#include "llvm/Target/TargetOptions.h"
#include <limits>
-#define GET_INSTRINFO_CTOR
+#define GET_INSTRINFO_CTOR_DTOR
#include "X86GenInstrInfo.inc"
using namespace llvm;
uint16_t Flags;
};
+// Pin the vtable to this file.
+void X86InstrInfo::anchor() {}
+
X86InstrInfo::X86InstrInfo(X86TargetMachine &tm)
: X86GenInstrInfo((tm.getSubtarget<X86Subtarget>().is64Bit()
? X86::ADJCALLSTACKDOWN64
return 0;
}
+inline static bool MaskRegClassContains(unsigned Reg) {
+ return X86::VK8RegClass.contains(Reg) ||
+ X86::VK16RegClass.contains(Reg) ||
+ X86::VK1RegClass.contains(Reg);
+}
static
unsigned copyPhysRegOpcode_AVX512(unsigned& DestReg, unsigned& SrcReg) {
if (X86::VR128XRegClass.contains(DestReg, SrcReg) ||
SrcReg = get512BitSuperRegister(SrcReg);
return X86::VMOVAPSZrr;
}
- if ((X86::VK8RegClass.contains(DestReg) ||
- X86::VK16RegClass.contains(DestReg)) &&
- (X86::VK8RegClass.contains(SrcReg) ||
- X86::VK16RegClass.contains(SrcReg)))
+ if (MaskRegClassContains(DestReg) &&
+ MaskRegClassContains(SrcReg))
return X86::KMOVWkk;
+ if (MaskRegClassContains(DestReg) &&
+ (X86::GR32RegClass.contains(SrcReg) ||
+ X86::GR16RegClass.contains(SrcReg) ||
+ X86::GR8RegClass.contains(SrcReg))) {
+ SrcReg = getX86SubSuperRegister(SrcReg, MVT::i32);
+ return X86::KMOVWkr;
+ }
+ if ((X86::GR32RegClass.contains(DestReg) ||
+ X86::GR16RegClass.contains(DestReg) ||
+ X86::GR8RegClass.contains(DestReg)) &&
+ MaskRegClassContains(SrcReg)) {
+ DestReg = getX86SubSuperRegister(DestReg, MVT::i32);
+ return X86::KMOVWrk;
+ }
return 0;
}
assert(X86::RFP80RegClass.hasSubClassEq(RC) && "Unknown 10-byte regclass");
return load ? X86::LD_Fp80m : X86::ST_FpP80m;
case 16: {
- assert(X86::VR128RegClass.hasSubClassEq(RC) && "Unknown 16-byte regclass");
+ assert((X86::VR128RegClass.hasSubClassEq(RC) ||
+ X86::VR128XRegClass.hasSubClassEq(RC))&& "Unknown 16-byte regclass");
// If stack is realigned we can use aligned stores.
if (isStackAligned)
return load ?
(HasAVX ? X86::VMOVUPSmr : X86::MOVUPSmr);
}
case 32:
- assert(X86::VR256RegClass.hasSubClassEq(RC) && "Unknown 32-byte regclass");
+ assert((X86::VR256RegClass.hasSubClassEq(RC) ||
+ X86::VR256XRegClass.hasSubClassEq(RC)) && "Unknown 32-byte regclass");
// If stack is realigned we can use aligned stores.
if (isStackAligned)
return load ? X86::VMOVAPSYrm : X86::VMOVAPSYmr;
bool HasAVX = TM.getSubtarget<X86Subtarget>().hasAVX();
MachineInstrBuilder MIB(*MI->getParent()->getParent(), MI);
switch (MI->getOpcode()) {
+ case X86::MOV32r0:
+ return Expand2AddrUndef(MIB, get(X86::XOR32rr));
case X86::SETB_C8r:
return Expand2AddrUndef(MIB, get(X86::SBB8rr));
case X86::SETB_C16r:
MI->addRegisterKilled(Reg, TRI, true);
}
-static MachineInstr* foldPatchpoint(MachineFunction &MF,
- MachineInstr *MI,
- const SmallVectorImpl<unsigned> &Ops,
- int FrameIndex,
- const TargetInstrInfo &TII) {
- bool hasDef = MI->getOperand(0).isReg() && MI->getOperand(0).isDef() &&
- !MI->getOperand(0).isImplicit();
- unsigned StartIdx = hasDef ? 1 : 0;
-
- MachineInstr *NewMI =
- MF.CreateMachineInstr(TII.get(MI->getOpcode()), MI->getDebugLoc(), true);
- MachineInstrBuilder MIB(MF, NewMI);
-
- bool isPatchPoint = MI->getOpcode() == TargetOpcode::PATCHPOINT;
- StartIdx = isPatchPoint ?
- StartIdx + MI->getOperand(StartIdx+3).getImm() + 5 :
- StartIdx + 2;
-
- // No need to fold return, the meta data, and function arguments
- for (unsigned i = 0; i < StartIdx; ++i)
- MIB.addOperand(MI->getOperand(i));
-
- for (unsigned i = StartIdx; i < MI->getNumOperands(); ++i) {
- MachineOperand &MO = MI->getOperand(i);
- if (std::find(Ops.begin(), Ops.end(), i) != Ops.end()) {
- MIB.addOperand(MachineOperand::CreateImm(StackMaps::IndirectMemRefOp));
- MIB.addOperand(MachineOperand::CreateFI(FrameIndex));
- addOffset(MIB, 0);
- }
- else
- MIB.addOperand(MO);
- }
- return NewMI;
-}
-
MachineInstr*
X86InstrInfo::foldMemoryOperandImpl(MachineFunction &MF, MachineInstr *MI,
const SmallVectorImpl<unsigned> &Ops,
int FrameIndex) const {
- // Special case stack map and patch point intrinsics.
- if (MI->getOpcode() == TargetOpcode::STACKMAP
- || MI->getOpcode() == TargetOpcode::PATCHPOINT) {
- return foldPatchpoint(MF, MI, Ops, FrameIndex, *this);
- }
// Check switch flag
if (NoFusing) return NULL;
MachineInstr *MI,
const SmallVectorImpl<unsigned> &Ops,
MachineInstr *LoadMI) const {
+ // If loading from a FrameIndex, fold directly from the FrameIndex.
+ unsigned NumOps = LoadMI->getDesc().getNumOperands();
+ int FrameIndex;
+ if (isLoadFromStackSlot(LoadMI, FrameIndex))
+ return foldMemoryOperandImpl(MF, MI, Ops, FrameIndex);
+
// Check switch flag
if (NoFusing) return NULL;
return NULL;
// Folding a normal load. Just copy the load's address operands.
- unsigned NumOps = LoadMI->getDesc().getNumOperands();
for (unsigned i = NumOps - X86::AddrNumOperands; i != NumOps; ++i)
MOs.push_back(LoadMI->getOperand(i));
break;