mlir::fxpmath::detail::UniformBinaryOpInfo Struct Reference

#include <UniformKernelUtils.h>

Collaboration diagram for mlir::fxpmath::detail::UniformBinaryOpInfo:

Public Member Functions
	UniformBinaryOpInfo (Operation *op, Value lhs, Value rhs, Optional< APFloat > clampMin, Optional< APFloat > clampMax)
bool	isValid () const
	Returns whether this info is valid (all types defined, etc). More...
Type	getQuantizedResultType () const
	Gets the final quantized result type of the result. More...
bool	isSameStorageType () const
	Returns whether the storage type of all operands is identical. More...
bool	isFixedPointPOT (int &lhsLog2Scale, int &rhsLog2Scale, int &resultLog2Scale) const
std::pair< IntegerAttr, IntegerAttr >	getClampMinMax (IntegerType ty) const
	Gets the result integer clamp range given the result quantized type. More...

Public Attributes
Operation *	op
Value	lhs
Value	rhs
Optional< APFloat >	clampMin
Optional< APFloat >	clampMax
quant::UniformQuantizedType	lhsType
quant::UniformQuantizedType	rhsType
quant::UniformQuantizedType	resultType
Type	lhsStorageType
Type	rhsStorageType
Type	resultStorageType

Detailed Description

Helper class for operating on binary operations where all operands and the result are a UniformQuantizedType.

Constructor & Destructor Documentation

UniformBinaryOpInfo()

mlir::fxpmath::detail::UniformBinaryOpInfo::UniformBinaryOpInfo ( Operation *  op, Value  lhs, Value  rhs, Optional< APFloat >  clampMin, Optional< APFloat >  clampMax  )

inline

Member Function Documentation

getClampMinMax()

std::pair<IntegerAttr, IntegerAttr> mlir::fxpmath::detail::UniformBinaryOpInfo::getClampMinMax ( IntegerType  ty ) const

inline

Gets the result integer clamp range given the result quantized type.

getQuantizedResultType()

Type mlir::fxpmath::detail::UniformBinaryOpInfo::getQuantizedResultType ( ) const

inline

Gets the final quantized result type of the result.

isFixedPointPOT()

bool mlir::fxpmath::detail::UniformBinaryOpInfo::isFixedPointPOT ( int &  lhsLog2Scale, int &  rhsLog2Scale, int &  resultLog2Scale  ) const

inline

Returns whether all operands and result are considered fixedpoint power of two, setting the lhs, rhs, and result log2 scale references.

isSameStorageType()

bool mlir::fxpmath::detail::UniformBinaryOpInfo::isSameStorageType ( ) const

inline

Returns whether the storage type of all operands is identical.

isValid()

bool mlir::fxpmath::detail::UniformBinaryOpInfo::isValid ( ) const

inline

Returns whether this info is valid (all types defined, etc).

Member Data Documentation

clampMax

Optional<APFloat> mlir::fxpmath::detail::UniformBinaryOpInfo::clampMax

clampMin

Optional<APFloat> mlir::fxpmath::detail::UniformBinaryOpInfo::clampMin

lhs

Value mlir::fxpmath::detail::UniformBinaryOpInfo::lhs

lhsStorageType

Type mlir::fxpmath::detail::UniformBinaryOpInfo::lhsStorageType

lhsType

quant::UniformQuantizedType mlir::fxpmath::detail::UniformBinaryOpInfo::lhsType

op

Operation* mlir::fxpmath::detail::UniformBinaryOpInfo::op

resultStorageType

Type mlir::fxpmath::detail::UniformBinaryOpInfo::resultStorageType

resultType

quant::UniformQuantizedType mlir::fxpmath::detail::UniformBinaryOpInfo::resultType

rhs

Value mlir::fxpmath::detail::UniformBinaryOpInfo::rhs

rhsStorageType

Type mlir::fxpmath::detail::UniformBinaryOpInfo::rhsStorageType

rhsType

quant::UniformQuantizedType mlir::fxpmath::detail::UniformBinaryOpInfo::rhsType

---

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

• lib/Dialect/FxpMathOps/Transforms/UniformKernelUtils.h