piwalker/dolphin
1
0
Fork 0
mirror of https://github.com/dolphin-emu/dolphin.git synced 2025-07-23 06:09:50 -06:00
Code Issues Packages Projects Releases Wiki Activity

Reverted because of some processor and performance issue. I will develop in a branch about SSSE3/SSE4.1.

Browse Source 
git-svn-id: https://dolphin-emu.googlecode.com/svn/trunk@5123 8ced0084-cf51-0410-be5f-012b33b47a6e
This commit is contained in:
nodchip
2010-02-24 23:58:48 +00:00
parent 1b43900f30
commit 609151c6e8
4 changed files with 30 additions and 189 deletions
Download Patch File Download Diff File Expand all files Collapse all files

34
Source/Core/VideoCommon/Src/VertexLoader_Position.cpp
View File

@ -18,9 +18,7 @@
#ifndef VERTEXLOADER_POSITION_H
#define VERTEXLOADER_POSITION_H
#include <nmmintrin.h>
#include "Common.h"
#include "CPUDetect.h"
#include "VideoCommon.h"
#include "VertexLoader.h"
#include "VertexLoader_Position.h"
@ -151,34 +149,18 @@ inline void Pos_ReadIndex_Short(int Index)
VertexManager::s_pCurBufferPointer += 12;
}
static const __m128i kMaskSwap32 = _mm_set_epi32(0xFFFFFFFFL, 0x08090A0BL, 0x04050607L, 0x00010203L);
template<bool three>
inline void Pos_ReadIndex_Float(int Index)
{
const u32* pData = (const u32 *)(cached_arraybases[ARRAY_POSITION] + (Index * arraystrides[ARRAY_POSITION]));
if (cpu_info.bSSE4_1) {
const __m128i a = _mm_loadu_si128((__m128i*)pData);
__m128i b = _mm_shuffle_epi8(a, kMaskSwap32);
if (!three) {
b = _mm_insert_epi32(b, 0, 2);
}
u8* p = VertexManager::s_pCurBufferPointer;
_mm_storeu_si128((__m128i*)p, b);
LOG_VTX();
p += 12;
VertexManager::s_pCurBufferPointer = p;
} else {
((u32*)VertexManager::s_pCurBufferPointer)[0] = Common::swap32(pData[0]);
((u32*)VertexManager::s_pCurBufferPointer)[1] = Common::swap32(pData[1]);
if (three)
((u32*)VertexManager::s_pCurBufferPointer)[2] = Common::swap32(pData[2]);
else
((float*)VertexManager::s_pCurBufferPointer)[2] = 0.0f;
LOG_VTX();
VertexManager::s_pCurBufferPointer += 12;
}
((u32*)VertexManager::s_pCurBufferPointer)[0] = Common::swap32(pData[0]);
((u32*)VertexManager::s_pCurBufferPointer)[1] = Common::swap32(pData[1]);
if (three)
((u32*)VertexManager::s_pCurBufferPointer)[2] = Common::swap32(pData[2]);
else
((float*)VertexManager::s_pCurBufferPointer)[2] = 0.0f;
LOG_VTX();
VertexManager::s_pCurBufferPointer += 12;
}
// ==============================================================================