Index: llvm/trunk/lib/Target/X86/X86FixupLEAs.cpp =================================================================== --- llvm/trunk/lib/Target/X86/X86FixupLEAs.cpp +++ llvm/trunk/lib/Target/X86/X86FixupLEAs.cpp @@ -9,6 +9,7 @@ // // This file defines the pass that finds instructions that can be // re-written as LEA instructions in order to reduce pipeline delays. +// When optimizing for size it replaces suitable LEAs with INC or DEC. // //===----------------------------------------------------------------------===// @@ -61,6 +62,11 @@ void processInstructionForSLM(MachineBasicBlock::iterator &I, MachineFunction::iterator MFI); + /// \brief Look for LEAs that add 1 to reg or subtract 1 from reg + /// and convert them to INC or DEC respectively. + bool fixupIncDec(MachineBasicBlock::iterator &I, + MachineFunction::iterator MFI) const; + /// \brief Determine if an instruction references a machine register /// and, if so, whether it reads or writes the register. RegUsageState usesRegister(MachineOperand &p, MachineBasicBlock::iterator I); @@ -89,6 +95,8 @@ private: MachineFunction *MF; const X86InstrInfo *TII; // Machine instruction info. + bool OptIncDec; + bool OptLEA; }; char FixupLEAPass::ID = 0; } @@ -150,7 +158,10 @@ bool FixupLEAPass::runOnMachineFunction(MachineFunction &Func) { MF = &Func; const X86Subtarget &ST = Func.getSubtarget(); - if (!ST.LEAusesAG() && !ST.slowLEA()) + OptIncDec = !ST.slowIncDec() || Func.getFunction()->optForMinSize(); + OptLEA = ST.LEAusesAG() || ST.slowLEA(); + + if (!OptLEA && !OptIncDec) return false; TII = ST.getInstrInfo(); @@ -222,6 +233,60 @@ return nullptr; } +static inline bool isLEA(const int opcode) { + return opcode == X86::LEA16r || opcode == X86::LEA32r || + opcode == X86::LEA64r || opcode == X86::LEA64_32r; +} + +/// isLEASimpleIncOrDec - Does this LEA have one these forms: +/// lea %reg, 1(%reg) +/// lea %reg, -1(%reg) +static inline bool isLEASimpleIncOrDec(MachineInstr *LEA) { + unsigned SrcReg = LEA->getOperand(1 + X86::AddrBaseReg).getReg(); + unsigned DstReg = LEA->getOperand(0).getReg(); + unsigned AddrDispOp = 1 + X86::AddrDisp; + return SrcReg == DstReg && + LEA->getOperand(1 + X86::AddrIndexReg).getReg() == 0 && + LEA->getOperand(1 + X86::AddrSegmentReg).getReg() == 0 && + LEA->getOperand(AddrDispOp).isImm() && + (LEA->getOperand(AddrDispOp).getImm() == 1 || + LEA->getOperand(AddrDispOp).getImm() == -1); +} + +bool FixupLEAPass::fixupIncDec(MachineBasicBlock::iterator &I, + MachineFunction::iterator MFI) const { + MachineInstr *MI = I; + int Opcode = MI->getOpcode(); + if (!isLEA(Opcode)) + return false; + + if (isLEASimpleIncOrDec(MI) && TII->isSafeToClobberEFLAGS(*MFI, I)) { + int NewOpcode; + bool isINC = MI->getOperand(4).getImm() == 1; + switch (Opcode) { + case X86::LEA16r: + NewOpcode = isINC ? X86::INC16r : X86::DEC16r; + break; + case X86::LEA32r: + case X86::LEA64_32r: + NewOpcode = isINC ? X86::INC32r : X86::DEC32r; + break; + case X86::LEA64r: + NewOpcode = isINC ? X86::INC64r : X86::DEC64r; + break; + } + + MachineInstr *NewMI = + BuildMI(*MFI, I, MI->getDebugLoc(), TII->get(NewOpcode)) + .addOperand(MI->getOperand(0)) + .addOperand(MI->getOperand(1)); + MFI->erase(I); + I = static_cast(NewMI); + return true; + } + return false; +} + void FixupLEAPass::processInstruction(MachineBasicBlock::iterator &I, MachineFunction::iterator MFI) { // Process a load, store, or LEA instruction. @@ -265,8 +330,7 @@ MachineFunction::iterator MFI) { MachineInstr *MI = I; const int opcode = MI->getOpcode(); - if (opcode != X86::LEA16r && opcode != X86::LEA32r && opcode != X86::LEA64r && - opcode != X86::LEA64_32r) + if (!isLEA(opcode)) return; if (MI->getOperand(5).getReg() != 0 || !MI->getOperand(4).isImm() || !TII->isSafeToClobberEFLAGS(*MFI, I)) @@ -280,7 +344,8 @@ return; int addrr_opcode, addri_opcode; switch (opcode) { - default: llvm_unreachable("Unexpected LEA instruction"); + default: + llvm_unreachable("Unexpected LEA instruction"); case X86::LEA16r: addrr_opcode = X86::ADD16rr; addri_opcode = X86::ADD16ri; @@ -330,10 +395,16 @@ MachineFunction::iterator MFI) { for (MachineBasicBlock::iterator I = MFI->begin(); I != MFI->end(); ++I) { - if (MF.getSubtarget().isSLM()) - processInstructionForSLM(I, MFI); - else - processInstruction(I, MFI); + if (OptIncDec) + if (fixupIncDec(I, MFI)) + continue; + + if (OptLEA) { + if (MF.getSubtarget().isSLM()) + processInstructionForSLM(I, MFI); + else + processInstruction(I, MFI); + } } return false; } Index: llvm/trunk/test/CodeGen/X86/fixup-lea.ll =================================================================== --- llvm/trunk/test/CodeGen/X86/fixup-lea.ll +++ llvm/trunk/test/CodeGen/X86/fixup-lea.ll @@ -0,0 +1,34 @@ +;RUN: llc < %s -march=x86 | FileCheck %s + +define void @foo(i32 inreg %dns) minsize { +entry: +; CHECK-LABEL: foo +; CHECK: dec + br label %for.body + +for.body: + %i.05 = phi i16 [ %dec, %for.body ], [ 0, %entry ] + %dec = add i16 %i.05, -1 + %conv = zext i16 %dec to i32 + %cmp = icmp slt i32 %conv, %dns + br i1 %cmp, label %for.body, label %for.end + +for.end: + ret void +} + +define void @bar(i32 inreg %dns) minsize { +entry: +; CHECK-LABEL: bar +; CHECK: inc + br label %for.body + +for.body: + %i.05 = phi i16 [ %inc, %for.body ], [ 0, %entry ] + %inc = add i16 %i.05, 1 + %conv = zext i16 %inc to i32 + %cmp = icmp slt i32 %conv, %dns + br i1 %cmp, label %for.body, label %for.end +for.end: + ret void +} Index: llvm/trunk/test/CodeGen/X86/lsr-static-addr.ll =================================================================== --- llvm/trunk/test/CodeGen/X86/lsr-static-addr.ll +++ llvm/trunk/test/CodeGen/X86/lsr-static-addr.ll @@ -18,7 +18,7 @@ ; ATOM-NEXT: movsd A(,%rax,8) ; ATOM-NEXT: mulsd ; ATOM-NEXT: movsd -; ATOM-NEXT: leaq 1(%rax), %rax +; ATOM-NEXT: incq %rax @A = external global [0 x double]