OpenE2K
/
glibc
2
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

Use p2align instead ALIGN

This commit is contained in:
Ondřej Bílka 2013-10-08 15:46:48 +02:00
parent 41500766f7
commit e7044ea76b
10 changed files with 301 additions and 323 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

13
ChangeLog
View File

@ -1,3 +1,16 @@
2013-10-08 Ondřej Bílka <neleai@seznam.cz>
* sysdeps/x86_64/memset.S (ALIGN): Macro removed.
Use .p2align directive instead, throughout.
* sysdeps/x86_64/multiarch/memcmp-sse4.S: Likewise.
* sysdeps/x86_64/multiarch/memcmp-ssse3.S: Likewise.
* sysdeps/x86_64/multiarch/memcpy-sse2-unaligned.S: Likewise.
* sysdeps/x86_64/multiarch/memcpy-ssse3-back.S: Likewise.
* sysdeps/x86_64/multiarch/memcpy-ssse3.S: Likewise.
* sysdeps/x86_64/multiarch/strcmp-sse2-unaligned.S: Likewise.
* sysdeps/x86_64/strchr.S: Likewise.
* sysdeps/x86_64/strrchr.S: Likewise.
2013-10-08 Siddhesh Poyarekar <siddhesh@redhat.com>
* sysdeps/ieee754/dbl-64/e_pow.c: Fix code formatting.

10
sysdeps/x86_64/memset.S
View File

@ -19,10 +19,6 @@
#include <sysdep.h>
#ifndef ALIGN
# define ALIGN(n) .p2align n
#endif
.text
#if !defined NOT_IN_libc
ENTRY(__bzero)
@ -71,12 +67,12 @@ L(entry_from_bzero):
L(return):
rep
ret
ALIGN (4)
.p2align 4
L(between_32_64_bytes):
movdqu %xmm8, 16(%rdi)
movdqu %xmm8, -32(%rdi,%rdx)
ret
ALIGN (4)
.p2align 4
L(loop_start):
leaq 64(%rdi), %rcx
movdqu %xmm8, (%rdi)
@ -92,7 +88,7 @@ L(loop_start):
andq $-64, %rdx
cmpq %rdx, %rcx
je L(return)
ALIGN (4)
.p2align 4
L(loop):
movdqa %xmm8, (%rcx)
movdqa %xmm8, 16(%rcx)

84
sysdeps/x86_64/multiarch/memcmp-sse4.S
View File

@ -25,10 +25,6 @@
# define MEMCMP __memcmp_sse4_1
# endif
# ifndef ALIGN
# define ALIGN(n) .p2align n
# endif
# define JMPTBL(I, B) (I - B)
# define BRANCH_TO_JMPTBL_ENTRY(TABLE, INDEX, SCALE) \
@ -60,7 +56,7 @@ ENTRY (MEMCMP)
BRANCH_TO_JMPTBL_ENTRY(L(table_64bytes), %rdx, 4)
# ifndef USE_AS_WMEMCMP
ALIGN (4)
.p2align 4
L(firstbyte):
movzbl (%rdi), %eax
movzbl (%rsi), %ecx
@ -68,7 +64,7 @@ L(firstbyte):
ret
# endif
ALIGN (4)
.p2align 4
L(79bytesormore):
movdqu (%rsi), %xmm1
movdqu (%rdi), %xmm2
@ -316,7 +312,7 @@ L(less32bytesin256):
add %rdx, %rdi
BRANCH_TO_JMPTBL_ENTRY(L(table_64bytes), %rdx, 4)
ALIGN (4)
.p2align 4
L(512bytesormore):
# ifdef DATA_CACHE_SIZE_HALF
mov $DATA_CACHE_SIZE_HALF, %R8_LP
@ -329,7 +325,7 @@ L(512bytesormore):
cmp %r8, %rdx
ja L(L2_L3_cache_unaglined)
sub $64, %rdx
ALIGN (4)
.p2align 4
L(64bytesormore_loop):
movdqu (%rdi), %xmm2
pxor (%rsi), %xmm2
@ -361,7 +357,7 @@ L(64bytesormore_loop):
L(L2_L3_cache_unaglined):
sub $64, %rdx
ALIGN (4)
.p2align 4
L(L2_L3_unaligned_128bytes_loop):
prefetchnta 0x1c0(%rdi)
prefetchnta 0x1c0(%rsi)
@ -396,7 +392,7 @@ L(L2_L3_unaligned_128bytes_loop):
/*
* This case is for machines which are sensitive for unaligned instructions.
*/
ALIGN (4)
.p2align 4
L(2aligned):
cmp $128, %rdx
ja L(128bytesormorein2aligned)
@ -444,7 +440,7 @@ L(less32bytesin64in2alinged):
add %rdx, %rdi
BRANCH_TO_JMPTBL_ENTRY(L(table_64bytes), %rdx, 4)
ALIGN (4)
.p2align 4
L(128bytesormorein2aligned):
cmp $512, %rdx
ja L(512bytesormorein2aligned)
@ -519,7 +515,7 @@ L(less32bytesin128in2aligned):
add %rdx, %rdi
BRANCH_TO_JMPTBL_ENTRY(L(table_64bytes), %rdx, 4)
ALIGN (4)
.p2align 4
L(256bytesormorein2aligned):
sub $256, %rdx
@ -632,7 +628,7 @@ L(less32bytesin256in2alinged):
add %rdx, %rdi
BRANCH_TO_JMPTBL_ENTRY(L(table_64bytes), %rdx, 4)
ALIGN (4)
.p2align 4
L(512bytesormorein2aligned):
# ifdef DATA_CACHE_SIZE_HALF
mov $DATA_CACHE_SIZE_HALF, %R8_LP
@ -646,7 +642,7 @@ L(512bytesormorein2aligned):
ja L(L2_L3_cache_aglined)
sub $64, %rdx
ALIGN (4)
.p2align 4
L(64bytesormore_loopin2aligned):
movdqa (%rdi), %xmm2
pxor (%rsi), %xmm2
@ -678,7 +674,7 @@ L(64bytesormore_loopin2aligned):
L(L2_L3_cache_aglined):
sub $64, %rdx
ALIGN (4)
.p2align 4
L(L2_L3_aligned_128bytes_loop):
prefetchnta 0x1c0(%rdi)
prefetchnta 0x1c0(%rsi)
@ -711,7 +707,7 @@ L(L2_L3_aligned_128bytes_loop):
BRANCH_TO_JMPTBL_ENTRY(L(table_64bytes), %rdx, 4)
ALIGN (4)
.p2align 4
L(64bytesormore_loop_end):
add $16, %rdi
add $16, %rsi
@ -806,7 +802,7 @@ L(8bytes):
xor %eax, %eax
ret
ALIGN (4)
.p2align 4
L(12bytes):
mov -12(%rdi), %rax
mov -12(%rsi), %rcx
@ -827,7 +823,7 @@ L(0bytes):
# ifndef USE_AS_WMEMCMP
/* unreal case for wmemcmp */
ALIGN (4)
.p2align 4
L(65bytes):
movdqu -65(%rdi), %xmm1
movdqu -65(%rsi), %xmm2
@ -864,7 +860,7 @@ L(9bytes):
sub %edx, %eax
ret
ALIGN (4)
.p2align 4
L(13bytes):
mov -13(%rdi), %rax
mov -13(%rsi), %rcx
@ -877,7 +873,7 @@ L(13bytes):
xor %eax, %eax
ret
ALIGN (4)
.p2align 4
L(5bytes):
mov -5(%rdi), %eax
mov -5(%rsi), %ecx
@ -888,7 +884,7 @@ L(5bytes):
sub %edx, %eax
ret
ALIGN (4)
.p2align 4
L(66bytes):
movdqu -66(%rdi), %xmm1
movdqu -66(%rsi), %xmm2
@ -929,7 +925,7 @@ L(10bytes):
sub %ecx, %eax
ret
ALIGN (4)
.p2align 4
L(14bytes):
mov -14(%rdi), %rax
mov -14(%rsi), %rcx
@ -942,7 +938,7 @@ L(14bytes):
xor %eax, %eax
ret
ALIGN (4)
.p2align 4
L(6bytes):
mov -6(%rdi), %eax
mov -6(%rsi), %ecx
@ -958,7 +954,7 @@ L(2bytes):
sub %ecx, %eax
ret
ALIGN (4)
.p2align 4
L(67bytes):
movdqu -67(%rdi), %xmm2
movdqu -67(%rsi), %xmm1
@ -997,7 +993,7 @@ L(11bytes):
xor %eax, %eax
ret
ALIGN (4)
.p2align 4
L(15bytes):
mov -15(%rdi), %rax
mov -15(%rsi), %rcx
@ -1010,7 +1006,7 @@ L(15bytes):
xor %eax, %eax
ret
ALIGN (4)
.p2align 4
L(7bytes):
mov -7(%rdi), %eax
mov -7(%rsi), %ecx
@ -1023,7 +1019,7 @@ L(7bytes):
xor %eax, %eax
ret
ALIGN (4)
.p2align 4
L(3bytes):
movzwl -3(%rdi), %eax
movzwl -3(%rsi), %ecx
@ -1036,7 +1032,7 @@ L(1bytes):
ret
# endif
ALIGN (4)
.p2align 4
L(68bytes):
movdqu -68(%rdi), %xmm2
movdqu -68(%rsi), %xmm1
@ -1079,7 +1075,7 @@ L(20bytes):
# ifndef USE_AS_WMEMCMP
/* unreal cases for wmemcmp */
ALIGN (4)
.p2align 4
L(69bytes):
movdqu -69(%rsi), %xmm1
movdqu -69(%rdi), %xmm2
@ -1115,7 +1111,7 @@ L(21bytes):
xor %eax, %eax
ret
ALIGN (4)
.p2align 4
L(70bytes):
movdqu -70(%rsi), %xmm1
movdqu -70(%rdi), %xmm2
@ -1151,7 +1147,7 @@ L(22bytes):
xor %eax, %eax
ret
ALIGN (4)
.p2align 4
L(71bytes):
movdqu -71(%rsi), %xmm1
movdqu -71(%rdi), %xmm2
@ -1188,7 +1184,7 @@ L(23bytes):
ret
# endif
ALIGN (4)
.p2align 4
L(72bytes):
movdqu -72(%rsi), %xmm1
movdqu -72(%rdi), %xmm2
@ -1227,7 +1223,7 @@ L(24bytes):
# ifndef USE_AS_WMEMCMP
/* unreal cases for wmemcmp */
ALIGN (4)
.p2align 4
L(73bytes):
movdqu -73(%rsi), %xmm1
movdqu -73(%rdi), %xmm2
@ -1265,7 +1261,7 @@ L(25bytes):
sub %ecx, %eax
ret
ALIGN (4)
.p2align 4
L(74bytes):
movdqu -74(%rsi), %xmm1
movdqu -74(%rdi), %xmm2
@ -1302,7 +1298,7 @@ L(26bytes):
movzwl -2(%rsi), %ecx
jmp L(diffin2bytes)
ALIGN (4)
.p2align 4
L(75bytes):
movdqu -75(%rsi), %xmm1
movdqu -75(%rdi), %xmm2
@ -1342,7 +1338,7 @@ L(27bytes):
xor %eax, %eax
ret
# endif
ALIGN (4)
.p2align 4
L(76bytes):
movdqu -76(%rsi), %xmm1
movdqu -76(%rdi), %xmm2
@ -1388,7 +1384,7 @@ L(28bytes):
# ifndef USE_AS_WMEMCMP
/* unreal cases for wmemcmp */
ALIGN (4)
.p2align 4
L(77bytes):
movdqu -77(%rsi), %xmm1
movdqu -77(%rdi), %xmm2
@ -1430,7 +1426,7 @@ L(29bytes):
xor %eax, %eax
ret
ALIGN (4)
.p2align 4
L(78bytes):
movdqu -78(%rsi), %xmm1
movdqu -78(%rdi), %xmm2
@ -1470,7 +1466,7 @@ L(30bytes):
xor %eax, %eax
ret
ALIGN (4)
.p2align 4
L(79bytes):
movdqu -79(%rsi), %xmm1
movdqu -79(%rdi), %xmm2
@ -1510,7 +1506,7 @@ L(31bytes):
xor %eax, %eax
ret
# endif
ALIGN (4)
.p2align 4
L(64bytes):
movdqu -64(%rdi), %xmm2
movdqu -64(%rsi), %xmm1
@ -1548,7 +1544,7 @@ L(32bytes):
/*
* Aligned 8 bytes to avoid 2 branch "taken" in one 16 alinged code block.
*/
ALIGN (3)
.p2align 3
L(less16bytes):
movsbq %dl, %rdx
mov (%rsi, %rdx), %rcx
@ -1585,7 +1581,7 @@ L(diffin2bytes):
sub %ecx, %eax
ret
ALIGN (4)
.p2align 4
L(end):
and $0xff, %eax
and $0xff, %ecx
@ -1599,7 +1595,7 @@ L(end):
neg %eax
ret
ALIGN (4)
.p2align 4
L(nequal_bigger):
ret
@ -1611,7 +1607,7 @@ L(unreal_case):
END (MEMCMP)
.section .rodata.sse4.1,"a",@progbits
ALIGN (3)
.p2align 3
# ifndef USE_AS_WMEMCMP
L(table_64bytes):
.int JMPTBL (L(0bytes), L(table_64bytes))

126
sysdeps/x86_64/multiarch/memcmp-ssse3.S
View File

@ -25,10 +25,6 @@
# define MEMCMP __memcmp_ssse3
# endif
# ifndef ALIGN
# define ALIGN(n) .p2align n
# endif
/* Warning!
wmemcmp has to use SIGNED comparison for elements.
memcmp has to use UNSIGNED comparison for elemnts.
@ -50,7 +46,7 @@ ENTRY (MEMCMP)
add %rcx, %rdi
jmp L(less48bytes)
ALIGN (4)
.p2align 4
/* ECX >= 32. */
L(48bytesormore):
movdqu (%rdi), %xmm3
@ -90,7 +86,7 @@ L(48bytesormore):
je L(shr_6)
jmp L(shr_7)
ALIGN (2)
.p2align 2
L(next_unaligned_table):
cmp $8, %edx
je L(shr_8)
@ -117,7 +113,7 @@ L(next_unaligned_table):
jmp L(shr_12)
# endif
ALIGN (4)
.p2align 4
L(shr_0):
cmp $80, %rcx
lea -48(%rcx), %rcx
@ -137,7 +133,7 @@ L(shr_0):
add %rcx, %rdi
jmp L(less48bytes)
ALIGN (4)
.p2align 4
L(shr_0_gobble):
movdqa (%rsi), %xmm0
xor %eax, %eax
@ -180,7 +176,7 @@ L(next):
# ifndef USE_AS_WMEMCMP
ALIGN (4)
.p2align 4
L(shr_1):
cmp $80, %rcx
lea -48(%rcx), %rcx
@ -207,7 +203,7 @@ L(shr_1):
add %rcx, %rdi
jmp L(less48bytes)
ALIGN (4)
.p2align 4
L(shr_1_gobble):
sub $32, %rcx
movdqa 16(%rsi), %xmm0
@ -258,7 +254,7 @@ L(shr_1_gobble_next):
jmp L(less48bytes)
ALIGN (4)
.p2align 4
L(shr_2):
cmp $80, %rcx
lea -48(%rcx), %rcx
@ -285,7 +281,7 @@ L(shr_2):
add %rcx, %rdi
jmp L(less48bytes)
ALIGN (4)
.p2align 4
L(shr_2_gobble):
sub $32, %rcx
movdqa 16(%rsi), %xmm0
@ -335,7 +331,7 @@ L(shr_2_gobble_next):
add %rcx, %rdi
jmp L(less48bytes)
ALIGN (4)
.p2align 4
L(shr_3):
cmp $80, %rcx
lea -48(%rcx), %rcx
@ -362,7 +358,7 @@ L(shr_3):
add %rcx, %rdi
jmp L(less48bytes)
ALIGN (4)
.p2align 4
L(shr_3_gobble):
sub $32, %rcx
movdqa 16(%rsi), %xmm0
@ -414,7 +410,7 @@ L(shr_3_gobble_next):
# endif
ALIGN (4)
.p2align 4
L(shr_4):
cmp $80, %rcx
lea -48(%rcx), %rcx
@ -441,7 +437,7 @@ L(shr_4):
add %rcx, %rdi
jmp L(less48bytes)
ALIGN (4)
.p2align 4
L(shr_4_gobble):
sub $32, %rcx
movdqa 16(%rsi), %xmm0
@ -493,7 +489,7 @@ L(shr_4_gobble_next):
# ifndef USE_AS_WMEMCMP
ALIGN (4)
.p2align 4
L(shr_5):
cmp $80, %rcx
lea -48(%rcx), %rcx
@ -520,7 +516,7 @@ L(shr_5):
add %rcx, %rdi
jmp L(less48bytes)
ALIGN (4)
.p2align 4
L(shr_5_gobble):
sub $32, %rcx
movdqa 16(%rsi), %xmm0
@ -570,7 +566,7 @@ L(shr_5_gobble_next):
add %rcx, %rdi
jmp L(less48bytes)
ALIGN (4)
.p2align 4
L(shr_6):
cmp $80, %rcx
lea -48(%rcx), %rcx
@ -597,7 +593,7 @@ L(shr_6):
add %rcx, %rdi
jmp L(less48bytes)
ALIGN (4)
.p2align 4
L(shr_6_gobble):
sub $32, %rcx
movdqa 16(%rsi), %xmm0
@ -647,7 +643,7 @@ L(shr_6_gobble_next):
add %rcx, %rdi
jmp L(less48bytes)
ALIGN (4)
.p2align 4
L(shr_7):
cmp $80, %rcx
lea -48(%rcx), %rcx
@ -674,7 +670,7 @@ L(shr_7):
add %rcx, %rdi
jmp L(less48bytes)
ALIGN (4)
.p2align 4
L(shr_7_gobble):
sub $32, %rcx
movdqa 16(%rsi), %xmm0
@ -726,7 +722,7 @@ L(shr_7_gobble_next):
# endif
ALIGN (4)
.p2align 4
L(shr_8):
cmp $80, %rcx
lea -48(%rcx), %rcx
@ -753,7 +749,7 @@ L(shr_8):
add %rcx, %rdi
jmp L(less48bytes)
ALIGN (4)
.p2align 4
L(shr_8_gobble):
sub $32, %rcx
movdqa 16(%rsi), %xmm0
@ -805,7 +801,7 @@ L(shr_8_gobble_next):
# ifndef USE_AS_WMEMCMP
ALIGN (4)
.p2align 4
L(shr_9):
cmp $80, %rcx
lea -48(%rcx), %rcx
@ -832,7 +828,7 @@ L(shr_9):
add %rcx, %rdi
jmp L(less48bytes)
ALIGN (4)
.p2align 4
L(shr_9_gobble):
sub $32, %rcx
movdqa 16(%rsi), %xmm0
@ -882,7 +878,7 @@ L(shr_9_gobble_next):
add %rcx, %rdi
jmp L(less48bytes)
ALIGN (4)
.p2align 4
L(shr_10):
cmp $80, %rcx
lea -48(%rcx), %rcx
@ -909,7 +905,7 @@ L(shr_10):
add %rcx, %rdi
jmp L(less48bytes)
ALIGN (4)
.p2align 4
L(shr_10_gobble):
sub $32, %rcx
movdqa 16(%rsi), %xmm0
@ -959,7 +955,7 @@ L(shr_10_gobble_next):
add %rcx, %rdi
jmp L(less48bytes)
ALIGN (4)
.p2align 4
L(shr_11):
cmp $80, %rcx
lea -48(%rcx), %rcx
@ -986,7 +982,7 @@ L(shr_11):
add %rcx, %rdi
jmp L(less48bytes)
ALIGN (4)
.p2align 4
L(shr_11_gobble):
sub $32, %rcx
movdqa 16(%rsi), %xmm0
@ -1038,7 +1034,7 @@ L(shr_11_gobble_next):
# endif
ALIGN (4)
.p2align 4
L(shr_12):
cmp $80, %rcx
lea -48(%rcx), %rcx
@ -1065,7 +1061,7 @@ L(shr_12):
add %rcx, %rdi
jmp L(less48bytes)
ALIGN (4)
.p2align 4
L(shr_12_gobble):
sub $32, %rcx
movdqa 16(%rsi), %xmm0
@ -1117,7 +1113,7 @@ L(shr_12_gobble_next):
# ifndef USE_AS_WMEMCMP
ALIGN (4)
.p2align 4
L(shr_13):
cmp $80, %rcx
lea -48(%rcx), %rcx
@ -1144,7 +1140,7 @@ L(shr_13):
add %rcx, %rdi
jmp L(less48bytes)
ALIGN (4)
.p2align 4
L(shr_13_gobble):
sub $32, %rcx
movdqa 16(%rsi), %xmm0
@ -1194,7 +1190,7 @@ L(shr_13_gobble_next):
add %rcx, %rdi
jmp L(less48bytes)
ALIGN (4)
.p2align 4
L(shr_14):
cmp $80, %rcx
lea -48(%rcx), %rcx
@ -1221,7 +1217,7 @@ L(shr_14):
add %rcx, %rdi
jmp L(less48bytes)
ALIGN (4)
.p2align 4
L(shr_14_gobble):
sub $32, %rcx
movdqa 16(%rsi), %xmm0
@ -1271,7 +1267,7 @@ L(shr_14_gobble_next):
add %rcx, %rdi
jmp L(less48bytes)
ALIGN (4)
.p2align 4
L(shr_15):
cmp $80, %rcx
lea -48(%rcx), %rcx
@ -1298,7 +1294,7 @@ L(shr_15):
add %rcx, %rdi
jmp L(less48bytes)
ALIGN (4)
.p2align 4
L(shr_15_gobble):
sub $32, %rcx
movdqa 16(%rsi), %xmm0
@ -1348,7 +1344,7 @@ L(shr_15_gobble_next):
add %rcx, %rdi
jmp L(less48bytes)
# endif
ALIGN (4)
.p2align 4
L(exit):
pmovmskb %xmm1, %r8d
sub $0xffff, %r8d
@ -1389,56 +1385,56 @@ L(less16bytes):
sub %edx, %eax
ret
ALIGN (4)
.p2align 4
L(Byte16):
movzbl -16(%rdi), %eax
movzbl -16(%rsi), %edx
sub %edx, %eax
ret
ALIGN (4)
.p2align 4
L(Byte17):
movzbl -15(%rdi), %eax
movzbl -15(%rsi), %edx
sub %edx, %eax
ret
ALIGN (4)
.p2align 4
L(Byte18):
movzbl -14(%rdi), %eax
movzbl -14(%rsi), %edx
sub %edx, %eax
ret
ALIGN (4)
.p2align 4
L(Byte19):
movzbl -13(%rdi), %eax
movzbl -13(%rsi), %edx
sub %edx, %eax
ret
ALIGN (4)
.p2align 4
L(Byte20):
movzbl -12(%rdi), %eax
movzbl -12(%rsi), %edx
sub %edx, %eax
ret
ALIGN (4)
.p2align 4
L(Byte21):
movzbl -11(%rdi), %eax
movzbl -11(%rsi), %edx
sub %edx, %eax
ret
ALIGN (4)
.p2align 4
L(Byte22):
movzbl -10(%rdi), %eax
movzbl -10(%rsi), %edx
sub %edx, %eax
ret
ALIGN (4)
.p2align 4
L(next_24_bytes):
lea 8(%rdi), %rdi
lea 8(%rsi), %rsi
@ -1479,14 +1475,14 @@ L(next_24_bytes):
jne L(find_diff)
ret
ALIGN (4)
.p2align 4
L(second_double_word):
mov -12(%rdi), %eax
cmp -12(%rsi), %eax
jne L(find_diff)
ret
ALIGN (4)
.p2align 4
L(next_two_double_words):
and $15, %dh
jz L(fourth_double_word)
@ -1495,7 +1491,7 @@ L(next_two_double_words):
jne L(find_diff)
ret
ALIGN (4)
.p2align 4
L(fourth_double_word):
mov -4(%rdi), %eax
cmp -4(%rsi), %eax
@ -1503,7 +1499,7 @@ L(fourth_double_word):
ret
# endif
ALIGN (4)
.p2align 4
L(less48bytes):
cmp $8, %ecx
jae L(more8bytes)
@ -1527,7 +1523,7 @@ L(less48bytes):
jmp L(4bytes)
# endif
ALIGN (4)
.p2align 4
L(more8bytes):
cmp $16, %ecx
jae L(more16bytes)
@ -1551,7 +1547,7 @@ L(more8bytes):
jmp L(12bytes)
# endif
ALIGN (4)
.p2align 4
L(more16bytes):
cmp $24, %ecx
jae L(more24bytes)
@ -1575,7 +1571,7 @@ L(more16bytes):
jmp L(20bytes)
# endif
ALIGN (4)
.p2align 4
L(more24bytes):
cmp $32, %ecx
jae L(more32bytes)
@ -1599,7 +1595,7 @@ L(more24bytes):
jmp L(28bytes)
# endif
ALIGN (4)
.p2align 4
L(more32bytes):
cmp $40, %ecx
jae L(more40bytes)
@ -1623,7 +1619,7 @@ L(more32bytes):
jmp L(36bytes)
# endif
ALIGN (4)
.p2align 4
L(more40bytes):
cmp $40, %ecx
je L(40bytes)
@ -1642,7 +1638,7 @@ L(more40bytes):
je L(46bytes)
jmp L(47bytes)
ALIGN (4)
.p2align 4
L(44bytes):
movl -44(%rdi), %eax
movl -44(%rsi), %ecx
@ -1702,7 +1698,7 @@ L(0bytes):
xor %eax, %eax
ret
# else
ALIGN (4)
.p2align 4
L(44bytes):
movl -44(%rdi), %eax
cmp -44(%rsi), %eax
@ -1753,7 +1749,7 @@ L(0bytes):
# endif
# ifndef USE_AS_WMEMCMP
ALIGN (4)
.p2align 4
L(45bytes):
movl -45(%rdi), %eax
movl -45(%rsi), %ecx
@ -1816,7 +1812,7 @@ L(1bytes):
xor %eax, %eax
ret
ALIGN (4)
.p2align 4
L(46bytes):
movl -46(%rdi), %eax
movl -46(%rsi), %ecx
@ -1882,7 +1878,7 @@ L(2bytes):
xor %eax, %eax
ret
ALIGN (4)
.p2align 4
L(47bytes):
movl -47(%rdi), %eax
movl -47(%rsi), %ecx
@ -1951,7 +1947,7 @@ L(3bytes):
xor %eax, %eax
ret
ALIGN (4)
.p2align 4
L(find_diff):
cmpb %cl, %al
jne L(set)
@ -1973,19 +1969,19 @@ L(set):
# else
/* for wmemcmp */
ALIGN (4)
.p2align 4
L(find_diff):
mov $1, %eax
jg L(find_diff_bigger)
neg %eax
ret
ALIGN (4)
.p2align 4
L(find_diff_bigger):
ret
# endif
ALIGN (4)
.p2align 4
L(equal):
xor %eax, %eax
ret

10
sysdeps/x86_64/multiarch/memcpy-sse2-unaligned.S
View File

@ -20,10 +20,6 @@
#include "asm-syntax.h"
#ifndef ALIGN
# define ALIGN(n) .p2align n
#endif
ENTRY(__memcpy_sse2_unaligned)
movq %rsi, %rax
@ -44,7 +40,7 @@ L(return):
movq %rdi, %rax
ret
.p2align 4,,10
ALIGN(4)
.p2align 4
.L31:
movdqu 16(%rsi), %xmm8
cmpq $64, %rdx
@ -77,7 +73,7 @@ L(return):
leaq 32(%r10), %r8
leaq 48(%r10), %rax
.p2align 4,,10
ALIGN(4)
.p2align 4
L(loop):
movdqu (%rcx,%r10), %xmm8
movdqa %xmm8, (%rcx)
@ -151,7 +147,7 @@ L(less_16):
.L3:
leaq -1(%rdx), %rax
.p2align 4,,10
ALIGN(4)
.p2align 4
.L11:
movzbl (%rsi,%rax), %edx
movb %dl, (%rdi,%rax)

86
sysdeps/x86_64/multiarch/memcpy-ssse3-back.S
View File

@ -31,10 +31,6 @@
# define MEMCPY_CHK __memcpy_chk_ssse3_back
#endif
#ifndef ALIGN
# define ALIGN(n) .p2align n
#endif
#define JMPTBL(I, B) I - B
/* Branch to an entry in a jump table. TABLE is a jump table with
@ -87,7 +83,7 @@ L(bk_write):
BRANCH_TO_JMPTBL_ENTRY (L(table_144_bytes_bwd), %rdx, 4)
#endif
ALIGN (4)
.p2align 4
L(144bytesormore):
#ifndef USE_AS_MEMMOVE
@ -119,7 +115,7 @@ L(144bytesormore):
jmp *%r9
ud2
ALIGN (4)
.p2align 4
L(copy_backward):
#ifdef DATA_CACHE_SIZE
mov $DATA_CACHE_SIZE, %RCX_LP
@ -149,7 +145,7 @@ L(copy_backward):
jmp *%r9
ud2
ALIGN (4)
.p2align 4
L(shl_0):
mov %rdx, %r9
@ -162,7 +158,7 @@ L(shl_0):
#endif
jae L(gobble_mem_fwd)
sub $0x80, %rdx
ALIGN (4)
.p2align 4
L(shl_0_loop):
movdqa (%rsi), %xmm1
movdqa %xmm1, (%rdi)
@ -190,7 +186,7 @@ L(shl_0_loop):
add %rdx, %rdi
BRANCH_TO_JMPTBL_ENTRY (L(table_144_bytes_fwd), %rdx, 4)
ALIGN (4)
.p2align 4
L(shl_0_bwd):
sub $0x80, %rdx
L(copy_backward_loop):
@ -221,7 +217,7 @@ L(copy_backward_loop):
sub %rdx, %rsi
BRANCH_TO_JMPTBL_ENTRY (L(table_144_bytes_bwd), %rdx, 4)
ALIGN (4)
.p2align 4
L(shl_1):
sub $0x80, %rdx
movaps -0x01(%rsi), %xmm1
@ -258,7 +254,7 @@ L(shl_1):
add %rdx, %rsi
BRANCH_TO_JMPTBL_ENTRY (L(table_144_bytes_fwd), %rdx, 4)
ALIGN (4)
.p2align 4
L(shl_1_bwd):
movaps -0x01(%rsi), %xmm1
@ -304,7 +300,7 @@ L(shl_1_bwd):
sub %rdx, %rsi
BRANCH_TO_JMPTBL_ENTRY (L(table_144_bytes_bwd), %rdx, 4)
ALIGN (4)
.p2align 4
L(shl_2):
sub $0x80, %rdx
movaps -0x02(%rsi), %xmm1
@ -341,7 +337,7 @@ L(shl_2):
add %rdx, %rsi
BRANCH_TO_JMPTBL_ENTRY (L(table_144_bytes_fwd), %rdx, 4)
ALIGN (4)
.p2align 4
L(shl_2_bwd):
movaps -0x02(%rsi), %xmm1
@ -387,7 +383,7 @@ L(shl_2_bwd):
sub %rdx, %rsi
BRANCH_TO_JMPTBL_ENTRY (L(table_144_bytes_bwd), %rdx, 4)
ALIGN (4)
.p2align 4
L(shl_3):
sub $0x80, %rdx
movaps -0x03(%rsi), %xmm1
@ -424,7 +420,7 @@ L(shl_3):
add %rdx, %rsi
BRANCH_TO_JMPTBL_ENTRY (L(table_144_bytes_fwd), %rdx, 4)
ALIGN (4)
.p2align 4
L(shl_3_bwd):
movaps -0x03(%rsi), %xmm1
@ -470,7 +466,7 @@ L(shl_3_bwd):
sub %rdx, %rsi
BRANCH_TO_JMPTBL_ENTRY (L(table_144_bytes_bwd), %rdx, 4)
ALIGN (4)
.p2align 4
L(shl_4):
sub $0x80, %rdx
movaps -0x04(%rsi), %xmm1
@ -507,7 +503,7 @@ L(shl_4):
add %rdx, %rsi
BRANCH_TO_JMPTBL_ENTRY (L(table_144_bytes_fwd), %rdx, 4)
ALIGN (4)
.p2align 4
L(shl_4_bwd):
movaps -0x04(%rsi), %xmm1
@ -553,7 +549,7 @@ L(shl_4_bwd):
sub %rdx, %rsi
BRANCH_TO_JMPTBL_ENTRY (L(table_144_bytes_bwd), %rdx, 4)
ALIGN (4)
.p2align 4
L(shl_5):
sub $0x80, %rdx
movaps -0x05(%rsi), %xmm1
@ -590,7 +586,7 @@ L(shl_5):
add %rdx, %rsi
BRANCH_TO_JMPTBL_ENTRY (L(table_144_bytes_fwd), %rdx, 4)
ALIGN (4)
.p2align 4
L(shl_5_bwd):
movaps -0x05(%rsi), %xmm1
@ -636,7 +632,7 @@ L(shl_5_bwd):
sub %rdx, %rsi
BRANCH_TO_JMPTBL_ENTRY (L(table_144_bytes_bwd), %rdx, 4)
ALIGN (4)
.p2align 4
L(shl_6):
sub $0x80, %rdx
movaps -0x06(%rsi), %xmm1
@ -673,7 +669,7 @@ L(shl_6):
add %rdx, %rsi
BRANCH_TO_JMPTBL_ENTRY (L(table_144_bytes_fwd), %rdx, 4)
ALIGN (4)
.p2align 4
L(shl_6_bwd):
movaps -0x06(%rsi), %xmm1
@ -719,7 +715,7 @@ L(shl_6_bwd):
sub %rdx, %rsi
BRANCH_TO_JMPTBL_ENTRY (L(table_144_bytes_bwd), %rdx, 4)
ALIGN (4)
.p2align 4
L(shl_7):
sub $0x80, %rdx
movaps -0x07(%rsi), %xmm1
@ -756,7 +752,7 @@ L(shl_7):
add %rdx, %rsi
BRANCH_TO_JMPTBL_ENTRY (L(table_144_bytes_fwd), %rdx, 4)
ALIGN (4)
.p2align 4
L(shl_7_bwd):
movaps -0x07(%rsi), %xmm1
@ -802,7 +798,7 @@ L(shl_7_bwd):
sub %rdx, %rsi
BRANCH_TO_JMPTBL_ENTRY (L(table_144_bytes_bwd), %rdx, 4)
ALIGN (4)
.p2align 4
L(shl_8):
sub $0x80, %rdx
movaps -0x08(%rsi), %xmm1
@ -839,7 +835,7 @@ L(shl_8):
add %rdx, %rsi
BRANCH_TO_JMPTBL_ENTRY (L(table_144_bytes_fwd), %rdx, 4)
ALIGN (4)
.p2align 4
L(shl_8_bwd):
movaps -0x08(%rsi), %xmm1
@ -886,7 +882,7 @@ L(shl_8_end_bwd):
sub %rdx, %rsi
BRANCH_TO_JMPTBL_ENTRY (L(table_144_bytes_bwd), %rdx, 4)
ALIGN (4)
.p2align 4
L(shl_9):
sub $0x80, %rdx
movaps -0x09(%rsi), %xmm1
@ -923,7 +919,7 @@ L(shl_9):
add %rdx, %rsi
BRANCH_TO_JMPTBL_ENTRY (L(table_144_bytes_fwd), %rdx, 4)
ALIGN (4)
.p2align 4
L(shl_9_bwd):
movaps -0x09(%rsi), %xmm1
@ -969,7 +965,7 @@ L(shl_9_bwd):
sub %rdx, %rsi
BRANCH_TO_JMPTBL_ENTRY (L(table_144_bytes_bwd), %rdx, 4)
ALIGN (4)
.p2align 4
L(shl_10):
sub $0x80, %rdx
movaps -0x0a(%rsi), %xmm1
@ -1006,7 +1002,7 @@ L(shl_10):
add %rdx, %rsi
BRANCH_TO_JMPTBL_ENTRY (L(table_144_bytes_fwd), %rdx, 4)
ALIGN (4)
.p2align 4
L(shl_10_bwd):
movaps -0x0a(%rsi), %xmm1
@ -1052,7 +1048,7 @@ L(shl_10_bwd):
sub %rdx, %rsi
BRANCH_TO_JMPTBL_ENTRY (L(table_144_bytes_bwd), %rdx, 4)
ALIGN (4)
.p2align 4
L(shl_11):
sub $0x80, %rdx
movaps -0x0b(%rsi), %xmm1
@ -1089,7 +1085,7 @@ L(shl_11):
add %rdx, %rsi
BRANCH_TO_JMPTBL_ENTRY (L(table_144_bytes_fwd), %rdx, 4)
ALIGN (4)
.p2align 4
L(shl_11_bwd):
movaps -0x0b(%rsi), %xmm1
@ -1135,7 +1131,7 @@ L(shl_11_bwd):
sub %rdx, %rsi
BRANCH_TO_JMPTBL_ENTRY (L(table_144_bytes_bwd), %rdx, 4)
ALIGN (4)
.p2align 4
L(shl_12):
sub $0x80, %rdx
movdqa -0x0c(%rsi), %xmm1
@ -1173,7 +1169,7 @@ L(shl_12):
add %rdx, %rsi
BRANCH_TO_JMPTBL_ENTRY (L(table_144_bytes_fwd), %rdx, 4)
ALIGN (4)
.p2align 4
L(shl_12_bwd):
movaps -0x0c(%rsi), %xmm1
@ -1219,7 +1215,7 @@ L(shl_12_bwd):
sub %rdx, %rsi
BRANCH_TO_JMPTBL_ENTRY (L(table_144_bytes_bwd), %rdx, 4)
ALIGN (4)
.p2align 4
L(shl_13):
sub $0x80, %rdx
movaps -0x0d(%rsi), %xmm1
@ -1256,7 +1252,7 @@ L(shl_13):
add %rdx, %rsi
BRANCH_TO_JMPTBL_ENTRY (L(table_144_bytes_fwd), %rdx, 4)
ALIGN (4)
.p2align 4
L(shl_13_bwd):
movaps -0x0d(%rsi), %xmm1
@ -1302,7 +1298,7 @@ L(shl_13_bwd):
sub %rdx, %rsi
BRANCH_TO_JMPTBL_ENTRY (L(table_144_bytes_bwd), %rdx, 4)
ALIGN (4)
.p2align 4
L(shl_14):
sub $0x80, %rdx
movaps -0x0e(%rsi), %xmm1
@ -1339,7 +1335,7 @@ L(shl_14):
add %rdx, %rsi
BRANCH_TO_JMPTBL_ENTRY (L(table_144_bytes_fwd), %rdx, 4)
ALIGN (4)
.p2align 4
L(shl_14_bwd):
movaps -0x0e(%rsi), %xmm1
@ -1385,7 +1381,7 @@ L(shl_14_bwd):
sub %rdx, %rsi
BRANCH_TO_JMPTBL_ENTRY (L(table_144_bytes_bwd), %rdx, 4)
ALIGN (4)
.p2align 4
L(shl_15):
sub $0x80, %rdx
movaps -0x0f(%rsi), %xmm1
@ -1422,7 +1418,7 @@ L(shl_15):
add %rdx, %rsi
BRANCH_TO_JMPTBL_ENTRY (L(table_144_bytes_fwd), %rdx, 4)
ALIGN (4)
.p2align 4
L(shl_15_bwd):
movaps -0x0f(%rsi), %xmm1
@ -1468,7 +1464,7 @@ L(shl_15_bwd):
sub %rdx, %rsi
BRANCH_TO_JMPTBL_ENTRY (L(table_144_bytes_bwd), %rdx, 4)
ALIGN (4)
.p2align 4
L(gobble_mem_fwd):
movdqu (%rsi), %xmm1
movdqu %xmm0, (%r8)
@ -1570,7 +1566,7 @@ L(gobble_mem_fwd_end):
add %rdx, %rdi
BRANCH_TO_JMPTBL_ENTRY (L(table_144_bytes_fwd), %rdx, 4)
ALIGN (4)
.p2align 4
L(gobble_mem_bwd):
add %rdx, %rsi
add %rdx, %rdi
@ -2833,7 +2829,7 @@ L(bwd_write_1bytes):
END (MEMCPY)
.section .rodata.ssse3,"a",@progbits
ALIGN (3)
.p2align 3
L(table_144_bytes_bwd):
.int JMPTBL (L(bwd_write_0bytes), L(table_144_bytes_bwd))
.int JMPTBL (L(bwd_write_1bytes), L(table_144_bytes_bwd))
@ -2980,7 +2976,7 @@ L(table_144_bytes_bwd):
.int JMPTBL (L(bwd_write_142bytes), L(table_144_bytes_bwd))
.int JMPTBL (L(bwd_write_143bytes), L(table_144_bytes_bwd))
ALIGN (3)
.p2align 3
L(table_144_bytes_fwd):
.int JMPTBL (L(fwd_write_0bytes), L(table_144_bytes_fwd))
.int JMPTBL (L(fwd_write_1bytes), L(table_144_bytes_fwd))
@ -3127,7 +3123,7 @@ L(table_144_bytes_fwd):
.int JMPTBL (L(fwd_write_142bytes), L(table_144_bytes_fwd))
.int JMPTBL (L(fwd_write_143bytes), L(table_144_bytes_fwd))
ALIGN (3)
.p2align 3
L(shl_table_fwd):
.int JMPTBL (L(shl_0), L(shl_table_fwd))
.int JMPTBL (L(shl_1), L(shl_table_fwd))
@ -3146,7 +3142,7 @@ L(shl_table_fwd):
.int JMPTBL (L(shl_14), L(shl_table_fwd))
.int JMPTBL (L(shl_15), L(shl_table_fwd))
ALIGN (3)
.p2align 3
L(shl_table_bwd):
.int JMPTBL (L(shl_0_bwd), L(shl_table_bwd))
.int JMPTBL (L(shl_1_bwd), L(shl_table_bwd))

254
sysdeps/x86_64/multiarch/memcpy-ssse3.S
View File

File diff suppressed because it is too large Load Diff

9
sysdeps/x86_64/multiarch/strcmp-sse2-unaligned.S
View File

@ -17,7 +17,6 @@
<http://www.gnu.org/licenses/>. */
#include "sysdep.h"
#define ALIGN(x) .p2align x
ENTRY ( __strcmp_sse2_unaligned)
movl %edi, %eax
@ -43,7 +42,7 @@ L(return):
subl %edx, %eax
ret
ALIGN (4)
.p2align 4
L(next_48_bytes):
movdqu 16(%rdi), %xmm6
movdqu 16(%rsi), %xmm3
@ -85,7 +84,7 @@ L(main_loop_header):
movq %rcx, %rsi
jmp L(loop_start)
ALIGN (4)
.p2align 4
L(loop):
addq $64, %rax
addq $64, %rdx
@ -141,7 +140,7 @@ L(back_to_loop):
subl %edx, %eax
ret
ALIGN (4)
.p2align 4
L(loop_cross_page):
xor %r10, %r10
movq %rdx, %r9
@ -191,7 +190,7 @@ L(loop_cross_page):
subl %edx, %eax
ret
ALIGN (4)
.p2align 4
L(cross_page_loop):
cmpb %cl, %al
jne L(different)

15
sysdeps/x86_64/strchr.S
View File

@ -19,11 +19,6 @@
#include <sysdep.h>
# ifndef ALIGN
# define ALIGN(n) .p2align n
# endif
.text
ENTRY (strchr)
movd %esi, %xmm1
@ -54,7 +49,7 @@ ENTRY (strchr)
#endif
ret
ALIGN(3)
.p2align 3
L(next_48_bytes):
movdqu 16(%rdi), %xmm0
movdqa %xmm0, %xmm4
@ -83,10 +78,10 @@ ENTRY (strchr)
L(loop_start):
/* We use this alignment to force loop be aligned to 8 but not
16 bytes. This gives better sheduling on AMD processors. */
ALIGN(4)
.p2align 4
pxor %xmm6, %xmm6
andq $-64, %rdi
ALIGN(3)
.p2align 3
L(loop64):
addq $64, %rdi
movdqa (%rdi), %xmm5
@ -129,7 +124,7 @@ L(loop64):
orq %rcx, %rax
salq $48, %rdx
orq %rdx, %rax
ALIGN(3)
.p2align 3
L(return):
bsfq %rax, %rax
#ifdef AS_STRCHRNUL
@ -141,7 +136,7 @@ L(return):
cmovne %rdx, %rax
#endif
ret
ALIGN(4)
.p2align 4
L(cross_page):
movq %rdi, %rdx

17
sysdeps/x86_64/strrchr.S
View File

@ -19,11 +19,6 @@
#include <sysdep.h>
# ifndef ALIGN
# define ALIGN(n) .p2align n
# endif
.text
ENTRY (strrchr)
movd %esi, %xmm1
@ -51,7 +46,7 @@ ENTRY (strrchr)
addq %rdi, %rax
ret
ALIGN(4)
.p2align 4
L(next_48_bytes):
movdqu 16(%rdi), %xmm4
movdqa %xmm4, %xmm5
@ -91,7 +86,7 @@ L(next_48_bytes):
leaq (%rdi,%rsi), %rax
ret
ALIGN(4)
.p2align 4
L(loop_header2):
testq %rsi, %rsi
movq %rdi, %rcx
@ -102,7 +97,7 @@ L(loop_header):
andq $-64, %rdi
jmp L(loop_entry)
ALIGN(4)
.p2align 4
L(loop64):
testq %rdx, %rdx
cmovne %rdx, %rsi
@ -163,18 +158,18 @@ L(loop_entry):
leaq (%rcx,%rsi), %rax
ret
ALIGN(4)
.p2align 4
L(no_c_found):
movl $1, %esi
xorl %ecx, %ecx
jmp L(loop_header)
ALIGN(4)
.p2align 4
L(exit):
xorl %eax, %eax
ret
ALIGN(4)
.p2align 4
L(cross_page):
movq %rdi, %rax
pxor %xmm0, %xmm0