memcpy-armv7a.S: Improve performance with misaligned buffers.

Adjust the conditions for entering the aligned copy loop to
improve performance on mutually misaligned buffer copies.

2013-07-01  Will Newton  <will.newton@linaro.org>

	* libc/machine/arm/memcpy-armv7a.S: Adjust entry to
	 aligned loop to improve misaligned copy performance.

diff --git a/newlib/ChangeLog b/newlib/ChangeLog
index d5b7c7b44f20d59fa27922c46d6e9099870f5eb4..5b73e71abffbc1e36d5c40c554d0fb6329253cfb 100644 (file)
--- a/newlib/ChangeLog
+++ b/newlib/ChangeLog
@@ -1,3 +1,8 @@
+2013-07-01  Will Newton  <will.newton@linaro.org>
+
+       * libc/machine/arm/memcpy-armv7a.S: Adjust entry to
+        aligned loop to improve misaligned copy performance.
+
 2013-06-26  Jeff Johnston  <jjohnstn@redhat.com>
 
        * libc/include/stdio.h: Specify std streams always in terms

diff --git a/newlib/libc/machine/arm/memcpy-armv7a.S b/newlib/libc/machine/arm/memcpy-armv7a.S
index de5bf9ad89a7d556c4e1f3a328046cbafcb4a234..cd7962e075a30cb90ec073d77b177c3536429b9b 100644 (file)
--- a/newlib/libc/machine/arm/memcpy-armv7a.S
+++ b/newlib/libc/machine/arm/memcpy-armv7a.S
@@ -242,8 +242,8 @@ def_fn memcpy p2align=6
 .Lcpy_not_short:
        /* At least 64 bytes to copy, but don't know the alignment yet.  */
        str     tmp2, [sp, #-FRAME_SIZE]!
-       and     tmp2, src, #3
-       and     tmp1, dst, #3
+       and     tmp2, src, #7
+       and     tmp1, dst, #7
        cmp     tmp1, tmp2
        bne     .Lcpy_notaligned