[S] Change in coreboot[main]: arch/x86/memcpy.c: Optimize code for 64bit

[Martin L Roth (Code Review)]

Martin L Roth has submitted this change. ( https://review.coreboot.org/c/coreboot/+/78646?usp=email )

Change subject: arch/x86/memcpy.c: Optimize code for 64bit
......................................................................

arch/x86/memcpy.c: Optimize code for 64bit

In 64bit movsq is available which moves memory in chunks of 8 bytes
rather than 4 bytes.

Linux uses the same code.

Signed-off-by: Arthur Heymans 
Change-Id: I65f178d2ed3aae54b0c1ce739c2b4af8738b9fcc
Reviewed-on: https://review.coreboot.org/c/coreboot/+/78646
Tested-by: build bot (Jenkins) 
Reviewed-by: Felix Held 
Reviewed-by: Martin Roth 
Reviewed-by: Jérémy Compostella 
Reviewed-by: Eric Lai 
---
M src/arch/x86/memcpy.c
1 file changed, 10 insertions(+), 3 deletions(-)

Approvals:
  Jérémy Compostella: Looks good to me, approved
  build bot (Jenkins): Verified
  Martin Roth: Looks good to me, approved
  Felix Held: Looks good to me, but someone else must approve
  Eric Lai: Looks good to me, but someone else must approve

diff --git a/src/arch/x86/memcpy.c b/src/arch/x86/memcpy.c
index 93002cd..d96a93c 100644
--- a/src/arch/x86/memcpy.c
+++ b/src/arch/x86/memcpy.c
@@ -14,19 +14,26 @@
    check_memory_region((unsigned long)dest, n, true, _RET_IP_);
 #endif

-	asm volatile(
 #if ENV_X86_64
-		"rep ; movsd\n\t"
+	asm volatile(
+		"rep ; movsq\n\t"
    	"mov %4,%%rcx\n\t"
+		"rep ; movsb\n\t"
+		: "=&c" (d0), "=&D" (d1), "=&S" (d2)
+		: "0" (n >> 3), "g" (n & 7), "1" (dest), "2" (src)
+		: "memory"
+	);
 #else
+	asm volatile(
    	"rep ; movsl\n\t"
    	"movl %4,%%ecx\n\t"
-#endif
    	"rep ; movsb\n\t"
    	: "=&c" (d0), "=&D" (d1), "=&S" (d2)
    	: "0" (n >> 2), "g" (n & 3), "1" (dest), "2" (src)
    	: "memory"
    );
+#endif
+

return dest;
 }