diff options
Diffstat (limited to 'sysdeps/x86_64/strcpy.S')
-rw-r--r-- | sysdeps/x86_64/strcpy.S | 138 |
1 files changed, 2 insertions, 136 deletions
diff --git a/sysdeps/x86_64/strcpy.S b/sysdeps/x86_64/strcpy.S index 17e8073550..05f19e6e94 100644 --- a/sysdeps/x86_64/strcpy.S +++ b/sysdeps/x86_64/strcpy.S @@ -16,140 +16,6 @@ License along with the GNU C Library; if not, see <https://www.gnu.org/licenses/>. */ -#include <sysdep.h> -#include "asm-syntax.h" - -#ifndef USE_AS_STPCPY -# define STRCPY strcpy -#endif - - .text -ENTRY (STRCPY) - movq %rsi, %rcx /* Source register. */ - andl $7, %ecx /* mask alignment bits */ - movq %rdi, %rdx /* Duplicate destination pointer. */ - - jz 5f /* aligned => start loop */ - - neg %ecx /* We need to align to 8 bytes. */ - addl $8,%ecx - /* Search the first bytes directly. */ -0: - movb (%rsi), %al /* Fetch a byte */ - testb %al, %al /* Is it NUL? */ - movb %al, (%rdx) /* Store it */ - jz 4f /* If it was NUL, done! */ - incq %rsi - incq %rdx - decl %ecx - jnz 0b - -5: - movq $0xfefefefefefefeff,%r8 - - /* Now the sources is aligned. Unfortunatly we cannot force - to have both source and destination aligned, so ignore the - alignment of the destination. */ - .p2align 4 -1: - /* 1st unroll. */ - movq (%rsi), %rax /* Read double word (8 bytes). */ - addq $8, %rsi /* Adjust pointer for next word. */ - movq %rax, %r9 /* Save a copy for NUL finding. */ - addq %r8, %r9 /* add the magic value to the word. We get - carry bits reported for each byte which - is *not* 0 */ - jnc 3f /* highest byte is NUL => return pointer */ - xorq %rax, %r9 /* (word+magic)^word */ - orq %r8, %r9 /* set all non-carry bits */ - incq %r9 /* add 1: if one carry bit was *not* set - the addition will not result in 0. */ - - jnz 3f /* found NUL => return pointer */ - - movq %rax, (%rdx) /* Write value to destination. */ - addq $8, %rdx /* Adjust pointer. */ - - /* 2nd unroll. */ - movq (%rsi), %rax /* Read double word (8 bytes). */ - addq $8, %rsi /* Adjust pointer for next word. */ - movq %rax, %r9 /* Save a copy for NUL finding. */ - addq %r8, %r9 /* add the magic value to the word. We get - carry bits reported for each byte which - is *not* 0 */ - jnc 3f /* highest byte is NUL => return pointer */ - xorq %rax, %r9 /* (word+magic)^word */ - orq %r8, %r9 /* set all non-carry bits */ - incq %r9 /* add 1: if one carry bit was *not* set - the addition will not result in 0. */ - - jnz 3f /* found NUL => return pointer */ - - movq %rax, (%rdx) /* Write value to destination. */ - addq $8, %rdx /* Adjust pointer. */ - - /* 3rd unroll. */ - movq (%rsi), %rax /* Read double word (8 bytes). */ - addq $8, %rsi /* Adjust pointer for next word. */ - movq %rax, %r9 /* Save a copy for NUL finding. */ - addq %r8, %r9 /* add the magic value to the word. We get - carry bits reported for each byte which - is *not* 0 */ - jnc 3f /* highest byte is NUL => return pointer */ - xorq %rax, %r9 /* (word+magic)^word */ - orq %r8, %r9 /* set all non-carry bits */ - incq %r9 /* add 1: if one carry bit was *not* set - the addition will not result in 0. */ - - jnz 3f /* found NUL => return pointer */ - - movq %rax, (%rdx) /* Write value to destination. */ - addq $8, %rdx /* Adjust pointer. */ - - /* 4th unroll. */ - movq (%rsi), %rax /* Read double word (8 bytes). */ - addq $8, %rsi /* Adjust pointer for next word. */ - movq %rax, %r9 /* Save a copy for NUL finding. */ - addq %r8, %r9 /* add the magic value to the word. We get - carry bits reported for each byte which - is *not* 0 */ - jnc 3f /* highest byte is NUL => return pointer */ - xorq %rax, %r9 /* (word+magic)^word */ - orq %r8, %r9 /* set all non-carry bits */ - incq %r9 /* add 1: if one carry bit was *not* set - the addition will not result in 0. */ - - jnz 3f /* found NUL => return pointer */ - - movq %rax, (%rdx) /* Write value to destination. */ - addq $8, %rdx /* Adjust pointer. */ - jmp 1b /* Next iteration. */ - - /* Do the last few bytes. %rax contains the value to write. - The loop is unrolled twice. */ - .p2align 4 -3: - /* Note that stpcpy needs to return with the value of the NUL - byte. */ - movb %al, (%rdx) /* 1st byte. */ - testb %al, %al /* Is it NUL. */ - jz 4f /* yes, finish. */ - incq %rdx /* Increment destination. */ - movb %ah, (%rdx) /* 2nd byte. */ - testb %ah, %ah /* Is it NUL?. */ - jz 4f /* yes, finish. */ - incq %rdx /* Increment destination. */ - shrq $16, %rax /* Shift... */ - jmp 3b /* and look at next two bytes in %rax. */ - -4: -#ifdef USE_AS_STPCPY - movq %rdx, %rax /* Destination is return value. */ -#else - movq %rdi, %rax /* Source is return value. */ -#endif - retq -END (STRCPY) -#ifndef USE_AS_STPCPY +#define STRCPY strcpy +#include "multiarch/strcpy-sse2.S" libc_hidden_builtin_def (strcpy) -#endif |