From 62d705d4a2764cf7d36300bcb4a26a806d0c6db1 Mon Sep 17 00:00:00 2001 From: Keir Fraser Date: Wed, 7 Oct 2009 07:45:39 +0100 Subject: [PATCH] Optimize memcpy for x86 arch. If source buffers does not start at a 64 bit boundary, copy a few bytes at the beginnig up to next 64-bit boundary and then does an aligned copy for the remaining data. This can reduce the copy cost by up to 50%. Signed-off-by: Jose Renato Santos --- xen/include/asm-x86/string.h | 18 +++++++++++++++++- 1 file changed, 17 insertions(+), 1 deletion(-) diff --git a/xen/include/asm-x86/string.h b/xen/include/asm-x86/string.h index 496b22754f..c3481216ee 100644 --- a/xen/include/asm-x86/string.h +++ b/xen/include/asm-x86/string.h @@ -96,13 +96,29 @@ static always_inline void * __constant_memcpy( } #define __HAVE_ARCH_MEMCPY +/* align source to a 64-bit boundary */ +static always_inline +void *__var_memcpy(void *t, const void *f, size_t n) +{ + int off = (unsigned long)f & 0x7; + /* just do alignment if needed and if size is worth */ + if ( (n > 32) && off ) { + size_t n1 = 8 - off; + __variable_memcpy(t, f, n1); + __variable_memcpy(t + n1, f + n1, n - n1); + return t; + } else { + return (__variable_memcpy(t, f, n)); + } +} + #define memcpy(t,f,n) (__memcpy((t),(f),(n))) static always_inline void *__memcpy(void *t, const void *f, size_t n) { return (__builtin_constant_p(n) ? __constant_memcpy((t),(f),(n)) : - __variable_memcpy((t),(f),(n))); + __var_memcpy((t),(f),(n))); } /* Some version of gcc don't have this builtin. It's non-critical anyway. */ -- 2.30.2