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

@@ -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"