glibc/glibc-upstream-2.34-220.patch
Arjun Shankar 601650f878 Import glibc-2.34-35.fc35 from f35
* Tue May 31 2022 Arjun Shankar <arjun@redhat.com> - 2.34-35
- Sync with upstream branch release/2.34/master,
  commit ff450cdbdee0b8cb6b9d653d6d2fa892de29be31:
- Fix deadlock when pthread_atfork handler calls pthread_atfork or dlclose
- x86: Fallback {str|wcs}cmp RTM in the ncmp overflow case [BZ #29127]
- string.h: fix __fortified_attr_access macro call [BZ #29162]
- linux: Add a getauxval test [BZ #23293]
- rtld: Use generic argv adjustment in ld.so [BZ #23293]
- S390: Enable static PIE

* Thu May 19 2022 Florian Weimer <fweimer@redhat.com> - 2.34-34
- Sync with upstream branch release/2.34/master,
  commit ede8d94d154157d269b18f3601440ac576c1f96a:
- csu: Implement and use _dl_early_allocate during static startup
- Linux: Introduce __brk_call for invoking the brk system call
- Linux: Implement a useful version of _startup_fatal
- ia64: Always define IA64_USE_NEW_STUB as a flag macro
- Linux: Define MMAP_CALL_INTERNAL
- i386: Honor I386_USE_SYSENTER for 6-argument Linux system calls
- i386: Remove OPTIMIZE_FOR_GCC_5 from Linux libc-do-syscall.S
- elf: Remove __libc_init_secure
- Linux: Consolidate auxiliary vector parsing (redo)
- Linux: Include <dl-auxv.h> in dl-sysdep.c only for SHARED
- Revert "Linux: Consolidate auxiliary vector parsing"
- Linux: Consolidate auxiliary vector parsing
- Linux: Assume that NEED_DL_SYSINFO_DSO is always defined
- Linux: Remove DL_FIND_ARG_COMPONENTS
- Linux: Remove HAVE_AUX_SECURE, HAVE_AUX_XID, HAVE_AUX_PAGESIZE
- elf: Merge dl-sysdep.c into the Linux version
- elf: Remove unused NEED_DL_BASE_ADDR and _dl_base_addr
- x86: Optimize {str|wcs}rchr-evex
- x86: Optimize {str|wcs}rchr-avx2
- x86: Optimize {str|wcs}rchr-sse2
- x86: Cleanup page cross code in memcmp-avx2-movbe.S
- x86: Remove memcmp-sse4.S
- x86: Small improvements for wcslen
- x86: Remove AVX str{n}casecmp
- x86: Add EVEX optimized str{n}casecmp
- x86: Add AVX2 optimized str{n}casecmp
- x86: Optimize str{n}casecmp TOLOWER logic in strcmp-sse42.S
- x86: Optimize str{n}casecmp TOLOWER logic in strcmp.S
- x86: Remove strspn-sse2.S and use the generic implementation
- x86: Remove strpbrk-sse2.S and use the generic implementation
- x87: Remove strcspn-sse2.S and use the generic implementation
- x86: Optimize strspn in strspn-c.c
- x86: Optimize strcspn and strpbrk in strcspn-c.c
- x86: Code cleanup in strchr-evex and comment justifying branch
- x86: Code cleanup in strchr-avx2 and comment justifying branch
- x86_64: Remove bcopy optimizations
- x86-64: Remove bzero weak alias in SS2 memset
- x86_64/multiarch: Sort sysdep_routines and put one entry per line
- x86: Improve L to support L(XXX_SYMBOL (YYY, ZZZ))
- fortify: Ensure that __glibc_fortify condition is a constant [BZ #29141]

* Thu May 12 2022 Florian Weimer <fweimer@redhat.com> - 2.34-33
- Sync with upstream branch release/2.34/master,
  commit 91c2e6c3db44297bf4cb3a2e3c40236c5b6a0b23:
- dlfcn: Implement the RTLD_DI_PHDR request type for dlinfo
- manual: Document the dlinfo function
- x86: Fix fallback for wcsncmp_avx2 in strcmp-avx2.S [BZ #28896]
- x86: Fix bug in strncmp-evex and strncmp-avx2 [BZ #28895]
- x86: Set .text section in memset-vec-unaligned-erms
- x86-64: Optimize bzero
- x86: Remove SSSE3 instruction for broadcast in memset.S (SSE2 Only)
- x86: Improve vec generation in memset-vec-unaligned-erms.S
- x86-64: Fix strcmp-evex.S
- x86-64: Fix strcmp-avx2.S
- x86: Optimize strcmp-evex.S
- x86: Optimize strcmp-avx2.S
- manual: Clarify that abbreviations of long options are allowed
- Add HWCAP2_AFP, HWCAP2_RPRES from Linux 5.17 to AArch64 bits/hwcap.h
- aarch64: Add HWCAP2_ECV from Linux 5.16
- Add SOL_MPTCP, SOL_MCTP from Linux 5.16 to bits/socket.h
- Update kernel version to 5.17 in tst-mman-consts.py
- Update kernel version to 5.16 in tst-mman-consts.py
- Update syscall lists for Linux 5.17
- Add ARPHRD_CAN, ARPHRD_MCTP to net/if_arp.h
- Update kernel version to 5.15 in tst-mman-consts.py
- Add PF_MCTP, AF_MCTP from Linux 5.15 to bits/socket.h

Resolves: #2091541
2022-06-06 16:33:33 +02:00

144 lines
4.6 KiB
Diff

commit 0ae1006967eef11909fbed0f6ecef2f260b133d3
Author: Noah Goldstein <goldstein.w.n@gmail.com>
Date: Wed Mar 23 16:57:22 2022 -0500
x86: Optimize strcspn and strpbrk in strcspn-c.c
Use _mm_cmpeq_epi8 and _mm_movemask_epi8 to get strlen instead of
_mm_cmpistri. Also change offset to unsigned to avoid unnecessary
sign extensions.
geometric_mean(N=20) of all benchmarks that dont fallback on
sse2/strlen; New / Original: .928
All string/memory tests pass.
Reviewed-by: H.J. Lu <hjl.tools@gmail.com>
(cherry picked from commit 30d627d477d7255345a4b713cf352ac32d644d61)
diff --git a/sysdeps/x86_64/multiarch/strcspn-c.c b/sysdeps/x86_64/multiarch/strcspn-c.c
index c56ddbd22f014653..2436b6dcd90d8efe 100644
--- a/sysdeps/x86_64/multiarch/strcspn-c.c
+++ b/sysdeps/x86_64/multiarch/strcspn-c.c
@@ -85,83 +85,74 @@ STRCSPN_SSE42 (const char *s, const char *a)
RETURN (NULL, strlen (s));
const char *aligned;
- __m128i mask;
- int offset = (int) ((size_t) a & 15);
+ __m128i mask, maskz, zero;
+ unsigned int maskz_bits;
+ unsigned int offset = (unsigned int) ((size_t) a & 15);
+ zero = _mm_set1_epi8 (0);
if (offset != 0)
{
/* Load masks. */
aligned = (const char *) ((size_t) a & -16L);
__m128i mask0 = _mm_load_si128 ((__m128i *) aligned);
-
- mask = __m128i_shift_right (mask0, offset);
+ maskz = _mm_cmpeq_epi8 (mask0, zero);
/* Find where the NULL terminator is. */
- int length = _mm_cmpistri (mask, mask, 0x3a);
- if (length == 16 - offset)
- {
- /* There is no NULL terminator. */
- __m128i mask1 = _mm_load_si128 ((__m128i *) (aligned + 16));
- int index = _mm_cmpistri (mask1, mask1, 0x3a);
- length += index;
-
- /* Don't use SSE4.2 if the length of A > 16. */
- if (length > 16)
- return STRCSPN_SSE2 (s, a);
-
- if (index != 0)
- {
- /* Combine mask0 and mask1. We could play games with
- palignr, but frankly this data should be in L1 now
- so do the merge via an unaligned load. */
- mask = _mm_loadu_si128 ((__m128i *) a);
- }
- }
+ maskz_bits = _mm_movemask_epi8 (maskz) >> offset;
+ if (maskz_bits != 0)
+ {
+ mask = __m128i_shift_right (mask0, offset);
+ offset = (unsigned int) ((size_t) s & 15);
+ if (offset)
+ goto start_unaligned;
+
+ aligned = s;
+ goto start_loop;
+ }
}
- else
- {
- /* A is aligned. */
- mask = _mm_load_si128 ((__m128i *) a);
- /* Find where the NULL terminator is. */
- int length = _mm_cmpistri (mask, mask, 0x3a);
- if (length == 16)
- {
- /* There is no NULL terminator. Don't use SSE4.2 if the length
- of A > 16. */
- if (a[16] != 0)
- return STRCSPN_SSE2 (s, a);
- }
+ /* A is aligned. */
+ mask = _mm_loadu_si128 ((__m128i *) a);
+ /* Find where the NULL terminator is. */
+ maskz = _mm_cmpeq_epi8 (mask, zero);
+ maskz_bits = _mm_movemask_epi8 (maskz);
+ if (maskz_bits == 0)
+ {
+ /* There is no NULL terminator. Don't use SSE4.2 if the length
+ of A > 16. */
+ if (a[16] != 0)
+ return STRCSPN_SSE2 (s, a);
}
- offset = (int) ((size_t) s & 15);
+ aligned = s;
+ offset = (unsigned int) ((size_t) s & 15);
if (offset != 0)
{
+ start_unaligned:
/* Check partial string. */
aligned = (const char *) ((size_t) s & -16L);
__m128i value = _mm_load_si128 ((__m128i *) aligned);
value = __m128i_shift_right (value, offset);
- int length = _mm_cmpistri (mask, value, 0x2);
+ unsigned int length = _mm_cmpistri (mask, value, 0x2);
/* No need to check ZFlag since ZFlag is always 1. */
- int cflag = _mm_cmpistrc (mask, value, 0x2);
+ unsigned int cflag = _mm_cmpistrc (mask, value, 0x2);
if (cflag)
RETURN ((char *) (s + length), length);
/* Find where the NULL terminator is. */
- int index = _mm_cmpistri (value, value, 0x3a);
+ unsigned int index = _mm_cmpistri (value, value, 0x3a);
if (index < 16 - offset)
RETURN (NULL, index);
aligned += 16;
}
- else
- aligned = s;
+start_loop:
while (1)
{
__m128i value = _mm_load_si128 ((__m128i *) aligned);
- int index = _mm_cmpistri (mask, value, 0x2);
- int cflag = _mm_cmpistrc (mask, value, 0x2);
- int zflag = _mm_cmpistrz (mask, value, 0x2);
+ unsigned int index = _mm_cmpistri (mask, value, 0x2);
+ unsigned int cflag = _mm_cmpistrc (mask, value, 0x2);
+ unsigned int zflag = _mm_cmpistrz (mask, value, 0x2);
if (cflag)
RETURN ((char *) (aligned + index), (size_t) (aligned + index - s));
if (zflag)