libvpx: Pull from upstream

source/libvpx/vp9/encoder/x86/vp9_avg_intrin_sse2.c

Index: source/libvpx/vp9/encoder/x86/vp9_avg_intrin_sse2.c
diff --git a/source/libvpx/vp9/encoder/x86/vp9_avg_intrin_sse2.c b/source/libvpx/vp9/encoder/x86/vp9_avg_intrin_sse2.c
index 56a91ed2d711f6919fb900a62c36786b0fe11bbe..4531d794a9ceffaebed35b795358a442f3f95a76 100644
--- a/source/libvpx/vp9/encoder/x86/vp9_avg_intrin_sse2.c
+++ b/source/libvpx/vp9/encoder/x86/vp9_avg_intrin_sse2.c
@@ -264,17 +264,18 @@ void vp9_hadamard_16x16_sse2(int16_t const *src_diff, int src_stride,
     __m128i b2 = _mm_add_epi16(coeff2, coeff3);
     __m128i b3 = _mm_sub_epi16(coeff2, coeff3);
 
+    b0 = _mm_srai_epi16(b0, 1);
+    b1 = _mm_srai_epi16(b1, 1);
+    b2 = _mm_srai_epi16(b2, 1);
+    b3 = _mm_srai_epi16(b3, 1);
+
     coeff0 = _mm_add_epi16(b0, b2);
     coeff1 = _mm_add_epi16(b1, b3);
-    coeff0 = _mm_srai_epi16(coeff0, 1);
-    coeff1 = _mm_srai_epi16(coeff1, 1);
     _mm_store_si128((__m128i *)coeff, coeff0);
     _mm_store_si128((__m128i *)(coeff + 64), coeff1);
 
     coeff2 = _mm_sub_epi16(b0, b2);
     coeff3 = _mm_sub_epi16(b1, b3);
-    coeff2 = _mm_srai_epi16(coeff2, 1);
-    coeff3 = _mm_srai_epi16(coeff3, 1);
     _mm_store_si128((__m128i *)(coeff + 128), coeff2);
     _mm_store_si128((__m128i *)(coeff + 192), coeff3);