From a9734a44925588be744f02ddccdb6fd73af4fc44 Mon Sep 17 00:00:00 2001 From: "fbarchard@google.com" Date: Fri, 12 Dec 2014 22:43:55 +0000 Subject: [PATCH] ARGBMirror for AVX had wrong loop counting. This fixes it to match windows, and reenables the function. BUG=269 TESTED=try bots R=tpsiaki@google.com Review URL: https://webrtc-codereview.appspot.com/33639004 git-svn-id: http://libyuv.googlecode.com/svn/trunk@1196 16f28f9a-4ce2-e073-06de-1de4eb20be90 --- README.chromium | 2 +- include/libyuv/row.h | 2 +- include/libyuv/version.h | 2 +- source/row_posix.cc | 2 +- 4 files changed, 4 insertions(+), 4 deletions(-) diff --git a/README.chromium b/README.chromium index 6d2203268..0f3ee080e 100644 --- a/README.chromium +++ b/README.chromium @@ -1,6 +1,6 @@ Name: libyuv URL: http://code.google.com/p/libyuv/ -Version: 1195 +Version: 1196 License: BSD License File: LICENSE diff --git a/include/libyuv/row.h b/include/libyuv/row.h index 620efbb5f..5c604e5f4 100644 --- a/include/libyuv/row.h +++ b/include/libyuv/row.h @@ -186,7 +186,7 @@ extern "C" { defined(CLANG_HAS_AVX2) || defined(GCC_HAS_AVX2)) #define HAS_ARGBCOPYALPHAROW_AVX2 #define HAS_ARGBCOPYYTOALPHAROW_AVX2 -//#define HAS_ARGBMIRRORROW_AVX2 +#define HAS_ARGBMIRRORROW_AVX2 #define HAS_ARGBPOLYNOMIALROW_AVX2 #define HAS_ARGBSHUFFLEROW_AVX2 //#define HAS_ARGBTOUVROW_AVX2 diff --git a/include/libyuv/version.h b/include/libyuv/version.h index 510d671a4..bda506fa9 100644 --- a/include/libyuv/version.h +++ b/include/libyuv/version.h @@ -11,6 +11,6 @@ #ifndef INCLUDE_LIBYUV_VERSION_H_ // NOLINT #define INCLUDE_LIBYUV_VERSION_H_ -#define LIBYUV_VERSION 1195 +#define LIBYUV_VERSION 1196 #endif // INCLUDE_LIBYUV_VERSION_H_ NOLINT diff --git a/source/row_posix.cc b/source/row_posix.cc index 7c433ee9f..e283f614a 100644 --- a/source/row_posix.cc +++ b/source/row_posix.cc @@ -2667,7 +2667,7 @@ void ARGBMirrorRow_AVX2(const uint8* src, uint8* dst, int width) { VMEMOPREG(vpermd,-0x20,0,2,4,ymm5,ymm0) // vpermd -0x20(%0,%2,4),ymm5,ymm0 "vmovdqu %%ymm0," MEMACCESS(1) " \n" "lea " MEMLEA(0x20,1) ",%1 \n" - "sub $0x20,%2 \n" + "sub $0x8,%2 \n" "jg 1b \n" "vzeroupper \n" : "+r"(src), // %0