X-Git-Url: http://plrg.eecs.uci.edu/git/?a=blobdiff_plain;f=lib%2FCodeGen%2FPrologEpilogInserter.cpp;h=32c932552bedd8c438026a978da767bcda7b7b5d;hb=b10946a5a938a433ca4d7301b8b5ff5a8c11a7ff;hp=05052d354df7acf8cd45f70d33e0e1e2f0e09087;hpb=08ede262a744f99429658fadb43662441bdcb42d;p=oota-llvm.git diff --git a/lib/CodeGen/PrologEpilogInserter.cpp b/lib/CodeGen/PrologEpilogInserter.cpp index 05052d354df..32c932552be 100644 --- a/lib/CodeGen/PrologEpilogInserter.cpp +++ b/lib/CodeGen/PrologEpilogInserter.cpp @@ -2,8 +2,8 @@ // // 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 is distributed under the University of Illinois Open Source +// License. See LICENSE.TXT for details. // //===----------------------------------------------------------------------===// // @@ -14,142 +14,224 @@ // This pass must be run after register allocation. After this pass is // executed, it is illegal to construct MO_FrameIndex operands. // +// This pass provides an optional shrink wrapping variant of prolog/epilog +// insertion, enabled via --shrink-wrap. See ShrinkWrapping.cpp. +// //===----------------------------------------------------------------------===// -#include "llvm/CodeGen/Passes.h" -#include "llvm/CodeGen/MachineFunctionPass.h" +#define DEBUG_TYPE "pei" +#include "PrologEpilogInserter.h" +#include "llvm/InlineAsm.h" +#include "llvm/CodeGen/MachineDominators.h" +#include "llvm/CodeGen/MachineLoopInfo.h" #include "llvm/CodeGen/MachineInstr.h" #include "llvm/CodeGen/MachineFrameInfo.h" +#include "llvm/CodeGen/MachineRegisterInfo.h" +#include "llvm/CodeGen/RegisterScavenging.h" #include "llvm/Target/TargetMachine.h" -#include "llvm/Target/MRegisterInfo.h" -#include "llvm/Target/TargetFrameInfo.h" +#include "llvm/Target/TargetOptions.h" +#include "llvm/Target/TargetRegisterInfo.h" +#include "llvm/Target/TargetFrameLowering.h" #include "llvm/Target/TargetInstrInfo.h" -#include "llvm/Support/Visibility.h" +#include "llvm/Support/CommandLine.h" +#include "llvm/Support/Compiler.h" +#include "llvm/Support/Debug.h" +#include "llvm/ADT/IndexedMap.h" +#include "llvm/ADT/SmallSet.h" +#include "llvm/ADT/Statistic.h" +#include "llvm/ADT/STLExtras.h" +#include + using namespace llvm; -namespace { - struct VISIBILITY_HIDDEN PEI : public MachineFunctionPass { - const char *getPassName() const { - return "Prolog/Epilog Insertion & Frame Finalization"; - } +char PEI::ID = 0; - /// runOnMachineFunction - Insert prolog/epilog code and replace abstract - /// frame indexes with appropriate references. - /// - bool runOnMachineFunction(MachineFunction &Fn) { - // Get MachineDebugInfo so that we can track the construction of the - // frame. - if (MachineDebugInfo *DI = getAnalysisToUpdate()) { - Fn.getFrameInfo()->setMachineDebugInfo(DI); - } - - // Scan the function for modified caller saved registers and insert spill - // code for any caller saved registers that are modified. Also calculate - // the MaxCallFrameSize and HasCalls variables for the function's frame - // information and eliminates call frame pseudo instructions. - calculateCallerSavedRegisters(Fn); - - // Add the code to save and restore the caller saved registers - saveCallerSavedRegisters(Fn); - - // Allow the target machine to make final modifications to the function - // before the frame layout is finalized. - Fn.getTarget().getRegisterInfo()->processFunctionBeforeFrameFinalized(Fn); - - // Calculate actual frame offsets for all of the abstract stack objects... - calculateFrameObjectOffsets(Fn); - - // Add prolog and epilog code to the function. This function is required - // to align the stack frame as necessary for any stack variables or - // called functions. Because of this, calculateCallerSavedRegisters - // must be called before this function in order to set the HasCalls - // and MaxCallFrameSize variables. - insertPrologEpilogCode(Fn); - - // Replace all MO_FrameIndex operands with physical register references - // and actual offsets. - // - replaceFrameIndices(Fn); - - return true; - } - - private: - void calculateCallerSavedRegisters(MachineFunction &Fn); - void saveCallerSavedRegisters(MachineFunction &Fn); - void calculateFrameObjectOffsets(MachineFunction &Fn); - void replaceFrameIndices(MachineFunction &Fn); - void insertPrologEpilogCode(MachineFunction &Fn); - }; -} +INITIALIZE_PASS_BEGIN(PEI, "prologepilog", + "Prologue/Epilogue Insertion", false, false) +INITIALIZE_PASS_DEPENDENCY(MachineLoopInfo) +INITIALIZE_PASS_DEPENDENCY(MachineDominatorTree) +INITIALIZE_PASS_END(PEI, "prologepilog", + "Prologue/Epilogue Insertion", false, false) +STATISTIC(NumVirtualFrameRegs, "Number of virtual frame regs encountered"); +STATISTIC(NumScavengedRegs, "Number of frame index regs scavenged"); +STATISTIC(NumBytesStackSpace, + "Number of bytes used for stack in all functions"); /// createPrologEpilogCodeInserter - This function returns a pass that inserts /// prolog and epilog code, and eliminates abstract frame references. /// FunctionPass *llvm::createPrologEpilogCodeInserter() { return new PEI(); } - -/// calculateCallerSavedRegisters - Scan the function for modified caller saved -/// registers. Also calculate the MaxCallFrameSize and HasCalls variables for -/// the function's frame information and eliminates call frame pseudo -/// instructions. +/// runOnMachineFunction - Insert prolog/epilog code and replace abstract +/// frame indexes with appropriate references. /// -void PEI::calculateCallerSavedRegisters(MachineFunction &Fn) { - const MRegisterInfo *RegInfo = Fn.getTarget().getRegisterInfo(); - const TargetFrameInfo *TFI = Fn.getTarget().getFrameInfo(); +bool PEI::runOnMachineFunction(MachineFunction &Fn) { + const Function* F = Fn.getFunction(); + const TargetRegisterInfo *TRI = Fn.getTarget().getRegisterInfo(); + const TargetFrameLowering *TFI = Fn.getTarget().getFrameLowering(); + + RS = TRI->requiresRegisterScavenging(Fn) ? new RegScavenger() : NULL; + FrameIndexVirtualScavenging = TRI->requiresFrameIndexScavenging(Fn); + + // Calculate the MaxCallFrameSize and AdjustsStack variables for the + // function's frame information. Also eliminates call frame pseudo + // instructions. + calculateCallsInformation(Fn); + + // Allow the target machine to make some adjustments to the function + // e.g. UsedPhysRegs before calculateCalleeSavedRegisters. + TFI->processFunctionBeforeCalleeSavedScan(Fn, RS); + + // Scan the function for modified callee saved registers and insert spill code + // for any callee saved registers that are modified. + calculateCalleeSavedRegisters(Fn); + + // Determine placement of CSR spill/restore code: + // - With shrink wrapping, place spills and restores to tightly + // enclose regions in the Machine CFG of the function where + // they are used. + // - Without shink wrapping (default), place all spills in the + // entry block, all restores in return blocks. + placeCSRSpillsAndRestores(Fn); + + // Add the code to save and restore the callee saved registers + if (!F->hasFnAttr(Attribute::Naked)) + insertCSRSpillsAndRestores(Fn); + + // Allow the target machine to make final modifications to the function + // before the frame layout is finalized. + TFI->processFunctionBeforeFrameFinalized(Fn); + + // Calculate actual frame offsets for all abstract stack objects... + calculateFrameObjectOffsets(Fn); + + // Add prolog and epilog code to the function. This function is required + // to align the stack frame as necessary for any stack variables or + // called functions. Because of this, calculateCalleeSavedRegisters() + // must be called before this function in order to set the AdjustsStack + // and MaxCallFrameSize variables. + if (!F->hasFnAttr(Attribute::Naked)) + insertPrologEpilogCode(Fn); + + // Replace all MO_FrameIndex operands with physical register references + // and actual offsets. + // + replaceFrameIndices(Fn); - // Get the callee saved register list... - const unsigned *CSRegs = RegInfo->getCalleeSaveRegs(); + // If register scavenging is needed, as we've enabled doing it as a + // post-pass, scavenge the virtual registers that frame index elimiation + // inserted. + if (TRI->requiresRegisterScavenging(Fn) && FrameIndexVirtualScavenging) + scavengeFrameVirtualRegs(Fn); - // Get the function call frame set-up and tear-down instruction opcode - int FrameSetupOpcode = RegInfo->getCallFrameSetupOpcode(); - int FrameDestroyOpcode = RegInfo->getCallFrameDestroyOpcode(); + delete RS; + clearAllSets(); + return true; +} - // Early exit for targets which have no callee saved registers and no call - // frame setup/destroy pseudo instructions. - if ((CSRegs == 0 || CSRegs[0] == 0) && - FrameSetupOpcode == -1 && FrameDestroyOpcode == -1) - return; +#if 0 +void PEI::getAnalysisUsage(AnalysisUsage &AU) const { + AU.setPreservesCFG(); + if (ShrinkWrapping || ShrinkWrapFunc != "") { + AU.addRequired(); + AU.addRequired(); + } + AU.addPreserved(); + AU.addPreserved(); + MachineFunctionPass::getAnalysisUsage(AU); +} +#endif + +/// calculateCallsInformation - Calculate the MaxCallFrameSize and AdjustsStack +/// variables for the function's frame information and eliminate call frame +/// pseudo instructions. +void PEI::calculateCallsInformation(MachineFunction &Fn) { + const TargetRegisterInfo *RegInfo = Fn.getTarget().getRegisterInfo(); + const TargetInstrInfo &TII = *Fn.getTarget().getInstrInfo(); + const TargetFrameLowering *TFI = Fn.getTarget().getFrameLowering(); + MachineFrameInfo *MFI = Fn.getFrameInfo(); unsigned MaxCallFrameSize = 0; - bool HasCalls = false; + bool AdjustsStack = MFI->adjustsStack(); + // Get the function call frame set-up and tear-down instruction opcode + int FrameSetupOpcode = TII.getCallFrameSetupOpcode(); + int FrameDestroyOpcode = TII.getCallFrameDestroyOpcode(); + + // Early exit for targets which have no call frame setup/destroy pseudo + // instructions. + if (FrameSetupOpcode == -1 && FrameDestroyOpcode == -1) + return; + + std::vector FrameSDOps; for (MachineFunction::iterator BB = Fn.begin(), E = Fn.end(); BB != E; ++BB) - for (MachineBasicBlock::iterator I = BB->begin(); I != BB->end(); ) + for (MachineBasicBlock::iterator I = BB->begin(); I != BB->end(); ++I) if (I->getOpcode() == FrameSetupOpcode || I->getOpcode() == FrameDestroyOpcode) { assert(I->getNumOperands() >= 1 && "Call Frame Setup/Destroy Pseudo" " instructions should have a single immediate argument!"); - unsigned Size = I->getOperand(0).getImmedValue(); + unsigned Size = I->getOperand(0).getImm(); if (Size > MaxCallFrameSize) MaxCallFrameSize = Size; - HasCalls = true; - RegInfo->eliminateCallFramePseudoInstr(Fn, *BB, I++); - } else { - ++I; + AdjustsStack = true; + FrameSDOps.push_back(I); + } else if (I->isInlineAsm()) { + // Some inline asm's need a stack frame, as indicated by operand 1. + unsigned ExtraInfo = I->getOperand(InlineAsm::MIOp_ExtraInfo).getImm(); + if (ExtraInfo & InlineAsm::Extra_IsAlignStack) + AdjustsStack = true; } - MachineFrameInfo *FFI = Fn.getFrameInfo(); - FFI->setHasCalls(HasCalls); - FFI->setMaxCallFrameSize(MaxCallFrameSize); + MFI->setAdjustsStack(AdjustsStack); + MFI->setMaxCallFrameSize(MaxCallFrameSize); + + for (std::vector::iterator + i = FrameSDOps.begin(), e = FrameSDOps.end(); i != e; ++i) { + MachineBasicBlock::iterator I = *i; + + // If call frames are not being included as part of the stack frame, and + // the target doesn't indicate otherwise, remove the call frame pseudos + // here. The sub/add sp instruction pairs are still inserted, but we don't + // need to track the SP adjustment for frame index elimination. + if (TFI->canSimplifyCallFramePseudos(Fn)) + RegInfo->eliminateCallFramePseudoInstr(Fn, *I->getParent(), I); + } +} + + +/// calculateCalleeSavedRegisters - Scan the function for modified callee saved +/// registers. +void PEI::calculateCalleeSavedRegisters(MachineFunction &Fn) { + const TargetRegisterInfo *RegInfo = Fn.getTarget().getRegisterInfo(); + const TargetFrameLowering *TFI = Fn.getTarget().getFrameLowering(); + MachineFrameInfo *MFI = Fn.getFrameInfo(); + + // Get the callee saved register list... + const unsigned *CSRegs = RegInfo->getCalleeSavedRegs(&Fn); + + // These are used to keep track the callee-save area. Initialize them. + MinCSFrameIndex = INT_MAX; + MaxCSFrameIndex = 0; + + // Early exit for targets which have no callee saved registers. + if (CSRegs == 0 || CSRegs[0] == 0) + return; + + // In Naked functions we aren't going to save any registers. + if (Fn.getFunction()->hasFnAttr(Attribute::Naked)) + return; - // Now figure out which *callee saved* registers are modified by the current - // function, thus needing to be saved and restored in the prolog/epilog. - // - const bool *PhysRegsUsed = Fn.getUsedPhysregs(); - const TargetRegisterClass* const *CSRegClasses = - RegInfo->getCalleeSaveRegClasses(); std::vector CSI; for (unsigned i = 0; CSRegs[i]; ++i) { unsigned Reg = CSRegs[i]; - if (PhysRegsUsed[Reg]) { - // If the reg is modified, save it! - CSI.push_back(CalleeSavedInfo(Reg, CSRegClasses[i])); + if (Fn.getRegInfo().isPhysRegUsed(Reg)) { + // If the reg is modified, save it! + CSI.push_back(CalleeSavedInfo(Reg)); } else { for (const unsigned *AliasSet = RegInfo->getAliasSet(Reg); *AliasSet; ++AliasSet) { // Check alias registers too. - if (PhysRegsUsed[*AliasSet]) { - CSI.push_back(CalleeSavedInfo(Reg, CSRegClasses[i])); + if (Fn.getRegInfo().isPhysRegUsed(*AliasSet)) { + CSI.push_back(CalleeSavedInfo(Reg)); break; } } @@ -157,201 +239,476 @@ void PEI::calculateCallerSavedRegisters(MachineFunction &Fn) { } if (CSI.empty()) - return; // Early exit if no caller saved registers are modified! + return; // Early exit if no callee saved registers are modified! unsigned NumFixedSpillSlots; - const std::pair *FixedSpillSlots = - TFI->getCalleeSaveSpillSlots(NumFixedSpillSlots); + const TargetFrameLowering::SpillSlot *FixedSpillSlots = + TFI->getCalleeSavedSpillSlots(NumFixedSpillSlots); // Now that we know which registers need to be saved and restored, allocate // stack slots for them. - for (unsigned i = 0, e = CSI.size(); i != e; ++i) { - unsigned Reg = CSI[i].getReg(); - const TargetRegisterClass *RC = CSI[i].getRegClass(); + for (std::vector::iterator + I = CSI.begin(), E = CSI.end(); I != E; ++I) { + unsigned Reg = I->getReg(); + const TargetRegisterClass *RC = RegInfo->getMinimalPhysRegClass(Reg); + + int FrameIdx; + if (RegInfo->hasReservedSpillSlot(Fn, Reg, FrameIdx)) { + I->setFrameIdx(FrameIdx); + continue; + } // Check to see if this physreg must be spilled to a particular stack slot // on this target. - const std::pair *FixedSlot = FixedSpillSlots; + const TargetFrameLowering::SpillSlot *FixedSlot = FixedSpillSlots; while (FixedSlot != FixedSpillSlots+NumFixedSpillSlots && - FixedSlot->first != Reg) + FixedSlot->Reg != Reg) ++FixedSlot; - int FrameIdx; - if (FixedSlot == FixedSpillSlots+NumFixedSpillSlots) { + if (FixedSlot == FixedSpillSlots + NumFixedSpillSlots) { // Nope, just spill it anywhere convenient. - FrameIdx = FFI->CreateStackObject(RC->getSize(), RC->getAlignment()); + unsigned Align = RC->getAlignment(); + unsigned StackAlign = TFI->getStackAlignment(); + + // We may not be able to satisfy the desired alignment specification of + // the TargetRegisterClass if the stack alignment is smaller. Use the + // min. + Align = std::min(Align, StackAlign); + FrameIdx = MFI->CreateStackObject(RC->getSize(), Align, true); + if ((unsigned)FrameIdx < MinCSFrameIndex) MinCSFrameIndex = FrameIdx; + if ((unsigned)FrameIdx > MaxCSFrameIndex) MaxCSFrameIndex = FrameIdx; } else { // Spill it to the stack where we must. - FrameIdx = FFI->CreateFixedObject(RC->getSize(), FixedSlot->second); + FrameIdx = MFI->CreateFixedObject(RC->getSize(), FixedSlot->Offset, true); } - CSI[i].setFrameIdx(FrameIdx); + + I->setFrameIdx(FrameIdx); } - FFI->setCalleeSavedInfo(CSI); + MFI->setCalleeSavedInfo(CSI); } -/// saveCallerSavedRegisters - Insert spill code for any caller saved registers -/// that are modified in the function. +/// insertCSRSpillsAndRestores - Insert spill and restore code for +/// callee saved registers used in the function, handling shrink wrapping. /// -void PEI::saveCallerSavedRegisters(MachineFunction &Fn) { +void PEI::insertCSRSpillsAndRestores(MachineFunction &Fn) { // Get callee saved register information. - MachineFrameInfo *FFI = Fn.getFrameInfo(); - const std::vector &CSI = FFI->getCalleeSavedInfo(); - - // Early exit if no caller saved registers are modified! - if (CSI.empty()) - return; + MachineFrameInfo *MFI = Fn.getFrameInfo(); + const std::vector &CSI = MFI->getCalleeSavedInfo(); - const MRegisterInfo *RegInfo = Fn.getTarget().getRegisterInfo(); + MFI->setCalleeSavedInfoValid(true); - // Now that we have a stack slot for each register to be saved, insert spill - // code into the entry block. - MachineBasicBlock *MBB = Fn.begin(); - MachineBasicBlock::iterator I = MBB->begin(); - for (unsigned i = 0, e = CSI.size(); i != e; ++i) { - // Insert the spill to the stack frame. - RegInfo->storeRegToStackSlot(*MBB, I, CSI[i].getReg(), CSI[i].getFrameIdx(), - CSI[i].getRegClass()); - } + // Early exit if no callee saved registers are modified! + if (CSI.empty()) + return; - // Add code to restore the callee-save registers in each exiting block. const TargetInstrInfo &TII = *Fn.getTarget().getInstrInfo(); - for (MachineFunction::iterator FI = Fn.begin(), E = Fn.end(); FI != E; ++FI) - // If last instruction is a return instruction, add an epilogue. - if (!FI->empty() && TII.isReturn(FI->back().getOpcode())) { - MBB = FI; + const TargetFrameLowering *TFI = Fn.getTarget().getFrameLowering(); + const TargetRegisterInfo *TRI = Fn.getTarget().getRegisterInfo(); + MachineBasicBlock::iterator I; + + if (! ShrinkWrapThisFunction) { + // Spill using target interface. + I = EntryBlock->begin(); + if (!TFI->spillCalleeSavedRegisters(*EntryBlock, I, CSI, TRI)) { + for (unsigned i = 0, e = CSI.size(); i != e; ++i) { + // Add the callee-saved register as live-in. + // It's killed at the spill. + EntryBlock->addLiveIn(CSI[i].getReg()); + + // Insert the spill to the stack frame. + unsigned Reg = CSI[i].getReg(); + const TargetRegisterClass *RC = TRI->getMinimalPhysRegClass(Reg); + TII.storeRegToStackSlot(*EntryBlock, I, Reg, true, + CSI[i].getFrameIdx(), RC, TRI); + } + } + + // Restore using target interface. + for (unsigned ri = 0, re = ReturnBlocks.size(); ri != re; ++ri) { + MachineBasicBlock* MBB = ReturnBlocks[ri]; I = MBB->end(); --I; // Skip over all terminator instructions, which are part of the return // sequence. MachineBasicBlock::iterator I2 = I; - while (I2 != MBB->begin() && TII.isTerminatorInstr((--I2)->getOpcode())) + while (I2 != MBB->begin() && (--I2)->getDesc().isTerminator()) I = I2; bool AtStart = I == MBB->begin(); MachineBasicBlock::iterator BeforeI = I; if (!AtStart) --BeforeI; - - // Restore all registers immediately before the return and any terminators - // that preceed it. - for (unsigned i = 0, e = CSI.size(); i != e; ++i) { - RegInfo->loadRegFromStackSlot(*MBB, I, CSI[i].getReg(), - CSI[i].getFrameIdx(), - CSI[i].getRegClass()); - assert(I != MBB->begin() && - "loadRegFromStackSlot didn't insert any code!"); - // Insert in reverse order. loadRegFromStackSlot can insert multiple - // instructions. - if (AtStart) - I = MBB->begin(); - else { - I = BeforeI; - ++I; + + // Restore all registers immediately before the return and any + // terminators that precede it. + if (!TFI->restoreCalleeSavedRegisters(*MBB, I, CSI, TRI)) { + for (unsigned i = 0, e = CSI.size(); i != e; ++i) { + unsigned Reg = CSI[i].getReg(); + const TargetRegisterClass *RC = TRI->getMinimalPhysRegClass(Reg); + TII.loadRegFromStackSlot(*MBB, I, Reg, + CSI[i].getFrameIdx(), + RC, TRI); + assert(I != MBB->begin() && + "loadRegFromStackSlot didn't insert any code!"); + // Insert in reverse order. loadRegFromStackSlot can insert + // multiple instructions. + if (AtStart) + I = MBB->begin(); + else { + I = BeforeI; + ++I; + } } } } + return; + } + + // Insert spills. + std::vector blockCSI; + for (CSRegBlockMap::iterator BI = CSRSave.begin(), + BE = CSRSave.end(); BI != BE; ++BI) { + MachineBasicBlock* MBB = BI->first; + CSRegSet save = BI->second; + + if (save.empty()) + continue; + + blockCSI.clear(); + for (CSRegSet::iterator RI = save.begin(), + RE = save.end(); RI != RE; ++RI) { + blockCSI.push_back(CSI[*RI]); + } + assert(blockCSI.size() > 0 && + "Could not collect callee saved register info"); + + I = MBB->begin(); + + // When shrink wrapping, use stack slot stores/loads. + for (unsigned i = 0, e = blockCSI.size(); i != e; ++i) { + // Add the callee-saved register as live-in. + // It's killed at the spill. + MBB->addLiveIn(blockCSI[i].getReg()); + + // Insert the spill to the stack frame. + unsigned Reg = blockCSI[i].getReg(); + const TargetRegisterClass *RC = TRI->getMinimalPhysRegClass(Reg); + TII.storeRegToStackSlot(*MBB, I, Reg, + true, + blockCSI[i].getFrameIdx(), + RC, TRI); + } + } + + for (CSRegBlockMap::iterator BI = CSRRestore.begin(), + BE = CSRRestore.end(); BI != BE; ++BI) { + MachineBasicBlock* MBB = BI->first; + CSRegSet restore = BI->second; + + if (restore.empty()) + continue; + + blockCSI.clear(); + for (CSRegSet::iterator RI = restore.begin(), + RE = restore.end(); RI != RE; ++RI) { + blockCSI.push_back(CSI[*RI]); + } + assert(blockCSI.size() > 0 && + "Could not find callee saved register info"); + + // If MBB is empty and needs restores, insert at the _beginning_. + if (MBB->empty()) { + I = MBB->begin(); + } else { + I = MBB->end(); + --I; + + // Skip over all terminator instructions, which are part of the + // return sequence. + if (! I->getDesc().isTerminator()) { + ++I; + } else { + MachineBasicBlock::iterator I2 = I; + while (I2 != MBB->begin() && (--I2)->getDesc().isTerminator()) + I = I2; + } + } + + bool AtStart = I == MBB->begin(); + MachineBasicBlock::iterator BeforeI = I; + if (!AtStart) + --BeforeI; + + // Restore all registers immediately before the return and any + // terminators that precede it. + for (unsigned i = 0, e = blockCSI.size(); i != e; ++i) { + unsigned Reg = blockCSI[i].getReg(); + const TargetRegisterClass *RC = TRI->getMinimalPhysRegClass(Reg); + TII.loadRegFromStackSlot(*MBB, I, Reg, + blockCSI[i].getFrameIdx(), + RC, TRI); + assert(I != MBB->begin() && + "loadRegFromStackSlot didn't insert any code!"); + // Insert in reverse order. loadRegFromStackSlot can insert + // multiple instructions. + if (AtStart) + I = MBB->begin(); + else { + I = BeforeI; + ++I; + } + } + } } +/// AdjustStackOffset - Helper function used to adjust the stack frame offset. +static inline void +AdjustStackOffset(MachineFrameInfo *MFI, int FrameIdx, + bool StackGrowsDown, int64_t &Offset, + unsigned &MaxAlign) { + // If the stack grows down, add the object size to find the lowest address. + if (StackGrowsDown) + Offset += MFI->getObjectSize(FrameIdx); + + unsigned Align = MFI->getObjectAlignment(FrameIdx); + + // If the alignment of this object is greater than that of the stack, then + // increase the stack alignment to match. + MaxAlign = std::max(MaxAlign, Align); + + // Adjust to alignment boundary. + Offset = (Offset + Align - 1) / Align * Align; + + if (StackGrowsDown) { + DEBUG(dbgs() << "alloc FI(" << FrameIdx << ") at SP[" << -Offset << "]\n"); + MFI->setObjectOffset(FrameIdx, -Offset); // Set the computed offset + } else { + DEBUG(dbgs() << "alloc FI(" << FrameIdx << ") at SP[" << Offset << "]\n"); + MFI->setObjectOffset(FrameIdx, Offset); + Offset += MFI->getObjectSize(FrameIdx); + } +} /// calculateFrameObjectOffsets - Calculate actual frame offsets for all of the /// abstract stack objects. /// void PEI::calculateFrameObjectOffsets(MachineFunction &Fn) { - const TargetFrameInfo &TFI = *Fn.getTarget().getFrameInfo(); + const TargetFrameLowering &TFI = *Fn.getTarget().getFrameLowering(); bool StackGrowsDown = - TFI.getStackGrowthDirection() == TargetFrameInfo::StackGrowsDown; + TFI.getStackGrowthDirection() == TargetFrameLowering::StackGrowsDown; // Loop over all of the stack objects, assigning sequential addresses... - MachineFrameInfo *FFI = Fn.getFrameInfo(); - - unsigned StackAlignment = TFI.getStackAlignment(); - unsigned MaxAlign = 0; + MachineFrameInfo *MFI = Fn.getFrameInfo(); // Start at the beginning of the local area. // The Offset is the distance from the stack top in the direction - // of stack growth -- so it's always positive. - int Offset = TFI.getOffsetOfLocalArea(); + // of stack growth -- so it's always nonnegative. + int LocalAreaOffset = TFI.getOffsetOfLocalArea(); if (StackGrowsDown) - Offset = -Offset; - assert(Offset >= 0 + LocalAreaOffset = -LocalAreaOffset; + assert(LocalAreaOffset >= 0 && "Local area offset should be in direction of stack growth"); + int64_t Offset = LocalAreaOffset; // If there are fixed sized objects that are preallocated in the local area, // non-fixed objects can't be allocated right at the start of local area. - // We currently don't support filling in holes in between fixed sized objects, - // so we adjust 'Offset' to point to the end of last fixed sized + // We currently don't support filling in holes in between fixed sized + // objects, so we adjust 'Offset' to point to the end of last fixed sized // preallocated object. - for (int i = FFI->getObjectIndexBegin(); i != 0; ++i) { - int FixedOff; + for (int i = MFI->getObjectIndexBegin(); i != 0; ++i) { + int64_t FixedOff; if (StackGrowsDown) { // The maximum distance from the stack pointer is at lower address of // the object -- which is given by offset. For down growing stack // the offset is negative, so we negate the offset to get the distance. - FixedOff = -FFI->getObjectOffset(i); + FixedOff = -MFI->getObjectOffset(i); } else { // The maximum distance from the start pointer is at the upper // address of the object. - FixedOff = FFI->getObjectOffset(i) + FFI->getObjectSize(i); + FixedOff = MFI->getObjectOffset(i) + MFI->getObjectSize(i); } if (FixedOff > Offset) Offset = FixedOff; } - for (unsigned i = 0, e = FFI->getObjectIndexEnd(); i != e; ++i) { - // If stack grows down, we need to add size of find the lowest - // address of the object. - if (StackGrowsDown) - Offset += FFI->getObjectSize(i); + // First assign frame offsets to stack objects that are used to spill + // callee saved registers. + if (StackGrowsDown) { + for (unsigned i = MinCSFrameIndex; i <= MaxCSFrameIndex; ++i) { + // If the stack grows down, we need to add the size to find the lowest + // address of the object. + Offset += MFI->getObjectSize(i); - unsigned Align = FFI->getObjectAlignment(i); - // If the alignment of this object is greater than that of the stack, then - // increase the stack alignment to match. - MaxAlign = std::max(MaxAlign, Align); - // Adjust to alignment boundary - Offset = (Offset+Align-1)/Align*Align; + unsigned Align = MFI->getObjectAlignment(i); + // Adjust to alignment boundary + Offset = (Offset+Align-1)/Align*Align; - if (StackGrowsDown) { - FFI->setObjectOffset(i, -Offset); // Set the computed offset - } else { - FFI->setObjectOffset(i, Offset); - Offset += FFI->getObjectSize(i); + MFI->setObjectOffset(i, -Offset); // Set the computed offset } + } else { + int MaxCSFI = MaxCSFrameIndex, MinCSFI = MinCSFrameIndex; + for (int i = MaxCSFI; i >= MinCSFI ; --i) { + unsigned Align = MFI->getObjectAlignment(i); + // Adjust to alignment boundary + Offset = (Offset+Align-1)/Align*Align; + + MFI->setObjectOffset(i, Offset); + Offset += MFI->getObjectSize(i); + } + } + + unsigned MaxAlign = MFI->getMaxAlignment(); + + // Make sure the special register scavenging spill slot is closest to the + // frame pointer if a frame pointer is required. + const TargetRegisterInfo *RegInfo = Fn.getTarget().getRegisterInfo(); + if (RS && TFI.hasFP(Fn) && RegInfo->useFPForScavengingIndex(Fn) && + !RegInfo->needsStackRealignment(Fn)) { + int SFI = RS->getScavengingFrameIndex(); + if (SFI >= 0) + AdjustStackOffset(MFI, SFI, StackGrowsDown, Offset, MaxAlign); } - // Align the final stack pointer offset, but only if there are calls in the - // function. This ensures that any calls to subroutines have their stack - // frames suitable aligned. - if (FFI->hasCalls()) - Offset = (Offset+StackAlignment-1)/StackAlignment*StackAlignment; + // FIXME: Once this is working, then enable flag will change to a target + // check for whether the frame is large enough to want to use virtual + // frame index registers. Functions which don't want/need this optimization + // will continue to use the existing code path. + if (MFI->getUseLocalStackAllocationBlock()) { + unsigned Align = MFI->getLocalFrameMaxAlign(); + + // Adjust to alignment boundary. + Offset = (Offset + Align - 1) / Align * Align; + + DEBUG(dbgs() << "Local frame base offset: " << Offset << "\n"); + + // Resolve offsets for objects in the local block. + for (unsigned i = 0, e = MFI->getLocalFrameObjectCount(); i != e; ++i) { + std::pair Entry = MFI->getLocalFrameObjectMap(i); + int64_t FIOffset = (StackGrowsDown ? -Offset : Offset) + Entry.second; + DEBUG(dbgs() << "alloc FI(" << Entry.first << ") at SP[" << + FIOffset << "]\n"); + MFI->setObjectOffset(Entry.first, FIOffset); + } + // Allocate the local block + Offset += MFI->getLocalFrameSize(); - // Set the final value of the stack pointer... - FFI->setStackSize(Offset+TFI.getOffsetOfLocalArea()); + MaxAlign = std::max(Align, MaxAlign); + } - // Remember the required stack alignment in case targets need it to perform - // dynamic stack alignment. - assert(FFI->getMaxAlignment() == MaxAlign && - "Stack alignment calculation broken!"); -} + // Make sure that the stack protector comes before the local variables on the + // stack. + SmallSet LargeStackObjs; + if (MFI->getStackProtectorIndex() >= 0) { + AdjustStackOffset(MFI, MFI->getStackProtectorIndex(), StackGrowsDown, + Offset, MaxAlign); + + // Assign large stack objects first. + for (unsigned i = 0, e = MFI->getObjectIndexEnd(); i != e; ++i) { + if (MFI->isObjectPreAllocated(i) && + MFI->getUseLocalStackAllocationBlock()) + continue; + if (i >= MinCSFrameIndex && i <= MaxCSFrameIndex) + continue; + if (RS && (int)i == RS->getScavengingFrameIndex()) + continue; + if (MFI->isDeadObjectIndex(i)) + continue; + if (MFI->getStackProtectorIndex() == (int)i) + continue; + if (!MFI->MayNeedStackProtector(i)) + continue; + + AdjustStackOffset(MFI, i, StackGrowsDown, Offset, MaxAlign); + LargeStackObjs.insert(i); + } + } + // Then assign frame offsets to stack objects that are not used to spill + // callee saved registers. + for (unsigned i = 0, e = MFI->getObjectIndexEnd(); i != e; ++i) { + if (MFI->isObjectPreAllocated(i) && + MFI->getUseLocalStackAllocationBlock()) + continue; + if (i >= MinCSFrameIndex && i <= MaxCSFrameIndex) + continue; + if (RS && (int)i == RS->getScavengingFrameIndex()) + continue; + if (MFI->isDeadObjectIndex(i)) + continue; + if (MFI->getStackProtectorIndex() == (int)i) + continue; + if (LargeStackObjs.count(i)) + continue; + + AdjustStackOffset(MFI, i, StackGrowsDown, Offset, MaxAlign); + } -/// insertPrologEpilogCode - Scan the function for modified caller saved -/// registers, insert spill code for these caller saved registers, then add + // Make sure the special register scavenging spill slot is closest to the + // stack pointer. + if (RS && (!TFI.hasFP(Fn) || RegInfo->needsStackRealignment(Fn) || + !RegInfo->useFPForScavengingIndex(Fn))) { + int SFI = RS->getScavengingFrameIndex(); + if (SFI >= 0) + AdjustStackOffset(MFI, SFI, StackGrowsDown, Offset, MaxAlign); + } + + if (!TFI.targetHandlesStackFrameRounding()) { + // If we have reserved argument space for call sites in the function + // immediately on entry to the current function, count it as part of the + // overall stack size. + if (MFI->adjustsStack() && TFI.hasReservedCallFrame(Fn)) + Offset += MFI->getMaxCallFrameSize(); + + // Round up the size to a multiple of the alignment. If the function has + // any calls or alloca's, align to the target's StackAlignment value to + // ensure that the callee's frame or the alloca data is suitably aligned; + // otherwise, for leaf functions, align to the TransientStackAlignment + // value. + unsigned StackAlign; + if (MFI->adjustsStack() || MFI->hasVarSizedObjects() || + (RegInfo->needsStackRealignment(Fn) && MFI->getObjectIndexEnd() != 0)) + StackAlign = TFI.getStackAlignment(); + else + StackAlign = TFI.getTransientStackAlignment(); + + // If the frame pointer is eliminated, all frame offsets will be relative to + // SP not FP. Align to MaxAlign so this works. + StackAlign = std::max(StackAlign, MaxAlign); + unsigned AlignMask = StackAlign - 1; + Offset = (Offset + AlignMask) & ~uint64_t(AlignMask); + } + + // Update frame info to pretend that this is part of the stack... + int64_t StackSize = Offset - LocalAreaOffset; + MFI->setStackSize(StackSize); + NumBytesStackSpace += StackSize; +} + +/// insertPrologEpilogCode - Scan the function for modified callee saved +/// registers, insert spill code for these callee saved registers, then add /// prolog and epilog code to the function. /// void PEI::insertPrologEpilogCode(MachineFunction &Fn) { + const TargetFrameLowering &TFI = *Fn.getTarget().getFrameLowering(); + // Add prologue to the function... - Fn.getTarget().getRegisterInfo()->emitPrologue(Fn); + TFI.emitPrologue(Fn); // Add epilogue to restore the callee-save registers in each exiting block - const TargetInstrInfo &TII = *Fn.getTarget().getInstrInfo(); for (MachineFunction::iterator I = Fn.begin(), E = Fn.end(); I != E; ++I) { // If last instruction is a return instruction, add an epilogue - if (!I->empty() && TII.isReturn(I->back().getOpcode())) - Fn.getTarget().getRegisterInfo()->emitEpilogue(Fn, *I); + if (!I->empty() && I->back().getDesc().isReturn()) + TFI.emitEpilogue(Fn, *I); } -} + // Emit additional code that is required to support segmented stacks, if + // we've been asked for it. This, when linked with a runtime with support + // for segmented stacks (libgcc is one), will result in allocating stack + // space in small chunks instead of one large contiguous block. + if (EnableSegmentedStacks) + TFI.adjustForSegmentedStacks(Fn); +} /// replaceFrameIndices - Replace all MO_FrameIndex operands with physical /// register references and actual offsets. @@ -361,15 +718,147 @@ void PEI::replaceFrameIndices(MachineFunction &Fn) { const TargetMachine &TM = Fn.getTarget(); assert(TM.getRegisterInfo() && "TM::getRegisterInfo() must be implemented!"); - const MRegisterInfo &MRI = *TM.getRegisterInfo(); + const TargetInstrInfo &TII = *Fn.getTarget().getInstrInfo(); + const TargetRegisterInfo &TRI = *TM.getRegisterInfo(); + const TargetFrameLowering *TFI = TM.getFrameLowering(); + bool StackGrowsDown = + TFI->getStackGrowthDirection() == TargetFrameLowering::StackGrowsDown; + int FrameSetupOpcode = TII.getCallFrameSetupOpcode(); + int FrameDestroyOpcode = TII.getCallFrameDestroyOpcode(); - for (MachineFunction::iterator BB = Fn.begin(), E = Fn.end(); BB != E; ++BB) - for (MachineBasicBlock::iterator I = BB->begin(); I != BB->end(); ++I) - for (unsigned i = 0, e = I->getNumOperands(); i != e; ++i) - if (I->getOperand(i).isFrameIndex()) { - // If this instruction has a FrameIndex operand, we need to use that - // target machine register info object to eliminate it. - MRI.eliminateFrameIndex(I); + for (MachineFunction::iterator BB = Fn.begin(), + E = Fn.end(); BB != E; ++BB) { +#ifndef NDEBUG + int SPAdjCount = 0; // frame setup / destroy count. +#endif + int SPAdj = 0; // SP offset due to call frame setup / destroy. + if (RS && !FrameIndexVirtualScavenging) RS->enterBasicBlock(BB); + + for (MachineBasicBlock::iterator I = BB->begin(); I != BB->end(); ) { + + if (I->getOpcode() == FrameSetupOpcode || + I->getOpcode() == FrameDestroyOpcode) { +#ifndef NDEBUG + // Track whether we see even pairs of them + SPAdjCount += I->getOpcode() == FrameSetupOpcode ? 1 : -1; +#endif + // Remember how much SP has been adjusted to create the call + // frame. + int Size = I->getOperand(0).getImm(); + + if ((!StackGrowsDown && I->getOpcode() == FrameSetupOpcode) || + (StackGrowsDown && I->getOpcode() == FrameDestroyOpcode)) + Size = -Size; + + SPAdj += Size; + + MachineBasicBlock::iterator PrevI = BB->end(); + if (I != BB->begin()) PrevI = prior(I); + TRI.eliminateCallFramePseudoInstr(Fn, *BB, I); + + // Visit the instructions created by eliminateCallFramePseudoInstr(). + if (PrevI == BB->end()) + I = BB->begin(); // The replaced instr was the first in the block. + else + I = llvm::next(PrevI); + continue; + } + + MachineInstr *MI = I; + bool DoIncr = true; + for (unsigned i = 0, e = MI->getNumOperands(); i != e; ++i) + if (MI->getOperand(i).isFI()) { + // Some instructions (e.g. inline asm instructions) can have + // multiple frame indices and/or cause eliminateFrameIndex + // to insert more than one instruction. We need the register + // scavenger to go through all of these instructions so that + // it can update its register information. We keep the + // iterator at the point before insertion so that we can + // revisit them in full. + bool AtBeginning = (I == BB->begin()); + if (!AtBeginning) --I; + + // If this instruction has a FrameIndex operand, we need to + // use that target machine register info object to eliminate + // it. + TRI.eliminateFrameIndex(MI, SPAdj, + FrameIndexVirtualScavenging ? NULL : RS); + + // Reset the iterator if we were at the beginning of the BB. + if (AtBeginning) { + I = BB->begin(); + DoIncr = false; + } + + MI = 0; break; } + + if (DoIncr && I != BB->end()) ++I; + + // Update register states. + if (RS && !FrameIndexVirtualScavenging && MI) RS->forward(MI); + } + + // If we have evenly matched pairs of frame setup / destroy instructions, + // make sure the adjustments come out to zero. If we don't have matched + // pairs, we can't be sure the missing bit isn't in another basic block + // due to a custom inserter playing tricks, so just asserting SPAdj==0 + // isn't sufficient. See tMOVCC on Thumb1, for example. + assert((SPAdjCount || SPAdj == 0) && + "Unbalanced call frame setup / destroy pairs?"); + } +} + +/// scavengeFrameVirtualRegs - Replace all frame index virtual registers +/// with physical registers. Use the register scavenger to find an +/// appropriate register to use. +void PEI::scavengeFrameVirtualRegs(MachineFunction &Fn) { + // Run through the instructions and find any virtual registers. + for (MachineFunction::iterator BB = Fn.begin(), + E = Fn.end(); BB != E; ++BB) { + RS->enterBasicBlock(BB); + + unsigned VirtReg = 0; + unsigned ScratchReg = 0; + int SPAdj = 0; + + // The instruction stream may change in the loop, so check BB->end() + // directly. + for (MachineBasicBlock::iterator I = BB->begin(); I != BB->end(); ) { + MachineInstr *MI = I; + for (unsigned i = 0, e = MI->getNumOperands(); i != e; ++i) { + if (MI->getOperand(i).isReg()) { + MachineOperand &MO = MI->getOperand(i); + unsigned Reg = MO.getReg(); + if (Reg == 0) + continue; + if (!TargetRegisterInfo::isVirtualRegister(Reg)) + continue; + + ++NumVirtualFrameRegs; + + // Have we already allocated a scratch register for this virtual? + if (Reg != VirtReg) { + // When we first encounter a new virtual register, it + // must be a definition. + assert(MI->getOperand(i).isDef() && + "frame index virtual missing def!"); + // Scavenge a new scratch register + VirtReg = Reg; + const TargetRegisterClass *RC = Fn.getRegInfo().getRegClass(Reg); + ScratchReg = RS->scavengeRegister(RC, I, SPAdj); + ++NumScavengedRegs; + } + // Replace this reference to the virtual register with the + // scratch register. + assert (ScratchReg && "Missing scratch register!"); + MI->getOperand(i).setReg(ScratchReg); + + } + } + RS->forward(I); + ++I; + } + } }