diff options
author | dim <dim@FreeBSD.org> | 2014-03-21 17:53:59 +0000 |
---|---|---|
committer | dim <dim@FreeBSD.org> | 2014-03-21 17:53:59 +0000 |
commit | 9cedb8bb69b89b0f0c529937247a6a80cabdbaec (patch) | |
tree | c978f0e9ec1ab92dc8123783f30b08a7fd1e2a39 /contrib/llvm/lib/Target/PowerPC/PPCFrameLowering.cpp | |
parent | 03fdc2934eb61c44c049a02b02aa974cfdd8a0eb (diff) | |
download | FreeBSD-src-9cedb8bb69b89b0f0c529937247a6a80cabdbaec.zip FreeBSD-src-9cedb8bb69b89b0f0c529937247a6a80cabdbaec.tar.gz |
MFC 261991:
Upgrade our copy of llvm/clang to 3.4 release. This version supports
all of the features in the current working draft of the upcoming C++
standard, provisionally named C++1y.
The code generator's performance is greatly increased, and the loop
auto-vectorizer is now enabled at -Os and -O2 in addition to -O3. The
PowerPC backend has made several major improvements to code generation
quality and compile time, and the X86, SPARC, ARM32, Aarch64 and SystemZ
backends have all seen major feature work.
Release notes for llvm and clang can be found here:
<http://llvm.org/releases/3.4/docs/ReleaseNotes.html>
<http://llvm.org/releases/3.4/tools/clang/docs/ReleaseNotes.html>
MFC 262121 (by emaste):
Update lldb for clang/llvm 3.4 import
This commit largely restores the lldb source to the upstream r196259
snapshot with the addition of threaded inferior support and a few bug
fixes.
Specific upstream lldb revisions restored include:
SVN git
181387 779e6ac
181703 7bef4e2
182099 b31044e
182650 f2dcf35
182683 0d91b80
183862 15c1774
183929 99447a6
184177 0b2934b
184948 4dc3761
184954 007e7bc
186990 eebd175
Sponsored by: DARPA, AFRL
MFC 262186 (by emaste):
Fix mismerge in r262121
A break statement was lost in the merge. The error had no functional
impact, but restore it to reduce the diff against upstream.
MFC 262303:
Pull in r197521 from upstream clang trunk (by rdivacky):
Use the integrated assembler by default on FreeBSD/ppc and ppc64.
Requested by: jhibbits
MFC 262611:
Pull in r196874 from upstream llvm trunk:
Fix a crash that occurs when PWD is invalid.
MCJIT needs to be able to run in hostile environments, even when PWD
is invalid. There's no need to crash MCJIT in this case.
The obvious fix is to simply leave MCContext's CompilationDir empty
when PWD can't be determined. This way, MCJIT clients,
and other clients that link with LLVM don't need a valid working directory.
If we do want to guarantee valid CompilationDir, that should be done
only for clients of getCompilationDir(). This is as simple as checking
for an empty string.
The only current use of getCompilationDir is EmitGenDwarfInfo, which
won't conceivably run with an invalid working dir. However, in the
purely hypothetically and untestable case that this happens, the
AT_comp_dir will be omitted from the compilation_unit DIE.
This should help fix assertions occurring with ports-mgmt/tinderbox,
when it is using jails, and sometimes invalidates clang's current
working directory.
Reported by: decke
MFC 262809:
Pull in r203007 from upstream clang trunk:
Don't produce an alias between destructors with different calling conventions.
Fixes pr19007.
(Please note that is an LLVM PR identifier, not a FreeBSD one.)
This should fix Firefox and/or libxul crashes (due to problems with
regparm/stdcall calling conventions) on i386.
Reported by: multiple users on freebsd-current
PR: bin/187103
MFC 263048:
Repair recognition of "CC" as an alias for the C++ compiler, since it
was silently broken by upstream for a Windows-specific use-case.
Apparently some versions of CMake still rely on this archaic feature...
Reported by: rakuco
MFC 263049:
Garbage collect the old way of adding the libstdc++ include directories
in clang's InitHeaderSearch.cpp. This has been superseded by David
Chisnall's commit in r255321.
Moreover, if libc++ is used, the libstdc++ include directories should
not be in the search path at all. These directories are now only used
if you pass -stdlib=libstdc++.
Diffstat (limited to 'contrib/llvm/lib/Target/PowerPC/PPCFrameLowering.cpp')
-rw-r--r-- | contrib/llvm/lib/Target/PowerPC/PPCFrameLowering.cpp | 635 |
1 files changed, 353 insertions, 282 deletions
diff --git a/contrib/llvm/lib/Target/PowerPC/PPCFrameLowering.cpp b/contrib/llvm/lib/Target/PowerPC/PPCFrameLowering.cpp index c845909..0ac2ced 100644 --- a/contrib/llvm/lib/Target/PowerPC/PPCFrameLowering.cpp +++ b/contrib/llvm/lib/Target/PowerPC/PPCFrameLowering.cpp @@ -26,17 +26,6 @@ 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[] = { @@ -215,11 +204,13 @@ 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<const PPCRegisterInfo*>(MF.getTarget().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 @@ -235,7 +226,7 @@ unsigned PPCFrameLowering::determineFrameLayout(MachineFunction &MF, 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); @@ -305,6 +296,12 @@ 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<const PPCRegisterInfo*>(MF.getTarget().getRegisterInfo()); + bool HasBP = RegInfo->hasBasePointer(MF); + unsigned BPReg = HasBP ? (unsigned) PPC::R30 : 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 +318,13 @@ 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; + } } } @@ -332,18 +336,29 @@ void PPCFrameLowering::emitPrologue(MachineFunction &MF) const { MachineFrameInfo *MFI = MF.getFrameInfo(); const PPCInstrInfo &TII = *static_cast<const PPCInstrInfo*>(MF.getTarget().getInstrInfo()); + const PPCRegisterInfo *RegInfo = + static_cast<const PPCRegisterInfo*>(MF.getTarget().getRegisterInfo()); MachineModuleInfo &MMI = MF.getMMI(); + const MCRegisterInfo *MRI = MMI.getContext().getRegisterInfo(); DebugLoc dl; bool needsFrameMoves = MMI.hasDebugInfo() || MF.getFunction()->needsUnwindTableEntry(); + // Get processor type. + bool isPPC64 = Subtarget.isPPC64(); + // Get the ABI. + bool isDarwinABI = Subtarget.isDarwinABI(); + bool isSVR4ABI = Subtarget.isSVR4ABI(); + assert((isDarwinABI || isSVR4ABI) && + "Currently only Darwin and SVR4 ABIs are supported for PowerPC."); + // Prepare for frame info. MCSymbol *FrameLabel = 0; // 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); @@ -357,26 +372,58 @@ 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<PPCFunctionInfo>(); bool MustSaveLR = FI->mustSaveLR(); - const SmallVector<unsigned, 3> &MustSaveCRs = FI->getMustSaveCRs(); - // Do we have a frame pointer for this function? + const SmallVectorImpl<unsigned> &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 = isPPC64 ? PPC::X30 : PPC::R30; + 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!"); @@ -386,136 +433,130 @@ void PPCFrameLowering::emitPrologue(MachineFunction &MF) const { } } - 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); } - - 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 + 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); - std::vector<MachineMove> &Moves = MMI.getFrameMoves(); + } else if (!isLargeFrame) { + BuildMI(MBB, MBBI, dl, StoreUpdtInst, SPReg) + .addReg(SPReg) + .addImm(NegFrameSize) + .addReg(SPReg); + + } 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. @@ -525,25 +566,26 @@ void PPCFrameLowering::emitPrologue(MachineFunction &MF) const { 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); + MMI.addFrameInst( + MCCFIInstruction::createDefCfaOffset(FrameLabel, NegFrameSize)); 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); + MMI.addFrameInst( + MCCFIInstruction::createOffset(FrameLabel, Reg, FPOffset)); + } + + if (HasBP) { + unsigned Reg = MRI->getDwarfRegNum(BPReg, true); + MMI.addFrameInst( + MCCFIInstruction::createOffset(FrameLabel, Reg, BPOffset)); } 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); + MMI.addFrameInst( + MCCFIInstruction::createOffset(FrameLabel, Reg, LROffset)); } } @@ -551,15 +593,9 @@ void PPCFrameLowering::emitPrologue(MachineFunction &MF) const { // 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(); @@ -567,10 +603,8 @@ void PPCFrameLowering::emitPrologue(MachineFunction &MF) const { // Mark effective beginning of when frame pointer is ready. BuildMI(MBB, MBBI, dl, TII.get(PPC::PROLOG_LABEL)).addSym(ReadyLabel); - MachineLocation FPDst(HasFP ? (isPPC64 ? PPC::X31 : PPC::R31) : - (isPPC64 ? PPC::X1 : PPC::R1)); - MachineLocation FPSrc(MachineLocation::VirtualFP); - Moves.push_back(MachineMove(ReadyLabel, FPDst, FPSrc)); + unsigned Reg = MRI->getDwarfRegNum(FPReg, true); + MMI.addFrameInst(MCCFIInstruction::createDefCfaRegister(ReadyLabel, Reg)); } } @@ -590,26 +624,21 @@ 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)) { + MMI.addFrameInst(MCCFIInstruction::createOffset( + Label, MRI->getDwarfRegNum(PPC::CR2, true), 8)); + continue; } int Offset = MFI->getObjectOffset(CSI[I].getFrameIdx()); - MachineLocation CSDst(MachineLocation::VirtualFP, Offset); - MachineLocation CSSrc(Reg); - Moves.push_back(MachineMove(Label, CSDst, CSSrc)); + MMI.addFrameInst(MCCFIInstruction::createOffset( + Label, MRI->getDwarfRegNum(Reg, true), Offset)); } } } @@ -620,6 +649,8 @@ void PPCFrameLowering::emitEpilogue(MachineFunction &MF, assert(MBBI != MBB.end() && "Returning block has no terminator"); const PPCInstrInfo &TII = *static_cast<const PPCInstrInfo*>(MF.getTarget().getInstrInfo()); + const PPCRegisterInfo *RegInfo = + static_cast<const PPCRegisterInfo*>(MF.getTarget().getRegisterInfo()); unsigned RetOpcode = MBBI->getOpcode(); DebugLoc dl; @@ -633,30 +664,49 @@ 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(); + // Check if the link register (LR) has been saved. PPCFunctionInfo *FI = MF.getInfo<PPCFunctionInfo>(); bool MustSaveLR = FI->mustSaveLR(); - const SmallVector<unsigned, 3> &MustSaveCRs = FI->getMustSaveCRs(); - // Do we have a frame pointer for this function? + const SmallVectorImpl<unsigned> &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 = isPPC64 ? PPC::X30 : PPC::R30; + 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!"); @@ -666,6 +716,19 @@ void PPCFrameLowering::emitEpilogue(MachineFunction &MF, } } + 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); + } + } + bool UsesTCRet = RetOpcode == PPC::TCRETURNri || RetOpcode == PPC::TCRETURNdi || RetOpcode == PPC::TCRETURNai || @@ -687,98 +750,76 @@ 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 +827,20 @@ void PPCFrameLowering::emitEpilogue(MachineFunction &MF, MF.getFunction()->getCallingConv() == CallingConv::Fast) { PPCFunctionInfo *FI = MF.getInfo<PPCFunctionInfo>(); 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 +888,8 @@ 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<const PPCRegisterInfo*>(MF.getTarget().getRegisterInfo()); // Save and clear the LR state. PPCFunctionInfo *FI = MF.getInfo<PPCFunctionInfo>(); @@ -879,6 +914,15 @@ PPCFrameLowering::processFunctionBeforeCalleeSavedScan(MachineFunction &MF, FI->setFramePointerSaveIndex(FPSI); } + int BPSI = FI->getBasePointerSaveIndex(); + if (!BPSI && RegInfo->hasBasePointer(MF)) { + int BPOffset = getBasePointerSaveOffset(isPPC64, isDarwinABI); + // 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 && @@ -1010,6 +1054,17 @@ void PPCFrameLowering::processFunctionBeforeFrameFinalized(MachineFunction &MF, FFI->setObjectOffset(FI, LowerBound + FFI->getObjectOffset(FI)); } + const PPCRegisterInfo *RegInfo = + static_cast<const PPCRegisterInfo*>(MF.getTarget().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) { @@ -1122,8 +1177,12 @@ 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)); @@ -1151,6 +1210,12 @@ PPCFrameLowering::spillCalleeSavedRegisters(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; + // CR2 through CR4 are the nonvolatile CR fields. bool IsCRField = PPC::CR2 <= Reg && Reg <= PPC::CR4; @@ -1212,7 +1277,7 @@ restoreCRs(bool isPPC64, bool is31, MBB.insert(MI, addFrameReference(BuildMI(*MF, DL, TII.get(PPC::LWZ), PPC::R12), CSI[CSIIndex].getFrameIdx())); - RestoreOp = PPC::MTCRF; + RestoreOp = PPC::MTOCRF; MoveReg = PPC::R12; } @@ -1300,6 +1365,12 @@ 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 |