Commit 455ae94c authored by fbarchard@google.com's avatar fbarchard@google.com

Make rotate SIMD allow unaligned pointers.

BUG=365
TESTED=libyuv_unittest
R=tpsiaki@google.com

Review URL: https://webrtc-codereview.appspot.com/22899004

git-svn-id: http://libyuv.googlecode.com/svn/trunk@1102 16f28f9a-4ce2-e073-06de-1de4eb20be90
parent 044f914c
Name: libyuv Name: libyuv
URL: http://code.google.com/p/libyuv/ URL: http://code.google.com/p/libyuv/
Version: 1101 Version: 1102
License: BSD License: BSD
License File: LICENSE License File: LICENSE
......
...@@ -11,6 +11,6 @@ ...@@ -11,6 +11,6 @@
#ifndef INCLUDE_LIBYUV_VERSION_H_ // NOLINT #ifndef INCLUDE_LIBYUV_VERSION_H_ // NOLINT
#define INCLUDE_LIBYUV_VERSION_H_ #define INCLUDE_LIBYUV_VERSION_H_
#define LIBYUV_VERSION 1101 #define LIBYUV_VERSION 1102
#endif // INCLUDE_LIBYUV_VERSION_H_ NOLINT #endif // INCLUDE_LIBYUV_VERSION_H_ NOLINT
This diff is collapsed.
...@@ -2970,10 +2970,10 @@ void MirrorRow_SSSE3(const uint8* src, uint8* dst, int width) { ...@@ -2970,10 +2970,10 @@ void MirrorRow_SSSE3(const uint8* src, uint8* dst, int width) {
"lea " MEMLEA(-0x10,0) ",%0 \n" "lea " MEMLEA(-0x10,0) ",%0 \n"
LABELALIGN LABELALIGN
"1: \n" "1: \n"
MEMOPREG(movdqa,0x00,0,2,1,xmm0) // movdqa (%0,%2),%%xmm0 MEMOPREG(movdqu,0x00,0,2,1,xmm0) // movdqa (%0,%2),%%xmm0
"pshufb %%xmm5,%%xmm0 \n" "pshufb %%xmm5,%%xmm0 \n"
"sub $0x10,%2 \n" "sub $0x10,%2 \n"
"movdqa %%xmm0," MEMACCESS(1) " \n" "movdqu %%xmm0," MEMACCESS(1) " \n"
"lea " MEMLEA(0x10,1) ",%1 \n" "lea " MEMLEA(0x10,1) ",%1 \n"
"jg 1b \n" "jg 1b \n"
: "+r"(src), // %0 : "+r"(src), // %0
...@@ -3039,7 +3039,7 @@ void MirrorUVRow_SSSE3(const uint8* src, uint8* dst_u, uint8* dst_v, ...@@ -3039,7 +3039,7 @@ void MirrorUVRow_SSSE3(const uint8* src, uint8* dst_u, uint8* dst_v,
"sub %1,%2 \n" "sub %1,%2 \n"
LABELALIGN LABELALIGN
"1: \n" "1: \n"
"movdqa " MEMACCESS(0) ",%%xmm0 \n" "movdqu " MEMACCESS(0) ",%%xmm0 \n"
"lea " MEMLEA(-0x10,0) ",%0 \n" "lea " MEMLEA(-0x10,0) ",%0 \n"
"pshufb %%xmm1,%%xmm0 \n" "pshufb %%xmm1,%%xmm0 \n"
"sub $8,%3 \n" "sub $8,%3 \n"
...@@ -3077,11 +3077,11 @@ void ARGBMirrorRow_SSSE3(const uint8* src, uint8* dst, int width) { ...@@ -3077,11 +3077,11 @@ void ARGBMirrorRow_SSSE3(const uint8* src, uint8* dst, int width) {
"movdqa %3,%%xmm5 \n" "movdqa %3,%%xmm5 \n"
LABELALIGN LABELALIGN
"1: \n" "1: \n"
"movdqa " MEMACCESS(0) ",%%xmm0 \n" "movdqu " MEMACCESS(0) ",%%xmm0 \n"
"pshufb %%xmm5,%%xmm0 \n" "pshufb %%xmm5,%%xmm0 \n"
"lea " MEMLEA(-0x10,0) ",%0 \n" "lea " MEMLEA(-0x10,0) ",%0 \n"
"sub $0x4,%2 \n" "sub $0x4,%2 \n"
"movdqa %%xmm0," MEMACCESS(1) " \n" "movdqu %%xmm0," MEMACCESS(1) " \n"
"lea " MEMLEA(0x10,1) ",%1 \n" "lea " MEMLEA(0x10,1) ",%1 \n"
"jg 1b \n" "jg 1b \n"
: "+r"(src), // %0 : "+r"(src), // %0
......
...@@ -3288,10 +3288,10 @@ void MirrorRow_SSSE3(const uint8* src, uint8* dst, int width) { ...@@ -3288,10 +3288,10 @@ void MirrorRow_SSSE3(const uint8* src, uint8* dst, int width) {
align 4 align 4
convertloop: convertloop:
movdqa xmm0, [eax + ecx] movdqu xmm0, [eax + ecx]
pshufb xmm0, xmm5 pshufb xmm0, xmm5
sub ecx, 16 sub ecx, 16
movdqa [edx], xmm0 movdqu [edx], xmm0
lea edx, [edx + 16] lea edx, [edx + 16]
jg convertloop jg convertloop
ret ret
...@@ -3381,7 +3381,7 @@ void MirrorUVRow_SSSE3(const uint8* src, uint8* dst_u, uint8* dst_v, ...@@ -3381,7 +3381,7 @@ void MirrorUVRow_SSSE3(const uint8* src, uint8* dst_u, uint8* dst_v,
align 4 align 4
convertloop: convertloop:
movdqa xmm0, [eax] movdqu xmm0, [eax]
lea eax, [eax - 16] lea eax, [eax - 16]
pshufb xmm0, xmm1 pshufb xmm0, xmm1
sub ecx, 8 sub ecx, 8
...@@ -3413,11 +3413,11 @@ void ARGBMirrorRow_SSSE3(const uint8* src, uint8* dst, int width) { ...@@ -3413,11 +3413,11 @@ void ARGBMirrorRow_SSSE3(const uint8* src, uint8* dst, int width) {
align 4 align 4
convertloop: convertloop:
movdqa xmm0, [eax] movdqu xmm0, [eax]
lea eax, [eax - 16] lea eax, [eax - 16]
pshufb xmm0, xmm5 pshufb xmm0, xmm5
sub ecx, 4 sub ecx, 4
movdqa [edx], xmm0 movdqu [edx], xmm0
lea edx, [edx + 16] lea edx, [edx + 16]
jg convertloop jg convertloop
ret ret
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment