diff options
Diffstat (limited to 'libc/sysdeps/x86_64/multiarch/memcmp-ssse3.S')
| -rw-r--r-- | libc/sysdeps/x86_64/multiarch/memcmp-ssse3.S | 126 |
1 files changed, 61 insertions, 65 deletions
diff --git a/libc/sysdeps/x86_64/multiarch/memcmp-ssse3.S b/libc/sysdeps/x86_64/multiarch/memcmp-ssse3.S index e319df926..e04f918df 100644 --- a/libc/sysdeps/x86_64/multiarch/memcmp-ssse3.S +++ b/libc/sysdeps/x86_64/multiarch/memcmp-ssse3.S @@ -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 |