arch/x86/lib/copy_page_64.S

   1 /* Written 2003 by Andi Kleen, based on a kernel by Evandro Menezes */
   2
   3 #include <linux/linkage.h>
   4 #include <asm/dwarf2.h>
   5 #include <asm/cpufeature.h>
   6 #include <asm/alternative-asm.h>
   7
   8 /*
   9  * Some CPUs run faster using the string copy instructions (sane microcode).
  10  * It is also a lot simpler. Use this when possible. But, don't use streaming
  11  * copy unless the CPU indicates X86_FEATURE_REP_GOOD. Could vary the
  12  * prefetch distance based on SMP/UP.
  13  */
  14         ALIGN
  15 ENTRY(copy_page)
  16         CFI_STARTPROC
  17         ALTERNATIVE "jmp copy_page_regs", "", X86_FEATURE_REP_GOOD
  18         movl    $4096/8, %ecx
  19         rep     movsq
  20         ret
  21         CFI_ENDPROC
  22 ENDPROC(copy_page)
  23
  24 ENTRY(copy_page_regs)
  25         CFI_STARTPROC
  26         subq    $2*8,   %rsp
  27         CFI_ADJUST_CFA_OFFSET 2*8
  28         movq    %rbx,   (%rsp)
  29         CFI_REL_OFFSET rbx, 0
  30         movq    %r12,   1*8(%rsp)
  31         CFI_REL_OFFSET r12, 1*8
  32
  33         movl    $(4096/64)-5,   %ecx
  34         .p2align 4
  35 .Loop64:
  36         dec     %rcx
  37         movq    0x8*0(%rsi), %rax
  38         movq    0x8*1(%rsi), %rbx
  39         movq    0x8*2(%rsi), %rdx
  40         movq    0x8*3(%rsi), %r8
  41         movq    0x8*4(%rsi), %r9
  42         movq    0x8*5(%rsi), %r10
  43         movq    0x8*6(%rsi), %r11
  44         movq    0x8*7(%rsi), %r12
  45
  46         prefetcht0 5*64(%rsi)
  47
  48         movq    %rax, 0x8*0(%rdi)
  49         movq    %rbx, 0x8*1(%rdi)
  50         movq    %rdx, 0x8*2(%rdi)
  51         movq    %r8,  0x8*3(%rdi)
  52         movq    %r9,  0x8*4(%rdi)
  53         movq    %r10, 0x8*5(%rdi)
  54         movq    %r11, 0x8*6(%rdi)
  55         movq    %r12, 0x8*7(%rdi)
  56
  57         leaq    64 (%rsi), %rsi
  58         leaq    64 (%rdi), %rdi
  59
  60         jnz     .Loop64
  61
  62         movl    $5, %ecx
  63         .p2align 4
  64 .Loop2:
  65         decl    %ecx
  66
  67         movq    0x8*0(%rsi), %rax
  68         movq    0x8*1(%rsi), %rbx
  69         movq    0x8*2(%rsi), %rdx
  70         movq    0x8*3(%rsi), %r8
  71         movq    0x8*4(%rsi), %r9
  72         movq    0x8*5(%rsi), %r10
  73         movq    0x8*6(%rsi), %r11
  74         movq    0x8*7(%rsi), %r12
  75
  76         movq    %rax, 0x8*0(%rdi)
  77         movq    %rbx, 0x8*1(%rdi)
  78         movq    %rdx, 0x8*2(%rdi)
  79         movq    %r8,  0x8*3(%rdi)
  80         movq    %r9,  0x8*4(%rdi)
  81         movq    %r10, 0x8*5(%rdi)
  82         movq    %r11, 0x8*6(%rdi)
  83         movq    %r12, 0x8*7(%rdi)
  84
  85         leaq    64(%rdi), %rdi
  86         leaq    64(%rsi), %rsi
  87         jnz     .Loop2
  88
  89         movq    (%rsp), %rbx
  90         CFI_RESTORE rbx
  91         movq    1*8(%rsp), %r12
  92         CFI_RESTORE r12
  93         addq    $2*8, %rsp
  94         CFI_ADJUST_CFA_OFFSET -2*8
  95         ret
  96         CFI_ENDPROC
  97 ENDPROC(copy_page_regs)