Safe Haskell	Safe-Inferred

LLVM.General.AST.Constant

Description

A representation of LLVM constants

Synopsis

Documentation

data Constant

http://llvm.org/docs/LangRef.html#constants

N.B. - http://llvm.org/docs/LangRef.html#constant-expressions

Although constant expressions and instructions have many similarites, there are important differences - so they're represented using different types in this AST. At the cost of making it harder to move an code back and forth between being constant and not, this approach embeds more of the rules of what IR is legal into the Haskell types.

Constructors

Int
Fields integerBits :: Word32 integerValue :: Integer
Float
Fields floatValue :: SomeFloat
Null
Fields constantType :: Type
Struct
Fields isPacked :: Bool memberValues :: [Constant]
Array
Fields memberType :: Type memberValues :: [Constant]
Vector
Fields memberValues :: [Constant]
Undef
Fields constantType :: Type
BlockAddress
Fields blockAddressFunction :: Name blockAddressBlock :: Name
GlobalReference Name
Add
Fields nsw :: Bool nuw :: Bool operand0 :: Constant operand1 :: Constant
FAdd
Fields operand0 :: Constant operand1 :: Constant
Sub
Fields nsw :: Bool nuw :: Bool operand0 :: Constant operand1 :: Constant
FSub
Fields operand0 :: Constant operand1 :: Constant
Mul
Fields nsw :: Bool nuw :: Bool operand0 :: Constant operand1 :: Constant
FMul
Fields operand0 :: Constant operand1 :: Constant
UDiv
Fields exact :: Bool operand0 :: Constant operand1 :: Constant
SDiv
Fields exact :: Bool operand0 :: Constant operand1 :: Constant
FDiv
Fields operand0 :: Constant operand1 :: Constant
URem
Fields operand0 :: Constant operand1 :: Constant
SRem
Fields operand0 :: Constant operand1 :: Constant
FRem
Fields operand0 :: Constant operand1 :: Constant
Shl
Fields nsw :: Bool nuw :: Bool operand0 :: Constant operand1 :: Constant
LShr
Fields exact :: Bool operand0 :: Constant operand1 :: Constant
AShr
Fields exact :: Bool operand0 :: Constant operand1 :: Constant
And
Fields operand0 :: Constant operand1 :: Constant
Or
Fields operand0 :: Constant operand1 :: Constant
Xor
Fields operand0 :: Constant operand1 :: Constant
GetElementPtr
Fields inBounds :: Bool address :: Constant indices :: [Constant]
Trunc
Fields operand0 :: Constant type' :: Type
ZExt
Fields operand0 :: Constant type' :: Type
SExt
Fields operand0 :: Constant type' :: Type
FPToUI
Fields operand0 :: Constant type' :: Type
FPToSI
Fields operand0 :: Constant type' :: Type
UIToFP
Fields operand0 :: Constant type' :: Type
SIToFP
Fields operand0 :: Constant type' :: Type
FPTrunc
Fields operand0 :: Constant type' :: Type
FPExt
Fields operand0 :: Constant type' :: Type
PtrToInt
Fields operand0 :: Constant type' :: Type
IntToPtr
Fields operand0 :: Constant type' :: Type
BitCast
Fields operand0 :: Constant type' :: Type
ICmp
Fields iPredicate :: IntegerPredicate operand0 :: Constant operand1 :: Constant
FCmp
Fields fpPredicate :: FloatingPointPredicate operand0 :: Constant operand1 :: Constant
Select
Fields condition' :: Constant trueValue :: Constant falseValue :: Constant
ExtractElement
Fields vector :: Constant index :: Constant
InsertElement
Fields vector :: Constant element :: Constant index :: Constant
ShuffleVector
Fields operand0 :: Constant operand1 :: Constant mask :: Constant
ExtractValue
Fields aggregate :: Constant indices' :: [Word32]
InsertValue
Fields aggregate :: Constant element :: Constant indices' :: [Word32]

Instances

Eq Constant
Ord Constant
Read Constant
Show Constant
PrettyShow Constant

signedIntegerValue :: Constant -> Integer

Since LLVM types don't include signedness, there's ambiguity in interpreting an constant as an Integer. The LLVM assembly printer prints integers as signed, but cheats for 1-bit integers and prints them as true or false. That way it circuments the otherwise awkward fact that a twos complement 1-bit number only has the values -1 and 0.

unsignedIntegerValue :: Constant -> Integer

This library's conversion from LLVM C++ objects will always produce integer constants as unsigned, so this function in many cases is not necessary. However, nothing's to keep stop direct construction of an Int with a negative integerValue. There's nothing in principle wrong with such a value - it has perfectly good low order bits like any integer, and will be used as such, likely producing the intended result if lowered to C++. If, however one wishes to interpret an Int of unknown provenance as unsigned, then this function will serve.