Replace %xmm8 with %xmm0

Since ld.so preserves vector registers now, we can use %xmm0 to avoid
the REX prefix.

	* sysdeps/x86_64/memset.S: Replace %xmm8 with %xmm0.

diff --git a/ChangeLog b/ChangeLog
index 73645c3..362acee 100644 (file)
--- a/ChangeLog
+++ b/ChangeLog
@@ -1,3 +1,7 @@
+2015-08-25  H.J. Lu  <hongjiu.lu@intel.com>
+
+       * sysdeps/x86_64/memset.S: Replace %xmm8 with %xmm0.
+
 2015-08-25  Ondřej Bílka  <neleai@seznam.cz>
 
        * debug/strcpy_chk.c: Improve performance.

diff --git a/sysdeps/x86_64/memset.S b/sysdeps/x86_64/memset.S
index e496254..3855cc8 100644 (file)
--- a/sysdeps/x86_64/memset.S
+++ b/sysdeps/x86_64/memset.S
@@ -24,7 +24,7 @@
 ENTRY(__bzero)
        movq    %rdi, %rax /* Set return value.  */
        movq    %rsi, %rdx /* Set n.  */
-       pxor    %xmm8, %xmm8
+       pxor    %xmm0, %xmm0
        jmp     L(entry_from_bzero)
 END(__bzero)
 weak_alias (__bzero, bzero)
@@ -33,10 +33,10 @@ weak_alias (__bzero, bzero)
 ENTRY(__memset_tail)
        movq    %rcx, %rax /* Set return value.  */
 
-       movd    %esi, %xmm8
-       punpcklbw       %xmm8, %xmm8
-       punpcklwd       %xmm8, %xmm8
-       pshufd  $0, %xmm8, %xmm8
+       movd    %esi, %xmm0
+       punpcklbw       %xmm0, %xmm0
+       punpcklwd       %xmm0, %xmm0
+       pshufd  $0, %xmm0, %xmm0
 
        jmp     L(entry_from_bzero)
 END(__memset_tail)
@@ -50,57 +50,57 @@ END_CHK (__memset_chk)
 #endif
 
 ENTRY (memset)
-       movd    %esi, %xmm8
+       movd    %esi, %xmm0
        movq    %rdi, %rax
-       punpcklbw       %xmm8, %xmm8
-       punpcklwd       %xmm8, %xmm8
-       pshufd  $0, %xmm8, %xmm8
+       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  %xmm8, (%rdi)
-       movdqu  %xmm8, -16(%rdi,%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  %xmm8, 16(%rdi)
-       movdqu  %xmm8, -32(%rdi,%rdx)
+       movdqu  %xmm0, 16(%rdi)
+       movdqu  %xmm0, -32(%rdi,%rdx)
        ret
        .p2align 4
 L(loop_start):
        leaq    64(%rdi), %rcx
-       movdqu  %xmm8, (%rdi)
+       movdqu  %xmm0, (%rdi)
        andq    $-64, %rcx
-       movdqu  %xmm8, -16(%rdi,%rdx)
-       movdqu  %xmm8, 16(%rdi)
-       movdqu  %xmm8, -32(%rdi,%rdx)
-       movdqu  %xmm8, 32(%rdi)
-       movdqu  %xmm8, -48(%rdi,%rdx)
-       movdqu  %xmm8, 48(%rdi)
-       movdqu  %xmm8, -64(%rdi,%rdx)
+       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  %xmm8, (%rcx)
-       movdqa  %xmm8, 16(%rcx)
-       movdqa  %xmm8, 32(%rcx)
-       movdqa  %xmm8, 48(%rcx)
+       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 %xmm8, %rcx
+       movq %xmm0, %rcx
        testb   $24, %dl
        jne     L(between8_16bytes)
        testb   $4, %dl