RPCS3/llvm-mirror
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-12-25 21:16:19 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity
7c05e5d4d8
llvm-mirror/lib/Target/X86/InstSelectSimple.cpp
Chris Lattner ab9628ad18 Teach the instruction selector how to transform 'array' GEP computations into X86 
scaled indexes.  This allows us to compile GEP's like this:

int* %test([10 x { int, { int } }]* %X, int %Idx) {
        %Idx = cast int %Idx to long
        %X = getelementptr [10 x { int, { int } }]* %X, long 0, long %Idx, ubyte 1, ubyte 0
        ret int* %X
}

Into a single address computation:

test:
        mov %EAX, DWORD PTR [%ESP + 4]
        mov %ECX, DWORD PTR [%ESP + 8]
        lea %EAX, DWORD PTR [%EAX + 8*%ECX + 4]
        ret

Before it generated:
test:
        mov %EAX, DWORD PTR [%ESP + 4]
        mov %ECX, DWORD PTR [%ESP + 8]
        shl %ECX, 3
        add %EAX, %ECX
        lea %EAX, DWORD PTR [%EAX + 4]
        ret

This is useful for things like int/float/double arrays, as the indexing can be folded into
the loads&stores, reducing register pressure and decreasing the pressure on the decode unit.
With these changes, I expect our performance on 256.bzip2 and gzip to improve a lot.  On
bzip2 for example, we go from this:

10665 asm-printer           - Number of machine instrs printed
   40 ra-local              - Number of loads/stores folded into instructions
 1708 ra-local              - Number of loads added
 1532 ra-local              - Number of stores added
 1354 twoaddressinstruction - Number of instructions added
 1354 twoaddressinstruction - Number of two-address instructions
 2794 x86-peephole          - Number of peephole optimization performed

to this:
9873 asm-printer           - Number of machine instrs printed
  41 ra-local              - Number of loads/stores folded into instructions
1710 ra-local              - Number of loads added
1521 ra-local              - Number of stores added
 789 twoaddressinstruction - Number of instructions added
 789 twoaddressinstruction - Number of two-address instructions
2142 x86-peephole          - Number of peephole optimization performed

... and these types of instructions are often in tight loops.

Linear scan is also helped, but not as much.  It goes from:

8787 asm-printer           - Number of machine instrs printed
2389 liveintervals         - Number of identity moves eliminated after coalescing
2288 liveintervals         - Number of interval joins performed
3522 liveintervals         - Number of intervals after coalescing
5810 liveintervals         - Number of original intervals
 700 spiller               - Number of loads added
 487 spiller               - Number of stores added
 303 spiller               - Number of register spills
1354 twoaddressinstruction - Number of instructions added
1354 twoaddressinstruction - Number of two-address instructions
 363 x86-peephole          - Number of peephole optimization performed

to:

7982 asm-printer           - Number of machine instrs printed
1759 liveintervals         - Number of identity moves eliminated after coalescing
1658 liveintervals         - Number of interval joins performed
3282 liveintervals         - Number of intervals after coalescing
4940 liveintervals         - Number of original intervals
 635 spiller               - Number of loads added
 452 spiller               - Number of stores added
 288 spiller               - Number of register spills
 789 twoaddressinstruction - Number of instructions added
 789 twoaddressinstruction - Number of two-address instructions
 258 x86-peephole          - Number of peephole optimization performed

Though I'm not complaining about the drop in the number of intervals.  :)

llvm-svn: 11820
2004-02-25 07:00:55 +00:00

2725 lines
104 KiB
C++
Raw Blame History

//===-- InstSelectSimple.cpp - A simple instruction selector for x86 ------===//
//
// The LLVM Compiler Infrastructure
//
// This file was developed by the LLVM research group and is distributed under
// the University of Illinois Open Source License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
//
// This file defines a simple peephole instruction selector for the x86 target
//
//===----------------------------------------------------------------------===//
#include "X86.h"
#include "X86InstrBuilder.h"
#include "X86InstrInfo.h"
#include "llvm/Constants.h"
#include "llvm/DerivedTypes.h"
#include "llvm/Function.h"
#include "llvm/Instructions.h"
#include "llvm/IntrinsicLowering.h"
#include "llvm/Pass.h"
#include "llvm/CodeGen/MachineConstantPool.h"
#include "llvm/CodeGen/MachineFrameInfo.h"
#include "llvm/CodeGen/MachineFunction.h"
#include "llvm/CodeGen/MachineInstrBuilder.h"
#include "llvm/CodeGen/SSARegMap.h"
#include "llvm/Target/MRegisterInfo.h"
#include "llvm/Target/TargetMachine.h"
#include "llvm/Support/GetElementPtrTypeIterator.h"
#include "llvm/Support/InstVisitor.h"
#include "llvm/Support/CFG.h"
#include "Support/Statistic.h"
using namespace llvm;
namespace {
Statistic<>
NumFPKill("x86-codegen", "Number of FP_REG_KILL instructions added");
}
/// BMI - A special BuildMI variant that takes an iterator to insert the
/// instruction at as well as a basic block. This is the version for when you
/// have a destination register in mind.
inline static MachineInstrBuilder BMI(MachineBasicBlock *MBB,
MachineBasicBlock::iterator I,
int Opcode, unsigned NumOperands,
unsigned DestReg) {
MachineInstr *MI = new MachineInstr(Opcode, NumOperands+1, true, true);
MBB->insert(I, MI);
return MachineInstrBuilder(MI).addReg(DestReg, MachineOperand::Def);
}
/// BMI - A special BuildMI variant that takes an iterator to insert the
/// instruction at as well as a basic block.
inline static MachineInstrBuilder BMI(MachineBasicBlock *MBB,
MachineBasicBlock::iterator I,
int Opcode, unsigned NumOperands) {
MachineInstr *MI = new MachineInstr(Opcode, NumOperands, true, true);
MBB->insert(I, MI);
return MachineInstrBuilder(MI);
}
namespace {
struct ISel : public FunctionPass, InstVisitor<ISel> {
TargetMachine &TM;
MachineFunction *F; // The function we are compiling into
MachineBasicBlock *BB; // The current MBB we are compiling
int VarArgsFrameIndex; // FrameIndex for start of varargs area
int ReturnAddressIndex; // FrameIndex for the return address
std::map<Value*, unsigned> RegMap; // Mapping between Val's and SSA Regs
// MBBMap - Mapping between LLVM BB -> Machine BB
std::map<const BasicBlock*, MachineBasicBlock*> MBBMap;
ISel(TargetMachine &tm) : TM(tm), F(0), BB(0) {}
/// runOnFunction - Top level implementation of instruction selection for
/// the entire function.
///
bool runOnFunction(Function &Fn) {
// First pass over the function, lower any unknown intrinsic functions
// with the IntrinsicLowering class.
LowerUnknownIntrinsicFunctionCalls(Fn);
F = &MachineFunction::construct(&Fn, TM);
// Create all of the machine basic blocks for the function...
for (Function::iterator I = Fn.begin(), E = Fn.end(); I != E; ++I)
F->getBasicBlockList().push_back(MBBMap[I] = new MachineBasicBlock(I));
BB = &F->front();
// Set up a frame object for the return address. This is used by the
// llvm.returnaddress & llvm.frameaddress intrinisics.
ReturnAddressIndex = F->getFrameInfo()->CreateFixedObject(4, -4);
// Copy incoming arguments off of the stack...
LoadArgumentsToVirtualRegs(Fn);
// Instruction select everything except PHI nodes
visit(Fn);
// Select the PHI nodes
SelectPHINodes();
// Insert the FP_REG_KILL instructions into blocks that need them.
InsertFPRegKills();
RegMap.clear();
MBBMap.clear();
F = 0;
// We always build a machine code representation for the function
return true;
}
virtual const char *getPassName() const {
return "X86 Simple Instruction Selection";
}
/// visitBasicBlock - This method is called when we are visiting a new basic
/// block. This simply creates a new MachineBasicBlock to emit code into
/// and adds it to the current MachineFunction. Subsequent visit* for
/// instructions will be invoked for all instructions in the basic block.
///
void visitBasicBlock(BasicBlock &LLVM_BB) {
BB = MBBMap[&LLVM_BB];
}
/// LowerUnknownIntrinsicFunctionCalls - This performs a prepass over the
/// function, lowering any calls to unknown intrinsic functions into the
/// equivalent LLVM code.
void LowerUnknownIntrinsicFunctionCalls(Function &F);
/// LoadArgumentsToVirtualRegs - Load all of the arguments to this function
/// from the stack into virtual registers.
///
void LoadArgumentsToVirtualRegs(Function &F);
/// SelectPHINodes - Insert machine code to generate phis. This is tricky
/// because we have to generate our sources into the source basic blocks,
/// not the current one.
///
void SelectPHINodes();
/// InsertFPRegKills - Insert FP_REG_KILL instructions into basic blocks
/// that need them. This only occurs due to the floating point stackifier
/// not being aggressive enough to handle arbitrary global stackification.
///
void InsertFPRegKills();
// Visitation methods for various instructions. These methods simply emit
// fixed X86 code for each instruction.
//
// Control flow operators
void visitReturnInst(ReturnInst &RI);
void visitBranchInst(BranchInst &BI);
struct ValueRecord {
Value *Val;
unsigned Reg;
const Type *Ty;
ValueRecord(unsigned R, const Type *T) : Val(0), Reg(R), Ty(T) {}
ValueRecord(Value *V) : Val(V), Reg(0), Ty(V->getType()) {}
};
void doCall(const ValueRecord &Ret, MachineInstr *CallMI,
const std::vector<ValueRecord> &Args);
void visitCallInst(CallInst &I);
void visitIntrinsicCall(Intrinsic::ID ID, CallInst &I);
// Arithmetic operators
void visitSimpleBinary(BinaryOperator &B, unsigned OpcodeClass);
void visitAdd(BinaryOperator &B) { visitSimpleBinary(B, 0); }
void visitSub(BinaryOperator &B) { visitSimpleBinary(B, 1); }
void doMultiply(MachineBasicBlock *MBB, MachineBasicBlock::iterator MBBI,
unsigned DestReg, const Type *DestTy,
unsigned Op0Reg, unsigned Op1Reg);
void doMultiplyConst(MachineBasicBlock *MBB,
MachineBasicBlock::iterator MBBI,
unsigned DestReg, const Type *DestTy,
unsigned Op0Reg, unsigned Op1Val);
void visitMul(BinaryOperator &B);
void visitDiv(BinaryOperator &B) { visitDivRem(B); }
void visitRem(BinaryOperator &B) { visitDivRem(B); }
void visitDivRem(BinaryOperator &B);
// Bitwise operators
void visitAnd(BinaryOperator &B) { visitSimpleBinary(B, 2); }
void visitOr (BinaryOperator &B) { visitSimpleBinary(B, 3); }
void visitXor(BinaryOperator &B) { visitSimpleBinary(B, 4); }
// Comparison operators...
void visitSetCondInst(SetCondInst &I);
unsigned EmitComparison(unsigned OpNum, Value *Op0, Value *Op1,
MachineBasicBlock *MBB,
MachineBasicBlock::iterator MBBI);
// Memory Instructions
void visitLoadInst(LoadInst &I);
void visitStoreInst(StoreInst &I);
void visitGetElementPtrInst(GetElementPtrInst &I);
void visitAllocaInst(AllocaInst &I);
void visitMallocInst(MallocInst &I);
void visitFreeInst(FreeInst &I);
// Other operators
void visitShiftInst(ShiftInst &I);
void visitPHINode(PHINode &I) {} // PHI nodes handled by second pass
void visitCastInst(CastInst &I);
void visitVANextInst(VANextInst &I);
void visitVAArgInst(VAArgInst &I);
void visitInstruction(Instruction &I) {
std::cerr << "Cannot instruction select: " << I;
abort();
}
/// promote32 - Make a value 32-bits wide, and put it somewhere.
///
void promote32(unsigned targetReg, const ValueRecord &VR);
// getGEPIndex - This is used to fold GEP instructions into X86 addressing
// expressions.
void getGEPIndex(MachineBasicBlock *MBB, MachineBasicBlock::iterator IP,
std::vector<Value*> &GEPOps,
std::vector<const Type*> &GEPTypes, unsigned &BaseReg,
unsigned &Scale, unsigned &IndexReg, unsigned &Disp);
/// isGEPFoldable - Return true if the specified GEP can be completely
/// folded into the addressing mode of a load/store or lea instruction.
bool isGEPFoldable(MachineBasicBlock *MBB,
Value *Src, User::op_iterator IdxBegin,
User::op_iterator IdxEnd, unsigned &BaseReg,
unsigned &Scale, unsigned &IndexReg, unsigned &Disp);
/// emitGEPOperation - Common code shared between visitGetElementPtrInst and
/// constant expression GEP support.
///
void emitGEPOperation(MachineBasicBlock *BB, MachineBasicBlock::iterator IP,
Value *Src, User::op_iterator IdxBegin,
User::op_iterator IdxEnd, unsigned TargetReg);
/// emitCastOperation - Common code shared between visitCastInst and
/// constant expression cast support.
void emitCastOperation(MachineBasicBlock *BB,MachineBasicBlock::iterator IP,
Value *Src, const Type *DestTy, unsigned TargetReg);
/// emitSimpleBinaryOperation - Common code shared between visitSimpleBinary
/// and constant expression support.
void emitSimpleBinaryOperation(MachineBasicBlock *BB,
MachineBasicBlock::iterator IP,
Value *Op0, Value *Op1,
unsigned OperatorClass, unsigned TargetReg);
void emitDivRemOperation(MachineBasicBlock *BB,
MachineBasicBlock::iterator IP,
unsigned Op0Reg, unsigned Op1Reg, bool isDiv,
const Type *Ty, unsigned TargetReg);
/// emitSetCCOperation - Common code shared between visitSetCondInst and
/// constant expression support.
void emitSetCCOperation(MachineBasicBlock *BB,
MachineBasicBlock::iterator IP,
Value *Op0, Value *Op1, unsigned Opcode,
unsigned TargetReg);
/// emitShiftOperation - Common code shared between visitShiftInst and
/// constant expression support.
void emitShiftOperation(MachineBasicBlock *MBB,
MachineBasicBlock::iterator IP,
Value *Op, Value *ShiftAmount, bool isLeftShift,
const Type *ResultTy, unsigned DestReg);
/// copyConstantToRegister - Output the instructions required to put the
/// specified constant into the specified register.
///
void copyConstantToRegister(MachineBasicBlock *MBB,
MachineBasicBlock::iterator MBBI,
Constant *C, unsigned Reg);
/// makeAnotherReg - This method returns the next register number we haven't
/// yet used.
///
/// Long values are handled somewhat specially. They are always allocated
/// as pairs of 32 bit integer values. The register number returned is the
/// lower 32 bits of the long value, and the regNum+1 is the upper 32 bits
/// of the long value.
///
unsigned makeAnotherReg(const Type *Ty) {
assert(dynamic_cast<const X86RegisterInfo*>(TM.getRegisterInfo()) &&
"Current target doesn't have X86 reg info??");
const X86RegisterInfo *MRI =
static_cast<const X86RegisterInfo*>(TM.getRegisterInfo());
if (Ty == Type::LongTy || Ty == Type::ULongTy) {
const TargetRegisterClass *RC = MRI->getRegClassForType(Type::IntTy);
// Create the lower part
F->getSSARegMap()->createVirtualRegister(RC);
// Create the upper part.
return F->getSSARegMap()->createVirtualRegister(RC)-1;
}
// Add the mapping of regnumber => reg class to MachineFunction
const TargetRegisterClass *RC = MRI->getRegClassForType(Ty);
return F->getSSARegMap()->createVirtualRegister(RC);
}
/// getReg - This method turns an LLVM value into a register number. This
/// is guaranteed to produce the same register number for a particular value
/// every time it is queried.
///
unsigned getReg(Value &V) { return getReg(&V); } // Allow references
unsigned getReg(Value *V) {
// Just append to the end of the current bb.
MachineBasicBlock::iterator It = BB->end();
return getReg(V, BB, It);
}
unsigned getReg(Value *V, MachineBasicBlock *MBB,
MachineBasicBlock::iterator IPt) {
unsigned &Reg = RegMap[V];
if (Reg == 0) {
Reg = makeAnotherReg(V->getType());
RegMap[V] = Reg;
}
// If this operand is a constant, emit the code to copy the constant into
// the register here...
//
if (Constant *C = dyn_cast<Constant>(V)) {
copyConstantToRegister(MBB, IPt, C, Reg);
RegMap.erase(V); // Assign a new name to this constant if ref'd again
} else if (GlobalValue *GV = dyn_cast<GlobalValue>(V)) {
// Move the address of the global into the register
BMI(MBB, IPt, X86::MOVri32, 1, Reg).addGlobalAddress(GV);
RegMap.erase(V); // Assign a new name to this address if ref'd again
}
return Reg;
}
};
}
/// TypeClass - Used by the X86 backend to group LLVM types by their basic X86
/// Representation.
///
enum TypeClass {
cByte, cShort, cInt, cFP, cLong
};
/// getClass - Turn a primitive type into a "class" number which is based on the
/// size of the type, and whether or not it is floating point.
///
static inline TypeClass getClass(const Type *Ty) {
switch (Ty->getPrimitiveID()) {
case Type::SByteTyID:
case Type::UByteTyID: return cByte; // Byte operands are class #0
case Type::ShortTyID:
case Type::UShortTyID: return cShort; // Short operands are class #1
case Type::IntTyID:
case Type::UIntTyID:
case Type::PointerTyID: return cInt; // Int's and pointers are class #2
case Type::FloatTyID:
case Type::DoubleTyID: return cFP; // Floating Point is #3
case Type::LongTyID:
case Type::ULongTyID: return cLong; // Longs are class #4
default:
assert(0 && "Invalid type to getClass!");
return cByte; // not reached
}
}
// getClassB - Just like getClass, but treat boolean values as bytes.
static inline TypeClass getClassB(const Type *Ty) {
if (Ty == Type::BoolTy) return cByte;
return getClass(Ty);
}
/// copyConstantToRegister - Output the instructions required to put the
/// specified constant into the specified register.
///
void ISel::copyConstantToRegister(MachineBasicBlock *MBB,
MachineBasicBlock::iterator IP,
Constant *C, unsigned R) {
if (ConstantExpr *CE = dyn_cast<ConstantExpr>(C)) {
unsigned Class = 0;
switch (CE->getOpcode()) {
case Instruction::GetElementPtr:
emitGEPOperation(MBB, IP, CE->getOperand(0),
CE->op_begin()+1, CE->op_end(), R);
return;
case Instruction::Cast:
emitCastOperation(MBB, IP, CE->getOperand(0), CE->getType(), R);
return;
case Instruction::Xor: ++Class; // FALL THROUGH
case Instruction::Or: ++Class; // FALL THROUGH
case Instruction::And: ++Class; // FALL THROUGH
case Instruction::Sub: ++Class; // FALL THROUGH
case Instruction::Add:
emitSimpleBinaryOperation(MBB, IP, CE->getOperand(0), CE->getOperand(1),
Class, R);
return;
case Instruction::Mul: {
unsigned Op0Reg = getReg(CE->getOperand(0), MBB, IP);
unsigned Op1Reg = getReg(CE->getOperand(1), MBB, IP);
doMultiply(MBB, IP, R, CE->getType(), Op0Reg, Op1Reg);
return;
}
case Instruction::Div:
case Instruction::Rem: {
unsigned Op0Reg = getReg(CE->getOperand(0), MBB, IP);
unsigned Op1Reg = getReg(CE->getOperand(1), MBB, IP);
emitDivRemOperation(MBB, IP, Op0Reg, Op1Reg,
CE->getOpcode() == Instruction::Div,
CE->getType(), R);
return;
}
case Instruction::SetNE:
case Instruction::SetEQ:
case Instruction::SetLT:
case Instruction::SetGT:
case Instruction::SetLE:
case Instruction::SetGE:
emitSetCCOperation(MBB, IP, CE->getOperand(0), CE->getOperand(1),
CE->getOpcode(), R);
return;
case Instruction::Shl:
case Instruction::Shr:
emitShiftOperation(MBB, IP, CE->getOperand(0), CE->getOperand(1),
CE->getOpcode() == Instruction::Shl, CE->getType(), R);
return;
default:
std::cerr << "Offending expr: " << C << "\n";
assert(0 && "Constant expression not yet handled!\n");
}
}
if (C->getType()->isIntegral()) {
unsigned Class = getClassB(C->getType());
if (Class == cLong) {
// Copy the value into the register pair.
uint64_t Val = cast<ConstantInt>(C)->getRawValue();
BMI(MBB, IP, X86::MOVri32, 1, R).addZImm(Val & 0xFFFFFFFF);
BMI(MBB, IP, X86::MOVri32, 1, R+1).addZImm(Val >> 32);
return;
}
assert(Class <= cInt && "Type not handled yet!");
static const unsigned IntegralOpcodeTab[] = {
X86::MOVri8, X86::MOVri16, X86::MOVri32
};
if (C->getType() == Type::BoolTy) {
BMI(MBB, IP, X86::MOVri8, 1, R).addZImm(C == ConstantBool::True);
} else {
ConstantInt *CI = cast<ConstantInt>(C);
BMI(MBB, IP, IntegralOpcodeTab[Class], 1, R).addZImm(CI->getRawValue());
}
} else if (ConstantFP *CFP = dyn_cast<ConstantFP>(C)) {
if (CFP->isExactlyValue(+0.0))
BMI(MBB, IP, X86::FLD0, 0, R);
else if (CFP->isExactlyValue(+1.0))
BMI(MBB, IP, X86::FLD1, 0, R);
else {
// Otherwise we need to spill the constant to memory...
MachineConstantPool *CP = F->getConstantPool();
unsigned CPI = CP->getConstantPoolIndex(CFP);
const Type *Ty = CFP->getType();
assert(Ty == Type::FloatTy || Ty == Type::DoubleTy && "Unknown FP type!");
unsigned LoadOpcode = Ty == Type::FloatTy ? X86::FLDr32 : X86::FLDr64;
addConstantPoolReference(BMI(MBB, IP, LoadOpcode, 4, R), CPI);
}
} else if (isa<ConstantPointerNull>(C)) {
// Copy zero (null pointer) to the register.
BMI(MBB, IP, X86::MOVri32, 1, R).addZImm(0);
} else if (ConstantPointerRef *CPR = dyn_cast<ConstantPointerRef>(C)) {
BMI(MBB, IP, X86::MOVri32, 1, R).addGlobalAddress(CPR->getValue());
} else {
std::cerr << "Offending constant: " << C << "\n";
assert(0 && "Type not handled yet!");
}
}
/// LoadArgumentsToVirtualRegs - Load all of the arguments to this function from
/// the stack into virtual registers.
///
void ISel::LoadArgumentsToVirtualRegs(Function &Fn) {
// Emit instructions to load the arguments... On entry to a function on the
// X86, the stack frame looks like this:
//
// [ESP] -- return address
// [ESP + 4] -- first argument (leftmost lexically)
// [ESP + 8] -- second argument, if first argument is four bytes in size
// ...
//
unsigned ArgOffset = 0; // Frame mechanisms handle retaddr slot
MachineFrameInfo *MFI = F->getFrameInfo();
for (Function::aiterator I = Fn.abegin(), E = Fn.aend(); I != E; ++I) {
unsigned Reg = getReg(*I);
int FI; // Frame object index
switch (getClassB(I->getType())) {
case cByte:
FI = MFI->CreateFixedObject(1, ArgOffset);
addFrameReference(BuildMI(BB, X86::MOVrm8, 4, Reg), FI);
break;
case cShort:
FI = MFI->CreateFixedObject(2, ArgOffset);
addFrameReference(BuildMI(BB, X86::MOVrm16, 4, Reg), FI);
break;
case cInt:
FI = MFI->CreateFixedObject(4, ArgOffset);
addFrameReference(BuildMI(BB, X86::MOVrm32, 4, Reg), FI);
break;
case cLong:
FI = MFI->CreateFixedObject(8, ArgOffset);
addFrameReference(BuildMI(BB, X86::MOVrm32, 4, Reg), FI);
addFrameReference(BuildMI(BB, X86::MOVrm32, 4, Reg+1), FI, 4);
ArgOffset += 4; // longs require 4 additional bytes
break;
case cFP:
unsigned Opcode;
if (I->getType() == Type::FloatTy) {
Opcode = X86::FLDr32;
FI = MFI->CreateFixedObject(4, ArgOffset);
} else {
Opcode = X86::FLDr64;
FI = MFI->CreateFixedObject(8, ArgOffset);
ArgOffset += 4; // doubles require 4 additional bytes
}
addFrameReference(BuildMI(BB, Opcode, 4, Reg), FI);
break;
default:
assert(0 && "Unhandled argument type!");
}
ArgOffset += 4; // Each argument takes at least 4 bytes on the stack...
}
// If the function takes variable number of arguments, add a frame offset for
// the start of the first vararg value... this is used to expand
// llvm.va_start.
if (Fn.getFunctionType()->isVarArg())
VarArgsFrameIndex = MFI->CreateFixedObject(1, ArgOffset);
}
/// SelectPHINodes - Insert machine code to generate phis. This is tricky
/// because we have to generate our sources into the source basic blocks, not
/// the current one.
///
void ISel::SelectPHINodes() {
const TargetInstrInfo &TII = TM.getInstrInfo();
const Function &LF = *F->getFunction(); // The LLVM function...
for (Function::const_iterator I = LF.begin(), E = LF.end(); I != E; ++I) {
const BasicBlock *BB = I;
MachineBasicBlock *MBB = MBBMap[I];
// Loop over all of the PHI nodes in the LLVM basic block...
MachineBasicBlock::iterator instr = MBB->begin();
for (BasicBlock::const_iterator I = BB->begin();
PHINode *PN = const_cast<PHINode*>(dyn_cast<PHINode>(I)); ++I) {
// Create a new machine instr PHI node, and insert it.
unsigned PHIReg = getReg(*PN);
MachineInstr *PhiMI = BuildMI(X86::PHI, PN->getNumOperands(), PHIReg);
MBB->insert(instr, PhiMI);
MachineInstr *LongPhiMI = 0;
if (PN->getType() == Type::LongTy || PN->getType() == Type::ULongTy) {
LongPhiMI = BuildMI(X86::PHI, PN->getNumOperands(), PHIReg+1);
MBB->insert(instr, LongPhiMI);
}
// PHIValues - Map of blocks to incoming virtual registers. We use this
// so that we only initialize one incoming value for a particular block,
// even if the block has multiple entries in the PHI node.
//
std::map<MachineBasicBlock*, unsigned> PHIValues;
for (unsigned i = 0, e = PN->getNumIncomingValues(); i != e; ++i) {
MachineBasicBlock *PredMBB = MBBMap[PN->getIncomingBlock(i)];
unsigned ValReg;
std::map<MachineBasicBlock*, unsigned>::iterator EntryIt =
PHIValues.lower_bound(PredMBB);
if (EntryIt != PHIValues.end() && EntryIt->first == PredMBB) {
// We already inserted an initialization of the register for this
// predecessor. Recycle it.
ValReg = EntryIt->second;
} else {
// Get the incoming value into a virtual register.
//
Value *Val = PN->getIncomingValue(i);
// If this is a constant or GlobalValue, we may have to insert code
// into the basic block to compute it into a virtual register.
if (isa<Constant>(Val) || isa<GlobalValue>(Val)) {
// Because we don't want to clobber any values which might be in
// physical registers with the computation of this constant (which
// might be arbitrarily complex if it is a constant expression),
// just insert the computation at the top of the basic block.
MachineBasicBlock::iterator PI = PredMBB->begin();
// Skip over any PHI nodes though!
while (PI != PredMBB->end() && PI->getOpcode() == X86::PHI)
++PI;
ValReg = getReg(Val, PredMBB, PI);
} else {
ValReg = getReg(Val);
}
// Remember that we inserted a value for this PHI for this predecessor
PHIValues.insert(EntryIt, std::make_pair(PredMBB, ValReg));
}
PhiMI->addRegOperand(ValReg);
PhiMI->addMachineBasicBlockOperand(PredMBB);
if (LongPhiMI) {
LongPhiMI->addRegOperand(ValReg+1);
LongPhiMI->addMachineBasicBlockOperand(PredMBB);
}
}
}
}
}
/// RequiresFPRegKill - The floating point stackifier pass cannot insert
/// compensation code on critical edges. As such, it requires that we kill all
/// FP registers on the exit from any blocks that either ARE critical edges, or
/// branch to a block that has incoming critical edges.
///
/// Note that this kill instruction will eventually be eliminated when
/// restrictions in the stackifier are relaxed.
///
static bool RequiresFPRegKill(const BasicBlock *BB) {
#if 0
for (succ_const_iterator SI = succ_begin(BB), E = succ_end(BB); SI!=E; ++SI) {
const BasicBlock *Succ = *SI;
pred_const_iterator PI = pred_begin(Succ), PE = pred_end(Succ);
++PI; // Block have at least one predecessory
if (PI != PE) { // If it has exactly one, this isn't crit edge
// If this block has more than one predecessor, check all of the
// predecessors to see if they have multiple successors. If so, then the
// block we are analyzing needs an FPRegKill.
for (PI = pred_begin(Succ); PI != PE; ++PI) {
const BasicBlock *Pred = *PI;
succ_const_iterator SI2 = succ_begin(Pred);
++SI2; // There must be at least one successor of this block.
if (SI2 != succ_end(Pred))
return true; // Yes, we must insert the kill on this edge.
}
}
}
// If we got this far, there is no need to insert the kill instruction.
return false;
#else
return true;
#endif
}
// InsertFPRegKills - Insert FP_REG_KILL instructions into basic blocks that
// need them. This only occurs due to the floating point stackifier not being
// aggressive enough to handle arbitrary global stackification.
//
// Currently we insert an FP_REG_KILL instruction into each block that uses or
// defines a floating point virtual register.
//
// When the global register allocators (like linear scan) finally update live
// variable analysis, we can keep floating point values in registers across
// portions of the CFG that do not involve critical edges. This will be a big
// win, but we are waiting on the global allocators before we can do this.
//
// With a bit of work, the floating point stackifier pass can be enhanced to
// break critical edges as needed (to make a place to put compensation code),
// but this will require some infrastructure improvements as well.
//
void ISel::InsertFPRegKills() {
SSARegMap &RegMap = *F->getSSARegMap();
for (MachineFunction::iterator BB = F->begin(), E = F->end(); BB != E; ++BB) {
for (MachineBasicBlock::iterator I = BB->begin(), E = BB->end(); I!=E; ++I)
for (unsigned i = 0, e = I->getNumOperands(); i != e; ++i)
if (I->getOperand(i).isRegister()) {
unsigned Reg = I->getOperand(i).getReg();
if (MRegisterInfo::isVirtualRegister(Reg))
if (RegMap.getRegClass(Reg)->getSize() == 10)
goto UsesFPReg;
}
// If we haven't found an FP register use or def in this basic block, check
// to see if any of our successors has an FP PHI node, which will cause a
// copy to be inserted into this block.
for (succ_const_iterator SI = succ_begin(BB->getBasicBlock()),
E = succ_end(BB->getBasicBlock()); SI != E; ++SI) {
MachineBasicBlock *SBB = MBBMap[*SI];
for (MachineBasicBlock::iterator I = SBB->begin();
I != SBB->end() && I->getOpcode() == X86::PHI; ++I) {
if (RegMap.getRegClass(I->getOperand(0).getReg())->getSize() == 10)
goto UsesFPReg;
}
}
continue;
UsesFPReg:
// Okay, this block uses an FP register. If the block has successors (ie,
// it's not an unwind/return), insert the FP_REG_KILL instruction.
if (BB->getBasicBlock()->getTerminator()->getNumSuccessors() &&
RequiresFPRegKill(BB->getBasicBlock())) {
BMI(BB, BB->getFirstTerminator(), X86::FP_REG_KILL, 0);
++NumFPKill;
}
}
}
// canFoldSetCCIntoBranch - Return the setcc instruction if we can fold it into
// the conditional branch instruction which is the only user of the cc
// instruction. This is the case if the conditional branch is the only user of
// the setcc, and if the setcc is in the same basic block as the conditional
// branch. We also don't handle long arguments below, so we reject them here as
// well.
//
static SetCondInst *canFoldSetCCIntoBranch(Value *V) {
if (SetCondInst *SCI = dyn_cast<SetCondInst>(V))
if (SCI->hasOneUse() && isa<BranchInst>(SCI->use_back()) &&
SCI->getParent() == cast<BranchInst>(SCI->use_back())->getParent()) {
const Type *Ty = SCI->getOperand(0)->getType();
if (Ty != Type::LongTy && Ty != Type::ULongTy)
return SCI;
}
return 0;
}
// Return a fixed numbering for setcc instructions which does not depend on the
// order of the opcodes.
//
static unsigned getSetCCNumber(unsigned Opcode) {
switch(Opcode) {
default: assert(0 && "Unknown setcc instruction!");
case Instruction::SetEQ: return 0;
case Instruction::SetNE: return 1;
case Instruction::SetLT: return 2;
case Instruction::SetGE: return 3;
case Instruction::SetGT: return 4;
case Instruction::SetLE: return 5;
}
}
// LLVM -> X86 signed X86 unsigned
// ----- ---------- ------------
// seteq -> sete sete
// setne -> setne setne
// setlt -> setl setb
// setge -> setge setae
// setgt -> setg seta
// setle -> setle setbe
// ----
// sets // Used by comparison with 0 optimization
// setns
static const unsigned SetCCOpcodeTab[2][8] = {
{ X86::SETEr, X86::SETNEr, X86::SETBr, X86::SETAEr, X86::SETAr, X86::SETBEr,
0, 0 },
{ X86::SETEr, X86::SETNEr, X86::SETLr, X86::SETGEr, X86::SETGr, X86::SETLEr,
X86::SETSr, X86::SETNSr },
};
// EmitComparison - This function emits a comparison of the two operands,
// returning the extended setcc code to use.
unsigned ISel::EmitComparison(unsigned OpNum, Value *Op0, Value *Op1,
MachineBasicBlock *MBB,
MachineBasicBlock::iterator IP) {
// The arguments are already supposed to be of the same type.
const Type *CompTy = Op0->getType();
unsigned Class = getClassB(CompTy);
unsigned Op0r = getReg(Op0, MBB, IP);
// Special case handling of: cmp R, i
if (Class == cByte || Class == cShort || Class == cInt)
if (ConstantInt *CI = dyn_cast<ConstantInt>(Op1)) {
uint64_t Op1v = cast<ConstantInt>(CI)->getRawValue();
// Mask off any upper bits of the constant, if there are any...
Op1v &= (1ULL << (8 << Class)) - 1;
// If this is a comparison against zero, emit more efficient code. We
// can't handle unsigned comparisons against zero unless they are == or
// !=. These should have been strength reduced already anyway.
if (Op1v == 0 && (CompTy->isSigned() || OpNum < 2)) {
static const unsigned TESTTab[] = {
X86::TESTrr8, X86::TESTrr16, X86::TESTrr32
};
BMI(MBB, IP, TESTTab[Class], 2).addReg(Op0r).addReg(Op0r);
if (OpNum == 2) return 6; // Map jl -> js
if (OpNum == 3) return 7; // Map jg -> jns
return OpNum;
}
static const unsigned CMPTab[] = {
X86::CMPri8, X86::CMPri16, X86::CMPri32
};
BMI(MBB, IP, CMPTab[Class], 2).addReg(Op0r).addZImm(Op1v);
return OpNum;
}
// Special case handling of comparison against +/- 0.0
if (ConstantFP *CFP = dyn_cast<ConstantFP>(Op1))
if (CFP->isExactlyValue(+0.0) || CFP->isExactlyValue(-0.0)) {
BMI(MBB, IP, X86::FTST, 1).addReg(Op0r);
BMI(MBB, IP, X86::FNSTSWr8, 0);
BMI(MBB, IP, X86::SAHF, 1);
return OpNum;
}
unsigned Op1r = getReg(Op1, MBB, IP);
switch (Class) {
default: assert(0 && "Unknown type class!");
// Emit: cmp <var1>, <var2> (do the comparison). We can
// compare 8-bit with 8-bit, 16-bit with 16-bit, 32-bit with
// 32-bit.
case cByte:
BMI(MBB, IP, X86::CMPrr8, 2).addReg(Op0r).addReg(Op1r);
break;
case cShort:
BMI(MBB, IP, X86::CMPrr16, 2).addReg(Op0r).addReg(Op1r);
break;
case cInt:
BMI(MBB, IP, X86::CMPrr32, 2).addReg(Op0r).addReg(Op1r);
break;
case cFP:
BMI(MBB, IP, X86::FpUCOM, 2).addReg(Op0r).addReg(Op1r);
BMI(MBB, IP, X86::FNSTSWr8, 0);
BMI(MBB, IP, X86::SAHF, 1);
break;
case cLong:
if (OpNum < 2) { // seteq, setne
unsigned LoTmp = makeAnotherReg(Type::IntTy);
unsigned HiTmp = makeAnotherReg(Type::IntTy);
unsigned FinalTmp = makeAnotherReg(Type::IntTy);
BMI(MBB, IP, X86::XORrr32, 2, LoTmp).addReg(Op0r).addReg(Op1r);
BMI(MBB, IP, X86::XORrr32, 2, HiTmp).addReg(Op0r+1).addReg(Op1r+1);
BMI(MBB, IP, X86::ORrr32, 2, FinalTmp).addReg(LoTmp).addReg(HiTmp);
break; // Allow the sete or setne to be generated from flags set by OR
} else {
// Emit a sequence of code which compares the high and low parts once
// each, then uses a conditional move to handle the overflow case. For
// example, a setlt for long would generate code like this:
//
// AL = lo(op1) < lo(op2) // Signedness depends on operands
// BL = hi(op1) < hi(op2) // Always unsigned comparison
// dest = hi(op1) == hi(op2) ? AL : BL;
//
// FIXME: This would be much better if we had hierarchical register
// classes! Until then, hardcode registers so that we can deal with their
// aliases (because we don't have conditional byte moves).
//
BMI(MBB, IP, X86::CMPrr32, 2).addReg(Op0r).addReg(Op1r);
BMI(MBB, IP, SetCCOpcodeTab[0][OpNum], 0, X86::AL);
BMI(MBB, IP, X86::CMPrr32, 2).addReg(Op0r+1).addReg(Op1r+1);
BMI(MBB, IP, SetCCOpcodeTab[CompTy->isSigned()][OpNum], 0, X86::BL);
BMI(MBB, IP, X86::IMPLICIT_DEF, 0, X86::BH);
BMI(MBB, IP, X86::IMPLICIT_DEF, 0, X86::AH);
BMI(MBB, IP, X86::CMOVErr16, 2, X86::BX).addReg(X86::BX).addReg(X86::AX);
// NOTE: visitSetCondInst knows that the value is dumped into the BL
// register at this point for long values...
return OpNum;
}
}
return OpNum;
}
/// SetCC instructions - Here we just emit boilerplate code to set a byte-sized
/// register, then move it to wherever the result should be.
///
void ISel::visitSetCondInst(SetCondInst &I) {
if (canFoldSetCCIntoBranch(&I)) return; // Fold this into a branch...
unsigned DestReg = getReg(I);
MachineBasicBlock::iterator MII = BB->end();
emitSetCCOperation(BB, MII, I.getOperand(0), I.getOperand(1), I.getOpcode(),
DestReg);
}
/// emitSetCCOperation - Common code shared between visitSetCondInst and
/// constant expression support.
void ISel::emitSetCCOperation(MachineBasicBlock *MBB,
MachineBasicBlock::iterator IP,
Value *Op0, Value *Op1, unsigned Opcode,
unsigned TargetReg) {
unsigned OpNum = getSetCCNumber(Opcode);
OpNum = EmitComparison(OpNum, Op0, Op1, MBB, IP);
const Type *CompTy = Op0->getType();
unsigned CompClass = getClassB(CompTy);
bool isSigned = CompTy->isSigned() && CompClass != cFP;
if (CompClass != cLong || OpNum < 2) {
// Handle normal comparisons with a setcc instruction...
BMI(MBB, IP, SetCCOpcodeTab[isSigned][OpNum], 0, TargetReg);
} else {
// Handle long comparisons by copying the value which is already in BL into
// the register we want...
BMI(MBB, IP, X86::MOVrr8, 1, TargetReg).addReg(X86::BL);
}
}
/// promote32 - Emit instructions to turn a narrow operand into a 32-bit-wide
/// operand, in the specified target register.
void ISel::promote32(unsigned targetReg, const ValueRecord &VR) {
bool isUnsigned = VR.Ty->isUnsigned();
// Make sure we have the register number for this value...
unsigned Reg = VR.Val ? getReg(VR.Val) : VR.Reg;
switch (getClassB(VR.Ty)) {
case cByte:
// Extend value into target register (8->32)
if (isUnsigned)
BuildMI(BB, X86::MOVZXr32r8, 1, targetReg).addReg(Reg);
else
BuildMI(BB, X86::MOVSXr32r8, 1, targetReg).addReg(Reg);
break;
case cShort:
// Extend value into target register (16->32)
if (isUnsigned)
BuildMI(BB, X86::MOVZXr32r16, 1, targetReg).addReg(Reg);
else
BuildMI(BB, X86::MOVSXr32r16, 1, targetReg).addReg(Reg);
break;
case cInt:
// Move value into target register (32->32)
BuildMI(BB, X86::MOVrr32, 1, targetReg).addReg(Reg);
break;
default:
assert(0 && "Unpromotable operand class in promote32");
}
}
/// 'ret' instruction - Here we are interested in meeting the x86 ABI. As such,
/// we have the following possibilities:
///
/// ret void: No return value, simply emit a 'ret' instruction
/// ret sbyte, ubyte : Extend value into EAX and return
/// ret short, ushort: Extend value into EAX and return
/// ret int, uint : Move value into EAX and return
/// ret pointer : Move value into EAX and return
/// ret long, ulong : Move value into EAX/EDX and return
/// ret float/double : Top of FP stack
///
void ISel::visitReturnInst(ReturnInst &I) {
if (I.getNumOperands() == 0) {
BuildMI(BB, X86::RET, 0); // Just emit a 'ret' instruction
return;
}
Value *RetVal = I.getOperand(0);
unsigned RetReg = getReg(RetVal);
switch (getClassB(RetVal->getType())) {
case cByte: // integral return values: extend or move into EAX and return
case cShort:
case cInt:
promote32(X86::EAX, ValueRecord(RetReg, RetVal->getType()));
// Declare that EAX is live on exit
BuildMI(BB, X86::IMPLICIT_USE, 2).addReg(X86::EAX).addReg(X86::ESP);
break;
case cFP: // Floats & Doubles: Return in ST(0)
BuildMI(BB, X86::FpSETRESULT, 1).addReg(RetReg);
// Declare that top-of-stack is live on exit
BuildMI(BB, X86::IMPLICIT_USE, 2).addReg(X86::ST0).addReg(X86::ESP);
break;
case cLong:
BuildMI(BB, X86::MOVrr32, 1, X86::EAX).addReg(RetReg);
BuildMI(BB, X86::MOVrr32, 1, X86::EDX).addReg(RetReg+1);
// Declare that EAX & EDX are live on exit
BuildMI(BB, X86::IMPLICIT_USE, 3).addReg(X86::EAX).addReg(X86::EDX)
.addReg(X86::ESP);
break;
default:
visitInstruction(I);
}
// Emit a 'ret' instruction
BuildMI(BB, X86::RET, 0);
}
// getBlockAfter - Return the basic block which occurs lexically after the
// specified one.
static inline BasicBlock *getBlockAfter(BasicBlock *BB) {
Function::iterator I = BB; ++I; // Get iterator to next block
return I != BB->getParent()->end() ? &*I : 0;
}
/// visitBranchInst - Handle conditional and unconditional branches here. Note
/// that since code layout is frozen at this point, that if we are trying to
/// jump to a block that is the immediate successor of the current block, we can
/// just make a fall-through (but we don't currently).
///
void ISel::visitBranchInst(BranchInst &BI) {
BasicBlock *NextBB = getBlockAfter(BI.getParent()); // BB after current one
if (!BI.isConditional()) { // Unconditional branch?
if (BI.getSuccessor(0) != NextBB)
BuildMI(BB, X86::JMP, 1).addPCDisp(BI.getSuccessor(0));
return;
}
// See if we can fold the setcc into the branch itself...
SetCondInst *SCI = canFoldSetCCIntoBranch(BI.getCondition());
if (SCI == 0) {
// Nope, cannot fold setcc into this branch. Emit a branch on a condition
// computed some other way...
unsigned condReg = getReg(BI.getCondition());
BuildMI(BB, X86::CMPri8, 2).addReg(condReg).addZImm(0);
if (BI.getSuccessor(1) == NextBB) {
if (BI.getSuccessor(0) != NextBB)
BuildMI(BB, X86::JNE, 1).addPCDisp(BI.getSuccessor(0));
} else {
BuildMI(BB, X86::JE, 1).addPCDisp(BI.getSuccessor(1));
if (BI.getSuccessor(0) != NextBB)
BuildMI(BB, X86::JMP, 1).addPCDisp(BI.getSuccessor(0));
}
return;
}
unsigned OpNum = getSetCCNumber(SCI->getOpcode());
MachineBasicBlock::iterator MII = BB->end();
OpNum = EmitComparison(OpNum, SCI->getOperand(0), SCI->getOperand(1), BB,MII);
const Type *CompTy = SCI->getOperand(0)->getType();
bool isSigned = CompTy->isSigned() && getClassB(CompTy) != cFP;
// LLVM -> X86 signed X86 unsigned
// ----- ---------- ------------
// seteq -> je je
// setne -> jne jne
// setlt -> jl jb
// setge -> jge jae
// setgt -> jg ja
// setle -> jle jbe
// ----
// js // Used by comparison with 0 optimization
// jns
static const unsigned OpcodeTab[2][8] = {
{ X86::JE, X86::JNE, X86::JB, X86::JAE, X86::JA, X86::JBE, 0, 0 },
{ X86::JE, X86::JNE, X86::JL, X86::JGE, X86::JG, X86::JLE,
X86::JS, X86::JNS },
};
if (BI.getSuccessor(0) != NextBB) {
BuildMI(BB, OpcodeTab[isSigned][OpNum], 1).addPCDisp(BI.getSuccessor(0));
if (BI.getSuccessor(1) != NextBB)
BuildMI(BB, X86::JMP, 1).addPCDisp(BI.getSuccessor(1));
} else {
// Change to the inverse condition...
if (BI.getSuccessor(1) != NextBB) {
OpNum ^= 1;
BuildMI(BB, OpcodeTab[isSigned][OpNum], 1).addPCDisp(BI.getSuccessor(1));
}
}
}
/// doCall - This emits an abstract call instruction, setting up the arguments
/// and the return value as appropriate. For the actual function call itself,
/// it inserts the specified CallMI instruction into the stream.
///
void ISel::doCall(const ValueRecord &Ret, MachineInstr *CallMI,
const std::vector<ValueRecord> &Args) {
// Count how many bytes are to be pushed on the stack...
unsigned NumBytes = 0;
if (!Args.empty()) {
for (unsigned i = 0, e = Args.size(); i != e; ++i)
switch (getClassB(Args[i].Ty)) {
case cByte: case cShort: case cInt:
NumBytes += 4; break;
case cLong:
NumBytes += 8; break;
case cFP:
NumBytes += Args[i].Ty == Type::FloatTy ? 4 : 8;
break;
default: assert(0 && "Unknown class!");
}
// Adjust the stack pointer for the new arguments...
BuildMI(BB, X86::ADJCALLSTACKDOWN, 1).addZImm(NumBytes);
// Arguments go on the stack in reverse order, as specified by the ABI.
unsigned ArgOffset = 0;
for (unsigned i = 0, e = Args.size(); i != e; ++i) {
unsigned ArgReg = Args[i].Val ? getReg(Args[i].Val) : Args[i].Reg;
switch (getClassB(Args[i].Ty)) {
case cByte:
case cShort: {
// Promote arg to 32 bits wide into a temporary register...
unsigned R = makeAnotherReg(Type::UIntTy);
promote32(R, Args[i]);
addRegOffset(BuildMI(BB, X86::MOVmr32, 5),
X86::ESP, ArgOffset).addReg(R);
break;
}
case cInt:
addRegOffset(BuildMI(BB, X86::MOVmr32, 5),
X86::ESP, ArgOffset).addReg(ArgReg);
break;
case cLong:
addRegOffset(BuildMI(BB, X86::MOVmr32, 5),
X86::ESP, ArgOffset).addReg(ArgReg);
addRegOffset(BuildMI(BB, X86::MOVmr32, 5),
X86::ESP, ArgOffset+4).addReg(ArgReg+1);
ArgOffset += 4; // 8 byte entry, not 4.
break;
case cFP:
if (Args[i].Ty == Type::FloatTy) {
addRegOffset(BuildMI(BB, X86::FSTr32, 5),
X86::ESP, ArgOffset).addReg(ArgReg);
} else {
assert(Args[i].Ty == Type::DoubleTy && "Unknown FP type!");
addRegOffset(BuildMI(BB, X86::FSTr64, 5),
X86::ESP, ArgOffset).addReg(ArgReg);
ArgOffset += 4; // 8 byte entry, not 4.
}
break;
default: assert(0 && "Unknown class!");
}
ArgOffset += 4;
}
} else {
BuildMI(BB, X86::ADJCALLSTACKDOWN, 1).addZImm(0);
}
BB->push_back(CallMI);
BuildMI(BB, X86::ADJCALLSTACKUP, 1).addZImm(NumBytes);
// If there is a return value, scavenge the result from the location the call
// leaves it in...
//
if (Ret.Ty != Type::VoidTy) {
unsigned DestClass = getClassB(Ret.Ty);
switch (DestClass) {
case cByte:
case cShort:
case cInt: {
// Integral results are in %eax, or the appropriate portion
// thereof.
static const unsigned regRegMove[] = {
X86::MOVrr8, X86::MOVrr16, X86::MOVrr32
};
static const unsigned AReg[] = { X86::AL, X86::AX, X86::EAX };
BuildMI(BB, regRegMove[DestClass], 1, Ret.Reg).addReg(AReg[DestClass]);
break;
}
case cFP: // Floating-point return values live in %ST(0)
BuildMI(BB, X86::FpGETRESULT, 1, Ret.Reg);
break;
case cLong: // Long values are left in EDX:EAX
BuildMI(BB, X86::MOVrr32, 1, Ret.Reg).addReg(X86::EAX);
BuildMI(BB, X86::MOVrr32, 1, Ret.Reg+1).addReg(X86::EDX);
break;
default: assert(0 && "Unknown class!");
}
}
}
/// visitCallInst - Push args on stack and do a procedure call instruction.
void ISel::visitCallInst(CallInst &CI) {
MachineInstr *TheCall;
if (Function *F = CI.getCalledFunction()) {
// Is it an intrinsic function call?
if (Intrinsic::ID ID = (Intrinsic::ID)F->getIntrinsicID()) {
visitIntrinsicCall(ID, CI); // Special intrinsics are not handled here
return;
}
// Emit a CALL instruction with PC-relative displacement.
TheCall = BuildMI(X86::CALLpcrel32, 1).addGlobalAddress(F, true);
} else { // Emit an indirect call...
unsigned Reg = getReg(CI.getCalledValue());
TheCall = BuildMI(X86::CALLr32, 1).addReg(Reg);
}
std::vector<ValueRecord> Args;
for (unsigned i = 1, e = CI.getNumOperands(); i != e; ++i)
Args.push_back(ValueRecord(CI.getOperand(i)));
unsigned DestReg = CI.getType() != Type::VoidTy ? getReg(CI) : 0;
doCall(ValueRecord(DestReg, CI.getType()), TheCall, Args);
}
/// LowerUnknownIntrinsicFunctionCalls - This performs a prepass over the
/// function, lowering any calls to unknown intrinsic functions into the
/// equivalent LLVM code.
void ISel::LowerUnknownIntrinsicFunctionCalls(Function &F) {
for (Function::iterator BB = F.begin(), E = F.end(); BB != E; ++BB)
for (BasicBlock::iterator I = BB->begin(), E = BB->end(); I != E; )
if (CallInst *CI = dyn_cast<CallInst>(I++))
if (Function *F = CI->getCalledFunction())
switch (F->getIntrinsicID()) {
case Intrinsic::not_intrinsic:
case Intrinsic::va_start:
case Intrinsic::va_copy:
case Intrinsic::va_end:
case Intrinsic::returnaddress:
case Intrinsic::frameaddress:
case Intrinsic::memcpy:
case Intrinsic::memset:
// We directly implement these intrinsics
break;
default:
// All other intrinsic calls we must lower.
Instruction *Before = CI->getPrev();
TM.getIntrinsicLowering().LowerIntrinsicCall(CI);
if (Before) { // Move iterator to instruction after call
I = Before; ++I;
} else {
I = BB->begin();
}
}
}
void ISel::visitIntrinsicCall(Intrinsic::ID ID, CallInst &CI) {
unsigned TmpReg1, TmpReg2;
switch (ID) {
case Intrinsic::va_start:
// Get the address of the first vararg value...
TmpReg1 = getReg(CI);
addFrameReference(BuildMI(BB, X86::LEAr32, 5, TmpReg1), VarArgsFrameIndex);
return;
case Intrinsic::va_copy:
TmpReg1 = getReg(CI);
TmpReg2 = getReg(CI.getOperand(1));
BuildMI(BB, X86::MOVrr32, 1, TmpReg1).addReg(TmpReg2);
return;
case Intrinsic::va_end: return; // Noop on X86
case Intrinsic::returnaddress:
case Intrinsic::frameaddress:
TmpReg1 = getReg(CI);
if (cast<Constant>(CI.getOperand(1))->isNullValue()) {
if (ID == Intrinsic::returnaddress) {
// Just load the return address
addFrameReference(BuildMI(BB, X86::MOVrm32, 4, TmpReg1),
ReturnAddressIndex);
} else {
addFrameReference(BuildMI(BB, X86::LEAr32, 4, TmpReg1),
ReturnAddressIndex, -4);
}
} else {
// Values other than zero are not implemented yet.
BuildMI(BB, X86::MOVri32, 1, TmpReg1).addZImm(0);
}
return;
case Intrinsic::memcpy: {
assert(CI.getNumOperands() == 5 && "Illegal llvm.memcpy call!");
unsigned Align = 1;
if (ConstantInt *AlignC = dyn_cast<ConstantInt>(CI.getOperand(4))) {
Align = AlignC->getRawValue();
if (Align == 0) Align = 1;
}
// Turn the byte code into # iterations
unsigned ByteReg;
unsigned CountReg;
unsigned Opcode;
switch (Align & 3) {
case 2: // WORD aligned
if (ConstantInt *I = dyn_cast<ConstantInt>(CI.getOperand(3))) {
CountReg = getReg(ConstantUInt::get(Type::UIntTy, I->getRawValue()/2));
} else {
CountReg = makeAnotherReg(Type::IntTy);
BuildMI(BB, X86::SHRri32, 2, CountReg).addReg(ByteReg).addZImm(1);
}
Opcode = X86::REP_MOVSW;
break;
case 0: // DWORD aligned
if (ConstantInt *I = dyn_cast<ConstantInt>(CI.getOperand(3))) {
CountReg = getReg(ConstantUInt::get(Type::UIntTy, I->getRawValue()/4));
} else {
CountReg = makeAnotherReg(Type::IntTy);
BuildMI(BB, X86::SHRri32, 2, CountReg).addReg(ByteReg).addZImm(2);
}
Opcode = X86::REP_MOVSD;
break;
case 1: // BYTE aligned
case 3: // BYTE aligned
CountReg = getReg(CI.getOperand(3));
Opcode = X86::REP_MOVSB;
break;
}
// No matter what the alignment is, we put the source in ESI, the
// destination in EDI, and the count in ECX.
TmpReg1 = getReg(CI.getOperand(1));
TmpReg2 = getReg(CI.getOperand(2));
BuildMI(BB, X86::MOVrr32, 1, X86::ECX).addReg(CountReg);
BuildMI(BB, X86::MOVrr32, 1, X86::EDI).addReg(TmpReg1);
BuildMI(BB, X86::MOVrr32, 1, X86::ESI).addReg(TmpReg2);
BuildMI(BB, Opcode, 0);
return;
}
case Intrinsic::memset: {
assert(CI.getNumOperands() == 5 && "Illegal llvm.memset call!");
unsigned Align = 1;
if (ConstantInt *AlignC = dyn_cast<ConstantInt>(CI.getOperand(4))) {
Align = AlignC->getRawValue();
if (Align == 0) Align = 1;
}
// Turn the byte code into # iterations
unsigned ByteReg;
unsigned CountReg;
unsigned Opcode;
if (ConstantInt *ValC = dyn_cast<ConstantInt>(CI.getOperand(2))) {
unsigned Val = ValC->getRawValue() & 255;
// If the value is a constant, then we can potentially use larger copies.
switch (Align & 3) {
case 2: // WORD aligned
if (ConstantInt *I = dyn_cast<ConstantInt>(CI.getOperand(3))) {
CountReg =getReg(ConstantUInt::get(Type::UIntTy, I->getRawValue()/2));
} else {
CountReg = makeAnotherReg(Type::IntTy);
BuildMI(BB, X86::SHRri32, 2, CountReg).addReg(ByteReg).addZImm(1);
}
BuildMI(BB, X86::MOVri16, 1, X86::AX).addZImm((Val << 8) | Val);
Opcode = X86::REP_STOSW;
break;
case 0: // DWORD aligned
if (ConstantInt *I = dyn_cast<ConstantInt>(CI.getOperand(3))) {
CountReg =getReg(ConstantUInt::get(Type::UIntTy, I->getRawValue()/4));
} else {
CountReg = makeAnotherReg(Type::IntTy);
BuildMI(BB, X86::SHRri32, 2, CountReg).addReg(ByteReg).addZImm(2);
}
Val = (Val << 8) | Val;
BuildMI(BB, X86::MOVri32, 1, X86::EAX).addZImm((Val << 16) | Val);
Opcode = X86::REP_STOSD;
break;
case 1: // BYTE aligned
case 3: // BYTE aligned
CountReg = getReg(CI.getOperand(3));
BuildMI(BB, X86::MOVri8, 1, X86::AL).addZImm(Val);
Opcode = X86::REP_STOSB;
break;
}
} else {
// If it's not a constant value we are storing, just fall back. We could
// try to be clever to form 16 bit and 32 bit values, but we don't yet.
unsigned ValReg = getReg(CI.getOperand(2));
BuildMI(BB, X86::MOVrr8, 1, X86::AL).addReg(ValReg);
CountReg = getReg(CI.getOperand(3));
Opcode = X86::REP_STOSB;
}
// No matter what the alignment is, we put the source in ESI, the
// destination in EDI, and the count in ECX.
TmpReg1 = getReg(CI.getOperand(1));
//TmpReg2 = getReg(CI.getOperand(2));
BuildMI(BB, X86::MOVrr32, 1, X86::ECX).addReg(CountReg);
BuildMI(BB, X86::MOVrr32, 1, X86::EDI).addReg(TmpReg1);
BuildMI(BB, Opcode, 0);
return;
}
default: assert(0 && "Error: unknown intrinsics should have been lowered!");
}
}
/// visitSimpleBinary - Implement simple binary operators for integral types...
/// OperatorClass is one of: 0 for Add, 1 for Sub, 2 for And, 3 for Or, 4 for
/// Xor.
void ISel::visitSimpleBinary(BinaryOperator &B, unsigned OperatorClass) {
unsigned DestReg = getReg(B);
MachineBasicBlock::iterator MI = BB->end();
emitSimpleBinaryOperation(BB, MI, B.getOperand(0), B.getOperand(1),
OperatorClass, DestReg);
}
/// emitSimpleBinaryOperation - Implement simple binary operators for integral
/// types... OperatorClass is one of: 0 for Add, 1 for Sub, 2 for And, 3 for
/// Or, 4 for Xor.
///
/// emitSimpleBinaryOperation - Common code shared between visitSimpleBinary
/// and constant expression support.
///
void ISel::emitSimpleBinaryOperation(MachineBasicBlock *MBB,
MachineBasicBlock::iterator IP,
Value *Op0, Value *Op1,
unsigned OperatorClass, unsigned DestReg) {
unsigned Class = getClassB(Op0->getType());
// sub 0, X -> neg X
if (OperatorClass == 1 && Class != cLong)
if (ConstantInt *CI = dyn_cast<ConstantInt>(Op0)) {
if (CI->isNullValue()) {
unsigned op1Reg = getReg(Op1, MBB, IP);
switch (Class) {
default: assert(0 && "Unknown class for this function!");
case cByte:
BMI(MBB, IP, X86::NEGr8, 1, DestReg).addReg(op1Reg);
return;
case cShort:
BMI(MBB, IP, X86::NEGr16, 1, DestReg).addReg(op1Reg);
return;
case cInt:
BMI(MBB, IP, X86::NEGr32, 1, DestReg).addReg(op1Reg);
return;
}
}
} else if (ConstantFP *CFP = dyn_cast<ConstantFP>(Op0))
if (CFP->isExactlyValue(-0.0)) {
// -0.0 - X === -X
unsigned op1Reg = getReg(Op1, MBB, IP);
BMI(MBB, IP, X86::FCHS, 1, DestReg).addReg(op1Reg);
return;
}
if (!isa<ConstantInt>(Op1) || Class == cLong) {
static const unsigned OpcodeTab[][4] = {
// Arithmetic operators
{ X86::ADDrr8, X86::ADDrr16, X86::ADDrr32, X86::FpADD }, // ADD
{ X86::SUBrr8, X86::SUBrr16, X86::SUBrr32, X86::FpSUB }, // SUB
// Bitwise operators
{ X86::ANDrr8, X86::ANDrr16, X86::ANDrr32, 0 }, // AND
{ X86:: ORrr8, X86:: ORrr16, X86:: ORrr32, 0 }, // OR
{ X86::XORrr8, X86::XORrr16, X86::XORrr32, 0 }, // XOR
};
bool isLong = false;
if (Class == cLong) {
isLong = true;
Class = cInt; // Bottom 32 bits are handled just like ints
}
unsigned Opcode = OpcodeTab[OperatorClass][Class];
assert(Opcode && "Floating point arguments to logical inst?");
unsigned Op0r = getReg(Op0, MBB, IP);
unsigned Op1r = getReg(Op1, MBB, IP);
BMI(MBB, IP, Opcode, 2, DestReg).addReg(Op0r).addReg(Op1r);
if (isLong) { // Handle the upper 32 bits of long values...
static const unsigned TopTab[] = {
X86::ADCrr32, X86::SBBrr32, X86::ANDrr32, X86::ORrr32, X86::XORrr32
};
BMI(MBB, IP, TopTab[OperatorClass], 2,
DestReg+1).addReg(Op0r+1).addReg(Op1r+1);
}
return;
}
// Special case: op Reg, <const>
ConstantInt *Op1C = cast<ConstantInt>(Op1);
unsigned Op0r = getReg(Op0, MBB, IP);
// xor X, -1 -> not X
if (OperatorClass == 4 && Op1C->isAllOnesValue()) {
static unsigned const NOTTab[] = { X86::NOTr8, X86::NOTr16, X86::NOTr32 };
BMI(MBB, IP, NOTTab[Class], 1, DestReg).addReg(Op0r);
return;
}
// add X, -1 -> dec X
if (OperatorClass == 0 && Op1C->isAllOnesValue()) {
static unsigned const DECTab[] = { X86::DECr8, X86::DECr16, X86::DECr32 };
BMI(MBB, IP, DECTab[Class], 1, DestReg).addReg(Op0r);
return;
}
// add X, 1 -> inc X
if (OperatorClass == 0 && Op1C->equalsInt(1)) {
static unsigned const DECTab[] = { X86::INCr8, X86::INCr16, X86::INCr32 };
BMI(MBB, IP, DECTab[Class], 1, DestReg).addReg(Op0r);
return;
}
static const unsigned OpcodeTab[][3] = {
// Arithmetic operators
{ X86::ADDri8, X86::ADDri16, X86::ADDri32 }, // ADD
{ X86::SUBri8, X86::SUBri16, X86::SUBri32 }, // SUB
// Bitwise operators
{ X86::ANDri8, X86::ANDri16, X86::ANDri32 }, // AND
{ X86:: ORri8, X86:: ORri16, X86:: ORri32 }, // OR
{ X86::XORri8, X86::XORri16, X86::XORri32 }, // XOR
};
assert(Class < 3 && "General code handles 64-bit integer types!");
unsigned Opcode = OpcodeTab[OperatorClass][Class];
uint64_t Op1v = cast<ConstantInt>(Op1C)->getRawValue();
// Mask off any upper bits of the constant, if there are any...
Op1v &= (1ULL << (8 << Class)) - 1;
BMI(MBB, IP, Opcode, 2, DestReg).addReg(Op0r).addZImm(Op1v);
}
/// doMultiply - Emit appropriate instructions to multiply together the
/// registers op0Reg and op1Reg, and put the result in DestReg. The type of the
/// result should be given as DestTy.
///
void ISel::doMultiply(MachineBasicBlock *MBB, MachineBasicBlock::iterator MBBI,
unsigned DestReg, const Type *DestTy,
unsigned op0Reg, unsigned op1Reg) {
unsigned Class = getClass(DestTy);
switch (Class) {
case cFP: // Floating point multiply
BMI(BB, MBBI, X86::FpMUL, 2, DestReg).addReg(op0Reg).addReg(op1Reg);
return;
case cInt:
case cShort:
BMI(BB, MBBI, Class == cInt ? X86::IMULrr32 : X86::IMULrr16, 2, DestReg)
.addReg(op0Reg).addReg(op1Reg);
return;
case cByte:
// Must use the MUL instruction, which forces use of AL...
BMI(MBB, MBBI, X86::MOVrr8, 1, X86::AL).addReg(op0Reg);
BMI(MBB, MBBI, X86::MULr8, 1).addReg(op1Reg);
BMI(MBB, MBBI, X86::MOVrr8, 1, DestReg).addReg(X86::AL);
return;
default:
case cLong: assert(0 && "doMultiply cannot operate on LONG values!");
}
}
// ExactLog2 - This function solves for (Val == 1 << (N-1)) and returns N. It
// returns zero when the input is not exactly a power of two.
static unsigned ExactLog2(unsigned Val) {
if (Val == 0) return 0;
unsigned Count = 0;
while (Val != 1) {
if (Val & 1) return 0;
Val >>= 1;
++Count;
}
return Count+1;
}
void ISel::doMultiplyConst(MachineBasicBlock *MBB,
MachineBasicBlock::iterator IP,
unsigned DestReg, const Type *DestTy,
unsigned op0Reg, unsigned ConstRHS) {
unsigned Class = getClass(DestTy);
// If the element size is exactly a power of 2, use a shift to get it.
if (unsigned Shift = ExactLog2(ConstRHS)) {
switch (Class) {
default: assert(0 && "Unknown class for this function!");
case cByte:
BMI(MBB, IP, X86::SHLri32, 2, DestReg).addReg(op0Reg).addZImm(Shift-1);
return;
case cShort:
BMI(MBB, IP, X86::SHLri32, 2, DestReg).addReg(op0Reg).addZImm(Shift-1);
return;
case cInt:
BMI(MBB, IP, X86::SHLri32, 2, DestReg).addReg(op0Reg).addZImm(Shift-1);
return;
}
}
if (Class == cShort) {
BMI(MBB, IP, X86::IMULrri16, 2, DestReg).addReg(op0Reg).addZImm(ConstRHS);
return;
} else if (Class == cInt) {
BMI(MBB, IP, X86::IMULrri32, 2, DestReg).addReg(op0Reg).addZImm(ConstRHS);
return;
}
// Most general case, emit a normal multiply...
static const unsigned MOVriTab[] = {
X86::MOVri8, X86::MOVri16, X86::MOVri32
};
unsigned TmpReg = makeAnotherReg(DestTy);
BMI(MBB, IP, MOVriTab[Class], 1, TmpReg).addZImm(ConstRHS);
// Emit a MUL to multiply the register holding the index by
// elementSize, putting the result in OffsetReg.
doMultiply(MBB, IP, DestReg, DestTy, op0Reg, TmpReg);
}
/// visitMul - Multiplies are not simple binary operators because they must deal
/// with the EAX register explicitly.
///
void ISel::visitMul(BinaryOperator &I) {
unsigned Op0Reg = getReg(I.getOperand(0));
unsigned DestReg = getReg(I);
// Simple scalar multiply?
if (I.getType() != Type::LongTy && I.getType() != Type::ULongTy) {
if (ConstantInt *CI = dyn_cast<ConstantInt>(I.getOperand(1))) {
unsigned Val = (unsigned)CI->getRawValue(); // Cannot be 64-bit constant
MachineBasicBlock::iterator MBBI = BB->end();
doMultiplyConst(BB, MBBI, DestReg, I.getType(), Op0Reg, Val);
} else {
unsigned Op1Reg = getReg(I.getOperand(1));
MachineBasicBlock::iterator MBBI = BB->end();
doMultiply(BB, MBBI, DestReg, I.getType(), Op0Reg, Op1Reg);
}
} else {
unsigned Op1Reg = getReg(I.getOperand(1));
// Long value. We have to do things the hard way...
// Multiply the two low parts... capturing carry into EDX
BuildMI(BB, X86::MOVrr32, 1, X86::EAX).addReg(Op0Reg);
BuildMI(BB, X86::MULr32, 1).addReg(Op1Reg); // AL*BL
unsigned OverflowReg = makeAnotherReg(Type::UIntTy);
BuildMI(BB, X86::MOVrr32, 1, DestReg).addReg(X86::EAX); // AL*BL
BuildMI(BB, X86::MOVrr32, 1, OverflowReg).addReg(X86::EDX); // AL*BL >> 32
MachineBasicBlock::iterator MBBI = BB->end();
unsigned AHBLReg = makeAnotherReg(Type::UIntTy); // AH*BL
BMI(BB, MBBI, X86::IMULrr32, 2, AHBLReg).addReg(Op0Reg+1).addReg(Op1Reg);
unsigned AHBLplusOverflowReg = makeAnotherReg(Type::UIntTy);
BuildMI(BB, X86::ADDrr32, 2, // AH*BL+(AL*BL >> 32)
AHBLplusOverflowReg).addReg(AHBLReg).addReg(OverflowReg);
MBBI = BB->end();
unsigned ALBHReg = makeAnotherReg(Type::UIntTy); // AL*BH
BMI(BB, MBBI, X86::IMULrr32, 2, ALBHReg).addReg(Op0Reg).addReg(Op1Reg+1);
BuildMI(BB, X86::ADDrr32, 2, // AL*BH + AH*BL + (AL*BL >> 32)
DestReg+1).addReg(AHBLplusOverflowReg).addReg(ALBHReg);
}
}
/// visitDivRem - Handle division and remainder instructions... these
/// instruction both require the same instructions to be generated, they just
/// select the result from a different register. Note that both of these
/// instructions work differently for signed and unsigned operands.
///
void ISel::visitDivRem(BinaryOperator &I) {
unsigned Op0Reg = getReg(I.getOperand(0));
unsigned Op1Reg = getReg(I.getOperand(1));
unsigned ResultReg = getReg(I);
MachineBasicBlock::iterator IP = BB->end();
emitDivRemOperation(BB, IP, Op0Reg, Op1Reg, I.getOpcode() == Instruction::Div,
I.getType(), ResultReg);
}
void ISel::emitDivRemOperation(MachineBasicBlock *BB,
MachineBasicBlock::iterator IP,
unsigned Op0Reg, unsigned Op1Reg, bool isDiv,
const Type *Ty, unsigned ResultReg) {
unsigned Class = getClass(Ty);
switch (Class) {
case cFP: // Floating point divide
if (isDiv) {
BMI(BB, IP, X86::FpDIV, 2, ResultReg).addReg(Op0Reg).addReg(Op1Reg);
} else { // Floating point remainder...
MachineInstr *TheCall =
BuildMI(X86::CALLpcrel32, 1).addExternalSymbol("fmod", true);
std::vector<ValueRecord> Args;
Args.push_back(ValueRecord(Op0Reg, Type::DoubleTy));
Args.push_back(ValueRecord(Op1Reg, Type::DoubleTy));
doCall(ValueRecord(ResultReg, Type::DoubleTy), TheCall, Args);
}
return;
case cLong: {
static const char *FnName[] =
{ "__moddi3", "__divdi3", "__umoddi3", "__udivdi3" };
unsigned NameIdx = Ty->isUnsigned()*2 + isDiv;
MachineInstr *TheCall =
BuildMI(X86::CALLpcrel32, 1).addExternalSymbol(FnName[NameIdx], true);
std::vector<ValueRecord> Args;
Args.push_back(ValueRecord(Op0Reg, Type::LongTy));
Args.push_back(ValueRecord(Op1Reg, Type::LongTy));
doCall(ValueRecord(ResultReg, Type::LongTy), TheCall, Args);
return;
}
case cByte: case cShort: case cInt:
break; // Small integrals, handled below...
default: assert(0 && "Unknown class!");
}
static const unsigned Regs[] ={ X86::AL , X86::AX , X86::EAX };
static const unsigned MovOpcode[]={ X86::MOVrr8, X86::MOVrr16, X86::MOVrr32 };
static const unsigned SarOpcode[]={ X86::SARri8, X86::SARri16, X86::SARri32 };
static const unsigned ClrOpcode[]={ X86::MOVri8, X86::MOVri16, X86::MOVri32 };
static const unsigned ExtRegs[] ={ X86::AH , X86::DX , X86::EDX };
static const unsigned DivOpcode[][4] = {
{ X86::DIVr8 , X86::DIVr16 , X86::DIVr32 , 0 }, // Unsigned division
{ X86::IDIVr8, X86::IDIVr16, X86::IDIVr32, 0 }, // Signed division
};
bool isSigned = Ty->isSigned();
unsigned Reg = Regs[Class];
unsigned ExtReg = ExtRegs[Class];
// Put the first operand into one of the A registers...
BMI(BB, IP, MovOpcode[Class], 1, Reg).addReg(Op0Reg);
if (isSigned) {
// Emit a sign extension instruction...
unsigned ShiftResult = makeAnotherReg(Ty);
BMI(BB, IP, SarOpcode[Class], 2, ShiftResult).addReg(Op0Reg).addZImm(31);
BMI(BB, IP, MovOpcode[Class], 1, ExtReg).addReg(ShiftResult);
} else {
// If unsigned, emit a zeroing instruction... (reg = 0)
BMI(BB, IP, ClrOpcode[Class], 2, ExtReg).addZImm(0);
}
// Emit the appropriate divide or remainder instruction...
BMI(BB, IP, DivOpcode[isSigned][Class], 1).addReg(Op1Reg);
// Figure out which register we want to pick the result out of...
unsigned DestReg = isDiv ? Reg : ExtReg;
// Put the result into the destination register...
BMI(BB, IP, MovOpcode[Class], 1, ResultReg).addReg(DestReg);
}
/// Shift instructions: 'shl', 'sar', 'shr' - Some special cases here
/// for constant immediate shift values, and for constant immediate
/// shift values equal to 1. Even the general case is sort of special,
/// because the shift amount has to be in CL, not just any old register.
///
void ISel::visitShiftInst(ShiftInst &I) {
MachineBasicBlock::iterator IP = BB->end ();
emitShiftOperation (BB, IP, I.getOperand (0), I.getOperand (1),
I.getOpcode () == Instruction::Shl, I.getType (),
getReg (I));
}
/// emitShiftOperation - Common code shared between visitShiftInst and
/// constant expression support.
void ISel::emitShiftOperation(MachineBasicBlock *MBB,
MachineBasicBlock::iterator IP,
Value *Op, Value *ShiftAmount, bool isLeftShift,
const Type *ResultTy, unsigned DestReg) {
unsigned SrcReg = getReg (Op, MBB, IP);
bool isSigned = ResultTy->isSigned ();
unsigned Class = getClass (ResultTy);
static const unsigned ConstantOperand[][4] = {
{ X86::SHRri8, X86::SHRri16, X86::SHRri32, X86::SHRDri32 }, // SHR
{ X86::SARri8, X86::SARri16, X86::SARri32, X86::SHRDri32 }, // SAR
{ X86::SHLri8, X86::SHLri16, X86::SHLri32, X86::SHLDri32 }, // SHL
{ X86::SHLri8, X86::SHLri16, X86::SHLri32, X86::SHLDri32 }, // SAL = SHL
};
static const unsigned NonConstantOperand[][4] = {
{ X86::SHRrr8, X86::SHRrr16, X86::SHRrr32 }, // SHR
{ X86::SARrr8, X86::SARrr16, X86::SARrr32 }, // SAR
{ X86::SHLrr8, X86::SHLrr16, X86::SHLrr32 }, // SHL
{ X86::SHLrr8, X86::SHLrr16, X86::SHLrr32 }, // SAL = SHL
};
// Longs, as usual, are handled specially...
if (Class == cLong) {
// If we have a constant shift, we can generate much more efficient code
// than otherwise...
//
if (ConstantUInt *CUI = dyn_cast<ConstantUInt>(ShiftAmount)) {
unsigned Amount = CUI->getValue();
if (Amount < 32) {
const unsigned *Opc = ConstantOperand[isLeftShift*2+isSigned];
if (isLeftShift) {
BMI(MBB, IP, Opc[3], 3,
DestReg+1).addReg(SrcReg+1).addReg(SrcReg).addZImm(Amount);
BMI(MBB, IP, Opc[2], 2, DestReg).addReg(SrcReg).addZImm(Amount);
} else {
BMI(MBB, IP, Opc[3], 3,
DestReg).addReg(SrcReg ).addReg(SrcReg+1).addZImm(Amount);
BMI(MBB, IP, Opc[2], 2, DestReg+1).addReg(SrcReg+1).addZImm(Amount);
}
} else { // Shifting more than 32 bits
Amount -= 32;
if (isLeftShift) {
BMI(MBB, IP, X86::SHLri32, 2,
DestReg + 1).addReg(SrcReg).addZImm(Amount);
BMI(MBB, IP, X86::MOVri32, 1,
DestReg).addZImm(0);
} else {
unsigned Opcode = isSigned ? X86::SARri32 : X86::SHRri32;
BMI(MBB, IP, Opcode, 2, DestReg).addReg(SrcReg+1).addZImm(Amount);
BMI(MBB, IP, X86::MOVri32, 1, DestReg+1).addZImm(0);
}
}
} else {
unsigned TmpReg = makeAnotherReg(Type::IntTy);
if (!isLeftShift && isSigned) {
// If this is a SHR of a Long, then we need to do funny sign extension
// stuff. TmpReg gets the value to use as the high-part if we are
// shifting more than 32 bits.
BMI(MBB, IP, X86::SARri32, 2, TmpReg).addReg(SrcReg).addZImm(31);
} else {
// Other shifts use a fixed zero value if the shift is more than 32
// bits.
BMI(MBB, IP, X86::MOVri32, 1, TmpReg).addZImm(0);
}
// Initialize CL with the shift amount...
unsigned ShiftAmountReg = getReg(ShiftAmount, MBB, IP);
BMI(MBB, IP, X86::MOVrr8, 1, X86::CL).addReg(ShiftAmountReg);
unsigned TmpReg2 = makeAnotherReg(Type::IntTy);
unsigned TmpReg3 = makeAnotherReg(Type::IntTy);
if (isLeftShift) {
// TmpReg2 = shld inHi, inLo
BMI(MBB, IP, X86::SHLDrr32, 2, TmpReg2).addReg(SrcReg+1).addReg(SrcReg);
// TmpReg3 = shl inLo, CL
BMI(MBB, IP, X86::SHLrr32, 1, TmpReg3).addReg(SrcReg);
// Set the flags to indicate whether the shift was by more than 32 bits.
BMI(MBB, IP, X86::TESTri8, 2).addReg(X86::CL).addZImm(32);
// DestHi = (>32) ? TmpReg3 : TmpReg2;
BMI(MBB, IP, X86::CMOVNErr32, 2,
DestReg+1).addReg(TmpReg2).addReg(TmpReg3);
// DestLo = (>32) ? TmpReg : TmpReg3;
BMI(MBB, IP, X86::CMOVNErr32, 2,
DestReg).addReg(TmpReg3).addReg(TmpReg);
} else {
// TmpReg2 = shrd inLo, inHi
BMI(MBB, IP, X86::SHRDrr32, 2, TmpReg2).addReg(SrcReg).addReg(SrcReg+1);
// TmpReg3 = s[ah]r inHi, CL
BMI(MBB, IP, isSigned ? X86::SARrr32 : X86::SHRrr32, 1, TmpReg3)
.addReg(SrcReg+1);
// Set the flags to indicate whether the shift was by more than 32 bits.
BMI(MBB, IP, X86::TESTri8, 2).addReg(X86::CL).addZImm(32);
// DestLo = (>32) ? TmpReg3 : TmpReg2;
BMI(MBB, IP, X86::CMOVNErr32, 2,
DestReg).addReg(TmpReg2).addReg(TmpReg3);
// DestHi = (>32) ? TmpReg : TmpReg3;
BMI(MBB, IP, X86::CMOVNErr32, 2,
DestReg+1).addReg(TmpReg3).addReg(TmpReg);
}
}
return;
}
if (ConstantUInt *CUI = dyn_cast<ConstantUInt>(ShiftAmount)) {
// The shift amount is constant, guaranteed to be a ubyte. Get its value.
assert(CUI->getType() == Type::UByteTy && "Shift amount not a ubyte?");
const unsigned *Opc = ConstantOperand[isLeftShift*2+isSigned];
BMI(MBB, IP, Opc[Class], 2,
DestReg).addReg(SrcReg).addZImm(CUI->getValue());
} else { // The shift amount is non-constant.
unsigned ShiftAmountReg = getReg (ShiftAmount, MBB, IP);
BMI(MBB, IP, X86::MOVrr8, 1, X86::CL).addReg(ShiftAmountReg);
const unsigned *Opc = NonConstantOperand[isLeftShift*2+isSigned];
BMI(MBB, IP, Opc[Class], 1, DestReg).addReg(SrcReg);
}
}
/// visitLoadInst - Implement LLVM load instructions in terms of the x86 'mov'
/// instruction. The load and store instructions are the only place where we
/// need to worry about the memory layout of the target machine.
///
void ISel::visitLoadInst(LoadInst &I) {
unsigned DestReg = getReg(I);
unsigned BaseReg = 0, Scale = 1, IndexReg = 0, Disp = 0;
Value *Addr = I.getOperand(0);
if (GetElementPtrInst *GEP = dyn_cast<GetElementPtrInst>(Addr)) {
if (isGEPFoldable(BB, GEP->getOperand(0), GEP->op_begin()+1, GEP->op_end(),
BaseReg, Scale, IndexReg, Disp))
Addr = 0; // Address is consumed!
} else if (ConstantExpr *CE = dyn_cast<ConstantExpr>(Addr)) {
if (CE->getOpcode() == Instruction::GetElementPtr)
if (isGEPFoldable(BB, CE->getOperand(0), CE->op_begin()+1, CE->op_end(),
BaseReg, Scale, IndexReg, Disp))
Addr = 0;
}
if (Addr) {
// If it's not foldable, reset addr mode.
BaseReg = getReg(Addr);
Scale = 1; IndexReg = 0; Disp = 0;
}
unsigned Class = getClassB(I.getType());
if (Class == cLong) {
addFullAddress(BuildMI(BB, X86::MOVrm32, 4, DestReg),
BaseReg, Scale, IndexReg, Disp);
addFullAddress(BuildMI(BB, X86::MOVrm32, 4, DestReg+1),
BaseReg, Scale, IndexReg, Disp+4);
return;
}
static const unsigned Opcodes[] = {
X86::MOVrm8, X86::MOVrm16, X86::MOVrm32, X86::FLDr32
};
unsigned Opcode = Opcodes[Class];
if (I.getType() == Type::DoubleTy) Opcode = X86::FLDr64;
addFullAddress(BuildMI(BB, Opcode, 4, DestReg),
BaseReg, Scale, IndexReg, Disp);
}
/// visitStoreInst - Implement LLVM store instructions in terms of the x86 'mov'
/// instruction.
///
void ISel::visitStoreInst(StoreInst &I) {
unsigned BaseReg = 0, Scale = 1, IndexReg = 0, Disp = 0;
Value *Addr = I.getOperand(1);
if (GetElementPtrInst *GEP = dyn_cast<GetElementPtrInst>(Addr)) {
if (isGEPFoldable(BB, GEP->getOperand(0), GEP->op_begin()+1, GEP->op_end(),
BaseReg, Scale, IndexReg, Disp))
Addr = 0; // Address is consumed!
} else if (ConstantExpr *CE = dyn_cast<ConstantExpr>(Addr)) {
if (CE->getOpcode() == Instruction::GetElementPtr)
if (isGEPFoldable(BB, CE->getOperand(0), CE->op_begin()+1, CE->op_end(),
BaseReg, Scale, IndexReg, Disp))
Addr = 0;
}
if (Addr) {
// If it's not foldable, reset addr mode.
BaseReg = getReg(Addr);
Scale = 1; IndexReg = 0; Disp = 0;
}
const Type *ValTy = I.getOperand(0)->getType();
unsigned Class = getClassB(ValTy);
if (ConstantInt *CI = dyn_cast<ConstantInt>(I.getOperand(0))) {
uint64_t Val = CI->getRawValue();
if (Class == cLong) {
addFullAddress(BuildMI(BB, X86::MOVmi32, 5),
BaseReg, Scale, IndexReg, Disp).addZImm(Val & ~0U);
addFullAddress(BuildMI(BB, X86::MOVmi32, 5),
BaseReg, Scale, IndexReg, Disp+4).addZImm(Val>>32);
} else {
static const unsigned Opcodes[] = {
X86::MOVmi8, X86::MOVmi16, X86::MOVmi32
};
unsigned Opcode = Opcodes[Class];
addFullAddress(BuildMI(BB, Opcode, 5),
BaseReg, Scale, IndexReg, Disp).addZImm(Val);
}
} else if (ConstantBool *CB = dyn_cast<ConstantBool>(I.getOperand(0))) {
addFullAddress(BuildMI(BB, X86::MOVmi8, 5),
BaseReg, Scale, IndexReg, Disp).addZImm(CB->getValue());
} else {
if (Class == cLong) {
unsigned ValReg = getReg(I.getOperand(0));
addFullAddress(BuildMI(BB, X86::MOVmr32, 5),
BaseReg, Scale, IndexReg, Disp).addReg(ValReg);
addFullAddress(BuildMI(BB, X86::MOVmr32, 5),
BaseReg, Scale, IndexReg, Disp+4).addReg(ValReg+1);
} else {
unsigned ValReg = getReg(I.getOperand(0));
static const unsigned Opcodes[] = {
X86::MOVmr8, X86::MOVmr16, X86::MOVmr32, X86::FSTr32
};
unsigned Opcode = Opcodes[Class];
if (ValTy == Type::DoubleTy) Opcode = X86::FSTr64;
addFullAddress(BuildMI(BB, Opcode, 1+4),
BaseReg, Scale, IndexReg, Disp).addReg(ValReg);
}
}
}
/// visitCastInst - Here we have various kinds of copying with or without
/// sign extension going on.
void ISel::visitCastInst(CastInst &CI) {
Value *Op = CI.getOperand(0);
// If this is a cast from a 32-bit integer to a Long type, and the only uses
// of the case are GEP instructions, then the cast does not need to be
// generated explicitly, it will be folded into the GEP.
if (CI.getType() == Type::LongTy &&
(Op->getType() == Type::IntTy || Op->getType() == Type::UIntTy)) {
bool AllUsesAreGEPs = true;
for (Value::use_iterator I = CI.use_begin(), E = CI.use_end(); I != E; ++I)
if (!isa<GetElementPtrInst>(*I)) {
AllUsesAreGEPs = false;
break;
}
// No need to codegen this cast if all users are getelementptr instrs...
if (AllUsesAreGEPs) return;
}
unsigned DestReg = getReg(CI);
MachineBasicBlock::iterator MI = BB->end();
emitCastOperation(BB, MI, Op, CI.getType(), DestReg);
}
/// emitCastOperation - Common code shared between visitCastInst and
/// constant expression cast support.
void ISel::emitCastOperation(MachineBasicBlock *BB,
MachineBasicBlock::iterator IP,
Value *Src, const Type *DestTy,
unsigned DestReg) {
unsigned SrcReg = getReg(Src, BB, IP);
const Type *SrcTy = Src->getType();
unsigned SrcClass = getClassB(SrcTy);
unsigned DestClass = getClassB(DestTy);
// Implement casts to bool by using compare on the operand followed by set if
// not zero on the result.
if (DestTy == Type::BoolTy) {
switch (SrcClass) {
case cByte:
BMI(BB, IP, X86::TESTrr8, 2).addReg(SrcReg).addReg(SrcReg);
break;
case cShort:
BMI(BB, IP, X86::TESTrr16, 2).addReg(SrcReg).addReg(SrcReg);
break;
case cInt:
BMI(BB, IP, X86::TESTrr32, 2).addReg(SrcReg).addReg(SrcReg);
break;
case cLong: {
unsigned TmpReg = makeAnotherReg(Type::IntTy);
BMI(BB, IP, X86::ORrr32, 2, TmpReg).addReg(SrcReg).addReg(SrcReg+1);
break;
}
case cFP:
BMI(BB, IP, X86::FTST, 1).addReg(SrcReg);
BMI(BB, IP, X86::FNSTSWr8, 0);
BMI(BB, IP, X86::SAHF, 1);
break;
}
// If the zero flag is not set, then the value is true, set the byte to
// true.
BMI(BB, IP, X86::SETNEr, 1, DestReg);
return;
}
static const unsigned RegRegMove[] = {
X86::MOVrr8, X86::MOVrr16, X86::MOVrr32, X86::FpMOV, X86::MOVrr32
};
// Implement casts between values of the same type class (as determined by
// getClass) by using a register-to-register move.
if (SrcClass == DestClass) {
if (SrcClass <= cInt || (SrcClass == cFP && SrcTy == DestTy)) {
BMI(BB, IP, RegRegMove[SrcClass], 1, DestReg).addReg(SrcReg);
} else if (SrcClass == cFP) {
if (SrcTy == Type::FloatTy) { // double -> float
assert(DestTy == Type::DoubleTy && "Unknown cFP member!");
BMI(BB, IP, X86::FpMOV, 1, DestReg).addReg(SrcReg);
} else { // float -> double
assert(SrcTy == Type::DoubleTy && DestTy == Type::FloatTy &&
"Unknown cFP member!");
// Truncate from double to float by storing to memory as short, then
// reading it back.
unsigned FltAlign = TM.getTargetData().getFloatAlignment();
int FrameIdx = F->getFrameInfo()->CreateStackObject(4, FltAlign);
addFrameReference(BMI(BB, IP, X86::FSTr32, 5), FrameIdx).addReg(SrcReg);
addFrameReference(BMI(BB, IP, X86::FLDr32, 5, DestReg), FrameIdx);
}
} else if (SrcClass == cLong) {
BMI(BB, IP, X86::MOVrr32, 1, DestReg).addReg(SrcReg);
BMI(BB, IP, X86::MOVrr32, 1, DestReg+1).addReg(SrcReg+1);
} else {
assert(0 && "Cannot handle this type of cast instruction!");
abort();
}
return;
}
// Handle cast of SMALLER int to LARGER int using a move with sign extension
// or zero extension, depending on whether the source type was signed.
if (SrcClass <= cInt && (DestClass <= cInt || DestClass == cLong) &&
SrcClass < DestClass) {
bool isLong = DestClass == cLong;
if (isLong) DestClass = cInt;
static const unsigned Opc[][4] = {
{ X86::MOVSXr16r8, X86::MOVSXr32r8, X86::MOVSXr32r16, X86::MOVrr32 }, // s
{ X86::MOVZXr16r8, X86::MOVZXr32r8, X86::MOVZXr32r16, X86::MOVrr32 } // u
};
bool isUnsigned = SrcTy->isUnsigned();
BMI(BB, IP, Opc[isUnsigned][SrcClass + DestClass - 1], 1,
DestReg).addReg(SrcReg);
if (isLong) { // Handle upper 32 bits as appropriate...
if (isUnsigned) // Zero out top bits...
BMI(BB, IP, X86::MOVri32, 1, DestReg+1).addZImm(0);
else // Sign extend bottom half...
BMI(BB, IP, X86::SARri32, 2, DestReg+1).addReg(DestReg).addZImm(31);
}
return;
}
// Special case long -> int ...
if (SrcClass == cLong && DestClass == cInt) {
BMI(BB, IP, X86::MOVrr32, 1, DestReg).addReg(SrcReg);
return;
}
// Handle cast of LARGER int to SMALLER int using a move to EAX followed by a
// move out of AX or AL.
if ((SrcClass <= cInt || SrcClass == cLong) && DestClass <= cInt
&& SrcClass > DestClass) {
static const unsigned AReg[] = { X86::AL, X86::AX, X86::EAX, 0, X86::EAX };
BMI(BB, IP, RegRegMove[SrcClass], 1, AReg[SrcClass]).addReg(SrcReg);
BMI(BB, IP, RegRegMove[DestClass], 1, DestReg).addReg(AReg[DestClass]);
return;
}
// Handle casts from integer to floating point now...
if (DestClass == cFP) {
// Promote the integer to a type supported by FLD. We do this because there
// are no unsigned FLD instructions, so we must promote an unsigned value to
// a larger signed value, then use FLD on the larger value.
//
const Type *PromoteType = 0;
unsigned PromoteOpcode;
unsigned RealDestReg = DestReg;
switch (SrcTy->getPrimitiveID()) {
case Type::BoolTyID:
case Type::SByteTyID:
// We don't have the facilities for directly loading byte sized data from
// memory (even signed). Promote it to 16 bits.
PromoteType = Type::ShortTy;
PromoteOpcode = X86::MOVSXr16r8;
break;
case Type::UByteTyID:
PromoteType = Type::ShortTy;
PromoteOpcode = X86::MOVZXr16r8;
break;
case Type::UShortTyID:
PromoteType = Type::IntTy;
PromoteOpcode = X86::MOVZXr32r16;
break;
case Type::UIntTyID: {
// Make a 64 bit temporary... and zero out the top of it...
unsigned TmpReg = makeAnotherReg(Type::LongTy);
BMI(BB, IP, X86::MOVrr32, 1, TmpReg).addReg(SrcReg);
BMI(BB, IP, X86::MOVri32, 1, TmpReg+1).addZImm(0);
SrcTy = Type::LongTy;
SrcClass = cLong;
SrcReg = TmpReg;
break;
}
case Type::ULongTyID:
// Don't fild into the read destination.
DestReg = makeAnotherReg(Type::DoubleTy);
break;
default: // No promotion needed...
break;
}
if (PromoteType) {
unsigned TmpReg = makeAnotherReg(PromoteType);
unsigned Opc = SrcTy->isSigned() ? X86::MOVSXr16r8 : X86::MOVZXr16r8;
BMI(BB, IP, Opc, 1, TmpReg).addReg(SrcReg);
SrcTy = PromoteType;
SrcClass = getClass(PromoteType);
SrcReg = TmpReg;
}
// Spill the integer to memory and reload it from there...
int FrameIdx =
F->getFrameInfo()->CreateStackObject(SrcTy, TM.getTargetData());
if (SrcClass == cLong) {
addFrameReference(BMI(BB, IP, X86::MOVmr32, 5), FrameIdx).addReg(SrcReg);
addFrameReference(BMI(BB, IP, X86::MOVmr32, 5),
FrameIdx, 4).addReg(SrcReg+1);
} else {
static const unsigned Op1[] = { X86::MOVmr8, X86::MOVmr16, X86::MOVmr32 };
addFrameReference(BMI(BB, IP, Op1[SrcClass], 5), FrameIdx).addReg(SrcReg);
}
static const unsigned Op2[] =
{ 0/*byte*/, X86::FILDr16, X86::FILDr32, 0/*FP*/, X86::FILDr64 };
addFrameReference(BMI(BB, IP, Op2[SrcClass], 5, DestReg), FrameIdx);
// We need special handling for unsigned 64-bit integer sources. If the
// input number has the "sign bit" set, then we loaded it incorrectly as a
// negative 64-bit number. In this case, add an offset value.
if (SrcTy == Type::ULongTy) {
// Emit a test instruction to see if the dynamic input value was signed.
BMI(BB, IP, X86::TESTrr32, 2).addReg(SrcReg+1).addReg(SrcReg+1);
// If the sign bit is set, get a pointer to an offset, otherwise get a
// pointer to a zero.
MachineConstantPool *CP = F->getConstantPool();
unsigned Zero = makeAnotherReg(Type::IntTy);
Constant *Null = Constant::getNullValue(Type::UIntTy);
addConstantPoolReference(BMI(BB, IP, X86::LEAr32, 5, Zero),
CP->getConstantPoolIndex(Null));
unsigned Offset = makeAnotherReg(Type::IntTy);
Constant *OffsetCst = ConstantUInt::get(Type::UIntTy, 0x5f800000);
addConstantPoolReference(BMI(BB, IP, X86::LEAr32, 5, Offset),
CP->getConstantPoolIndex(OffsetCst));
unsigned Addr = makeAnotherReg(Type::IntTy);
BMI(BB, IP, X86::CMOVSrr32, 2, Addr).addReg(Zero).addReg(Offset);
// Load the constant for an add. FIXME: this could make an 'fadd' that
// reads directly from memory, but we don't support these yet.
unsigned ConstReg = makeAnotherReg(Type::DoubleTy);
addDirectMem(BMI(BB, IP, X86::FLDr32, 4, ConstReg), Addr);
BMI(BB, IP, X86::FpADD, 2, RealDestReg).addReg(ConstReg).addReg(DestReg);
}
return;
}
// Handle casts from floating point to integer now...
if (SrcClass == cFP) {
// Change the floating point control register to use "round towards zero"
// mode when truncating to an integer value.
//
int CWFrameIdx = F->getFrameInfo()->CreateStackObject(2, 2);
addFrameReference(BMI(BB, IP, X86::FNSTCWm16, 4), CWFrameIdx);
// Load the old value of the high byte of the control word...
unsigned HighPartOfCW = makeAnotherReg(Type::UByteTy);
addFrameReference(BMI(BB, IP, X86::MOVrm8, 4, HighPartOfCW), CWFrameIdx, 1);
// Set the high part to be round to zero...
addFrameReference(BMI(BB, IP, X86::MOVmi8, 5), CWFrameIdx, 1).addZImm(12);
// Reload the modified control word now...
addFrameReference(BMI(BB, IP, X86::FLDCWm16, 4), CWFrameIdx);
// Restore the memory image of control word to original value
addFrameReference(BMI(BB, IP, X86::MOVmr8, 5),
CWFrameIdx, 1).addReg(HighPartOfCW);
// We don't have the facilities for directly storing byte sized data to
// memory. Promote it to 16 bits. We also must promote unsigned values to
// larger classes because we only have signed FP stores.
unsigned StoreClass = DestClass;
const Type *StoreTy = DestTy;
if (StoreClass == cByte || DestTy->isUnsigned())
switch (StoreClass) {
case cByte: StoreTy = Type::ShortTy; StoreClass = cShort; break;
case cShort: StoreTy = Type::IntTy; StoreClass = cInt; break;
case cInt: StoreTy = Type::LongTy; StoreClass = cLong; break;
// The following treatment of cLong may not be perfectly right,
// but it survives chains of casts of the form
// double->ulong->double.
case cLong: StoreTy = Type::LongTy; StoreClass = cLong; break;
default: assert(0 && "Unknown store class!");
}
// Spill the integer to memory and reload it from there...
int FrameIdx =
F->getFrameInfo()->CreateStackObject(StoreTy, TM.getTargetData());
static const unsigned Op1[] =
{ 0, X86::FISTr16, X86::FISTr32, 0, X86::FISTPr64 };
addFrameReference(BMI(BB, IP, Op1[StoreClass], 5), FrameIdx).addReg(SrcReg);
if (DestClass == cLong) {
addFrameReference(BMI(BB, IP, X86::MOVrm32, 4, DestReg), FrameIdx);
addFrameReference(BMI(BB, IP, X86::MOVrm32, 4, DestReg+1), FrameIdx, 4);
} else {
static const unsigned Op2[] = { X86::MOVrm8, X86::MOVrm16, X86::MOVrm32 };
addFrameReference(BMI(BB, IP, Op2[DestClass], 4, DestReg), FrameIdx);
}
// Reload the original control word now...
addFrameReference(BMI(BB, IP, X86::FLDCWm16, 4), CWFrameIdx);
return;
}
// Anything we haven't handled already, we can't (yet) handle at all.
assert(0 && "Unhandled cast instruction!");
abort();
}
/// visitVANextInst - Implement the va_next instruction...
///
void ISel::visitVANextInst(VANextInst &I) {
unsigned VAList = getReg(I.getOperand(0));
unsigned DestReg = getReg(I);
unsigned Size;
switch (I.getArgType()->getPrimitiveID()) {
default:
std::cerr << I;
assert(0 && "Error: bad type for va_next instruction!");
return;
case Type::PointerTyID:
case Type::UIntTyID:
case Type::IntTyID:
Size = 4;
break;
case Type::ULongTyID:
case Type::LongTyID:
case Type::DoubleTyID:
Size = 8;
break;
}
// Increment the VAList pointer...
BuildMI(BB, X86::ADDri32, 2, DestReg).addReg(VAList).addZImm(Size);
}
void ISel::visitVAArgInst(VAArgInst &I) {
unsigned VAList = getReg(I.getOperand(0));
unsigned DestReg = getReg(I);
switch (I.getType()->getPrimitiveID()) {
default:
std::cerr << I;
assert(0 && "Error: bad type for va_next instruction!");
return;
case Type::PointerTyID:
case Type::UIntTyID:
case Type::IntTyID:
addDirectMem(BuildMI(BB, X86::MOVrm32, 4, DestReg), VAList);
break;
case Type::ULongTyID:
case Type::LongTyID:
addDirectMem(BuildMI(BB, X86::MOVrm32, 4, DestReg), VAList);
addRegOffset(BuildMI(BB, X86::MOVrm32, 4, DestReg+1), VAList, 4);
break;
case Type::DoubleTyID:
addDirectMem(BuildMI(BB, X86::FLDr64, 4, DestReg), VAList);
break;
}
}
void ISel::visitGetElementPtrInst(GetElementPtrInst &I) {
// If this GEP instruction will be folded into all of its users, we don't need
// to explicitly calculate it!
unsigned A, B, C, D;
if (isGEPFoldable(0, I.getOperand(0), I.op_begin()+1, I.op_end(), A,B,C,D)) {
// Check all of the users of the instruction to see if they are loads and
// stores.
bool AllWillFold = true;
for (Value::use_iterator UI = I.use_begin(), E = I.use_end(); UI != E; ++UI)
if (cast<Instruction>(*UI)->getOpcode() != Instruction::Load)
if (cast<Instruction>(*UI)->getOpcode() != Instruction::Store ||
cast<Instruction>(*UI)->getOperand(0) == &I) {
AllWillFold = false;
break;
}
// If the instruction is foldable, and will be folded into all users, don't
// emit it!
if (AllWillFold) return;
}
unsigned outputReg = getReg(I);
emitGEPOperation(BB, BB->end(), I.getOperand(0),
I.op_begin()+1, I.op_end(), outputReg);
}
/// getGEPIndex - Inspect the getelementptr operands specified with GEPOps and
/// GEPTypes (the derived types being stepped through at each level). On return
/// from this function, if some indexes of the instruction are representable as
/// an X86 lea instruction, the machine operands are put into the Ops
/// instruction and the consumed indexes are poped from the GEPOps/GEPTypes
/// lists. Otherwise, GEPOps.size() is returned. If this returns a an
/// addressing mode that only partially consumes the input, the BaseReg input of
/// the addressing mode must be left free.
///
/// Note that there is one fewer entry in GEPTypes than there is in GEPOps.
///
void ISel::getGEPIndex(MachineBasicBlock *MBB, MachineBasicBlock::iterator IP,
std::vector<Value*> &GEPOps,
std::vector<const Type*> &GEPTypes, unsigned &BaseReg,
unsigned &Scale, unsigned &IndexReg, unsigned &Disp) {
const TargetData &TD = TM.getTargetData();
// Clear out the state we are working with...
BaseReg = 0; // No base register
Scale = 1; // Unit scale
IndexReg = 0; // No index register
Disp = 0; // No displacement
// While there are GEP indexes that can be folded into the current address,
// keep processing them.
while (!GEPTypes.empty()) {
if (const StructType *StTy = dyn_cast<StructType>(GEPTypes.back())) {
// It's a struct access. CUI is the index into the structure,
// which names the field. This index must have unsigned type.
const ConstantUInt *CUI = cast<ConstantUInt>(GEPOps.back());
// Use the TargetData structure to pick out what the layout of the
// structure is in memory. Since the structure index must be constant, we
// can get its value and use it to find the right byte offset from the
// StructLayout class's list of structure member offsets.
Disp += TD.getStructLayout(StTy)->MemberOffsets[CUI->getValue()];
GEPOps.pop_back(); // Consume a GEP operand
GEPTypes.pop_back();
} else {
// It's an array or pointer access: [ArraySize x ElementType].
const SequentialType *SqTy = cast<SequentialType>(GEPTypes.back());
Value *idx = GEPOps.back();
// idx is the index into the array. Unlike with structure
// indices, we may not know its actual value at code-generation
// time.
assert(idx->getType() == Type::LongTy && "Bad GEP array index!");
// If idx is a constant, fold it into the offset.
unsigned TypeSize = TD.getTypeSize(SqTy->getElementType());
if (ConstantSInt *CSI = dyn_cast<ConstantSInt>(idx)) {
Disp += TypeSize*CSI->getValue();
} else {
// If the index reg is already taken, we can't handle this index.
if (IndexReg) return;
// If this is a size that we can handle, then add the index as
switch (TypeSize) {
case 1: case 2: case 4: case 8:
// These are all acceptable scales on X86.
Scale = TypeSize;
break;
default:
// Otherwise, we can't handle this scale
return;
}
if (CastInst *CI = dyn_cast<CastInst>(idx))
if (CI->getOperand(0)->getType() == Type::IntTy ||
CI->getOperand(0)->getType() == Type::UIntTy)
idx = CI->getOperand(0);
IndexReg = MBB ? getReg(idx, MBB, IP) : 1;
}
GEPOps.pop_back(); // Consume a GEP operand
GEPTypes.pop_back();
}
}
// GEPTypes is empty, which means we have a single operand left. See if we
// can set it as the base register.
//
// FIXME: When addressing modes are more powerful/correct, we could load
// global addresses directly as 32-bit immediates.
assert(BaseReg == 0);
BaseReg = MBB ? getReg(GEPOps[0], MBB, IP) : 1;
GEPOps.pop_back(); // Consume the last GEP operand
}
/// isGEPFoldable - Return true if the specified GEP can be completely
/// folded into the addressing mode of a load/store or lea instruction.
bool ISel::isGEPFoldable(MachineBasicBlock *MBB,
Value *Src, User::op_iterator IdxBegin,
User::op_iterator IdxEnd, unsigned &BaseReg,
unsigned &Scale, unsigned &IndexReg, unsigned &Disp) {
if (ConstantPointerRef *CPR = dyn_cast<ConstantPointerRef>(Src))
Src = CPR->getValue();
std::vector<Value*> GEPOps;
GEPOps.resize(IdxEnd-IdxBegin+1);
GEPOps[0] = Src;
std::copy(IdxBegin, IdxEnd, GEPOps.begin()+1);
std::vector<const Type*> GEPTypes;
GEPTypes.assign(gep_type_begin(Src->getType(), IdxBegin, IdxEnd),
gep_type_end(Src->getType(), IdxBegin, IdxEnd));
MachineBasicBlock::iterator IP;
if (MBB) IP = MBB->end();
getGEPIndex(MBB, IP, GEPOps, GEPTypes, BaseReg, Scale, IndexReg, Disp);
// We can fold it away iff the getGEPIndex call eliminated all operands.
return GEPOps.empty();
}
void ISel::emitGEPOperation(MachineBasicBlock *MBB,
MachineBasicBlock::iterator IP,
Value *Src, User::op_iterator IdxBegin,
User::op_iterator IdxEnd, unsigned TargetReg) {
const TargetData &TD = TM.getTargetData();
if (ConstantPointerRef *CPR = dyn_cast<ConstantPointerRef>(Src))
Src = CPR->getValue();
std::vector<Value*> GEPOps;
GEPOps.resize(IdxEnd-IdxBegin+1);
GEPOps[0] = Src;
std::copy(IdxBegin, IdxEnd, GEPOps.begin()+1);
std::vector<const Type*> GEPTypes;
GEPTypes.assign(gep_type_begin(Src->getType(), IdxBegin, IdxEnd),
gep_type_end(Src->getType(), IdxBegin, IdxEnd));
// Keep emitting instructions until we consume the entire GEP instruction.
while (!GEPOps.empty()) {
unsigned OldSize = GEPOps.size();
unsigned BaseReg, Scale, IndexReg, Disp;
getGEPIndex(MBB, IP, GEPOps, GEPTypes, BaseReg, Scale, IndexReg, Disp);
if (GEPOps.size() != OldSize) {
// getGEPIndex consumed some of the input. Build an LEA instruction here.
unsigned NextTarget = 0;
if (!GEPOps.empty()) {
assert(BaseReg == 0 &&
"getGEPIndex should have left the base register open for chaining!");
NextTarget = BaseReg = makeAnotherReg(Type::UIntTy);
}
if (IndexReg == 0 && Disp == 0)
BMI(MBB, IP, X86::MOVrr32, 1, TargetReg).addReg(BaseReg);
else
addFullAddress(BMI(MBB, IP, X86::LEAr32, 5, TargetReg),
BaseReg, Scale, IndexReg, Disp);
--IP;
TargetReg = NextTarget;
} else if (GEPTypes.empty()) {
// The getGEPIndex operation didn't want to build an LEA. Check to see if
// all operands are consumed but the base pointer. If so, just load it
// into the register.
if (GlobalValue *GV = dyn_cast<GlobalValue>(GEPOps[0])) {
BMI(MBB, IP, X86::MOVri32, 1, TargetReg).addGlobalAddress(GV);
} else {
unsigned BaseReg = getReg(GEPOps[0], MBB, IP);
BMI(MBB, IP, X86::MOVrr32, 1, TargetReg).addReg(BaseReg);
}
break; // we are now done
} else {
// It's an array or pointer access: [ArraySize x ElementType].
const SequentialType *SqTy = cast<SequentialType>(GEPTypes.back());
Value *idx = GEPOps.back();
GEPOps.pop_back(); // Consume a GEP operand
GEPTypes.pop_back();
// idx is the index into the array. Unlike with structure
// indices, we may not know its actual value at code-generation
// time.
assert(idx->getType() == Type::LongTy && "Bad GEP array index!");
// Most GEP instructions use a [cast (int/uint) to LongTy] as their
// operand on X86. Handle this case directly now...
if (CastInst *CI = dyn_cast<CastInst>(idx))
if (CI->getOperand(0)->getType() == Type::IntTy ||
CI->getOperand(0)->getType() == Type::UIntTy)
idx = CI->getOperand(0);
// We want to add BaseReg to(idxReg * sizeof ElementType). First, we
// must find the size of the pointed-to type (Not coincidentally, the next
// type is the type of the elements in the array).
const Type *ElTy = SqTy->getElementType();
unsigned elementSize = TD.getTypeSize(ElTy);
// If idxReg is a constant, we don't need to perform the multiply!
if (ConstantSInt *CSI = dyn_cast<ConstantSInt>(idx)) {
if (!CSI->isNullValue()) {
unsigned Offset = elementSize*CSI->getValue();
unsigned Reg = makeAnotherReg(Type::UIntTy);
BMI(MBB, IP, X86::ADDri32, 2, TargetReg).addReg(Reg).addZImm(Offset);
--IP; // Insert the next instruction before this one.
TargetReg = Reg; // Codegen the rest of the GEP into this
}
} else if (elementSize == 1) {
// If the element size is 1, we don't have to multiply, just add
unsigned idxReg = getReg(idx, MBB, IP);
unsigned Reg = makeAnotherReg(Type::UIntTy);
BMI(MBB, IP, X86::ADDrr32, 2, TargetReg).addReg(Reg).addReg(idxReg);
--IP; // Insert the next instruction before this one.
TargetReg = Reg; // Codegen the rest of the GEP into this
} else {
unsigned idxReg = getReg(idx, MBB, IP);
unsigned OffsetReg = makeAnotherReg(Type::UIntTy);
// Make sure we can back the iterator up to point to the first
// instruction emitted.
MachineBasicBlock::iterator BeforeIt = IP;
if (IP == MBB->begin())
BeforeIt = MBB->end();
else
--BeforeIt;
doMultiplyConst(MBB, IP, OffsetReg, Type::IntTy, idxReg, elementSize);
// Emit an ADD to add OffsetReg to the basePtr.
unsigned Reg = makeAnotherReg(Type::UIntTy);
BMI(MBB, IP, X86::ADDrr32, 2, TargetReg).addReg(Reg).addReg(OffsetReg);
// Step to the first instruction of the multiply.
if (BeforeIt == MBB->end())
IP = MBB->begin();
else
IP = ++BeforeIt;
TargetReg = Reg; // Codegen the rest of the GEP into this
}
}
}
}
/// visitAllocaInst - If this is a fixed size alloca, allocate space from the
/// frame manager, otherwise do it the hard way.
///
void ISel::visitAllocaInst(AllocaInst &I) {
// Find the data size of the alloca inst's getAllocatedType.
const Type *Ty = I.getAllocatedType();
unsigned TySize = TM.getTargetData().getTypeSize(Ty);
// If this is a fixed size alloca in the entry block for the function,
// statically stack allocate the space.
//
if (ConstantUInt *CUI = dyn_cast<ConstantUInt>(I.getArraySize())) {
if (I.getParent() == I.getParent()->getParent()->begin()) {
TySize *= CUI->getValue(); // Get total allocated size...
unsigned Alignment = TM.getTargetData().getTypeAlignment(Ty);
// Create a new stack object using the frame manager...
int FrameIdx = F->getFrameInfo()->CreateStackObject(TySize, Alignment);
addFrameReference(BuildMI(BB, X86::LEAr32, 5, getReg(I)), FrameIdx);
return;
}
}
// Create a register to hold the temporary result of multiplying the type size
// constant by the variable amount.
unsigned TotalSizeReg = makeAnotherReg(Type::UIntTy);
unsigned SrcReg1 = getReg(I.getArraySize());
// TotalSizeReg = mul <numelements>, <TypeSize>
MachineBasicBlock::iterator MBBI = BB->end();
doMultiplyConst(BB, MBBI, TotalSizeReg, Type::UIntTy, SrcReg1, TySize);
// AddedSize = add <TotalSizeReg>, 15
unsigned AddedSizeReg = makeAnotherReg(Type::UIntTy);
BuildMI(BB, X86::ADDri32, 2, AddedSizeReg).addReg(TotalSizeReg).addZImm(15);
// AlignedSize = and <AddedSize>, ~15
unsigned AlignedSize = makeAnotherReg(Type::UIntTy);
BuildMI(BB, X86::ANDri32, 2, AlignedSize).addReg(AddedSizeReg).addZImm(~15);
// Subtract size from stack pointer, thereby allocating some space.
BuildMI(BB, X86::SUBrr32, 2, X86::ESP).addReg(X86::ESP).addReg(AlignedSize);
// Put a pointer to the space into the result register, by copying
// the stack pointer.
BuildMI(BB, X86::MOVrr32, 1, getReg(I)).addReg(X86::ESP);
// Inform the Frame Information that we have just allocated a variable-sized
// object.
F->getFrameInfo()->CreateVariableSizedObject();
}
/// visitMallocInst - Malloc instructions are code generated into direct calls
/// to the library malloc.
///
void ISel::visitMallocInst(MallocInst &I) {
unsigned AllocSize = TM.getTargetData().getTypeSize(I.getAllocatedType());
unsigned Arg;
if (ConstantUInt *C = dyn_cast<ConstantUInt>(I.getOperand(0))) {
Arg = getReg(ConstantUInt::get(Type::UIntTy, C->getValue() * AllocSize));
} else {
Arg = makeAnotherReg(Type::UIntTy);
unsigned Op0Reg = getReg(I.getOperand(0));
MachineBasicBlock::iterator MBBI = BB->end();
doMultiplyConst(BB, MBBI, Arg, Type::UIntTy, Op0Reg, AllocSize);
}
std::vector<ValueRecord> Args;
Args.push_back(ValueRecord(Arg, Type::UIntTy));
MachineInstr *TheCall = BuildMI(X86::CALLpcrel32,
1).addExternalSymbol("malloc", true);
doCall(ValueRecord(getReg(I), I.getType()), TheCall, Args);
}
/// visitFreeInst - Free instructions are code gen'd to call the free libc
/// function.
///
void ISel::visitFreeInst(FreeInst &I) {
std::vector<ValueRecord> Args;
Args.push_back(ValueRecord(I.getOperand(0)));
MachineInstr *TheCall = BuildMI(X86::CALLpcrel32,
1).addExternalSymbol("free", true);
doCall(ValueRecord(0, Type::VoidTy), TheCall, Args);
}
/// createX86SimpleInstructionSelector - This pass converts an LLVM function
/// into a machine code representation is a very simple peep-hole fashion. The
/// generated code sucks but the implementation is nice and simple.
///
FunctionPass *llvm::createX86SimpleInstructionSelector(TargetMachine &TM) {
return new ISel(TM);
}
Reference in New Issue View Git Blame Copy Permalink