|
MLIR 24.0.0git
|
Convert math ops marked with fast (afn) to native OpenCL intrinsics. More...
Public Member Functions | |
| ConvertNativeFuncPattern (MLIRContext *context, StringRef nativeFunc, PatternBenefit benefit=1) | |
| std::string | getMangledNativeFuncName (const ArrayRef< Type > operandTypes) const |
| LogicalResult | matchAndRewrite (Op op, typename Op::Adaptor adaptor, ConversionPatternRewriter &rewriter) const override |
Public Attributes | |
| const StringRef | nativeFunc |
Convert math ops marked with fast (afn) to native OpenCL intrinsics.
Definition at line 63 of file MathToXeVM.cpp.
|
inline |
Definition at line 65 of file MathToXeVM.cpp.
References nativeFunc.
|
inline |
Definition at line 70 of file MathToXeVM.cpp.
References nativeFunc.
Referenced by matchAndRewrite().
|
inlineoverride |
Definition at line 95 of file MathToXeVM.cpp.
References mlir::arith::AttrConvertFastMathToLLVM< SourceOp, TargetOp >::getAttrs(), mlir::OpState::getLoc(), getMangledNativeFuncName(), mlir::NamedAttribute::getName(), mlir::NamedAttribute::getValue(), isSizeOneVector(), isSPIRVCompatibleFloatOrVec(), mlir::LLVM::lookupOrCreateFn(), and success().
| const StringRef ConvertNativeFuncPattern< Op >::nativeFunc |
Definition at line 160 of file MathToXeVM.cpp.
Referenced by ConvertNativeFuncPattern(), and getMangledNativeFuncName().