From 6a4ff519c8c21c8acbedb5e2db4d92c14e99d9f1 Mon Sep 17 00:00:00 2001 From: Michael Brown Date: Tue, 26 Jun 2012 12:42:24 +0100 Subject: [PATCH] [libc] Simplify memcpy() implementation The "rep" prefix can be used with an iteration count of zero, which allows the variable-length memcpy() to be implemented without using any conditional jumps. Signed-off-by: Michael Brown --- src/arch/x86/core/x86_string.c | 26 ++++++++++---------------- 1 file changed, 10 insertions(+), 16 deletions(-) diff --git a/src/arch/x86/core/x86_string.c b/src/arch/x86/core/x86_string.c index 5838ebace..8c3a4d2a0 100644 --- a/src/arch/x86/core/x86_string.c +++ b/src/arch/x86/core/x86_string.c @@ -43,21 +43,15 @@ void * __memcpy ( void *dest, const void *src, size_t len ) { * moves. Using movsl rather than movsb speeds these up by * around 32%. */ - if ( len >> 2 ) { - __asm__ __volatile__ ( "rep movsl" - : "=&D" ( edi ), "=&S" ( esi ), - "=&c" ( discard_ecx ) - : "0" ( edi ), "1" ( esi ), - "2" ( len >> 2 ) - : "memory" ); - } - if ( len & 0x02 ) { - __asm__ __volatile__ ( "movsw" : "=&D" ( edi ), "=&S" ( esi ) - : "0" ( edi ), "1" ( esi ) : "memory" ); - } - if ( len & 0x01 ) { - __asm__ __volatile__ ( "movsb" : "=&D" ( edi ), "=&S" ( esi ) - : "0" ( edi ), "1" ( esi ) : "memory" ); - } + __asm__ __volatile__ ( "rep movsl" + : "=&D" ( edi ), "=&S" ( esi ), + "=&c" ( discard_ecx ) + : "0" ( edi ), "1" ( esi ), "2" ( len >> 2 ) + : "memory" ); + __asm__ __volatile__ ( "rep movsb" + : "=&D" ( edi ), "=&S" ( esi ), + "=&c" ( discard_ecx ) + : "0" ( edi ), "1" ( esi ), "2" ( len & 3 ) + : "memory" ); return dest; }