glibc/sysdeps/i386/i686/multiarch/strrchr-sse2-bsf.S

283 lines
4.7 KiB
ArmAsm

/* strrchr with SSE2 with bsf and bsr
Copyright (C) 2011-2013 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, see
<http://www.gnu.org/licenses/>. */
#ifndef NOT_IN_libc
# include <sysdep.h>
# define CFI_PUSH(REG) \
cfi_adjust_cfa_offset (4); \
cfi_rel_offset (REG, 0)
# define CFI_POP(REG) \
cfi_adjust_cfa_offset (-4); \
cfi_restore (REG)
# define PUSH(REG) pushl REG; CFI_PUSH (REG)
# define POP(REG) popl REG; CFI_POP (REG)
# define PARMS 4
# define STR1 PARMS
# define STR2 STR1+4
.text
ENTRY (__strrchr_sse2_bsf)
mov STR1(%esp), %ecx
movd STR2(%esp), %xmm1
PUSH (%edi)
pxor %xmm2, %xmm2
mov %ecx, %edi
punpcklbw %xmm1, %xmm1
punpcklbw %xmm1, %xmm1
/* ECX has OFFSET. */
and $63, %ecx
cmp $48, %ecx
pshufd $0, %xmm1, %xmm1
ja L(crosscashe)
/* unaligned string. */
movdqu (%edi), %xmm0
pcmpeqb %xmm0, %xmm2
pcmpeqb %xmm1, %xmm0
/* Find where NULL is. */
pmovmskb %xmm2, %edx
/* Check if there is a match. */
pmovmskb %xmm0, %eax
test %eax, %eax
jnz L(unaligned_match1)
test %edx, %edx
jnz L(return_null)
and $-16, %edi
add $16, %edi
PUSH (%esi)
PUSH (%ebx)
xor %ebx, %ebx
jmp L(loop)
CFI_POP (%esi)
CFI_POP (%ebx)
.p2align 4
L(unaligned_return_value1):
bsf %edx, %ecx
mov $2, %edx
shl %cl, %edx
sub $1, %edx
and %edx, %eax
jz L(return_null)
bsr %eax, %eax
add %edi, %eax
POP (%edi)
ret
CFI_PUSH (%edi)
.p2align 4
L(unaligned_match1):
test %edx, %edx
jnz L(unaligned_return_value1)
PUSH (%esi)
PUSH (%ebx)
mov %eax, %ebx
lea 16(%edi), %esi
and $-16, %edi
add $16, %edi
jmp L(loop)
CFI_POP (%esi)
CFI_POP (%ebx)
.p2align 4
L(crosscashe):
/* Hancle unaligned string. */
and $15, %ecx
and $-16, %edi
pxor %xmm3, %xmm3
movdqa (%edi), %xmm0
pcmpeqb %xmm0, %xmm3
pcmpeqb %xmm1, %xmm0
/* Find where NULL is. */
pmovmskb %xmm3, %edx
/* Check if there is a match. */
pmovmskb %xmm0, %eax
/* Remove the leading bytes. */
shr %cl, %edx
shr %cl, %eax
test %eax, %eax
jnz L(unaligned_match)
test %edx, %edx
jnz L(return_null)
add $16, %edi
PUSH (%esi)
PUSH (%ebx)
xor %ebx, %ebx
jmp L(loop)
CFI_POP (%esi)
CFI_POP (%ebx)
.p2align 4
L(unaligned_return_value):
add %ecx, %edi
bsf %edx, %ecx
mov $2, %edx
shl %cl, %edx
sub $1, %edx
and %edx, %eax
jz L(return_null)
bsr %eax, %eax
add %edi, %eax
POP (%edi)
ret
CFI_PUSH (%edi)
.p2align 4
L(unaligned_match):
test %edx, %edx
jnz L(unaligned_return_value)
PUSH (%esi)
PUSH (%ebx)
mov %eax, %ebx
add $16, %edi
lea (%edi, %ecx), %esi
/* Loop start on aligned string. */
.p2align 4
L(loop):
movdqa (%edi), %xmm0
pcmpeqb %xmm0, %xmm2
add $16, %edi
pcmpeqb %xmm1, %xmm0
pmovmskb %xmm2, %ecx
pmovmskb %xmm0, %eax
or %eax, %ecx
jnz L(matches)
movdqa (%edi), %xmm0
pcmpeqb %xmm0, %xmm2
add $16, %edi
pcmpeqb %xmm1, %xmm0
pmovmskb %xmm2, %ecx
pmovmskb %xmm0, %eax
or %eax, %ecx
jnz L(matches)
movdqa (%edi), %xmm0
pcmpeqb %xmm0, %xmm2
add $16, %edi
pcmpeqb %xmm1, %xmm0
pmovmskb %xmm2, %ecx
pmovmskb %xmm0, %eax
or %eax, %ecx
jnz L(matches)
movdqa (%edi), %xmm0
pcmpeqb %xmm0, %xmm2
add $16, %edi
pcmpeqb %xmm1, %xmm0
pmovmskb %xmm2, %ecx
pmovmskb %xmm0, %eax
or %eax, %ecx
jz L(loop)
L(matches):
test %eax, %eax
jnz L(match)
L(return_value):
test %ebx, %ebx
jz L(return_null_1)
bsr %ebx, %eax
add %esi, %eax
POP (%ebx)
POP (%esi)
sub $16, %eax
POP (%edi)
ret
CFI_PUSH (%edi)
CFI_PUSH (%ebx)
CFI_PUSH (%esi)
.p2align 4
L(match):
pmovmskb %xmm2, %ecx
test %ecx, %ecx
jnz L(return_value_1)
mov %eax, %ebx
mov %edi, %esi
jmp L(loop)
.p2align 4
L(return_value_1):
bsf %ecx, %ecx
mov $2, %edx
shl %cl, %edx
sub $1, %edx
and %edx, %eax
jz L(return_value)
POP (%ebx)
POP (%esi)
bsr %eax, %eax
add %edi, %eax
sub $16, %eax
POP (%edi)
ret
CFI_PUSH (%edi)
/* Return NULL. */
.p2align 4
L(return_null):
xor %eax, %eax
POP (%edi)
ret
CFI_PUSH (%edi)
CFI_PUSH (%ebx)
CFI_PUSH (%esi)
/* Return NULL. */
.p2align 4
L(return_null_1):
POP (%ebx)
POP (%esi)
POP (%edi)
xor %eax, %eax
ret
END (__strrchr_sse2_bsf)
#endif