//===-- PeepholeOptimizer.cpp - X86 Peephole Optimizer --------------------===//
+//
+// 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 contains a peephole optimizer for the X86.
//
#include "X86.h"
#include "llvm/CodeGen/MachineFunctionPass.h"
#include "llvm/CodeGen/MachineInstrBuilder.h"
+#include "llvm/Target/MRegisterInfo.h"
+#include "Support/Statistic.h"
+#include "Support/STLExtras.h"
+
+using namespace llvm;
namespace {
+ Statistic<> NumPHOpts("x86-peephole",
+ "Number of peephole optimization performed");
struct PH : public MachineFunctionPass {
virtual bool runOnMachineFunction(MachineFunction &MF);
};
}
-FunctionPass *createX86PeepholeOptimizerPass() { return new PH(); }
+FunctionPass *llvm::createX86PeepholeOptimizerPass() { return new PH(); }
bool PH::runOnMachineFunction(MachineFunction &MF) {
bool Changed = false;
for (MachineFunction::iterator BI = MF.begin(), E = MF.end(); BI != E; ++BI)
for (MachineBasicBlock::iterator I = BI->begin(); I != BI->end(); )
- if (PeepholeOptimize(*BI, I))
+ if (PeepholeOptimize(*BI, I)) {
Changed = true;
- else
+ ++NumPHOpts;
+ } else
++I;
return Changed;
bool PH::PeepholeOptimize(MachineBasicBlock &MBB,
MachineBasicBlock::iterator &I) {
- MachineInstr *MI = *I;
- MachineInstr *Next = (I+1 != MBB.end()) ? *(I+1) : 0;
+ assert(I != MBB.end());
+ MachineBasicBlock::iterator NextI = next(I);
+
+ MachineInstr *MI = I;
+ MachineInstr *Next = (NextI != MBB.end()) ? &*NextI : (MachineInstr*)0;
unsigned Size = 0;
switch (MI->getOpcode()) {
case X86::MOVrr8:
case X86::MOVrr32: // Destroy X = X copies...
if (MI->getOperand(0).getReg() == MI->getOperand(1).getReg()) {
I = MBB.erase(I);
- delete MI;
return true;
}
return false;
+ // A large number of X86 instructions have forms which take an 8-bit
+ // immediate despite the fact that the operands are 16 or 32 bits. Because
+ // this can save three bytes of code size (and icache space), we want to
+ // shrink them if possible.
+ case X86::IMULrri16: case X86::IMULrri32:
+ assert(MI->getNumOperands() == 3 && "These should all have 3 operands!");
+ if (MI->getOperand(2).isImmediate()) {
+ int Val = MI->getOperand(2).getImmedValue();
+ // If the value is the same when signed extended from 8 bits...
+ if (Val == (signed int)(signed char)Val) {
+ unsigned Opcode;
+ switch (MI->getOpcode()) {
+ default: assert(0 && "Unknown opcode value!");
+ case X86::IMULrri16: Opcode = X86::IMULrri16b; break;
+ case X86::IMULrri32: Opcode = X86::IMULrri32b; break;
+ }
+ unsigned R0 = MI->getOperand(0).getReg();
+ unsigned R1 = MI->getOperand(1).getReg();
+ I = MBB.insert(MBB.erase(I),
+ BuildMI(Opcode, 2, R0).addReg(R1).addZImm((char)Val));
+ return true;
+ }
+ }
+ return false;
+
+ case X86::IMULrmi16: case X86::IMULrmi32:
+ assert(MI->getNumOperands() == 6 && "These should all have 6 operands!");
+ if (MI->getOperand(5).isImmediate()) {
+ int Val = MI->getOperand(5).getImmedValue();
+ // If the value is the same when signed extended from 8 bits...
+ if (Val == (signed int)(signed char)Val) {
+ unsigned Opcode;
+ switch (MI->getOpcode()) {
+ default: assert(0 && "Unknown opcode value!");
+ case X86::IMULrmi16: Opcode = X86::IMULrmi16b; break;
+ case X86::IMULrmi32: Opcode = X86::IMULrmi32b; break;
+ }
+ unsigned R0 = MI->getOperand(0).getReg();
+ unsigned R1 = MI->getOperand(1).getReg();
+ unsigned Scale = MI->getOperand(2).getImmedValue();
+ unsigned R2 = MI->getOperand(3).getReg();
+ unsigned Offset = MI->getOperand(4).getImmedValue();
+ I = MBB.insert(MBB.erase(I),
+ BuildMI(Opcode, 5, R0).addReg(R1).addZImm(Scale).
+ addReg(R2).addSImm(Offset).addZImm((char)Val));
+ return true;
+ }
+ }
+ return false;
+
+ case X86::ADDri16: case X86::ADDri32:
+ case X86::ADDmi16: case X86::ADDmi32:
+ case X86::SUBri16: case X86::SUBri32:
+ case X86::ANDri16: case X86::ANDri32:
+ case X86::ORri16: case X86::ORri32:
+ case X86::XORri16: case X86::XORri32:
+ assert(MI->getNumOperands() == 2 && "These should all have 2 operands!");
+ if (MI->getOperand(1).isImmediate()) {
+ int Val = MI->getOperand(1).getImmedValue();
+ // If the value is the same when signed extended from 8 bits...
+ if (Val == (signed int)(signed char)Val) {
+ unsigned Opcode;
+ switch (MI->getOpcode()) {
+ default: assert(0 && "Unknown opcode value!");
+ case X86::ADDri16: Opcode = X86::ADDri16b; break;
+ case X86::ADDri32: Opcode = X86::ADDri32b; break;
+ case X86::ADDmi16: Opcode = X86::ADDmi16b; break;
+ case X86::ADDmi32: Opcode = X86::ADDmi32b; break;
+ case X86::SUBri16: Opcode = X86::SUBri16b; break;
+ case X86::SUBri32: Opcode = X86::SUBri32b; break;
+ case X86::ANDri16: Opcode = X86::ANDri16b; break;
+ case X86::ANDri32: Opcode = X86::ANDri32b; break;
+ case X86::ORri16: Opcode = X86::ORri16b; break;
+ case X86::ORri32: Opcode = X86::ORri32b; break;
+ case X86::XORri16: Opcode = X86::XORri16b; break;
+ case X86::XORri32: Opcode = X86::XORri32b; break;
+ }
+ unsigned R0 = MI->getOperand(0).getReg();
+ I = MBB.insert(MBB.erase(I),
+ BuildMI(Opcode, 1, R0, MOTy::UseAndDef).addZImm((char)Val));
+ return true;
+ }
+ }
+ return false;
+
+
+ case X86::ANDmi16: case X86::ANDmi32:
+ assert(MI->getNumOperands() == 5 && "These should all have 5 operands!");
+ if (MI->getOperand(4).isImmediate()) {
+ int Val = MI->getOperand(4).getImmedValue();
+ // If the value is the same when signed extended from 8 bits...
+ if (Val == (signed int)(signed char)Val) {
+ unsigned Opcode;
+ switch (MI->getOpcode()) {
+ default: assert(0 && "Unknown opcode value!");
+ case X86::ANDmi16: Opcode = X86::ANDmi16b; break;
+ case X86::ANDmi32: Opcode = X86::ANDmi32b; break;
+ }
+ unsigned R0 = MI->getOperand(0).getReg();
+ unsigned Scale = MI->getOperand(1).getImmedValue();
+ unsigned R1 = MI->getOperand(2).getReg();
+ unsigned Offset = MI->getOperand(3).getImmedValue();
+ I = MBB.insert(MBB.erase(I),
+ BuildMI(Opcode, 5).addReg(R0).addZImm(Scale).
+ addReg(R1).addSImm(Offset).addZImm((char)Val));
+ return true;
+ }
+ }
+ return false;
+
#if 0
- case X86::MOVir32: Size++;
- case X86::MOVir16: Size++;
- case X86::MOVir8:
+ case X86::MOVri32: Size++;
+ case X86::MOVri16: Size++;
+ case X86::MOVri8:
// FIXME: We can only do this transformation if we know that flags are not
// used here, because XOR clobbers the flags!
if (MI->getOperand(1).isImmediate()) { // avoid mov EAX, <value>
if (Val == 0) { // mov EAX, 0 -> xor EAX, EAX
static const unsigned Opcode[] ={X86::XORrr8,X86::XORrr16,X86::XORrr32};
unsigned Reg = MI->getOperand(0).getReg();
- *I = BuildMI(Opcode[Size], 2, Reg).addReg(Reg).addReg(Reg);
- delete MI;
+ I = MBB.insert(MBB.erase(I),
+ BuildMI(Opcode[Size], 2, Reg).addReg(Reg).addReg(Reg));
return true;
} else if (Val == -1) { // mov EAX, -1 -> or EAX, -1
// TODO: 'or Reg, -1' has a smaller encoding than 'mov Reg, -1'
if (Next->getOpcode() == X86::BSWAPr32 &&
MI->getOperand(0).getReg() == Next->getOperand(0).getReg()) {
I = MBB.erase(MBB.erase(I));
- delete MI;
- delete Next;
return true;
}
return false;
return false;
}
}
+
+namespace {
+ class UseDefChains : public MachineFunctionPass {
+ std::vector<MachineInstr*> DefiningInst;
+ public:
+ // getDefinition - Return the machine instruction that defines the specified
+ // SSA virtual register.
+ MachineInstr *getDefinition(unsigned Reg) {
+ assert(MRegisterInfo::isVirtualRegister(Reg) &&
+ "use-def chains only exist for SSA registers!");
+ assert(Reg - MRegisterInfo::FirstVirtualRegister < DefiningInst.size() &&
+ "Unknown register number!");
+ assert(DefiningInst[Reg-MRegisterInfo::FirstVirtualRegister] &&
+ "Unknown register number!");
+ return DefiningInst[Reg-MRegisterInfo::FirstVirtualRegister];
+ }
+
+ // setDefinition - Update the use-def chains to indicate that MI defines
+ // register Reg.
+ void setDefinition(unsigned Reg, MachineInstr *MI) {
+ if (Reg-MRegisterInfo::FirstVirtualRegister >= DefiningInst.size())
+ DefiningInst.resize(Reg-MRegisterInfo::FirstVirtualRegister+1);
+ DefiningInst[Reg-MRegisterInfo::FirstVirtualRegister] = MI;
+ }
+
+ // removeDefinition - Update the use-def chains to forget about Reg
+ // entirely.
+ void removeDefinition(unsigned Reg) {
+ assert(getDefinition(Reg)); // Check validity
+ DefiningInst[Reg-MRegisterInfo::FirstVirtualRegister] = 0;
+ }
+
+ virtual bool runOnMachineFunction(MachineFunction &MF) {
+ for (MachineFunction::iterator BI = MF.begin(), E = MF.end(); BI!=E; ++BI)
+ for (MachineBasicBlock::iterator I = BI->begin(); I != BI->end(); ++I) {
+ for (unsigned i = 0, e = I->getNumOperands(); i != e; ++i) {
+ MachineOperand &MO = I->getOperand(i);
+ if (MO.isRegister() && MO.isDef() && !MO.isUse() &&
+ MRegisterInfo::isVirtualRegister(MO.getReg()))
+ setDefinition(MO.getReg(), I);
+ }
+ }
+ return false;
+ }
+
+ virtual void getAnalysisUsage(AnalysisUsage &AU) const {
+ AU.setPreservesAll();
+ MachineFunctionPass::getAnalysisUsage(AU);
+ }
+
+ virtual void releaseMemory() {
+ std::vector<MachineInstr*>().swap(DefiningInst);
+ }
+ };
+
+ RegisterAnalysis<UseDefChains> X("use-def-chains",
+ "use-def chain construction for machine code");
+}
+
+
+namespace {
+ Statistic<> NumSSAPHOpts("x86-ssa-peephole",
+ "Number of SSA peephole optimization performed");
+
+ /// SSAPH - This pass is an X86-specific, SSA-based, peephole optimizer. This
+ /// pass is really a bad idea: a better instruction selector should completely
+ /// supersume it. However, that will take some time to develop, and the
+ /// simple things this can do are important now.
+ class SSAPH : public MachineFunctionPass {
+ UseDefChains *UDC;
+ public:
+ virtual bool runOnMachineFunction(MachineFunction &MF);
+
+ bool PeepholeOptimize(MachineBasicBlock &MBB,
+ MachineBasicBlock::iterator &I);
+
+ virtual const char *getPassName() const {
+ return "X86 SSA-based Peephole Optimizer";
+ }
+
+ /// Propagate - Set MI[DestOpNo] = Src[SrcOpNo], optionally change the
+ /// opcode of the instruction, then return true.
+ bool Propagate(MachineInstr *MI, unsigned DestOpNo,
+ MachineInstr *Src, unsigned SrcOpNo, unsigned NewOpcode = 0){
+ MI->getOperand(DestOpNo) = Src->getOperand(SrcOpNo);
+ if (NewOpcode) MI->setOpcode(NewOpcode);
+ return true;
+ }
+
+ /// OptimizeAddress - If we can fold the addressing arithmetic for this
+ /// memory instruction into the instruction itself, do so and return true.
+ bool OptimizeAddress(MachineInstr *MI, unsigned OpNo);
+
+ /// getDefininingInst - If the specified operand is a read of an SSA
+ /// register, return the machine instruction defining it, otherwise, return
+ /// null.
+ MachineInstr *getDefiningInst(MachineOperand &MO) {
+ if (MO.isDef() || !MO.isRegister() ||
+ !MRegisterInfo::isVirtualRegister(MO.getReg())) return 0;
+ return UDC->getDefinition(MO.getReg());
+ }
+
+ virtual void getAnalysisUsage(AnalysisUsage &AU) const {
+ AU.addRequired<UseDefChains>();
+ AU.addPreserved<UseDefChains>();
+ MachineFunctionPass::getAnalysisUsage(AU);
+ }
+ };
+}
+
+FunctionPass *llvm::createX86SSAPeepholeOptimizerPass() { return new SSAPH(); }
+
+bool SSAPH::runOnMachineFunction(MachineFunction &MF) {
+ bool Changed = false;
+ bool LocalChanged;
+
+ UDC = &getAnalysis<UseDefChains>();
+
+ do {
+ LocalChanged = false;
+
+ for (MachineFunction::iterator BI = MF.begin(), E = MF.end(); BI != E; ++BI)
+ for (MachineBasicBlock::iterator I = BI->begin(); I != BI->end(); )
+ if (PeepholeOptimize(*BI, I)) {
+ LocalChanged = true;
+ ++NumSSAPHOpts;
+ } else
+ ++I;
+ Changed |= LocalChanged;
+ } while (LocalChanged);
+
+ return Changed;
+}
+
+static bool isValidScaleAmount(unsigned Scale) {
+ return Scale == 1 || Scale == 2 || Scale == 4 || Scale == 8;
+}
+
+/// OptimizeAddress - If we can fold the addressing arithmetic for this
+/// memory instruction into the instruction itself, do so and return true.
+bool SSAPH::OptimizeAddress(MachineInstr *MI, unsigned OpNo) {
+ MachineOperand &BaseRegOp = MI->getOperand(OpNo+0);
+ MachineOperand &ScaleOp = MI->getOperand(OpNo+1);
+ MachineOperand &IndexRegOp = MI->getOperand(OpNo+2);
+ MachineOperand &DisplacementOp = MI->getOperand(OpNo+3);
+
+ unsigned BaseReg = BaseRegOp.hasAllocatedReg() ? BaseRegOp.getReg() : 0;
+ unsigned Scale = ScaleOp.getImmedValue();
+ unsigned IndexReg = IndexRegOp.hasAllocatedReg() ? IndexRegOp.getReg() : 0;
+
+ bool Changed = false;
+
+ // If the base register is unset, and the index register is set with a scale
+ // of 1, move it to be the base register.
+ if (BaseRegOp.hasAllocatedReg() && BaseReg == 0 &&
+ Scale == 1 && IndexReg != 0) {
+ BaseRegOp.setReg(IndexReg);
+ IndexRegOp.setReg(0);
+ return true;
+ }
+
+ // Attempt to fold instructions used by the base register into the instruction
+ if (MachineInstr *DefInst = getDefiningInst(BaseRegOp)) {
+ switch (DefInst->getOpcode()) {
+ case X86::MOVri32:
+ // If there is no displacement set for this instruction set one now.
+ // FIXME: If we can fold two immediates together, we should do so!
+ if (DisplacementOp.isImmediate() && !DisplacementOp.getImmedValue()) {
+ if (DefInst->getOperand(1).isImmediate()) {
+ BaseRegOp.setReg(0);
+ return Propagate(MI, OpNo+3, DefInst, 1);
+ }
+ }
+ break;
+
+ case X86::ADDrr32:
+ // If the source is a register-register add, and we do not yet have an
+ // index register, fold the add into the memory address.
+ if (IndexReg == 0) {
+ BaseRegOp = DefInst->getOperand(1);
+ IndexRegOp = DefInst->getOperand(2);
+ ScaleOp.setImmedValue(1);
+ return true;
+ }
+ break;
+
+ case X86::SHLri32:
+ // If this shift could be folded into the index portion of the address if
+ // it were the index register, move it to the index register operand now,
+ // so it will be folded in below.
+ if ((Scale == 1 || (IndexReg == 0 && IndexRegOp.hasAllocatedReg())) &&
+ DefInst->getOperand(2).getImmedValue() < 4) {
+ std::swap(BaseRegOp, IndexRegOp);
+ ScaleOp.setImmedValue(1); Scale = 1;
+ std::swap(IndexReg, BaseReg);
+ Changed = true;
+ break;
+ }
+ }
+ }
+
+ // Attempt to fold instructions used by the index into the instruction
+ if (MachineInstr *DefInst = getDefiningInst(IndexRegOp)) {
+ switch (DefInst->getOpcode()) {
+ case X86::SHLri32: {
+ // Figure out what the resulting scale would be if we folded this shift.
+ unsigned ResScale = Scale * (1 << DefInst->getOperand(2).getImmedValue());
+ if (isValidScaleAmount(ResScale)) {
+ IndexRegOp = DefInst->getOperand(1);
+ ScaleOp.setImmedValue(ResScale);
+ return true;
+ }
+ break;
+ }
+ }
+ }
+
+ return Changed;
+}
+
+bool SSAPH::PeepholeOptimize(MachineBasicBlock &MBB,
+ MachineBasicBlock::iterator &I) {
+ MachineBasicBlock::iterator NextI = next(I);
+
+ MachineInstr *MI = I;
+ MachineInstr *Next = (NextI != MBB.end()) ? &*NextI : (MachineInstr*)0;
+
+ bool Changed = false;
+
+ // Scan the operands of this instruction. If any operands are
+ // register-register copies, replace the operand with the source.
+ for (unsigned i = 0, e = MI->getNumOperands(); i != e; ++i)
+ // Is this an SSA register use?
+ if (MachineInstr *DefInst = getDefiningInst(MI->getOperand(i)))
+ // If the operand is a vreg-vreg copy, it is always safe to replace the
+ // source value with the input operand.
+ if (DefInst->getOpcode() == X86::MOVrr8 ||
+ DefInst->getOpcode() == X86::MOVrr16 ||
+ DefInst->getOpcode() == X86::MOVrr32) {
+ // Don't propagate physical registers into PHI nodes...
+ if (MI->getOpcode() != X86::PHI ||
+ (DefInst->getOperand(1).isRegister() &&
+ MRegisterInfo::isVirtualRegister(DefInst->getOperand(1).getReg())))
+ Changed = Propagate(MI, i, DefInst, 1);
+ }
+
+
+ // Perform instruction specific optimizations.
+ switch (MI->getOpcode()) {
+
+ // Register to memory stores. Format: <base,scale,indexreg,immdisp>, srcreg
+ case X86::MOVrm32: case X86::MOVrm16: case X86::MOVrm8:
+ case X86::MOVmi32: case X86::MOVmi16: case X86::MOVmi8:
+ // Check to see if we can fold the source instruction into this one...
+ if (MachineInstr *SrcInst = getDefiningInst(MI->getOperand(4))) {
+ switch (SrcInst->getOpcode()) {
+ // Fold the immediate value into the store, if possible.
+ case X86::MOVri8: return Propagate(MI, 4, SrcInst, 1, X86::MOVmi8);
+ case X86::MOVri16: return Propagate(MI, 4, SrcInst, 1, X86::MOVmi16);
+ case X86::MOVri32: return Propagate(MI, 4, SrcInst, 1, X86::MOVmi32);
+ default: break;
+ }
+ }
+
+ // If we can optimize the addressing expression, do so now.
+ if (OptimizeAddress(MI, 0))
+ return true;
+ break;
+
+ case X86::MOVmr32:
+ case X86::MOVmr16:
+ case X86::MOVmr8:
+ // If we can optimize the addressing expression, do so now.
+ if (OptimizeAddress(MI, 1))
+ return true;
+ break;
+
+ default: break;
+ }
+
+ return Changed;
+}