diff options
Diffstat (limited to 'sysdeps/x86_64/memset.S')
| -rw-r--r-- | sysdeps/x86_64/memset.S | 137 |
1 files changed, 33 insertions, 104 deletions
diff --git a/sysdeps/x86_64/memset.S b/sysdeps/x86_64/memset.S index 4cf0da0fb8..b342679576 100644 --- a/sysdeps/x86_64/memset.S +++ b/sysdeps/x86_64/memset.S @@ -1,6 +1,6 @@ /* memset/bzero -- set memory area to CH/0 Optimized version for x86-64. - Copyright (C) 2002-2016 Free Software Foundation, Inc. + Copyright (C) 2002-2018 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 @@ -19,114 +19,43 @@ #include <sysdep.h> - .text -#if IS_IN (libc) -ENTRY(__bzero) - movq %rdi, %rax /* Set return value. */ - movq %rsi, %rdx /* Set n. */ - pxor %xmm0, %xmm0 - jmp L(entry_from_bzero) -END(__bzero) -weak_alias (__bzero, bzero) - -/* Like memset but takes additional parameter with return value. */ -ENTRY(__memset_tail) - movq %rcx, %rax /* Set return value. */ - - movd %esi, %xmm0 - punpcklbw %xmm0, %xmm0 - punpcklwd %xmm0, %xmm0 - pshufd $0, %xmm0, %xmm0 - - jmp L(entry_from_bzero) -END(__memset_tail) +#define VEC_SIZE 16 +#define VEC(i) xmm##i +/* Don't use movups and movaps since it will get larger nop paddings for + alignment. */ +#define VMOVU movdqu +#define VMOVA movdqa + +#define MEMSET_VDUP_TO_VEC0_AND_SET_RETURN(d, r) \ + movd d, %xmm0; \ + movq r, %rax; \ + punpcklbw %xmm0, %xmm0; \ + punpcklwd %xmm0, %xmm0; \ + pshufd $0, %xmm0, %xmm0 + +#define WMEMSET_VDUP_TO_VEC0_AND_SET_RETURN(d, r) \ + movd d, %xmm0; \ + movq r, %rax; \ + pshufd $0, %xmm0, %xmm0 + +#define SECTION(p) p + +#ifndef MEMSET_SYMBOL +# define MEMSET_CHK_SYMBOL(p,s) p +# define MEMSET_SYMBOL(p,s) memset #endif -#if defined PIC && IS_IN (libc) -ENTRY_CHK (__memset_chk) - cmpq %rdx, %rcx - jb HIDDEN_JUMPTARGET (__chk_fail) -END_CHK (__memset_chk) +#ifndef WMEMSET_SYMBOL +# define WMEMSET_CHK_SYMBOL(p,s) p +# define WMEMSET_SYMBOL(p,s) __wmemset #endif -ENTRY (memset) - movd %esi, %xmm0 - movq %rdi, %rax - punpcklbw %xmm0, %xmm0 - punpcklwd %xmm0, %xmm0 - pshufd $0, %xmm0, %xmm0 -L(entry_from_bzero): - cmpq $64, %rdx - ja L(loop_start) - cmpq $16, %rdx - jbe L(less_16_bytes) - cmpq $32, %rdx - movdqu %xmm0, (%rdi) - movdqu %xmm0, -16(%rdi,%rdx) - ja L(between_32_64_bytes) -L(return): - rep - ret - .p2align 4 -L(between_32_64_bytes): - movdqu %xmm0, 16(%rdi) - movdqu %xmm0, -32(%rdi,%rdx) - ret - .p2align 4 -L(loop_start): - leaq 64(%rdi), %rcx - movdqu %xmm0, (%rdi) - andq $-64, %rcx - movdqu %xmm0, -16(%rdi,%rdx) - movdqu %xmm0, 16(%rdi) - movdqu %xmm0, -32(%rdi,%rdx) - movdqu %xmm0, 32(%rdi) - movdqu %xmm0, -48(%rdi,%rdx) - movdqu %xmm0, 48(%rdi) - movdqu %xmm0, -64(%rdi,%rdx) - addq %rdi, %rdx - andq $-64, %rdx - cmpq %rdx, %rcx - je L(return) - .p2align 4 -L(loop): - movdqa %xmm0, (%rcx) - movdqa %xmm0, 16(%rcx) - movdqa %xmm0, 32(%rcx) - movdqa %xmm0, 48(%rcx) - addq $64, %rcx - cmpq %rcx, %rdx - jne L(loop) - rep - ret -L(less_16_bytes): - movq %xmm0, %rcx - testb $24, %dl - jne L(between8_16bytes) - testb $4, %dl - jne L(between4_7bytes) - testb $1, %dl - je L(odd_byte) - movb %cl, (%rdi) -L(odd_byte): - testb $2, %dl - je L(return) - movw %cx, -2(%rax,%rdx) - ret -L(between4_7bytes): - movl %ecx, (%rdi) - movl %ecx, -4(%rdi,%rdx) - ret -L(between8_16bytes): - movq %rcx, (%rdi) - movq %rcx, -8(%rdi,%rdx) - ret +#include "multiarch/memset-vec-unaligned-erms.S" -END (memset) libc_hidden_builtin_def (memset) -#if defined PIC && IS_IN (libc) && !defined USE_MULTIARCH -strong_alias (__memset_chk, __memset_zero_constant_len_parameter) - .section .gnu.warning.__memset_zero_constant_len_parameter - .string "memset used with constant zero length parameter; this could be due to transposed parameters" +#if IS_IN (libc) +libc_hidden_def (__wmemset) +weak_alias (__wmemset, wmemset) +libc_hidden_weak (wmemset) #endif |