Commit Graph

76 Commits

Author SHA1 Message Date
das
4d2ad621bf - Define LDBL_NBIT to be a mask indicating the position of the integer
bit in a long double.  For architectures that don't have such a bit,
  LDBL_NBIT is 0.  This makes it possible to say `mantissa & ~LDBL_NBIT'
  in places that previously used an #ifdef to select the right expression.
  The optimizer should dispense with the extra arithmetic when LDBL_NBIT
  is 0 anyway.
- Add an XXX comment for the big endian case.
2005-03-07 04:55:40 +00:00
das
b45580b648 Update my email address. 2005-02-06 03:23:31 +00:00
das
87888496cc Replace the ldexp() implementation in libc with a renamed copy of the
scalbn() implementation from libm.  (The two functions are defined to
be identical, but ldexp() lives in libc for backwards compatibility.)
The old ldexp() implementation...
  - was more complicated than this one
  - set errno instead of raising FP exceptions
  - got some corner cases wrong
    (e.g. ldexp(1.0, 2000) in round-to-zero mode)

The new implementation lives in libc/gen instead of
libc/$MACHINE_ARCH/gen, since we don't need N copies of a
machine-independent file.  The amd64 and i386 platforms
retain their fast and correct MD implementations and
override this one.
2005-01-22 06:03:40 +00:00
das
0bc7903301 Eliminate gdtoa.mk and move its contents to ${MACHINE_ARCH}/Makefile.inc.
The purpose of having a separate file involved an abandoned scheme that
would have kept contrib/gdtoa out of the include path for the rest of libc.
2005-01-15 05:23:58 +00:00
ru
5db2b9d5b3 For variables that are only checked with defined(), don't provide
any fake value.
2004-10-24 15:33:08 +00:00
marcel
7f85950ec5 Unbreak memmove(). Return the dst argument. While here, fix the END
macro to actually reference memmove, not memcpy.
2004-09-04 00:23:15 +00:00
marcel
a035e5e43c Unbreak memcpy(). memcpy() is defined to return the dst argument.
By using r8 instead of r14 to do the swap, we put the dst argument
in the return register. Since bcopy() doesn't clobber r8, we don't
have to do anything else. This fixes ports/textproc/aspell.
2004-09-04 00:04:58 +00:00
marcel
52736a0a13 Make profiling actually work. The gcc compiler emits a call to the
_mcount() stub when profiling is enabled. Emit this code sequence
for assembly routines as welli (MCOUNT definition in <machine/asm.h>.
We do not pass the GOT entry however as the 4th argument, because it's
not used. The _mcount() stub calls __mcount(), which does the actual
work. Define _MCOUNT_DECL to define __mcount. We do not have an
implementation of mcount(), so we define MCOUNT as empty, but have a
weak alias to _mcount() in _mcount.S.
Note that the _mcount() stub in the kernel is slightly different from
the stub in userland. This is because we do not have to worry about
nested routines in the kernel.
2004-08-25 07:42:34 +00:00
dfr
2f90ca8b3c Add support for TLS in statically linked programs. 2004-08-15 16:18:52 +00:00
das
86c293bf54 Make FLT_ROUNDS correctly reflect the dynamic rounding mode. 2004-07-19 08:17:25 +00:00
das
1a69fc3370 Replace seven nominally MD implementations of frexp() that are broken
for subnormals with one implementation that works.
2004-07-18 21:23:39 +00:00
das
65d8d759b1 Implement the classification macros isfinite(), isinf(), isnan(), and
isnormal() the hard way, rather than relying on fpclassify().  This is
a lose in the sense that we need a total of 12 functions, but it is
necessary for binary compatibility because we have never bumped libm's
major version number.  In particular, isinf(), isnan(), and isnanf()
were BSD libc functions before they were C99 macros, so we can't
reimplement them in terms of fpclassify() without adding a dependency
on libc.so.5.  I have tried to arrange things so that programs that
could be compiled in FreeBSD 4.X will generate the same external
references when compiled in 5.X.  At the same time, the new macros
should remain C99-compliant.

The isinf() and isnan() functions remain in libc for historical
reasons; however, I have moved the functions that implement the macros
isfinite() and isnormal() to libm where they belong.  Moreover,
half a dozen MD versions of isinf() and isnan() have been replaced
with MI versions that work equally well.

Prodded by:	kris
2004-07-09 03:32:40 +00:00
das
0a50563417 Fix some aliasing problems. 2004-02-16 10:02:40 +00:00
cperciva
371e76013c style cleanup: Remove duplicate $FreeBSD$ tags.
These files had tags after teh copyright notice,
inside the comment block (incorrect, removed),
and outside the comment block (correct).

Approved by:	rwatson (mentor)
2004-02-10 20:45:28 +00:00
das
daf3a2f509 Define LDBL_MANH_SIZE and LDBL_MANL_SIZE to be the sizes of the
high and low words of the mantissa in bits, respectively.
2004-01-18 07:57:02 +00:00
nectar
e87a51f854 Add required headers. 2004-01-06 19:40:28 +00:00
marcel
6ae3d37f99 Virtual addresses in headers of ELF files for dynamic objects need
to be relocated before they can be used as pointers.
2003-11-08 05:29:49 +00:00
marcel
e03bb88101 The FP status register allows for 6 traps to be masked. One of them,
the denormal/unnormal trap, is not a standard IEEE trap. We did
not exclude it from being returned by fpgetmask(), nor did we make
sure that fpsetmask() didn't clobber it. Since the non-IEEE trap
is not part of fp_except_t, users of ifpgetmask()/fpsetmask() would
be confronted with unexpected behaviour, one of which is a SIGFPE
for denormal/unnormal FP results.

This commit makes sure that we don't leak the denormal/unnormal mask
bit in fp_except_t and also that we don't clobber it.
2003-10-22 09:00:07 +00:00
bms
44aa51e3ae Add the mlockall() and munlockall() system calls.
- All those diffs to syscalls.master for each architecture *are*
   necessary. This needed clarification; the stub code generation for
   mlockall() was disabled, which would prevent applications from
   linking to this API (suggested by mux)
 - Giant has been quoshed. It is no longer held by the code, as
   the required locking has been pushed down within vm_map.c.
 - Callers must specify VM_MAP_WIRE_HOLESOK or VM_MAP_WIRE_NOHOLES
   to express their intention explicitly.
 - Inspected at the vmstat, top and vm pager sysctl stats level.
   Paging-in activity is occurring correctly, using a test harness.
 - The RES size for a process may appear to be greater than its SIZE.
   This is believed to be due to mappings of the same shared library
   page being wired twice. Further exploration is needed.
 - Believed to back out of allocations and locks correctly
   (tested with WITNESS, MUTEX_PROFILING, INVARIANTS and DIAGNOSTIC).

PR:             kern/43426, standards/54223
Reviewed by:    jake, alc
Approved by:    jake (mentor)
MFC after:	2 weeks
2003-08-11 07:14:08 +00:00
marcel
32aa925e11 o There are 6 trap disable bits in ar.fpsr, not five. Even though we
didn't provide a constant for one of them (non-IEEE denormal trap),
  in an attempt to not support it probably, it's not we are left with
  the lower 5 bits.
o Properly mask the passed or returned fp_except_t. Not doing so
  causes instant core dumps by trying to write an invalid value to
  ar.fpsr. Now that we're masking, stop using exclusive-or to invert
  bits.

This fixes the illegal instruction fault encountered when building
mozilla.
2003-08-09 17:07:24 +00:00
marcel
8781039ce8 Override the default stubs for getcontext(2) and swapcontext(2) so
that we can flush the register stack prior to entering the kernel.
This avoids having dirty registers and saves us from having to
manually write them to the backing store from within the kernel.
In that respect, flushing the RSE is both functionally required as
well as performance optimal.

On average we had 18 dirty registers when getcontext(2) was called
from libthr. Since libthr does not switch back to a context created
by getcontext(2), not having dealt with the dirty registers was
harmless.
2003-08-02 00:49:36 +00:00
marcel
08b0cb60b3 The END() must expand to the .endp directive with the same name as
on the corresponding .proc directive, or the .endp must not have a
name at all.
While here, remove an artificial dependency in Ovfork.S by performing
manual register renaming.
2003-08-01 22:17:12 +00:00
marcel
3c92e21eeb Revert previous commit. We don't use setjmp()/longjmp() for context
switching anymore, so there's no need to save and restore GP. This
change breaks threaded applications linked against libc_r. Pull the
tier 2 card again: relink. This will link against libthr instead.
2003-07-25 22:36:48 +00:00
marcel
8e72bcec56 Implement signalcontext(). Needed by libpthread (aka libkse). 2003-06-24 05:06:42 +00:00
marcel
a16e92bfa3 o Fix a cut-n-paste bug. We were clobbering rp with gp...
o  Make sure the arguments to ctx_wrapper() are loaded from the
   backing store by forcing an underflow. Do this by making all
   registers in the register frame local.
2003-06-02 00:16:39 +00:00
marcel
d5ebc72a62 Rough implement of makecontext() needed to support libthr.
o  Up to 8 arguments are allowed. This is the number of arguments
   passed in registers. Subsequent registers are passed on the stack.
   Trying to deal with this is not easy in C and likely forces us to
   use assembly code. Let's avoid that for now. There's no indication
   that more than 8 arguments is a strong requirement (Linux also has
   an 8 argument limit).
o  We expect that the stack base is 16-byte aligned and the stack
   size is a multiple of 16-byte. We bomb out if this is not the case.
   We probably want to be less strict by enforcing it ourselves. For
   now it's better to not hide gross alignment bogons by silently
   correcting it.
2003-05-31 19:42:51 +00:00
marcel
5d3af2c5ab Revamp of the syscall path, exception and context handling. The
prime objectives are:
o  Implement a syscall path based on the epc inststruction (see
   sys/ia64/ia64/syscall.s).
o  Revisit the places were we need to save and restore registers
   and define those contexts in terms of the register sets (see
   sys/ia64/include/_regset.h).

Secundairy objectives:
o  Remove the requirement to use contigmalloc for kernel stacks.
o  Better handling of the high FP registers for SMP systems.
o  Switch to the new cpu_switch() and cpu_throw() semantics.
o  Add a good unwinder to reconstruct contexts for the rare
   cases we need to (see sys/contrib/ia64/libuwx)

Many files are affected by this change. Functionally it boils
down to:
o  The EPC syscall doesn't preserve registers it does not need
   to preserve and places the arguments differently on the stack.
   This affects libc and truss.
o  The address of the kernel page directory (kptdir) had to
   be unstaticized for use by the nested TLB fault handler.
   The name has been changed to ia64_kptdir to avoid conflicts.
   The renaming affects libkvm.
o  The trapframe only contains the special registers and the
   scratch registers. For syscalls using the EPC syscall path
   no scratch registers are saved. This affects all places where
   the trapframe is accessed. Most notably the unaligned access
   handler, the signal delivery code and the debugger.
o  Context switching only partly saves the special registers
   and the preserved registers. This affects cpu_switch() and
   triggered the move to the new semantics, which additionally
   affects cpu_throw().
o  The high FP registers are either in the PCB or on some
   CPU. context switching for them is done lazily. This affects
   trap().
o  The mcontext has room for all registers, but not all of them
   have to be defined in all cases. This mostly affects signal
   delivery code now. The *context syscalls are as of yet still
   unimplemented.

Many details went into the removal of the requirement to use
contigmalloc for kernel stacks. The details are mostly CPU
specific and limited to exception_save() and exception_restore().
The few places where we create, destroy or switch stacks were
mostly simplified by not having to construct physical addresses
and additionally saving the virtual addresses for later use.

Besides more efficient context saving and restoring, which of
course yields a noticable speedup, this also fixes the dreaded
SMP bootup problem as a side-effect. The details of which are
still not fully understood.

This change includes all the necessary backward compatibility
code to have it handle older userland binaries that use the
break instruction for syscalls. Support for break-based syscalls
has been pessimized in favor of a clean implementation. Due to
the overall better performance of the kernel, this will still
be notived as an improvement if it's noticed at all.

Approved by: re@ (jhb)
2003-05-16 21:26:42 +00:00
das
d807e3cae2 Add a comment describing why it's important for the values in this
file to be correct, and how to generate them automatically.

Caused much pain and suffering for:	peter
2003-05-08 13:50:44 +00:00
das
2d9148b622 Add __ldtoa(), a wrapper around gdtoa() to make it look like dtoa().
In support of this, add some MD macros to assist in converting long
doubles to the format expected by gdtoa().

Reviewed by:	silence on standards@
2003-04-05 22:10:13 +00:00
das
d02cfc3692 Replace our ancient dtoa/strtod implementation with the gdtoa
package, a more recent, generalized set of routines.  Among the
changes:
- Declare strtof() and strtold() in stdlib.h.
- Add glue to libc to support these routines for all kinds
  of ``long double''.
- Update printf() to reflect the fact that dtoa works slightly
  differently now.

As soon as I see that nothing has blown up, I will kill
src/lib/libc/stdlib/strtod.c.  Soon printf() will be able
to use the new routines to output long doubles without loss
of precision, but numerous bugs in the existing code must
be addressed first.

Reviewed by:	bde (briefly), mike (mentor), obrien
2003-03-12 20:30:00 +00:00
marcel
d4ee62b07a Fix threaded applications on ia64 that are linked dynamicly. We did
not save (restore) the global pointer (GP) in the jmpbuf in setjmp
(longjmp) because it's not needed in general. GP is considered a
scratch register at callsites and hence is always restored after a
call (when it's possible that the call resolves to a symbol in a
different loadmodule; otherwise GP does not have to be saved and
restored at all), including calls to setjmp/longjmp. There's just
one problem with this now that we use setjmp/longjmp for context
switching: A new context must have GP defined properly for the
thread's entry point. This means that we need to put GP in the
jmpbuf and consequently that we have to restore is in longjmp.
This automaticly requires us to save it as well.

When setjmp/longjmp isn't used for context switching, this can be
reverted again.
2003-03-05 04:39:24 +00:00
marcel
55f069454e ABI breaker: Move the J_SIGMASK field in the jmpbuf before
the J_SIG0 field. While here, rename J_SIG0 to J_SIGSET and
remove J_SIG1. The main reason for this change is that the
128-bit sigset_t is now aligned on a 16-byte boundary, which
allows us to use 16-byte atomic loads and stores on CPUs that
support it. The removal of J_SIG1 is done to avoid confusion:
it is never accessed and should not be. Renaming J_SIG0 to
J_SIGSET is the icing on the cake that's better done now than
later.
2003-03-05 03:30:54 +00:00
obrien
49328f91e2 Use __FBSDID. 2003-03-03 01:09:46 +00:00
mike
39b6d14ab4 ia64 actually uses 80-bit long doubles and must support big and little
endian at compile-time.

Reviewed by:	das
2003-02-26 16:04:34 +00:00
mike
1998abeb23 o Implement C99 classification macros isfinite(), isinf(), isnan(),
isnormal().  The current isinf() and isnan() are perserved for
  binary compatibility with 5.0, but new programs will use the macros.
o Implement C99 comparison macros isgreater(), isgreaterequal(),
  isless(), islessequal(), islessgreater(), isunordered().

Submitted by:	David Schultz <dschultz@uclink.Berkeley.EDU>
2003-02-12 20:03:41 +00:00
mike
b4e3f2f94a Implement fpclassify():
o Add a MD header private to libc called _fpmath.h; this header
  contains bitfield layouts of MD floating-point types.
o Add a MI header private to libc called fpmath.h; this header
  contains bitfield layouts of MI floating-point types.
o Add private libc variables to lib/libc/$arch/gen/infinity.c for
  storing NaN values.
o Add __double_t and __float_t to <machine/_types.h>, and provide
  double_t and float_t typedefs in <math.h>.
o Add some C99 manifest constants (FP_ILOGB0, FP_ILOGBNAN, HUGE_VALF,
  HUGE_VALL, INFINITY, NAN, and return values for fpclassify()) to
  <math.h> and others (FLT_EVAL_METHOD, DECIMAL_DIG) to <float.h> via
  <machine/float.h>.
o Add C99 macro fpclassify() which calls __fpclassify{d,f,l}() based
  on the size of its argument.  __fpclassifyl() is never called on
  alpha because (sizeof(long double) == sizeof(double)), which is good
  since __fpclassifyl() can't deal with such a small `long double'.

This was developed by David Schultz and myself with input from bde and
fenner.

PR:		23103
Submitted by:	David Schultz <dschultz@uclink.Berkeley.EDU>
		(significant portions)
Reviewed by:	bde, fenner (earlier versions)
2003-02-08 20:37:55 +00:00
marcel
e2486fdf1a Implement missing fpgetround() and fpsetround(). 2003-01-11 07:24:54 +00:00
schweikh
d3367c5f5d Correct typos, mostly s/ a / an / where appropriate. Some whitespace cleanup,
especially in troff files.
2003-01-01 18:49:04 +00:00
marcel
34ba8a6132 Sort SRCS. 2002-11-16 01:41:33 +00:00
marcel
52879ad635 o Fix _longjmp() to return 1 when the return value is given as 0.
o  Remove the unwanted smartness in _longjmp() where it compares
   the current ar.bspstore with the saved ar.bspstore and restores
   ar.rnat based on it. This either avoids saving ar.rnat in the
   jmp_buf or is the consequence of not saving ar.rnat. All this
   complexity breaks libc_r where we use longjmp() to switch to
   different threads and the current ar.bspstore has no relation
   to the saved ar.bspstore. Thus: we save ar.rnat in setjmp()
   and simply restore ar.bspstore and ar.rnat in longjmp().

This code needs a cleanup.
2002-11-14 06:40:23 +00:00
marcel
0489f5e22b ia64 ABI breaker:
Don't force 16-byte alignment at run-time. Do it at compile-time.
This saves us the pointer fiddling by the setjmp functions and
reduces complexity. While here, increase the jmp_buf by 16 bytes
to an even 512 bytes. Coincidentally, due to the way alignment
was handled prior to this change, the jmp_buf has not changed in
size, but only in how the space is used. Prior to this change
the 16 bytes were reserved for enforcing alignment; now they are
reserved by us for future extensions.
Therefore, this ABI breaker is relatively save: the failure is
always an alignment trap.
2002-11-11 08:11:44 +00:00
archie
d93f84495b Re-apply the previously backed-out commit that fixes the problem where
HUGE_VAL is not properly aligned on some architectures. The previous
fix now works because the two versions of 'math.h' (include/math.h
and lib/msun/src/math.h) have since been merged into one.

PR:	bin/43544
2002-10-31 23:05:20 +00:00
archie
207fb06f80 Revert previous commit to unbreak world until we figure out the
right way to do it.
2002-09-20 15:43:26 +00:00
archie
59b24d359e Fix a problem with the definition of HUGE_VAL causing the gcc warning
"cast increases required alignment of target type" on some platforms.

Reviewed by:	bde
2002-09-19 19:47:27 +00:00
marcel
e1124db23a We cannot use an alloc with only inputs and/or locals. The kernel
assumes that the parameters are passed in output registers. Remove
the alloc entirely, but don't depend on the kernel not trashing
our registers.
2002-08-23 03:47:50 +00:00
jake
d6dcf86c81 Generate the normal asm stubs for all sysv system calls. Use these instead
of C wrappers for the *sys indirect system calls.  The indirect system calls
are horribly broken on sparc64.

Submitted by:	tmm
2002-05-23 23:51:57 +00:00
obrien
1b12a0abfe Our default syscall.S is identical to what was explicitly being done for
Alpha and IA-64.  So there is no need for special handling.
2002-05-15 21:17:59 +00:00
marcel
7de93ed35c Implement _Unwind_FindTableEntry(). This function is part of GCC
for some configurations, but not for FreeBSD (yet?). Have one in
libc in the mean time.
2002-04-13 04:06:34 +00:00
obrien
b1189e8930 Fix the style of the SCM ID's.
I believe have made all of libc .h's as consistent as possible.
2002-03-22 23:42:05 +00:00
obrien
d90536e35b Fix the style of the SCM ID's.
I believe have made all of libc .c's as consistent as possible.
2002-03-22 21:53:29 +00:00