From: Dale Johannesen Date: Thu, 26 Jun 2008 01:51:13 +0000 (+0000) Subject: Fixes the last x86-64 test failure in compat.exp: X-Git-Url: http://demsky.eecs.uci.edu/git/?a=commitdiff_plain;h=b5dae003252d8e650a32bfdf33cba5aed8e41e40;p=oota-llvm.git Fixes the last x86-64 test failure in compat.exp: <16 x float> is 64-byte aligned (for some reason), which gets us into the stack realignment code. The computation changing FP-relative offsets to SP-relative was broken, assiging a spill temp to a location also used for parameter passing. This fixes it by rounding up the stack frame to a multiple of the largest alignment (I concluded it wasn't fixable without doing this, but I'm not very sure.) git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@52750 91177308-0d34-0410-b5e6-96231b3b80d8 --- diff --git a/include/llvm/Target/TargetRegisterInfo.h b/include/llvm/Target/TargetRegisterInfo.h index 08b9bfc7649..2deede235f7 100644 --- a/include/llvm/Target/TargetRegisterInfo.h +++ b/include/llvm/Target/TargetRegisterInfo.h @@ -516,6 +516,13 @@ public: return !hasFP(MF); } + // needsStackRealignment - true if storage within the function requires the + // stack pointer to be aligned more than the normal calling convention calls + // for. + virtual bool needsStackRealignment(const MachineFunction &MF) const { + return false; + } + /// getCallFrameSetup/DestroyOpcode - These methods return the opcode of the /// frame setup/destroy instructions if they exist (-1 otherwise). Some /// targets use pseudo instructions in order to abstract away the difference diff --git a/lib/CodeGen/PrologEpilogInserter.cpp b/lib/CodeGen/PrologEpilogInserter.cpp index c52da321c1e..10ba9c0256f 100644 --- a/lib/CodeGen/PrologEpilogInserter.cpp +++ b/lib/CodeGen/PrologEpilogInserter.cpp @@ -459,15 +459,19 @@ void PEI::calculateFrameObjectOffsets(MachineFunction &Fn) { // 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 suitable 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())) { + (FFI->hasCalls() || FFI->hasVarSizedObjects() || + RegInfo->needsStackRealignment(Fn))) { // 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 = TFI.getStackAlignment() - 1; + unsigned AlignMask = std::max(TFI.getStackAlignment(),MaxAlign) - 1; Offset = (Offset + AlignMask) & ~uint64_t(AlignMask); } diff --git a/lib/Target/X86/X86RegisterInfo.cpp b/lib/Target/X86/X86RegisterInfo.cpp index 0bac85094ca..b48cee3ac33 100644 --- a/lib/Target/X86/X86RegisterInfo.cpp +++ b/lib/Target/X86/X86RegisterInfo.cpp @@ -302,11 +302,9 @@ X86RegisterInfo::getFrameIndexOffset(MachineFunction &MF, int FI) const { // Skip the saved EBP Offset += SlotSize; else { - unsigned MaxAlign = MF.getFrameInfo()->getMaxAlignment(); - uint64_t FrameSize = - (StackSize - SlotSize + MaxAlign - 1)/MaxAlign*MaxAlign; - - return Offset + FrameSize - SlotSize; + unsigned Align = MF.getFrameInfo()->getObjectAlignment(FI); + assert( (-(Offset + StackSize)) % Align == 0); + return Offset + StackSize; } // FIXME: Support tail calls diff --git a/utils/TableGen/RegisterInfoEmitter.cpp b/utils/TableGen/RegisterInfoEmitter.cpp index 92773359a9c..2f253ff34e5 100644 --- a/utils/TableGen/RegisterInfoEmitter.cpp +++ b/utils/TableGen/RegisterInfoEmitter.cpp @@ -64,6 +64,8 @@ void RegisterInfoEmitter::runHeader(std::ostream &OS) { << " virtual int getDwarfRegNumFull(unsigned RegNum, " << "unsigned Flavour) const;\n" << " virtual int getDwarfRegNum(unsigned RegNum, bool isEH) const = 0;\n" + << " virtual bool needsStackRealignment(const MachineFunction &) const\n" + << " { return false; }\n" << " unsigned getSubReg(unsigned RegNo, unsigned Index) const;\n" << "};\n\n";