From 78ad8d1f98a4328890b41b5b9099fe63920aa2a6 Mon Sep 17 00:00:00 2001 From: "fbarchard@google.com" Date: Mon, 21 Oct 2013 18:04:24 +0000 Subject: [PATCH] Polynomial AVX2 on gcc use vex128 vmovq instead of SSE2 movq to avoid stall. BUG=265 TEST=unittest polynomial R=ryanpetrie@google.com Review URL: https://webrtc-codereview.appspot.com/2679004 git-svn-id: http://libyuv.googlecode.com/svn/trunk@816 16f28f9a-4ce2-e073-06de-1de4eb20be90 --- source/row_posix.cc | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/source/row_posix.cc b/source/row_posix.cc index 6908be213..d90cf1393 100644 --- a/source/row_posix.cc +++ b/source/row_posix.cc @@ -6068,7 +6068,7 @@ void ARGBPolynomialRow_AVX2(const uint8* src_argb, "vpermq $0xd8,%%ymm0,%%ymm0 \n" "vpackuswb %%xmm0,%%xmm0,%%xmm0 \n" "sub $0x2,%2 \n" - "movq %%xmm0,"MEMACCESS(1)" \n" + "vmovq %%xmm0,"MEMACCESS(1)" \n" "lea "MEMLEA(0x8,1)",%1 \n" "jg 1b \n" "vzeroupper \n"