x86-64: Handle byte-wise tail copying in memcpy() without a loop

While hard to measure, reducing the number of possibly/likely
mis-predicted branches can generally be expected to be slightly
better.

Other than apparent at the first glance, this also doesn't grow
the function size (the alignment gap to the next function just
gets smaller).

Signed-off-by: Jan Beulich <jbeulich@suse.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Link: http://lkml.kernel.org/r/4F218584020000780006F422@nat28.tlf.novell.com
Signed-off-by: Ingo Molnar <mingo@elte.hu>

diff --git a/arch/x86/lib/memcpy_64.S b/arch/x86/lib/memcpy_64.S
index 1235b04..1c273be 100644 (file)
--- a/arch/x86/lib/memcpy_64.S
+++ b/arch/x86/lib/memcpy_64.S
@@ -164,18 +164,19 @@ ENTRY(memcpy)
        retq
        .p2align 4
 .Lless_3bytes:
-       cmpl $0, %edx
-       je .Lend
+       subl $1, %edx
+       jb .Lend
        /*
         * Move data from 1 bytes to 3 bytes.
         */
-.Lloop_1:
-       movb (%rsi), %r8b
-       movb %r8b, (%rdi)
-       incq %rdi
-       incq %rsi
-       decl %edx
-       jnz .Lloop_1
+       movzbl (%rsi), %ecx
+       jz .Lstore_1byte
+       movzbq 1(%rsi), %r8
+       movzbq (%rsi, %rdx), %r9
+       movb %r8b, 1(%rdi)
+       movb %r9b, (%rdi, %rdx)
+.Lstore_1byte:
+       movb %cl, (%rdi)
 
 .Lend:
        retq