X-Git-Url: http://demsky.eecs.uci.edu/git/?a=blobdiff_plain;f=lib%2FTarget%2FPowerPC%2FPPCFrameLowering.cpp;h=dc87a6ca366e45670ddfc692e656a28c8b9b6c60;hb=06e8dbef25fde1bc8354b1f3c3caa65ef5bd05df;hp=9ec10f62f5953d3e91e86e3c8224c298a85f139a;hpb=fb6fe0aea2d1adde6d5e86f43797b5795ff2dc36;p=oota-llvm.git diff --git a/lib/Target/PowerPC/PPCFrameLowering.cpp b/lib/Target/PowerPC/PPCFrameLowering.cpp index 9ec10f62f59..dc87a6ca366 100644 --- a/lib/Target/PowerPC/PPCFrameLowering.cpp +++ b/lib/Target/PowerPC/PPCFrameLowering.cpp @@ -15,6 +15,7 @@ #include "PPCInstrBuilder.h" #include "PPCInstrInfo.h" #include "PPCMachineFunctionInfo.h" +#include "PPCSubtarget.h" #include "llvm/CodeGen/MachineFrameInfo.h" #include "llvm/CodeGen/MachineFunction.h" #include "llvm/CodeGen/MachineInstrBuilder.h" @@ -26,17 +27,6 @@ using namespace llvm; -// FIXME This disables some code that aligns the stack to a boundary bigger than -// the default (16 bytes on Darwin) when there is a stack local of greater -// alignment. This does not currently work, because the delta between old and -// new stack pointers is added to offsets that reference incoming parameters -// after the prolog is generated, and the code that does that doesn't handle a -// variable delta. You don't want to do that anyway; a better approach is to -// reserve another register that retains to the incoming stack pointer, and -// reference parameters relative to that. -#define ALIGN_STACK 0 - - /// VRRegNo - Map from a numbered VR register to its enum value. /// static const uint16_t VRRegNo[] = { @@ -46,6 +36,167 @@ static const uint16_t VRRegNo[] = { PPC::V24, PPC::V25, PPC::V26, PPC::V27, PPC::V28, PPC::V29, PPC::V30, PPC::V31 }; +PPCFrameLowering::PPCFrameLowering(const PPCSubtarget &STI) + : TargetFrameLowering(TargetFrameLowering::StackGrowsDown, + (STI.hasQPX() || STI.isBGQ()) ? 32 : 16, 0), + Subtarget(STI) {} + +// With the SVR4 ABI, callee-saved registers have fixed offsets on the stack. +const PPCFrameLowering::SpillSlot *PPCFrameLowering::getCalleeSavedSpillSlots( + unsigned &NumEntries) const { + if (Subtarget.isDarwinABI()) { + NumEntries = 1; + if (Subtarget.isPPC64()) { + static const SpillSlot darwin64Offsets = {PPC::X31, -8}; + return &darwin64Offsets; + } else { + static const SpillSlot darwinOffsets = {PPC::R31, -4}; + return &darwinOffsets; + } + } + + // Early exit if not using the SVR4 ABI. + if (!Subtarget.isSVR4ABI()) { + NumEntries = 0; + return nullptr; + } + + // Note that the offsets here overlap, but this is fixed up in + // processFunctionBeforeFrameFinalized. + + static const SpillSlot Offsets[] = { + // Floating-point register save area offsets. + {PPC::F31, -8}, + {PPC::F30, -16}, + {PPC::F29, -24}, + {PPC::F28, -32}, + {PPC::F27, -40}, + {PPC::F26, -48}, + {PPC::F25, -56}, + {PPC::F24, -64}, + {PPC::F23, -72}, + {PPC::F22, -80}, + {PPC::F21, -88}, + {PPC::F20, -96}, + {PPC::F19, -104}, + {PPC::F18, -112}, + {PPC::F17, -120}, + {PPC::F16, -128}, + {PPC::F15, -136}, + {PPC::F14, -144}, + + // General register save area offsets. + {PPC::R31, -4}, + {PPC::R30, -8}, + {PPC::R29, -12}, + {PPC::R28, -16}, + {PPC::R27, -20}, + {PPC::R26, -24}, + {PPC::R25, -28}, + {PPC::R24, -32}, + {PPC::R23, -36}, + {PPC::R22, -40}, + {PPC::R21, -44}, + {PPC::R20, -48}, + {PPC::R19, -52}, + {PPC::R18, -56}, + {PPC::R17, -60}, + {PPC::R16, -64}, + {PPC::R15, -68}, + {PPC::R14, -72}, + + // CR save area offset. We map each of the nonvolatile CR fields + // to the slot for CR2, which is the first of the nonvolatile CR + // fields to be assigned, so that we only allocate one save slot. + // See PPCRegisterInfo::hasReservedSpillSlot() for more information. + {PPC::CR2, -4}, + + // VRSAVE save area offset. + {PPC::VRSAVE, -4}, + + // Vector register save area + {PPC::V31, -16}, + {PPC::V30, -32}, + {PPC::V29, -48}, + {PPC::V28, -64}, + {PPC::V27, -80}, + {PPC::V26, -96}, + {PPC::V25, -112}, + {PPC::V24, -128}, + {PPC::V23, -144}, + {PPC::V22, -160}, + {PPC::V21, -176}, + {PPC::V20, -192}}; + + static const SpillSlot Offsets64[] = { + // Floating-point register save area offsets. + {PPC::F31, -8}, + {PPC::F30, -16}, + {PPC::F29, -24}, + {PPC::F28, -32}, + {PPC::F27, -40}, + {PPC::F26, -48}, + {PPC::F25, -56}, + {PPC::F24, -64}, + {PPC::F23, -72}, + {PPC::F22, -80}, + {PPC::F21, -88}, + {PPC::F20, -96}, + {PPC::F19, -104}, + {PPC::F18, -112}, + {PPC::F17, -120}, + {PPC::F16, -128}, + {PPC::F15, -136}, + {PPC::F14, -144}, + + // General register save area offsets. + {PPC::X31, -8}, + {PPC::X30, -16}, + {PPC::X29, -24}, + {PPC::X28, -32}, + {PPC::X27, -40}, + {PPC::X26, -48}, + {PPC::X25, -56}, + {PPC::X24, -64}, + {PPC::X23, -72}, + {PPC::X22, -80}, + {PPC::X21, -88}, + {PPC::X20, -96}, + {PPC::X19, -104}, + {PPC::X18, -112}, + {PPC::X17, -120}, + {PPC::X16, -128}, + {PPC::X15, -136}, + {PPC::X14, -144}, + + // VRSAVE save area offset. + {PPC::VRSAVE, -4}, + + // Vector register save area + {PPC::V31, -16}, + {PPC::V30, -32}, + {PPC::V29, -48}, + {PPC::V28, -64}, + {PPC::V27, -80}, + {PPC::V26, -96}, + {PPC::V25, -112}, + {PPC::V24, -128}, + {PPC::V23, -144}, + {PPC::V22, -160}, + {PPC::V21, -176}, + {PPC::V20, -192}}; + + if (Subtarget.isPPC64()) { + NumEntries = array_lengthof(Offsets64); + + return Offsets64; + } else { + NumEntries = array_lengthof(Offsets); + + return Offsets; + } +} + /// RemoveVRSaveCode - We have found that this function does not need any code /// to manipulate the VRSAVE register, even though it uses vector registers. /// This can happen when the only registers used are known to be live in or out @@ -103,7 +254,7 @@ static void RemoveVRSaveCode(MachineInstr *MI) { // transform this into the appropriate ORI instruction. static void HandleVRSaveUpdate(MachineInstr *MI, const TargetInstrInfo &TII) { MachineFunction *MF = MI->getParent()->getParent(); - const TargetRegisterInfo *TRI = MF->getTarget().getRegisterInfo(); + const TargetRegisterInfo *TRI = MF->getSubtarget().getRegisterInfo(); DebugLoc dl = MI->getDebugLoc(); unsigned UsedRegMask = 0; @@ -215,11 +366,13 @@ unsigned PPCFrameLowering::determineFrameLayout(MachineFunction &MF, unsigned FrameSize = UseEstimate ? MFI->estimateStackSize(MF) : MFI->getStackSize(); - // Get the alignments provided by the target, and the maximum alignment - // (if any) of the fixed frame objects. - unsigned MaxAlign = MFI->getMaxAlignment(); - unsigned TargetAlign = getStackAlignment(); - unsigned AlignMask = TargetAlign - 1; // + // Get stack alignments. The frame must be aligned to the greatest of these: + unsigned TargetAlign = getStackAlignment(); // alignment required per the ABI + unsigned MaxAlign = MFI->getMaxAlignment(); // algmt required by data in frame + unsigned AlignMask = std::max(MaxAlign, TargetAlign) - 1; + + const PPCRegisterInfo *RegInfo = + static_cast(MF.getSubtarget().getRegisterInfo()); // If we are a leaf function, and use up to 224 bytes of stack space, // don't have a frame pointer, calls, or dynamic alloca then we do not need @@ -231,11 +384,11 @@ unsigned PPCFrameLowering::determineFrameLayout(MachineFunction &MF, if (!DisableRedZone && (Subtarget.isPPC64() || // 32-bit SVR4, no stack- !Subtarget.isSVR4ABI() || // allocated locals. - FrameSize == 0) && + FrameSize == 0) && FrameSize <= 224 && // Fits in red zone. !MFI->hasVarSizedObjects() && // No dynamic alloca. !MFI->adjustsStack() && // No calls. - (!ALIGN_STACK || MaxAlign <= TargetAlign)) { // No special alignment. + !RegInfo->hasBasePointer(MF)) { // No special alignment. // No need for frame if (UpdateMF) MFI->setStackSize(0); @@ -245,9 +398,10 @@ unsigned PPCFrameLowering::determineFrameLayout(MachineFunction &MF, // Get the maximum call frame size of all the calls. unsigned maxCallFrameSize = MFI->getMaxCallFrameSize(); - // Maximum call frame needs to be at least big enough for linkage and 8 args. - unsigned minCallFrameSize = getMinCallFrameSize(Subtarget.isPPC64(), - Subtarget.isDarwinABI()); + // Maximum call frame needs to be at least big enough for linkage area. + unsigned minCallFrameSize = getLinkageSize(Subtarget.isPPC64(), + Subtarget.isDarwinABI(), + Subtarget.isELFv2ABI()); maxCallFrameSize = std::max(maxCallFrameSize, minCallFrameSize); // If we have dynamic alloca then maxCallFrameSize needs to be aligned so @@ -290,8 +444,8 @@ bool PPCFrameLowering::needsFP(const MachineFunction &MF) const { // Naked functions have no stack frame pushed, so we don't have a frame // pointer. - if (MF.getFunction()->getAttributes().hasAttribute(AttributeSet::FunctionIndex, - Attribute::Naked)) + if (MF.getFunction()->getAttributes().hasAttribute( + AttributeSet::FunctionIndex, Attribute::Naked)) return false; return MF.getTarget().Options.DisableFramePointerElim(MF) || @@ -305,6 +459,12 @@ void PPCFrameLowering::replaceFPWithRealFP(MachineFunction &MF) const { unsigned FPReg = is31 ? PPC::R31 : PPC::R1; unsigned FP8Reg = is31 ? PPC::X31 : PPC::X1; + const PPCRegisterInfo *RegInfo = + static_cast(MF.getSubtarget().getRegisterInfo()); + bool HasBP = RegInfo->hasBasePointer(MF); + unsigned BPReg = HasBP ? (unsigned) RegInfo->getBaseRegister(MF) : FPReg; + unsigned BP8Reg = HasBP ? (unsigned) PPC::X30 : FPReg; + for (MachineFunction::iterator BI = MF.begin(), BE = MF.end(); BI != BE; ++BI) for (MachineBasicBlock::iterator MBBI = BI->end(); MBBI != BI->begin(); ) { @@ -321,6 +481,13 @@ void PPCFrameLowering::replaceFPWithRealFP(MachineFunction &MF) const { case PPC::FP8: MO.setReg(FP8Reg); break; + case PPC::BP: + MO.setReg(BPReg); + break; + case PPC::BP8: + MO.setReg(BP8Reg); + break; + } } } @@ -331,19 +498,29 @@ void PPCFrameLowering::emitPrologue(MachineFunction &MF) const { MachineBasicBlock::iterator MBBI = MBB.begin(); MachineFrameInfo *MFI = MF.getFrameInfo(); const PPCInstrInfo &TII = - *static_cast(MF.getTarget().getInstrInfo()); + *static_cast(MF.getSubtarget().getInstrInfo()); + const PPCRegisterInfo *RegInfo = + static_cast(MF.getSubtarget().getRegisterInfo()); MachineModuleInfo &MMI = MF.getMMI(); + const MCRegisterInfo *MRI = MMI.getContext().getRegisterInfo(); DebugLoc dl; bool needsFrameMoves = MMI.hasDebugInfo() || MF.getFunction()->needsUnwindTableEntry(); + bool isPIC = MF.getTarget().getRelocationModel() == Reloc::PIC_; - // Prepare for frame info. - MCSymbol *FrameLabel = 0; + // Get processor type. + bool isPPC64 = Subtarget.isPPC64(); + // Get the ABI. + bool isDarwinABI = Subtarget.isDarwinABI(); + bool isSVR4ABI = Subtarget.isSVR4ABI(); + bool isELFv2ABI = Subtarget.isELFv2ABI(); + assert((isDarwinABI || isSVR4ABI) && + "Currently only Darwin and SVR4 ABIs are supported for PowerPC."); // Scan the prolog, looking for an UPDATE_VRSAVE instruction. If we find it, // process it. - if (!Subtarget.isSVR4ABI()) + if (!isSVR4ABI) for (unsigned i = 0; MBBI != MBB.end(); ++i, ++MBBI) { if (MBBI->getOpcode() == PPC::UPDATE_VRSAVE) { HandleVRSaveUpdate(MBBI, TII); @@ -357,226 +534,251 @@ void PPCFrameLowering::emitPrologue(MachineFunction &MF) const { // Work out frame sizes. unsigned FrameSize = determineFrameLayout(MF); int NegFrameSize = -FrameSize; + if (!isInt<32>(NegFrameSize)) + llvm_unreachable("Unhandled stack size!"); if (MFI->isFrameAddressTaken()) replaceFPWithRealFP(MF); - // Get processor type. - bool isPPC64 = Subtarget.isPPC64(); - // Get operating system - bool isDarwinABI = Subtarget.isDarwinABI(); // Check if the link register (LR) must be saved. PPCFunctionInfo *FI = MF.getInfo(); bool MustSaveLR = FI->mustSaveLR(); - const SmallVector &MustSaveCRs = FI->getMustSaveCRs(); - // Do we have a frame pointer for this function? + const SmallVectorImpl &MustSaveCRs = FI->getMustSaveCRs(); + // Do we have a frame pointer and/or base pointer for this function? bool HasFP = hasFP(MF); + bool HasBP = RegInfo->hasBasePointer(MF); + + unsigned SPReg = isPPC64 ? PPC::X1 : PPC::R1; + unsigned BPReg = RegInfo->getBaseRegister(MF); + unsigned FPReg = isPPC64 ? PPC::X31 : PPC::R31; + unsigned LRReg = isPPC64 ? PPC::LR8 : PPC::LR; + unsigned ScratchReg = isPPC64 ? PPC::X0 : PPC::R0; + unsigned TempReg = isPPC64 ? PPC::X12 : PPC::R12; // another scratch reg + // ...(R12/X12 is volatile in both Darwin & SVR4, & can't be a function arg.) + const MCInstrDesc& MFLRInst = TII.get(isPPC64 ? PPC::MFLR8 + : PPC::MFLR ); + const MCInstrDesc& StoreInst = TII.get(isPPC64 ? PPC::STD + : PPC::STW ); + const MCInstrDesc& StoreUpdtInst = TII.get(isPPC64 ? PPC::STDU + : PPC::STWU ); + const MCInstrDesc& StoreUpdtIdxInst = TII.get(isPPC64 ? PPC::STDUX + : PPC::STWUX); + const MCInstrDesc& LoadImmShiftedInst = TII.get(isPPC64 ? PPC::LIS8 + : PPC::LIS ); + const MCInstrDesc& OrImmInst = TII.get(isPPC64 ? PPC::ORI8 + : PPC::ORI ); + const MCInstrDesc& OrInst = TII.get(isPPC64 ? PPC::OR8 + : PPC::OR ); + const MCInstrDesc& SubtractCarryingInst = TII.get(isPPC64 ? PPC::SUBFC8 + : PPC::SUBFC); + const MCInstrDesc& SubtractImmCarryingInst = TII.get(isPPC64 ? PPC::SUBFIC8 + : PPC::SUBFIC); + + // Regarding this assert: Even though LR is saved in the caller's frame (i.e., + // LROffset is positive), that slot is callee-owned. Because PPC32 SVR4 has no + // Red Zone, an asynchronous event (a form of "callee") could claim a frame & + // overwrite it, so PPC32 SVR4 must claim at least a minimal frame to save LR. + assert((isPPC64 || !isSVR4ABI || !(!FrameSize && (MustSaveLR || HasFP))) && + "FrameSize must be >0 to save/restore the FP or LR for 32-bit SVR4."); int LROffset = PPCFrameLowering::getReturnSaveOffset(isPPC64, isDarwinABI); int FPOffset = 0; if (HasFP) { - if (Subtarget.isSVR4ABI()) { + if (isSVR4ABI) { MachineFrameInfo *FFI = MF.getFrameInfo(); int FPIndex = FI->getFramePointerSaveIndex(); assert(FPIndex && "No Frame Pointer Save Slot!"); FPOffset = FFI->getObjectOffset(FPIndex); } else { - FPOffset = PPCFrameLowering::getFramePointerSaveOffset(isPPC64, isDarwinABI); + FPOffset = + PPCFrameLowering::getFramePointerSaveOffset(isPPC64, isDarwinABI); } } - if (isPPC64) { - if (MustSaveLR) - BuildMI(MBB, MBBI, dl, TII.get(PPC::MFLR8), PPC::X0); - - if (!MustSaveCRs.empty()) { - MachineInstrBuilder MIB = - BuildMI(MBB, MBBI, dl, TII.get(PPC::MFCR8), PPC::X12); - for (unsigned i = 0, e = MustSaveCRs.size(); i != e; ++i) - MIB.addReg(MustSaveCRs[i], RegState::ImplicitKill); + int BPOffset = 0; + if (HasBP) { + if (isSVR4ABI) { + MachineFrameInfo *FFI = MF.getFrameInfo(); + int BPIndex = FI->getBasePointerSaveIndex(); + assert(BPIndex && "No Base Pointer Save Slot!"); + BPOffset = FFI->getObjectOffset(BPIndex); + } else { + BPOffset = + PPCFrameLowering::getBasePointerSaveOffset(isPPC64, + isDarwinABI, + isPIC); } - - if (HasFP) - BuildMI(MBB, MBBI, dl, TII.get(PPC::STD)) - .addReg(PPC::X31) - .addImm(FPOffset/4) - .addReg(PPC::X1); - - if (MustSaveLR) - BuildMI(MBB, MBBI, dl, TII.get(PPC::STD)) - .addReg(PPC::X0) - .addImm(LROffset / 4) - .addReg(PPC::X1); - - if (!MustSaveCRs.empty()) - BuildMI(MBB, MBBI, dl, TII.get(PPC::STW8)) - .addReg(PPC::X12, getKillRegState(true)) - .addImm(8) - .addReg(PPC::X1); - } else { - if (MustSaveLR) - BuildMI(MBB, MBBI, dl, TII.get(PPC::MFLR), PPC::R0); - - if (HasFP) - // FIXME: On PPC32 SVR4, FPOffset is negative and access to negative - // offsets of R1 is not allowed. - BuildMI(MBB, MBBI, dl, TII.get(PPC::STW)) - .addReg(PPC::R31) - .addImm(FPOffset) - .addReg(PPC::R1); - - assert(MustSaveCRs.empty() && - "Prologue CR saving supported only in 64-bit mode"); - - if (MustSaveLR) - BuildMI(MBB, MBBI, dl, TII.get(PPC::STW)) - .addReg(PPC::R0) - .addImm(LROffset) - .addReg(PPC::R1); } - // Skip if a leaf routine. - if (!FrameSize) return; - // Get stack alignments. - unsigned TargetAlign = getStackAlignment(); unsigned MaxAlign = MFI->getMaxAlignment(); + if (HasBP && MaxAlign > 1) + assert(isPowerOf2_32(MaxAlign) && isInt<16>(MaxAlign) && + "Invalid alignment!"); + + // Frames of 32KB & larger require special handling because they cannot be + // indexed into with a simple STDU/STWU/STD/STW immediate offset operand. + bool isLargeFrame = !isInt<16>(NegFrameSize); + + if (MustSaveLR) + BuildMI(MBB, MBBI, dl, MFLRInst, ScratchReg); + + assert((isPPC64 || MustSaveCRs.empty()) && + "Prologue CR saving supported only in 64-bit mode"); + + if (!MustSaveCRs.empty()) { // will only occur for PPC64 + // FIXME: In the ELFv2 ABI, we are not required to save all CR fields. + // If only one or two CR fields are clobbered, it could be more + // efficient to use mfocrf to selectively save just those fields. + MachineInstrBuilder MIB = + BuildMI(MBB, MBBI, dl, TII.get(PPC::MFCR8), TempReg); + for (unsigned i = 0, e = MustSaveCRs.size(); i != e; ++i) + MIB.addReg(MustSaveCRs[i], RegState::ImplicitKill); + } + + if (HasFP) + // FIXME: On PPC32 SVR4, we must not spill before claiming the stackframe. + BuildMI(MBB, MBBI, dl, StoreInst) + .addReg(FPReg) + .addImm(FPOffset) + .addReg(SPReg); + + if (HasBP) + // FIXME: On PPC32 SVR4, we must not spill before claiming the stackframe. + BuildMI(MBB, MBBI, dl, StoreInst) + .addReg(BPReg) + .addImm(BPOffset) + .addReg(SPReg); + + if (MustSaveLR) + // FIXME: On PPC32 SVR4, we must not spill before claiming the stackframe. + BuildMI(MBB, MBBI, dl, StoreInst) + .addReg(ScratchReg) + .addImm(LROffset) + .addReg(SPReg); + + if (!MustSaveCRs.empty()) // will only occur for PPC64 + BuildMI(MBB, MBBI, dl, TII.get(PPC::STW8)) + .addReg(TempReg, getKillRegState(true)) + .addImm(8) + .addReg(SPReg); + + // Skip the rest if this is a leaf function & all spills fit in the Red Zone. + if (!FrameSize) return; // Adjust stack pointer: r1 += NegFrameSize. // If there is a preferred stack alignment, align R1 now - if (!isPPC64) { - // PPC32. - if (ALIGN_STACK && MaxAlign > TargetAlign) { - assert(isPowerOf2_32(MaxAlign) && isInt<16>(MaxAlign) && - "Invalid alignment!"); - assert(isInt<16>(NegFrameSize) && "Unhandled stack size and alignment!"); - - BuildMI(MBB, MBBI, dl, TII.get(PPC::RLWINM), PPC::R0) - .addReg(PPC::R1) + + if (HasBP) { + // Save a copy of r1 as the base pointer. + BuildMI(MBB, MBBI, dl, OrInst, BPReg) + .addReg(SPReg) + .addReg(SPReg); + } + + if (HasBP && MaxAlign > 1) { + if (isPPC64) + BuildMI(MBB, MBBI, dl, TII.get(PPC::RLDICL), ScratchReg) + .addReg(SPReg) + .addImm(0) + .addImm(64 - Log2_32(MaxAlign)); + else // PPC32... + BuildMI(MBB, MBBI, dl, TII.get(PPC::RLWINM), ScratchReg) + .addReg(SPReg) .addImm(0) .addImm(32 - Log2_32(MaxAlign)) .addImm(31); - BuildMI(MBB, MBBI, dl, TII.get(PPC::SUBFIC) ,PPC::R0) - .addReg(PPC::R0, RegState::Kill) + if (!isLargeFrame) { + BuildMI(MBB, MBBI, dl, SubtractImmCarryingInst, ScratchReg) + .addReg(ScratchReg, RegState::Kill) .addImm(NegFrameSize); - BuildMI(MBB, MBBI, dl, TII.get(PPC::STWUX), PPC::R1) - .addReg(PPC::R1, RegState::Kill) - .addReg(PPC::R1) - .addReg(PPC::R0); - } else if (isInt<16>(NegFrameSize)) { - BuildMI(MBB, MBBI, dl, TII.get(PPC::STWU), PPC::R1) - .addReg(PPC::R1) - .addImm(NegFrameSize) - .addReg(PPC::R1); } else { - BuildMI(MBB, MBBI, dl, TII.get(PPC::LIS), PPC::R0) + BuildMI(MBB, MBBI, dl, LoadImmShiftedInst, TempReg) .addImm(NegFrameSize >> 16); - BuildMI(MBB, MBBI, dl, TII.get(PPC::ORI), PPC::R0) - .addReg(PPC::R0, RegState::Kill) + BuildMI(MBB, MBBI, dl, OrImmInst, TempReg) + .addReg(TempReg, RegState::Kill) .addImm(NegFrameSize & 0xFFFF); - BuildMI(MBB, MBBI, dl, TII.get(PPC::STWUX), PPC::R1) - .addReg(PPC::R1, RegState::Kill) - .addReg(PPC::R1) - .addReg(PPC::R0); + BuildMI(MBB, MBBI, dl, SubtractCarryingInst, ScratchReg) + .addReg(ScratchReg, RegState::Kill) + .addReg(TempReg, RegState::Kill); } - } else { // PPC64. - if (ALIGN_STACK && MaxAlign > TargetAlign) { - assert(isPowerOf2_32(MaxAlign) && isInt<16>(MaxAlign) && - "Invalid alignment!"); - assert(isInt<16>(NegFrameSize) && "Unhandled stack size and alignment!"); - - BuildMI(MBB, MBBI, dl, TII.get(PPC::RLDICL), PPC::X0) - .addReg(PPC::X1) - .addImm(0) - .addImm(64 - Log2_32(MaxAlign)); - BuildMI(MBB, MBBI, dl, TII.get(PPC::SUBFIC8), PPC::X0) - .addReg(PPC::X0) - .addImm(NegFrameSize); - BuildMI(MBB, MBBI, dl, TII.get(PPC::STDUX), PPC::X1) - .addReg(PPC::X1, RegState::Kill) - .addReg(PPC::X1) - .addReg(PPC::X0); - } else if (isInt<16>(NegFrameSize)) { - BuildMI(MBB, MBBI, dl, TII.get(PPC::STDU), PPC::X1) - .addReg(PPC::X1) - .addImm(NegFrameSize / 4) - .addReg(PPC::X1); - } else { - BuildMI(MBB, MBBI, dl, TII.get(PPC::LIS8), PPC::X0) - .addImm(NegFrameSize >> 16); - BuildMI(MBB, MBBI, dl, TII.get(PPC::ORI8), PPC::X0) - .addReg(PPC::X0, RegState::Kill) - .addImm(NegFrameSize & 0xFFFF); - BuildMI(MBB, MBBI, dl, TII.get(PPC::STDUX), PPC::X1) - .addReg(PPC::X1, RegState::Kill) - .addReg(PPC::X1) - .addReg(PPC::X0); - } - } + BuildMI(MBB, MBBI, dl, StoreUpdtIdxInst, SPReg) + .addReg(SPReg, RegState::Kill) + .addReg(SPReg) + .addReg(ScratchReg); + + } else if (!isLargeFrame) { + BuildMI(MBB, MBBI, dl, StoreUpdtInst, SPReg) + .addReg(SPReg) + .addImm(NegFrameSize) + .addReg(SPReg); - std::vector &Moves = MMI.getFrameMoves(); + } else { + BuildMI(MBB, MBBI, dl, LoadImmShiftedInst, ScratchReg) + .addImm(NegFrameSize >> 16); + BuildMI(MBB, MBBI, dl, OrImmInst, ScratchReg) + .addReg(ScratchReg, RegState::Kill) + .addImm(NegFrameSize & 0xFFFF); + BuildMI(MBB, MBBI, dl, StoreUpdtIdxInst, SPReg) + .addReg(SPReg, RegState::Kill) + .addReg(SPReg) + .addReg(ScratchReg); + } // Add the "machine moves" for the instructions we generated above, but in // reverse order. if (needsFrameMoves) { - // Mark effective beginning of when frame pointer becomes valid. - FrameLabel = MMI.getContext().CreateTempSymbol(); - BuildMI(MBB, MBBI, dl, TII.get(PPC::PROLOG_LABEL)).addSym(FrameLabel); - // Show update of SP. - if (NegFrameSize) { - MachineLocation SPDst(MachineLocation::VirtualFP); - MachineLocation SPSrc(MachineLocation::VirtualFP, NegFrameSize); - Moves.push_back(MachineMove(FrameLabel, SPDst, SPSrc)); - } else { - MachineLocation SP(isPPC64 ? PPC::X31 : PPC::R31); - Moves.push_back(MachineMove(FrameLabel, SP, SP)); - } + assert(NegFrameSize); + unsigned CFIIndex = MMI.addFrameInst( + MCCFIInstruction::createDefCfaOffset(nullptr, NegFrameSize)); + BuildMI(MBB, MBBI, dl, TII.get(TargetOpcode::CFI_INSTRUCTION)) + .addCFIIndex(CFIIndex); if (HasFP) { - MachineLocation FPDst(MachineLocation::VirtualFP, FPOffset); - MachineLocation FPSrc(isPPC64 ? PPC::X31 : PPC::R31); - Moves.push_back(MachineMove(FrameLabel, FPDst, FPSrc)); + unsigned Reg = MRI->getDwarfRegNum(FPReg, true); + CFIIndex = MMI.addFrameInst( + MCCFIInstruction::createOffset(nullptr, Reg, FPOffset)); + BuildMI(MBB, MBBI, dl, TII.get(TargetOpcode::CFI_INSTRUCTION)) + .addCFIIndex(CFIIndex); + } + + if (HasBP) { + unsigned Reg = MRI->getDwarfRegNum(BPReg, true); + CFIIndex = MMI.addFrameInst( + MCCFIInstruction::createOffset(nullptr, Reg, BPOffset)); + BuildMI(MBB, MBBI, dl, TII.get(TargetOpcode::CFI_INSTRUCTION)) + .addCFIIndex(CFIIndex); } if (MustSaveLR) { - MachineLocation LRDst(MachineLocation::VirtualFP, LROffset); - MachineLocation LRSrc(isPPC64 ? PPC::LR8 : PPC::LR); - Moves.push_back(MachineMove(FrameLabel, LRDst, LRSrc)); + unsigned Reg = MRI->getDwarfRegNum(LRReg, true); + CFIIndex = MMI.addFrameInst( + MCCFIInstruction::createOffset(nullptr, Reg, LROffset)); + BuildMI(MBB, MBBI, dl, TII.get(TargetOpcode::CFI_INSTRUCTION)) + .addCFIIndex(CFIIndex); } } - MCSymbol *ReadyLabel = 0; - // If there is a frame pointer, copy R1 into R31 if (HasFP) { - if (!isPPC64) { - BuildMI(MBB, MBBI, dl, TII.get(PPC::OR), PPC::R31) - .addReg(PPC::R1) - .addReg(PPC::R1); - } else { - BuildMI(MBB, MBBI, dl, TII.get(PPC::OR8), PPC::X31) - .addReg(PPC::X1) - .addReg(PPC::X1); - } + BuildMI(MBB, MBBI, dl, OrInst, FPReg) + .addReg(SPReg) + .addReg(SPReg); if (needsFrameMoves) { - ReadyLabel = MMI.getContext().CreateTempSymbol(); - // Mark effective beginning of when frame pointer is ready. - BuildMI(MBB, MBBI, dl, TII.get(PPC::PROLOG_LABEL)).addSym(ReadyLabel); + unsigned Reg = MRI->getDwarfRegNum(FPReg, true); + unsigned CFIIndex = MMI.addFrameInst( + MCCFIInstruction::createDefCfaRegister(nullptr, Reg)); - MachineLocation FPDst(HasFP ? (isPPC64 ? PPC::X31 : PPC::R31) : - (isPPC64 ? PPC::X1 : PPC::R1)); - MachineLocation FPSrc(MachineLocation::VirtualFP); - Moves.push_back(MachineMove(ReadyLabel, FPDst, FPSrc)); + BuildMI(MBB, MBBI, dl, TII.get(TargetOpcode::CFI_INSTRUCTION)) + .addCFIIndex(CFIIndex); } } if (needsFrameMoves) { - MCSymbol *Label = HasFP ? ReadyLabel : FrameLabel; - // Add callee saved registers to move list. const std::vector &CSI = MFI->getCalleeSavedInfo(); for (unsigned I = 0, E = CSI.size(); I != E; ++I) { @@ -590,26 +792,29 @@ void PPCFrameLowering::emitPrologue(MachineFunction &MF) const { // For SVR4, don't emit a move for the CR spill slot if we haven't // spilled CRs. - if (Subtarget.isSVR4ABI() - && (PPC::CR2 <= Reg && Reg <= PPC::CR4) - && MustSaveCRs.empty()) - continue; + if (isSVR4ABI && (PPC::CR2 <= Reg && Reg <= PPC::CR4) + && MustSaveCRs.empty()) + continue; // For 64-bit SVR4 when we have spilled CRs, the spill location // is SP+8, not a frame-relative slot. - if (Subtarget.isSVR4ABI() - && Subtarget.isPPC64() - && (PPC::CR2 <= Reg && Reg <= PPC::CR4)) { - MachineLocation CSDst(PPC::X1, 8); - MachineLocation CSSrc(PPC::CR2); - Moves.push_back(MachineMove(Label, CSDst, CSSrc)); - continue; + if (isSVR4ABI && isPPC64 && (PPC::CR2 <= Reg && Reg <= PPC::CR4)) { + // In the ELFv1 ABI, only CR2 is noted in CFI and stands in for + // the whole CR word. In the ELFv2 ABI, every CR that was + // actually saved gets its own CFI record. + unsigned CRReg = isELFv2ABI? Reg : (unsigned) PPC::CR2; + unsigned CFIIndex = MMI.addFrameInst(MCCFIInstruction::createOffset( + nullptr, MRI->getDwarfRegNum(CRReg, true), 8)); + BuildMI(MBB, MBBI, dl, TII.get(TargetOpcode::CFI_INSTRUCTION)) + .addCFIIndex(CFIIndex); + continue; } int Offset = MFI->getObjectOffset(CSI[I].getFrameIdx()); - MachineLocation CSDst(MachineLocation::VirtualFP, Offset); - MachineLocation CSSrc(Reg); - Moves.push_back(MachineMove(Label, CSDst, CSSrc)); + unsigned CFIIndex = MMI.addFrameInst(MCCFIInstruction::createOffset( + nullptr, MRI->getDwarfRegNum(Reg, true), Offset)); + BuildMI(MBB, MBBI, dl, TII.get(TargetOpcode::CFI_INSTRUCTION)) + .addCFIIndex(CFIIndex); } } } @@ -619,7 +824,9 @@ void PPCFrameLowering::emitEpilogue(MachineFunction &MF, MachineBasicBlock::iterator MBBI = MBB.getLastNonDebugInstr(); assert(MBBI != MBB.end() && "Returning block has no terminator"); const PPCInstrInfo &TII = - *static_cast(MF.getTarget().getInstrInfo()); + *static_cast(MF.getSubtarget().getInstrInfo()); + const PPCRegisterInfo *RegInfo = + static_cast(MF.getSubtarget().getRegisterInfo()); unsigned RetOpcode = MBBI->getOpcode(); DebugLoc dl; @@ -633,36 +840,72 @@ void PPCFrameLowering::emitEpilogue(MachineFunction &MF, RetOpcode == PPC::TCRETURNai8) && "Can only insert epilog into returning blocks"); - // Get alignment info so we know how to restore r1 + // Get alignment info so we know how to restore the SP. const MachineFrameInfo *MFI = MF.getFrameInfo(); - unsigned TargetAlign = getStackAlignment(); - unsigned MaxAlign = MFI->getMaxAlignment(); // Get the number of bytes allocated from the FrameInfo. int FrameSize = MFI->getStackSize(); // Get processor type. bool isPPC64 = Subtarget.isPPC64(); - // Get operating system + // Get the ABI. bool isDarwinABI = Subtarget.isDarwinABI(); + bool isSVR4ABI = Subtarget.isSVR4ABI(); + bool isPIC = MF.getTarget().getRelocationModel() == Reloc::PIC_; + // Check if the link register (LR) has been saved. PPCFunctionInfo *FI = MF.getInfo(); bool MustSaveLR = FI->mustSaveLR(); - const SmallVector &MustSaveCRs = FI->getMustSaveCRs(); - // Do we have a frame pointer for this function? + const SmallVectorImpl &MustSaveCRs = FI->getMustSaveCRs(); + // Do we have a frame pointer and/or base pointer for this function? bool HasFP = hasFP(MF); + bool HasBP = RegInfo->hasBasePointer(MF); + + unsigned SPReg = isPPC64 ? PPC::X1 : PPC::R1; + unsigned BPReg = RegInfo->getBaseRegister(MF); + unsigned FPReg = isPPC64 ? PPC::X31 : PPC::R31; + unsigned ScratchReg = isPPC64 ? PPC::X0 : PPC::R0; + unsigned TempReg = isPPC64 ? PPC::X12 : PPC::R12; // another scratch reg + const MCInstrDesc& MTLRInst = TII.get( isPPC64 ? PPC::MTLR8 + : PPC::MTLR ); + const MCInstrDesc& LoadInst = TII.get( isPPC64 ? PPC::LD + : PPC::LWZ ); + const MCInstrDesc& LoadImmShiftedInst = TII.get( isPPC64 ? PPC::LIS8 + : PPC::LIS ); + const MCInstrDesc& OrImmInst = TII.get( isPPC64 ? PPC::ORI8 + : PPC::ORI ); + const MCInstrDesc& AddImmInst = TII.get( isPPC64 ? PPC::ADDI8 + : PPC::ADDI ); + const MCInstrDesc& AddInst = TII.get( isPPC64 ? PPC::ADD8 + : PPC::ADD4 ); int LROffset = PPCFrameLowering::getReturnSaveOffset(isPPC64, isDarwinABI); int FPOffset = 0; if (HasFP) { - if (Subtarget.isSVR4ABI()) { + if (isSVR4ABI) { MachineFrameInfo *FFI = MF.getFrameInfo(); int FPIndex = FI->getFramePointerSaveIndex(); assert(FPIndex && "No Frame Pointer Save Slot!"); FPOffset = FFI->getObjectOffset(FPIndex); } else { - FPOffset = PPCFrameLowering::getFramePointerSaveOffset(isPPC64, isDarwinABI); + FPOffset = + PPCFrameLowering::getFramePointerSaveOffset(isPPC64, isDarwinABI); + } + } + + int BPOffset = 0; + if (HasBP) { + if (isSVR4ABI) { + MachineFrameInfo *FFI = MF.getFrameInfo(); + int BPIndex = FI->getBasePointerSaveIndex(); + assert(BPIndex && "No Base Pointer Save Slot!"); + BPOffset = FFI->getObjectOffset(BPIndex); + } else { + BPOffset = + PPCFrameLowering::getBasePointerSaveOffset(isPPC64, + isDarwinABI, + isPIC); } } @@ -687,98 +930,76 @@ void PPCFrameLowering::emitEpilogue(MachineFunction &MF, FrameSize += StackAdj; } + // Frames of 32KB & larger require special handling because they cannot be + // indexed into with a simple LD/LWZ immediate offset operand. + bool isLargeFrame = !isInt<16>(FrameSize); + if (FrameSize) { - // The loaded (or persistent) stack pointer value is offset by the 'stwu' - // on entry to the function. Add this offset back now. - if (!isPPC64) { - // If this function contained a fastcc call and GuaranteedTailCallOpt is - // enabled (=> hasFastCall()==true) the fastcc call might contain a tail - // call which invalidates the stack pointer value in SP(0). So we use the - // value of R31 in this case. - if (FI->hasFastCall() && isInt<16>(FrameSize)) { - assert(hasFP(MF) && "Expecting a valid the frame pointer."); - BuildMI(MBB, MBBI, dl, TII.get(PPC::ADDI), PPC::R1) - .addReg(PPC::R31).addImm(FrameSize); - } else if(FI->hasFastCall()) { - BuildMI(MBB, MBBI, dl, TII.get(PPC::LIS), PPC::R0) - .addImm(FrameSize >> 16); - BuildMI(MBB, MBBI, dl, TII.get(PPC::ORI), PPC::R0) - .addReg(PPC::R0, RegState::Kill) - .addImm(FrameSize & 0xFFFF); - BuildMI(MBB, MBBI, dl, TII.get(PPC::ADD4)) - .addReg(PPC::R1) - .addReg(PPC::R31) - .addReg(PPC::R0); - } else if (isInt<16>(FrameSize) && - (!ALIGN_STACK || TargetAlign >= MaxAlign) && - !MFI->hasVarSizedObjects()) { - BuildMI(MBB, MBBI, dl, TII.get(PPC::ADDI), PPC::R1) - .addReg(PPC::R1).addImm(FrameSize); + // In the prologue, the loaded (or persistent) stack pointer value is offset + // by the STDU/STDUX/STWU/STWUX instruction. Add this offset back now. + + // If this function contained a fastcc call and GuaranteedTailCallOpt is + // enabled (=> hasFastCall()==true) the fastcc call might contain a tail + // call which invalidates the stack pointer value in SP(0). So we use the + // value of R31 in this case. + if (FI->hasFastCall()) { + assert(HasFP && "Expecting a valid frame pointer."); + if (!isLargeFrame) { + BuildMI(MBB, MBBI, dl, AddImmInst, SPReg) + .addReg(FPReg).addImm(FrameSize); } else { - BuildMI(MBB, MBBI, dl, TII.get(PPC::LWZ),PPC::R1) - .addImm(0).addReg(PPC::R1); - } - } else { - if (FI->hasFastCall() && isInt<16>(FrameSize)) { - assert(hasFP(MF) && "Expecting a valid the frame pointer."); - BuildMI(MBB, MBBI, dl, TII.get(PPC::ADDI8), PPC::X1) - .addReg(PPC::X31).addImm(FrameSize); - } else if(FI->hasFastCall()) { - BuildMI(MBB, MBBI, dl, TII.get(PPC::LIS8), PPC::X0) + BuildMI(MBB, MBBI, dl, LoadImmShiftedInst, ScratchReg) .addImm(FrameSize >> 16); - BuildMI(MBB, MBBI, dl, TII.get(PPC::ORI8), PPC::X0) - .addReg(PPC::X0, RegState::Kill) + BuildMI(MBB, MBBI, dl, OrImmInst, ScratchReg) + .addReg(ScratchReg, RegState::Kill) .addImm(FrameSize & 0xFFFF); - BuildMI(MBB, MBBI, dl, TII.get(PPC::ADD8)) - .addReg(PPC::X1) - .addReg(PPC::X31) - .addReg(PPC::X0); - } else if (isInt<16>(FrameSize) && TargetAlign >= MaxAlign && - !MFI->hasVarSizedObjects()) { - BuildMI(MBB, MBBI, dl, TII.get(PPC::ADDI8), PPC::X1) - .addReg(PPC::X1).addImm(FrameSize); - } else { - BuildMI(MBB, MBBI, dl, TII.get(PPC::LD), PPC::X1) - .addImm(0).addReg(PPC::X1); + BuildMI(MBB, MBBI, dl, AddInst) + .addReg(SPReg) + .addReg(FPReg) + .addReg(ScratchReg); } + } else if (!isLargeFrame && !HasBP && !MFI->hasVarSizedObjects()) { + BuildMI(MBB, MBBI, dl, AddImmInst, SPReg) + .addReg(SPReg) + .addImm(FrameSize); + } else { + BuildMI(MBB, MBBI, dl, LoadInst, SPReg) + .addImm(0) + .addReg(SPReg); } - } - if (isPPC64) { - if (MustSaveLR) - BuildMI(MBB, MBBI, dl, TII.get(PPC::LD), PPC::X0) - .addImm(LROffset/4).addReg(PPC::X1); + } - if (!MustSaveCRs.empty()) - BuildMI(MBB, MBBI, dl, TII.get(PPC::LWZ8), PPC::X12) - .addImm(8).addReg(PPC::X1); + if (MustSaveLR) + BuildMI(MBB, MBBI, dl, LoadInst, ScratchReg) + .addImm(LROffset) + .addReg(SPReg); - if (HasFP) - BuildMI(MBB, MBBI, dl, TII.get(PPC::LD), PPC::X31) - .addImm(FPOffset/4).addReg(PPC::X1); + assert((isPPC64 || MustSaveCRs.empty()) && + "Epilogue CR restoring supported only in 64-bit mode"); - if (!MustSaveCRs.empty()) - for (unsigned i = 0, e = MustSaveCRs.size(); i != e; ++i) - BuildMI(MBB, MBBI, dl, TII.get(PPC::MTCRF8), MustSaveCRs[i]) - .addReg(PPC::X12, getKillRegState(i == e-1)); + if (!MustSaveCRs.empty()) // will only occur for PPC64 + BuildMI(MBB, MBBI, dl, TII.get(PPC::LWZ8), TempReg) + .addImm(8) + .addReg(SPReg); - if (MustSaveLR) - BuildMI(MBB, MBBI, dl, TII.get(PPC::MTLR8)).addReg(PPC::X0); - } else { - if (MustSaveLR) - BuildMI(MBB, MBBI, dl, TII.get(PPC::LWZ), PPC::R0) - .addImm(LROffset).addReg(PPC::R1); + if (HasFP) + BuildMI(MBB, MBBI, dl, LoadInst, FPReg) + .addImm(FPOffset) + .addReg(SPReg); - assert(MustSaveCRs.empty() && - "Epilogue CR restoring supported only in 64-bit mode"); + if (HasBP) + BuildMI(MBB, MBBI, dl, LoadInst, BPReg) + .addImm(BPOffset) + .addReg(SPReg); - if (HasFP) - BuildMI(MBB, MBBI, dl, TII.get(PPC::LWZ), PPC::R31) - .addImm(FPOffset).addReg(PPC::R1); + if (!MustSaveCRs.empty()) // will only occur for PPC64 + for (unsigned i = 0, e = MustSaveCRs.size(); i != e; ++i) + BuildMI(MBB, MBBI, dl, TII.get(PPC::MTOCRF8), MustSaveCRs[i]) + .addReg(TempReg, getKillRegState(i == e-1)); - if (MustSaveLR) - BuildMI(MBB, MBBI, dl, TII.get(PPC::MTLR)).addReg(PPC::R0); - } + if (MustSaveLR) + BuildMI(MBB, MBBI, dl, MTLRInst).addReg(ScratchReg); // Callee pop calling convention. Pop parameter/linkage area. Used for tail // call optimization @@ -786,27 +1007,20 @@ void PPCFrameLowering::emitEpilogue(MachineFunction &MF, MF.getFunction()->getCallingConv() == CallingConv::Fast) { PPCFunctionInfo *FI = MF.getInfo(); unsigned CallerAllocatedAmt = FI->getMinReservedArea(); - unsigned StackReg = isPPC64 ? PPC::X1 : PPC::R1; - unsigned FPReg = isPPC64 ? PPC::X31 : PPC::R31; - unsigned TmpReg = isPPC64 ? PPC::X0 : PPC::R0; - unsigned ADDIInstr = isPPC64 ? PPC::ADDI8 : PPC::ADDI; - unsigned ADDInstr = isPPC64 ? PPC::ADD8 : PPC::ADD4; - unsigned LISInstr = isPPC64 ? PPC::LIS8 : PPC::LIS; - unsigned ORIInstr = isPPC64 ? PPC::ORI8 : PPC::ORI; if (CallerAllocatedAmt && isInt<16>(CallerAllocatedAmt)) { - BuildMI(MBB, MBBI, dl, TII.get(ADDIInstr), StackReg) - .addReg(StackReg).addImm(CallerAllocatedAmt); + BuildMI(MBB, MBBI, dl, AddImmInst, SPReg) + .addReg(SPReg).addImm(CallerAllocatedAmt); } else { - BuildMI(MBB, MBBI, dl, TII.get(LISInstr), TmpReg) + BuildMI(MBB, MBBI, dl, LoadImmShiftedInst, ScratchReg) .addImm(CallerAllocatedAmt >> 16); - BuildMI(MBB, MBBI, dl, TII.get(ORIInstr), TmpReg) - .addReg(TmpReg, RegState::Kill) + BuildMI(MBB, MBBI, dl, OrImmInst, ScratchReg) + .addReg(ScratchReg, RegState::Kill) .addImm(CallerAllocatedAmt & 0xFFFF); - BuildMI(MBB, MBBI, dl, TII.get(ADDInstr)) - .addReg(StackReg) + BuildMI(MBB, MBBI, dl, AddInst) + .addReg(SPReg) .addReg(FPReg) - .addReg(TmpReg); + .addReg(ScratchReg); } } else if (RetOpcode == PPC::TCRETURNdi) { MBBI = MBB.getLastNonDebugInstr(); @@ -854,7 +1068,8 @@ static bool MustSaveLR(const MachineFunction &MF, unsigned LR) { void PPCFrameLowering::processFunctionBeforeCalleeSavedScan(MachineFunction &MF, RegScavenger *) const { - const TargetRegisterInfo *RegInfo = MF.getTarget().getRegisterInfo(); + const PPCRegisterInfo *RegInfo = + static_cast(MF.getSubtarget().getRegisterInfo()); // Save and clear the LR state. PPCFunctionInfo *FI = MF.getInfo(); @@ -867,6 +1082,7 @@ PPCFrameLowering::processFunctionBeforeCalleeSavedScan(MachineFunction &MF, int FPSI = FI->getFramePointerSaveIndex(); bool isPPC64 = Subtarget.isPPC64(); bool isDarwinABI = Subtarget.isDarwinABI(); + bool isPIC = MF.getTarget().getRelocationModel() == Reloc::PIC_; MachineFrameInfo *MFI = MF.getFrameInfo(); // If the frame pointer save index hasn't been defined yet. @@ -879,6 +1095,15 @@ PPCFrameLowering::processFunctionBeforeCalleeSavedScan(MachineFunction &MF, FI->setFramePointerSaveIndex(FPSI); } + int BPSI = FI->getBasePointerSaveIndex(); + if (!BPSI && RegInfo->hasBasePointer(MF)) { + int BPOffset = getBasePointerSaveOffset(isPPC64, isDarwinABI, isPIC); + // Allocate the frame index for the base pointer save area. + BPSI = MFI->CreateFixedObject(isPPC64? 8 : 4, BPOffset, true); + // Save the result. + FI->setBasePointerSaveIndex(BPSI); + } + // Reserve stack space to move the linkage area to in case of a tail call. int TCSPDelta = 0; if (MF.getTarget().Options.GuaranteedTailCallOpt && @@ -886,9 +1111,9 @@ PPCFrameLowering::processFunctionBeforeCalleeSavedScan(MachineFunction &MF, MFI->CreateFixedObject(-1 * TCSPDelta, TCSPDelta, true); } - // For 32-bit SVR4, allocate the nonvolatile CR spill slot iff the + // For 32-bit SVR4, allocate the nonvolatile CR spill slot iff the // function uses CR 2, 3, or 4. - if (!isPPC64 && !isDarwinABI && + if (!isPPC64 && !isDarwinABI && (MRI.isPhysRegUsed(PPC::CR2) || MRI.isPhysRegUsed(PPC::CR3) || MRI.isPhysRegUsed(PPC::CR4))) { @@ -976,7 +1201,7 @@ void PPCFrameLowering::processFunctionBeforeFrameFinalized(MachineFunction &MF, } PPCFunctionInfo *PFI = MF.getInfo(); - const TargetRegisterInfo *TRI = MF.getTarget().getRegisterInfo(); + const TargetRegisterInfo *TRI = MF.getSubtarget().getRegisterInfo(); int64_t LowerBound = 0; @@ -1010,6 +1235,17 @@ void PPCFrameLowering::processFunctionBeforeFrameFinalized(MachineFunction &MF, FFI->setObjectOffset(FI, LowerBound + FFI->getObjectOffset(FI)); } + const PPCRegisterInfo *RegInfo = + static_cast(MF.getSubtarget().getRegisterInfo()); + if (RegInfo->hasBasePointer(MF)) { + HasGPSaveArea = true; + + int FI = PFI->getBasePointerSaveIndex(); + assert(FI && "No Base Pointer Save Slot!"); + + FFI->setObjectOffset(FI, LowerBound + FFI->getObjectOffset(FI)); + } + // General register save area starts right below the Floating-point // register save area. if (HasGPSaveArea || HasG8SaveArea) { @@ -1051,10 +1287,10 @@ void PPCFrameLowering::processFunctionBeforeFrameFinalized(MachineFunction &MF, unsigned Reg = CSI[i].getReg(); if ((Subtarget.isSVR4ABI() && Reg == PPC::CR2) - // Leave Darwin logic as-is. - || (!Subtarget.isSVR4ABI() && - (PPC::CRBITRCRegClass.contains(Reg) || - PPC::CRRCRegClass.contains(Reg)))) { + // Leave Darwin logic as-is. + || (!Subtarget.isSVR4ABI() && + (PPC::CRBITRCRegClass.contains(Reg) || + PPC::CRRCRegClass.contains(Reg)))) { int FI = CSI[i].getFrameIdx(); FFI->setObjectOffset(FI, LowerBound + FFI->getObjectOffset(FI)); @@ -1122,8 +1358,12 @@ PPCFrameLowering::addScavengingSpillSlot(MachineFunction &MF, RC->getAlignment(), false)); + // Might we have over-aligned allocas? + bool HasAlVars = MFI->hasVarSizedObjects() && + MFI->getMaxAlignment() > getStackAlignment(); + // These kinds of spills might need two registers. - if (spillsCR(MF) || spillsVRSAVE(MF)) + if (spillsCR(MF) || spillsVRSAVE(MF) || HasAlVars) RS->addScavengingFrameIndex(MFI->CreateStackObject(RC->getSize(), RC->getAlignment(), false)); @@ -1131,11 +1371,11 @@ PPCFrameLowering::addScavengingSpillSlot(MachineFunction &MF, } } -bool +bool PPCFrameLowering::spillCalleeSavedRegisters(MachineBasicBlock &MBB, - MachineBasicBlock::iterator MI, - const std::vector &CSI, - const TargetRegisterInfo *TRI) const { + MachineBasicBlock::iterator MI, + const std::vector &CSI, + const TargetRegisterInfo *TRI) const { // Currently, this function only handles SVR4 32- and 64-bit ABIs. // Return false otherwise to maintain pre-existing behavior. @@ -1144,13 +1384,19 @@ PPCFrameLowering::spillCalleeSavedRegisters(MachineBasicBlock &MBB, MachineFunction *MF = MBB.getParent(); const PPCInstrInfo &TII = - *static_cast(MF->getTarget().getInstrInfo()); + *static_cast(MF->getSubtarget().getInstrInfo()); DebugLoc DL; bool CRSpilled = false; MachineInstrBuilder CRMIB; - + for (unsigned i = 0, e = CSI.size(); i != e; ++i) { unsigned Reg = CSI[i].getReg(); + // Only Darwin actually uses the VRSAVE register, but it can still appear + // here if, for example, @llvm.eh.unwind.init() is used. If we're not on + // Darwin, ignore it. + if (Reg == PPC::VRSAVE && !Subtarget.isDarwinABI()) + continue; + // CR2 through CR4 are the nonvolatile CR fields. bool IsCRField = PPC::CR2 <= Reg && Reg <= PPC::CR4; @@ -1170,22 +1416,23 @@ PPCFrameLowering::spillCalleeSavedRegisters(MachineBasicBlock &MBB, FuncInfo->addMustSaveCR(Reg); } else { CRSpilled = true; + FuncInfo->setSpillsCR(); - // 32-bit: FP-relative. Note that we made sure CR2-CR4 all have - // the same frame index in PPCRegisterInfo::hasReservedSpillSlot. - CRMIB = BuildMI(*MF, DL, TII.get(PPC::MFCR), PPC::R12) + // 32-bit: FP-relative. Note that we made sure CR2-CR4 all have + // the same frame index in PPCRegisterInfo::hasReservedSpillSlot. + CRMIB = BuildMI(*MF, DL, TII.get(PPC::MFCR), PPC::R12) .addReg(Reg, RegState::ImplicitKill); - MBB.insert(MI, CRMIB); - MBB.insert(MI, addFrameReference(BuildMI(*MF, DL, TII.get(PPC::STW)) - .addReg(PPC::R12, - getKillRegState(true)), - CSI[i].getFrameIdx())); + MBB.insert(MI, CRMIB); + MBB.insert(MI, addFrameReference(BuildMI(*MF, DL, TII.get(PPC::STW)) + .addReg(PPC::R12, + getKillRegState(true)), + CSI[i].getFrameIdx())); } } else { const TargetRegisterClass *RC = TRI->getMinimalPhysRegClass(Reg); TII.storeRegToStackSlot(MBB, MI, Reg, true, - CSI[i].getFrameIdx(), RC, TRI); + CSI[i].getFrameIdx(), RC, TRI); } } return true; @@ -1194,12 +1441,12 @@ PPCFrameLowering::spillCalleeSavedRegisters(MachineBasicBlock &MBB, static void restoreCRs(bool isPPC64, bool is31, bool CR2Spilled, bool CR3Spilled, bool CR4Spilled, - MachineBasicBlock &MBB, MachineBasicBlock::iterator MI, - const std::vector &CSI, unsigned CSIIndex) { + MachineBasicBlock &MBB, MachineBasicBlock::iterator MI, + const std::vector &CSI, unsigned CSIIndex) { MachineFunction *MF = MBB.getParent(); const PPCInstrInfo &TII = - *static_cast(MF->getTarget().getInstrInfo()); + *static_cast(MF->getSubtarget().getInstrInfo()); DebugLoc DL; unsigned RestoreOp, MoveReg; @@ -1209,12 +1456,12 @@ restoreCRs(bool isPPC64, bool is31, else { // 32-bit: FP-relative MBB.insert(MI, addFrameReference(BuildMI(*MF, DL, TII.get(PPC::LWZ), - PPC::R12), - CSI[CSIIndex].getFrameIdx())); - RestoreOp = PPC::MTCRF; + PPC::R12), + CSI[CSIIndex].getFrameIdx())); + RestoreOp = PPC::MTOCRF; MoveReg = PPC::R12; } - + if (CR2Spilled) MBB.insert(MI, BuildMI(*MF, DL, TII.get(RestoreOp), PPC::CR2) .addReg(MoveReg, getKillRegState(!CR3Spilled && !CR4Spilled))); @@ -1232,7 +1479,7 @@ void PPCFrameLowering:: eliminateCallFramePseudoInstr(MachineFunction &MF, MachineBasicBlock &MBB, MachineBasicBlock::iterator I) const { const PPCInstrInfo &TII = - *static_cast(MF.getTarget().getInstrInfo()); + *static_cast(MF.getSubtarget().getInstrInfo()); if (MF.getTarget().Options.GuaranteedTailCallOpt && I->getOpcode() == PPC::ADJCALLSTACKUP) { // Add (actually subtract) back the amount the callee popped on return. @@ -1269,11 +1516,11 @@ eliminateCallFramePseudoInstr(MachineFunction &MF, MachineBasicBlock &MBB, MBB.erase(I); } -bool +bool PPCFrameLowering::restoreCalleeSavedRegisters(MachineBasicBlock &MBB, - MachineBasicBlock::iterator MI, - const std::vector &CSI, - const TargetRegisterInfo *TRI) const { + MachineBasicBlock::iterator MI, + const std::vector &CSI, + const TargetRegisterInfo *TRI) const { // Currently, this function only handles SVR4 32- and 64-bit ABIs. // Return false otherwise to maintain pre-existing behavior. @@ -1282,7 +1529,7 @@ PPCFrameLowering::restoreCalleeSavedRegisters(MachineBasicBlock &MBB, MachineFunction *MF = MBB.getParent(); const PPCInstrInfo &TII = - *static_cast(MF->getTarget().getInstrInfo()); + *static_cast(MF->getSubtarget().getInstrInfo()); bool CR2Spilled = false; bool CR3Spilled = false; bool CR4Spilled = false; @@ -1299,6 +1546,12 @@ PPCFrameLowering::restoreCalleeSavedRegisters(MachineBasicBlock &MBB, for (unsigned i = 0, e = CSI.size(); i != e; ++i) { unsigned Reg = CSI[i].getReg(); + // Only Darwin actually uses the VRSAVE register, but it can still appear + // here if, for example, @llvm.eh.unwind.init() is used. If we're not on + // Darwin, ignore it. + if (Reg == PPC::VRSAVE && !Subtarget.isDarwinABI()) + continue; + if (Reg == PPC::CR2) { CR2Spilled = true; // The spill slot is associated only with CR2, which is the @@ -1315,20 +1568,20 @@ PPCFrameLowering::restoreCalleeSavedRegisters(MachineBasicBlock &MBB, // When we first encounter a non-CR register after seeing at // least one CR register, restore all spilled CRs together. if ((CR2Spilled || CR3Spilled || CR4Spilled) - && !(PPC::CR2 <= Reg && Reg <= PPC::CR4)) { + && !(PPC::CR2 <= Reg && Reg <= PPC::CR4)) { bool is31 = needsFP(*MF); restoreCRs(Subtarget.isPPC64(), is31, CR2Spilled, CR3Spilled, CR4Spilled, - MBB, I, CSI, CSIIndex); - CR2Spilled = CR3Spilled = CR4Spilled = false; + MBB, I, CSI, CSIIndex); + CR2Spilled = CR3Spilled = CR4Spilled = false; } // Default behavior for non-CR saves. const TargetRegisterClass *RC = TRI->getMinimalPhysRegClass(Reg); TII.loadRegFromStackSlot(MBB, I, Reg, CSI[i].getFrameIdx(), - RC, TRI); + RC, TRI); assert(I != MBB.begin() && - "loadRegFromStackSlot didn't insert any code!"); + "loadRegFromStackSlot didn't insert any code!"); } // Insert in reverse order. @@ -1337,16 +1590,15 @@ PPCFrameLowering::restoreCalleeSavedRegisters(MachineBasicBlock &MBB, else { I = BeforeI; ++I; - } + } } // If we haven't yet spilled the CRs, do so now. if (CR2Spilled || CR3Spilled || CR4Spilled) { - bool is31 = needsFP(*MF); + bool is31 = needsFP(*MF); restoreCRs(Subtarget.isPPC64(), is31, CR2Spilled, CR3Spilled, CR4Spilled, - MBB, I, CSI, CSIIndex); + MBB, I, CSI, CSIIndex); } return true; } -