llvm/lib/Transforms/Scalar/InstructionCombining.cpp
2003-05-26 23:41:32 +00:00

1135 lines
43 KiB
C++

//===- InstructionCombining.cpp - Combine multiple instructions -----------===//
//
// InstructionCombining - Combine instructions to form fewer, simple
// instructions. This pass does not modify the CFG This pass is where algebraic
// simplification happens.
//
// This pass combines things like:
// %Y = add int 1, %X
// %Z = add int 1, %Y
// into:
// %Z = add int 2, %X
//
// This is a simple worklist driven algorithm.
//
//===----------------------------------------------------------------------===//
#include "llvm/Transforms/Scalar.h"
#include "llvm/Transforms/Utils/BasicBlockUtils.h"
#include "llvm/Transforms/Utils/Local.h"
#include "llvm/ConstantHandling.h"
#include "llvm/Instructions.h"
#include "llvm/Pass.h"
#include "llvm/DerivedTypes.h"
#include "llvm/Support/InstIterator.h"
#include "llvm/Support/InstVisitor.h"
#include "Support/Statistic.h"
#include <algorithm>
namespace {
Statistic<> NumCombined ("instcombine", "Number of insts combined");
Statistic<> NumConstProp("instcombine", "Number of constant folds");
Statistic<> NumDeadInst ("instcombine", "Number of dead inst eliminated");
class InstCombiner : public FunctionPass,
public InstVisitor<InstCombiner, Instruction*> {
// Worklist of all of the instructions that need to be simplified.
std::vector<Instruction*> WorkList;
void AddUsesToWorkList(Instruction &I) {
// The instruction was simplified, add all users of the instruction to
// the work lists because they might get more simplified now...
//
for (Value::use_iterator UI = I.use_begin(), UE = I.use_end();
UI != UE; ++UI)
WorkList.push_back(cast<Instruction>(*UI));
}
// removeFromWorkList - remove all instances of I from the worklist.
void removeFromWorkList(Instruction *I);
public:
virtual bool runOnFunction(Function &F);
virtual void getAnalysisUsage(AnalysisUsage &AU) const {
AU.setPreservesCFG();
}
// Visitation implementation - Implement instruction combining for different
// instruction types. The semantics are as follows:
// Return Value:
// null - No change was made
// I - Change was made, I is still valid, I may be dead though
// otherwise - Change was made, replace I with returned instruction
//
Instruction *visitAdd(BinaryOperator &I);
Instruction *visitSub(BinaryOperator &I);
Instruction *visitMul(BinaryOperator &I);
Instruction *visitDiv(BinaryOperator &I);
Instruction *visitRem(BinaryOperator &I);
Instruction *visitAnd(BinaryOperator &I);
Instruction *visitOr (BinaryOperator &I);
Instruction *visitXor(BinaryOperator &I);
Instruction *visitSetCondInst(BinaryOperator &I);
Instruction *visitShiftInst(ShiftInst &I);
Instruction *visitCastInst(CastInst &CI);
Instruction *visitPHINode(PHINode &PN);
Instruction *visitGetElementPtrInst(GetElementPtrInst &GEP);
Instruction *visitAllocationInst(AllocationInst &AI);
// visitInstruction - Specify what to return for unhandled instructions...
Instruction *visitInstruction(Instruction &I) { return 0; }
// InsertNewInstBefore - insert an instruction New before instruction Old
// in the program. Add the new instruction to the worklist.
//
void InsertNewInstBefore(Instruction *New, Instruction &Old) {
assert(New && New->getParent() == 0 &&
"New instruction already inserted into a basic block!");
BasicBlock *BB = Old.getParent();
BB->getInstList().insert(&Old, New); // Insert inst
WorkList.push_back(New); // Add to worklist
}
// ReplaceInstUsesWith - This method is to be used when an instruction is
// found to be dead, replacable with another preexisting expression. Here
// we add all uses of I to the worklist, replace all uses of I with the new
// value, then return I, so that the inst combiner will know that I was
// modified.
//
Instruction *ReplaceInstUsesWith(Instruction &I, Value *V) {
AddUsesToWorkList(I); // Add all modified instrs to worklist
I.replaceAllUsesWith(V);
return &I;
}
// SimplifyCommutative - This performs a few simplifications for commutative
// operators...
bool SimplifyCommutative(BinaryOperator &I);
};
RegisterOpt<InstCombiner> X("instcombine", "Combine redundant instructions");
}
// getComplexity: Assign a complexity or rank value to LLVM Values...
// 0 -> Constant, 1 -> Other, 2 -> Argument, 2 -> Unary, 3 -> OtherInst
static unsigned getComplexity(Value *V) {
if (isa<Instruction>(V)) {
if (BinaryOperator::isNeg(V) || BinaryOperator::isNot(V))
return 2;
return 3;
}
if (isa<Argument>(V)) return 2;
return isa<Constant>(V) ? 0 : 1;
}
// isOnlyUse - Return true if this instruction will be deleted if we stop using
// it.
static bool isOnlyUse(Value *V) {
return V->use_size() == 1 || isa<Constant>(V);
}
// SimplifyCommutative - This performs a few simplifications for commutative
// operators:
//
// 1. Order operands such that they are listed from right (least complex) to
// left (most complex). This puts constants before unary operators before
// binary operators.
//
// 2. Transform: (op (op V, C1), C2) ==> (op V, (op C1, C2))
// 3. Transform: (op (op V1, C1), (op V2, C2)) ==> (op (op V1, V2), (op C1,C2))
//
bool InstCombiner::SimplifyCommutative(BinaryOperator &I) {
bool Changed = false;
if (getComplexity(I.getOperand(0)) < getComplexity(I.getOperand(1)))
Changed = !I.swapOperands();
if (!I.isAssociative()) return Changed;
Instruction::BinaryOps Opcode = I.getOpcode();
if (BinaryOperator *Op = dyn_cast<BinaryOperator>(I.getOperand(0)))
if (Op->getOpcode() == Opcode && isa<Constant>(Op->getOperand(1))) {
if (isa<Constant>(I.getOperand(1))) {
Constant *Folded = ConstantFoldBinaryInstruction(I.getOpcode(),
cast<Constant>(I.getOperand(1)), cast<Constant>(Op->getOperand(1)));
assert(Folded && "Couldn't constant fold commutative operand?");
I.setOperand(0, Op->getOperand(0));
I.setOperand(1, Folded);
return true;
} else if (BinaryOperator *Op1=dyn_cast<BinaryOperator>(I.getOperand(1)))
if (Op1->getOpcode() == Opcode && isa<Constant>(Op1->getOperand(1)) &&
isOnlyUse(Op) && isOnlyUse(Op1)) {
Constant *C1 = cast<Constant>(Op->getOperand(1));
Constant *C2 = cast<Constant>(Op1->getOperand(1));
// Fold (op (op V1, C1), (op V2, C2)) ==> (op (op V1, V2), (op C1,C2))
Constant *Folded = ConstantFoldBinaryInstruction(I.getOpcode(),C1,C2);
assert(Folded && "Couldn't constant fold commutative operand?");
Instruction *New = BinaryOperator::create(Opcode, Op->getOperand(0),
Op1->getOperand(0),
Op1->getName(), &I);
WorkList.push_back(New);
I.setOperand(0, New);
I.setOperand(1, Folded);
return true;
}
}
return Changed;
}
// dyn_castNegVal - Given a 'sub' instruction, return the RHS of the instruction
// if the LHS is a constant zero (which is the 'negate' form).
//
static inline Value *dyn_castNegVal(Value *V) {
if (BinaryOperator::isNeg(V))
return BinaryOperator::getNegArgument(cast<BinaryOperator>(V));
// Constants can be considered to be negated values if they can be folded...
if (Constant *C = dyn_cast<Constant>(V))
return *Constant::getNullValue(V->getType()) - *C;
return 0;
}
static inline Value *dyn_castNotVal(Value *V) {
if (BinaryOperator::isNot(V))
return BinaryOperator::getNotArgument(cast<BinaryOperator>(V));
// Constants can be considered to be not'ed values...
if (ConstantIntegral *C = dyn_cast<ConstantIntegral>(V))
return *ConstantIntegral::getAllOnesValue(C->getType()) ^ *C;
return 0;
}
// dyn_castFoldableMul - If this value is a multiply that can be folded into
// other computations (because it has a constant operand), return the
// non-constant operand of the multiply.
//
static inline Value *dyn_castFoldableMul(Value *V) {
if (V->use_size() == 1 && V->getType()->isInteger())
if (Instruction *I = dyn_cast<Instruction>(V))
if (I->getOpcode() == Instruction::Mul)
if (isa<Constant>(I->getOperand(1)))
return I->getOperand(0);
return 0;
}
// dyn_castMaskingAnd - If this value is an And instruction masking a value with
// a constant, return the constant being anded with.
//
static inline Constant *dyn_castMaskingAnd(Value *V) {
if (Instruction *I = dyn_cast<Instruction>(V))
if (I->getOpcode() == Instruction::And)
return dyn_cast<Constant>(I->getOperand(1));
// If this is a constant, it acts just like we were masking with it.
return dyn_cast<Constant>(V);
}
// Log2 - Calculate the log base 2 for the specified value if it is exactly a
// power of 2.
static unsigned Log2(uint64_t Val) {
assert(Val > 1 && "Values 0 and 1 should be handled elsewhere!");
unsigned Count = 0;
while (Val != 1) {
if (Val & 1) return 0; // Multiple bits set?
Val >>= 1;
++Count;
}
return Count;
}
Instruction *InstCombiner::visitAdd(BinaryOperator &I) {
bool Changed = SimplifyCommutative(I);
Value *LHS = I.getOperand(0), *RHS = I.getOperand(1);
// Eliminate 'add int %X, 0'
if (RHS == Constant::getNullValue(I.getType()))
return ReplaceInstUsesWith(I, LHS);
// -A + B --> B - A
if (Value *V = dyn_castNegVal(LHS))
return BinaryOperator::create(Instruction::Sub, RHS, V);
// A + -B --> A - B
if (!isa<Constant>(RHS))
if (Value *V = dyn_castNegVal(RHS))
return BinaryOperator::create(Instruction::Sub, LHS, V);
// X*C + X --> X * (C+1)
if (dyn_castFoldableMul(LHS) == RHS) {
Constant *CP1 = *cast<Constant>(cast<Instruction>(LHS)->getOperand(1)) +
*ConstantInt::get(I.getType(), 1);
assert(CP1 && "Couldn't constant fold C + 1?");
return BinaryOperator::create(Instruction::Mul, RHS, CP1);
}
// X + X*C --> X * (C+1)
if (dyn_castFoldableMul(RHS) == LHS) {
Constant *CP1 = *cast<Constant>(cast<Instruction>(RHS)->getOperand(1)) +
*ConstantInt::get(I.getType(), 1);
assert(CP1 && "Couldn't constant fold C + 1?");
return BinaryOperator::create(Instruction::Mul, LHS, CP1);
}
// (A & C1)+(B & C2) -> (A & C1)|(B & C2) iff C1&C2 == 0
if (Constant *C1 = dyn_castMaskingAnd(LHS))
if (Constant *C2 = dyn_castMaskingAnd(RHS))
if ((*C1 & *C2)->isNullValue())
return BinaryOperator::create(Instruction::Or, LHS, RHS);
return Changed ? &I : 0;
}
Instruction *InstCombiner::visitSub(BinaryOperator &I) {
Value *Op0 = I.getOperand(0), *Op1 = I.getOperand(1);
if (Op0 == Op1) // sub X, X -> 0
return ReplaceInstUsesWith(I, Constant::getNullValue(I.getType()));
// If this is a 'B = x-(-A)', change to B = x+A...
if (Value *V = dyn_castNegVal(Op1))
return BinaryOperator::create(Instruction::Add, Op0, V);
// Replace (-1 - A) with (~A)...
if (ConstantInt *C = dyn_cast<ConstantInt>(Op0))
if (C->isAllOnesValue())
return BinaryOperator::createNot(Op1);
if (BinaryOperator *Op1I = dyn_cast<BinaryOperator>(Op1))
if (Op1I->use_size() == 1) {
// Replace (x - (y - z)) with (x + (z - y)) if the (y - z) subexpression
// is not used by anyone else...
//
if (Op1I->getOpcode() == Instruction::Sub) {
// Swap the two operands of the subexpr...
Value *IIOp0 = Op1I->getOperand(0), *IIOp1 = Op1I->getOperand(1);
Op1I->setOperand(0, IIOp1);
Op1I->setOperand(1, IIOp0);
// Create the new top level add instruction...
return BinaryOperator::create(Instruction::Add, Op0, Op1);
}
// Replace (A - (A & B)) with (A & ~B) if this is the only use of (A&B)...
//
if (Op1I->getOpcode() == Instruction::And &&
(Op1I->getOperand(0) == Op0 || Op1I->getOperand(1) == Op0)) {
Value *OtherOp = Op1I->getOperand(Op1I->getOperand(0) == Op0);
Instruction *NewNot = BinaryOperator::createNot(OtherOp, "B.not", &I);
return BinaryOperator::create(Instruction::And, Op0, NewNot);
}
// X - X*C --> X * (1-C)
if (dyn_castFoldableMul(Op1I) == Op0) {
Constant *CP1 = *ConstantInt::get(I.getType(), 1) -
*cast<Constant>(cast<Instruction>(Op1)->getOperand(1));
assert(CP1 && "Couldn't constant fold 1-C?");
return BinaryOperator::create(Instruction::Mul, Op0, CP1);
}
}
// X*C - X --> X * (C-1)
if (dyn_castFoldableMul(Op0) == Op1) {
Constant *CP1 = *cast<Constant>(cast<Instruction>(Op0)->getOperand(1)) -
*ConstantInt::get(I.getType(), 1);
assert(CP1 && "Couldn't constant fold C - 1?");
return BinaryOperator::create(Instruction::Mul, Op1, CP1);
}
return 0;
}
Instruction *InstCombiner::visitMul(BinaryOperator &I) {
bool Changed = SimplifyCommutative(I);
Value *Op0 = I.getOperand(0);
// Simplify mul instructions with a constant RHS...
if (Constant *Op1 = dyn_cast<Constant>(I.getOperand(1))) {
if (ConstantInt *CI = dyn_cast<ConstantInt>(Op1)) {
const Type *Ty = CI->getType();
uint64_t Val = Ty->isSigned() ?
(uint64_t)cast<ConstantSInt>(CI)->getValue() :
cast<ConstantUInt>(CI)->getValue();
switch (Val) {
case 0:
return ReplaceInstUsesWith(I, Op1); // Eliminate 'mul double %X, 0'
case 1:
return ReplaceInstUsesWith(I, Op0); // Eliminate 'mul int %X, 1'
case 2: // Convert 'mul int %X, 2' to 'add int %X, %X'
return BinaryOperator::create(Instruction::Add, Op0, Op0, I.getName());
}
if (uint64_t C = Log2(Val)) // Replace X*(2^C) with X << C
return new ShiftInst(Instruction::Shl, Op0,
ConstantUInt::get(Type::UByteTy, C));
} else {
ConstantFP *Op1F = cast<ConstantFP>(Op1);
if (Op1F->isNullValue())
return ReplaceInstUsesWith(I, Op1);
// "In IEEE floating point, x*1 is not equivalent to x for nans. However,
// ANSI says we can drop signals, so we can do this anyway." (from GCC)
if (Op1F->getValue() == 1.0)
return ReplaceInstUsesWith(I, Op0); // Eliminate 'mul double %X, 1.0'
}
}
if (Value *Op0v = dyn_castNegVal(Op0)) // -X * -Y = X*Y
if (Value *Op1v = dyn_castNegVal(I.getOperand(1)))
return BinaryOperator::create(Instruction::Mul, Op0v, Op1v);
return Changed ? &I : 0;
}
Instruction *InstCombiner::visitDiv(BinaryOperator &I) {
// div X, 1 == X
if (ConstantInt *RHS = dyn_cast<ConstantInt>(I.getOperand(1))) {
if (RHS->equalsInt(1))
return ReplaceInstUsesWith(I, I.getOperand(0));
// Check to see if this is an unsigned division with an exact power of 2,
// if so, convert to a right shift.
if (ConstantUInt *C = dyn_cast<ConstantUInt>(RHS))
if (uint64_t Val = C->getValue()) // Don't break X / 0
if (uint64_t C = Log2(Val))
return new ShiftInst(Instruction::Shr, I.getOperand(0),
ConstantUInt::get(Type::UByteTy, C));
}
// 0 / X == 0, we don't need to preserve faults!
if (ConstantInt *LHS = dyn_cast<ConstantInt>(I.getOperand(0)))
if (LHS->equalsInt(0))
return ReplaceInstUsesWith(I, Constant::getNullValue(I.getType()));
return 0;
}
Instruction *InstCombiner::visitRem(BinaryOperator &I) {
if (ConstantInt *RHS = dyn_cast<ConstantInt>(I.getOperand(1))) {
if (RHS->equalsInt(1)) // X % 1 == 0
return ReplaceInstUsesWith(I, Constant::getNullValue(I.getType()));
// Check to see if this is an unsigned remainder with an exact power of 2,
// if so, convert to a bitwise and.
if (ConstantUInt *C = dyn_cast<ConstantUInt>(RHS))
if (uint64_t Val = C->getValue()) // Don't break X % 0 (divide by zero)
if (Log2(Val))
return BinaryOperator::create(Instruction::And, I.getOperand(0),
ConstantUInt::get(I.getType(), Val-1));
}
// 0 % X == 0, we don't need to preserve faults!
if (ConstantInt *LHS = dyn_cast<ConstantInt>(I.getOperand(0)))
if (LHS->equalsInt(0))
return ReplaceInstUsesWith(I, Constant::getNullValue(I.getType()));
return 0;
}
// isMaxValueMinusOne - return true if this is Max-1
static bool isMaxValueMinusOne(const ConstantInt *C) {
if (const ConstantUInt *CU = dyn_cast<ConstantUInt>(C)) {
// Calculate -1 casted to the right type...
unsigned TypeBits = C->getType()->getPrimitiveSize()*8;
uint64_t Val = ~0ULL; // All ones
Val >>= 64-TypeBits; // Shift out unwanted 1 bits...
return CU->getValue() == Val-1;
}
const ConstantSInt *CS = cast<ConstantSInt>(C);
// Calculate 0111111111..11111
unsigned TypeBits = C->getType()->getPrimitiveSize()*8;
int64_t Val = INT64_MAX; // All ones
Val >>= 64-TypeBits; // Shift out unwanted 1 bits...
return CS->getValue() == Val-1;
}
// isMinValuePlusOne - return true if this is Min+1
static bool isMinValuePlusOne(const ConstantInt *C) {
if (const ConstantUInt *CU = dyn_cast<ConstantUInt>(C))
return CU->getValue() == 1;
const ConstantSInt *CS = cast<ConstantSInt>(C);
// Calculate 1111111111000000000000
unsigned TypeBits = C->getType()->getPrimitiveSize()*8;
int64_t Val = -1; // All ones
Val <<= TypeBits-1; // Shift over to the right spot
return CS->getValue() == Val+1;
}
Instruction *InstCombiner::visitAnd(BinaryOperator &I) {
bool Changed = SimplifyCommutative(I);
Value *Op0 = I.getOperand(0), *Op1 = I.getOperand(1);
// and X, X = X and X, 0 == 0
if (Op0 == Op1 || Op1 == Constant::getNullValue(I.getType()))
return ReplaceInstUsesWith(I, Op1);
// and X, -1 == X
if (ConstantIntegral *RHS = dyn_cast<ConstantIntegral>(Op1))
if (RHS->isAllOnesValue())
return ReplaceInstUsesWith(I, Op0);
Value *Op0NotVal = dyn_castNotVal(Op0);
Value *Op1NotVal = dyn_castNotVal(Op1);
// (~A & ~B) == (~(A | B)) - Demorgan's Law
if (Op0NotVal && Op1NotVal && isOnlyUse(Op0) && isOnlyUse(Op1)) {
Instruction *Or = BinaryOperator::create(Instruction::Or, Op0NotVal,
Op1NotVal,I.getName()+".demorgan",
&I);
WorkList.push_back(Or);
return BinaryOperator::createNot(Or);
}
if (Op0NotVal == Op1 || Op1NotVal == Op0) // A & ~A == ~A & A == 0
return ReplaceInstUsesWith(I, Constant::getNullValue(I.getType()));
return Changed ? &I : 0;
}
Instruction *InstCombiner::visitOr(BinaryOperator &I) {
bool Changed = SimplifyCommutative(I);
Value *Op0 = I.getOperand(0), *Op1 = I.getOperand(1);
// or X, X = X or X, 0 == X
if (Op0 == Op1 || Op1 == Constant::getNullValue(I.getType()))
return ReplaceInstUsesWith(I, Op0);
// or X, -1 == -1
if (ConstantIntegral *RHS = dyn_cast<ConstantIntegral>(Op1))
if (RHS->isAllOnesValue())
return ReplaceInstUsesWith(I, Op1);
Value *Op0NotVal = dyn_castNotVal(Op0);
Value *Op1NotVal = dyn_castNotVal(Op1);
if (Op1 == Op0NotVal) // ~A | A == -1
return ReplaceInstUsesWith(I,
ConstantIntegral::getAllOnesValue(I.getType()));
if (Op0 == Op1NotVal) // A | ~A == -1
return ReplaceInstUsesWith(I,
ConstantIntegral::getAllOnesValue(I.getType()));
// (~A | ~B) == (~(A & B)) - Demorgan's Law
if (Op0NotVal && Op1NotVal && isOnlyUse(Op0) && isOnlyUse(Op1)) {
Instruction *And = BinaryOperator::create(Instruction::And, Op0NotVal,
Op1NotVal,I.getName()+".demorgan",
&I);
WorkList.push_back(And);
return BinaryOperator::createNot(And);
}
return Changed ? &I : 0;
}
Instruction *InstCombiner::visitXor(BinaryOperator &I) {
bool Changed = SimplifyCommutative(I);
Value *Op0 = I.getOperand(0), *Op1 = I.getOperand(1);
// xor X, X = 0
if (Op0 == Op1)
return ReplaceInstUsesWith(I, Constant::getNullValue(I.getType()));
if (ConstantIntegral *Op1C = dyn_cast<ConstantIntegral>(Op1)) {
// xor X, 0 == X
if (Op1C->isNullValue())
return ReplaceInstUsesWith(I, Op0);
// Is this a "NOT" instruction?
if (Op1C->isAllOnesValue()) {
// xor (xor X, -1), -1 = not (not X) = X
if (Value *X = dyn_castNotVal(Op0))
return ReplaceInstUsesWith(I, X);
// xor (setcc A, B), true = not (setcc A, B) = setncc A, B
if (SetCondInst *SCI = dyn_cast<SetCondInst>(Op0))
if (SCI->use_size() == 1)
return new SetCondInst(SCI->getInverseCondition(),
SCI->getOperand(0), SCI->getOperand(1));
}
}
if (Value *X = dyn_castNotVal(Op0)) // ~A ^ A == -1
if (X == Op1)
return ReplaceInstUsesWith(I,
ConstantIntegral::getAllOnesValue(I.getType()));
if (Value *X = dyn_castNotVal(Op1)) // A ^ ~A == -1
if (X == Op0)
return ReplaceInstUsesWith(I,
ConstantIntegral::getAllOnesValue(I.getType()));
if (Instruction *Op1I = dyn_cast<Instruction>(Op1))
if (Op1I->getOpcode() == Instruction::Or)
if (Op1I->getOperand(0) == Op0) { // B^(B|A) == (A|B)^B
cast<BinaryOperator>(Op1I)->swapOperands();
I.swapOperands();
std::swap(Op0, Op1);
} else if (Op1I->getOperand(1) == Op0) { // B^(A|B) == (A|B)^B
I.swapOperands();
std::swap(Op0, Op1);
}
if (Instruction *Op0I = dyn_cast<Instruction>(Op0))
if (Op0I->getOpcode() == Instruction::Or && Op0I->use_size() == 1) {
if (Op0I->getOperand(0) == Op1) // (B|A)^B == (A|B)^B
cast<BinaryOperator>(Op0I)->swapOperands();
if (Op0I->getOperand(1) == Op1) { // (A|B)^B == A & ~B
Value *NotB = BinaryOperator::createNot(Op1, Op1->getName()+".not", &I);
WorkList.push_back(cast<Instruction>(NotB));
return BinaryOperator::create(Instruction::And, Op0I->getOperand(0),
NotB);
}
}
// (A & C1)^(B & C2) -> (A & C1)|(B & C2) iff C1^C2 == 0
if (Constant *C1 = dyn_castMaskingAnd(Op0))
if (Constant *C2 = dyn_castMaskingAnd(Op1))
if ((*C1 & *C2)->isNullValue())
return BinaryOperator::create(Instruction::Or, Op0, Op1);
return Changed ? &I : 0;
}
// AddOne, SubOne - Add or subtract a constant one from an integer constant...
static Constant *AddOne(ConstantInt *C) {
Constant *Result = *C + *ConstantInt::get(C->getType(), 1);
assert(Result && "Constant folding integer addition failed!");
return Result;
}
static Constant *SubOne(ConstantInt *C) {
Constant *Result = *C - *ConstantInt::get(C->getType(), 1);
assert(Result && "Constant folding integer addition failed!");
return Result;
}
// isTrueWhenEqual - Return true if the specified setcondinst instruction is
// true when both operands are equal...
//
static bool isTrueWhenEqual(Instruction &I) {
return I.getOpcode() == Instruction::SetEQ ||
I.getOpcode() == Instruction::SetGE ||
I.getOpcode() == Instruction::SetLE;
}
Instruction *InstCombiner::visitSetCondInst(BinaryOperator &I) {
bool Changed = SimplifyCommutative(I);
Value *Op0 = I.getOperand(0), *Op1 = I.getOperand(1);
const Type *Ty = Op0->getType();
// setcc X, X
if (Op0 == Op1)
return ReplaceInstUsesWith(I, ConstantBool::get(isTrueWhenEqual(I)));
// setcc <global*>, 0 - Global value addresses are never null!
if (isa<GlobalValue>(Op0) && isa<ConstantPointerNull>(Op1))
return ReplaceInstUsesWith(I, ConstantBool::get(!isTrueWhenEqual(I)));
// setcc's with boolean values can always be turned into bitwise operations
if (Ty == Type::BoolTy) {
// If this is <, >, or !=, we can change this into a simple xor instruction
if (!isTrueWhenEqual(I))
return BinaryOperator::create(Instruction::Xor, Op0, Op1, I.getName());
// Otherwise we need to make a temporary intermediate instruction and insert
// it into the instruction stream. This is what we are after:
//
// seteq bool %A, %B -> ~(A^B)
// setle bool %A, %B -> ~A | B
// setge bool %A, %B -> A | ~B
//
if (I.getOpcode() == Instruction::SetEQ) { // seteq case
Instruction *Xor = BinaryOperator::create(Instruction::Xor, Op0, Op1,
I.getName()+"tmp");
InsertNewInstBefore(Xor, I);
return BinaryOperator::createNot(Xor, I.getName());
}
// Handle the setXe cases...
assert(I.getOpcode() == Instruction::SetGE ||
I.getOpcode() == Instruction::SetLE);
if (I.getOpcode() == Instruction::SetGE)
std::swap(Op0, Op1); // Change setge -> setle
// Now we just have the SetLE case.
Instruction *Not = BinaryOperator::createNot(Op0, I.getName()+"tmp");
InsertNewInstBefore(Not, I);
return BinaryOperator::create(Instruction::Or, Not, Op1, I.getName());
}
// Check to see if we are doing one of many comparisons against constant
// integers at the end of their ranges...
//
if (ConstantInt *CI = dyn_cast<ConstantInt>(Op1)) {
// Check to see if we are comparing against the minimum or maximum value...
if (CI->isMinValue()) {
if (I.getOpcode() == Instruction::SetLT) // A < MIN -> FALSE
return ReplaceInstUsesWith(I, ConstantBool::False);
if (I.getOpcode() == Instruction::SetGE) // A >= MIN -> TRUE
return ReplaceInstUsesWith(I, ConstantBool::True);
if (I.getOpcode() == Instruction::SetLE) // A <= MIN -> A == MIN
return BinaryOperator::create(Instruction::SetEQ, Op0,Op1, I.getName());
if (I.getOpcode() == Instruction::SetGT) // A > MIN -> A != MIN
return BinaryOperator::create(Instruction::SetNE, Op0,Op1, I.getName());
} else if (CI->isMaxValue()) {
if (I.getOpcode() == Instruction::SetGT) // A > MAX -> FALSE
return ReplaceInstUsesWith(I, ConstantBool::False);
if (I.getOpcode() == Instruction::SetLE) // A <= MAX -> TRUE
return ReplaceInstUsesWith(I, ConstantBool::True);
if (I.getOpcode() == Instruction::SetGE) // A >= MAX -> A == MAX
return BinaryOperator::create(Instruction::SetEQ, Op0,Op1, I.getName());
if (I.getOpcode() == Instruction::SetLT) // A < MAX -> A != MAX
return BinaryOperator::create(Instruction::SetNE, Op0,Op1, I.getName());
// Comparing against a value really close to min or max?
} else if (isMinValuePlusOne(CI)) {
if (I.getOpcode() == Instruction::SetLT) // A < MIN+1 -> A == MIN
return BinaryOperator::create(Instruction::SetEQ, Op0,
SubOne(CI), I.getName());
if (I.getOpcode() == Instruction::SetGE) // A >= MIN-1 -> A != MIN
return BinaryOperator::create(Instruction::SetNE, Op0,
SubOne(CI), I.getName());
} else if (isMaxValueMinusOne(CI)) {
if (I.getOpcode() == Instruction::SetGT) // A > MAX-1 -> A == MAX
return BinaryOperator::create(Instruction::SetEQ, Op0,
AddOne(CI), I.getName());
if (I.getOpcode() == Instruction::SetLE) // A <= MAX-1 -> A != MAX
return BinaryOperator::create(Instruction::SetNE, Op0,
AddOne(CI), I.getName());
}
}
return Changed ? &I : 0;
}
Instruction *InstCombiner::visitShiftInst(ShiftInst &I) {
assert(I.getOperand(1)->getType() == Type::UByteTy);
Value *Op0 = I.getOperand(0), *Op1 = I.getOperand(1);
// shl X, 0 == X and shr X, 0 == X
// shl 0, X == 0 and shr 0, X == 0
if (Op1 == Constant::getNullValue(Type::UByteTy) ||
Op0 == Constant::getNullValue(Op0->getType()))
return ReplaceInstUsesWith(I, Op0);
// If this is a shift of a shift, see if we can fold the two together...
if (ShiftInst *Op0SI = dyn_cast<ShiftInst>(Op0)) {
if (isa<Constant>(Op1) && isa<Constant>(Op0SI->getOperand(1))) {
ConstantUInt *ShiftAmt1C = cast<ConstantUInt>(Op0SI->getOperand(1));
unsigned ShiftAmt1 = ShiftAmt1C->getValue();
unsigned ShiftAmt2 = cast<ConstantUInt>(Op1)->getValue();
// Check for (A << c1) << c2 and (A >> c1) >> c2
if (I.getOpcode() == Op0SI->getOpcode()) {
unsigned Amt = ShiftAmt1+ShiftAmt2; // Fold into one big shift...
return new ShiftInst(I.getOpcode(), Op0SI->getOperand(0),
ConstantUInt::get(Type::UByteTy, Amt));
}
if (I.getType()->isUnsigned()) { // Check for (A << c1) >> c2 or visaversa
// Calculate bitmask for what gets shifted off the edge...
Constant *C = ConstantIntegral::getAllOnesValue(I.getType());
if (I.getOpcode() == Instruction::Shr)
C = *C >> *ShiftAmt1C;
else
C = *C << *ShiftAmt1C;
assert(C && "Couldn't constant fold shift expression?");
Instruction *Mask =
BinaryOperator::create(Instruction::And, Op0SI->getOperand(0),
C, Op0SI->getOperand(0)->getName()+".mask",&I);
WorkList.push_back(Mask);
// Figure out what flavor of shift we should use...
if (ShiftAmt1 == ShiftAmt2)
return ReplaceInstUsesWith(I, Mask); // (A << c) >> c === A & c2
else if (ShiftAmt1 < ShiftAmt2) {
return new ShiftInst(I.getOpcode(), Mask,
ConstantUInt::get(Type::UByteTy, ShiftAmt2-ShiftAmt1));
} else {
return new ShiftInst(Op0SI->getOpcode(), Mask,
ConstantUInt::get(Type::UByteTy, ShiftAmt1-ShiftAmt2));
}
}
}
}
// shl uint X, 32 = 0 and shr ubyte Y, 9 = 0, ... just don't eliminate shr of
// a signed value.
//
if (ConstantUInt *CUI = dyn_cast<ConstantUInt>(Op1)) {
unsigned TypeBits = Op0->getType()->getPrimitiveSize()*8;
if (CUI->getValue() >= TypeBits &&
(!Op0->getType()->isSigned() || I.getOpcode() == Instruction::Shl))
return ReplaceInstUsesWith(I, Constant::getNullValue(Op0->getType()));
// Check to see if we are shifting left by 1. If so, turn it into an add
// instruction.
if (I.getOpcode() == Instruction::Shl && CUI->equalsInt(1))
// Convert 'shl int %X, 1' to 'add int %X, %X'
return BinaryOperator::create(Instruction::Add, Op0, Op0, I.getName());
}
// shr int -1, X = -1 (for any arithmetic shift rights of ~0)
if (ConstantSInt *CSI = dyn_cast<ConstantSInt>(Op0))
if (I.getOpcode() == Instruction::Shr && CSI->isAllOnesValue())
return ReplaceInstUsesWith(I, CSI);
return 0;
}
// isEliminableCastOfCast - Return true if it is valid to eliminate the CI
// instruction.
//
static inline bool isEliminableCastOfCast(const CastInst &CI,
const CastInst *CSrc) {
assert(CI.getOperand(0) == CSrc);
const Type *SrcTy = CSrc->getOperand(0)->getType();
const Type *MidTy = CSrc->getType();
const Type *DstTy = CI.getType();
// It is legal to eliminate the instruction if casting A->B->A if the sizes
// are identical and the bits don't get reinterpreted (for example
// int->float->int would not be allowed)
if (SrcTy == DstTy && SrcTy->isLosslesslyConvertibleTo(MidTy))
return true;
// Allow free casting and conversion of sizes as long as the sign doesn't
// change...
if (SrcTy->isIntegral() && MidTy->isIntegral() && DstTy->isIntegral()) {
unsigned SrcSize = SrcTy->getPrimitiveSize();
unsigned MidSize = MidTy->getPrimitiveSize();
unsigned DstSize = DstTy->getPrimitiveSize();
// Cases where we are monotonically decreasing the size of the type are
// always ok, regardless of what sign changes are going on.
//
if (SrcSize >= MidSize && MidSize >= DstSize)
return true;
// Cases where the source and destination type are the same, but the middle
// type is bigger are noops.
//
if (SrcSize == DstSize && MidSize > SrcSize)
return true;
// If we are monotonically growing, things are more complex.
//
if (SrcSize <= MidSize && MidSize <= DstSize) {
// We have eight combinations of signedness to worry about. Here's the
// table:
static const int SignTable[8] = {
// CODE, SrcSigned, MidSigned, DstSigned, Comment
1, // U U U Always ok
1, // U U S Always ok
3, // U S U Ok iff SrcSize != MidSize
3, // U S S Ok iff SrcSize != MidSize
0, // S U U Never ok
2, // S U S Ok iff MidSize == DstSize
1, // S S U Always ok
1, // S S S Always ok
};
// Choose an action based on the current entry of the signtable that this
// cast of cast refers to...
unsigned Row = SrcTy->isSigned()*4+MidTy->isSigned()*2+DstTy->isSigned();
switch (SignTable[Row]) {
case 0: return false; // Never ok
case 1: return true; // Always ok
case 2: return MidSize == DstSize; // Ok iff MidSize == DstSize
case 3: // Ok iff SrcSize != MidSize
return SrcSize != MidSize || SrcTy == Type::BoolTy;
default: assert(0 && "Bad entry in sign table!");
}
}
}
// Otherwise, we cannot succeed. Specifically we do not want to allow things
// like: short -> ushort -> uint, because this can create wrong results if
// the input short is negative!
//
return false;
}
// CastInst simplification
//
Instruction *InstCombiner::visitCastInst(CastInst &CI) {
// If the user is casting a value to the same type, eliminate this cast
// instruction...
if (CI.getType() == CI.getOperand(0)->getType())
return ReplaceInstUsesWith(CI, CI.getOperand(0));
// If casting the result of another cast instruction, try to eliminate this
// one!
//
if (CastInst *CSrc = dyn_cast<CastInst>(CI.getOperand(0))) {
if (isEliminableCastOfCast(CI, CSrc)) {
// This instruction now refers directly to the cast's src operand. This
// has a good chance of making CSrc dead.
CI.setOperand(0, CSrc->getOperand(0));
return &CI;
}
// If this is an A->B->A cast, and we are dealing with integral types, try
// to convert this into a logical 'and' instruction.
//
if (CSrc->getOperand(0)->getType() == CI.getType() &&
CI.getType()->isInteger() && CSrc->getType()->isInteger() &&
CI.getType()->isUnsigned() && CSrc->getType()->isUnsigned() &&
CSrc->getType()->getPrimitiveSize() < CI.getType()->getPrimitiveSize()){
assert(CSrc->getType() != Type::ULongTy &&
"Cannot have type bigger than ulong!");
uint64_t AndValue = (1ULL << CSrc->getType()->getPrimitiveSize()*8)-1;
Constant *AndOp = ConstantUInt::get(CI.getType(), AndValue);
return BinaryOperator::create(Instruction::And, CSrc->getOperand(0),
AndOp);
}
}
return 0;
}
// PHINode simplification
//
Instruction *InstCombiner::visitPHINode(PHINode &PN) {
// If the PHI node only has one incoming value, eliminate the PHI node...
if (PN.getNumIncomingValues() == 1)
return ReplaceInstUsesWith(PN, PN.getIncomingValue(0));
// Otherwise if all of the incoming values are the same for the PHI, replace
// the PHI node with the incoming value.
//
Value *InVal = 0;
for (unsigned i = 0, e = PN.getNumIncomingValues(); i != e; ++i)
if (PN.getIncomingValue(i) != &PN) // Not the PHI node itself...
if (InVal && PN.getIncomingValue(i) != InVal)
return 0; // Not the same, bail out.
else
InVal = PN.getIncomingValue(i);
// The only case that could cause InVal to be null is if we have a PHI node
// that only has entries for itself. In this case, there is no entry into the
// loop, so kill the PHI.
//
if (InVal == 0) InVal = Constant::getNullValue(PN.getType());
// All of the incoming values are the same, replace the PHI node now.
return ReplaceInstUsesWith(PN, InVal);
}
Instruction *InstCombiner::visitGetElementPtrInst(GetElementPtrInst &GEP) {
// Is it 'getelementptr %P, long 0' or 'getelementptr %P'
// If so, eliminate the noop.
if ((GEP.getNumOperands() == 2 &&
GEP.getOperand(1) == Constant::getNullValue(Type::LongTy)) ||
GEP.getNumOperands() == 1)
return ReplaceInstUsesWith(GEP, GEP.getOperand(0));
// Combine Indices - If the source pointer to this getelementptr instruction
// is a getelementptr instruction, combine the indices of the two
// getelementptr instructions into a single instruction.
//
if (GetElementPtrInst *Src = dyn_cast<GetElementPtrInst>(GEP.getOperand(0))) {
std::vector<Value *> Indices;
// Can we combine the two pointer arithmetics offsets?
if (Src->getNumOperands() == 2 && isa<Constant>(Src->getOperand(1)) &&
isa<Constant>(GEP.getOperand(1))) {
// Replace: gep (gep %P, long C1), long C2, ...
// With: gep %P, long (C1+C2), ...
Value *Sum = *cast<Constant>(Src->getOperand(1)) +
*cast<Constant>(GEP.getOperand(1));
assert(Sum && "Constant folding of longs failed!?");
GEP.setOperand(0, Src->getOperand(0));
GEP.setOperand(1, Sum);
AddUsesToWorkList(*Src); // Reduce use count of Src
return &GEP;
} else if (Src->getNumOperands() == 2) {
// Replace: gep (gep %P, long B), long A, ...
// With: T = long A+B; gep %P, T, ...
//
Value *Sum = BinaryOperator::create(Instruction::Add, Src->getOperand(1),
GEP.getOperand(1),
Src->getName()+".sum", &GEP);
GEP.setOperand(0, Src->getOperand(0));
GEP.setOperand(1, Sum);
WorkList.push_back(cast<Instruction>(Sum));
return &GEP;
} else if (*GEP.idx_begin() == Constant::getNullValue(Type::LongTy) &&
Src->getNumOperands() != 1) {
// Otherwise we can do the fold if the first index of the GEP is a zero
Indices.insert(Indices.end(), Src->idx_begin(), Src->idx_end());
Indices.insert(Indices.end(), GEP.idx_begin()+1, GEP.idx_end());
} else if (Src->getOperand(Src->getNumOperands()-1) ==
Constant::getNullValue(Type::LongTy)) {
// If the src gep ends with a constant array index, merge this get into
// it, even if we have a non-zero array index.
Indices.insert(Indices.end(), Src->idx_begin(), Src->idx_end()-1);
Indices.insert(Indices.end(), GEP.idx_begin(), GEP.idx_end());
}
if (!Indices.empty())
return new GetElementPtrInst(Src->getOperand(0), Indices, GEP.getName());
} else if (GlobalValue *GV = dyn_cast<GlobalValue>(GEP.getOperand(0))) {
// GEP of global variable. If all of the indices for this GEP are
// constants, we can promote this to a constexpr instead of an instruction.
// Scan for nonconstants...
std::vector<Constant*> Indices;
User::op_iterator I = GEP.idx_begin(), E = GEP.idx_end();
for (; I != E && isa<Constant>(*I); ++I)
Indices.push_back(cast<Constant>(*I));
if (I == E) { // If they are all constants...
Constant *CE =
ConstantExpr::getGetElementPtr(ConstantPointerRef::get(GV), Indices);
// Replace all uses of the GEP with the new constexpr...
return ReplaceInstUsesWith(GEP, CE);
}
}
return 0;
}
Instruction *InstCombiner::visitAllocationInst(AllocationInst &AI) {
// Convert: malloc Ty, C - where C is a constant != 1 into: malloc [C x Ty], 1
if (AI.isArrayAllocation()) // Check C != 1
if (const ConstantUInt *C = dyn_cast<ConstantUInt>(AI.getArraySize())) {
const Type *NewTy = ArrayType::get(AI.getAllocatedType(), C->getValue());
AllocationInst *New = 0;
// Create and insert the replacement instruction...
if (isa<MallocInst>(AI))
New = new MallocInst(NewTy, 0, AI.getName(), &AI);
else {
assert(isa<AllocaInst>(AI) && "Unknown type of allocation inst!");
New = new AllocaInst(NewTy, 0, AI.getName(), &AI);
}
// Scan to the end of the allocation instructions, to skip over a block of
// allocas if possible...
//
BasicBlock::iterator It = New;
while (isa<AllocationInst>(*It)) ++It;
// Now that I is pointing to the first non-allocation-inst in the block,
// insert our getelementptr instruction...
//
std::vector<Value*> Idx(2, Constant::getNullValue(Type::LongTy));
Value *V = new GetElementPtrInst(New, Idx, New->getName()+".sub", It);
// Now make everything use the getelementptr instead of the original
// allocation.
ReplaceInstUsesWith(AI, V);
return &AI;
}
return 0;
}
void InstCombiner::removeFromWorkList(Instruction *I) {
WorkList.erase(std::remove(WorkList.begin(), WorkList.end(), I),
WorkList.end());
}
bool InstCombiner::runOnFunction(Function &F) {
bool Changed = false;
WorkList.insert(WorkList.end(), inst_begin(F), inst_end(F));
while (!WorkList.empty()) {
Instruction *I = WorkList.back(); // Get an instruction from the worklist
WorkList.pop_back();
// Check to see if we can DCE or ConstantPropagate the instruction...
// Check to see if we can DIE the instruction...
if (isInstructionTriviallyDead(I)) {
// Add operands to the worklist...
for (unsigned i = 0, e = I->getNumOperands(); i != e; ++i)
if (Instruction *Op = dyn_cast<Instruction>(I->getOperand(i)))
WorkList.push_back(Op);
++NumDeadInst;
BasicBlock::iterator BBI = I;
if (dceInstruction(BBI)) {
removeFromWorkList(I);
continue;
}
}
// Instruction isn't dead, see if we can constant propagate it...
if (Constant *C = ConstantFoldInstruction(I)) {
// Add operands to the worklist...
for (unsigned i = 0, e = I->getNumOperands(); i != e; ++i)
if (Instruction *Op = dyn_cast<Instruction>(I->getOperand(i)))
WorkList.push_back(Op);
ReplaceInstUsesWith(*I, C);
++NumConstProp;
BasicBlock::iterator BBI = I;
if (dceInstruction(BBI)) {
removeFromWorkList(I);
continue;
}
}
// Now that we have an instruction, try combining it to simplify it...
if (Instruction *Result = visit(*I)) {
++NumCombined;
// Should we replace the old instruction with a new one?
if (Result != I) {
// Instructions can end up on the worklist more than once. Make sure
// we do not process an instruction that has been deleted.
removeFromWorkList(I);
ReplaceInstWithInst(I, Result);
} else {
BasicBlock::iterator II = I;
// If the instruction was modified, it's possible that it is now dead.
// if so, remove it.
if (dceInstruction(II)) {
// Instructions may end up in the worklist more than once. Erase them
// all.
removeFromWorkList(I);
Result = 0;
}
}
if (Result) {
WorkList.push_back(Result);
AddUsesToWorkList(*Result);
}
Changed = true;
}
}
return Changed;
}
Pass *createInstructionCombiningPass() {
return new InstCombiner();
}