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

[dxbc] Serialize hull shader fork/join phase invocations 

We cannot run these in parallel in case the hull shader's output vertex
count, and thus the invocation count, is less than the fork/join phase
invocation count.
This commit is contained in:
Philip Rebohle 2018-04-02 16:22:19 +02:00
parent 2973c18055
commit eb85f61084
1 changed files with 11 additions and 16 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

27
src/dxbc/dxbc_compiler.cpp
View File

@ -5620,15 +5620,15 @@ namespace dxvk {
this->emitHsControlPointPhase(m_hs.cpPhase);
this->emitHsPhaseBarrier();
// Fork-join phases (will run in parallel)
// Fork-join phases and output setup
this->emitHsInvocationBlockBegin(1);
for (const auto& phase : m_hs.forkPhases)
this->emitHsForkJoinPhase(phase);
for (const auto& phase : m_hs.joinPhases)
this->emitHsForkJoinPhase(phase);
// Output setup phase
this->emitHsInvocationBlockBegin(1);
this->emitOutputSetup();
this->emitHsInvocationBlockEnd();
this->emitMainFunctionEnd();
@ -5692,19 +5692,14 @@ namespace dxvk {
void DxbcCompiler::emitHsForkJoinPhase(
const DxbcCompilerHsForkJoinPhase& phase) {
this->emitHsInvocationBlockBegin(phase.instanceCount);
uint32_t invocationId = m_module.opLoad(
getScalarTypeId(DxbcScalarType::Uint32),
m_hs.builtinInvocationId);
m_module.opFunctionCall(
m_module.defVoidType(),
phase.functionId, 1,
&invocationId);
this->emitHsInvocationBlockEnd();
this->emitHsPhaseBarrier();
for (uint32_t i = 0; i < phase.instanceCount; i++) {
uint32_t invocationId = m_module.constu32(i);
m_module.opFunctionCall(
m_module.defVoidType(),
phase.functionId, 1,
&invocationId);
}
}