maxjhandsome
/
llvm-project
forked from OSchip/llvm-project
1
0
Fork 0
Code Issues Pull Requests Packages Releases Wiki Activity

AMDGPU: Use Register

This commit is contained in:
Matt Arsenault 2020-04-21 15:06:53 -04:00
parent 704293b168
commit 7dece2fde3
2 changed files with 26 additions and 26 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
llvm/lib/Target/AMDGPU/AMDGPUInstructionSelector.cpp
View File

@ -2933,7 +2933,7 @@ AMDGPUInstructionSelector::selectSmrdImm32(MachineOperand &Root) const {
return None;
const GEPInfo &GEPInfo = AddrInfo[0];
unsigned PtrReg = GEPInfo.SgprParts[0];
Register PtrReg = GEPInfo.SgprParts[0];
Optional<int64_t> EncodedImm =
AMDGPU::getSMRDEncodedLiteralOffset32(STI, GEPInfo.Imm);
if (!EncodedImm)
@ -2967,7 +2967,7 @@ AMDGPUInstructionSelector::selectSmrdSgpr(MachineOperand &Root) const {
// It is OK to select this using a sgpr offset, because we have already
// failed trying to select this load into one of the _IMM variants since
// the _IMM Patterns are considered before the _SGPR patterns.
unsigned PtrReg = GEPInfo.SgprParts[0];
Register PtrReg = GEPInfo.SgprParts[0];
Register OffsetReg = MRI->createVirtualRegister(&AMDGPU::SReg_32RegClass);
BuildMI(*MBB, MI, MI->getDebugLoc(), TII.get(AMDGPU::S_MOV_B32), OffsetReg)
.addImm(GEPInfo.Imm);

48
llvm/lib/Target/AMDGPU/SIFrameLowering.cpp
View File

@ -41,10 +41,10 @@ static ArrayRef<MCPhysReg> getAllSGPR128(const GCNSubtarget &ST,
// but we would then have to make sure that we were in fact saving at least one
// callee-save register in the prologue, which is additional complexity that
// doesn't seem worth the benefit.
static unsigned findScratchNonCalleeSaveRegister(MachineRegisterInfo &MRI,
LivePhysRegs &LiveRegs,
const TargetRegisterClass &RC,
bool Unused = false) {
static MCRegister findScratchNonCalleeSaveRegister(MachineRegisterInfo &MRI,
LivePhysRegs &LiveRegs,
const TargetRegisterClass &RC,
bool Unused = false) {
// Mark callee saved registers as used so we will not choose them.
const MCPhysReg *CSRegs = MRI.getCalleeSavedRegs();
for (unsigned i = 0; CSRegs[i]; ++i)
@ -53,12 +53,12 @@ static unsigned findScratchNonCalleeSaveRegister(MachineRegisterInfo &MRI,
if (Unused) {
// We are looking for a register that can be used throughout the entire
// function, so any use is unacceptable.
for (unsigned Reg : RC) {
for (MCRegister Reg : RC) {
if (!MRI.isPhysRegUsed(Reg) && LiveRegs.available(MRI, Reg))
return Reg;
}
} else {
for (unsigned Reg : RC) {
for (MCRegister Reg : RC) {
if (LiveRegs.available(MRI, Reg))
return Reg;
}
@ -70,7 +70,7 @@ static unsigned findScratchNonCalleeSaveRegister(MachineRegisterInfo &MRI,
if (!Unused)
report_fatal_error("failed to find free scratch register");
return AMDGPU::NoRegister;
return MCRegister();
}
static MCPhysReg findUnusedSGPRNonCalleeSaved(MachineRegisterInfo &MRI) {
@ -85,8 +85,8 @@ static MCPhysReg findUnusedSGPRNonCalleeSaved(MachineRegisterInfo &MRI) {
// use.
static void buildPrologSpill(LivePhysRegs &LiveRegs, MachineBasicBlock &MBB,
MachineBasicBlock::iterator I,
const SIInstrInfo *TII, unsigned SpillReg,
unsigned ScratchRsrcReg, unsigned SPReg, int FI) {
const SIInstrInfo *TII, Register SpillReg,
Register ScratchRsrcReg, Register SPReg, int FI) {
MachineFunction *MF = MBB.getParent();
MachineFrameInfo &MFI = MF->getFrameInfo();
@ -133,8 +133,8 @@ static void buildPrologSpill(LivePhysRegs &LiveRegs, MachineBasicBlock &MBB,
static void buildEpilogReload(LivePhysRegs &LiveRegs, MachineBasicBlock &MBB,
MachineBasicBlock::iterator I,
const SIInstrInfo *TII, unsigned SpillReg,
unsigned ScratchRsrcReg, unsigned SPReg, int FI) {
const SIInstrInfo *TII, Register SpillReg,
Register ScratchRsrcReg, Register SPReg, int FI) {
MachineFunction *MF = MBB.getParent();
MachineFrameInfo &MFI = MF->getFrameInfo();
int64_t Offset = MFI.getObjectOffset(FI);
@ -595,8 +595,8 @@ void SIFrameLowering::emitPrologue(MachineFunction &MF,
const SIInstrInfo *TII = ST.getInstrInfo();
const SIRegisterInfo &TRI = TII->getRegisterInfo();
unsigned StackPtrReg = FuncInfo->getStackPtrOffsetReg();
unsigned FramePtrReg = FuncInfo->getFrameOffsetReg();
Register StackPtrReg = FuncInfo->getStackPtrOffsetReg();
Register FramePtrReg = FuncInfo->getFrameOffsetReg();
LivePhysRegs LiveRegs;
MachineBasicBlock::iterator MBBI = MBB.begin();
@ -607,10 +607,10 @@ void SIFrameLowering::emitPrologue(MachineFunction &MF,
uint32_t RoundedSize = NumBytes;
// To avoid clobbering VGPRs in lanes that weren't active on function entry,
// turn on all lanes before doing the spill to memory.
unsigned ScratchExecCopy = AMDGPU::NoRegister;
Register ScratchExecCopy;
// Emit the copy if we need an FP, and are using a free SGPR to save it.
if (FuncInfo->SGPRForFPSaveRestoreCopy != AMDGPU::NoRegister) {
if (FuncInfo->SGPRForFPSaveRestoreCopy) {
BuildMI(MBB, MBBI, DL, TII->get(AMDGPU::COPY), FuncInfo->SGPRForFPSaveRestoreCopy)
.addReg(FramePtrReg)
.setMIFlag(MachineInstr::FrameSetup);
@ -621,7 +621,7 @@ void SIFrameLowering::emitPrologue(MachineFunction &MF,
if (!Reg.FI.hasValue())
continue;
if (ScratchExecCopy == AMDGPU::NoRegister) {
if (!ScratchExecCopy) {
if (LiveRegs.empty()) {
LiveRegs.init(TRI);
LiveRegs.addLiveIns(MBB);
@ -650,7 +650,7 @@ void SIFrameLowering::emitPrologue(MachineFunction &MF,
if (ScratchExecCopy != AMDGPU::NoRegister) {
// FIXME: Split block and make terminator.
unsigned ExecMov = ST.isWave32() ? AMDGPU::S_MOV_B32 : AMDGPU::S_MOV_B64;
unsigned Exec = ST.isWave32() ? AMDGPU::EXEC_LO : AMDGPU::EXEC;
MCRegister Exec = ST.isWave32() ? AMDGPU::EXEC_LO : AMDGPU::EXEC;
BuildMI(MBB, MBBI, DL, TII->get(ExecMov), Exec)
.addReg(ScratchExecCopy, RegState::Kill);
LiveRegs.addReg(ScratchExecCopy);
@ -685,7 +685,7 @@ void SIFrameLowering::emitPrologue(MachineFunction &MF,
LiveRegs.addReg(FuncInfo->SGPRForFPSaveRestoreCopy);
}
unsigned ScratchSPReg = findScratchNonCalleeSaveRegister(
Register ScratchSPReg = findScratchNonCalleeSaveRegister(
MRI, LiveRegs, AMDGPU::SReg_32_XM0RegClass);
assert(ScratchSPReg != AMDGPU::NoRegister &&
ScratchSPReg != FuncInfo->SGPRForFPSaveRestoreCopy);
@ -718,11 +718,11 @@ void SIFrameLowering::emitPrologue(MachineFunction &MF,
.setMIFlag(MachineInstr::FrameSetup);
}
assert((!HasFP || (FuncInfo->SGPRForFPSaveRestoreCopy != AMDGPU::NoRegister ||
assert((!HasFP || (FuncInfo->SGPRForFPSaveRestoreCopy ||
FuncInfo->FramePointerSaveIndex)) &&
"Needed to save FP but didn't save it anywhere");
assert((HasFP || (FuncInfo->SGPRForFPSaveRestoreCopy == AMDGPU::NoRegister &&
assert((HasFP || (!FuncInfo->SGPRForFPSaveRestoreCopy &&
!FuncInfo->FramePointerSaveIndex)) &&
"Saved FP but didn't need it");
}
@ -747,14 +747,14 @@ void SIFrameLowering::emitEpilogue(MachineFunction &MF,
: NumBytes;
if (RoundedSize != 0 && hasFP(MF)) {
const unsigned StackPtrReg = FuncInfo->getStackPtrOffsetReg();
const Register StackPtrReg = FuncInfo->getStackPtrOffsetReg();
BuildMI(MBB, MBBI, DL, TII->get(AMDGPU::S_SUB_U32), StackPtrReg)
.addReg(StackPtrReg)
.addImm(RoundedSize * ST.getWavefrontSize())
.setMIFlag(MachineInstr::FrameDestroy);
}
if (FuncInfo->SGPRForFPSaveRestoreCopy != AMDGPU::NoRegister) {
if (FuncInfo->SGPRForFPSaveRestoreCopy) {
BuildMI(MBB, MBBI, DL, TII->get(AMDGPU::COPY), FuncInfo->getFrameOffsetReg())
.addReg(FuncInfo->SGPRForFPSaveRestoreCopy)
.setMIFlag(MachineInstr::FrameSetup);
@ -775,7 +775,7 @@ void SIFrameLowering::emitEpilogue(MachineFunction &MF,
.addImm(Spill[0].Lane);
}
unsigned ScratchExecCopy = AMDGPU::NoRegister;
Register ScratchExecCopy;
for (const SIMachineFunctionInfo::SGPRSpillVGPRCSR &Reg
: FuncInfo->getSGPRSpillVGPRs()) {
if (!Reg.FI.hasValue())
@ -1016,7 +1016,7 @@ MachineBasicBlock::iterator SIFrameLowering::eliminateCallFramePseudoInstr(
Amount = alignTo(Amount, getStackAlign());
assert(isUInt<32>(Amount) && "exceeded stack address space size");
const SIMachineFunctionInfo *MFI = MF.getInfo<SIMachineFunctionInfo>();
unsigned SPReg = MFI->getStackPtrOffsetReg();
Register SPReg = MFI->getStackPtrOffsetReg();
unsigned Op = IsDestroy ? AMDGPU::S_SUB_U32 : AMDGPU::S_ADD_U32;
BuildMI(MBB, I, DL, TII->get(Op), SPReg)