summaryrefslogtreecommitdiff
path: root/lib/Target/X86/X86CallFrameOptimization.cpp
diff options
context:
space:
mode:
Diffstat (limited to 'lib/Target/X86/X86CallFrameOptimization.cpp')
-rw-r--r--lib/Target/X86/X86CallFrameOptimization.cpp131
1 files changed, 85 insertions, 46 deletions
diff --git a/lib/Target/X86/X86CallFrameOptimization.cpp b/lib/Target/X86/X86CallFrameOptimization.cpp
index 765af67de160..522dc7926b94 100644
--- a/lib/Target/X86/X86CallFrameOptimization.cpp
+++ b/lib/Target/X86/X86CallFrameOptimization.cpp
@@ -34,14 +34,14 @@
#include "llvm/CodeGen/MachineInstrBuilder.h"
#include "llvm/CodeGen/MachineOperand.h"
#include "llvm/CodeGen/MachineRegisterInfo.h"
+#include "llvm/CodeGen/TargetInstrInfo.h"
+#include "llvm/CodeGen/TargetRegisterInfo.h"
#include "llvm/IR/DebugLoc.h"
#include "llvm/IR/Function.h"
#include "llvm/MC/MCDwarf.h"
#include "llvm/Support/CommandLine.h"
#include "llvm/Support/ErrorHandling.h"
#include "llvm/Support/MathExtras.h"
-#include "llvm/Target/TargetInstrInfo.h"
-#include "llvm/Target/TargetRegisterInfo.h"
#include <cassert>
#include <cstddef>
#include <cstdint>
@@ -56,18 +56,27 @@ static cl::opt<bool>
cl::desc("Avoid optimizing x86 call frames for size"),
cl::init(false), cl::Hidden);
+namespace llvm {
+void initializeX86CallFrameOptimizationPass(PassRegistry &);
+}
+
namespace {
class X86CallFrameOptimization : public MachineFunctionPass {
public:
- X86CallFrameOptimization() : MachineFunctionPass(ID) {}
+ X86CallFrameOptimization() : MachineFunctionPass(ID) {
+ initializeX86CallFrameOptimizationPass(
+ *PassRegistry::getPassRegistry());
+ }
bool runOnMachineFunction(MachineFunction &MF) override;
+ static char ID;
+
private:
// Information we know about a particular call site
struct CallContext {
- CallContext() : FrameSetup(nullptr), MovVector(4, nullptr) {}
+ CallContext() : FrameSetup(nullptr), ArgStoreVector(4, nullptr) {}
// Iterator referring to the frame setup instruction
MachineBasicBlock::iterator FrameSetup;
@@ -81,8 +90,8 @@ private:
// The total displacement of all passed parameters
int64_t ExpectedDist = 0;
- // The sequence of movs used to pass the parameters
- SmallVector<MachineInstr *, 4> MovVector;
+ // The sequence of storing instructions used to pass the parameters
+ SmallVector<MachineInstr *, 4> ArgStoreVector;
// True if this call site has no stack parameters
bool NoStackParams = false;
@@ -120,12 +129,12 @@ private:
MachineRegisterInfo *MRI;
unsigned SlotSize;
unsigned Log2SlotSize;
- static char ID;
};
-char X86CallFrameOptimization::ID = 0;
-
} // end anonymous namespace
+char X86CallFrameOptimization::ID = 0;
+INITIALIZE_PASS(X86CallFrameOptimization, DEBUG_TYPE,
+ "X86 Call Frame Optimization", false, false)
// This checks whether the transformation is legal.
// Also returns false in cases where it's potentially legal, but
@@ -139,7 +148,7 @@ bool X86CallFrameOptimization::isLegal(MachineFunction &MF) {
// is a danger of that being generated.
if (STI->isTargetDarwin() &&
(!MF.getLandingPads().empty() ||
- (MF.getFunction()->needsUnwindTableEntry() && !TFL->hasFP(MF))))
+ (MF.getFunction().needsUnwindTableEntry() && !TFL->hasFP(MF))))
return false;
// It is not valid to change the stack pointer outside the prolog/epilog
@@ -234,7 +243,7 @@ bool X86CallFrameOptimization::runOnMachineFunction(MachineFunction &MF) {
assert(isPowerOf2_32(SlotSize) && "Expect power of 2 stack slot size");
Log2SlotSize = Log2_32(SlotSize);
- if (skipFunction(*MF.getFunction()) || !isLegal(MF))
+ if (skipFunction(MF.getFunction()) || !isLegal(MF))
return false;
unsigned FrameSetupOpcode = TII->getCallFrameSetupOpcode();
@@ -271,11 +280,27 @@ X86CallFrameOptimization::classifyInstruction(
if (MI == MBB.end())
return Exit;
- // The instructions we actually care about are movs onto the stack
- int Opcode = MI->getOpcode();
- if (Opcode == X86::MOV32mi || Opcode == X86::MOV32mr ||
- Opcode == X86::MOV64mi32 || Opcode == X86::MOV64mr)
- return Convert;
+ // The instructions we actually care about are movs onto the stack or special
+ // cases of constant-stores to stack
+ switch (MI->getOpcode()) {
+ case X86::AND16mi8:
+ case X86::AND32mi8:
+ case X86::AND64mi8: {
+ MachineOperand ImmOp = MI->getOperand(X86::AddrNumOperands);
+ return ImmOp.getImm() == 0 ? Convert : Exit;
+ }
+ case X86::OR16mi8:
+ case X86::OR32mi8:
+ case X86::OR64mi8: {
+ MachineOperand ImmOp = MI->getOperand(X86::AddrNumOperands);
+ return ImmOp.getImm() == -1 ? Convert : Exit;
+ }
+ case X86::MOV32mi:
+ case X86::MOV32mr:
+ case X86::MOV64mi32:
+ case X86::MOV64mr:
+ return Convert;
+ }
// Not all calling conventions have only stack MOVs between the stack
// adjust and the call.
@@ -354,32 +379,40 @@ void X86CallFrameOptimization::collectCallInfo(MachineFunction &MF,
++I;
unsigned StackPtr = RegInfo.getStackRegister();
+ auto StackPtrCopyInst = MBB.end();
// SelectionDAG (but not FastISel) inserts a copy of ESP into a virtual
- // register here. If it's there, use that virtual register as stack pointer
- // instead.
- if (I->isCopy() && I->getOperand(0).isReg() && I->getOperand(1).isReg() &&
- I->getOperand(1).getReg() == StackPtr) {
- Context.SPCopy = &*I++;
- StackPtr = Context.SPCopy->getOperand(0).getReg();
- }
+ // register. If it's there, use that virtual register as stack pointer
+ // instead. Also, we need to locate this instruction so that we can later
+ // safely ignore it while doing the conservative processing of the call chain.
+ // The COPY can be located anywhere between the call-frame setup
+ // instruction and its first use. We use the call instruction as a boundary
+ // because it is usually cheaper to check if an instruction is a call than
+ // checking if an instruction uses a register.
+ for (auto J = I; !J->isCall(); ++J)
+ if (J->isCopy() && J->getOperand(0).isReg() && J->getOperand(1).isReg() &&
+ J->getOperand(1).getReg() == StackPtr) {
+ StackPtrCopyInst = J;
+ Context.SPCopy = &*J++;
+ StackPtr = Context.SPCopy->getOperand(0).getReg();
+ break;
+ }
// Scan the call setup sequence for the pattern we're looking for.
// We only handle a simple case - a sequence of store instructions that
// push a sequence of stack-slot-aligned values onto the stack, with
// no gaps between them.
if (MaxAdjust > 4)
- Context.MovVector.resize(MaxAdjust, nullptr);
+ Context.ArgStoreVector.resize(MaxAdjust, nullptr);
- InstClassification Classification;
DenseSet<unsigned int> UsedRegs;
- while ((Classification = classifyInstruction(MBB, I, RegInfo, UsedRegs)) !=
- Exit) {
- if (Classification == Skip) {
- ++I;
+ for (InstClassification Classification = Skip; Classification != Exit; ++I) {
+ // If this is the COPY of the stack pointer, it's ok to ignore.
+ if (I == StackPtrCopyInst)
+ continue;
+ Classification = classifyInstruction(MBB, I, RegInfo, UsedRegs);
+ if (Classification != Convert)
continue;
- }
-
// We know the instruction has a supported store opcode.
// We only want movs of the form:
// mov imm/reg, k(%StackPtr)
@@ -407,13 +440,13 @@ void X86CallFrameOptimization::collectCallInfo(MachineFunction &MF,
return;
StackDisp >>= Log2SlotSize;
- assert((size_t)StackDisp < Context.MovVector.size() &&
+ assert((size_t)StackDisp < Context.ArgStoreVector.size() &&
"Function call has more parameters than the stack is adjusted for.");
// If the same stack slot is being filled twice, something's fishy.
- if (Context.MovVector[StackDisp] != nullptr)
+ if (Context.ArgStoreVector[StackDisp] != nullptr)
return;
- Context.MovVector[StackDisp] = &*I;
+ Context.ArgStoreVector[StackDisp] = &*I;
for (const MachineOperand &MO : I->uses()) {
if (!MO.isReg())
@@ -422,10 +455,10 @@ void X86CallFrameOptimization::collectCallInfo(MachineFunction &MF,
if (RegInfo.isPhysicalRegister(Reg))
UsedRegs.insert(Reg);
}
-
- ++I;
}
+ --I;
+
// We now expect the end of the sequence. If we stopped early,
// or reached the end of the block without finding a call, bail.
if (I == MBB.end() || !I->isCall())
@@ -436,14 +469,14 @@ void X86CallFrameOptimization::collectCallInfo(MachineFunction &MF,
return;
// Now, go through the vector, and see that we don't have any gaps,
- // but only a series of MOVs.
- auto MMI = Context.MovVector.begin(), MME = Context.MovVector.end();
+ // but only a series of storing instructions.
+ auto MMI = Context.ArgStoreVector.begin(), MME = Context.ArgStoreVector.end();
for (; MMI != MME; ++MMI, Context.ExpectedDist += SlotSize)
if (*MMI == nullptr)
break;
// If the call had no parameters, do nothing
- if (MMI == Context.MovVector.begin())
+ if (MMI == Context.ArgStoreVector.begin())
return;
// We are either at the last parameter, or a gap.
@@ -466,17 +499,23 @@ void X86CallFrameOptimization::adjustCallSequence(MachineFunction &MF,
DebugLoc DL = FrameSetup->getDebugLoc();
bool Is64Bit = STI->is64Bit();
- // Now, iterate through the vector in reverse order, and replace the movs
- // with pushes. MOVmi/MOVmr doesn't have any defs, so no need to
+ // Now, iterate through the vector in reverse order, and replace the store to
+ // stack with pushes. MOVmi/MOVmr doesn't have any defs, so no need to
// replace uses.
for (int Idx = (Context.ExpectedDist >> Log2SlotSize) - 1; Idx >= 0; --Idx) {
- MachineBasicBlock::iterator MOV = *Context.MovVector[Idx];
- MachineOperand PushOp = MOV->getOperand(X86::AddrNumOperands);
+ MachineBasicBlock::iterator Store = *Context.ArgStoreVector[Idx];
+ MachineOperand PushOp = Store->getOperand(X86::AddrNumOperands);
MachineBasicBlock::iterator Push = nullptr;
unsigned PushOpcode;
- switch (MOV->getOpcode()) {
+ switch (Store->getOpcode()) {
default:
llvm_unreachable("Unexpected Opcode!");
+ case X86::AND16mi8:
+ case X86::AND32mi8:
+ case X86::AND64mi8:
+ case X86::OR16mi8:
+ case X86::OR32mi8:
+ case X86::OR64mi8:
case X86::MOV32mi:
case X86::MOV64mi32:
PushOpcode = Is64Bit ? X86::PUSH64i32 : X86::PUSHi32;
@@ -497,7 +536,7 @@ void X86CallFrameOptimization::adjustCallSequence(MachineFunction &MF,
// If storing a 32-bit vreg on 64-bit targets, extend to a 64-bit vreg
// in preparation for the PUSH64. The upper 32 bits can be undef.
- if (Is64Bit && MOV->getOpcode() == X86::MOV32mr) {
+ if (Is64Bit && Store->getOpcode() == X86::MOV32mr) {
unsigned UndefReg = MRI->createVirtualRegister(&X86::GR64RegClass);
Reg = MRI->createVirtualRegister(&X86::GR64RegClass);
BuildMI(MBB, Context.Call, DL, TII->get(X86::IMPLICIT_DEF), UndefReg);
@@ -541,7 +580,7 @@ void X86CallFrameOptimization::adjustCallSequence(MachineFunction &MF,
MBB, std::next(Push), DL,
MCCFIInstruction::createAdjustCfaOffset(nullptr, SlotSize));
- MBB.erase(MOV);
+ MBB.erase(Store);
}
// The stack-pointer copy is no longer used in the call sequences.
generated by cgit v1.2.3 (git 2.25.1) at 2025-11-06 09:41:06 +0000