MLIR  20.0.0git
Public Member Functions | List of all members
mlir::arith::WideIntEmulationConverter Class Reference

Converts integer types that are too wide for the target by splitting them in two halves and thus turning into supported ones, i.e., i2*N --> iN, where N is the widest integer bitwidth supported by the target. More...

#include "mlir/Dialect/Arith/Transforms/WideIntEmulationConverter.h"

+ Inheritance diagram for mlir::arith::WideIntEmulationConverter:

Public Member Functions

 WideIntEmulationConverter (unsigned widestIntSupportedByTarget)
 
unsigned getMaxTargetIntBitWidth () const
 
- Public Member Functions inherited from mlir::TypeConverter
virtual ~TypeConverter ()=default
 
 TypeConverter ()=default
 
 TypeConverter (const TypeConverter &other)
 
TypeConverteroperator= (const TypeConverter &other)
 
template<typename FnT , typename T = typename llvm::function_traits< std::decay_t<FnT>>::template arg_t<0>>
void addConversion (FnT &&callback)
 Register a conversion function. More...
 
template<typename FnT , typename T = typename llvm::function_traits< std::decay_t<FnT>>::template arg_t<1>>
void addArgumentMaterialization (FnT &&callback)
 All of the following materializations require function objects that are convertible to the following form: Value(OpBuilder &, T, ValueRange, Location), where T is any subclass of Type. More...
 
template<typename FnT , typename T = typename llvm::function_traits< std::decay_t<FnT>>::template arg_t<1>>
void addSourceMaterialization (FnT &&callback)
 This method registers a materialization that will be called when converting a legal replacement value back to an illegal source type. More...
 
template<typename FnT , typename T = typename llvm::function_traits< std::decay_t<FnT>>::template arg_t<1>>
void addTargetMaterialization (FnT &&callback)
 This method registers a materialization that will be called when converting an illegal (source) value to a legal (target) type. More...
 
template<typename FnT , typename T = typename llvm::function_traits<std::decay_t<FnT>>::template arg_t<0>, typename A = typename llvm::function_traits<std::decay_t<FnT>>::template arg_t<1>>
void addTypeAttributeConversion (FnT &&callback)
 Register a conversion function for attributes within types. More...
 
LogicalResult convertType (Type t, SmallVectorImpl< Type > &results) const
 Convert the given type. More...
 
Type convertType (Type t) const
 This hook simplifies defining 1-1 type conversions. More...
 
template<typename TargetType >
TargetType convertType (Type t) const
 Attempts a 1-1 type conversion, expecting the result type to be TargetType. More...
 
LogicalResult convertTypes (TypeRange types, SmallVectorImpl< Type > &results) const
 Convert the given set of types, filling 'results' as necessary. More...
 
bool isLegal (Type type) const
 Return true if the given type is legal for this type converter, i.e. More...
 
template<typename RangeT >
std::enable_if_t<!std::is_convertible< RangeT, Type >::value &&!std::is_convertible< RangeT, Operation * >::value, bool > isLegal (RangeT &&range) const
 Return true if all of the given types are legal for this type converter. More...
 
bool isLegal (Operation *op) const
 Return true if the given operation has legal operand and result types. More...
 
bool isLegal (Region *region) const
 Return true if the types of block arguments within the region are legal. More...
 
bool isSignatureLegal (FunctionType ty) const
 Return true if the inputs and outputs of the given function type are legal. More...
 
LogicalResult convertSignatureArg (unsigned inputNo, Type type, SignatureConversion &result) const
 This method allows for converting a specific argument of a signature. More...
 
LogicalResult convertSignatureArgs (TypeRange types, SignatureConversion &result, unsigned origInputOffset=0) const
 
std::optional< SignatureConversionconvertBlockSignature (Block *block) const
 This function converts the type signature of the given block, by invoking 'convertSignatureArg' for each argument. More...
 
Value materializeArgumentConversion (OpBuilder &builder, Location loc, Type resultType, ValueRange inputs) const
 Materialize a conversion from a set of types into one result type by generating a cast sequence of some kind. More...
 
Value materializeSourceConversion (OpBuilder &builder, Location loc, Type resultType, ValueRange inputs) const
 
Value materializeTargetConversion (OpBuilder &builder, Location loc, Type resultType, ValueRange inputs, Type originalType={}) const
 
SmallVector< ValuematerializeTargetConversion (OpBuilder &builder, Location loc, TypeRange resultType, ValueRange inputs, Type originalType={}) const
 
std::optional< AttributeconvertTypeAttribute (Type type, Attribute attr) const
 Convert an attribute present attr from within the type type using the registered conversion functions. More...
 

Detailed Description

Converts integer types that are too wide for the target by splitting them in two halves and thus turning into supported ones, i.e., i2*N --> iN, where N is the widest integer bitwidth supported by the target.

Currently, we only handle power-of-two integer types and support conversions of integers twice as wide as the maximum supported by the target. Wide integers are represented as vectors, e.g., i64 --> vector<2xi32>, where the first element is the low half of the original integer, and the second element the high half.

Definition at line 23 of file WideIntEmulationConverter.h.

Constructor & Destructor Documentation

◆ WideIntEmulationConverter()

arith::WideIntEmulationConverter::WideIntEmulationConverter ( unsigned  widestIntSupportedByTarget)
explicit

Member Function Documentation

◆ getMaxTargetIntBitWidth()

unsigned mlir::arith::WideIntEmulationConverter::getMaxTargetIntBitWidth ( ) const
inline

The documentation for this class was generated from the following files: