xutaxkamay/css_enhanced_waf
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity

Autovectorize when possible

Browse source
This commit is contained in:
unknown 2024-08-28 20:32:22 +02:00
parent cd9787a4bd
commit ff2be3b56e
1 changed files with 40 additions and 85 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

125
materialsystem/cmatqueuedrendercontext.cpp
View file

@ -6,6 +6,7 @@
#include "pch_materialsystem.h"
#include "platform.h"
#include "tier1/functors.h"
#include "itextureinternal.h"
@ -24,80 +25,6 @@ ConVar mat_report_queue_status( "mat_report_queue_status", "0", FCVAR_MATERIAL_S
//
//-----------------------------------------------------------------------------
#if defined( _WIN32 )
void FastCopy( byte *pDest, const byte *pSrc, size_t nBytes )
{
if ( !nBytes )
{
return;
}
#if !defined( _X360 )
if ( (size_t)pDest % 16 == 0 && (size_t)pSrc % 16 == 0 )
{
const int BYTES_PER_FULL = 128;
int nBytesFull = nBytes - ( nBytes % BYTES_PER_FULL );
for ( byte *pLimit = pDest + nBytesFull; pDest < pLimit; pDest += BYTES_PER_FULL, pSrc += BYTES_PER_FULL )
{
#ifdef __i386__
__asm
{
mov esi, pSrc
mov edi, pDest
movaps xmm0, [esi + 0]
movaps xmm1, [esi + 16]
movaps xmm2, [esi + 32]
movaps xmm3, [esi + 48]
movaps xmm4, [esi + 64]
movaps xmm5, [esi + 80]
movaps xmm6, [esi + 96]
movaps xmm7, [esi + 112]
movntps [edi + 0], xmm0
movntps [edi + 16], xmm1
movntps [edi + 32], xmm2
movntps [edi + 48], xmm3
movntps [edi + 64], xmm4
movntps [edi + 80], xmm5
movntps [edi + 96], xmm6
movntps [edi + 112], xmm7
}
#else
memcpy( pDest, pSrc, BYTES_PER_FULL);
#endif
}
nBytes -= nBytesFull;
}
if ( nBytes )
{
memcpy( pDest, pSrc, nBytes );
}
#else
if ( (size_t)pDest % 4 == 0 && nBytes % 4 == 0 )
{
XMemCpyStreaming_WriteCombined( pDest, pSrc, nBytes );
}
else
{
// work around a bug in memcpy
if ((size_t)pDest % 2 == 0 && nBytes == 4)
{
*(reinterpret_cast<short *>(pDest)) = *(reinterpret_cast<const short *>(pSrc));
*(reinterpret_cast<short *>(pDest)+1) = *(reinterpret_cast<const short *>(pSrc)+1);
}
else
{
memcpy( pDest, pSrc, nBytes );
}
}
#endif
}
#else
#define FastCopy memcpy
#endif
//-----------------------------------------------------------------------------
//
//-----------------------------------------------------------------------------
@ -440,10 +367,7 @@ public:
}
Assert( pDest );
if ( pDest )
{
FastCopy( (byte *)pDest, pVertexData, nBytesVerts );
}
memcpy( (byte *)pDest, pVertexData, nBytesVerts );
}
if ( pIndexData && pIndexData != &gm_ScratchIndexBuffer[0] && desc.m_nIndexSize )
@ -451,24 +375,55 @@ public:
if ( !desc.m_nFirstVertex )
{
// AssertMsg(desc.m_pIndices & 0x03 == 0,"desc.m_pIndices is misaligned in CMatQueuedMesh::ExecuteDefferedBuild\n");
FastCopy( (byte *)desc.m_pIndices, (byte *)pIndexData, nIndices * sizeof(*pIndexData) );
// memcpy( (byte *)desc.m_pIndices, (byte *)pIndexData, nIndices * sizeof(*pIndexData) );
// Let it autovectorize.
for (int i = 0; i < nIndices; i++)
{
desc.m_pIndices[i] = pIndexData[i];
}
}
else
{
static ALIGN16 uint16 tempIndices[256];
static constexpr auto INDICES_TO_AUTOVECTORIZE = 256;
// original method
int i = 0;
int firstVertex = desc.m_nFirstVertex;
uint16 DECL_ALIGN(512) firstVertexMatrix[INDICES_TO_AUTOVECTORIZE];
for (int i = 0; i < INDICES_TO_AUTOVECTORIZE; i++)
{
firstVertexMatrix[i] = firstVertex;
}
while ( i < nIndices )
{
int nToCopy = min( (int)ARRAYSIZE(tempIndices), nIndices - i );
for ( int j = 0; j < nToCopy; j++ )
int nToCopy = nIndices - i;
auto pSrc = pIndexData + i;
auto pDst = desc.m_pIndices + i;
if (nToCopy < INDICES_TO_AUTOVECTORIZE)
{
tempIndices[j] = pIndexData[i+j] + desc.m_nFirstVertex;
for (int j = 0; j < nToCopy; j++)
{
pDst[j] = pSrc[j] + firstVertexMatrix[j];
}
i += nToCopy;
}
else
{
for (int j = 0; j < INDICES_TO_AUTOVECTORIZE; j++)
{
pDst[j] = pSrc[j] + firstVertexMatrix[j];
}
i += INDICES_TO_AUTOVECTORIZE;
}
FastCopy( (byte *)(desc.m_pIndices + i), (byte *)tempIndices, nToCopy * sizeof(uint16) );
i += nToCopy;
}
}
}