sys/crypto/skein ${SRCTOP}/sys/crypto/skein/${MACHINE_ARCH}
@ -101,6 +103,8 @@ CFLAGS+= -DRMD160_ASM
.endif .endif
.if exists(${MACHINE_ARCH}/skein_block_asm.s) .if exists(${MACHINE_ARCH}/skein_block_asm.s)
AFLAGS += --strip-local-absolute AFLAGS += --strip-local-absolute
# Fully unroll all loops in the assembly optimized version
AFLAGS+= --defsym SKEIN_LOOP=0
SRCS+= skein_block_asm.s SRCS+= skein_block_asm.s
CFLAGS+= -DSKEIN_ASM -DSKEIN_USE_ASM=1792 # list of block functions to replace with assembly: 256+512+1024 = 1792 CFLAGS+= -DSKEIN_ASM -DSKEIN_USE_ASM=1792 # list of block functions to replace with assembly: 256+512+1024 = 1792
.endif .endif

4
sys/modules/crypto/Makefile
View File

@ -19,11 +19,15 @@ SRCS += camellia.c camellia-api.c
SRCS += des_ecb.c des_enc.c des_setkey.c SRCS += des_ecb.c des_enc.c des_setkey.c
SRCS += sha1.c sha256c.c sha512c.c SRCS += sha1.c sha256c.c sha512c.c
SRCS += skein.c skein_block.c SRCS += skein.c skein_block.c
# unroll the 256 and 512 loops, half unroll the 1024
CFLAGS+= -DSKEIN_LOOP=995
.if exists(${MACHINE_ARCH}/skein_block_asm.s) .if exists(${MACHINE_ARCH}/skein_block_asm.s)
.PATH: ${SRCTOP}/sys/crypto/skein/${MACHINE_ARCH} .PATH: ${SRCTOP}/sys/crypto/skein/${MACHINE_ARCH}
SRCS += skein_block_asm.s SRCS += skein_block_asm.s
CFLAGS += -DSKEIN_ASM -DSKEIN_USE_ASM=1792 # list of block functions to replace with assembly: 256+512+1024 = 1792 CFLAGS += -DSKEIN_ASM -DSKEIN_USE_ASM=1792 # list of block functions to replace with assembly: 256+512+1024 = 1792
ACFLAGS += -DELF -Wa,--noexecstack ACFLAGS += -DELF -Wa,--noexecstack
# Fully unroll all loops in the assembly optimized version
AFLAGS+= --defsym SKEIN_LOOP=0
.endif .endif
SRCS += siphash.c SRCS += siphash.c
SRCS += gmac.c gfmult.c SRCS += gmac.c gfmult.c