sysdeps/x86_64/multiarch/memcpy-sse2-unaligned.S

   1 /* memcpy with unaliged loads
   2    Copyright (C) 2013-2015 Free Software Foundation, Inc.
   3    This file is part of the GNU C Library.
   4
   5    The GNU C Library is free software; you can redistribute it and/or
   6    modify it under the terms of the GNU Lesser General Public
   7    License as published by the Free Software Foundation; either
   8    version 2.1 of the License, or (at your option) any later version.
   9
  10    The GNU C Library is distributed in the hope that it will be useful,
  11    but WITHOUT ANY WARRANTY; without even the implied warranty of
  12    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  13    Lesser General Public License for more details.
  14
  15    You should have received a copy of the GNU Lesser General Public
  16    License along with the GNU C Library; if not, see
  17    <http://www.gnu.org/licenses/>.  */
  18
  19 #include <sysdep.h>
  20
  21 #include "asm-syntax.h"
  22
  23
  24 ENTRY(__memcpy_sse2_unaligned)
  25         movq    %rsi, %rax
  26         leaq    (%rdx,%rdx), %rcx
  27         subq    %rdi, %rax
  28         subq    %rdx, %rax
  29         cmpq    %rcx, %rax
  30         jb      L(overlapping)
  31         cmpq    $16, %rdx
  32         jbe     L(less_16)
  33         movdqu  (%rsi), %xmm8
  34         cmpq    $32, %rdx
  35         movdqu  %xmm8, (%rdi)
  36         movdqu  -16(%rsi,%rdx), %xmm8
  37         movdqu  %xmm8, -16(%rdi,%rdx)
  38         ja      .L31
  39 L(return):
  40         movq    %rdi, %rax
  41         ret
  42         .p2align 4,,10
  43         .p2align 4
  44 .L31:
  45         movdqu  16(%rsi), %xmm8
  46         cmpq    $64, %rdx
  47         movdqu  %xmm8, 16(%rdi)
  48         movdqu  -32(%rsi,%rdx), %xmm8
  49         movdqu  %xmm8, -32(%rdi,%rdx)
  50         jbe     L(return)
  51         movdqu  32(%rsi), %xmm8
  52         cmpq    $128, %rdx
  53         movdqu  %xmm8, 32(%rdi)
  54         movdqu  -48(%rsi,%rdx), %xmm8
  55         movdqu  %xmm8, -48(%rdi,%rdx)
  56         movdqu  48(%rsi), %xmm8
  57         movdqu  %xmm8, 48(%rdi)
  58         movdqu  -64(%rsi,%rdx), %xmm8
  59         movdqu  %xmm8, -64(%rdi,%rdx)
  60         jbe     L(return)
  61         leaq    64(%rdi), %rcx
  62         addq    %rdi, %rdx
  63         andq    $-64, %rdx
  64         andq    $-64, %rcx
  65         movq    %rcx, %rax
  66         subq    %rdi, %rax
  67         addq    %rax, %rsi
  68         cmpq    %rdx, %rcx
  69         je      L(return)
  70         movq    %rsi, %r10
  71         subq    %rcx, %r10
  72         leaq    16(%r10), %r9
  73         leaq    32(%r10), %r8
  74         leaq    48(%r10), %rax
  75         .p2align 4,,10
  76         .p2align 4
  77 L(loop):
  78         movdqu  (%rcx,%r10), %xmm8
  79         movdqa  %xmm8, (%rcx)
  80         movdqu  (%rcx,%r9), %xmm8
  81         movdqa  %xmm8, 16(%rcx)
  82         movdqu  (%rcx,%r8), %xmm8
  83         movdqa  %xmm8, 32(%rcx)
  84         movdqu  (%rcx,%rax), %xmm8
  85         movdqa  %xmm8, 48(%rcx)
  86         addq    $64, %rcx
  87         cmpq    %rcx, %rdx
  88         jne     L(loop)
  89         jmp     L(return)
  90 L(overlapping):
  91         cmpq    %rsi, %rdi
  92         jae     .L3
  93         testq   %rdx, %rdx
  94         .p2align 4,,5
  95         je      L(return)
  96         movq    %rdx, %r9
  97         leaq    16(%rsi), %rcx
  98         leaq    16(%rdi), %r8
  99         shrq    $4, %r9
 100         movq    %r9, %rax
 101         salq    $4, %rax
 102         cmpq    %rcx, %rdi
 103         setae   %cl
 104         cmpq    %r8, %rsi
 105         setae   %r8b
 106         orl     %r8d, %ecx
 107         cmpq    $15, %rdx
 108         seta    %r8b
 109         testb   %r8b, %cl
 110         je      .L16
 111         testq   %rax, %rax
 112         je      .L16
 113         xorl    %ecx, %ecx
 114         xorl    %r8d, %r8d
 115 .L7:
 116         movdqu  (%rsi,%rcx), %xmm8
 117         addq    $1, %r8
 118         movdqu  %xmm8, (%rdi,%rcx)
 119         addq    $16, %rcx
 120         cmpq    %r8, %r9
 121         ja      .L7
 122         cmpq    %rax, %rdx
 123         je      L(return)
 124 .L21:
 125         movzbl  (%rsi,%rax), %ecx
 126         movb    %cl, (%rdi,%rax)
 127         addq    $1, %rax
 128         cmpq    %rax, %rdx
 129         ja      .L21
 130         jmp     L(return)
 131 L(less_16):
 132         testb   $24, %dl
 133         jne     L(between_9_16)
 134         testb   $4, %dl
 135         .p2align 4,,5
 136         jne     L(between_5_8)
 137         testq   %rdx, %rdx
 138         .p2align 4,,2
 139         je      L(return)
 140         movzbl  (%rsi), %eax
 141         testb   $2, %dl
 142         movb    %al, (%rdi)
 143         je      L(return)
 144         movzwl  -2(%rsi,%rdx), %eax
 145         movw    %ax, -2(%rdi,%rdx)
 146         jmp     L(return)
 147 .L3:
 148         leaq    -1(%rdx), %rax
 149         .p2align 4,,10
 150         .p2align 4
 151 .L11:
 152         movzbl  (%rsi,%rax), %edx
 153         movb    %dl, (%rdi,%rax)
 154         subq    $1, %rax
 155         jmp     .L11
 156 L(between_9_16):
 157         movq    (%rsi), %rax
 158         movq    %rax, (%rdi)
 159         movq    -8(%rsi,%rdx), %rax
 160         movq    %rax, -8(%rdi,%rdx)
 161         jmp     L(return)
 162 .L16:
 163         xorl    %eax, %eax
 164         jmp     .L21
 165 L(between_5_8):
 166         movl    (%rsi), %eax
 167         movl    %eax, (%rdi)
 168         movl    -4(%rsi,%rdx), %eax
 169         movl    %eax, -4(%rdi,%rdx)
 170         jmp     L(return)
 171 END(__memcpy_sse2_unaligned)