X-Git-Url: http://demsky.eecs.uci.edu/git/?a=blobdiff_plain;f=lib%2FCodeGen%2FPrologEpilogInserter.cpp;h=4b1e979a62cf6e7caa3f7d3144b3436cf6079d8e;hb=021e3b6444e9179751002db7b20de96455e03171;hp=226e1ba2833771fe1c694dd6908a96c798111c0b;hpb=94188d4e67cf1c570ad87dbabf198931033d628e;p=oota-llvm.git diff --git a/lib/CodeGen/PrologEpilogInserter.cpp b/lib/CodeGen/PrologEpilogInserter.cpp index 226e1ba2833..4b1e979a62c 100644 --- a/lib/CodeGen/PrologEpilogInserter.cpp +++ b/lib/CodeGen/PrologEpilogInserter.cpp @@ -19,33 +19,47 @@ // //===----------------------------------------------------------------------===// +#define DEBUG_TYPE "pei" #include "PrologEpilogInserter.h" +#include "llvm/ADT/IndexedMap.h" +#include "llvm/ADT/STLExtras.h" +#include "llvm/ADT/SmallSet.h" +#include "llvm/ADT/Statistic.h" #include "llvm/CodeGen/MachineDominators.h" -#include "llvm/CodeGen/MachineLoopInfo.h" -#include "llvm/CodeGen/MachineInstr.h" #include "llvm/CodeGen/MachineFrameInfo.h" -#include "llvm/CodeGen/MachineModuleInfo.h" +#include "llvm/CodeGen/MachineInstr.h" +#include "llvm/CodeGen/MachineLoopInfo.h" #include "llvm/CodeGen/MachineRegisterInfo.h" #include "llvm/CodeGen/RegisterScavenging.h" +#include "llvm/InlineAsm.h" +#include "llvm/Support/CommandLine.h" +#include "llvm/Support/Compiler.h" +#include "llvm/Support/Debug.h" +#include "llvm/Target/TargetFrameLowering.h" +#include "llvm/Target/TargetInstrInfo.h" #include "llvm/Target/TargetMachine.h" +#include "llvm/Target/TargetOptions.h" #include "llvm/Target/TargetRegisterInfo.h" -#include "llvm/Target/TargetFrameInfo.h" -#include "llvm/Target/TargetInstrInfo.h" -#include "llvm/Support/Compiler.h" -#include "llvm/ADT/STLExtras.h" #include using namespace llvm; char PEI::ID = 0; - -static RegisterPass -X("prologepilog", "Prologue/Epilogue Insertion"); - -/// createPrologEpilogCodeInserter - This function returns a pass that inserts -/// prolog and epilog code, and eliminates abstract frame references. -/// -FunctionPass *llvm::createPrologEpilogCodeInserter() { return new PEI(); } +char &llvm::PrologEpilogCodeInserterID = PEI::ID; + +INITIALIZE_PASS_BEGIN(PEI, "prologepilog", + "Prologue/Epilogue Insertion", false, false) +INITIALIZE_PASS_DEPENDENCY(MachineLoopInfo) +INITIALIZE_PASS_DEPENDENCY(MachineDominatorTree) +INITIALIZE_PASS_DEPENDENCY(TargetPassConfig) +INITIALIZE_PASS_END(PEI, "prologepilog", + "Prologue/Epilogue Insertion & Frame Finalization", + 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"); /// runOnMachineFunction - Insert prolog/epilog code and replace abstract /// frame indexes with appropriate references. @@ -53,50 +67,51 @@ FunctionPass *llvm::createPrologEpilogCodeInserter() { return new PEI(); } bool PEI::runOnMachineFunction(MachineFunction &Fn) { const Function* F = Fn.getFunction(); const TargetRegisterInfo *TRI = Fn.getTarget().getRegisterInfo(); - RS = TRI->requiresRegisterScavenging(Fn) ? new RegScavenger() : NULL; + const TargetFrameLowering *TFI = Fn.getTarget().getFrameLowering(); - // Get MachineModuleInfo so that we can track the construction of the - // frame. - if (MachineModuleInfo *MMI = getAnalysisIfAvailable()) - Fn.getFrameInfo()->setMachineModuleInfo(MMI); + assert(!Fn.getRegInfo().getNumVirtRegs() && "Regalloc must assign all vregs"); - // Calculate the MaxCallFrameSize and HasCalls variables for the function's - // frame information. Also eliminates call frame pseudo instructions. + 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. - TRI->processFunctionBeforeCalleeSavedScan(Fn, RS); + 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 + // - With shrink wrapping, place spills and restores to tightly // enclose regions in the Machine CFG of the function where - // they are used. Without shrink wrapping - // - default (no shrink wrapping), place all spills in the + // 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)) + if (!F->getFnAttributes().hasAttribute(Attribute::Naked)) insertCSRSpillsAndRestores(Fn); // Allow the target machine to make final modifications to the function // before the frame layout is finalized. - TRI->processFunctionBeforeFrameFinalized(Fn); + 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 HasCalls + // 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)) + if (!F->getFnAttributes().hasAttribute(Attribute::Naked)) insertPrologEpilogCode(Fn); // Replace all MO_FrameIndex operands with physical register references @@ -104,36 +119,35 @@ bool PEI::runOnMachineFunction(MachineFunction &Fn) { // replaceFrameIndices(Fn); + // 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); + + // Clear any vregs created by virtual scavenging. + Fn.getRegInfo().clearVirtRegs(); + delete RS; clearAllSets(); return true; } -#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 HasCalls +/// 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 = RegInfo->getCallFrameSetupOpcode(); - int FrameDestroyOpcode = RegInfo->getCallFrameDestroyOpcode(); + int FrameSetupOpcode = TII.getCallFrameSetupOpcode(); + int FrameDestroyOpcode = TII.getCallFrameDestroyOpcode(); // Early exit for targets which have no call frame setup/destroy pseudo // instructions. @@ -149,26 +163,27 @@ void PEI::calculateCallsInformation(MachineFunction &Fn) { " instructions should have a single immediate argument!"); unsigned Size = I->getOperand(0).getImm(); if (Size > MaxCallFrameSize) MaxCallFrameSize = Size; - HasCalls = true; + AdjustsStack = true; FrameSDOps.push_back(I); - } else if (I->getOpcode() == TargetInstrInfo::INLINEASM) { - // An InlineAsm might be a call; assume it is to get the stack frame - // aligned correctly for calls. - HasCalls = true; + } 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 - // there is no dynamic allocation (therefore referencing frame slots off - // sp), leave the pseudo ops alone. We'll eliminate them later. - if (RegInfo->hasReservedCallFrame(Fn) || RegInfo->hasFP(Fn)) + // 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); } } @@ -178,11 +193,11 @@ void PEI::calculateCallsInformation(MachineFunction &Fn) { /// registers. void PEI::calculateCalleeSavedRegisters(MachineFunction &Fn) { const TargetRegisterInfo *RegInfo = Fn.getTarget().getRegisterInfo(); - const TargetFrameInfo *TFI = Fn.getTarget().getFrameInfo(); - MachineFrameInfo *FFI = Fn.getFrameInfo(); + const TargetFrameLowering *TFI = Fn.getTarget().getFrameLowering(); + MachineFrameInfo *MFI = Fn.getFrameInfo(); // Get the callee saved register list... - const unsigned *CSRegs = RegInfo->getCalleeSavedRegs(&Fn); + const uint16_t *CSRegs = RegInfo->getCalleeSavedRegs(&Fn); // These are used to keep track the callee-save area. Initialize them. MinCSFrameIndex = INT_MAX; @@ -192,25 +207,16 @@ void PEI::calculateCalleeSavedRegisters(MachineFunction &Fn) { if (CSRegs == 0 || CSRegs[0] == 0) return; - // Figure out which *callee saved* registers are modified by the current - // function, thus needing to be saved and restored in the prolog/epilog. - const TargetRegisterClass * const *CSRegClasses = - RegInfo->getCalleeSavedRegClasses(&Fn); + // In Naked functions we aren't going to save any registers. + if (Fn.getFunction()->getFnAttributes().hasAttribute(Attribute::Naked)) + return; std::vector CSI; for (unsigned i = 0; CSRegs[i]; ++i) { unsigned Reg = CSRegs[i]; if (Fn.getRegInfo().isPhysRegUsed(Reg)) { // If the reg is modified, save it! - CSI.push_back(CalleeSavedInfo(Reg, CSRegClasses[i])); - } else { - for (const unsigned *AliasSet = RegInfo->getAliasSet(Reg); - *AliasSet; ++AliasSet) { // Check alias registers too. - if (Fn.getRegInfo().isPhysRegUsed(*AliasSet)) { - CSI.push_back(CalleeSavedInfo(Reg, CSRegClasses[i])); - break; - } - } + CSI.push_back(CalleeSavedInfo(Reg)); } } @@ -218,7 +224,7 @@ void PEI::calculateCalleeSavedRegisters(MachineFunction &Fn) { return; // Early exit if no callee saved registers are modified! unsigned NumFixedSpillSlots; - const std::pair *FixedSpillSlots = + const TargetFrameLowering::SpillSlot *FixedSpillSlots = TFI->getCalleeSavedSpillSlots(NumFixedSpillSlots); // Now that we know which registers need to be saved and restored, allocate @@ -226,7 +232,7 @@ void PEI::calculateCalleeSavedRegisters(MachineFunction &Fn) { for (std::vector::iterator I = CSI.begin(), E = CSI.end(); I != E; ++I) { unsigned Reg = I->getReg(); - const TargetRegisterClass *RC = I->getRegClass(); + const TargetRegisterClass *RC = RegInfo->getMinimalPhysRegClass(Reg); int FrameIdx; if (RegInfo->hasReservedSpillSlot(Fn, Reg, FrameIdx)) { @@ -236,9 +242,9 @@ void PEI::calculateCalleeSavedRegisters(MachineFunction &Fn) { // 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; if (FixedSlot == FixedSpillSlots + NumFixedSpillSlots) { @@ -250,18 +256,18 @@ void PEI::calculateCalleeSavedRegisters(MachineFunction &Fn) { // the TargetRegisterClass if the stack alignment is smaller. Use the // min. Align = std::min(Align, StackAlign); - FrameIdx = FFI->CreateStackObject(RC->getSize(), Align); + 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); } I->setFrameIdx(FrameIdx); } - FFI->setCalleeSavedInfo(CSI); + MFI->setCalleeSavedInfo(CSI); } /// insertCSRSpillsAndRestores - Insert spill and restore code for @@ -269,30 +275,34 @@ void PEI::calculateCalleeSavedRegisters(MachineFunction &Fn) { /// void PEI::insertCSRSpillsAndRestores(MachineFunction &Fn) { // Get callee saved register information. - MachineFrameInfo *FFI = Fn.getFrameInfo(); - const std::vector &CSI = FFI->getCalleeSavedInfo(); + MachineFrameInfo *MFI = Fn.getFrameInfo(); + const std::vector &CSI = MFI->getCalleeSavedInfo(); - FFI->setCalleeSavedInfoValid(true); + MFI->setCalleeSavedInfoValid(true); // Early exit if no callee saved registers are modified! if (CSI.empty()) return; const TargetInstrInfo &TII = *Fn.getTarget().getInstrInfo(); + const TargetFrameLowering *TFI = Fn.getTarget().getFrameLowering(); + const TargetRegisterInfo *TRI = Fn.getTarget().getRegisterInfo(); MachineBasicBlock::iterator I; - if (! ShrinkWrapThisFunction) { + if (!ShrinkWrapThisFunction) { // Spill using target interface. I = EntryBlock->begin(); - if (!TII.spillCalleeSavedRegisters(*EntryBlock, I, CSI)) { + 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. - TII.storeRegToStackSlot(*EntryBlock, I, CSI[i].getReg(), true, - CSI[i].getFrameIdx(), CSI[i].getRegClass()); + unsigned Reg = CSI[i].getReg(); + const TargetRegisterClass *RC = TRI->getMinimalPhysRegClass(Reg); + TII.storeRegToStackSlot(*EntryBlock, I, Reg, true, + CSI[i].getFrameIdx(), RC, TRI); } } @@ -304,7 +314,7 @@ void PEI::insertCSRSpillsAndRestores(MachineFunction &Fn) { // Skip over all terminator instructions, which are part of the return // sequence. MachineBasicBlock::iterator I2 = I; - while (I2 != MBB->begin() && (--I2)->getDesc().isTerminator()) + while (I2 != MBB->begin() && (--I2)->isTerminator()) I = I2; bool AtStart = I == MBB->begin(); @@ -313,12 +323,14 @@ void PEI::insertCSRSpillsAndRestores(MachineFunction &Fn) { --BeforeI; // Restore all registers immediately before the return and any - // terminators that preceed it. - if (!TII.restoreCalleeSavedRegisters(*MBB, I, CSI)) { + // terminators that precede it. + if (!TFI->restoreCalleeSavedRegisters(*MBB, I, CSI, TRI)) { for (unsigned i = 0, e = CSI.size(); i != e; ++i) { - TII.loadRegFromStackSlot(*MBB, I, CSI[i].getReg(), + unsigned Reg = CSI[i].getReg(); + const TargetRegisterClass *RC = TRI->getMinimalPhysRegClass(Reg); + TII.loadRegFromStackSlot(*MBB, I, Reg, CSI[i].getFrameIdx(), - CSI[i].getRegClass()); + RC, TRI); assert(I != MBB->begin() && "loadRegFromStackSlot didn't insert any code!"); // Insert in reverse order. loadRegFromStackSlot can insert @@ -362,10 +374,12 @@ void PEI::insertCSRSpillsAndRestores(MachineFunction &Fn) { MBB->addLiveIn(blockCSI[i].getReg()); // Insert the spill to the stack frame. - TII.storeRegToStackSlot(*MBB, I, blockCSI[i].getReg(), + unsigned Reg = blockCSI[i].getReg(); + const TargetRegisterClass *RC = TRI->getMinimalPhysRegClass(Reg); + TII.storeRegToStackSlot(*MBB, I, Reg, true, blockCSI[i].getFrameIdx(), - blockCSI[i].getRegClass()); + RC, TRI); } } @@ -394,11 +408,11 @@ void PEI::insertCSRSpillsAndRestores(MachineFunction &Fn) { // Skip over all terminator instructions, which are part of the // return sequence. - if (! I->getDesc().isTerminator()) { + if (! I->isTerminator()) { ++I; } else { MachineBasicBlock::iterator I2 = I; - while (I2 != MBB->begin() && (--I2)->getDesc().isTerminator()) + while (I2 != MBB->begin() && (--I2)->isTerminator()) I = I2; } } @@ -409,11 +423,13 @@ void PEI::insertCSRSpillsAndRestores(MachineFunction &Fn) { --BeforeI; // Restore all registers immediately before the return and any - // terminators that preceed it. + // terminators that precede it. for (unsigned i = 0, e = blockCSI.size(); i != e; ++i) { - TII.loadRegFromStackSlot(*MBB, I, blockCSI[i].getReg(), + unsigned Reg = blockCSI[i].getReg(); + const TargetRegisterClass *RC = TRI->getMinimalPhysRegClass(Reg); + TII.loadRegFromStackSlot(*MBB, I, Reg, blockCSI[i].getFrameIdx(), - blockCSI[i].getRegClass()); + RC, TRI); assert(I != MBB->begin() && "loadRegFromStackSlot didn't insert any code!"); // Insert in reverse order. loadRegFromStackSlot can insert @@ -430,14 +446,14 @@ void PEI::insertCSRSpillsAndRestores(MachineFunction &Fn) { /// AdjustStackOffset - Helper function used to adjust the stack frame offset. static inline void -AdjustStackOffset(MachineFrameInfo *FFI, int FrameIdx, +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 += FFI->getObjectSize(FrameIdx); + Offset += MFI->getObjectSize(FrameIdx); - unsigned Align = FFI->getObjectAlignment(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. @@ -447,10 +463,12 @@ AdjustStackOffset(MachineFrameInfo *FFI, int FrameIdx, Offset = (Offset + Align - 1) / Align * Align; if (StackGrowsDown) { - FFI->setObjectOffset(FrameIdx, -Offset); // Set the computed offset + DEBUG(dbgs() << "alloc FI(" << FrameIdx << ") at SP[" << -Offset << "]\n"); + MFI->setObjectOffset(FrameIdx, -Offset); // Set the computed offset } else { - FFI->setObjectOffset(FrameIdx, Offset); - Offset += FFI->getObjectSize(FrameIdx); + DEBUG(dbgs() << "alloc FI(" << FrameIdx << ") at SP[" << Offset << "]\n"); + MFI->setObjectOffset(FrameIdx, Offset); + Offset += MFI->getObjectSize(FrameIdx); } } @@ -458,41 +476,40 @@ AdjustStackOffset(MachineFrameInfo *FFI, int FrameIdx, /// 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 MaxAlign = FFI->getMaxAlignment(); + 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 nonnegative. - int64_t Offset = TFI.getOffsetOfLocalArea(); + 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 // preallocated object. - for (int i = FFI->getObjectIndexBegin(); i != 0; ++i) { + 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; } @@ -501,119 +518,179 @@ void PEI::calculateFrameObjectOffsets(MachineFunction &Fn) { // callee saved registers. if (StackGrowsDown) { for (unsigned i = MinCSFrameIndex; i <= MaxCSFrameIndex; ++i) { - // If stack grows down, we need to add size of find the lowest + // If the stack grows down, we need to add the size to find the lowest // address of the object. - Offset += FFI->getObjectSize(i); + 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); + unsigned Align = MFI->getObjectAlignment(i); // Adjust to alignment boundary Offset = (Offset+Align-1)/Align*Align; - FFI->setObjectOffset(i, -Offset); // Set the computed offset + MFI->setObjectOffset(i, -Offset); // Set the computed offset } } else { int MaxCSFI = MaxCSFrameIndex, MinCSFI = MinCSFrameIndex; for (int i = MaxCSFI; i >= MinCSFI ; --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); + unsigned Align = MFI->getObjectAlignment(i); // Adjust to alignment boundary Offset = (Offset+Align-1)/Align*Align; - FFI->setObjectOffset(i, Offset); - Offset += FFI->getObjectSize(i); + 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 && RegInfo->hasFP(Fn)) { + if (RS && TFI.hasFP(Fn) && RegInfo->useFPForScavengingIndex(Fn) && + !RegInfo->needsStackRealignment(Fn)) { int SFI = RS->getScavengingFrameIndex(); if (SFI >= 0) - AdjustStackOffset(FFI, SFI, StackGrowsDown, Offset, MaxAlign); + AdjustStackOffset(MFI, SFI, StackGrowsDown, Offset, MaxAlign); + } + + // 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(); + + MaxAlign = std::max(Align, MaxAlign); } // Make sure that the stack protector comes before the local variables on the // stack. - if (FFI->getStackProtectorIndex() >= 0) - AdjustStackOffset(FFI, FFI->getStackProtectorIndex(), StackGrowsDown, + 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 = FFI->getObjectIndexEnd(); i != e; ++i) { + 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 (FFI->isDeadObjectIndex(i)) + if (MFI->isDeadObjectIndex(i)) + continue; + if (MFI->getStackProtectorIndex() == (int)i) continue; - if (FFI->getStackProtectorIndex() == (int)i) + if (LargeStackObjs.count(i)) continue; - AdjustStackOffset(FFI, i, StackGrowsDown, Offset, MaxAlign); + AdjustStackOffset(MFI, i, StackGrowsDown, Offset, MaxAlign); } // Make sure the special register scavenging spill slot is closest to the // stack pointer. - if (RS && !RegInfo->hasFP(Fn)) { + if (RS && (!TFI.hasFP(Fn) || RegInfo->needsStackRealignment(Fn) || + !RegInfo->useFPForScavengingIndex(Fn))) { int SFI = RS->getScavengingFrameIndex(); if (SFI >= 0) - AdjustStackOffset(FFI, SFI, StackGrowsDown, Offset, MaxAlign); + AdjustStackOffset(MFI, SFI, StackGrowsDown, Offset, MaxAlign); } - // Round up the size to a multiple of the alignment, but only if there are - // calls or alloca's in the function. This ensures that any calls to - // subroutines have their stack frames suitably aligned. - // Also do this if we need runtime alignment of the stack. In this case - // offsets will be relative to SP not FP; round up the stack size so this - // works. - if (!RegInfo->targetHandlesStackFrameRounding() && - (FFI->hasCalls() || FFI->hasVarSizedObjects() || - (RegInfo->needsStackRealignment(Fn) && - FFI->getObjectIndexEnd() != 0))) { + 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 (RegInfo->hasReservedCallFrame(Fn)) - Offset += FFI->getMaxCallFrameSize(); - - unsigned AlignMask = std::max(TFI.getStackAlignment(), MaxAlign) - 1; + 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... - FFI->setStackSize(Offset+TFI.getOffsetOfLocalArea()); - - // Remember the required stack alignment in case targets need it to perform - // dynamic stack alignment. - FFI->setMaxAlignment(MaxAlign); + 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 TargetRegisterInfo *TRI = Fn.getTarget().getRegisterInfo(); + const TargetFrameLowering &TFI = *Fn.getTarget().getFrameLowering(); // Add prologue to the function... - TRI->emitPrologue(Fn); + TFI.emitPrologue(Fn); // Add epilogue to restore the callee-save registers in each exiting block 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() && I->back().getDesc().isReturn()) - TRI->emitEpilogue(Fn, *I); + if (!I->empty() && I->back().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 (Fn.getTarget().Options.EnableSegmentedStacks) + TFI.adjustForSegmentedStacks(Fn); +} /// replaceFrameIndices - Replace all MO_FrameIndex operands with physical /// register references and actual offsets. @@ -623,22 +700,30 @@ void PEI::replaceFrameIndices(MachineFunction &Fn) { const TargetMachine &TM = Fn.getTarget(); assert(TM.getRegisterInfo() && "TM::getRegisterInfo() must be implemented!"); + const TargetInstrInfo &TII = *Fn.getTarget().getInstrInfo(); const TargetRegisterInfo &TRI = *TM.getRegisterInfo(); - const TargetFrameInfo *TFI = TM.getFrameInfo(); + const TargetFrameLowering *TFI = TM.getFrameLowering(); bool StackGrowsDown = - TFI->getStackGrowthDirection() == TargetFrameInfo::StackGrowsDown; - int FrameSetupOpcode = TRI.getCallFrameSetupOpcode(); - int FrameDestroyOpcode = TRI.getCallFrameDestroyOpcode(); + TFI->getStackGrowthDirection() == TargetFrameLowering::StackGrowsDown; + int FrameSetupOpcode = TII.getCallFrameSetupOpcode(); + int FrameDestroyOpcode = TII.getCallFrameDestroyOpcode(); 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) RS->enterBasicBlock(BB); + 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(); @@ -657,7 +742,7 @@ void PEI::replaceFrameIndices(MachineFunction &Fn) { if (PrevI == BB->end()) I = BB->begin(); // The replaced instr was the first in the block. else - I = next(PrevI); + I = llvm::next(PrevI); continue; } @@ -678,8 +763,8 @@ void PEI::replaceFrameIndices(MachineFunction &Fn) { // If this instruction has a FrameIndex operand, we need to // use that target machine register info object to eliminate // it. - - TRI.eliminateFrameIndex(MI, SPAdj, RS); + TRI.eliminateFrameIndex(MI, SPAdj, + FrameIndexVirtualScavenging ? NULL : RS); // Reset the iterator if we were at the beginning of the BB. if (AtBeginning) { @@ -694,10 +779,72 @@ void PEI::replaceFrameIndices(MachineFunction &Fn) { if (DoIncr && I != BB->end()) ++I; // Update register states. - if (RS && MI) RS->forward(MI); + if (RS && !FrameIndexVirtualScavenging && MI) RS->forward(MI); } - assert(SPAdj == 0 && "Unbalanced call frame setup / destroy pairs?"); + // 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. +/// +/// FIXME: Iterating over the instruction stream is unnecessary. We can simply +/// iterate over the vreg use list, which at this point only contains machine +/// operands for which eliminateFrameIndex need a new scratch reg. +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; + } + } +}