Mine

#define SDL_memcpy4(dst, src, len)				\
do {								\
	int ecx, edi, esi;					\
	__asm__ __volatile__ (					\
		"cld\n\t"					\
		"rep ; movsl"					\
		: "=&c" (ecx), "=&D" (edi), "=&S" (esi)		\
		: "0" ((unsigned)(len)), "1" (dst), "2" (src)	\
		: "memory" );					\
} while(0)