[llvm] r238035 - Reapply r238011 with a fix for the trap instruction.
Quentin Colombet
qcolombet at apple.com
Fri May 22 11:10:47 PDT 2015
Author: qcolombet
Date: Fri May 22 13:10:47 2015
New Revision: 238035
URL: http://llvm.org/viewvc/llvm-project?rev=238035&view=rev
Log:
Reapply r238011 with a fix for the trap instruction.
The problem was that I slipped a change required for shrink-wrapping, namely I
used getFirstTerminator instead of the getLastNonDebugInstr that was here before
the refactoring, whereas the surrounding code is not yet patched for that.
Original message:
[X86] Refactor the prologue emission to prepare for shrink-wrapping.
- Add a late pass to expand pseudo instructions (tail call and EH returns).
Instead of doing it in the prologue emission.
- Factor some static methods in X86FrameLowering to ease code sharing.
NFC.
Related to <rdar://problem/20821487>
Added:
llvm/trunk/lib/Target/X86/X86ExpandPseudo.cpp
- copied unchanged from r238010, llvm/trunk/lib/Target/X86/X86ExpandPseudo.cpp
llvm/trunk/test/CodeGen/X86/x86-framelowering-trap.ll
Modified:
llvm/trunk/lib/Target/X86/CMakeLists.txt
llvm/trunk/lib/Target/X86/X86.h
llvm/trunk/lib/Target/X86/X86FrameLowering.cpp
llvm/trunk/lib/Target/X86/X86FrameLowering.h
llvm/trunk/lib/Target/X86/X86RegisterInfo.cpp
llvm/trunk/lib/Target/X86/X86TargetMachine.cpp
Modified: llvm/trunk/lib/Target/X86/CMakeLists.txt
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/CMakeLists.txt?rev=238035&r1=238034&r2=238035&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/CMakeLists.txt (original)
+++ llvm/trunk/lib/Target/X86/CMakeLists.txt Fri May 22 13:10:47 2015
@@ -15,6 +15,7 @@ add_public_tablegen_target(X86CommonTabl
set(sources
X86AsmPrinter.cpp
X86CallFrameOptimization.cpp
+ X86ExpandPseudo.cpp
X86FastISel.cpp
X86FloatingPoint.cpp
X86FrameLowering.cpp
Modified: llvm/trunk/lib/Target/X86/X86.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86.h?rev=238035&r1=238034&r2=238035&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86.h (original)
+++ llvm/trunk/lib/Target/X86/X86.h Fri May 22 13:10:47 2015
@@ -75,6 +75,11 @@ FunctionPass *createX86CallFrameOptimiza
/// preparation.
FunctionPass *createX86WinEHStatePass();
+/// Return a Machine IR pass that expands X86-specific pseudo
+/// instructions into a sequence of actual instructions. This pass
+/// must run after prologue/epilogue insertion and before lowering
+/// the MachineInstr to MC.
+FunctionPass *createX86ExpandPseudoPass();
} // End llvm namespace
#endif
Modified: llvm/trunk/lib/Target/X86/X86FrameLowering.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86FrameLowering.cpp?rev=238035&r1=238034&r2=238035&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86FrameLowering.cpp (original)
+++ llvm/trunk/lib/Target/X86/X86FrameLowering.cpp Fri May 22 13:10:47 2015
@@ -205,11 +205,12 @@ static bool isEAXLiveIn(MachineFunction
/// emitSPUpdate - Emit a series of instructions to increment / decrement the
/// stack pointer by a constant value.
-static
-void emitSPUpdate(MachineBasicBlock &MBB, MachineBasicBlock::iterator &MBBI,
- unsigned StackPtr, int64_t NumBytes,
- bool Is64BitTarget, bool Is64BitStackPtr, bool UseLEA,
- const TargetInstrInfo &TII, const TargetRegisterInfo &TRI) {
+void X86FrameLowering::emitSPUpdate(MachineBasicBlock &MBB,
+ MachineBasicBlock::iterator &MBBI,
+ unsigned StackPtr, int64_t NumBytes,
+ bool Is64BitTarget, bool Is64BitStackPtr,
+ bool UseLEA, const TargetInstrInfo &TII,
+ const TargetRegisterInfo &TRI) {
bool isSub = NumBytes < 0;
uint64_t Offset = isSub ? -NumBytes : NumBytes;
unsigned Opc;
@@ -312,13 +313,10 @@ void mergeSPUpdatesUp(MachineBasicBlock
}
}
-/// mergeSPUpdates - Checks the instruction before/after the passed
-/// instruction. If it is an ADD/SUB/LEA instruction it is deleted argument and
-/// the stack adjustment is returned as a positive value for ADD/LEA and a
-/// negative for SUB.
-static int mergeSPUpdates(MachineBasicBlock &MBB,
- MachineBasicBlock::iterator &MBBI, unsigned StackPtr,
- bool doMergeWithPrevious) {
+int X86FrameLowering::mergeSPUpdates(MachineBasicBlock &MBB,
+ MachineBasicBlock::iterator &MBBI,
+ unsigned StackPtr,
+ bool doMergeWithPrevious) {
if ((doMergeWithPrevious && MBBI == MBB.begin()) ||
(!doMergeWithPrevious && MBBI == MBB.end()))
return 0;
@@ -967,6 +965,17 @@ void X86FrameLowering::emitPrologue(Mach
}
}
+bool X86FrameLowering::useLEAForSPInProlog(const MachineFunction &MF) const {
+ // We can't use LEA instructions for adjusting the stack pointer if this is a
+ // leaf function in the Win64 ABI. Only ADD instructions may be used to
+ // deallocate the stack.
+ // This means that we can use LEA for SP in two situations:
+ // 1. We *aren't* using the Win64 ABI which means we are free to use LEA.
+ // 2. We *have* a frame pointer which means we are permitted to use LEA.
+ return MF.getSubtarget<X86Subtarget>().useLeaForSP() &&
+ (!MF.getTarget().getMCAsmInfo()->usesWindowsCFI() || hasFP(MF));
+}
+
void X86FrameLowering::emitEpilogue(MachineFunction &MF,
MachineBasicBlock &MBB) const {
const MachineFrameInfo *MFI = MF.getFrameInfo();
@@ -976,12 +985,10 @@ void X86FrameLowering::emitEpilogue(Mach
const TargetInstrInfo &TII = *STI.getInstrInfo();
MachineBasicBlock::iterator MBBI = MBB.getLastNonDebugInstr();
assert(MBBI != MBB.end() && "Returning block has no instructions");
- unsigned RetOpcode = MBBI->getOpcode();
DebugLoc DL = MBBI->getDebugLoc();
bool Is64Bit = STI.is64Bit();
// standard x86_64 and NaCl use 64-bit frame/stack pointers, x32 - 32-bit.
const bool Uses64BitFramePtr = STI.isTarget64BitLP64() || STI.isTargetNaCl64();
- bool HasFP = hasFP(MF);
const bool Is64BitILP32 = STI.isTarget64BitILP32();
unsigned SlotSize = RegInfo->getSlotSize();
unsigned FramePtr = RegInfo->getFrameRegister(MF);
@@ -992,22 +999,9 @@ void X86FrameLowering::emitEpilogue(Mach
bool IsWinEH = MF.getTarget().getMCAsmInfo()->usesWindowsCFI();
bool NeedsWinEH = IsWinEH && MF.getFunction()->needsUnwindTableEntry();
- bool UseLEAForSP = false;
+ bool UseLEAForSP = useLEAForSPInProlog(MF);
- // We can't use LEA instructions for adjusting the stack pointer if this is a
- // leaf function in the Win64 ABI. Only ADD instructions may be used to
- // deallocate the stack.
- if (STI.useLeaForSP()) {
- if (!IsWinEH) {
- // We *aren't* using the Win64 ABI which means we are free to use LEA.
- UseLEAForSP = true;
- } else if (HasFP) {
- // We *have* a frame pointer which means we are permitted to use LEA.
- UseLEAForSP = true;
- }
- }
-
- switch (RetOpcode) {
+ switch (MBBI->getOpcode()) {
default:
llvm_unreachable("Can only insert epilogue into returning blocks");
case X86::RETQ:
@@ -1112,88 +1106,15 @@ void X86FrameLowering::emitEpilogue(Mach
if (NeedsWinEH)
BuildMI(MBB, MBBI, DL, TII.get(X86::SEH_Epilogue));
- // We're returning from function via eh_return.
- if (RetOpcode == X86::EH_RETURN || RetOpcode == X86::EH_RETURN64) {
- MBBI = MBB.getLastNonDebugInstr();
- MachineOperand &DestAddr = MBBI->getOperand(0);
- assert(DestAddr.isReg() && "Offset should be in register!");
- BuildMI(MBB, MBBI, DL,
- TII.get(Uses64BitFramePtr ? X86::MOV64rr : X86::MOV32rr),
- StackPtr).addReg(DestAddr.getReg());
- } else if (RetOpcode == X86::TCRETURNri || RetOpcode == X86::TCRETURNdi ||
- RetOpcode == X86::TCRETURNmi ||
- RetOpcode == X86::TCRETURNri64 || RetOpcode == X86::TCRETURNdi64 ||
- RetOpcode == X86::TCRETURNmi64) {
- bool isMem = RetOpcode == X86::TCRETURNmi || RetOpcode == X86::TCRETURNmi64;
- // Tail call return: adjust the stack pointer and jump to callee.
- MBBI = MBB.getLastNonDebugInstr();
- MachineOperand &JumpTarget = MBBI->getOperand(0);
- MachineOperand &StackAdjust = MBBI->getOperand(isMem ? 5 : 1);
- assert(StackAdjust.isImm() && "Expecting immediate value.");
-
- // Adjust stack pointer.
- int StackAdj = StackAdjust.getImm();
- int MaxTCDelta = X86FI->getTCReturnAddrDelta();
- int Offset = 0;
- assert(MaxTCDelta <= 0 && "MaxTCDelta should never be positive");
-
- // Incoporate the retaddr area.
- Offset = StackAdj-MaxTCDelta;
- assert(Offset >= 0 && "Offset should never be negative");
-
- if (Offset) {
- // Check for possible merge with preceding ADD instruction.
- Offset += mergeSPUpdates(MBB, MBBI, StackPtr, true);
- emitSPUpdate(MBB, MBBI, StackPtr, Offset, Is64Bit, Uses64BitFramePtr,
- UseLEAForSP, TII, *RegInfo);
- }
-
- // Jump to label or value in register.
- bool IsWin64 = STI.isTargetWin64();
- if (RetOpcode == X86::TCRETURNdi || RetOpcode == X86::TCRETURNdi64) {
- unsigned Op = (RetOpcode == X86::TCRETURNdi)
- ? X86::TAILJMPd
- : (IsWin64 ? X86::TAILJMPd64_REX : X86::TAILJMPd64);
- MachineInstrBuilder MIB = BuildMI(MBB, MBBI, DL, TII.get(Op));
- if (JumpTarget.isGlobal())
- MIB.addGlobalAddress(JumpTarget.getGlobal(), JumpTarget.getOffset(),
- JumpTarget.getTargetFlags());
- else {
- assert(JumpTarget.isSymbol());
- MIB.addExternalSymbol(JumpTarget.getSymbolName(),
- JumpTarget.getTargetFlags());
- }
- } else if (RetOpcode == X86::TCRETURNmi || RetOpcode == X86::TCRETURNmi64) {
- unsigned Op = (RetOpcode == X86::TCRETURNmi)
- ? X86::TAILJMPm
- : (IsWin64 ? X86::TAILJMPm64_REX : X86::TAILJMPm64);
- MachineInstrBuilder MIB = BuildMI(MBB, MBBI, DL, TII.get(Op));
- for (unsigned i = 0; i != 5; ++i)
- MIB.addOperand(MBBI->getOperand(i));
- } else if (RetOpcode == X86::TCRETURNri64) {
- BuildMI(MBB, MBBI, DL,
- TII.get(IsWin64 ? X86::TAILJMPr64_REX : X86::TAILJMPr64))
- .addReg(JumpTarget.getReg(), RegState::Kill);
- } else {
- BuildMI(MBB, MBBI, DL, TII.get(X86::TAILJMPr)).
- addReg(JumpTarget.getReg(), RegState::Kill);
- }
-
- MachineInstr *NewMI = std::prev(MBBI);
- NewMI->copyImplicitOps(MF, MBBI);
-
- // Delete the pseudo instruction TCRETURN.
- MBB.erase(MBBI);
- } else if ((RetOpcode == X86::RETQ || RetOpcode == X86::RETL ||
- RetOpcode == X86::RETIQ || RetOpcode == X86::RETIL) &&
- (X86FI->getTCReturnAddrDelta() < 0)) {
- // Add the return addr area delta back since we are not tail calling.
- int delta = -1*X86FI->getTCReturnAddrDelta();
- MBBI = MBB.getLastNonDebugInstr();
+ // Add the return addr area delta back since we are not tail calling.
+ int Offset = -1 * X86FI->getTCReturnAddrDelta();
+ assert(Offset >= 0 && "TCDelta should never be positive");
+ if (Offset) {
+ MBBI = MBB.getFirstTerminator();
// Check for possible merge with preceding ADD instruction.
- delta += mergeSPUpdates(MBB, MBBI, StackPtr, true);
- emitSPUpdate(MBB, MBBI, StackPtr, delta, Is64Bit, Uses64BitFramePtr,
+ Offset += mergeSPUpdates(MBB, MBBI, StackPtr, true);
+ emitSPUpdate(MBB, MBBI, StackPtr, Offset, Is64Bit, Uses64BitFramePtr,
UseLEAForSP, TII, *RegInfo);
}
}
Modified: llvm/trunk/lib/Target/X86/X86FrameLowering.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86FrameLowering.h?rev=238035&r1=238034&r2=238035&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86FrameLowering.h (original)
+++ llvm/trunk/lib/Target/X86/X86FrameLowering.h Fri May 22 13:10:47 2015
@@ -79,6 +79,26 @@ public:
MachineBasicBlock &MBB,
MachineBasicBlock::iterator MI) const override;
+ /// Check the instruction before/after the passed instruction. If
+ /// it is an ADD/SUB/LEA instruction it is deleted argument and the
+ /// stack adjustment is returned as a positive value for ADD/LEA and
+ /// a negative for SUB.
+ static int mergeSPUpdates(MachineBasicBlock &MBB,
+ MachineBasicBlock::iterator &MBBI,
+ unsigned StackPtr, bool doMergeWithPrevious);
+
+ /// Emit a series of instructions to increment / decrement the stack
+ /// pointer by a constant value.
+ static void emitSPUpdate(MachineBasicBlock &MBB,
+ MachineBasicBlock::iterator &MBBI, unsigned StackPtr,
+ int64_t NumBytes, bool Is64BitTarget,
+ bool Is64BitStackPtr, bool UseLEA,
+ const TargetInstrInfo &TII,
+ const TargetRegisterInfo &TRI);
+
+ /// Check that LEA can be use on SP in a prologue sequence for \p MF.
+ bool useLEAForSPInProlog(const MachineFunction &MF) const;
+
private:
/// convertArgMovsToPushes - This method tries to convert a call sequence
/// that uses sub and mov instructions to put the argument onto the stack
Modified: llvm/trunk/lib/Target/X86/X86RegisterInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86RegisterInfo.cpp?rev=238035&r1=238034&r2=238035&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86RegisterInfo.cpp (original)
+++ llvm/trunk/lib/Target/X86/X86RegisterInfo.cpp Fri May 22 13:10:47 2015
@@ -492,7 +492,8 @@ X86RegisterInfo::eliminateFrameIndex(Mac
unsigned BasePtr;
unsigned Opc = MI.getOpcode();
- bool AfterFPPop = Opc == X86::TAILJMPm64 || Opc == X86::TAILJMPm;
+ bool AfterFPPop = Opc == X86::TAILJMPm64 || Opc == X86::TAILJMPm ||
+ Opc == X86::TCRETURNmi || Opc == X86::TCRETURNmi64;
if (hasBasePointer(MF))
BasePtr = (FrameIndex < 0 ? FramePtr : getBaseRegister());
else if (needsStackRealignment(MF))
Modified: llvm/trunk/lib/Target/X86/X86TargetMachine.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86TargetMachine.cpp?rev=238035&r1=238034&r2=238035&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86TargetMachine.cpp (original)
+++ llvm/trunk/lib/Target/X86/X86TargetMachine.cpp Fri May 22 13:10:47 2015
@@ -187,6 +187,7 @@ public:
void addPreRegAlloc() override;
void addPostRegAlloc() override;
void addPreEmitPass() override;
+ void addPreSched2() override;
};
} // namespace
@@ -235,6 +236,8 @@ void X86PassConfig::addPostRegAlloc() {
addPass(createX86FloatingPointStackifierPass());
}
+void X86PassConfig::addPreSched2() { addPass(createX86ExpandPseudoPass()); }
+
void X86PassConfig::addPreEmitPass() {
if (getOptLevel() != CodeGenOpt::None)
addPass(createExecutionDependencyFixPass(&X86::VR128RegClass));
Added: llvm/trunk/test/CodeGen/X86/x86-framelowering-trap.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/x86-framelowering-trap.ll?rev=238035&view=auto
==============================================================================
--- llvm/trunk/test/CodeGen/X86/x86-framelowering-trap.ll (added)
+++ llvm/trunk/test/CodeGen/X86/x86-framelowering-trap.ll Fri May 22 13:10:47 2015
@@ -0,0 +1,15 @@
+; RUN: llc %s -o - | FileCheck %s
+target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
+target triple = "x86_64-unknown-linux-gnu"
+
+; CHECK-LABEL: bar:
+; CHECK: ud2
+; CHECK-NEXT: retq
+define void @bar() {
+entry:
+ call void @llvm.trap()
+ ret void
+}
+
+; Function Attrs: noreturn nounwind
+declare void @llvm.trap()
More information about the llvm-commits
mailing list