diff options
| author | xsacha <xsacha@gmail.com> | 2011-01-08 04:59:26 +0000 |
|---|---|---|
| committer | xsacha <xsacha@gmail.com> | 2011-01-08 04:59:26 +0000 |
| commit | 3cf8003a55439a1c1a108d7fc67e9d2c26a508c3 (patch) | |
| tree | d7c9efbd0c07a3f14c1b6df80c247e34b594aa18 /Source/Core/VideoCommon/Src/TextureDecoder.cpp | |
| parent | 807671e32f6b30096a7dc19ac796b6a0cafb9825 (diff) | |
From my last commit: Fix build on Linux. Use SSSE3 instead of SSE3.
Remove some unused vars from the SSE2 CMPR.
git-svn-id: https://dolphin-emu.googlecode.com/svn/trunk@6781 8ced0084-cf51-0410-be5f-012b33b47a6e
Diffstat (limited to 'Source/Core/VideoCommon/Src/TextureDecoder.cpp')
| -rw-r--r-- | Source/Core/VideoCommon/Src/TextureDecoder.cpp | 6 |
1 files changed, 2 insertions, 4 deletions
diff --git a/Source/Core/VideoCommon/Src/TextureDecoder.cpp b/Source/Core/VideoCommon/Src/TextureDecoder.cpp index 76b2abaa08..bcf14877f9 100644 --- a/Source/Core/VideoCommon/Src/TextureDecoder.cpp +++ b/Source/Core/VideoCommon/Src/TextureDecoder.cpp @@ -1451,7 +1451,7 @@ PC_TexFormat TexDecoder_Decode_RGBA(u32 * dst, const u8 * src, int width, int he u32 *newdst = dst+(y+iy)*width+x; #if _M_SSE >= 0x301 // Produces a ~40% speed improvement over reference C implementation - if (cpu_info.bSSE3) + if (cpu_info.bSSSE3) { const __m128i mask = _mm_set_epi8(128,128,6,7,128,128,4,5,128,128,2,3,128,128,0,1); const __m128i valV = _mm_shuffle_epi8(_mm_loadl_epi64((const __m128i*)src),mask); @@ -1510,7 +1510,7 @@ PC_TexFormat TexDecoder_Decode_RGBA(u32 * dst, const u8 * src, int width, int he else { // TODO: Vectorise (Either 4-way branch or do both and select is better than this) - unsigned __int32 *vals = (unsigned __int32*) &valV; + u32 *vals = (u32*) &valV; int r,g,b,a; for (int i=0; i < 4; ++i) { @@ -1867,7 +1867,6 @@ PC_TexFormat TexDecoder_Decode_RGBA(u32 * dst, const u8 * src, int width, int he u32 dxt1sel = dxttmp[3]; __m128i argb888x4; - const __m128i lowMask = _mm_srli_si128( allFFs128, 8 ); __m128i c1 = _mm_unpackhi_epi16(dxt, dxt); c1 = _mm_slli_si128(c1, 8); const __m128i c0 = _mm_or_si128(c1, _mm_srli_si128(_mm_slli_si128(_mm_unpacklo_epi16(dxt, dxt), 8), 8)); @@ -1889,7 +1888,6 @@ PC_TexFormat TexDecoder_Decode_RGBA(u32 * dst, const u8 * src, int width, int he const __m128i gtmp = _mm_srli_epi32(c0, 3); const __m128i g0 = _mm_and_si128(gtmp, low6mask); // low3mask == _mm_set_epi32(0x00000300, 0x00000300, 0x00000300, 0x00000300) - const __m128i low3mask = _mm_slli_epi32(_mm_srli_epi32(allFFs128, 32 - 3), 8); const __m128i g1 = _mm_and_si128(_mm_srli_epi32(gtmp, 6), _mm_set_epi32(0x00000300, 0x00000300, 0x00000300, 0x00000300)); argb888x4 = _mm_or_si128(g0, g1); // red: |