diff options
Diffstat (limited to 'sysdeps/x86_64/memcmp.S')
| -rw-r--r-- | sysdeps/x86_64/memcmp.S | 84 |
1 files changed, 42 insertions, 42 deletions
diff --git a/sysdeps/x86_64/memcmp.S b/sysdeps/x86_64/memcmp.S index 165f42e17d..a9fe13ae58 100644 --- a/sysdeps/x86_64/memcmp.S +++ b/sysdeps/x86_64/memcmp.S @@ -26,7 +26,7 @@ ENTRY (memcmp) jz L(finz) cmpq $1, %rdx jle L(finr1b) - subq %rdi, %rsi + subq %rdi, %rsi movq %rdx, %r10 cmpq $32, %r10 jge L(gt32) @@ -37,7 +37,7 @@ L(small): movzbl (%rdi), %eax movzbl (%rdi, %rsi), %edx subq $1, %r10 - je L(finz1) + je L(finz1) addq $1, %rdi subl %edx, %eax jnz L(exit) @@ -47,7 +47,7 @@ L(s2b): movzwl (%rdi), %eax movzwl (%rdi, %rsi), %edx subq $2, %r10 - je L(fin2_7) + je L(fin2_7) addq $2, %rdi cmpl %edx, %eax jnz L(fin2_7) @@ -57,7 +57,7 @@ L(s4b): movl (%rdi), %eax movl (%rdi, %rsi), %edx subq $4, %r10 - je L(fin2_7) + je L(fin2_7) addq $4, %rdi cmpl %edx, %eax jnz L(fin2_7) @@ -67,7 +67,7 @@ L(s8b): movq (%rdi), %rax movq (%rdi, %rsi), %rdx subq $8, %r10 - je L(fin2_7) + je L(fin2_7) addq $8, %rdi cmpq %rdx, %rax jnz L(fin2_7) @@ -76,11 +76,11 @@ L(s16b): movdqu (%rdi, %rsi), %xmm0 pcmpeqb %xmm0, %xmm1 pmovmskb %xmm1, %edx - xorl %eax, %eax + xorl %eax, %eax subl $0xffff, %edx jz L(finz) - bsfl %edx, %ecx - leaq (%rdi, %rcx), %rcx + bsfl %edx, %ecx + leaq (%rdi, %rcx), %rcx movzbl (%rcx), %eax movzbl (%rsi, %rcx), %edx jmp L(finz1) @@ -88,7 +88,7 @@ L(s16b): .p2align 4,, 4 L(finr1b): movzbl (%rdi), %eax - movzbl (%rsi), %edx + movzbl (%rsi), %edx L(finz1): subl %edx, %eax L(exit): @@ -98,24 +98,24 @@ L(exit): L(fin2_7): cmpq %rdx, %rax jz L(finz) - movq %rax, %r11 - subq %rdx, %r11 + movq %rax, %r11 + subq %rdx, %r11 bsfq %r11, %rcx - sarq $3, %rcx + sarq $3, %rcx salq $3, %rcx - sarq %cl, %rax + sarq %cl, %rax movzbl %al, %eax - sarq %cl, %rdx + sarq %cl, %rdx movzbl %dl, %edx subl %edx, %eax - ret + ret .p2align 4,, 4 L(finz): xorl %eax, %eax ret - /* For blocks bigger than 32 bytes + /* For blocks bigger than 32 bytes 1. Advance one of the addr pointer to be 16B aligned. 2. Treat the case of both addr pointers aligned to 16B separately to avoid movdqu. @@ -128,10 +128,10 @@ L(finz): L(gt32): movq %rdx, %r11 addq %rdi, %r11 - movq %rdi, %r8 + movq %rdi, %r8 andq $15, %r8 - jz L(16am) + jz L(16am) /* Both pointers may be misaligned. */ movdqu (%rdi), %xmm1 movdqu (%rdi, %rsi), %xmm0 @@ -156,8 +156,8 @@ L(16am): L(A32): movq %r11, %r10 andq $-32, %r10 - cmpq %r10, %rdi - jge L(mt16) + cmpq %r10, %rdi + jge L(mt16) /* Pre-unroll to be ready for unrolled 64B loop. */ testq $32, %rdi jz L(A64) @@ -167,7 +167,7 @@ L(A32): subl $0xffff, %edx jnz L(neq) addq $16, %rdi - + movdqu (%rdi,%rsi), %xmm0 pcmpeqb (%rdi), %xmm0 pmovmskb %xmm0, %edx @@ -178,9 +178,9 @@ L(A32): L(A64): movq %r11, %r10 andq $-64, %r10 - cmpq %r10, %rdi - jge L(mt32) - + cmpq %r10, %rdi + jge L(mt32) + L(A64main): movdqu (%rdi,%rsi), %xmm0 pcmpeqb (%rdi), %xmm0 @@ -188,7 +188,7 @@ L(A64main): subl $0xffff, %edx jnz L(neq) addq $16, %rdi - + movdqu (%rdi,%rsi), %xmm0 pcmpeqb (%rdi), %xmm0 pmovmskb %xmm0, %edx @@ -216,8 +216,8 @@ L(A64main): L(mt32): movq %r11, %r10 andq $-32, %r10 - cmpq %r10, %rdi - jge L(mt16) + cmpq %r10, %rdi + jge L(mt16) L(A32main): movdqu (%rdi,%rsi), %xmm0 @@ -226,7 +226,7 @@ L(A32main): subl $0xffff, %edx jnz L(neq) addq $16, %rdi - + movdqu (%rdi,%rsi), %xmm0 pcmpeqb (%rdi), %xmm0 pmovmskb %xmm0, %edx @@ -239,23 +239,23 @@ L(A32main): L(mt16): subq %rdi, %r11 je L(finz) - movq %r11, %r10 - jmp L(small) + movq %r11, %r10 + jmp L(small) .p2align 4,, 4 L(neq): - bsfl %edx, %ecx + bsfl %edx, %ecx movzbl (%rdi, %rcx), %eax - addq %rdi, %rsi + addq %rdi, %rsi movzbl (%rsi,%rcx), %edx jmp L(finz1) .p2align 4,, 4 L(ATR): movq %r11, %r10 - andq $-32, %r10 - cmpq %r10, %rdi - jge L(mt16) + andq $-32, %r10 + cmpq %r10, %rdi + jge L(mt16) testq $16, %rdi jz L(ATR32) @@ -290,7 +290,7 @@ L(ATR32): L(ATR64): cmpq %rdi, %r10 - je L(mt32) + je L(mt32) L(ATR64main): movdqa (%rdi,%rsi), %xmm0 @@ -324,9 +324,9 @@ L(ATR64main): jne L(ATR64main) movq %r11, %r10 - andq $-32, %r10 - cmpq %r10, %rdi - jge L(mt16) + andq $-32, %r10 + cmpq %r10, %rdi + jge L(mt16) L(ATR32res): movdqa (%rdi,%rsi), %xmm0 @@ -343,13 +343,13 @@ L(ATR32res): jnz L(neq) addq $16, %rdi - cmpq %r10, %rdi + cmpq %r10, %rdi jne L(ATR32res) subq %rdi, %r11 je L(finz) - movq %r11, %r10 - jmp L(small) + movq %r11, %r10 + jmp L(small) /* Align to 16byte to improve instruction fetch. */ .p2align 4,, 4 END(memcmp) |