LLVM  mainline
InstCombineCasts.cpp File Reference
#include "InstCombineInternal.h"
#include "llvm/Analysis/ConstantFolding.h"
#include "llvm/IR/DataLayout.h"
#include "llvm/IR/PatternMatch.h"
#include "llvm/Analysis/TargetLibraryInfo.h"
Include dependency graph for InstCombineCasts.cpp:

Go to the source code of this file.

## Defines

#define DEBUG_TYPE   "instcombine"

## Functions

static ValuedecomposeSimpleLinearExpr (Value *Val, unsigned &Scale, uint64_t &Offset)
Analyze 'Val', seeing if it is a simple linear expression.
static Instruction::CastOps isEliminableCastPair (const CastInst *CI, unsigned opcode, Type *DstTy, const DataLayout &DL)
This function is a wrapper around CastInst::isEliminableCastPair.
static bool canEvaluateTruncated (Value *V, Type *Ty, InstCombiner &IC, Instruction *CxtI)
Return true if we can evaluate the specified expression tree as type Ty instead of its larger type, and arrive with the same value.
static InstructionfoldVecTruncToExtElt (TruncInst &Trunc, InstCombiner &IC, const DataLayout &DL)
Given a vector that is bitcast to an integer, optionally logically right-shifted, and truncated, convert it to an extractelement.
static bool canEvaluateZExtd (Value *V, Type *Ty, unsigned &BitsToClear, InstCombiner &IC, Instruction *CxtI)
Determine if the specified value can be computed in the specified wider type and produce the same low bits.
static bool canEvaluateSExtd (Value *V, Type *Ty)
Return true if we can take the specified value and return it as type Ty without inserting any new casts and without changing the value of the common low bits.
static ConstantfitsInFPType (ConstantFP *CFP, const fltSemantics &Sem)
Return a Constant* for the specified floating-point constant if it fits in the specified FP type without changing its value.
static ValuelookThroughFPExtensions (Value *V)
If this is a floating-point extension instruction, look through it until we get the source value.
static InstructionoptimizeVectorResize (Value *InVal, VectorType *DestTy, InstCombiner &IC)
This input value (which is known to have vector type) is being zero extended or truncated to the specified vector type.
static bool isMultipleOfTypeSize (unsigned Value, Type *Ty)
static unsigned getTypeSizeIndex (unsigned Value, Type *Ty)
static bool collectInsertionElements (Value *V, unsigned Shift, SmallVectorImpl< Value * > &Elements, Type *VecEltTy, bool isBigEndian)
V is a value which is inserted into a vector of VecEltTy.
static ValueoptimizeIntegerToVectorInsertions (BitCastInst &CI, InstCombiner &IC)
If the input is an 'or' instruction, we may be doing shifts and ors to assemble the elements of the vector manually.
static InstructioncanonicalizeBitCastExtElt (BitCastInst &BitCast, InstCombiner &IC, const DataLayout &DL)
Canonicalize scalar bitcasts of extracted elements into a bitcast of the vector followed by extract element.

## Define Documentation

 #define DEBUG_TYPE   "instcombine"

Definition at line 22 of file InstCombineCasts.cpp.

## Function Documentation

 static bool canEvaluateSExtd ( Value * V, Type * Ty )  [static]

Return true if we can take the specified value and return it as type Ty without inserting any new casts and without changing the value of the common low bits.

This is used by code that tries to promote integer operations to a wider types will allow us to eliminate the extension.

This function works on both vectors and scalars.

Definition at line 1060 of file InstCombineCasts.cpp.

Referenced by llvm::InstCombiner::visitSExt().

 static bool canEvaluateTruncated ( Value * V, Type * Ty, InstCombiner & IC, Instruction * CxtI )  [static]

Return true if we can evaluate the specified expression tree as type Ty instead of its larger type, and arrive with the same value.

This is used by code that tries to eliminate truncates.

Ty will always be a type smaller than V. We should return true if trunc(V) can be computed by computing V in the smaller type. If V is an instruction, then trunc(inst(x,y)) can be computed as inst(trunc(x),trunc(y)), which only makes sense if x and y can be efficiently truncated.

This function works on both vectors and scalars.

Definition at line 331 of file InstCombineCasts.cpp.

Referenced by llvm::InstCombiner::visitTrunc().

 static bool canEvaluateZExtd ( Value * V, Type * Ty, unsigned & BitsToClear, InstCombiner & IC, Instruction * CxtI )  [static]

Determine if the specified value can be computed in the specified wider type and produce the same low bits.

If not, return false.

If this function returns true, it can also return a non-zero number of bits (in BitsToClear) which indicates that the value it computes is correct for the zero extend, but that the additional BitsToClear bits need to be zero'd out. For example, to promote something like:

B = trunc i64 A to i32 C = lshr i32 B, 8 E = zext i32 C to i64

CanEvaluateZExtd for the 'lshr' will return true, and BitsToClear will be set to 8 to indicate that the promoted value needs to have bits 24-31 cleared in addition to bits 32-63. Since an 'and' will be generated to clear the top bits anyway, doing this has no extra cost.

This function works on both vectors and scalars.

Definition at line 724 of file InstCombineCasts.cpp.

Referenced by llvm::InstCombiner::visitZExt().

 static Instruction* canonicalizeBitCastExtElt ( BitCastInst & BitCast, InstCombiner & IC, const DataLayout & DL )  [static]

Canonicalize scalar bitcasts of extracted elements into a bitcast of the vector followed by extract element.

The backend tends to handle bitcasts of vectors better than bitcasts of scalars because vector registers are usually not type-specific like scalar integer or scalar floating-point.

Definition at line 1768 of file InstCombineCasts.cpp.

Referenced by llvm::InstCombiner::visitBitCast().

 static bool collectInsertionElements ( Value * V, unsigned Shift, SmallVectorImpl< Value * > & Elements, Type * VecEltTy, bool isBigEndian )  [static]

V is a value which is inserted into a vector of VecEltTy.

Look through the value to see if we can decompose it into insertions into the vector. See the example in the comment for OptimizeIntegerToVectorInsertions for the pattern this handles. The type of V is always a non-zero multiple of VecEltTy's size. Shift is the number of bits between the lsb of V and the lsb of the vector.

This returns false if the pattern can't be matched or true if it can, filling in Elements with the elements found here.

Definition at line 1630 of file InstCombineCasts.cpp.

Referenced by optimizeIntegerToVectorInsertions().

 static Value* decomposeSimpleLinearExpr ( Value * Val, unsigned & Scale, uint64_t & Offset )  [static]

Analyze 'Val', seeing if it is a simple linear expression.

If so, decompose it, returning some value X, such that Val is X*Scale+Offset.

Definition at line 28 of file InstCombineCasts.cpp.

 static Constant* fitsInFPType ( ConstantFP * CFP, const fltSemantics & Sem )  [static]

Return a Constant* for the specified floating-point constant if it fits in the specified FP type without changing its value.

Definition at line 1218 of file InstCombineCasts.cpp.

Referenced by lookThroughFPExtensions().

 static Instruction* foldVecTruncToExtElt ( TruncInst & Trunc, InstCombiner & IC, const DataLayout & DL )  [static]

Given a vector that is bitcast to an integer, optionally logically right-shifted, and truncated, convert it to an extractelement.

Example (big endian): trunc (lshr (bitcast <4 x i32> X to i128), 32) to i32 ---> extractelement <4 x i32> X, 1

Definition at line 439 of file InstCombineCasts.cpp.

Referenced by llvm::InstCombiner::visitTrunc().

 static unsigned getTypeSizeIndex ( unsigned Value, Type * Ty )  [static]

Definition at line 1616 of file InstCombineCasts.cpp.

References llvm::Type::getPrimitiveSizeInBits().

Referenced by collectInsertionElements().

 static Instruction::CastOps isEliminableCastPair ( const CastInst * CI, unsigned opcode, Type * DstTy, const DataLayout & DL )  [static]

This function is a wrapper around CastInst::isEliminableCastPair.

It simply extracts arguments and returns what that function returns.

Parameters:
 CI First cast instruction opcode Opcode for the second cast DstTy Target type for the second cast

Definition at line 234 of file InstCombineCasts.cpp.

Referenced by llvm::InstCombiner::commonCastTransforms(), and foldConstantCastPair().

 static bool isMultipleOfTypeSize ( unsigned Value, Type * Ty )  [static]

Definition at line 1612 of file InstCombineCasts.cpp.

References llvm::Type::getPrimitiveSizeInBits().

Referenced by collectInsertionElements().

 static Value* lookThroughFPExtensions ( Value * V )  [static]

If this is a floating-point extension instruction, look through it until we get the source value.

Definition at line 1229 of file InstCombineCasts.cpp.

Referenced by llvm::InstCombiner::visitFPTrunc().

 static Value* optimizeIntegerToVectorInsertions ( BitCastInst & CI, InstCombiner & IC )  [static]

If the input is an 'or' instruction, we may be doing shifts and ors to assemble the elements of the vector manually.

Try to rip the code out and replace it with insertelements. This is to optimize code like this:

tmp37 = bitcast float inc to i32 tmp38 = zext i32 tmp37 to i64 tmp31 = bitcast float inc5 to i32 tmp32 = zext i32 tmp31 to i64 tmp33 = shl i64 tmp32, 32 ins35 = or i64 tmp33, tmp38 tmp43 = bitcast i64 ins35 to <2 x float>

Into two insertelements that do "buildvector{%inc, %inc5}".

Definition at line 1739 of file InstCombineCasts.cpp.

Referenced by llvm::InstCombiner::visitBitCast().

 static Instruction* optimizeVectorResize ( Value * InVal, VectorType * DestTy, InstCombiner & IC )  [static]

This input value (which is known to have vector type) is being zero extended or truncated to the specified vector type.

Try to replace it with a shuffle (and vector/vector bitcast) if possible.

The source and destination vector types may have different element types.

Definition at line 1560 of file InstCombineCasts.cpp.

Referenced by llvm::InstCombiner::visitBitCast().