This is the mail archive of the
newlib@sourceware.org
mailing list for the newlib project.
[PATCH] memcpy-armv7a.S: Improve performance with misaligned buffers.
- From: Will Newton <will dot newton at linaro dot org>
- To: newlib at sourceware dot org
- Cc: patches at linaro dot org
- Date: Fri, 28 Jun 2013 11:25:16 +0100
- Subject: [PATCH] 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.
2013-06-28 Will Newton <will.newton@linaro.org>
* libc/machine/arm/memcpy-armv7a.S: Adjust entry to
aligned loop to improve misaligned copy performance.
---
newlib/libc/machine/arm/memcpy-armv7a.S | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/newlib/libc/machine/arm/memcpy-armv7a.S b/newlib/libc/machine/arm/memcpy-armv7a.S
index de5bf9a..cd7962e 100644
--- 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
--
1.8.1.4