mirrors
/
dxvk
mirror of https://github.com/doitsujin/dxvk
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

[dxso] Add bit selectors to D3D9ShaderSpecConstantManager 

Avoids doing double bitfield extractions which isn't
optimized right now.
This commit is contained in:
Philip Rebohle 2022-08-01 16:31:13 +02:00 committed by Joshie
parent 9619377723
commit d498551a23
2 changed files with 32 additions and 41 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

15
src/d3d9/d3d9_spec_constants.h
View File

@ -90,17 +90,22 @@ namespace dxvk {
class D3D9ShaderSpecConstantManager {
public:
uint32_t get(SpirvModule &module, D3D9SpecConstantId id) {
return get(module, id, 0, 32);
}
uint32_t get(SpirvModule &module, D3D9SpecConstantId id, uint32_t bitOffset, uint32_t bitCount) {
const auto &layout = D3D9SpecializationInfo::Layout[id];
uint32_t val = getSpecConstDword(module, layout.dwordOffset);
if (layout.sizeInBits == 32)
bitCount = std::min(bitCount, layout.sizeInBits - bitOffset);
if (bitCount == 32)
return val;
return module.opBitFieldUExtract(
module.defIntType(32, 0),
val,
module.consti32(layout.bitOffset),
module.consti32(layout.sizeInBits));
module.defIntType(32, 0), val,
module.consti32(bitOffset + layout.bitOffset),
module.consti32(bitCount));
}
private:

58
src/dxso/dxso_compiler.cpp
View File

@ -1015,7 +1015,7 @@ namespace dxvk {
uint32_t uvec4Type = getVectorTypeId({ DxsoScalarType::Uint32, 4 });
// If not SWVP, spec const this
uint32_t bitfield;
uint32_t bit;
if (m_layout->bitmaskCount != 1) {
std::array<uint32_t, 2> indices = { m_module.constu32(0), m_module.constu32(reg.id.num / 128) };
@ -1026,23 +1026,22 @@ namespace dxvk {
m_module.defPointerType(accessType, spv::StorageClassUniform),
m_cBoolBuffer, indexCount, indices.data());
bitfield = m_module.opLoad(accessType, ptrId);
}
else {
bitfield = m_spec.get(m_module,
m_programInfo.type() == DxsoProgramType::VertexShader
? SpecVertexShaderBools
: SpecPixelShaderBools);
}
uint32_t bitfield = m_module.opLoad(accessType, ptrId);
uint32_t bitIdx = m_module.consti32(reg.id.num % 32);
uint32_t bitIdx = m_module.consti32(reg.id.num % 32);
if (m_layout->bitmaskCount != 1) {
uint32_t index = (reg.id.num % 128) / 32;
bitfield = m_module.opCompositeExtract(uintType, bitfield, 1, &index);
bit = m_module.opBitFieldUExtract(
uintType, bitfield, bitIdx, m_module.consti32(1));
}
else {
bit = m_spec.get(m_module,
m_programInfo.type() == DxsoProgramType::VertexShader
? SpecVertexShaderBools
: SpecPixelShaderBools,
reg.id.num, 1);
}
uint32_t bit = m_module.opBitFieldUExtract(
uintType, bitfield, bitIdx, m_module.consti32(1));
result.id = m_module.opINotEqual(
getVectorTypeId(result.type),
@ -2762,10 +2761,7 @@ void DxsoCompiler::emitControlFlowGenericLoop(
// Of course it does...
uint32_t bool_t = m_module.defBoolType();
uint32_t shouldProj = m_module.opBitFieldUExtract(
m_module.defIntType(32, 0), m_spec.get(m_module, SpecProjectionType),
m_module.consti32(samplerIdx), m_module.consti32(1));
uint32_t shouldProj = m_spec.get(m_module, SpecProjectionType, samplerIdx, 1);
shouldProj = m_module.opINotEqual(bool_t, shouldProj, m_module.constu32(0));
uint32_t bvec4_t = m_module.defVectorType(bool_t, 4);
@ -2929,9 +2925,7 @@ void DxsoCompiler::emitControlFlowGenericLoop(
uint32_t fetch4 = 0;
if (m_programInfo.type() == DxsoProgramType::PixelShader && samplerType != SamplerTypeTexture3D) {
fetch4 = m_module.opBitFieldUExtract(
m_module.defIntType(32, 0), m_spec.get(m_module, SpecFetch4),
m_module.consti32(samplerIdx), m_module.consti32(1));
fetch4 = m_spec.get(m_module, SpecFetch4, samplerIdx, 1);
uint32_t bool_t = m_module.defBoolType();
fetch4 = m_module.opINotEqual(bool_t, fetch4, m_module.constu32(0));
@ -2962,10 +2956,7 @@ void DxsoCompiler::emitControlFlowGenericLoop(
fetch4,
imageOperands);
uint32_t shouldProj = m_module.opBitFieldUExtract(
m_module.defIntType(32, 0), m_spec.get(m_module, SpecProjectionType),
m_module.consti32(samplerIdx), m_module.consti32(1));
uint32_t shouldProj = m_spec.get(m_module, SpecProjectionType, samplerIdx, 1);
shouldProj = m_module.opINotEqual(m_module.defBoolType(), shouldProj, m_module.constu32(0));
// Depth -> .x
@ -3022,11 +3013,11 @@ void DxsoCompiler::emitControlFlowGenericLoop(
};
auto SampleType = [&](DxsoSamplerType samplerType) {
uint32_t typeId = m_module.defIntType(32, 0);
uint32_t offset = m_module.consti32(m_programInfo.type() == DxsoProgramTypes::VertexShader ? samplerIdx + caps::MaxTexturesPS : samplerIdx);
uint32_t bitCnt = m_module.consti32(1);
uint32_t bitOffset = m_programInfo.type() == DxsoProgramTypes::VertexShader
? samplerIdx + caps::MaxTexturesPS
: samplerIdx;
uint32_t isNull = m_module.opBitFieldUExtract(typeId, m_spec.get(m_module, SpecSamplerNull), offset, bitCnt);
uint32_t isNull = m_spec.get(m_module, SpecSamplerNull, bitOffset, 1);
isNull = m_module.opINotEqual(m_module.defBoolType(), isNull, m_module.constu32(0));
// Only do the check for depth comp. samplers
@ -3036,7 +3027,7 @@ void DxsoCompiler::emitControlFlowGenericLoop(
uint32_t depthLabel = m_module.allocateId();
uint32_t endLabel = m_module.allocateId();
uint32_t isDepth = m_module.opBitFieldUExtract(typeId, m_spec.get(m_module, SpecSamplerDepthMode), offset, bitCnt);
uint32_t isDepth = m_spec.get(m_module, SpecSamplerDepthMode, bitOffset, 1);
isDepth = m_module.opINotEqual(m_module.defBoolType(), isDepth, m_module.constu32(0));
m_module.opSelectionMerge(endLabel, spv::SelectionControlMaskNone);
@ -3071,12 +3062,7 @@ void DxsoCompiler::emitControlFlowGenericLoop(
}};
uint32_t switchEndLabel = m_module.allocateId();
uint32_t typeId = m_module.defIntType(32, 0);
uint32_t offset = m_module.consti32(samplerIdx * 2);
uint32_t bitCnt = m_module.consti32(2);
uint32_t type = m_module.opBitFieldUExtract(typeId, m_spec.get(m_module, SpecSamplerType), offset, bitCnt);
uint32_t type = m_spec.get(m_module, SpecSamplerType, samplerIdx * 2, 2);
m_module.opSelectionMerge(switchEndLabel, spv::SelectionControlMaskNone);
m_module.opSwitch(type,