Errno is not set and the testcases will fail.
Now the scalbln-aliases are removed in i386/m68
and the wrappers are used when calling the scalbln-functions.
On ia64 only scalblnf has its own implementation.
For scalbln and scalblnl the ieee754/dbl-64 and ieee754/ldbl-96 are used, thus
the wrappers are needed, too.
Since long is 4 bytes for x32, we should use 3 bytes for __pad1 when
a long __pad1 is replaced by a byte __rwelision and __pad1.
* sysdeps/x86/nptl/bits/pthreadtypes.h (pthread_rwlock_t): Use
3 bytes for __pad1 for x32.
(__PTHREAD_RWLOCK_ELISION_EXTRA): Likewise.
In this patch we take advantage of HSW memory bandwidth, manage to
reduce miss branch prediction by avoiding using branch instructions and
force destination to be aligned with avx & avx2 instruction.
The CPU2006 403.gcc benchmark indicates this patch improves performance
from 26% to 59%.
* sysdeps/x86_64/multiarch/Makefile: Add memset-avx2.
* sysdeps/x86_64/multiarch/memset-avx2.S: New file.
* sysdeps/x86_64/multiarch/memset.S: Likewise.
* sysdeps/x86_64/multiarch/memset_chk.S: Likewise.
* sysdeps/x86_64/multiarch/rtld-memset.S: Likewise.
Due to my bad review suggestion for the fix for BZ #15089 a check
was removed from systrim to prevent sbrk being called with a zero
argument. Add the check back to avoid this useless work.
ChangeLog:
2014-06-19 Will Newton <will.newton@linaro.org>
* malloc/malloc.c (systrim): If extra is zero then return
early.
Implementation of strchr for AArch64. Speedups taken from micro-bench
show the improvements relative to the standard C code.
The use of LD1 means we have identical code for both big- and
little-endian systems.
This patch fixes __ieee754_logl (-LDBL_MAX) on x86_64 and x86 not to
subtract 1 from its argument and so cause spurious overflow in
FE_DOWNWARD mode. (For any argument strictly less than -1, it doesn't
matter whether or not 1 is subtracted before computing log1p, as long
as the result doesn't overflow to -Inf.)
Tested x86_64 and x86. (This particular case lacks test coverage,
since the testsuite doesn't cover -lieee, but it will be covered by
tests after the following patch to test pow in all rounding modes,
which was the context in which this bug was found.)
[BZ #17022]
* sysdeps/i386/fpu/e_logl.S (__ieee754_logl): Do not subtract 1
from arguments -2 or below.
* sysdeps/i386/i686/fpu/e_logl.S (__ieee754_logl): Likewise.
* sysdeps/x86_64/fpu/e_logl.S (__ieee754_logl): Likewise.
The glibc makefiles have a standard variable, $(rtld-prefix), to run
the dynamic linker with a default --library-path option; this is used
as the basis of lots of other variables for running programs compiled
with the newly built library.
A few places however use $(elf-objpfx)ld.so or
$(elf-objpfx)${rtld-installed-name} directly, with such a
--library-path option. This patch makes such places use
$(rtld-prefix) instead. I'm not aware of any significance in these
cases to the choice of ld.so or ${rtld-installed-name} when running
the dynamic linker, or to whether $(patsubst
%,:%,$(sysdep-library-path)) is included in the library-path as it is
in $(rtld-prefix) and just one of the places being changed.
Tested x86_64.
* elf/Makefile ($(objpfx)tst-unused-dep.out): Use $(rtld-prefix).
* iconvdata/Makefile ($(inst_gconvdir)/gconv-modules)
[$(cross-compiling) = no]: Likewise.
* sysdeps/s390/s390-64/Makefile ($(inst_gconvdir)/gconv-modules)
[$(cross-compiling) = no]: Likewise.
localedata/ChangeLog:
* Makefile (LOCALEDEF): Use $(rtld-prefix).
This patch fixes few failures in nearbyintl() where the fraction part is
close to 0.5.i The new tests added report few extra failures in
nearbyint_downward and nearbyint_towardzero which is a known issue.
Fixes#17031.
Add the missing fallback file for elide.h to fix non x86 builds.
Sorry about that. This is just a noop macro file that makes
all elision code to be optimized out.
With the recent tuning the C version of rwlocks is basically the same
performance as the x86 assembler version for uncontended locks (with a
a few cycles near the run-to-run variability). For others it should not
matter anyways.
So remove the assembler code and use the C version like other
architectures.
This patch relies on the C version of the rwlocks posted earlier.
With C rwlocks it is very straight forward to do adaptive elision
using TSX. It is based on the infrastructure added earlier
for mutexes, but uses its own elision macros. The macros
are fairly general purpose and could be used for other
elision purposes too.
This version is much cleaner than the earlier assembler based
version, and in particular implements adaptation which makes
it safer.
I changed the behavior slightly to not require any changes
in the test suite and fully conform to all expected
behaviors (generally at the cost of not eliding in
various situations). In particular this means the timedlock
variants are not elided. Nested trylock aborts.
One difference of the C versions to the assembler wr/rdlock
is that the C compiler saves some registers which are unnecessary
for the fast path in the prologue of the functions. Split the
uncontended fast path out into a separate function. Only when contention is
detected is the full featured function called. This makes
the fast path code (nearly) identical to the assembler version,
and gives uncontended performance within a few cycles.
v2: Rename some functions and add space.
The implementation of __get_nprocs uses a stactic variable to cache
the value of the current number of processors. The caching breaks when
'time (NULL) == 0':
$ cat nproc.c
#include <stdio.h>
#include <time.h>
#include <sys/time.h>
int main(int argc, char *argv[])
{
time_t t;
struct timeval tv = {0, 0};
printf("settimeofday({0, 0}, NULL) = %d\n", settimeofday(&tv, NULL));
t = time(NULL);
printf("Time: %d, CPUs: %d\n", (unsigned int)t, get_nprocs());
return 0;
}
$ gcc -O3 nproc.c
$ ./a.out
settimeofday({0, 0}, NULL) = -1
Time: 1401311578, CPUs: 4
$ sudo ./a.out
settimeofday({0, 0}, NULL) = 0
Time: 0, CPUs: 0
The problem is with the condition used to check whether a cached
value should be returned or not:
static int cached_result;
static time_t timestamp;
time_t now = time (NULL);
time_t prev = timestamp;
atomic_read_barrier ();
if (now == prev)
return cached_result;
This patch fixes the problem by ensuring that 'cached_result' has
been set at least once before returning it.
Continuing the series of patches to clean up conformtest expectations
for "POSIX" (1995/6) based on review of the expectations against the
standard, this patch cleans up expectations for sys/mman.h, sys/stat.h
and sys/types.h. Tested x86_64; no new XFAILs needed.
* conform/data/sys/mman.h-data [POSIX] (size_t): Do not require
type.
[POSIX] (off_t): Likewise.
* conform/data/sys/stat.h-data (S_IRGRP): Require constant.
[POSIX] (S_ISBLK): Require macro.
[POSIX] (S_ISCHR): Likewise.
[POSIX] (S_ISDIR): Likewise.
[POSIX] (S_ISFIFO): Likewise.
[POSIX] (S_ISREG): Likewise.
[POSIX || XPG3 || XPG4 || UNIX98] (S_TYPEISTMO): Do not list
optional-macro.
* conform/data/sys/types.h-data [POSIX] (blkcnt_t): Do not require
type.
[POSIX] (time_t): Likewise.
[POSIX] (timer_t): Likewise.
POSIX requires that we make a copy, so we allocate a new string
and free it in posix_spawn_file_actions_destroy.
Reported by David Reid, Alex Gaynor, and Glyph Lefkowitz. This bug
may have security implications.
