fedora core 6 1.2949 + vserver 2.2.0

[linux-2.6.git] / arch / x86_64 / lib / memcpy.S

diff --git a/arch/x86_64/lib/memcpy.S b/arch/x86_64/lib/memcpy.S
index c6c4649..0ea0ddc 100644 (file)
--- a/arch/x86_64/lib/memcpy.S
+++ b/arch/x86_64/lib/memcpy.S
@@ -1,6 +1,9 @@
 /* Copyright 2002 Andi Kleen */
-       
-       #include <asm/cpufeature.h>             
+
+#include <linux/linkage.h>
+#include <asm/dwarf2.h>
+#include <asm/cpufeature.h>
+
 /*
  * memcpy - Copy a memory block.
  *
@@ -13,22 +16,36 @@
  * rax original destination
  */    
 
-       .globl __memcpy
-       .globl memcpy
-       .p2align 4
-__memcpy:
-memcpy:                
+       ALIGN
+memcpy_c:
+       CFI_STARTPROC
+       movq %rdi,%rax
+       movl %edx,%ecx
+       shrl $3,%ecx
+       andl $7,%edx
+       rep movsq
+       movl %edx,%ecx
+       rep movsb
+       ret
+       CFI_ENDPROC
+ENDPROC(memcpy_c)
+
+ENTRY(__memcpy)
+ENTRY(memcpy)
+       CFI_STARTPROC
        pushq %rbx
+       CFI_ADJUST_CFA_OFFSET 8
+       CFI_REL_OFFSET rbx, 0
        movq %rdi,%rax
 
        movl %edx,%ecx
        shrl $6,%ecx
        jz .Lhandle_tail
-       
+
        .p2align 4
 .Lloop_64:
        decl %ecx
-       
+
        movq (%rsi),%r11
        movq 8(%rsi),%r8
 
@@ -40,7 +57,7 @@ memcpy:
 
        movq %r9,2*8(%rdi)
        movq %r10,3*8(%rdi)
-               
+
        movq 4*8(%rsi),%r11
        movq 5*8(%rsi),%r8
 
@@ -63,10 +80,10 @@ memcpy:
        shrl $3,%ecx
        jz   .Lhandle_7
        .p2align 4
-.Lloop_8: 
+.Lloop_8:
        decl %ecx
        movq (%rsi),%r8
-       movq %r8,(%rdi) 
+       movq %r8,(%rdi)
        leaq 8(%rdi),%rdi
        leaq 8(%rsi),%rsi
        jnz  .Lloop_8
@@ -78,44 +95,35 @@ memcpy:
        .p2align 4
 .Lloop_1:
        movb (%rsi),%r8b
-       movb %r8b,(%rdi) 
+       movb %r8b,(%rdi)
        incq %rdi
        incq %rsi
        decl %ecx
        jnz .Lloop_1
-       
-.Lende:        
+
+.Lende:
        popq %rbx
+       CFI_ADJUST_CFA_OFFSET -8
+       CFI_RESTORE rbx
        ret
 .Lfinal:
-       
-       /* C stepping K8 run faster using the string copy instructions.
+       CFI_ENDPROC
+ENDPROC(memcpy)
+ENDPROC(__memcpy)
+
+       /* Some CPUs run faster using the string copy instructions.
           It is also a lot simpler. Use this when possible */
-       
-       .section .altinstructions,"a"
-       .align 8
-       .quad  memcpy
-       .quad  memcpy_c
-       .byte  X86_FEATURE_K8_C
-       .byte  .Lfinal-memcpy
-       .byte  memcpy_c_end-memcpy_c    
-       .previous
 
        .section .altinstr_replacement,"ax"
- /* rdi        destination
-  * rsi source
-  * rdx count
-  */                   
-memcpy_c:
-       movq %rdi,%rax
-       movl %edx,%ecx
-       shrl $3,%ecx
-       andl $7,%edx    
-       rep 
-       movsq 
-       movl %edx,%ecx
-       rep
-       movsb
-       ret
-memcpy_c_end:
+1:     .byte 0xeb                              /* jmp <disp8> */
+       .byte (memcpy_c - memcpy) - (2f - 1b)   /* offset */
+2:
+       .previous
+       .section .altinstructions,"a"
+       .align 8
+       .quad memcpy
+       .quad 1b
+       .byte X86_FEATURE_REP_GOOD
+       .byte .Lfinal - memcpy
+       .byte 2b - 1b
        .previous