Optimize memmove

This speeds up memmove by copying a word at a time.

Ran the unit test on Peppy:
  > runtest
  ...
  Running test_memmove... (speed gain: 2156 -> 592 us) OK
  ...

Ran make buildall:
  ...
  Running test_memmove... (speed gain: 143918 -> 32367 us) OK
  ...

TEST=Described above.
BUG=chrome-os-partner:23720
BRANCH=none
Signed-off-by: Daisuke Nojiri <dnojiri@chromium.org>
Tested-by: Daisuke Nojiri <dnojiri@google.com>
Change-Id: I6a3ac6aed27a404c3bef227b6c886a59414b51d7
Reviewed-on: https://chromium-review.googlesource.com/186020
Reviewed-by: Vic Yang <victoryang@chromium.org>
Reviewed-by: Randall Spangler <rspangler@chromium.org>

1 files changed, 36 insertions, 8 deletions

diff --git a/common/util.c b/common/util.c
index 628aef9e82..5158d8aa06 100644
--- a/common/util.c
+++ b/common/util.c
@@ -238,17 +238,45 @@ void *memmove(void *dest, const void *src, int len)
 		 * memcpy(). */
 		return memcpy(dest, src, len);
 	} else {
-		/* Copy from end, so we don't overwrite the source */
+		/* Need to copy from tail because there is overlap. */
 		char *d = (char *)dest + len;
 		const char *s = (const char *)src + len;
-		/*
-		 * TODO(crosbug.com/p/23720): if src/dest are aligned, copy a
-		 * word at a time instead.
-		 */
-		while (len > 0) {
-			*(--d) = *(--s);
-			len--;
+		uint32_t *dw;
+		const uint32_t *sw;
+		char *head;
+		char * const tail = (char *)dest;
+		/* Set 'body' to the last word boundary */
+		uint32_t * const body = (uint32_t *)(((uintptr_t)tail+3) & ~3);
+
+		if (((uintptr_t)dest & 3) != ((uintptr_t)src & 3)) {
+			/* Misaligned. no body, no tail. */
+			head = tail;
+		} else {
+			/* Aligned */
+			if ((uintptr_t)tail > ((uintptr_t)d & ~3))
+				/* Shorter than the first word boundary */
+				head = tail;
+			else
+				/* Set 'head' to the first word boundary */
+				head = (char *)((uintptr_t)d & ~3);
 		}
+
+		/* Copy head */
+		while (d > head)
+			*(--d) = *(--s);
+
+		/* Copy body */
+		dw = (uint32_t *)d;
+		sw = (uint32_t *)s;
+		while (dw > body)
+			*(--dw) = *(--sw);
+
+		/* Copy tail */
+		d = (char *)dw;
+		s = (const char *)sw;
+		while (d > tail)
+			*(--d) = *(--s);
+
 		return dest;
 	}
 }