diff options
| author | ArielG-NV <159081215+ArielG-NV@users.noreply.github.com> | 2024-08-26 15:11:41 -0400 |
|---|---|---|
| committer | GitHub <noreply@github.com> | 2024-08-26 12:11:41 -0700 |
| commit | e1c6fecd90142761aaecbf4e281beb87893fc531 (patch) | |
| tree | 30cf06cf6a7ea4a55b2d3e4ecd5d4f13f1f31e60 /source/slang/slang-emit-spirv.cpp | |
| parent | 76999788902a8c50e8e5d0e867763e5ea2f10042 (diff) | |
Implement `-fvk-use-dx-layout` (#4912)
* Implement `-fvk-use-dx-layout`
Fixes: #4126
Changes:
* Added fvk-use-dx-layout
* Modified `HLSLConstantBufferLayoutRulesImpl` for correctness (ex: Array is always 16 byte aligned)
* Added kFXCShaderResourceLayoutRulesFamilyImpl and kFXCConstantBufferLayoutRulesFamilyImpl to handle fvk-use-dx-layout
* Added `ConstantBufferLayoutRules` to manage constant buffer rules
* Added `alignCompositeElementOfNonAggregate`/`alignCompositeElementOfAggregate` to handle forced alignment of composites for ConstantBuffers
* `StructuredBuffer` rules are mostly equal to `scalar` layout, not much was needed to be changed to support this behavior.
* seperate legacy constant buffer and how Slang does constant-buffer normally
* undo an addition
* remove accidental test
* Address review and fix
Address review and remove GLSL support since GLSL requires a seperate legalization (need to linearlize structs like with `legalizeMetalIR` to assign explicit offsets)
* comments
* remove aggregate and non-aggregate logic
We don't need this distinction for the logic
---------
Co-authored-by: Yong He <yonghe@outlook.com>
Diffstat (limited to 'source/slang/slang-emit-spirv.cpp')
| -rw-r--r-- | source/slang/slang-emit-spirv.cpp | 18 |
1 files changed, 9 insertions, 9 deletions
diff --git a/source/slang/slang-emit-spirv.cpp b/source/slang/slang-emit-spirv.cpp index 20cbbac02..8f4501c2b 100644 --- a/source/slang/slang-emit-spirv.cpp +++ b/source/slang/slang-emit-spirv.cpp @@ -3993,7 +3993,7 @@ struct SPIRVEmitContext auto rule = IRTypeLayoutRules::get(layoutRuleName); IRSizeAndAlignment elementSizeAlignment; getSizeAndAlignment(m_targetProgram->getOptionSet(), rule, matrixType->getElementType(), &elementSizeAlignment); - + IRIntegerValue matrixMinorVectorCount = 0; // Reminder: the meaning of row/column major layout // in our semantics is the *opposite* of what GLSL/SPIRV // calls them, because what they call "columns" @@ -4007,10 +4007,7 @@ struct SPIRVEmitContext spvStructID, SpvLiteralInteger::from32(id), SpvDecorationRowMajor); - - auto vectorSize = rule->getVectorSizeAndAlignment(elementSizeAlignment, getIntVal(matrixType->getRowCount())); - vectorSize = rule->alignCompositeElement(vectorSize); - matrixStride = vectorSize.getStride(); + matrixMinorVectorCount = getIntVal(matrixType->getRowCount()); } else { @@ -4020,12 +4017,15 @@ struct SPIRVEmitContext spvStructID, SpvLiteralInteger::from32(id), SpvDecorationColMajor); - - auto vectorSize = rule->getVectorSizeAndAlignment(elementSizeAlignment, getIntVal(matrixType->getColumnCount())); - vectorSize = rule->alignCompositeElement(vectorSize); - matrixStride = vectorSize.getStride(); + matrixMinorVectorCount = getIntVal(matrixType->getColumnCount()); } + // We need the size of our vector. To get the stride we need to know how 'big' + // each vector element is inside an array, due to this we align our vector + // as if a composite. + auto vectorSize = rule->getVectorSizeAndAlignment(elementSizeAlignment, matrixMinorVectorCount); + vectorSize = rule->alignCompositeElement(vectorSize); + matrixStride = vectorSize.getStride(); emitOpMemberDecorateMatrixStride( getSection(SpvLogicalSectionID::Annotations), nullptr, |
