diff options
| author | Jingning Han <jingning@google.com> | 2015-03-30 15:02:54 -0700 |
|---|---|---|
| committer | Gerrit Code Review <gerrit@gerrit.golo.chromium.org> | 2015-03-30 15:54:08 -0700 |
| commit | 34a996ac1e69abd0caf8310a0753a2bebfdbce6c (patch) | |
| tree | 8884494bbd50cadfd489ece520ba8e7bf936ad22 | |
| parent | 26d3d3af6ace50ee20c474771a580e38f86c7864 (diff) | |
| download | libvpx-34a996ac1e69abd0caf8310a0753a2bebfdbce6c.tar libvpx-34a996ac1e69abd0caf8310a0753a2bebfdbce6c.tar.gz libvpx-34a996ac1e69abd0caf8310a0753a2bebfdbce6c.tar.bz2 libvpx-34a996ac1e69abd0caf8310a0753a2bebfdbce6c.zip | |
Fix 8x8 Hadamard SSE2 implementation
This commit fixes the SSE2 version 8x8 Hadamard transform
alignment and makes it consistent with the C version.
Change-Id: I1304e5f97e0e5ef2d798fe38081609c39f5bfe74
| -rw-r--r-- | vp9/encoder/x86/vp9_avg_intrin_sse2.c | 28 |
1 files changed, 14 insertions, 14 deletions
diff --git a/vp9/encoder/x86/vp9_avg_intrin_sse2.c b/vp9/encoder/x86/vp9_avg_intrin_sse2.c index ee3ead49c..f2c7c645f 100644 --- a/vp9/encoder/x86/vp9_avg_intrin_sse2.c +++ b/vp9/encoder/x86/vp9_avg_intrin_sse2.c @@ -87,13 +87,13 @@ static void hadamard_col8_sse2(__m128i *in, int iter) { if (iter == 0) { b0 = _mm_add_epi16(a0, a4); - b1 = _mm_add_epi16(a1, a5); - b2 = _mm_add_epi16(a2, a6); - b3 = _mm_add_epi16(a3, a7); - b4 = _mm_sub_epi16(a0, a4); - b5 = _mm_sub_epi16(a1, a5); - b6 = _mm_sub_epi16(a2, a6); - b7 = _mm_sub_epi16(a3, a7); + b7 = _mm_add_epi16(a1, a5); + b3 = _mm_add_epi16(a2, a6); + b4 = _mm_add_epi16(a3, a7); + b2 = _mm_sub_epi16(a0, a4); + b6 = _mm_sub_epi16(a1, a5); + b1 = _mm_sub_epi16(a2, a6); + b5 = _mm_sub_epi16(a3, a7); a0 = _mm_unpacklo_epi16(b0, b1); a1 = _mm_unpacklo_epi16(b2, b3); @@ -114,13 +114,13 @@ static void hadamard_col8_sse2(__m128i *in, int iter) { b7 = _mm_unpackhi_epi32(a6, a7); in[0] = _mm_unpacklo_epi64(b0, b1); - in[7] = _mm_unpackhi_epi64(b0, b1); - in[3] = _mm_unpacklo_epi64(b2, b3); - in[4] = _mm_unpackhi_epi64(b2, b3); - in[2] = _mm_unpacklo_epi64(b4, b5); - in[6] = _mm_unpackhi_epi64(b4, b5); - in[1] = _mm_unpacklo_epi64(b6, b7); - in[5] = _mm_unpackhi_epi64(b6, b7); + in[1] = _mm_unpackhi_epi64(b0, b1); + in[2] = _mm_unpacklo_epi64(b2, b3); + in[3] = _mm_unpackhi_epi64(b2, b3); + in[4] = _mm_unpacklo_epi64(b4, b5); + in[5] = _mm_unpackhi_epi64(b4, b5); + in[6] = _mm_unpacklo_epi64(b6, b7); + in[7] = _mm_unpackhi_epi64(b6, b7); } else { in[0] = _mm_add_epi16(a0, a4); in[7] = _mm_add_epi16(a1, a5); |