By using a combination of mask-and-add instead of the shift-based
index calculation the routines can share the same table as other
variants with no performance degradation.
The tables change name because of other changes in downstream AOR.
Reviewed-by: Wilco Dijkstra <Wilco.Dijkstra@arm.com>
The CORE-MATH exp2m1f implementation showed slight worse latency
when using x86_64 baseline ABI. This patch adds a ifunc variant
with similar performance for x86_64-v3.
Reviewed-by: Noah Goldstein <goldstein.w.n@gmail.com>
Reviewed-by: DJ Delorie <dj@redhat.com>
The CORE-MATH exp10m1f implementation showed slight worse latency
when using x86_64 baseline ABI. This patch adds a ifunc variant
with similar performance for x86_64-v3.
Reviewed-by: Noah Goldstein <goldstein.w.n@gmail.com>
Reviewed-by: DJ Delorie <dj@redhat.com>
The CORE-MATH implementation is correctly rounded (for any rounding mode)
and shows better performance compared to the generic exp2m1f.
The code was adapted to glibc style and to use the definition of
math_config.h (to handle errno, overflow, and underflow). The
only change is to handle FLT_MAX_EXP for FE_DOWNWARD or FE_TOWARDZERO.
The benchmark inputs are based on exp2f ones.
Benchtest on x64_64 (Ryzen 9 5900X, gcc 14.2.1), aarch64 (Neoverse-N1,
gcc 13.3.1), and powerpc (POWER10, gcc 13.2.1):
Latency master patched improvement
x86_64 40.6042 48.7104 -19.96%
x86_64v2 40.7506 35.9032 11.90%
x86_64v3 35.2301 31.7956 9.75%
i686 102.094 94.6657 7.28%
aarch64 18.2704 15.1387 17.14%
power10 11.9444 8.2402 31.01%
reciprocal-throughput master patched improvement
x86_64 20.8683 16.1428 22.64%
x86_64v2 19.5076 10.4474 46.44%
x86_64v3 19.2106 10.4014 45.86%
i686 56.4054 59.3004 -5.13%
aarch64 12.0781 7.3953 38.77%
power10 6.5306 5.9388 9.06%
The generic implementation calls __ieee754_exp2f and x86_64 provides
an optimized ifunc version (built with -mfma -mavx2, not correctly
rounded). This explains the performance difference for x86_64.
Same for i686, where the ABI provides an optimized __ieee754_exp2f
version built with '-msse2 -mfpmath=sse'. When built wth same
flags, the new algorithm shows a better performance:
master patched improvement
latency 102.094 91.2823 10.59%
reciprocal-throughput 56.4054 52.7984 6.39%
Signed-off-by: Alexei Sibidanov <sibid@uvic.ca>
Signed-off-by: Paul Zimmermann <Paul.Zimmermann@inria.fr>
Signed-off-by: Adhemerval Zanella <adhemerval.zanella@linaro.org>
Reviewed-by: DJ Delorie <dj@redhat.com>
The CORE-MATH implementation is correctly rounded (for any rounding mode)
and shows better performance compared to the generic exp10m1f.
The code was adapted to glibc style and to use the definition of
math_config.h (to handle errno, overflow, and underflow). I mostly
fixed some small issues in corner cases (sNaN handling, -INFINITY,
a specific overflow check).
Benchtest on x64_64 (Ryzen 9 5900X, gcc 14.2.1), aarch64 (Neoverse-N1,
gcc 13.3.1), and powerpc (POWER10, gcc 13.2.1):
Latency master patched improvement
x86_64 45.4690 49.5845 -9.05%
x86_64v2 46.1604 36.2665 21.43%
x86_64v3 37.8442 31.0359 17.99%
i686 121.367 93.0079 23.37%
aarch64 21.1126 15.0165 28.87%
power10 12.7426 8.4929 33.35%
reciprocal-throughput master patched improvement
x86_64 19.6005 17.4005 11.22%
x86_64v2 19.6008 11.1977 42.87%
x86_64v3 17.5427 10.2898 41.34%
i686 59.4215 60.9675 -2.60%
aarch64 13.9814 7.9173 43.37%
power10 6.7814 6.4258 5.24%
The generic implementation calls __ieee754_exp10f which has an
optimized version, although it is not correctly rounded, which is
the main culprit of the the latency difference for x86_64 and
throughp for i686.
Signed-off-by: Alexei Sibidanov <sibid@uvic.ca>
Signed-off-by: Paul Zimmermann <Paul.Zimmermann@inria.fr>
Signed-off-by: Adhemerval Zanella <adhemerval.zanella@linaro.org>
Reviewed-by: DJ Delorie <dj@redhat.com>
Also remove the use of builtins in favor of standard names, compiler
already inline them (if supported) with current compiler options.
It also fixes and issue where __builtin_roundeven is not support on
gcc older than version 10.
Checked on x86_64-linux-gnu and i686-linux_gnu.
Signed-off-by: Adhemerval Zanella <adhemerval.zanella@linaro.org>
Reviewed-by: DJ Delorie <dj@redhat.com>
This will be required by the rseq extensible ABI implementation on all
Linux architectures exposing the '__rseq_size' and '__rseq_offset'
symbols to set the initial value of the 'cpu_id' field which can be used
by applications to test if rseq is available and registered. As long as
the symbols are exposed it is valid for an application to perform this
test even if rseq is not yet implemented in libc for this architecture.
Both code paths are compile tested with build-many-glibcs.py but I don't
have access to any hardware to run the tests.
Signed-off-by: Michael Jeanson <mjeanson@efficios.com>
Reviewed-by: Arjun Shankar <arjun@redhat.com>
Save lr in a non-volatile register before scv in clone/clone3.
For clone, the non-volatile register was unused and already
saved/restored. Remove the dead code from clone.
Signed-off-by: Sachin Monga <smonga@linux.ibm.com>
Reviewed-by: Peter Bergner <bergner@linux.ibm.com>
In both routines, reduce register pressure such that GCC 14 emits no
spills for erf and fewer spills for erfc. Also use more efficient
comparison for the special-case in erf.
Benchtests show erf improves by 6.4%, erfc by 1.0%.
This avoids -Werror build issues in strace, which bundles UAPI
headers, but does not include them as system headers.
Fixes commit c444cc1d83
("Linux: Add missing scheduler constants to <sched.h>").
Reviewed-by: Carlos O'Donell <carlos@redhat.com>
Add basic tests of pthread_mutexattr_gettype and
pthread_mutexattr_settype with each valid mutex kind, plus test for
EINVAL with an invalid mutex kind.
Tested for x86_64.
The pthread_timedjoin_np and pthread_clockjoin_np functions do not
check that a valid time has been specified. The documentation for
these functions in the glibc manual isn't sufficiently detailed to say
if they should, but consistency with POSIX functions such as
pthread_mutex_timedlock and pthread_cond_timedwait strongly indicates
that an EINVAL error is appropriate (even if there might be some
ambiguity about exactly where such a check should go in relation to
other checks for whether the thread exists, whether it's immediately
joinable, etc.). Copy the logic for such a check used in
pthread_rwlock_common.c.
pthread_join_common had some logic calling valid_nanoseconds before
commit 9e92278ffa, "nptl: Remove
clockwait_tid"; I haven't checked exactly what cases that detected.
Tested for x86_64 and x86.
The commit 'sparc: Use Linux kABI for syscall return'
(86c5d2cf0c) did not take into account
a subtle sparc syscall kABI constraint. For syscalls that might block
indefinitely, on an interrupt (like SIGCONT) the kernel will set the
instruction pointer to just before the syscall:
arch/sparc/kernel/signal_64.c
476 static void do_signal(struct pt_regs *regs, unsigned long orig_i0)
477 {
[...]
525 if (restart_syscall) {
526 switch (regs->u_regs[UREG_I0]) {
527 case ERESTARTNOHAND:
528 case ERESTARTSYS:
529 case ERESTARTNOINTR:
530 /* replay the system call when we are done */
531 regs->u_regs[UREG_I0] = orig_i0;
532 regs->tpc -= 4;
533 regs->tnpc -= 4;
534 pt_regs_clear_syscall(regs);
535 fallthrough;
536 case ERESTART_RESTARTBLOCK:
537 regs->u_regs[UREG_G1] = __NR_restart_syscall;
538 regs->tpc -= 4;
539 regs->tnpc -= 4;
540 pt_regs_clear_syscall(regs);
541 }
However, on a SIGCONT it seems that 'g1' register is being clobbered after the
syscall returns. Before 86c5d2cf0c, the 'g1' was always placed jus
before the 'ta' instruction which then reloads the syscall number and restarts
the syscall.
On master, where 'g1' might be placed before 'ta':
$ cat test.c
#include <unistd.h>
int main ()
{
pause ();
}
$ gcc test.c -o test
$ strace -f ./t
[...]
ppoll(NULL, 0, NULL, NULL, 0
On another terminal
$ kill -STOP 2262828
$ strace -f ./t
[...]
--- SIGSTOP {si_signo=SIGSTOP, si_code=SI_USER, si_pid=2521813, si_uid=8289} ---
--- stopped by SIGSTOP ---
And then
$ kill -CONT 2262828
Results in:
--- SIGCONT {si_signo=SIGCONT, si_code=SI_USER, si_pid=2521813, si_uid=8289} ---
restart_syscall(<... resuming interrupted ppoll ...>) = -1 EINTR (Interrupted system call)
Where the expected behaviour would be:
$ strace -f ./t
[...]
ppoll(NULL, 0, NULL, NULL, 0) = ? ERESTARTNOHAND (To be restarted if no handler)
--- SIGSTOP {si_signo=SIGSTOP, si_code=SI_USER, si_pid=2521813, si_uid=8289} ---
--- stopped by SIGSTOP ---
--- SIGCONT {si_signo=SIGCONT, si_code=SI_USER, si_pid=2521813, si_uid=8289} ---
ppoll(NULL, 0, NULL, NULL, 0
Just moving the 'g1' setting near the syscall asm is not suffice,
the compiler might optimize it away (as I saw on cancellation.c by
trying this fix). Instead, I have change the inline asm to put the
'g1' setup in ithe asm block. This would require to change the asm
constraint for INTERNAL_SYSCALL_NCS, since the syscall number is not
constant.
Checked on sparc64-linux-gnu.
Reported-by: René Rebe <rene@exactcode.de>
Tested-by: Sam James <sam@gentoo.org>
Reviewed-by: Sam James <sam@gentoo.org>
The CORE-MATH implementation is correctly rounded (for any rounding mode).
This can be checked by exhaustive tests in a few minutes since there are
less than 2^32 values to check against for example GNU MPFR.
This patch also adds some bench values for tgammaf.
Tested on x86_64 and x86 (cfarm26).
With the initial GNU libc code it gave on an Intel(R) Core(TM) i7-8700:
"tgammaf": {
"": {
"duration": 3.50188e+09,
"iterations": 2e+07,
"max": 602.891,
"min": 65.1415,
"mean": 175.094
}
}
With the new code:
"tgammaf": {
"": {
"duration": 3.30825e+09,
"iterations": 5e+07,
"max": 211.592,
"min": 32.0325,
"mean": 66.1649
}
}
With the initial GNU libc code it gave on cfarm26 (i686):
"tgammaf": {
"": {
"duration": 3.70505e+09,
"iterations": 6e+06,
"max": 2420.23,
"min": 243.154,
"mean": 617.509
}
}
With the new code:
"tgammaf": {
"": {
"duration": 3.24497e+09,
"iterations": 1.8e+07,
"max": 1238.15,
"min": 101.155,
"mean": 180.276
}
}
Signed-off-by: Alexei Sibidanov <sibid@uvic.ca>
Signed-off-by: Paul Zimmermann <Paul.Zimmermann@inria.fr>
Changes in v2:
- include <math.h> (fix the linknamespace failures)
- restored original benchtests/strcoll-inputs/filelist#en_US.UTF-8 file
- restored original wrapper code (math/w_tgammaf_compat.c),
except for the dealing with the sign
- removed the tgammaf/float entries in all libm-test-ulps files
- address other comments from Joseph Myers
(https://sourceware.org/pipermail/libc-alpha/2024-July/158736.html)
Changes in v3:
- pass NULL argument for signgam from w_tgammaf_compat.c
- use of math_narrow_eval
- added more comments
Changes in v4:
- initialize local_signgam to 0 in math/w_tgamma_template.c
- replace sysdeps/ieee754/dbl-64/gamma_productf.c by dummy file
Changes in v5:
- do not mention local_signgam any more in math/w_tgammaf_compat.c
- initialize local_signgam to 1 instead of 0 in w_tgamma_template.c
and added comment
Changes in v6:
- pass NULL as 2nd argument of __ieee754_gammaf_r in
w_tgammaf_compat.c, and check for NULL in e_gammaf_r.c
Changes in v7:
- added Signed-off-by line for Alexei Sibidanov (author of the code)
Changes in v8:
- added Signed-off-by line for Paul Zimmermann (submitted of the patch)
Changes in v9:
- address comments from review by Adhemerval Zanella
Reviewed-by: Adhemerval Zanella <adhemerval.zanella@linaro.org>
Linux 6.11 adds the new flag for pwritev2 (commit
c34fc6f26ab86d03a2d47446f42b6cd492dfdc56).
Checked on x86_64-linux-gnu on 6.11 kernel.
Reviewed-by: H.J. Lu <hjl.tools@gmail.com>
This patch updates the kernel version in the tests tst-mount-consts.py,
and tst-sched-consts.py to 6.11.
There are no new constants covered by these tests in 6.11.
Tested with build-many-glibcs.py.
Reviewed-by: H.J. Lu <hjl.tools@gmail.com>
This request the page to be never written out to swap, it will be zeroed
under memory pressure (so kernel can just drop the page), it is inherited
by fork, it is not counted against @code{mlock} budget, and if there is
no enough memory to service a page faults there is no fatal error (so not
signal is sent).
Tested with build-many-glibcs.py.
Reviewed-by: H.J. Lu <hjl.tools@gmail.com>
Linux 6.11 adds some more PIDFD_* constants for 'pidfs: allow retrieval
of namespace file descriptors'
(5b08bd408534bfb3a7cf5778da5b27d4e4fffe12).
Tested with build-many-glibcs.py.
Reviewed-by: H.J. Lu <hjl.tools@gmail.com>
Linux 6.11 changes for syscall are:
* fstat/newfstatat for loongarch (it should be safe to add since
255dc1e4ed that undefine them).
* clone3 for nios2, which only adds the entry point but defined
__ARCH_BROKEN_SYS_CLONE3 (the syscall will always return ENOSYS).
* uretprobe for x86_64 and x32.
Update syscall-names.list and regenerate the arch-syscall.h headers
with build-many-glibcs.py update-syscalls.
Tested with build-many-glibcs.py.
Reviewed-by: H.J. Lu <hjl.tools@gmail.com>
GCC mainline produces a -Wheader-guard error building for x86_64-gnu.
Fix what seems to be incorrect macro naming in the #ifndef
conditional.
Tested with build-many-glibc.py for x86_64-gnu (GCC mainline).
Message-ID: <fd800046-5ecb-ebd5-4df1-29d4eb3d5433@redhat.com>
The recursive lock used on abort does not synchronize with a new process
creation (either by fork-like interfaces or posix_spawn ones), nor it
is reinitialized after fork().
Also, the SIGABRT unblock before raise() shows another race condition,
where a fork or posix_spawn() call by another thread, just after the
recursive lock release and before the SIGABRT signal, might create
programs with a non-expected signal mask. With the default option
(without POSIX_SPAWN_SETSIGDEF), the process can see SIG_DFL for
SIGABRT, where it should be SIG_IGN.
To fix the AS-safe, raise() does not change the process signal mask,
and an AS-safe lock is used if a SIGABRT is installed or the process
is blocked or ignored. With the signal mask change removal,
there is no need to use a recursive loc. The lock is also taken on
both _Fork() and posix_spawn(), to avoid the spawn process to see the
abort handler as SIG_DFL.
A read-write lock is used to avoid serialize _Fork and posix_spawn
execution. Both sigaction (SIGABRT) and abort() requires to lock
as writer (since both change the disposition).
The fallback is also simplified: there is no need to use a loop of
ABORT_INSTRUCTION after _exit() (if the syscall does not terminate the
process, the system is broken).
The proposed fix changes how setjmp works on a SIGABRT handler, where
glibc does not save the signal mask. So usage like the below will now
always abort.
static volatile int chk_fail_ok;
static jmp_buf chk_fail_buf;
static void
handler (int sig)
{
if (chk_fail_ok)
{
chk_fail_ok = 0;
longjmp (chk_fail_buf, 1);
}
else
_exit (127);
}
[...]
signal (SIGABRT, handler);
[....]
chk_fail_ok = 1;
if (! setjmp (chk_fail_buf))
{
// Something that can calls abort, like a failed fortify function.
chk_fail_ok = 0;
printf ("FAIL\n");
}
Such cases will need to use sigsetjmp instead.
The _dl_start_profile calls sigaction through _profil, and to avoid
pulling abort() on loader the call is replaced with __libc_sigaction.
Checked on x86_64-linux-gnu and aarch64-linux-gnu.
Reviewed-by: DJ Delorie <dj@redhat.com>
The BZ#24967 fix (1bdda52fe9) missed the time for
architectures that define USE_IFUNC_TIME. Although it is not
an issue, since there is no pointer mangling, there is also no need
to call dl_vdso_vsym since the vDSO setup was already done by the
loader.
Checked on x86_64-linux-gnu and i686-linux-gnu.
The BZ#24967 fix (1bdda52fe9) missed the gettimeofday for
architectures that define USE_IFUNC_GETTIMEOFDAY. Although it is not
an issue, since there is no pointer mangling, there is also no need
to call dl_vdso_vsym since the vDSO setup was already done by the
loader.
Checked on x86_64-linux-gnu and i686-linux-gnu.
Building the s390 specific iconv modules - utf16-utf32-z9.c, utf8-utf32-z9.c
and utf8-utf16-z9.c - with -fno-omit-frame-pointer leads to a build error
"error: %r11 cannot be used in 'asm' here" as r11 is needed as frame-pointer.
The cuXY-instructions need two even-odd register pairs. Therefore the register
pinning is used. This patch just uses a different register pair.
Reviewed-by: Florian Weimer <fweimer@redhat.com>
Several copies of the licenses in files contained whitespace related
problems. Two cases are addressed here, the first is two spaces
after a period which appears between "PURPOSE." and "See". The other
is a space after the last forward slash in the URL. Both issues are
corrected and the licenses now match the official textual description
of the license (and the other license in the sources).
Since these whitespaces changes do not alter the paragraph structure of
the license, nor create new sentences, they do not change the license.
The sparc clone mitigation (faeaa3bc9f) added the use of
flushw, which is not support by LEON/sparcv8. As discussed on
the libc-alpha, 'ta 3' is a working alternative [1].
[1] https://sourceware.org/pipermail/libc-alpha/2024-August/158905.html
Checked with a build for sparcv8-linux-gnu targetting leon.
Acked-by: John Paul Adrian Glaubitz <glaubitz@physik.fu-berlin.de>
LEON2/LEON3 are both sparcv8, which does not support branch hints
(bne,pn) nor the return instruction.
Checked with a build for sparcv8-linux-gnu targetting leon. I also
checked some cancellation tests with qemu-system (targeting LEON3).
Acked-by: John Paul Adrian Glaubitz <glaubitz@physik.fu-berlin.de>
GCC aligns global data to 16 bytes if their size is >= 16 bytes. This patch
changes the exp2f_data struct slightly so that the fields are better aligned.
As a result on targets that support them, load-pair instructions accessing
poly_scaled and invln2_scaled are now 16-byte aligned.
Reviewed-by: Adhemerval Zanella <adhemerval.zanella@linaro.org>
The loop should be aligned to 32-bytes so that it can ideally run out
the DSB. This is particularly important on Skylake-Server where
deficiencies in it's DSB implementation make it prone to not being
able to run loops out of the DSB.
For example running strcmp-evex on 200Mb string:
32-byte aligned loop:
- 43,399,578,766 idq.dsb_uops
not 32-byte aligned loop:
- 6,060,139,704 idq.dsb_uops
This results in a 25% performance degradation for the non-aligned
version.
The fix is to just ensure the code layout is such that the loop is
aligned. (Which was previously the case but was accidentally dropped
in 84e7c46df).
NB: The fix was actually 64-byte alignment. This is because 64-byte
alignment generally produces more stable performance than 32-byte
aligned code (cache line crosses can affect perf), so if we are going
past 16-byte alignmnent, might as well go to 64. 64-byte alignment
also matches most other functions we over-align, so it creates a
common point of optimization.
Times are reported as ratio of Time_With_Patch /
Time_Without_Patch. Lower is better.
The values being reported is the geometric mean of the ratio across
all tests in bench-strcmp and bench-strncmp.
Note this patch is only attempting to improve the Skylake-Server
strcmp for long strings. The rest of the numbers are only to test for
regressions.
Tigerlake Results Strings <= 512:
strcmp : 1.026
strncmp: 0.949
Tigerlake Results Strings > 512:
strcmp : 0.994
strncmp: 0.998
Skylake-Server Results Strings <= 512:
strcmp : 0.945
strncmp: 0.943
Skylake-Server Results Strings > 512:
strcmp : 0.778
strncmp: 1.000
The 2.6% regression on TGL-strcmp is due to slowdowns caused by
changes in alignment of code handling small sizes (most on the
page-cross logic). These should be safe to ignore because 1) We
previously only 16-byte aligned the function so this behavior is not
new and was essentially up to chance before this patch and 2) this
type of alignment related regression on small sizes really only comes
up in tight micro-benchmark loops and is unlikely to have any affect
on realworld performance.
Reviewed-by: H.J. Lu <hjl.tools@gmail.com>
This hides the inconsistent TCB state (missing robust mutex list) from
signal handlers.
Reviewed-by: Adhemerval Zanella <adhemerval.zanella@linaro.org>
In Linux 6.11, fstat and newfstatat are added back. To avoid the messy
usage of the fstat, newfstatat, and statx system calls, we will continue
using statx only in glibc, maintaining consistency with previous versions of
the LoongArch-specific glibc implementation.
Signed-off-by: caiyinyu <caiyinyu@loongson.cn>
Reviewed-by: Xi Ruoyao <xry111@xry111.site>
Suggested-by: Florian Weimer <fweimer@redhat.com>
This operation can be simplified to use simpler multiply-round-convert
sequence, which uses fewer instructions and constants.
Reviewed-by: Wilco Dijkstra <Wilco.Dijkstra@arm.com>
Rearrange operations so MOV is not necessary in reduction or around
the special-case handler. Reduce memory access by using more indexed
MLAs in polynomial.
Reviewed-by: Wilco Dijkstra <Wilco.Dijkstra@arm.com>
log1pf is quite register-intensive - use fewer registers for the
polynomial, and make various changes to shorten dependency chains in
parent routines. There is now no spilling with GCC 14. Accuracy moves
around a little - comments adjusted accordingly but does not require
regen-ulps.
Use the helper in log1pf as well, instead of having separate
implementations. The more accurate polynomial means special-casing can
be simplified, and the shorter dependency chain avoids the usual dance
around v0, which is otherwise difficult.
There is a small duplication of vectors containing 1.0f (or 0x3f800000) -
GCC is not currently able to efficiently handle values which fit in FMOV
but not MOVI, and are reinterpreted to integer. There may be potential
for more optimisation if this is fixed.
Reviewed-by: Wilco Dijkstra <Wilco.Dijkstra@arm.com>
