diff options
| author | nodchip <nodchip@gmail.com> | 2010-02-24 09:56:41 +0000 |
|---|---|---|
| committer | nodchip <nodchip@gmail.com> | 2010-02-24 09:56:41 +0000 |
| commit | 47fb73b71a27d1a1aaa9a7dc160de573c92a9ee3 (patch) | |
| tree | d773daf9d479ece37ce0f44ba00d810ea5ed4015 /Source/Core/VideoCommon/Src/VertexLoader_TextCoord.cpp | |
| parent | dc9d533d4f4d228052b74f828c9b78b5f73d6c1b (diff) | |
Added SSSE3/SSE4.1 code for speed up. The code does not work in this revision because cpu_info is not initialized properly. I will fix the issue in another commit.
git-svn-id: https://dolphin-emu.googlecode.com/svn/trunk@5119 8ced0084-cf51-0410-be5f-012b33b47a6e
Diffstat (limited to 'Source/Core/VideoCommon/Src/VertexLoader_TextCoord.cpp')
| -rw-r--r-- | Source/Core/VideoCommon/Src/VertexLoader_TextCoord.cpp | 29 |
1 files changed, 23 insertions, 6 deletions
diff --git a/Source/Core/VideoCommon/Src/VertexLoader_TextCoord.cpp b/Source/Core/VideoCommon/Src/VertexLoader_TextCoord.cpp index 00f4a31667..036d0e13cf 100644 --- a/Source/Core/VideoCommon/Src/VertexLoader_TextCoord.cpp +++ b/Source/Core/VideoCommon/Src/VertexLoader_TextCoord.cpp @@ -18,7 +18,9 @@ #ifndef VERTEXLOADER_TEXCOORD_H #define VERTEXLOADER_TEXCOORD_H +#include <nmmintrin.h> #include "Common.h" +#include "CPUDetect.h" #include "VideoCommon.h" #include "VertexLoader.h" #include "VertexLoader_Position.h" @@ -308,15 +310,30 @@ void LOADERDECL TexCoord_ReadIndex16_Float1() VertexManager::s_pCurBufferPointer += 4; tcIndex++; } -void LOADERDECL TexCoord_ReadIndex16_Float2() + +static const __m128i kMaskSwap32 = _mm_set_epi32(0xFFFFFFFFL, 0xFFFFFFFFL, 0x04050607L, 0x00010203L); +void LOADERDECL TexCoord_ReadIndex16_Float2() { u16 Index = DataReadU16(); const u32 *pData = (const u32 *)(cached_arraybases[ARRAY_TEXCOORD0+tcIndex] + (Index * arraystrides[ARRAY_TEXCOORD0+tcIndex])); - ((u32*)VertexManager::s_pCurBufferPointer)[0] = Common::swap32(pData[0]); - ((u32*)VertexManager::s_pCurBufferPointer)[1] = Common::swap32(pData[1]); - LOG_TEX2(); - VertexManager::s_pCurBufferPointer += 8; - tcIndex++; + + if (cpu_info.bSSSE3) { + const __m128i a = _mm_loadl_epi64((__m128i*)pData); + const __m128i b = _mm_shuffle_epi8(a, kMaskSwap32); + u8* p = VertexManager::s_pCurBufferPointer; + _mm_storel_epi64((__m128i*)p, b); + LOG_TEX2(); + p += 8; + VertexManager::s_pCurBufferPointer = p; + tcIndex++; + + } else { + ((u32*)VertexManager::s_pCurBufferPointer)[0] = Common::swap32(pData[0]); + ((u32*)VertexManager::s_pCurBufferPointer)[1] = Common::swap32(pData[1]); + LOG_TEX2(); + VertexManager::s_pCurBufferPointer += 8; + tcIndex++; + } } #endif |