VectorizationState Struct Reference

Contains the vectorization state and related methods used across the vectorization process of a given operation. More...

Public Member Functions
	VectorizationState (RewriterBase &rewriter)
LogicalResult	initState (RewriterBase &rewriter, LinalgOp linalgOp, ArrayRef< int64_t > inputVectorSizes, ArrayRef< bool > inputScalableVecDims, bool assumeDynamicDimsMatchVecSizes=false)
	Initializes the vectorization state, including the computation of the canonical vector shape for vectorization.
ArrayRef< int64_t >	getCanonicalVecShape () const
	Returns the canonical vector shape used to vectorize the iteration space.
ArrayRef< bool >	getScalableVecDims () const
	Returns the vector dimensions that are scalable in the canonical vector shape.
VectorType	getCanonicalVecType (Type elementType, std::optional< AffineMap > dimPermutation=std::nullopt) const
	Returns a vector type of the provided elementType with the canonical vector shape and the corresponding fixed/scalable dimensions bit.
Operation *	maskOperation (RewriterBase &rewriter, Operation *opToMask, LinalgOp linalgOp, std::optional< AffineMap > maybeIndexingMap=std::nullopt)
	Masks an operation with the canonical vector mask if the operation needs masking.
Value	getOrCreateMaskFor (RewriterBase &rewriter, LinalgOp linalgOp, std::optional< AffineMap > maybeMaskingMap=std::nullopt)
	Create or retrieve an existing mask value for the canonical vector iteration space.

Detailed Description

Contains the vectorization state and related methods used across the vectorization process of a given operation.

Definition at line 206 of file Vectorization.cpp.

Constructor & Destructor Documentation

VectorizationState()

VectorizationState::VectorizationState ( RewriterBase & rewriter )

inline

Definition at line 207 of file Vectorization.cpp.

Member Function Documentation

getCanonicalVecShape()

ArrayRef< int64_t > VectorizationState::getCanonicalVecShape ( ) const

inline

Returns the canonical vector shape used to vectorize the iteration space.

Definition at line 217 of file Vectorization.cpp.

getCanonicalVecType()

VectorType VectorizationState::getCanonicalVecType ( Type elementType, std::optional< AffineMap > dimPermutation = std::nullopt ) const

inline

Returns a vector type of the provided elementType with the canonical vector shape and the corresponding fixed/scalable dimensions bit.

If dimPermutation is provided, the canonical vector dimensions are permuted accordingly.

Definition at line 227 of file Vectorization.cpp.

References mlir::applyPermutationMap(), and vectorShape().

Referenced by getOrCreateMaskFor().

getOrCreateMaskFor()

Value VectorizationState::getOrCreateMaskFor	(	RewriterBase &	rewriter,
		LinalgOp	linalgOp,
		std::optional< AffineMap >	maybeMaskingMap = std::nullopt )

Create or retrieve an existing mask value for the canonical vector iteration space.

If maybeMaskingMap the mask is permuted using that permutation map. If a new mask is created, it will be cached for future users. Returns a null value if no mask is needed.

Definition at line 410 of file Vectorization.cpp.

References mlir::applyPermutationMap(), getCanonicalVecType(), mlir::Builder::getContext(), mlir::Builder::getI1Type(), mlir::AffineMap::getMultiDimIdentityMap(), and getOrCreateMaskFor().

Referenced by getOrCreateMaskFor(), and maskOperation().

getScalableVecDims()

ArrayRef< bool > VectorizationState::getScalableVecDims ( ) const

inline

Returns the vector dimensions that are scalable in the canonical vector shape.

Definition at line 221 of file Vectorization.cpp.

initState()

LogicalResult VectorizationState::initState	(	RewriterBase &	rewriter,
		LinalgOp	linalgOp,
		ArrayRef< int64_t >	inputVectorSizes,
		ArrayRef< bool >	inputScalableVecDims,
		bool	assumeDynamicDimsMatchVecSizes = false )

Initializes the vectorization state, including the computation of the canonical vector shape for vectorization.

Definition at line 368 of file Vectorization.cpp.

References initState(), mlir::OpBuilder::setInsertionPoint(), and success().

Referenced by initState().

maskOperation()

Operation * VectorizationState::maskOperation	(	RewriterBase &	rewriter,
		Operation *	opToMask,
		LinalgOp	linalgOp,
		std::optional< AffineMap >	maybeIndexingMap = std::nullopt )

Masks an operation with the canonical vector mask if the operation needs masking.

Returns the masked operation or the original operation if masking is not needed. If provided, the canonical mask for this operation is permuted using maybeIndexingMap.

Definition at line 488 of file Vectorization.cpp.

References mlir::Builder::getBoolArrayAttr(), getOrCreateMaskFor(), mlir::AffineMap::getPermutationMap(), mlir::AffineMap::getResult(), mlir::Operation::getResults(), mlir::vector::maskOperation(), maskOperation(), mlir::RewriterBase::modifyOpInPlace(), and mlir::RewriterBase::replaceAllUsesExcept().

Referenced by maskOperation().

---

The documentation for this struct was generated from the following file:

• lib/Dialect/Linalg/Transforms/Vectorization.cpp