AuroraMiddleware/glibc
1
0
Fork 0
mirror of https://sourceware.org/git/glibc.git synced 2024-11-09 23:00:07 +00:00
Code Issues Packages Projects Releases Wiki Activity

Optimized strnlen and wcscmp for x86-64

Browse Source
This commit is contained in:
Liubov Dmitrieva 2011-10-23 14:56:04 -04:00 committed by Ulrich Drepper
parent 6ddf460fa5
commit ce7dd29f28
10 changed files with 1003 additions and 319 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

14
ChangeLog
View File

@ -1,3 +1,17 @@
2011-10-20 Liubov Dmitrieva <liubov.dmitrieva@gmail.com>
* sysdeps/x86_64/multiarch/Makefile (sysdep_routines): Add
strnlen-sse2-no-bsf.
Rename strlen-no-bsf to strlen-sse2-no-bsf.
* sysdeps/x86_64/multiarch/strlen-no-bsf.S: Rename to
* sysdeps/x86_64/multiarch/strlen-sse2-no-bsf.S:
Add strnlen support.
(USE_AS_STRNLEN): New macro.
* sysdeps/x86_64/multiarch/strnlen-sse2-no-bsf.S: New file.
* sysdeps/x86_64/multiarch/strcat-ssse3.S: Update.
Rename strlen-no-bsf.S to strlen-sse2-no-bsf.S
* sysdeps/x86_64/wcslen.S: New file.
2011-10-20 Michael Zolotukhin <michael.v.zolotukhin@gmail.com>
* sysdeps/i386/i686/multiarch/memcpy-ssse3.S: Update.

4
NEWS
View File

@ -26,8 +26,8 @@ Version 2.15
* Improved strcpy, strncpy, stpcpy, stpncpy for SSE2 and SSSE3 on x86-64.
Contributed by HJ Lu.
* Optimized strcat and strncat on x86-64 and optimized wcscmp on x86-32 and
x86-64.
* Optimized strcat, strncat, wcslen, strnlen on x86-64 and optimized
wcscmp on x86-32 and x86-64.
Contributed by Liubov Dmitrieva.
* Optimized strchr and strrchr for SSE on x86-32.

4
sysdeps/x86_64/multiarch/Makefile
View File

@ -9,13 +9,13 @@ sysdep_routines += strncat-c stpncpy-c strncpy-c strcmp-ssse3 strncmp-ssse3 \
strend-sse4 memcmp-sse4 memcpy-ssse3 mempcpy-ssse3 \
memmove-ssse3 memcpy-ssse3-back mempcpy-ssse3-back \
memmove-ssse3-back strcasestr-nonascii strcasecmp_l-ssse3 \
strncase_l-ssse3 strlen-sse4 strlen-no-bsf memset-x86-64 \
strncase_l-ssse3 strlen-sse4 strlen-sse2-no-bsf memset-x86-64 \
strcpy-ssse3 strncpy-ssse3 stpcpy-ssse3 stpncpy-ssse3 \
strcpy-sse2-unaligned strncpy-sse2-unaligned \
stpcpy-sse2-unaligned stpncpy-sse2-unaligned \
strcat-sse2-unaligned strncat-sse2-unaligned \
strcat-ssse3 strncat-ssse3 strlen-sse2-pminub \
strrchr-sse2-no-bsf strchr-sse2-no-bsf \
strnlen-sse2-no-bsf strrchr-sse2-no-bsf strchr-sse2-no-bsf \
memcmp-ssse3 wmemcmp-sse4 wmemcmp-ssse3 wmemcmp-c
ifeq (yes,$(config-cflags-sse4))
sysdep_routines += strcspn-c strpbrk-c strspn-c strstr-c strcasestr-c varshift

2
sysdeps/x86_64/multiarch/strcat-ssse3.S
View File

@ -35,7 +35,7 @@ ENTRY (STRCAT)
# endif
# define RETURN jmp L(StartStrcpyPart)
# include "strlen-no-bsf.S"
# include "strlen-sse2-no-bsf.S"
# undef RETURN

313
sysdeps/x86_64/multiarch/strlen-no-bsf.S
View File

@ -1,313 +0,0 @@
/* strlen SSE2 without bsf
Copyright (C) 2010, 2011 Free Software Foundation, Inc.
Contributed by Intel Corporation.
This file is part of the GNU C Library.
The GNU C Library is free software; you can redistribute it and/or
modify it under the terms of the GNU Lesser General Public
License as published by the Free Software Foundation; either
version 2.1 of the License, or (at your option) any later version.
The GNU C Library is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
Lesser General Public License for more details.
You should have received a copy of the GNU Lesser General Public
License along with the GNU C Library; if not, write to the Free
Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
02111-1307 USA. */
#if (defined SHARED || defined USE_AS_STRCAT) && !defined NOT_IN_libc
# ifndef USE_AS_STRCAT
# include <sysdep.h>
# define RETURN ret
atom_text_section
ENTRY (__strlen_no_bsf)
# endif
xor %eax, %eax
cmpb $0, (%rdi)
jz L(exit_tail0)
cmpb $0, 1(%rdi)
jz L(exit_tail1)
cmpb $0, 2(%rdi)
jz L(exit_tail2)
cmpb $0, 3(%rdi)
jz L(exit_tail3)
cmpb $0, 4(%rdi)
jz L(exit_tail4)
cmpb $0, 5(%rdi)
jz L(exit_tail5)
cmpb $0, 6(%rdi)
jz L(exit_tail6)
cmpb $0, 7(%rdi)
jz L(exit_tail7)
cmpb $0, 8(%rdi)
jz L(exit_tail8)
cmpb $0, 9(%rdi)
jz L(exit_tail9)
cmpb $0, 10(%rdi)
jz L(exit_tail10)
cmpb $0, 11(%rdi)
jz L(exit_tail11)
cmpb $0, 12(%rdi)
jz L(exit_tail12)
cmpb $0, 13(%rdi)
jz L(exit_tail13)
cmpb $0, 14(%rdi)
jz L(exit_tail14)
cmpb $0, 15(%rdi)
jz L(exit_tail15)
pxor %xmm0, %xmm0
mov %rdi, %rcx
mov %rdi, %rax
and $-16, %rax
add $16, %rax
add $16, %rcx
pcmpeqb (%rax), %xmm0
pmovmskb %xmm0, %edx
pxor %xmm1, %xmm1
test %edx, %edx
lea 16(%rax), %rax
jnz L(exit)
pcmpeqb (%rax), %xmm1
pmovmskb %xmm1, %edx
pxor %xmm2, %xmm2
test %edx, %edx
lea 16(%rax), %rax
jnz L(exit)
pcmpeqb (%rax), %xmm2
pmovmskb %xmm2, %edx
pxor %xmm3, %xmm3
test %edx, %edx
lea 16(%rax), %rax
jnz L(exit)
pcmpeqb (%rax), %xmm3
pmovmskb %xmm3, %edx
test %edx, %edx
lea 16(%rax), %rax
jnz L(exit)
pcmpeqb (%rax), %xmm0
pmovmskb %xmm0, %edx
test %edx, %edx
lea 16(%rax), %rax
jnz L(exit)
pcmpeqb (%rax), %xmm1
pmovmskb %xmm1, %edx
test %edx, %edx
lea 16(%rax), %rax
jnz L(exit)
pcmpeqb (%rax), %xmm2
pmovmskb %xmm2, %edx
test %edx, %edx
lea 16(%rax), %rax
jnz L(exit)
pcmpeqb (%rax), %xmm3
pmovmskb %xmm3, %edx
test %edx, %edx
lea 16(%rax), %rax
jnz L(exit)
pcmpeqb (%rax), %xmm0
pmovmskb %xmm0, %edx
test %edx, %edx
lea 16(%rax), %rax
jnz L(exit)
pcmpeqb (%rax), %xmm1
pmovmskb %xmm1, %edx
test %edx, %edx
lea 16(%rax), %rax
jnz L(exit)
pcmpeqb (%rax), %xmm2
pmovmskb %xmm2, %edx
test %edx, %edx
lea 16(%rax), %rax
jnz L(exit)
pcmpeqb (%rax), %xmm3
pmovmskb %xmm3, %edx
test %edx, %edx
lea 16(%rax), %rax
jnz L(exit)
pcmpeqb (%rax), %xmm0
pmovmskb %xmm0, %edx
test %edx, %edx
lea 16(%rax), %rax
jnz L(exit)
pcmpeqb (%rax), %xmm1
pmovmskb %xmm1, %edx
test %edx, %edx
lea 16(%rax), %rax
jnz L(exit)
pcmpeqb (%rax), %xmm2
pmovmskb %xmm2, %edx
test %edx, %edx
lea 16(%rax), %rax
jnz L(exit)
pcmpeqb (%rax), %xmm3
pmovmskb %xmm3, %edx
test %edx, %edx
lea 16(%rax), %rax
jnz L(exit)
and $-0x40, %rax
L(aligned_64):
pcmpeqb (%rax), %xmm0
pcmpeqb 16(%rax), %xmm1
pcmpeqb 32(%rax), %xmm2
pcmpeqb 48(%rax), %xmm3
pmovmskb %xmm0, %edx
pmovmskb %xmm1, %r11d
pmovmskb %xmm2, %r10d
pmovmskb %xmm3, %r9d
or %edx, %r9d
or %r11d, %r9d
or %r10d, %r9d
lea 64(%rax), %rax
jz L(aligned_64)
test %edx, %edx
jnz L(aligned_64_exit_16)
test %r11d, %r11d
jnz L(aligned_64_exit_32)
test %r10d, %r10d
jnz L(aligned_64_exit_48)
L(aligned_64_exit_64):
pmovmskb %xmm3, %edx
jmp L(aligned_64_exit)
L(aligned_64_exit_48):
lea -16(%rax), %rax
mov %r10d, %edx
jmp L(aligned_64_exit)
L(aligned_64_exit_32):
lea -32(%rax), %rax
mov %r11d, %edx
jmp L(aligned_64_exit)
L(aligned_64_exit_16):
lea -48(%rax), %rax
L(aligned_64_exit):
L(exit):
sub %rcx, %rax
test %dl, %dl
jz L(exit_high)
test $0x01, %dl
jnz L(exit_tail0)
test $0x02, %dl
jnz L(exit_tail1)
test $0x04, %dl
jnz L(exit_tail2)
test $0x08, %dl
jnz L(exit_tail3)
test $0x10, %dl
jnz L(exit_tail4)
test $0x20, %dl
jnz L(exit_tail5)
test $0x40, %dl
jnz L(exit_tail6)
add $7, %eax
L(exit_tail0):
RETURN
L(exit_high):
add $8, %eax
test $0x01, %dh
jnz L(exit_tail0)
test $0x02, %dh
jnz L(exit_tail1)
test $0x04, %dh
jnz L(exit_tail2)
test $0x08, %dh
jnz L(exit_tail3)
test $0x10, %dh
jnz L(exit_tail4)
test $0x20, %dh
jnz L(exit_tail5)
test $0x40, %dh
jnz L(exit_tail6)
add $7, %eax
RETURN
.p2align 4
L(exit_tail1):
add $1, %eax
RETURN
L(exit_tail2):
add $2, %eax
RETURN
L(exit_tail3):
add $3, %eax
RETURN
L(exit_tail4):
add $4, %eax
RETURN
L(exit_tail5):
add $5, %eax
RETURN
L(exit_tail6):
add $6, %eax
RETURN
L(exit_tail7):
add $7, %eax
RETURN
L(exit_tail8):
add $8, %eax
RETURN
L(exit_tail9):
add $9, %eax
RETURN
L(exit_tail10):
add $10, %eax
RETURN
L(exit_tail11):
add $11, %eax
RETURN
L(exit_tail12):
add $12, %eax
RETURN
L(exit_tail13):
add $13, %eax
RETURN
L(exit_tail14):
add $14, %eax
RETURN
L(exit_tail15):
add $15, %eax
# ifndef USE_AS_STRCAT
RETURN
END (__strlen_no_bsf)
# endif
#endif

686
sysdeps/x86_64/multiarch/strlen-sse2-no-bsf.S Normal file
View File

@ -0,0 +1,686 @@
/* strlen SSE2 without bsf
Copyright (C) 2010, 2011 Free Software Foundation, Inc.
Contributed by Intel Corporation.
This file is part of the GNU C Library.
The GNU C Library is free software; you can redistribute it and/or
modify it under the terms of the GNU Lesser General Public
License as published by the Free Software Foundation; either
version 2.1 of the License, or (at your option) any later version.
The GNU C Library is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
Lesser General Public License for more details.
You should have received a copy of the GNU Lesser General Public
License along with the GNU C Library; if not, write to the Free
Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
02111-1307 USA. */
/* only for strlen case we don't use optimized version for STATIC build just for SHARED */
#if (defined SHARED || defined USE_AS_STRCAT || defined USE_AS_STRNLEN) && !defined NOT_IN_libc
# ifndef USE_AS_STRCAT
# include <sysdep.h>
# define RETURN ret
# ifndef STRLEN
# define STRLEN __strlen_sse2_no_bsf
# endif
atom_text_section
ENTRY (STRLEN)
# endif
xor %eax, %eax
# ifdef USE_AS_STRNLEN
mov %rsi, %r8
sub $4, %rsi
jbe L(len_less4_prolog)
# endif
cmpb $0, (%rdi)
jz L(exit_tail0)
cmpb $0, 1(%rdi)
jz L(exit_tail1)
cmpb $0, 2(%rdi)
jz L(exit_tail2)
cmpb $0, 3(%rdi)
jz L(exit_tail3)
# ifdef USE_AS_STRNLEN
sub $4, %rsi
jbe L(len_less8_prolog)
# endif
cmpb $0, 4(%rdi)
jz L(exit_tail4)
cmpb $0, 5(%rdi)
jz L(exit_tail5)
cmpb $0, 6(%rdi)
jz L(exit_tail6)
cmpb $0, 7(%rdi)
jz L(exit_tail7)
# ifdef USE_AS_STRNLEN
sub $4, %rsi
jbe L(len_less12_prolog)
# endif
cmpb $0, 8(%rdi)
jz L(exit_tail8)
cmpb $0, 9(%rdi)
jz L(exit_tail9)
cmpb $0, 10(%rdi)
jz L(exit_tail10)
cmpb $0, 11(%rdi)
jz L(exit_tail11)
# ifdef USE_AS_STRNLEN
sub $4, %rsi
jbe L(len_less16_prolog)
# endif
cmpb $0, 12(%rdi)
jz L(exit_tail12)
cmpb $0, 13(%rdi)
jz L(exit_tail13)
cmpb $0, 14(%rdi)
jz L(exit_tail14)
cmpb $0, 15(%rdi)
jz L(exit_tail15)
pxor %xmm0, %xmm0
lea 16(%rdi), %rcx
lea 16(%rdi), %rax
and $-16, %rax
# ifdef USE_AS_STRNLEN
and $15, %rdi
add %rdi, %rsi
sub $64, %rsi
jbe L(len_less64)
# endif
pcmpeqb (%rax), %xmm0
pmovmskb %xmm0, %edx
pxor %xmm1, %xmm1
test %edx, %edx
lea 16(%rax), %rax
jnz L(exit)
pcmpeqb (%rax), %xmm1
pmovmskb %xmm1, %edx
pxor %xmm2, %xmm2
test %edx, %edx
lea 16(%rax), %rax
jnz L(exit)
pcmpeqb (%rax), %xmm2
pmovmskb %xmm2, %edx
pxor %xmm3, %xmm3
test %edx, %edx
lea 16(%rax), %rax
jnz L(exit)
pcmpeqb (%rax), %xmm3
pmovmskb %xmm3, %edx
test %edx, %edx
lea 16(%rax), %rax
jnz L(exit)
# ifdef USE_AS_STRNLEN
sub $64, %rsi
jbe L(len_less64)
# endif
pcmpeqb (%rax), %xmm0
pmovmskb %xmm0, %edx
test %edx, %edx
lea 16(%rax), %rax
jnz L(exit)
pcmpeqb (%rax), %xmm1
pmovmskb %xmm1, %edx
test %edx, %edx
lea 16(%rax), %rax
jnz L(exit)
pcmpeqb (%rax), %xmm2
pmovmskb %xmm2, %edx
test %edx, %edx
lea 16(%rax), %rax
jnz L(exit)
pcmpeqb (%rax), %xmm3
pmovmskb %xmm3, %edx
test %edx, %edx
lea 16(%rax), %rax
jnz L(exit)
# ifdef USE_AS_STRNLEN
sub $64, %rsi
jbe L(len_less64)
# endif
pcmpeqb (%rax), %xmm0
pmovmskb %xmm0, %edx
test %edx, %edx
lea 16(%rax), %rax
jnz L(exit)
pcmpeqb (%rax), %xmm1
pmovmskb %xmm1, %edx
test %edx, %edx
lea 16(%rax), %rax
jnz L(exit)
pcmpeqb (%rax), %xmm2
pmovmskb %xmm2, %edx
test %edx, %edx
lea 16(%rax), %rax
jnz L(exit)
pcmpeqb (%rax), %xmm3
pmovmskb %xmm3, %edx
test %edx, %edx
lea 16(%rax), %rax
jnz L(exit)
# ifdef USE_AS_STRNLEN
sub $64, %rsi
jbe L(len_less64)
# endif
pcmpeqb (%rax), %xmm0
pmovmskb %xmm0, %edx
test %edx, %edx
lea 16(%rax), %rax
jnz L(exit)
pcmpeqb (%rax), %xmm1
pmovmskb %xmm1, %edx
test %edx, %edx
lea 16(%rax), %rax
jnz L(exit)
pcmpeqb (%rax), %xmm2
pmovmskb %xmm2, %edx
test %edx, %edx
lea 16(%rax), %rax
jnz L(exit)
pcmpeqb (%rax), %xmm3
pmovmskb %xmm3, %edx
test %edx, %edx
lea 16(%rax), %rax
jnz L(exit)
# ifdef USE_AS_STRNLEN
mov %rax, %rdx
and $63, %rdx
add %rdx, %rsi
# endif
and $-0x40, %rax
.p2align 4
L(aligned_64):
# ifdef USE_AS_STRNLEN
sub $64, %rsi
jbe L(len_less64)
# endif
pcmpeqb (%rax), %xmm0
pcmpeqb 16(%rax), %xmm1
pcmpeqb 32(%rax), %xmm2
pcmpeqb 48(%rax), %xmm3
pmovmskb %xmm0, %edx
pmovmskb %xmm1, %r11d
pmovmskb %xmm2, %r10d
pmovmskb %xmm3, %r9d
or %edx, %r9d
or %r11d, %r9d
or %r10d, %r9d
lea 64(%rax), %rax
jz L(aligned_64)
test %edx, %edx
jnz L(aligned_64_exit_16)
test %r11d, %r11d
jnz L(aligned_64_exit_32)
test %r10d, %r10d
jnz L(aligned_64_exit_48)
L(aligned_64_exit_64):
pmovmskb %xmm3, %edx
jmp L(aligned_64_exit)
L(aligned_64_exit_48):
lea -16(%rax), %rax
mov %r10d, %edx
jmp L(aligned_64_exit)
L(aligned_64_exit_32):
lea -32(%rax), %rax
mov %r11d, %edx
jmp L(aligned_64_exit)
L(aligned_64_exit_16):
lea -48(%rax), %rax
L(aligned_64_exit):
L(exit):
sub %rcx, %rax
test %dl, %dl
jz L(exit_high)
test $0x01, %dl
jnz L(exit_tail0)
test $0x02, %dl
jnz L(exit_tail1)
test $0x04, %dl
jnz L(exit_tail2)
test $0x08, %dl
jnz L(exit_tail3)
test $0x10, %dl
jnz L(exit_tail4)
test $0x20, %dl
jnz L(exit_tail5)
test $0x40, %dl
jnz L(exit_tail6)
add $7, %eax
L(exit_tail0):
RETURN
L(exit_high):
add $8, %eax
test $0x01, %dh
jnz L(exit_tail0)
test $0x02, %dh
jnz L(exit_tail1)
test $0x04, %dh
jnz L(exit_tail2)
test $0x08, %dh
jnz L(exit_tail3)
test $0x10, %dh
jnz L(exit_tail4)
test $0x20, %dh
jnz L(exit_tail5)
test $0x40, %dh
jnz L(exit_tail6)
add $7, %eax
RETURN
# ifdef USE_AS_STRNLEN
.p2align 4
L(len_less64):
pxor %xmm0, %xmm0
add $64, %rsi
pcmpeqb (%rax), %xmm0
pmovmskb %xmm0, %edx
pxor %xmm1, %xmm1
lea 16(%rax), %rax
test %edx, %edx
jnz L(strnlen_exit)
sub $16, %rsi
jbe L(return_start_len)
pcmpeqb (%rax), %xmm1
pmovmskb %xmm1, %edx
lea 16(%rax), %rax
test %edx, %edx
jnz L(strnlen_exit)
sub $16, %rsi
jbe L(return_start_len)
pcmpeqb (%rax), %xmm0
pmovmskb %xmm0, %edx
lea 16(%rax), %rax
test %edx, %edx
jnz L(strnlen_exit)
sub $16, %rsi
jbe L(return_start_len)
pcmpeqb (%rax), %xmm1
pmovmskb %xmm1, %edx
lea 16(%rax), %rax
test %edx, %edx
jnz L(strnlen_exit)
mov %r8, %rax
ret
.p2align 4
L(strnlen_exit):
sub %rcx, %rax
test %dl, %dl
jz L(strnlen_exit_high)
mov %dl, %cl
and $15, %cl
jz L(strnlen_exit_8)
test $0x01, %dl
jnz L(exit_tail0)
test $0x02, %dl
jnz L(strnlen_exit_tail1)
test $0x04, %dl
jnz L(strnlen_exit_tail2)
sub $4, %rsi
jb L(return_start_len)
lea 3(%eax), %eax
ret
.p2align 4
L(strnlen_exit_8):
test $0x10, %dl
jnz L(strnlen_exit_tail4)
test $0x20, %dl
jnz L(strnlen_exit_tail5)
test $0x40, %dl
jnz L(strnlen_exit_tail6)
sub $8, %rsi
jb L(return_start_len)
lea 7(%eax), %eax
ret
.p2align 4
L(strnlen_exit_high):
mov %dh, %ch
and $15, %ch
jz L(strnlen_exit_high_8)
test $0x01, %dh
jnz L(strnlen_exit_tail8)
test $0x02, %dh
jnz L(strnlen_exit_tail9)
test $0x04, %dh
jnz L(strnlen_exit_tail10)
sub $12, %rsi
jb L(return_start_len)
lea 11(%eax), %eax
ret
.p2align 4
L(strnlen_exit_high_8):
test $0x10, %dh
jnz L(strnlen_exit_tail12)
test $0x20, %dh
jnz L(strnlen_exit_tail13)
test $0x40, %dh
jnz L(strnlen_exit_tail14)
sub $16, %rsi
jb L(return_start_len)
lea 15(%eax), %eax
ret
.p2align 4
L(strnlen_exit_tail1):
sub $2, %rsi
jb L(return_start_len)
lea 1(%eax), %eax
ret
.p2align 4
L(strnlen_exit_tail2):
sub $3, %rsi
jb L(return_start_len)
lea 2(%eax), %eax
ret
.p2align 4
L(strnlen_exit_tail4):
sub $5, %rsi
jb L(return_start_len)
lea 4(%eax), %eax
ret
.p2align 4
L(strnlen_exit_tail5):
sub $6, %rsi
jb L(return_start_len)
lea 5(%eax), %eax
ret
.p2align 4
L(strnlen_exit_tail6):
sub $7, %rsi
jb L(return_start_len)
lea 6(%eax), %eax
ret
.p2align 4
L(strnlen_exit_tail8):
sub $9, %rsi
jb L(return_start_len)
lea 8(%eax), %eax
ret
.p2align 4
L(strnlen_exit_tail9):
sub $10, %rsi
jb L(return_start_len)
lea 9(%eax), %eax
ret
.p2align 4
L(strnlen_exit_tail10):
sub $11, %rsi
jb L(return_start_len)
lea 10(%eax), %eax
ret
.p2align 4
L(strnlen_exit_tail12):
sub $13, %rsi
jb L(return_start_len)
lea 12(%eax), %eax
ret
.p2align 4
L(strnlen_exit_tail13):
sub $14, %rsi
jb L(return_start_len)
lea 13(%eax), %eax
ret
.p2align 4
L(strnlen_exit_tail14):
sub $15, %rsi
jb L(return_start_len)
lea 14(%eax), %eax
ret
.p2align 4
L(return_start_len):
mov %r8, %rax
ret
/* for prolog only */
.p2align 4
L(len_less4_prolog):
add $4, %rsi
jz L(exit_tail0)
cmpb $0, (%rdi)
jz L(exit_tail0)
cmp $1, %esi
je L(exit_tail1)
cmpb $0, 1(%rdi)
jz L(exit_tail1)
cmp $2, %esi
je L(exit_tail2)
cmpb $0, 2(%rdi)
jz L(exit_tail2)
cmp $3, %esi
je L(exit_tail3)
cmpb $0, 3(%rdi)
jz L(exit_tail3)
mov $4, %eax
ret
.p2align 4
L(len_less8_prolog):
add $4, %rsi
cmpb $0, 4(%rdi)
jz L(exit_tail4)
cmp $1, %esi
je L(exit_tail5)
cmpb $0, 5(%rdi)
jz L(exit_tail5)
cmp $2, %esi
je L(exit_tail6)
cmpb $0, 6(%rdi)
jz L(exit_tail6)
cmp $3, %esi
je L(exit_tail7)
cmpb $0, 7(%rdi)
jz L(exit_tail7)
mov $8, %eax
ret
.p2align 4
L(len_less12_prolog):
add $4, %rsi
cmpb $0, 8(%rdi)
jz L(exit_tail8)
cmp $1, %esi
je L(exit_tail9)
cmpb $0, 9(%rdi)
jz L(exit_tail9)
cmp $2, %esi
je L(exit_tail10)
cmpb $0, 10(%rdi)
jz L(exit_tail10)
cmp $3, %esi
je L(exit_tail11)
cmpb $0, 11(%rdi)
jz L(exit_tail11)
mov $12, %eax
ret
.p2align 4
L(len_less16_prolog):
add $4, %rsi
cmpb $0, 12(%rdi)
jz L(exit_tail12)
cmp $1, %esi
je L(exit_tail13)
cmpb $0, 13(%rdi)
jz L(exit_tail13)
cmp $2, %esi
je L(exit_tail14)
cmpb $0, 14(%rdi)
jz L(exit_tail14)
cmp $3, %esi
je L(exit_tail15)
cmpb $0, 15(%rdi)
jz L(exit_tail15)
mov $16, %eax
ret
# endif
.p2align 4
L(exit_tail1):
add $1, %eax
RETURN
.p2align 4
L(exit_tail2):
add $2, %eax
RETURN
.p2align 4
L(exit_tail3):
add $3, %eax
RETURN
.p2align 4
L(exit_tail4):
add $4, %eax
RETURN
.p2align 4
L(exit_tail5):
add $5, %eax
RETURN
.p2align 4
L(exit_tail6):
add $6, %eax
RETURN
.p2align 4
L(exit_tail7):
add $7, %eax
RETURN
.p2align 4
L(exit_tail8):
add $8, %eax
RETURN
.p2align 4
L(exit_tail9):
add $9, %eax
RETURN
.p2align 4
L(exit_tail10):
add $10, %eax
RETURN
.p2align 4
L(exit_tail11):
add $11, %eax
RETURN
.p2align 4
L(exit_tail12):
add $12, %eax
RETURN
.p2align 4
L(exit_tail13):
add $13, %eax
RETURN
.p2align 4
L(exit_tail14):
add $14, %eax
RETURN
.p2align 4
L(exit_tail15):
add $15, %eax
# ifndef USE_AS_STRCAT
RETURN
END (STRLEN)
# endif
#endif

2
sysdeps/x86_64/multiarch/strlen.S
View File

@ -42,7 +42,7 @@ ENTRY(strlen)
ret
2: testl $bit_Slow_BSF, __cpu_features+FEATURE_OFFSET+index_Slow_BSF(%rip)
jz 3f
leaq __strlen_no_bsf(%rip), %rax
leaq __strlen_sse2_no_bsf(%rip), %rax
3: ret
END(strlen)

3
sysdeps/x86_64/multiarch/strnlen-sse2-no-bsf.S Normal file
View File

@ -0,0 +1,3 @@
#define USE_AS_STRNLEN
#define STRLEN __strnlen_sse2_no_bsf
#include "strlen-sse2-no-bsf.S"

55
sysdeps/x86_64/multiarch/strnlen.S Normal file
View File

@ -0,0 +1,55 @@
/* multiple version of strnlen
Copyright (C) 2011 Free Software Foundation, Inc.
This file is part of the GNU C Library.
The GNU C Library is free software; you can redistribute it and/or
modify it under the terms of the GNU Lesser General Public
License as published by the Free Software Foundation; either
version 2.1 of the License, or (at your option) any later version.
The GNU C Library is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
Lesser General Public License for more details.
You should have received a copy of the GNU Lesser General Public
License along with the GNU C Library; if not, write to the Free
Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
02111-1307 USA. */
#include <sysdep.h>
#include <init-arch.h>
/* Define multiple versions only for the definition in libc. */
#ifndef NOT_IN_libc
.text
ENTRY(__strnlen)
.type __strnlen, @gnu_indirect_function
cmpl $0, __cpu_features+KIND_OFFSET(%rip)
jne 1f
call __init_cpu_features
1: leaq __strnlen_sse2(%rip), %rax
testl $bit_Slow_BSF, __cpu_features+FEATURE_OFFSET+index_Slow_BSF(%rip)
jz 2f
leaq __strnlen_sse2_no_bsf(%rip), %rax
2: ret
END(__strnlen)
# undef ENTRY
# define ENTRY(name) \
.type __strnlen_sse2, @function; \
.align 16; \
__strnlen_sse2: cfi_startproc; \
CALL_MCOUNT
# undef END
# define END(name) \
cfi_endproc; .size __strnlen_sse2, .-__strnlen_sse2
# undef libc_hidden_def
# define libc_hidden_def(name) \
.globl __GI_strnlen; __GI_strnlen = __strnlen_sse2
#endif
#include "../strnlen.S"

239
sysdeps/x86_64/wcslen.S Normal file
View File

@ -0,0 +1,239 @@
/* Optimized wcslen for x86-64 with SSE2.
Copyright (C) 2011 Free Software Foundation, Inc.
Contributed by Intel Corporation.
This file is part of the GNU C Library.
The GNU C Library is free software; you can redistribute it and/or
modify it under the terms of the GNU Lesser General Public
License as published by the Free Software Foundation; either
version 2.1 of the License, or (at your option) any later version.
The GNU C Library is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
Lesser General Public License for more details.
You should have received a copy of the GNU Lesser General Public
License along with the GNU C Library; if not, write to the Free
Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
02111-1307 USA. */
#include <sysdep.h>
.text
ENTRY (__wcslen)
cmpl $0, (%rdi)
jz L(exit_tail0)
cmpl $0, 4(%rdi)
jz L(exit_tail1)
cmpl $0, 8(%rdi)
jz L(exit_tail2)
cmpl $0, 12(%rdi)
jz L(exit_tail3)
cmpl $0, 16(%rdi)
jz L(exit_tail4)
cmpl $0, 20(%rdi)
jz L(exit_tail5)
cmpl $0, 24(%rdi)
jz L(exit_tail6)
cmpl $0, 28(%rdi)
jz L(exit_tail7)
pxor %xmm0, %xmm0
lea 32(%rdi), %rax
lea 16(%rdi), %rcx
and $-16, %rax
pcmpeqd (%rax), %xmm0
pmovmskb %xmm0, %edx
pxor %xmm1, %xmm1
test %edx, %edx
lea 16(%rax), %rax
jnz L(exit)
pcmpeqd (%rax), %xmm1
pmovmskb %xmm1, %edx
pxor %xmm2, %xmm2
test %edx, %edx
lea 16(%rax), %rax
jnz L(exit)
pcmpeqd (%rax), %xmm2
pmovmskb %xmm2, %edx
pxor %xmm3, %xmm3
test %edx, %edx
lea 16(%rax), %rax
jnz L(exit)
pcmpeqd (%rax), %xmm3
pmovmskb %xmm3, %edx
test %edx, %edx
lea 16(%rax), %rax
jnz L(exit)
pcmpeqd (%rax), %xmm0
pmovmskb %xmm0, %edx
test %edx, %edx
lea 16(%rax), %rax
jnz L(exit)
pcmpeqd (%rax), %xmm1
pmovmskb %xmm1, %edx
test %edx, %edx
lea 16(%rax), %rax
jnz L(exit)
pcmpeqd (%rax), %xmm2
pmovmskb %xmm2, %edx
test %edx, %edx
lea 16(%rax), %rax
jnz L(exit)
pcmpeqd (%rax), %xmm3
pmovmskb %xmm3, %edx
test %edx, %edx
lea 16(%rax), %rax
jnz L(exit)
pcmpeqd (%rax), %xmm0
pmovmskb %xmm0, %edx
test %edx, %edx
lea 16(%rax), %rax
jnz L(exit)
pcmpeqd (%rax), %xmm1
pmovmskb %xmm1, %edx
test %edx, %edx
lea 16(%rax), %rax
jnz L(exit)
pcmpeqd (%rax), %xmm2
pmovmskb %xmm2, %edx
test %edx, %edx
lea 16(%rax), %rax
jnz L(exit)
pcmpeqd (%rax), %xmm3
pmovmskb %xmm3, %edx
test %edx, %edx
lea 16(%rax), %rax
jnz L(exit)
and $-0x40, %rax
.p2align 4
L(aligned_64_loop):
movaps (%rax), %xmm0
movaps 16(%rax), %xmm1
movaps 32(%rax), %xmm2
movaps 48(%rax), %xmm6
pminub %xmm1, %xmm0
pminub %xmm6, %xmm2
pminub %xmm0, %xmm2
pcmpeqd %xmm3, %xmm2
pmovmskb %xmm2, %edx
test %edx, %edx
lea 64(%rax), %rax
jz L(aligned_64_loop)
pcmpeqd -64(%rax), %xmm3
pmovmskb %xmm3, %edx
test %edx, %edx
lea 48(%rcx), %rcx
jnz L(exit)
pcmpeqd %xmm1, %xmm3
pmovmskb %xmm3, %edx
test %edx, %edx
lea -16(%rcx), %rcx
jnz L(exit)
pcmpeqd -32(%rax), %xmm3
pmovmskb %xmm3, %edx
test %edx, %edx
lea -16(%rcx), %rcx
jnz L(exit)
pcmpeqd %xmm6, %xmm3
pmovmskb %xmm3, %edx
test %edx, %edx
lea -16(%rcx), %rcx
jnz L(exit)
jmp L(aligned_64_loop)
.p2align 4
L(exit):
sub %rcx, %rax
shr $2, %rax
test %dl, %dl
jz L(exit_high)
mov %dl, %cl
and $15, %cl
jz L(exit_1)
ret
.p2align 4
L(exit_high):
mov %dh, %ch
and $15, %ch
jz L(exit_3)
add $2, %rax
ret
.p2align 4
L(exit_1):
add $1, %rax
ret
.p2align 4
L(exit_3):
add $3, %rax
ret
.p2align 4
L(exit_tail0):
xor %rax, %rax
ret
.p2align 4
L(exit_tail1):
mov $1, %rax
ret
.p2align 4
L(exit_tail2):
mov $2, %rax
ret
.p2align 4
L(exit_tail3):
mov $3, %rax
ret
.p2align 4
L(exit_tail4):
mov $4, %rax
ret
.p2align 4
L(exit_tail5):
mov $5, %rax
ret
.p2align 4
L(exit_tail6):
mov $6, %rax
ret
.p2align 4
L(exit_tail7):
mov $7, %rax
ret
END (__wcslen)
weak_alias(__wcslen, wcslen)