swr/rast: Use different handing for stream masks

Reviewed-by: Bruce Cherniak <bruce.cherniak@intel.com>
This commit is contained in:
George Kyriazis 2018-04-16 23:18:28 -05:00
parent 6b1c852ebc
commit 7986519d50
5 changed files with 11 additions and 6 deletions

View File

@ -209,6 +209,9 @@ unsigned char _BitScanReverse(unsigned int *Index, unsigned int Mask)
return (Mask != 0);
}
#define _BitScanForward64 _BitScanForward
#define _BitScanReverse64 _BitScanReverse
inline
void *AlignedMalloc(size_t size, size_t alignment)
{

View File

@ -976,14 +976,14 @@ void SetupPipeline(DRAW_CONTEXT *pDC)
if (pState->state.soState.soEnable)
{
uint32_t streamMasks = 0;
uint64_t streamMasks = 0;
for (uint32_t i = 0; i < 4; ++i)
{
streamMasks |= pState->state.soState.streamMasks[i];
}
DWORD maxAttrib;
if (_BitScanReverse(&maxAttrib, streamMasks))
if (_BitScanReverse64(&maxAttrib, streamMasks))
{
pState->state.feNumAttributes = std::max(pState->state.feNumAttributes, (uint32_t)(maxAttrib + 1));
}

View File

@ -528,10 +528,10 @@ static void StreamOut(
for (uint32_t primIndex = 0; primIndex < numPrims; ++primIndex)
{
DWORD slot = 0;
uint32_t soMask = soState.streamMasks[streamIndex];
uint64_t soMask = soState.streamMasks[streamIndex];
// Write all entries into primitive data buffer for SOS.
while (_BitScanForward(&slot, soMask))
while (_BitScanForward64(&slot, soMask))
{
simd4scalar attrib[MAX_NUM_VERTS_PER_PRIM]; // prim attribs (always 4 wide)
uint32_t paSlot = slot + soState.vertexAttribOffset[streamIndex];
@ -551,7 +551,7 @@ static void StreamOut(
_mm_store_ps((float*)pPrimDataAttrib, attrib[v]);
}
soMask &= ~(1 << slot);
soMask &= ~(uint64_t(1) << slot);
}
// Update pPrimData pointer

View File

@ -702,7 +702,7 @@ struct SWR_STREAMOUT_STATE
// The stream masks specify which attributes are sent to which streams.
// These masks help the FE to setup the pPrimData buffer that is passed
// the Stream Output Shader (SOS) function.
uint32_t streamMasks[MAX_SO_STREAMS];
uint64_t streamMasks[MAX_SO_STREAMS];
// Number of attributes, including position, per vertex that are streamed out.
// This should match number of bits in stream mask.

View File

@ -313,6 +313,7 @@ struct StreamOutJit : public Builder
JitManager::DumpToFile(soFunc, "SoFunc_optimized");
return soFunc;
}
};
@ -333,6 +334,7 @@ PFN_SO_FUNC JitStreamoutFunc(HANDLE hJitMgr, const HANDLE hFunc)
pJitMgr->DumpAsm(func, "SoFunc_optimized");
return pfnStreamOut;
}