Add a strncmp testcase to cover cases where one of strings ends on the
page boundary with the maximum string length less than the number bytes
of each AVX2 loop iteration and different offsets from page boundary.
The updated string/test-strncmp fails on Intel Core i7-8559U without
ommit 1c6432316bc434a72108d7b0c7cfbfdde64c3124
Author: Sunil K Pandey <skpgkp1@gmail.com>
Date: Fri Jun 12 08:57:16 2020 -0700
Fix avx2 strncmp offset compare condition check [BZ #25933]
Similarly to Maciej's changes to fix the build of rawmemchr in the
presence of GCC 11's -Wstringop-overread, also disable that option in
two string function tests that have similar warnings and other string
function warnings already disabled.
Tested with build-many-glibcs.py for aarch64-linux-gnu and
arm-linux-gnueabi that it fixes building the glibc testsuite.
Fix a compilation error:
In function '__rawmemchr',
inlined from '__rawmemchr' at rawmemchr.c:27:1:
rawmemchr.c:36:12: error: 'memchr' specified bound 18446744073709551615 exceeds maximum object size 9223372036854775807 [-Werror=stringop-overread]
36 | return memchr (s, c, (size_t)-1);
| ^~~~~~~~~~~~~~~~~~~~~~~~~
cc1: all warnings being treated as errors
../o-iterator.mk:9: recipe for target '.../string/rawmemchr.o' failed
introduced with GCC 11 commit d14c547abd48 ("Add -Wstringop-overread
for reading past the end by string functions.").
Without msgfmt libc.mo files are not generated and its loading failure
is silent ignored with xsetlocale.
Also unset LANGUAGE environment variable to avoid it taking precedence
when loading the message catalog. Although not strictly required
(since the test is issued with test-container and it sets a strict
environment variable) it follows other tests that deal with
translation.
Checked on x86_64-linux-gnu.
* string/tst-strsignal.c (do_test): Actually check that RT signals are
available by comparing SIGRTMAX to SIGRTMIN. Check that SIGRTMAX is 64
before testing for a message reporting 65 for SIGRTMAX+1.
The strerrorname_np returns error number name (e.g. "EINVAL" for EINVAL)
while strerrordesc_np returns string describing error number (e.g
"Invalid argument" for EINVAL). Different than strerror,
strerrordesc_np does not attempt to translate the return description,
both functions return NULL for an invalid error number.
They should be used instead of sys_errlist and sys_nerr, both are
thread and async-signal safe. These functions are GNU extensions.
Checked on x86-64-linux-gnu, i686-linux-gnu, powerpc64le-linux-gnu,
and s390x-linux-gnu.
Tested-by: Carlos O'Donell <carlos@redhat.com>
Reviewed-by: Carlos O'Donell <carlos@redhat.com>
The sigabbrev_np returns the abbreviated signal name (e.g. "HUP" for
SIGHUP) while sigdescr_np returns the string describing the error
number (e.g "Hangup" for SIGHUP). Different than strsignal,
sigdescr_np does not attempt to translate the return description and
both functions return NULL for an invalid signal number.
They should be used instead of sys_siglist or sys_sigabbrev and they
are both thread and async-signal safe. They are added as GNU
extensions on string.h header (same as strsignal).
Checked on x86-64-linux-gnu, i686-linux-gnu, powerpc64le-linux-gnu,
and s390x-linux-gnu.
Tested-by: Carlos O'Donell <carlos@redhat.com>
Reviewed-by: Carlos O'Donell <carlos@redhat.com>
Checked on x86-64-linux-gnu, i686-linux-gnu, powerpc64le-linux-gnu,
and s390x-linux-gnu.
Tested-by: Carlos O'Donell <carlos@redhat.com>
Reviewed-by: Carlos O'Donell <carlos@redhat.com>
Checked on x86-64-linux-gnu, i686-linux-gnu, powerpc64le-linux-gnu,
and s390x-linux-gnu.
Tested-by: Carlos O'Donell <carlos@redhat.com>
Reviewed-by: Carlos O'Donell <carlos@redhat.com>
Checked on x86-64-linux-gnu, i686-linux-gnu, powerpc64le-linux-gnu,
and s390x-linux-gnu.
Tested-by: Carlos O'Donell <carlos@redhat.com>
Reviewed-by: Carlos O'Donell <carlos@redhat.com>
Use snprintf instead of mempcpy plus itoa_word and remove unused
definitions. There is no potential for infinite recursion because
snprintf only use strerror_r for the %m specifier.
Checked on x86-64-linux-gnu, i686-linux-gnu, powerpc64le-linux-gnu,
and s390x-linux-gnu.
Tested-by: Carlos O'Donell <carlos@redhat.com>
Reviewed-by: Carlos O'Donell <carlos@redhat.com>
The buffer allocation uses the same strategy of strsignal.
Checked on x86-64-linux-gnu, i686-linux-gnu, powerpc64le-linux-gnu,
and s390x-linux-gnu.
Tested-by: Carlos O'Donell <carlos@redhat.com>
Reviewed-by: Carlos O'Donell <carlos@redhat.com>
If the thread is terminated then __libc_thread_freeres will free the
storage via __glibc_tls_internal_free.
It is only within the calling thread that this matters. It makes
strerror MT-safe.
Checked on x86-64-linux-gnu, i686-linux-gnu, powerpc64le-linux-gnu,
and s390x-linux-gnu.
Tested-by: Carlos O'Donell <carlos@redhat.com>
Reviewed-by: Carlos O'Donell <carlos@redhat.com>
The per-thread state is refactored two use two strategies:
1. The default one uses a TLS structure, which will be placed in the
static TLS space (using __thread keyword).
2. Linux allocates via struct pthread and access it through THREAD_*
macros.
The default strategy has the disadvantage of increasing libc.so static
TLS consumption and thus decreasing the possible surplus used in
some scenarios (which might be mitigated by BZ#25051 fix).
It is used only on Hurd, where accessing the thread storage in the in
single thread case is not straightforward (afaiu, Hurd developers could
correct me here).
The fallback static allocation used for allocation failure is also
removed: defining its size is problematic without synchronizing with
translated messages (to avoid partial translation) and the resulting
usage is not thread-safe.
Checked on x86-64-linux-gnu, i686-linux-gnu, powerpc64le-linux-gnu,
and s390x-linux-gnu.
Tested-by: Carlos O'Donell <carlos@redhat.com>
Reviewed-by: Carlos O'Donell <carlos@redhat.com>
The symbol is deprecated by strerror since its usage imposes some issues
such as copy relocations.
Its internal name is also changed to _sys_errlist_internal to avoid
static linking usage. The compat code is also refactored by removing
the over enginered errlist-compat.c generation from manual entried and
extra comment token in linker script file. It disantangle the code
generation from manual and simplify both Linux and Hurd compat code.
The definitions from errlist.c are moved to errlist.h and a new test
is added to avoid a new errno entry without an associated one in manual.
Checked on x86_64-linux-gnu and i686-linux-gnu. I also run a check-abi
on all affected platforms.
Tested-by: Carlos O'Donell <carlos@redhat.com>
Reviewed-by: Carlos O'Donell <carlos@redhat.com>
The symbol was deprecated by strsignal and its usage imposes issues
such as copy relocations.
Its internal name is changed to __sys_siglist and __sys_sigabbrev to
avoid static linking usage. The compat code is also refactored, since
both Linux and Hurd usage the same strategy: export the same array with
different object sizes.
The libSegfault change avoids calling strsignal on the SIGFAULT signal
handler (the current usage is already sketchy, adding a call that
potentially issue locale internal function is even sketchier).
Checked on x86_64-linux-gnu and i686-linux-gnu. I also run a check-abi
on all affected platforms.
Reviewed-by: Carlos O'Donell <carlos@redhat.com>
GCC 8 relaxed what kind of expressions can be used in initializers,
and the previous use of static const variables relied on that. Switch
to wide (non-int) enum constants instead, which is another GCC
extension that is more widely implemented.
Reviewed-by: Carlos O'Donell <carlos@redhat.com>
Tested-by: Carlos O'Donell <carlos@redhat.com>
Adds the access attribute newly introduced in GCC 10 to the subset of
function declarations that are already covered by _FORTIFY_SOURCE and
that don't have corresponding GCC built-in equivalents.
Reviewed-by: DJ Delorie <dj@redhat.com>
Without the asm redirects, strchr et al. are not const-correct.
libc++ has a wrapper header that works with and without
__CORRECT_ISO_CPP_STRING_H_PROTO (using a Clang extension). But when
Clang is used with libstdc++ or just C headers, the overloaded functions
with the correct types are not declared.
This change does not impact current GCC (with libstdc++ or libc++).
If the specified needle crosses a page-boundary, the s390-z15 ifunc variant of
strstr truncates the needle which results in invalid results.
This is fixed by loading the needle beyond the page boundary to v18 instead of v16.
The bug is sometimes observable in test-strstr.c in check1 and check2 as the
haystack and needle is stored on stack. Thus the needle can be on a page boundary.
check2 is now extended to test haystack / needles located on stack, at end of page
and on two pages.
This bug was introduced with commit 6f47401bd5
("S390: Add arch13 strstr ifunc variant.") and is already released in glibc 2.30.
As for gettimeofday, time will be implemented based on clock_gettime
on all platforms and internal code should use clock_gettime
directly. In addition to removing a layer of indirection, this will
allow us to remove the PLT-bypass gunk for gettimeofday.
The changed code always assumes __clock_gettime (CLOCK_REALTIME)
or __clock_gettime (CLOCK_REALTIME_COARSE) (for Linux case) cannot
fail, using the same rationale for gettimeofday change. And internal
helper was added (time_now).
Checked on x86_64-linux-gnu, i686-linux-gnu, powerpc64le-linux-gnu,
powerpc64-linux-gnu, and powerpc-linux-gnu.
Reviewed-by: Lukasz Majewski <lukma@denx.de>
Commit 69fd157a3 "time: Add padding for the timespec if required"
caused a breakage in the glibc tests as the endian.h include file was
kept in the networking headers while the __USE_MISC #ifdefs had been
removed. This resulted in namespace violations in the networking
headers.
This patche restores the __USE_MISC conditionals in endian.h to fix the
test failures.
* string/endian.h: Restore the __USE_MISC conditionals.
string/tester.c contains code that correctly triggers various GCC
warnings about dubious uses of string functions (uses that are being
deliberately tested there), and duly disables those warnings around
the relevant code.
A change in GCC mainline resulted in this code failing to compile with
a -Warray-bounds error, despite the location with the error having
-Warray-bounds already disabled. This has been reported as
<https://gcc.gnu.org/bugzilla/show_bug.cgi?id=91890>. This patch
avoids that problem and possible future issues with these diagnostics
by moving all the warning disabling in this file to top level, as
suggested by Florian in
<https://sourceware.org/ml/libc-alpha/2019-10/msg00033.html>, rather
than only doing it locally around specific function calls.
Tested with build-many-glibcs.py for aarch64-linux-gnu with GCC
mainline (with only the conform/ failures noted in
<https://sourceware.org/ml/libc-alpha/2019-10/msg00043.html>).
* string/tester.c: Ignore -Warray-bounds and
-Wmemset-transposed-args at top level.
[__GNUC_PREREQ (7, 0)]: Ignore -Wrestrict and -Wstringop-overflow=
at top level.
[__GNUC_PREREQ (8, 0)]: Ignore -Wstringop-truncation at top level.
(test_stpncpy): Do not ignore warnings here.
(test_strncat): Likewise.
(test_strncpy): Likewise.
(test_memset): Likewise.
With only two exceptions (sys/types.h and sys/param.h, both of which
historically might have defined BYTE_ORDER) the public headers that
include <endian.h> only want to be able to test __BYTE_ORDER against
__*_ENDIAN.
This patch creates a new bits/endian.h that can be included by any
header that wants to be able to test __BYTE_ORDER and/or
__FLOAT_WORD_ORDER against the __*_ENDIAN constants, or needs
__LONG_LONG_PAIR. It only defines macros in the implementation
namespace.
The existing bits/endian.h (which could not be included independently
of endian.h, and only defines __BYTE_ORDER and maybe __FLOAT_WORD_ORDER)
is renamed to bits/endianness.h. I also took the opportunity to
canonicalize the form of this header, which we are stuck with having
one copy of per architecture. Since they are so short, this means git
doesn’t understand that they were renamed from existing headers, sigh.
endian.h itself is a nonstandard header and its only remaining use
from a standard header is guarded by __USE_MISC, so I dropped the
__USE_MISC conditionals from around all of the public-namespace things
it defines. (This means, an application that requests strict library
conformance but includes endian.h will still see the definition of
BYTE_ORDER.)
A few changes to specific bits/endian(ness).h variants deserve
mention:
- sysdeps/unix/sysv/linux/ia64/bits/endian.h is moved to
sysdeps/ia64/bits/endianness.h. If I remember correctly, ia64 did
have selectable endianness, but we have assembly code in
sysdeps/ia64 that assumes it’s little-endian, so there is no reason
to treat the ia64 endianness.h as linux-specific.
- The C-SKY port does not fully support big-endian mode, the compile
will error out if __CSKYBE__ is defined.
- The PowerPC port had extra logic in its bits/endian.h to detect a
broken compiler, which strikes me as unnecessary, so I removed it.
- The only files that defined __FLOAT_WORD_ORDER always defined it to
the same value as __BYTE_ORDER, so I removed those definitions.
The SH bits/endian(ness).h had comments inconsistent with the
actual setting of __FLOAT_WORD_ORDER, which I also removed.
- I *removed* copyright boilerplate from the few bits/endian(ness).h
headers that had it; these files record a single fact in a fashion
dictated by an external spec, so I do not think they are copyrightable.
As long as I was changing every copy of ieee754.h in the tree, I
noticed that only the MIPS variant includes float.h, because it uses
LDBL_MANT_DIG to decide among three different versions of
ieee854_long_double. This patch makes it not include float.h when
GCC’s intrinsic __LDBL_MANT_DIG__ is available.
* string/endian.h: Unconditionally define LITTLE_ENDIAN,
BIG_ENDIAN, PDP_ENDIAN, and BYTE_ORDER. Condition byteswapping
macros only on !__ASSEMBLER__. Move the definitions of
__BIG_ENDIAN, __LITTLE_ENDIAN, __PDP_ENDIAN, __FLOAT_WORD_ORDER,
and __LONG_LONG_PAIR to...
* string/bits/endian.h: ...this new file, which includes
the renamed header bits/endianness.h for the definition of
__BYTE_ORDER and possibly __FLOAT_WORD_ORDER.
* string/Makefile: Install bits/endianness.h.
* include/bits/endian.h: New wrapper.
* bits/endian.h: Rename to bits/endianness.h.
Add multiple-include guard. Rewrite the comment explaining what
the machine-specific variants of this file should do.
* sysdeps/unix/sysv/linux/ia64/bits/endian.h:
Move to sysdeps/ia64.
* sysdeps/aarch64/bits/endian.h
* sysdeps/alpha/bits/endian.h
* sysdeps/arm/bits/endian.h
* sysdeps/csky/bits/endian.h
* sysdeps/hppa/bits/endian.h
* sysdeps/ia64/bits/endian.h
* sysdeps/m68k/bits/endian.h
* sysdeps/microblaze/bits/endian.h
* sysdeps/mips/bits/endian.h
* sysdeps/nios2/bits/endian.h
* sysdeps/powerpc/bits/endian.h
* sysdeps/riscv/bits/endian.h
* sysdeps/s390/bits/endian.h
* sysdeps/sh/bits/endian.h
* sysdeps/sparc/bits/endian.h
* sysdeps/x86/bits/endian.h:
Rename to endianness.h; canonicalize form of file; remove
redundant definitions of __FLOAT_WORD_ORDER.
* sysdeps/powerpc/bits/endianness.h: Remove logic to check for
broken compilers.
* ctype/ctype.h
* sysdeps/aarch64/nptl/bits/pthreadtypes-arch.h
* sysdeps/arm/nptl/bits/pthreadtypes-arch.h
* sysdeps/csky/nptl/bits/pthreadtypes-arch.h
* sysdeps/ia64/ieee754.h
* sysdeps/ieee754/ieee754.h
* sysdeps/ieee754/ldbl-128/ieee754.h
* sysdeps/ieee754/ldbl-128ibm/ieee754.h
* sysdeps/m68k/nptl/bits/pthreadtypes-arch.h
* sysdeps/microblaze/nptl/bits/pthreadtypes-arch.h
* sysdeps/mips/ieee754/ieee754.h
* sysdeps/mips/nptl/bits/pthreadtypes-arch.h
* sysdeps/nios2/nptl/bits/pthreadtypes-arch.h
* sysdeps/nptl/pthread.h
* sysdeps/riscv/nptl/bits/pthreadtypes-arch.h
* sysdeps/sh/nptl/bits/pthreadtypes-arch.h
* sysdeps/sparc/sparc32/ieee754.h
* sysdeps/unix/sysv/linux/generic/bits/stat.h
* sysdeps/unix/sysv/linux/generic/bits/statfs.h
* sysdeps/unix/sysv/linux/sys/acct.h
* wctype/bits/wctype-wchar.h:
Include bits/endian.h, not endian.h.
* sysdeps/unix/sysv/linux/hppa/pthread.h: Don’t include endian.h.
* sysdeps/mips/ieee754/ieee754.h: Use __LDBL_MANT_DIG__
in ifdefs, instead of LDBL_MANT_DIG. Only include float.h
when __LDBL_MANT_DIG__ is not predefined, in which case
define __LDBL_MANT_DIG__ to equal LDBL_MANT_DIG.
Use the generic C memset/memcpy/memmove in benchtests since comparing
against a slow byte-oriented implementation makes no sense.
Reviewed-by: Adhemerval Zanella <adhemerval.zanella@linaro.org>
2019-08-29 Wilco Dijkstra <wdijkstr@arm.com>
* benchtests/bench-memcpy.c (simple_memcpy): Remove.
(generic_memcpy): Include generic C memcpy.
* benchtests/bench-memmove.c (simple_memmove): Remove.
(generic_memmove): Include generic C memmove.
* benchtests/bench-memset.c (simple_memset): Remove.
(generic_memset): Include generic C memset.
* benchtests/bench-memset-large.c (simple_memset): Remove.
(generic_memset): Include generic C memset.
* benchtests/bench-memset-walk.c (simple_memset): Remove.
(generic_memset): Include generic C memset.
* string/memcpy.c (MEMCPY): Add defines to enable redirection.
* string/memset.c (MEMSET): Likewise.
* sysdeps/x86_64/memcopy.h: Remove empty file.
It doesn't make sense to remove all the internal uses of time.
It's still a standard ISO C function, and its callers don't need
sub-second resolution and would be unnecessarily complicated if
they had to declare a struct timespec instead of just a time_t.
However, a handful of places were using the vestigial "result"
argument instead of the return value, which is slightly less
efficient and also looks strange. Correct this.
* misc/syslog.c (__vsyslog_internal)
* time/getdate.c (__getdate_r)
* time/tst_wcsftime.c (main):
Use return value of time, not its argument.
* string/strfry.c (strfry)
* sysdeps/mach/sleep.c (__sleep):
Remove unnecessary casts of NULL in calls to time.
C2X adds the memccpy, strdup and strndup functions. This patch duly
adds __GLIBC_USE (ISOC2X) to the conditions under which <string.h>
declares them.
Tested for x86_64.
* string/string.h (memccpy): Also declare if [__GLIBC_USE (ISOC2X)].
(strdup): Likewise.
(strndup): Likewise.
This patch fixes the following gcc 9 warnings for "make xcheck" / "make bench":
-string/tst-strcasestr.c:
../include/bits/../../misc/bits/error.h:42:5: error: ‘%s’ directive argument is null [-Werror=format-overflow=]
-argp/argp-test.c:
argp-test.c:130:20: error: ‘%d’ directive writing between 1 and 11 bytes into a region of size 10 [-Werror=format-overflow=]
argp-test.c:130:19: note: directive argument in the range [-2147483648, 122]
argp-test.c:130:5: note: ‘sprintf’ output between 2 and 12 bytes into a destination of size 10
-nss/tst-field.c:
tst-field.c:52:7: error: ‘%s’ directive argument is null [-Werror=format-overflow=]
-benchtests/bench-strstr.c:
../include/bits/../../misc/bits/error.h:42:5: error: ‘%s’ directive argument is null [-Werror=format-overflow=]
-benchtests/bench-malloc-simple.c:
bench-malloc-simple.c:93:16: error: iteration 3 invokes undefined behavior [-Werror=aggressive-loop-optimizations]
ChangeLog:
[BZ #24556]
* string/test-strcasestr.c (check_result): Add NULL check.
* nss/tst-field.c (check_rewrite): Likewise.
* benchtests/bench-strstr.c (do_one_test): Likewise.
* string/test-strstr.c (check_result): Likewise.
* argp/argp-test.c (popt): Increase size of buf to 12.
* benchtests/bench-malloc-simple.c (bench):
Do not initialize tests array out of bounds.
This patch significantly improves performance of memmem using a novel
modified Horspool algorithm. Needles up to size 256 use a bad-character
table indexed by hashed pairs of characters to quickly skip past mismatches.
Long needles use a self-adapting filtering step to avoid comparing the whole
needle repeatedly.
By limiting the needle length to 256, the shift table only requires 8 bits
per entry, lowering preprocessing overhead and minimizing cache effects.
This limit also implies worst-case performance is linear.
Small needles up to size 2 use a dedicated linear search. Very long needles
use the Two-Way algorithm (to avoid increasing stack size or slowing down
the common case, inlining is disabled).
The performance gain is 6.6 times on English text on AArch64 using random
needles with average size 8.
Tested against GLIBC testsuite and randomized tests.
Reviewed-by: Szabolcs Nagy <szabolcs.nagy@arm.com>
* string/memmem.c (__memmem): Rewrite to improve performance.
This patch significantly improves performance of strstr using a novel
modified Horspool algorithm. Needles up to size 256 use a bad-character
table indexed by hashed pairs of characters to quickly skip past mismatches.
Long needles use a self-adapting filtering step to avoid comparing the whole
needle repeatedly.
By limiting the needle length to 256, the shift table only requires 8 bits
per entry, lowering preprocessing overhead and minimizing cache effects.
This limit also implies worst-case performance is linear.
Small needles up to size 3 use a dedicated linear search. Very long needles
use the Two-Way algorithm.
The performance gain using the improved bench-strstr on Cortex-A72 is 5.8
times basic_strstr and 3.7 times twoway_strstr.
Tested against GLIBC testsuite, randomized tests and the GNULIB strstr test
(https://git.savannah.gnu.org/cgit/gnulib.git/tree/tests/test-strstr.c).
Reviewed-by: Szabolcs Nagy <szabolcs.nagy@arm.com>
* string/str-two-way.h (two_way_short_needle): Add inline to avoid
warning.
(two_way_long_needle): Block inlining.
* string/strstr.c (strstr2): Add new function.
(strstr3): Likewise.
(STRSTR): Completely rewrite strstr to improve performance.
Commit 1294b1892e ("Add support for sqrt asm redirects") added the
-fno-math-errno flag to build most of the glibc in order to enable GCC
to inline math functions. Due to GCC bug #88576, saving and restoring
errno around calls to malloc are optimized-out. In turn this causes
strerror to set errno to ENOMEM if it get passed an invalid error number
and if malloc sets errno to ENOMEM (which might happen even if it
succeeds). This is not allowed by POSIX.
This patch changes the build flags, building only libm with
-fno-math-errno and all the remaining code with -fno-math-errno. This
should be safe as libm doesn't contain any code saving and restoring
errno around malloc. This patch can probably be reverted once the GCC
bug is fixed and available in stable releases.
Tested on x86-64, no regression in the testsuite.
Changelog:
[BZ #24024]
* Makeconfig: Build libm with -fno-math-errno but build the remaining
code with -fmath-errno.
* string/Makefile [$(build-shared)] (tests): Add test-strerror-errno.
[$(build-shared)] (LDLIBS-test-strerror-errno): New variable.
* string/test-strerror-errno.c: New file.
The generic strstr in GLIBC 2.28 fails to match huge needles. The optimized
AVAILABLE macro reads ahead a large fixed amount to reduce the overhead of
repeatedly checking for the end of the string. However if the needle length
is larger than this, two_way_long_needle may confuse this as meaning the end
of the string and return NULL. This is fixed by adding the needle length to
the amount to read ahead.
[BZ #23637]
* string/test-strstr.c (pr23637): New function.
(test_main): Add tests with longer needles.
* string/strcasestr.c (AVAILABLE): Fix readahead distance.
* string/strstr.c (AVAILABLE): Likewise.
As done in commit 284f42bc77, memcmp
can be used after memchr to avoid the initialization overhead of the
two-way algorithm for the first match. This has shown improvement
>40% for first match.
Looking at the benchtests, both strstr and strcasestr spend a lot of time
in a slow initialization loop handling one character per iteration.
This can be simplified and use the much faster strlen/strnlen/strchr/memcmp.
Read ahead a few cachelines to reduce the number of strnlen calls, which
improves performance by ~3-4%. This patch improves the time taken for the
full strstr benchtest by >40%.
* string/strcasestr.c (STRCASESTR): Simplify and speedup first match.
* string/strstr.c (AVAILABLE): Likewise.
On s390x, the test string/tst-xbzero-opt is failing if build with gcc head:
FAIL: no clear/prepare: expected 32 got 0
FAIL: no clear/test: expected some got 0
FAIL: ordinary clear/prepare: expected 32 got 0
INFO: ordinary clear/test: found 0 patterns (memset not eliminated)
PASS: explicit clear/prepare: expected 32 got 32
PASS: explicit clear/test: expected 0 got 0
In setup_no_clear / setup_ordinary_clear, GCC is omitting the memcpy loop
in prepare_test_buffer. Thus count_test_patterns does not find any of the
test_pattern.
This patch calls use_test_buffer in order to force the compiler to really copy
the pattern to buf.
ChangeLog:
* string/tst-xbzero-opt.c (use_test_buffer): New function.
(prepare_test_buffer): Call use_test_buffer as compiler barrier.
Add <bits/indirect-return.h> and include it in <ucontext.h>.
__INDIRECT_RETURN defined in <bits/indirect-return.h> indicates if
swapcontext requires special compiler treatment. The default
__INDIRECT_RETURN is empty.
On x86, when shadow stack is enabled, __INDIRECT_RETURN is defined
with indirect_return attribute, which has been added to GCC 9, to
indicate that swapcontext returns via indirect branch. Otherwise
__INDIRECT_RETURN is defined with returns_twice attribute.
When shadow stack is enabled, remove always_inline attribute from
prepare_test_buffer in string/tst-xbzero-opt.c to avoid:
tst-xbzero-opt.c: In function ‘prepare_test_buffer’:
tst-xbzero-opt.c:105:1: error: function ‘prepare_test_buffer’ can never be inlined because it uses setjmp
prepare_test_buffer (unsigned char *buf)
when indirect_return attribute isn't available.
Reviewed-by: Carlos O'Donell <carlos@redhat.com>
* bits/indirect-return.h: New file.
* misc/sys/cdefs.h (__glibc_has_attribute): New.
* sysdeps/x86/bits/indirect-return.h: Likewise.
* stdlib/Makefile (headers): Add bits/indirect-return.h.
* stdlib/ucontext.h: Include <bits/indirect-return.h>.
(swapcontext): Add __INDIRECT_RETURN.
* string/tst-xbzero-opt.c (ALWAYS_INLINE): New.
(prepare_test_buffer): Use it.
Improve strstr performance. Strstr tends to be slow because it uses
many calls to memchr and a slow byte loop to scan for the next match.
Performance is significantly improved by using strnlen on larger blocks
and using strchr to search for the next matching character. strcasestr
can also use strnlen to scan ahead, and memmem can use memchr to check
for the next match.
On the GLIBC bench tests the performance gains on Cortex-A72 are:
strstr: +25%
strcasestr: +4.3%
memmem: +18%
On a 256KB dataset strstr performance improves by 67%, strcasestr by 47%.
Reviewd-by: Adhemerval Zanella <adhemerval.zanella@linaro.org>
