[LiveRegUnits] Enhanced the register liveness check #66061
Closed
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Currently there's only a provision to check whether a register is contained in the LiveRegUnits BitVector. Introducing the method which is available in LivePhysRegs to check whether the register is not reserved as well. The naming convention has been retained like in LivePhysRegs.
Member
|
@llvm/pr-subscribers-backend-aarch64 ChangesCurrently there's only a provision to check whether a register is contained in the LiveRegUnits BitVector. Introducing the method which is available in LivePhysRegs to check whether the register is not reserved as well. The naming convention has been retained like in LivePhysRegs. The motivation for this PR is to separate the commits in [AMDGPU] [SIFrameLowering] Use LiveRegUnits instead of LivePhysRegs into separate PRs. This will be used to replace the use of LivePhysRegs with LiveRegUnits in AMDGPU back-end's SIFrameLowering in a coming PRPatch is 23.02 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/66061.diff 16 Files Affected:
diff --git a/llvm/include/llvm/CodeGen/LiveRegUnits.h b/llvm/include/llvm/CodeGen/LiveRegUnits.h
index e96165d6b3bbf3a..e38aab6d514750d 100644
--- a/llvm/include/llvm/CodeGen/LiveRegUnits.h
+++ b/llvm/include/llvm/CodeGen/LiveRegUnits.h
@@ -112,13 +112,13 @@ class LiveRegUnits {
/// The regmask has the same format as the one in the RegMask machine operand.
void addRegsInMask(const uint32_t *RegMask);
- /// Returns true if no part of physical register \p Reg is live.
- bool available(MCPhysReg Reg) const {
- for (MCRegUnit Unit : TRI->regunits(Reg)) {
- if (Units.test(Unit))
- return false;
- }
- return true;
+ /// Returns true if no part of physical register \p Reg is live or reserved.
+ bool available(const MachineRegisterInfo &MRI, MCPhysReg Reg) const;
+
+ /// Returns true if any part of physical register \p Reg is live
+ bool contains(MCPhysReg Reg) const {
+ return llvm::any_of(TRI->regunits(Reg),
+ [&](MCRegUnit Unit) { return Units.test(Unit); });
}
/// Updates liveness when stepping backwards over the instruction \p MI.
diff --git a/llvm/include/llvm/CodeGen/MachineOutliner.h b/llvm/include/llvm/CodeGen/MachineOutliner.h
index d0ff02fea4ff9b7..f6ecb890accd1f9 100644
--- a/llvm/include/llvm/CodeGen/MachineOutliner.h
+++ b/llvm/include/llvm/CodeGen/MachineOutliner.h
@@ -156,7 +156,7 @@ struct Candidate {
const TargetRegisterInfo &TRI) {
if (!FromEndOfBlockToStartOfSeqWasSet)
initFromEndOfBlockToStartOfSeq(TRI);
- return FromEndOfBlockToStartOfSeq.available(Reg);
+ return !FromEndOfBlockToStartOfSeq.contains(Reg);
}
/// \returns True if `isAvailableAcrossAndOutOfSeq` fails for any register
@@ -166,7 +166,7 @@ struct Candidate {
if (!FromEndOfBlockToStartOfSeqWasSet)
initFromEndOfBlockToStartOfSeq(TRI);
return any_of(Regs, [&](Register Reg) {
- return !FromEndOfBlockToStartOfSeq.available(Reg);
+ return FromEndOfBlockToStartOfSeq.contains(Reg);
});
}
@@ -181,7 +181,7 @@ struct Candidate {
bool isAvailableInsideSeq(Register Reg, const TargetRegisterInfo &TRI) {
if (!InSeqWasSet)
initInSeq(TRI);
- return InSeq.available(Reg);
+ return !InSeq.contains(Reg);
}
/// The number of instructions that would be saved by outlining every
diff --git a/llvm/lib/CodeGen/DeadMachineInstructionElim.cpp b/llvm/lib/CodeGen/DeadMachineInstructionElim.cpp
index 6a7de3b241feeea..e3c5c57ec068c94 100644
--- a/llvm/lib/CodeGen/DeadMachineInstructionElim.cpp
+++ b/llvm/lib/CodeGen/DeadMachineInstructionElim.cpp
@@ -79,7 +79,7 @@ bool DeadMachineInstructionElim::isDead(const MachineInstr *MI) const {
Register Reg = MO.getReg();
if (Reg.isPhysical()) {
// Don't delete live physreg defs, or any reserved register defs.
- if (!LivePhysRegs.available(Reg) || MRI->isReserved(Reg))
+ if (!LivePhysRegs.available(*MRI, Reg))
return false;
} else {
if (MO.isDead()) {
diff --git a/llvm/lib/CodeGen/LiveRegUnits.cpp b/llvm/lib/CodeGen/LiveRegUnits.cpp
index 34de09dd2944bc7..d711cd44f260da4 100644
--- a/llvm/lib/CodeGen/LiveRegUnits.cpp
+++ b/llvm/lib/CodeGen/LiveRegUnits.cpp
@@ -41,6 +41,11 @@ void LiveRegUnits::addRegsInMask(const uint32_t *RegMask) {
}
}
+bool LiveRegUnits::available(const MachineRegisterInfo &MRI,
+ MCPhysReg Reg) const {
+ return !MRI.isReserved(Reg) && !contains(Reg);
+}
+
void LiveRegUnits::stepBackward(const MachineInstr &MI) {
// Remove defined registers and regmask kills from the set.
for (const MachineOperand &MOP : MI.operands()) {
diff --git a/llvm/lib/CodeGen/MachineSink.cpp b/llvm/lib/CodeGen/MachineSink.cpp
index b4cbb93d758ef2f..c73d17dc2f1a669 100644
--- a/llvm/lib/CodeGen/MachineSink.cpp
+++ b/llvm/lib/CodeGen/MachineSink.cpp
@@ -1627,7 +1627,7 @@ static bool aliasWithRegsInLiveIn(MachineBasicBlock &MBB, unsigned Reg,
const TargetRegisterInfo *TRI) {
LiveRegUnits LiveInRegUnits(*TRI);
LiveInRegUnits.addLiveIns(MBB);
- return !LiveInRegUnits.available(Reg);
+ return LiveInRegUnits.contains(Reg);
}
static MachineBasicBlock *
@@ -1680,7 +1680,7 @@ static void clearKillFlags(MachineInstr *MI, MachineBasicBlock &CurBB,
for (auto U : UsedOpsInCopy) {
MachineOperand &MO = MI->getOperand(U);
Register SrcReg = MO.getReg();
- if (!UsedRegUnits.available(SrcReg)) {
+ if (UsedRegUnits.contains(SrcReg)) {
MachineBasicBlock::iterator NI = std::next(MI->getIterator());
for (MachineInstr &UI : make_range(NI, CurBB.end())) {
if (UI.killsRegister(SrcReg, TRI)) {
@@ -1725,7 +1725,7 @@ static bool hasRegisterDependency(MachineInstr *MI,
if (!Reg)
continue;
if (MO.isDef()) {
- if (!ModifiedRegUnits.available(Reg) || !UsedRegUnits.available(Reg)) {
+ if (ModifiedRegUnits.contains(Reg) || UsedRegUnits.contains(Reg)) {
HasRegDependency = true;
break;
}
@@ -1736,7 +1736,7 @@ static bool hasRegisterDependency(MachineInstr *MI,
// it's not perfectly clear if skipping the internal read is safe in all
// other targets.
} else if (MO.isUse()) {
- if (!ModifiedRegUnits.available(Reg)) {
+ if (ModifiedRegUnits.contains(Reg)) {
HasRegDependency = true;
break;
}
diff --git a/llvm/lib/CodeGen/RegisterScavenging.cpp b/llvm/lib/CodeGen/RegisterScavenging.cpp
index e6ff5701bc4bd20..938fc9ba215c74e 100644
--- a/llvm/lib/CodeGen/RegisterScavenging.cpp
+++ b/llvm/lib/CodeGen/RegisterScavenging.cpp
@@ -109,7 +109,7 @@ void RegScavenger::backward() {
bool RegScavenger::isRegUsed(Register Reg, bool includeReserved) const {
if (isReserved(Reg))
return includeReserved;
- return !LiveUnits.available(Reg);
+ return LiveUnits.contains(Reg);
}
Register RegScavenger::FindUnusedReg(const TargetRegisterClass *RC) const {
@@ -164,8 +164,8 @@ findSurvivorBackwards(const MachineRegisterInfo &MRI,
if (I == To) {
// See if one of the registers in RC wasn't used so far.
for (MCPhysReg Reg : AllocationOrder) {
- if (!MRI.isReserved(Reg) && Used.available(Reg) &&
- LiveOut.available(Reg))
+ if (!MRI.isReserved(Reg) && !Used.contains(Reg) &&
+ !LiveOut.contains(Reg))
return std::make_pair(Reg, MBB.end());
}
// Otherwise we will continue up to InstrLimit instructions to find
@@ -186,10 +186,10 @@ findSurvivorBackwards(const MachineRegisterInfo &MRI,
MI.getFlag(MachineInstr::FrameSetup))
break;
- if (Survivor == 0 || !Used.available(Survivor)) {
+ if (Survivor == 0 || Used.contains(Survivor)) {
MCPhysReg AvilableReg = 0;
for (MCPhysReg Reg : AllocationOrder) {
- if (!MRI.isReserved(Reg) && Used.available(Reg)) {
+ if (Used.available(MRI, Reg)) {
AvilableReg = Reg;
break;
}
diff --git a/llvm/lib/Target/AArch64/AArch64A57FPLoadBalancing.cpp b/llvm/lib/Target/AArch64/AArch64A57FPLoadBalancing.cpp
index ea8e30269ece25f..17e5a60438f903a 100644
--- a/llvm/lib/Target/AArch64/AArch64A57FPLoadBalancing.cpp
+++ b/llvm/lib/Target/AArch64/AArch64A57FPLoadBalancing.cpp
@@ -518,7 +518,7 @@ int AArch64A57FPLoadBalancing::scavengeRegister(Chain *G, Color C,
unsigned RegClassID = ChainBegin->getDesc().operands()[0].RegClass;
auto Ord = RCI.getOrder(TRI->getRegClass(RegClassID));
for (auto Reg : Ord) {
- if (!Units.available(Reg))
+ if (Units.contains(Reg))
continue;
if (C == getColor(Reg))
return Reg;
diff --git a/llvm/lib/Target/AArch64/AArch64FalkorHWPFFix.cpp b/llvm/lib/Target/AArch64/AArch64FalkorHWPFFix.cpp
index 32686b25f21184b..7d481efb0136e7b 100644
--- a/llvm/lib/Target/AArch64/AArch64FalkorHWPFFix.cpp
+++ b/llvm/lib/Target/AArch64/AArch64FalkorHWPFFix.cpp
@@ -748,7 +748,7 @@ void FalkorHWPFFix::runOnLoop(MachineLoop &L, MachineFunction &Fn) {
}
for (unsigned ScratchReg : AArch64::GPR64RegClass) {
- if (!LR.available(ScratchReg) || MRI.isReserved(ScratchReg))
+ if (!LR.available(MRI, ScratchReg))
continue;
LoadInfo NewLdI(LdI);
diff --git a/llvm/lib/Target/AArch64/AArch64InstrInfo.cpp b/llvm/lib/Target/AArch64/AArch64InstrInfo.cpp
index a41ac0e44a7700b..61294bb2043bc94 100644
--- a/llvm/lib/Target/AArch64/AArch64InstrInfo.cpp
+++ b/llvm/lib/Target/AArch64/AArch64InstrInfo.cpp
@@ -7838,8 +7838,8 @@ AArch64InstrInfo::getOutlinableRanges(MachineBasicBlock &MBB,
// where these registers are dead. We will only outline from those ranges.
LiveRegUnits LRU(getRegisterInfo());
auto AreAllUnsafeRegsDead = [&LRU]() {
- return LRU.available(AArch64::W16) && LRU.available(AArch64::W17) &&
- LRU.available(AArch64::NZCV);
+ return !LRU.contains(AArch64::W16) && !LRU.contains(AArch64::W17) &&
+ !LRU.contains(AArch64::NZCV);
};
// We need to know if LR is live across an outlining boundary later on in
@@ -7909,7 +7909,7 @@ AArch64InstrInfo::getOutlinableRanges(MachineBasicBlock &MBB,
CreateNewRangeStartingAt(MI.getIterator());
continue;
}
- LRAvailableEverywhere &= LRU.available(AArch64::LR);
+ LRAvailableEverywhere &= !LRU.contains(AArch64::LR);
RangeBegin = MI.getIterator();
++RangeLen;
}
diff --git a/llvm/lib/Target/AArch64/AArch64LoadStoreOptimizer.cpp b/llvm/lib/Target/AArch64/AArch64LoadStoreOptimizer.cpp
index 41af5522d967dbf..14d1e873dbac2dd 100644
--- a/llvm/lib/Target/AArch64/AArch64LoadStoreOptimizer.cpp
+++ b/llvm/lib/Target/AArch64/AArch64LoadStoreOptimizer.cpp
@@ -1265,7 +1265,7 @@ bool AArch64LoadStoreOpt::findMatchingStore(
BaseReg == AArch64InstrInfo::getLdStBaseOp(MI).getReg() &&
AArch64InstrInfo::getLdStOffsetOp(MI).isImm() &&
isLdOffsetInRangeOfSt(LoadMI, MI, TII) &&
- ModifiedRegUnits.available(getLdStRegOp(MI).getReg())) {
+ !ModifiedRegUnits.contains(getLdStRegOp(MI).getReg())) {
StoreI = MBBI;
return true;
}
@@ -1278,7 +1278,7 @@ bool AArch64LoadStoreOpt::findMatchingStore(
// Otherwise, if the base register is modified, we have no match, so
// return early.
- if (!ModifiedRegUnits.available(BaseReg))
+ if (ModifiedRegUnits.contains(BaseReg))
return false;
// If we encounter a store aliased with the load, return early.
@@ -1510,7 +1510,7 @@ static std::optional tryToFindRegisterToRename(
auto *RegClass = TRI->getMinimalPhysRegClass(Reg);
for (const MCPhysReg &PR : *RegClass) {
- if (DefinedInBB.available(PR) && UsedInBetween.available(PR) &&
+ if (!DefinedInBB.contains(PR) && !UsedInBetween.contains(PR) &&
!RegInfo.isReserved(PR) && !AnySubOrSuperRegCalleePreserved(PR) &&
CanBeUsedForAllClasses(PR)) {
DefinedInBB.addReg(PR);
@@ -1615,9 +1615,9 @@ AArch64LoadStoreOpt::findMatchingInsn(MachineBasicBlock::iterator I,
// can't be paired: bail and keep looking.
if (IsPreLdSt) {
bool IsOutOfBounds = MIOffset != TII->getMemScale(MI);
- bool IsBaseRegUsed = !UsedRegUnits.available(
+ bool IsBaseRegUsed = UsedRegUnits.contains(
AArch64InstrInfo::getLdStBaseOp(MI).getReg());
- bool IsBaseRegModified = !ModifiedRegUnits.available(
+ bool IsBaseRegModified = ModifiedRegUnits.contains(
AArch64InstrInfo::getLdStBaseOp(MI).getReg());
// If the stored value and the address of the second instruction is
// the same, it needs to be using the updated register and therefore
@@ -1694,16 +1694,16 @@ AArch64LoadStoreOpt::findMatchingInsn(MachineBasicBlock::iterator I,
// ldr x2 [x3]
// ldr x4 [x2, #8],
// the first and third ldr cannot be converted to ldp x1, x4, [x2]
- if (!ModifiedRegUnits.available(BaseReg))
+ if (ModifiedRegUnits.contains(BaseReg))
return E;
// If the Rt of the second instruction was not modified or used between
// the two instructions and none of the instructions between the second
// and first alias with the second, we can combine the second into the
// first.
- if (ModifiedRegUnits.available(getLdStRegOp(MI).getReg()) &&
+ if (!ModifiedRegUnits.contains(getLdStRegOp(MI).getReg()) &&
!(MI.mayLoad() &&
- !UsedRegUnits.available(getLdStRegOp(MI).getReg())) &&
+ UsedRegUnits.contains(getLdStRegOp(MI).getReg())) &&
!mayAlias(MI, MemInsns, AA)) {
Flags.setMergeForward(false);
@@ -1716,10 +1716,10 @@ AArch64LoadStoreOpt::findMatchingInsn(MachineBasicBlock::iterator I,
// first and the second alias with the first, we can combine the first
// into the second.
if (!(MayLoad &&
- !UsedRegUnits.available(getLdStRegOp(FirstMI).getReg())) &&
+ UsedRegUnits.contains(getLdStRegOp(FirstMI).getReg())) &&
!mayAlias(FirstMI, MemInsns, AA)) {
- if (ModifiedRegUnits.available(getLdStRegOp(FirstMI).getReg())) {
+ if (!ModifiedRegUnits.contains(getLdStRegOp(FirstMI).getReg())) {
Flags.setMergeForward(true);
Flags.clearRenameReg();
return MBBI;
@@ -1761,7 +1761,7 @@ AArch64LoadStoreOpt::findMatchingInsn(MachineBasicBlock::iterator I,
// Otherwise, if the base register is modified, we have no match, so
// return early.
- if (!ModifiedRegUnits.available(BaseReg))
+ if (ModifiedRegUnits.contains(BaseReg))
return E;
// Update list of instructions that read/write memory.
@@ -1987,8 +1987,8 @@ MachineBasicBlock::iterator AArch64LoadStoreOpt::findMatchingUpdateInsnForward(
// return early.
// If we are optimizing SP, do not allow instructions that may load or store
// in between the load and the optimized value update.
- if (!ModifiedRegUnits.available(BaseReg) ||
- !UsedRegUnits.available(BaseReg) ||
+ if (ModifiedRegUnits.contains(BaseReg) ||
+ UsedRegUnits.contains(BaseReg) ||
(BaseRegSP && MBBI->mayLoadOrStore()))
return E;
}
@@ -2062,8 +2062,8 @@ MachineBasicBlock::iterator AArch64LoadStoreOpt::findMatchingUpdateInsnBackward(
// Otherwise, if the base register is used or modified, we have no match, so
// return early.
- if (!ModifiedRegUnits.available(BaseReg) ||
- !UsedRegUnits.available(BaseReg))
+ if (ModifiedRegUnits.contains(BaseReg) ||
+ UsedRegUnits.contains(BaseReg))
return E;
// Keep track if we have a memory access before an SP pre-increment, in this
// case we need to validate later that the update amount respects the red
diff --git a/llvm/lib/Target/AArch64/AArch64RedundantCopyElimination.cpp b/llvm/lib/Target/AArch64/AArch64RedundantCopyElimination.cpp
index 1494312886a40de..a4fc06cce5bd450 100644
--- a/llvm/lib/Target/AArch64/AArch64RedundantCopyElimination.cpp
+++ b/llvm/lib/Target/AArch64/AArch64RedundantCopyElimination.cpp
@@ -190,7 +190,7 @@ bool AArch64RedundantCopyElimination::knownRegValInBlock(
// register of the compare is not modified (including a self-clobbering
// compare) between the compare and conditional branch we known the value
// of the 1st source operand.
- if (PredI.getOperand(2).isImm() && DomBBClobberedRegs.available(SrcReg) &&
+ if (PredI.getOperand(2).isImm() && !DomBBClobberedRegs.contains(SrcReg) &&
SrcReg != DstReg) {
// We've found the instruction that sets NZCV.
int32_t KnownImm = PredI.getOperand(2).getImm();
@@ -210,7 +210,7 @@ bool AArch64RedundantCopyElimination::knownRegValInBlock(
// The destination register must not be modified between the NZCV setting
// instruction and the conditional branch.
- if (!DomBBClobberedRegs.available(DstReg))
+ if (DomBBClobberedRegs.contains(DstReg))
return Res;
FirstUse = PredI;
@@ -254,7 +254,7 @@ bool AArch64RedundantCopyElimination::knownRegValInBlock(
// The destination register of the NZCV setting instruction must not be
// modified before the conditional branch.
- if (!DomBBClobberedRegs.available(DstReg))
+ if (DomBBClobberedRegs.contains(DstReg))
return false;
// We've found the instruction that sets NZCV whose DstReg == 0.
@@ -323,12 +323,12 @@ bool AArch64RedundantCopyElimination::optimizeBlock(MachineBasicBlock *MBB) {
MCPhysReg CopyDstReg = PredI->getOperand(0).getReg();
MCPhysReg CopySrcReg = PredI->getOperand(1).getReg();
for (auto &KnownReg : KnownRegs) {
- if (!OptBBClobberedRegs.available(KnownReg.Reg))
+ if (OptBBClobberedRegs.contains(KnownReg.Reg))
continue;
// If we have X = COPY Y, and Y is known to be zero, then now X is
// known to be zero.
if (CopySrcReg == KnownReg.Reg &&
- OptBBClobberedRegs.available(CopyDstReg)) {
+ !OptBBClobberedRegs.contains(CopyDstReg)) {
KnownRegs.push_back(RegImm(CopyDstReg, KnownReg.Imm));
if (SeenFirstUse)
FirstUse = PredI;
@@ -337,7 +337,7 @@ bool AArch64RedundantCopyElimination::optimizeBlock(MachineBasicBlock *MBB) {
// If we have X = COPY Y, and X is known to be zero, then now Y is
// known to be zero.
if (CopyDstReg == KnownReg.Reg &&
- OptBBClobberedRegs.available(CopySrcReg)) {
+ !OptBBClobberedRegs.contains(CopySrcReg)) {
KnownRegs.push_back(RegImm(CopySrcReg, KnownReg.Imm));
if (SeenFirstUse)
FirstUse = PredI;
@@ -354,7 +354,7 @@ bool AArch64RedundantCopyElimination::optimizeBlock(MachineBasicBlock *MBB) {
OptBBUsedRegs, TRI);
// Stop if all of the known-zero regs have been clobbered.
if (all_of(KnownRegs, [&](RegImm KnownReg) {
- return !OptBBClobberedRegs.available(KnownReg.Reg);
+ return OptBBClobberedRegs.contains(KnownReg.Reg);
}))
break;
}
diff --git a/llvm/lib/Target/AArch64/GISel/AArch64PostSelectOptimize.cpp b/llvm/lib/Target/AArch64/GISel/AArch64PostSelectOptimize.cpp
index 94584e20f5ab3fb..ff3404b6584acc5 100644
--- a/llvm/lib/Target/AArch64/GISel/AArch64PostSelectOptimize.cpp
+++ b/llvm/lib/Target/AArch64/GISel/AArch64PostSelectOptimize.cpp
@@ -198,7 +198,7 @@ bool AArch64PostSelectOptimize::optimizeNZCVDefs(MachineBasicBlock &MBB) {
LRU.addLiveOuts(MBB);
for (auto &II : instructionsWithoutDebug(MBB.rbegin(), MBB.rend())) {
- bool NZCVDead = LRU.available(AArch64::NZCV);
+ bool NZCVDead = !LRU.contains(AArch64::NZCV);
if (NZCVDead && II.definesRegister(AArch64::NZCV)) {
// The instruction defines NZCV, but NZCV is dead.
unsigned NewOpc = getNonFlagSettingVariant(II.getOpcode());
diff --git a/llvm/lib/Target/ARM/ARMBaseInstrInfo.cpp b/llvm/lib/Target/ARM/ARMBaseInstrInfo.cpp
index 1ffdde0360cf623..f087ce71f8ba73c 100644
--- a/llvm/lib/Target/ARM/ARMBaseInstrInfo.cpp
+++ b/llvm/lib/Target/ARM/ARMBaseInstrInfo.cpp
@@ -6259,8 +6259,8 @@ bool ARMBaseInstrInfo::isMBBSafeToOutlineFrom(MachineBasicBlock ...
|
Member
|
@llvm/pr-subscribers-backend-risc-v ChangesCurrently there's only a provision to check whether a register is contained in the LiveRegUnits BitVector. Introducing the method which is available in LivePhysRegs to check whether the register is not reserved as well. The naming convention has been retained like in LivePhysRegs. The motivation for this PR is to separate the commits in [AMDGPU] [SIFrameLowering] Use LiveRegUnits instead of LivePhysRegs into separate PRs. This will be used to replace the use of LivePhysRegs with LiveRegUnits in AMDGPU back-end's SIFrameLowering in a coming PRPatch is 23.02 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/66061.diff 16 Files Affected:
diff --git a/llvm/include/llvm/CodeGen/LiveRegUnits.h b/llvm/include/llvm/CodeGen/LiveRegUnits.h
index e96165d6b3bbf3a..e38aab6d514750d 100644
--- a/llvm/include/llvm/CodeGen/LiveRegUnits.h
+++ b/llvm/include/llvm/CodeGen/LiveRegUnits.h
@@ -112,13 +112,13 @@ class LiveRegUnits {
/// The regmask has the same format as the one in the RegMask machine operand.
void addRegsInMask(const uint32_t *RegMask);
- /// Returns true if no part of physical register \p Reg is live.
- bool available(MCPhysReg Reg) const {
- for (MCRegUnit Unit : TRI->regunits(Reg)) {
- if (Units.test(Unit))
- return false;
- }
- return true;
+ /// Returns true if no part of physical register \p Reg is live or reserved.
+ bool available(const MachineRegisterInfo &MRI, MCPhysReg Reg) const;
+
+ /// Returns true if any part of physical register \p Reg is live
+ bool contains(MCPhysReg Reg) const {
+ return llvm::any_of(TRI->regunits(Reg),
+ [&](MCRegUnit Unit) { return Units.test(Unit); });
}
/// Updates liveness when stepping backwards over the instruction \p MI.
diff --git a/llvm/include/llvm/CodeGen/MachineOutliner.h b/llvm/include/llvm/CodeGen/MachineOutliner.h
index d0ff02fea4ff9b7..f6ecb890accd1f9 100644
--- a/llvm/include/llvm/CodeGen/MachineOutliner.h
+++ b/llvm/include/llvm/CodeGen/MachineOutliner.h
@@ -156,7 +156,7 @@ struct Candidate {
const TargetRegisterInfo &TRI) {
if (!FromEndOfBlockToStartOfSeqWasSet)
initFromEndOfBlockToStartOfSeq(TRI);
- return FromEndOfBlockToStartOfSeq.available(Reg);
+ return !FromEndOfBlockToStartOfSeq.contains(Reg);
}
/// \returns True if `isAvailableAcrossAndOutOfSeq` fails for any register
@@ -166,7 +166,7 @@ struct Candidate {
if (!FromEndOfBlockToStartOfSeqWasSet)
initFromEndOfBlockToStartOfSeq(TRI);
return any_of(Regs, [&](Register Reg) {
- return !FromEndOfBlockToStartOfSeq.available(Reg);
+ return FromEndOfBlockToStartOfSeq.contains(Reg);
});
}
@@ -181,7 +181,7 @@ struct Candidate {
bool isAvailableInsideSeq(Register Reg, const TargetRegisterInfo &TRI) {
if (!InSeqWasSet)
initInSeq(TRI);
- return InSeq.available(Reg);
+ return !InSeq.contains(Reg);
}
/// The number of instructions that would be saved by outlining every
diff --git a/llvm/lib/CodeGen/DeadMachineInstructionElim.cpp b/llvm/lib/CodeGen/DeadMachineInstructionElim.cpp
index 6a7de3b241feeea..e3c5c57ec068c94 100644
--- a/llvm/lib/CodeGen/DeadMachineInstructionElim.cpp
+++ b/llvm/lib/CodeGen/DeadMachineInstructionElim.cpp
@@ -79,7 +79,7 @@ bool DeadMachineInstructionElim::isDead(const MachineInstr *MI) const {
Register Reg = MO.getReg();
if (Reg.isPhysical()) {
// Don't delete live physreg defs, or any reserved register defs.
- if (!LivePhysRegs.available(Reg) || MRI->isReserved(Reg))
+ if (!LivePhysRegs.available(*MRI, Reg))
return false;
} else {
if (MO.isDead()) {
diff --git a/llvm/lib/CodeGen/LiveRegUnits.cpp b/llvm/lib/CodeGen/LiveRegUnits.cpp
index 34de09dd2944bc7..d711cd44f260da4 100644
--- a/llvm/lib/CodeGen/LiveRegUnits.cpp
+++ b/llvm/lib/CodeGen/LiveRegUnits.cpp
@@ -41,6 +41,11 @@ void LiveRegUnits::addRegsInMask(const uint32_t *RegMask) {
}
}
+bool LiveRegUnits::available(const MachineRegisterInfo &MRI,
+ MCPhysReg Reg) const {
+ return !MRI.isReserved(Reg) && !contains(Reg);
+}
+
void LiveRegUnits::stepBackward(const MachineInstr &MI) {
// Remove defined registers and regmask kills from the set.
for (const MachineOperand &MOP : MI.operands()) {
diff --git a/llvm/lib/CodeGen/MachineSink.cpp b/llvm/lib/CodeGen/MachineSink.cpp
index b4cbb93d758ef2f..c73d17dc2f1a669 100644
--- a/llvm/lib/CodeGen/MachineSink.cpp
+++ b/llvm/lib/CodeGen/MachineSink.cpp
@@ -1627,7 +1627,7 @@ static bool aliasWithRegsInLiveIn(MachineBasicBlock &MBB, unsigned Reg,
const TargetRegisterInfo *TRI) {
LiveRegUnits LiveInRegUnits(*TRI);
LiveInRegUnits.addLiveIns(MBB);
- return !LiveInRegUnits.available(Reg);
+ return LiveInRegUnits.contains(Reg);
}
static MachineBasicBlock *
@@ -1680,7 +1680,7 @@ static void clearKillFlags(MachineInstr *MI, MachineBasicBlock &CurBB,
for (auto U : UsedOpsInCopy) {
MachineOperand &MO = MI->getOperand(U);
Register SrcReg = MO.getReg();
- if (!UsedRegUnits.available(SrcReg)) {
+ if (UsedRegUnits.contains(SrcReg)) {
MachineBasicBlock::iterator NI = std::next(MI->getIterator());
for (MachineInstr &UI : make_range(NI, CurBB.end())) {
if (UI.killsRegister(SrcReg, TRI)) {
@@ -1725,7 +1725,7 @@ static bool hasRegisterDependency(MachineInstr *MI,
if (!Reg)
continue;
if (MO.isDef()) {
- if (!ModifiedRegUnits.available(Reg) || !UsedRegUnits.available(Reg)) {
+ if (ModifiedRegUnits.contains(Reg) || UsedRegUnits.contains(Reg)) {
HasRegDependency = true;
break;
}
@@ -1736,7 +1736,7 @@ static bool hasRegisterDependency(MachineInstr *MI,
// it's not perfectly clear if skipping the internal read is safe in all
// other targets.
} else if (MO.isUse()) {
- if (!ModifiedRegUnits.available(Reg)) {
+ if (ModifiedRegUnits.contains(Reg)) {
HasRegDependency = true;
break;
}
diff --git a/llvm/lib/CodeGen/RegisterScavenging.cpp b/llvm/lib/CodeGen/RegisterScavenging.cpp
index e6ff5701bc4bd20..938fc9ba215c74e 100644
--- a/llvm/lib/CodeGen/RegisterScavenging.cpp
+++ b/llvm/lib/CodeGen/RegisterScavenging.cpp
@@ -109,7 +109,7 @@ void RegScavenger::backward() {
bool RegScavenger::isRegUsed(Register Reg, bool includeReserved) const {
if (isReserved(Reg))
return includeReserved;
- return !LiveUnits.available(Reg);
+ return LiveUnits.contains(Reg);
}
Register RegScavenger::FindUnusedReg(const TargetRegisterClass *RC) const {
@@ -164,8 +164,8 @@ findSurvivorBackwards(const MachineRegisterInfo &MRI,
if (I == To) {
// See if one of the registers in RC wasn't used so far.
for (MCPhysReg Reg : AllocationOrder) {
- if (!MRI.isReserved(Reg) && Used.available(Reg) &&
- LiveOut.available(Reg))
+ if (!MRI.isReserved(Reg) && !Used.contains(Reg) &&
+ !LiveOut.contains(Reg))
return std::make_pair(Reg, MBB.end());
}
// Otherwise we will continue up to InstrLimit instructions to find
@@ -186,10 +186,10 @@ findSurvivorBackwards(const MachineRegisterInfo &MRI,
MI.getFlag(MachineInstr::FrameSetup))
break;
- if (Survivor == 0 || !Used.available(Survivor)) {
+ if (Survivor == 0 || Used.contains(Survivor)) {
MCPhysReg AvilableReg = 0;
for (MCPhysReg Reg : AllocationOrder) {
- if (!MRI.isReserved(Reg) && Used.available(Reg)) {
+ if (Used.available(MRI, Reg)) {
AvilableReg = Reg;
break;
}
diff --git a/llvm/lib/Target/AArch64/AArch64A57FPLoadBalancing.cpp b/llvm/lib/Target/AArch64/AArch64A57FPLoadBalancing.cpp
index ea8e30269ece25f..17e5a60438f903a 100644
--- a/llvm/lib/Target/AArch64/AArch64A57FPLoadBalancing.cpp
+++ b/llvm/lib/Target/AArch64/AArch64A57FPLoadBalancing.cpp
@@ -518,7 +518,7 @@ int AArch64A57FPLoadBalancing::scavengeRegister(Chain *G, Color C,
unsigned RegClassID = ChainBegin->getDesc().operands()[0].RegClass;
auto Ord = RCI.getOrder(TRI->getRegClass(RegClassID));
for (auto Reg : Ord) {
- if (!Units.available(Reg))
+ if (Units.contains(Reg))
continue;
if (C == getColor(Reg))
return Reg;
diff --git a/llvm/lib/Target/AArch64/AArch64FalkorHWPFFix.cpp b/llvm/lib/Target/AArch64/AArch64FalkorHWPFFix.cpp
index 32686b25f21184b..7d481efb0136e7b 100644
--- a/llvm/lib/Target/AArch64/AArch64FalkorHWPFFix.cpp
+++ b/llvm/lib/Target/AArch64/AArch64FalkorHWPFFix.cpp
@@ -748,7 +748,7 @@ void FalkorHWPFFix::runOnLoop(MachineLoop &L, MachineFunction &Fn) {
}
for (unsigned ScratchReg : AArch64::GPR64RegClass) {
- if (!LR.available(ScratchReg) || MRI.isReserved(ScratchReg))
+ if (!LR.available(MRI, ScratchReg))
continue;
LoadInfo NewLdI(LdI);
diff --git a/llvm/lib/Target/AArch64/AArch64InstrInfo.cpp b/llvm/lib/Target/AArch64/AArch64InstrInfo.cpp
index a41ac0e44a7700b..61294bb2043bc94 100644
--- a/llvm/lib/Target/AArch64/AArch64InstrInfo.cpp
+++ b/llvm/lib/Target/AArch64/AArch64InstrInfo.cpp
@@ -7838,8 +7838,8 @@ AArch64InstrInfo::getOutlinableRanges(MachineBasicBlock &MBB,
// where these registers are dead. We will only outline from those ranges.
LiveRegUnits LRU(getRegisterInfo());
auto AreAllUnsafeRegsDead = [&LRU]() {
- return LRU.available(AArch64::W16) && LRU.available(AArch64::W17) &&
- LRU.available(AArch64::NZCV);
+ return !LRU.contains(AArch64::W16) && !LRU.contains(AArch64::W17) &&
+ !LRU.contains(AArch64::NZCV);
};
// We need to know if LR is live across an outlining boundary later on in
@@ -7909,7 +7909,7 @@ AArch64InstrInfo::getOutlinableRanges(MachineBasicBlock &MBB,
CreateNewRangeStartingAt(MI.getIterator());
continue;
}
- LRAvailableEverywhere &= LRU.available(AArch64::LR);
+ LRAvailableEverywhere &= !LRU.contains(AArch64::LR);
RangeBegin = MI.getIterator();
++RangeLen;
}
diff --git a/llvm/lib/Target/AArch64/AArch64LoadStoreOptimizer.cpp b/llvm/lib/Target/AArch64/AArch64LoadStoreOptimizer.cpp
index 41af5522d967dbf..14d1e873dbac2dd 100644
--- a/llvm/lib/Target/AArch64/AArch64LoadStoreOptimizer.cpp
+++ b/llvm/lib/Target/AArch64/AArch64LoadStoreOptimizer.cpp
@@ -1265,7 +1265,7 @@ bool AArch64LoadStoreOpt::findMatchingStore(
BaseReg == AArch64InstrInfo::getLdStBaseOp(MI).getReg() &&
AArch64InstrInfo::getLdStOffsetOp(MI).isImm() &&
isLdOffsetInRangeOfSt(LoadMI, MI, TII) &&
- ModifiedRegUnits.available(getLdStRegOp(MI).getReg())) {
+ !ModifiedRegUnits.contains(getLdStRegOp(MI).getReg())) {
StoreI = MBBI;
return true;
}
@@ -1278,7 +1278,7 @@ bool AArch64LoadStoreOpt::findMatchingStore(
// Otherwise, if the base register is modified, we have no match, so
// return early.
- if (!ModifiedRegUnits.available(BaseReg))
+ if (ModifiedRegUnits.contains(BaseReg))
return false;
// If we encounter a store aliased with the load, return early.
@@ -1510,7 +1510,7 @@ static std::optional tryToFindRegisterToRename(
auto *RegClass = TRI->getMinimalPhysRegClass(Reg);
for (const MCPhysReg &PR : *RegClass) {
- if (DefinedInBB.available(PR) && UsedInBetween.available(PR) &&
+ if (!DefinedInBB.contains(PR) && !UsedInBetween.contains(PR) &&
!RegInfo.isReserved(PR) && !AnySubOrSuperRegCalleePreserved(PR) &&
CanBeUsedForAllClasses(PR)) {
DefinedInBB.addReg(PR);
@@ -1615,9 +1615,9 @@ AArch64LoadStoreOpt::findMatchingInsn(MachineBasicBlock::iterator I,
// can't be paired: bail and keep looking.
if (IsPreLdSt) {
bool IsOutOfBounds = MIOffset != TII->getMemScale(MI);
- bool IsBaseRegUsed = !UsedRegUnits.available(
+ bool IsBaseRegUsed = UsedRegUnits.contains(
AArch64InstrInfo::getLdStBaseOp(MI).getReg());
- bool IsBaseRegModified = !ModifiedRegUnits.available(
+ bool IsBaseRegModified = ModifiedRegUnits.contains(
AArch64InstrInfo::getLdStBaseOp(MI).getReg());
// If the stored value and the address of the second instruction is
// the same, it needs to be using the updated register and therefore
@@ -1694,16 +1694,16 @@ AArch64LoadStoreOpt::findMatchingInsn(MachineBasicBlock::iterator I,
// ldr x2 [x3]
// ldr x4 [x2, #8],
// the first and third ldr cannot be converted to ldp x1, x4, [x2]
- if (!ModifiedRegUnits.available(BaseReg))
+ if (ModifiedRegUnits.contains(BaseReg))
return E;
// If the Rt of the second instruction was not modified or used between
// the two instructions and none of the instructions between the second
// and first alias with the second, we can combine the second into the
// first.
- if (ModifiedRegUnits.available(getLdStRegOp(MI).getReg()) &&
+ if (!ModifiedRegUnits.contains(getLdStRegOp(MI).getReg()) &&
!(MI.mayLoad() &&
- !UsedRegUnits.available(getLdStRegOp(MI).getReg())) &&
+ UsedRegUnits.contains(getLdStRegOp(MI).getReg())) &&
!mayAlias(MI, MemInsns, AA)) {
Flags.setMergeForward(false);
@@ -1716,10 +1716,10 @@ AArch64LoadStoreOpt::findMatchingInsn(MachineBasicBlock::iterator I,
// first and the second alias with the first, we can combine the first
// into the second.
if (!(MayLoad &&
- !UsedRegUnits.available(getLdStRegOp(FirstMI).getReg())) &&
+ UsedRegUnits.contains(getLdStRegOp(FirstMI).getReg())) &&
!mayAlias(FirstMI, MemInsns, AA)) {
- if (ModifiedRegUnits.available(getLdStRegOp(FirstMI).getReg())) {
+ if (!ModifiedRegUnits.contains(getLdStRegOp(FirstMI).getReg())) {
Flags.setMergeForward(true);
Flags.clearRenameReg();
return MBBI;
@@ -1761,7 +1761,7 @@ AArch64LoadStoreOpt::findMatchingInsn(MachineBasicBlock::iterator I,
// Otherwise, if the base register is modified, we have no match, so
// return early.
- if (!ModifiedRegUnits.available(BaseReg))
+ if (ModifiedRegUnits.contains(BaseReg))
return E;
// Update list of instructions that read/write memory.
@@ -1987,8 +1987,8 @@ MachineBasicBlock::iterator AArch64LoadStoreOpt::findMatchingUpdateInsnForward(
// return early.
// If we are optimizing SP, do not allow instructions that may load or store
// in between the load and the optimized value update.
- if (!ModifiedRegUnits.available(BaseReg) ||
- !UsedRegUnits.available(BaseReg) ||
+ if (ModifiedRegUnits.contains(BaseReg) ||
+ UsedRegUnits.contains(BaseReg) ||
(BaseRegSP && MBBI->mayLoadOrStore()))
return E;
}
@@ -2062,8 +2062,8 @@ MachineBasicBlock::iterator AArch64LoadStoreOpt::findMatchingUpdateInsnBackward(
// Otherwise, if the base register is used or modified, we have no match, so
// return early.
- if (!ModifiedRegUnits.available(BaseReg) ||
- !UsedRegUnits.available(BaseReg))
+ if (ModifiedRegUnits.contains(BaseReg) ||
+ UsedRegUnits.contains(BaseReg))
return E;
// Keep track if we have a memory access before an SP pre-increment, in this
// case we need to validate later that the update amount respects the red
diff --git a/llvm/lib/Target/AArch64/AArch64RedundantCopyElimination.cpp b/llvm/lib/Target/AArch64/AArch64RedundantCopyElimination.cpp
index 1494312886a40de..a4fc06cce5bd450 100644
--- a/llvm/lib/Target/AArch64/AArch64RedundantCopyElimination.cpp
+++ b/llvm/lib/Target/AArch64/AArch64RedundantCopyElimination.cpp
@@ -190,7 +190,7 @@ bool AArch64RedundantCopyElimination::knownRegValInBlock(
// register of the compare is not modified (including a self-clobbering
// compare) between the compare and conditional branch we known the value
// of the 1st source operand.
- if (PredI.getOperand(2).isImm() && DomBBClobberedRegs.available(SrcReg) &&
+ if (PredI.getOperand(2).isImm() && !DomBBClobberedRegs.contains(SrcReg) &&
SrcReg != DstReg) {
// We've found the instruction that sets NZCV.
int32_t KnownImm = PredI.getOperand(2).getImm();
@@ -210,7 +210,7 @@ bool AArch64RedundantCopyElimination::knownRegValInBlock(
// The destination register must not be modified between the NZCV setting
// instruction and the conditional branch.
- if (!DomBBClobberedRegs.available(DstReg))
+ if (DomBBClobberedRegs.contains(DstReg))
return Res;
FirstUse = PredI;
@@ -254,7 +254,7 @@ bool AArch64RedundantCopyElimination::knownRegValInBlock(
// The destination register of the NZCV setting instruction must not be
// modified before the conditional branch.
- if (!DomBBClobberedRegs.available(DstReg))
+ if (DomBBClobberedRegs.contains(DstReg))
return false;
// We've found the instruction that sets NZCV whose DstReg == 0.
@@ -323,12 +323,12 @@ bool AArch64RedundantCopyElimination::optimizeBlock(MachineBasicBlock *MBB) {
MCPhysReg CopyDstReg = PredI->getOperand(0).getReg();
MCPhysReg CopySrcReg = PredI->getOperand(1).getReg();
for (auto &KnownReg : KnownRegs) {
- if (!OptBBClobberedRegs.available(KnownReg.Reg))
+ if (OptBBClobberedRegs.contains(KnownReg.Reg))
continue;
// If we have X = COPY Y, and Y is known to be zero, then now X is
// known to be zero.
if (CopySrcReg == KnownReg.Reg &&
- OptBBClobberedRegs.available(CopyDstReg)) {
+ !OptBBClobberedRegs.contains(CopyDstReg)) {
KnownRegs.push_back(RegImm(CopyDstReg, KnownReg.Imm));
if (SeenFirstUse)
FirstUse = PredI;
@@ -337,7 +337,7 @@ bool AArch64RedundantCopyElimination::optimizeBlock(MachineBasicBlock *MBB) {
// If we have X = COPY Y, and X is known to be zero, then now Y is
// known to be zero.
if (CopyDstReg == KnownReg.Reg &&
- OptBBClobberedRegs.available(CopySrcReg)) {
+ !OptBBClobberedRegs.contains(CopySrcReg)) {
KnownRegs.push_back(RegImm(CopySrcReg, KnownReg.Imm));
if (SeenFirstUse)
FirstUse = PredI;
@@ -354,7 +354,7 @@ bool AArch64RedundantCopyElimination::optimizeBlock(MachineBasicBlock *MBB) {
OptBBUsedRegs, TRI);
// Stop if all of the known-zero regs have been clobbered.
if (all_of(KnownRegs, [&](RegImm KnownReg) {
- return !OptBBClobberedRegs.available(KnownReg.Reg);
+ return OptBBClobberedRegs.contains(KnownReg.Reg);
}))
break;
}
diff --git a/llvm/lib/Target/AArch64/GISel/AArch64PostSelectOptimize.cpp b/llvm/lib/Target/AArch64/GISel/AArch64PostSelectOptimize.cpp
index 94584e20f5ab3fb..ff3404b6584acc5 100644
--- a/llvm/lib/Target/AArch64/GISel/AArch64PostSelectOptimize.cpp
+++ b/llvm/lib/Target/AArch64/GISel/AArch64PostSelectOptimize.cpp
@@ -198,7 +198,7 @@ bool AArch64PostSelectOptimize::optimizeNZCVDefs(MachineBasicBlock &MBB) {
LRU.addLiveOuts(MBB);
for (auto &II : instructionsWithoutDebug(MBB.rbegin(), MBB.rend())) {
- bool NZCVDead = LRU.available(AArch64::NZCV);
+ bool NZCVDead = !LRU.contains(AArch64::NZCV);
if (NZCVDead && II.definesRegister(AArch64::NZCV)) {
// The instruction defines NZCV, but NZCV is dead.
unsigned NewOpc = getNonFlagSettingVariant(II.getOpcode());
diff --git a/llvm/lib/Target/ARM/ARMBaseInstrInfo.cpp b/llvm/lib/Target/ARM/ARMBaseInstrInfo.cpp
index 1ffdde0360cf623..f087ce71f8ba73c 100644
--- a/llvm/lib/Target/ARM/ARMBaseInstrInfo.cpp
+++ b/llvm/lib/Target/ARM/ARMBaseInstrInfo.cpp
@@ -6259,8 +6259,8 @@ bool ARMBaseInstrInfo::isMBBSafeToOutlineFrom(MachineBasicBlock ...
|
Member
|
@llvm/pr-subscribers-backend-arm ChangesCurrently there's only a provision to check whether a register is contained in the LiveRegUnits BitVector. Introducing the method which is available in LivePhysRegs to check whether the register is not reserved as well. The naming convention has been retained like in LivePhysRegs. The motivation for this PR is to separate the commits in [AMDGPU] [SIFrameLowering] Use LiveRegUnits instead of LivePhysRegs into separate PRs. This will be used to replace the use of LivePhysRegs with LiveRegUnits in AMDGPU back-end's SIFrameLowering in a coming PRPatch is 23.02 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/66061.diff 16 Files Affected:
diff --git a/llvm/include/llvm/CodeGen/LiveRegUnits.h b/llvm/include/llvm/CodeGen/LiveRegUnits.h
index e96165d6b3bbf3a..e38aab6d514750d 100644
--- a/llvm/include/llvm/CodeGen/LiveRegUnits.h
+++ b/llvm/include/llvm/CodeGen/LiveRegUnits.h
@@ -112,13 +112,13 @@ class LiveRegUnits {
/// The regmask has the same format as the one in the RegMask machine operand.
void addRegsInMask(const uint32_t *RegMask);
- /// Returns true if no part of physical register \p Reg is live.
- bool available(MCPhysReg Reg) const {
- for (MCRegUnit Unit : TRI->regunits(Reg)) {
- if (Units.test(Unit))
- return false;
- }
- return true;
+ /// Returns true if no part of physical register \p Reg is live or reserved.
+ bool available(const MachineRegisterInfo &MRI, MCPhysReg Reg) const;
+
+ /// Returns true if any part of physical register \p Reg is live
+ bool contains(MCPhysReg Reg) const {
+ return llvm::any_of(TRI->regunits(Reg),
+ [&](MCRegUnit Unit) { return Units.test(Unit); });
}
/// Updates liveness when stepping backwards over the instruction \p MI.
diff --git a/llvm/include/llvm/CodeGen/MachineOutliner.h b/llvm/include/llvm/CodeGen/MachineOutliner.h
index d0ff02fea4ff9b7..f6ecb890accd1f9 100644
--- a/llvm/include/llvm/CodeGen/MachineOutliner.h
+++ b/llvm/include/llvm/CodeGen/MachineOutliner.h
@@ -156,7 +156,7 @@ struct Candidate {
const TargetRegisterInfo &TRI) {
if (!FromEndOfBlockToStartOfSeqWasSet)
initFromEndOfBlockToStartOfSeq(TRI);
- return FromEndOfBlockToStartOfSeq.available(Reg);
+ return !FromEndOfBlockToStartOfSeq.contains(Reg);
}
/// \returns True if `isAvailableAcrossAndOutOfSeq` fails for any register
@@ -166,7 +166,7 @@ struct Candidate {
if (!FromEndOfBlockToStartOfSeqWasSet)
initFromEndOfBlockToStartOfSeq(TRI);
return any_of(Regs, [&](Register Reg) {
- return !FromEndOfBlockToStartOfSeq.available(Reg);
+ return FromEndOfBlockToStartOfSeq.contains(Reg);
});
}
@@ -181,7 +181,7 @@ struct Candidate {
bool isAvailableInsideSeq(Register Reg, const TargetRegisterInfo &TRI) {
if (!InSeqWasSet)
initInSeq(TRI);
- return InSeq.available(Reg);
+ return !InSeq.contains(Reg);
}
/// The number of instructions that would be saved by outlining every
diff --git a/llvm/lib/CodeGen/DeadMachineInstructionElim.cpp b/llvm/lib/CodeGen/DeadMachineInstructionElim.cpp
index 6a7de3b241feeea..e3c5c57ec068c94 100644
--- a/llvm/lib/CodeGen/DeadMachineInstructionElim.cpp
+++ b/llvm/lib/CodeGen/DeadMachineInstructionElim.cpp
@@ -79,7 +79,7 @@ bool DeadMachineInstructionElim::isDead(const MachineInstr *MI) const {
Register Reg = MO.getReg();
if (Reg.isPhysical()) {
// Don't delete live physreg defs, or any reserved register defs.
- if (!LivePhysRegs.available(Reg) || MRI->isReserved(Reg))
+ if (!LivePhysRegs.available(*MRI, Reg))
return false;
} else {
if (MO.isDead()) {
diff --git a/llvm/lib/CodeGen/LiveRegUnits.cpp b/llvm/lib/CodeGen/LiveRegUnits.cpp
index 34de09dd2944bc7..d711cd44f260da4 100644
--- a/llvm/lib/CodeGen/LiveRegUnits.cpp
+++ b/llvm/lib/CodeGen/LiveRegUnits.cpp
@@ -41,6 +41,11 @@ void LiveRegUnits::addRegsInMask(const uint32_t *RegMask) {
}
}
+bool LiveRegUnits::available(const MachineRegisterInfo &MRI,
+ MCPhysReg Reg) const {
+ return !MRI.isReserved(Reg) && !contains(Reg);
+}
+
void LiveRegUnits::stepBackward(const MachineInstr &MI) {
// Remove defined registers and regmask kills from the set.
for (const MachineOperand &MOP : MI.operands()) {
diff --git a/llvm/lib/CodeGen/MachineSink.cpp b/llvm/lib/CodeGen/MachineSink.cpp
index b4cbb93d758ef2f..c73d17dc2f1a669 100644
--- a/llvm/lib/CodeGen/MachineSink.cpp
+++ b/llvm/lib/CodeGen/MachineSink.cpp
@@ -1627,7 +1627,7 @@ static bool aliasWithRegsInLiveIn(MachineBasicBlock &MBB, unsigned Reg,
const TargetRegisterInfo *TRI) {
LiveRegUnits LiveInRegUnits(*TRI);
LiveInRegUnits.addLiveIns(MBB);
- return !LiveInRegUnits.available(Reg);
+ return LiveInRegUnits.contains(Reg);
}
static MachineBasicBlock *
@@ -1680,7 +1680,7 @@ static void clearKillFlags(MachineInstr *MI, MachineBasicBlock &CurBB,
for (auto U : UsedOpsInCopy) {
MachineOperand &MO = MI->getOperand(U);
Register SrcReg = MO.getReg();
- if (!UsedRegUnits.available(SrcReg)) {
+ if (UsedRegUnits.contains(SrcReg)) {
MachineBasicBlock::iterator NI = std::next(MI->getIterator());
for (MachineInstr &UI : make_range(NI, CurBB.end())) {
if (UI.killsRegister(SrcReg, TRI)) {
@@ -1725,7 +1725,7 @@ static bool hasRegisterDependency(MachineInstr *MI,
if (!Reg)
continue;
if (MO.isDef()) {
- if (!ModifiedRegUnits.available(Reg) || !UsedRegUnits.available(Reg)) {
+ if (ModifiedRegUnits.contains(Reg) || UsedRegUnits.contains(Reg)) {
HasRegDependency = true;
break;
}
@@ -1736,7 +1736,7 @@ static bool hasRegisterDependency(MachineInstr *MI,
// it's not perfectly clear if skipping the internal read is safe in all
// other targets.
} else if (MO.isUse()) {
- if (!ModifiedRegUnits.available(Reg)) {
+ if (ModifiedRegUnits.contains(Reg)) {
HasRegDependency = true;
break;
}
diff --git a/llvm/lib/CodeGen/RegisterScavenging.cpp b/llvm/lib/CodeGen/RegisterScavenging.cpp
index e6ff5701bc4bd20..938fc9ba215c74e 100644
--- a/llvm/lib/CodeGen/RegisterScavenging.cpp
+++ b/llvm/lib/CodeGen/RegisterScavenging.cpp
@@ -109,7 +109,7 @@ void RegScavenger::backward() {
bool RegScavenger::isRegUsed(Register Reg, bool includeReserved) const {
if (isReserved(Reg))
return includeReserved;
- return !LiveUnits.available(Reg);
+ return LiveUnits.contains(Reg);
}
Register RegScavenger::FindUnusedReg(const TargetRegisterClass *RC) const {
@@ -164,8 +164,8 @@ findSurvivorBackwards(const MachineRegisterInfo &MRI,
if (I == To) {
// See if one of the registers in RC wasn't used so far.
for (MCPhysReg Reg : AllocationOrder) {
- if (!MRI.isReserved(Reg) && Used.available(Reg) &&
- LiveOut.available(Reg))
+ if (!MRI.isReserved(Reg) && !Used.contains(Reg) &&
+ !LiveOut.contains(Reg))
return std::make_pair(Reg, MBB.end());
}
// Otherwise we will continue up to InstrLimit instructions to find
@@ -186,10 +186,10 @@ findSurvivorBackwards(const MachineRegisterInfo &MRI,
MI.getFlag(MachineInstr::FrameSetup))
break;
- if (Survivor == 0 || !Used.available(Survivor)) {
+ if (Survivor == 0 || Used.contains(Survivor)) {
MCPhysReg AvilableReg = 0;
for (MCPhysReg Reg : AllocationOrder) {
- if (!MRI.isReserved(Reg) && Used.available(Reg)) {
+ if (Used.available(MRI, Reg)) {
AvilableReg = Reg;
break;
}
diff --git a/llvm/lib/Target/AArch64/AArch64A57FPLoadBalancing.cpp b/llvm/lib/Target/AArch64/AArch64A57FPLoadBalancing.cpp
index ea8e30269ece25f..17e5a60438f903a 100644
--- a/llvm/lib/Target/AArch64/AArch64A57FPLoadBalancing.cpp
+++ b/llvm/lib/Target/AArch64/AArch64A57FPLoadBalancing.cpp
@@ -518,7 +518,7 @@ int AArch64A57FPLoadBalancing::scavengeRegister(Chain *G, Color C,
unsigned RegClassID = ChainBegin->getDesc().operands()[0].RegClass;
auto Ord = RCI.getOrder(TRI->getRegClass(RegClassID));
for (auto Reg : Ord) {
- if (!Units.available(Reg))
+ if (Units.contains(Reg))
continue;
if (C == getColor(Reg))
return Reg;
diff --git a/llvm/lib/Target/AArch64/AArch64FalkorHWPFFix.cpp b/llvm/lib/Target/AArch64/AArch64FalkorHWPFFix.cpp
index 32686b25f21184b..7d481efb0136e7b 100644
--- a/llvm/lib/Target/AArch64/AArch64FalkorHWPFFix.cpp
+++ b/llvm/lib/Target/AArch64/AArch64FalkorHWPFFix.cpp
@@ -748,7 +748,7 @@ void FalkorHWPFFix::runOnLoop(MachineLoop &L, MachineFunction &Fn) {
}
for (unsigned ScratchReg : AArch64::GPR64RegClass) {
- if (!LR.available(ScratchReg) || MRI.isReserved(ScratchReg))
+ if (!LR.available(MRI, ScratchReg))
continue;
LoadInfo NewLdI(LdI);
diff --git a/llvm/lib/Target/AArch64/AArch64InstrInfo.cpp b/llvm/lib/Target/AArch64/AArch64InstrInfo.cpp
index a41ac0e44a7700b..61294bb2043bc94 100644
--- a/llvm/lib/Target/AArch64/AArch64InstrInfo.cpp
+++ b/llvm/lib/Target/AArch64/AArch64InstrInfo.cpp
@@ -7838,8 +7838,8 @@ AArch64InstrInfo::getOutlinableRanges(MachineBasicBlock &MBB,
// where these registers are dead. We will only outline from those ranges.
LiveRegUnits LRU(getRegisterInfo());
auto AreAllUnsafeRegsDead = [&LRU]() {
- return LRU.available(AArch64::W16) && LRU.available(AArch64::W17) &&
- LRU.available(AArch64::NZCV);
+ return !LRU.contains(AArch64::W16) && !LRU.contains(AArch64::W17) &&
+ !LRU.contains(AArch64::NZCV);
};
// We need to know if LR is live across an outlining boundary later on in
@@ -7909,7 +7909,7 @@ AArch64InstrInfo::getOutlinableRanges(MachineBasicBlock &MBB,
CreateNewRangeStartingAt(MI.getIterator());
continue;
}
- LRAvailableEverywhere &= LRU.available(AArch64::LR);
+ LRAvailableEverywhere &= !LRU.contains(AArch64::LR);
RangeBegin = MI.getIterator();
++RangeLen;
}
diff --git a/llvm/lib/Target/AArch64/AArch64LoadStoreOptimizer.cpp b/llvm/lib/Target/AArch64/AArch64LoadStoreOptimizer.cpp
index 41af5522d967dbf..14d1e873dbac2dd 100644
--- a/llvm/lib/Target/AArch64/AArch64LoadStoreOptimizer.cpp
+++ b/llvm/lib/Target/AArch64/AArch64LoadStoreOptimizer.cpp
@@ -1265,7 +1265,7 @@ bool AArch64LoadStoreOpt::findMatchingStore(
BaseReg == AArch64InstrInfo::getLdStBaseOp(MI).getReg() &&
AArch64InstrInfo::getLdStOffsetOp(MI).isImm() &&
isLdOffsetInRangeOfSt(LoadMI, MI, TII) &&
- ModifiedRegUnits.available(getLdStRegOp(MI).getReg())) {
+ !ModifiedRegUnits.contains(getLdStRegOp(MI).getReg())) {
StoreI = MBBI;
return true;
}
@@ -1278,7 +1278,7 @@ bool AArch64LoadStoreOpt::findMatchingStore(
// Otherwise, if the base register is modified, we have no match, so
// return early.
- if (!ModifiedRegUnits.available(BaseReg))
+ if (ModifiedRegUnits.contains(BaseReg))
return false;
// If we encounter a store aliased with the load, return early.
@@ -1510,7 +1510,7 @@ static std::optional tryToFindRegisterToRename(
auto *RegClass = TRI->getMinimalPhysRegClass(Reg);
for (const MCPhysReg &PR : *RegClass) {
- if (DefinedInBB.available(PR) && UsedInBetween.available(PR) &&
+ if (!DefinedInBB.contains(PR) && !UsedInBetween.contains(PR) &&
!RegInfo.isReserved(PR) && !AnySubOrSuperRegCalleePreserved(PR) &&
CanBeUsedForAllClasses(PR)) {
DefinedInBB.addReg(PR);
@@ -1615,9 +1615,9 @@ AArch64LoadStoreOpt::findMatchingInsn(MachineBasicBlock::iterator I,
// can't be paired: bail and keep looking.
if (IsPreLdSt) {
bool IsOutOfBounds = MIOffset != TII->getMemScale(MI);
- bool IsBaseRegUsed = !UsedRegUnits.available(
+ bool IsBaseRegUsed = UsedRegUnits.contains(
AArch64InstrInfo::getLdStBaseOp(MI).getReg());
- bool IsBaseRegModified = !ModifiedRegUnits.available(
+ bool IsBaseRegModified = ModifiedRegUnits.contains(
AArch64InstrInfo::getLdStBaseOp(MI).getReg());
// If the stored value and the address of the second instruction is
// the same, it needs to be using the updated register and therefore
@@ -1694,16 +1694,16 @@ AArch64LoadStoreOpt::findMatchingInsn(MachineBasicBlock::iterator I,
// ldr x2 [x3]
// ldr x4 [x2, #8],
// the first and third ldr cannot be converted to ldp x1, x4, [x2]
- if (!ModifiedRegUnits.available(BaseReg))
+ if (ModifiedRegUnits.contains(BaseReg))
return E;
// If the Rt of the second instruction was not modified or used between
// the two instructions and none of the instructions between the second
// and first alias with the second, we can combine the second into the
// first.
- if (ModifiedRegUnits.available(getLdStRegOp(MI).getReg()) &&
+ if (!ModifiedRegUnits.contains(getLdStRegOp(MI).getReg()) &&
!(MI.mayLoad() &&
- !UsedRegUnits.available(getLdStRegOp(MI).getReg())) &&
+ UsedRegUnits.contains(getLdStRegOp(MI).getReg())) &&
!mayAlias(MI, MemInsns, AA)) {
Flags.setMergeForward(false);
@@ -1716,10 +1716,10 @@ AArch64LoadStoreOpt::findMatchingInsn(MachineBasicBlock::iterator I,
// first and the second alias with the first, we can combine the first
// into the second.
if (!(MayLoad &&
- !UsedRegUnits.available(getLdStRegOp(FirstMI).getReg())) &&
+ UsedRegUnits.contains(getLdStRegOp(FirstMI).getReg())) &&
!mayAlias(FirstMI, MemInsns, AA)) {
- if (ModifiedRegUnits.available(getLdStRegOp(FirstMI).getReg())) {
+ if (!ModifiedRegUnits.contains(getLdStRegOp(FirstMI).getReg())) {
Flags.setMergeForward(true);
Flags.clearRenameReg();
return MBBI;
@@ -1761,7 +1761,7 @@ AArch64LoadStoreOpt::findMatchingInsn(MachineBasicBlock::iterator I,
// Otherwise, if the base register is modified, we have no match, so
// return early.
- if (!ModifiedRegUnits.available(BaseReg))
+ if (ModifiedRegUnits.contains(BaseReg))
return E;
// Update list of instructions that read/write memory.
@@ -1987,8 +1987,8 @@ MachineBasicBlock::iterator AArch64LoadStoreOpt::findMatchingUpdateInsnForward(
// return early.
// If we are optimizing SP, do not allow instructions that may load or store
// in between the load and the optimized value update.
- if (!ModifiedRegUnits.available(BaseReg) ||
- !UsedRegUnits.available(BaseReg) ||
+ if (ModifiedRegUnits.contains(BaseReg) ||
+ UsedRegUnits.contains(BaseReg) ||
(BaseRegSP && MBBI->mayLoadOrStore()))
return E;
}
@@ -2062,8 +2062,8 @@ MachineBasicBlock::iterator AArch64LoadStoreOpt::findMatchingUpdateInsnBackward(
// Otherwise, if the base register is used or modified, we have no match, so
// return early.
- if (!ModifiedRegUnits.available(BaseReg) ||
- !UsedRegUnits.available(BaseReg))
+ if (ModifiedRegUnits.contains(BaseReg) ||
+ UsedRegUnits.contains(BaseReg))
return E;
// Keep track if we have a memory access before an SP pre-increment, in this
// case we need to validate later that the update amount respects the red
diff --git a/llvm/lib/Target/AArch64/AArch64RedundantCopyElimination.cpp b/llvm/lib/Target/AArch64/AArch64RedundantCopyElimination.cpp
index 1494312886a40de..a4fc06cce5bd450 100644
--- a/llvm/lib/Target/AArch64/AArch64RedundantCopyElimination.cpp
+++ b/llvm/lib/Target/AArch64/AArch64RedundantCopyElimination.cpp
@@ -190,7 +190,7 @@ bool AArch64RedundantCopyElimination::knownRegValInBlock(
// register of the compare is not modified (including a self-clobbering
// compare) between the compare and conditional branch we known the value
// of the 1st source operand.
- if (PredI.getOperand(2).isImm() && DomBBClobberedRegs.available(SrcReg) &&
+ if (PredI.getOperand(2).isImm() && !DomBBClobberedRegs.contains(SrcReg) &&
SrcReg != DstReg) {
// We've found the instruction that sets NZCV.
int32_t KnownImm = PredI.getOperand(2).getImm();
@@ -210,7 +210,7 @@ bool AArch64RedundantCopyElimination::knownRegValInBlock(
// The destination register must not be modified between the NZCV setting
// instruction and the conditional branch.
- if (!DomBBClobberedRegs.available(DstReg))
+ if (DomBBClobberedRegs.contains(DstReg))
return Res;
FirstUse = PredI;
@@ -254,7 +254,7 @@ bool AArch64RedundantCopyElimination::knownRegValInBlock(
// The destination register of the NZCV setting instruction must not be
// modified before the conditional branch.
- if (!DomBBClobberedRegs.available(DstReg))
+ if (DomBBClobberedRegs.contains(DstReg))
return false;
// We've found the instruction that sets NZCV whose DstReg == 0.
@@ -323,12 +323,12 @@ bool AArch64RedundantCopyElimination::optimizeBlock(MachineBasicBlock *MBB) {
MCPhysReg CopyDstReg = PredI->getOperand(0).getReg();
MCPhysReg CopySrcReg = PredI->getOperand(1).getReg();
for (auto &KnownReg : KnownRegs) {
- if (!OptBBClobberedRegs.available(KnownReg.Reg))
+ if (OptBBClobberedRegs.contains(KnownReg.Reg))
continue;
// If we have X = COPY Y, and Y is known to be zero, then now X is
// known to be zero.
if (CopySrcReg == KnownReg.Reg &&
- OptBBClobberedRegs.available(CopyDstReg)) {
+ !OptBBClobberedRegs.contains(CopyDstReg)) {
KnownRegs.push_back(RegImm(CopyDstReg, KnownReg.Imm));
if (SeenFirstUse)
FirstUse = PredI;
@@ -337,7 +337,7 @@ bool AArch64RedundantCopyElimination::optimizeBlock(MachineBasicBlock *MBB) {
// If we have X = COPY Y, and X is known to be zero, then now Y is
// known to be zero.
if (CopyDstReg == KnownReg.Reg &&
- OptBBClobberedRegs.available(CopySrcReg)) {
+ !OptBBClobberedRegs.contains(CopySrcReg)) {
KnownRegs.push_back(RegImm(CopySrcReg, KnownReg.Imm));
if (SeenFirstUse)
FirstUse = PredI;
@@ -354,7 +354,7 @@ bool AArch64RedundantCopyElimination::optimizeBlock(MachineBasicBlock *MBB) {
OptBBUsedRegs, TRI);
// Stop if all of the known-zero regs have been clobbered.
if (all_of(KnownRegs, [&](RegImm KnownReg) {
- return !OptBBClobberedRegs.available(KnownReg.Reg);
+ return OptBBClobberedRegs.contains(KnownReg.Reg);
}))
break;
}
diff --git a/llvm/lib/Target/AArch64/GISel/AArch64PostSelectOptimize.cpp b/llvm/lib/Target/AArch64/GISel/AArch64PostSelectOptimize.cpp
index 94584e20f5ab3fb..ff3404b6584acc5 100644
--- a/llvm/lib/Target/AArch64/GISel/AArch64PostSelectOptimize.cpp
+++ b/llvm/lib/Target/AArch64/GISel/AArch64PostSelectOptimize.cpp
@@ -198,7 +198,7 @@ bool AArch64PostSelectOptimize::optimizeNZCVDefs(MachineBasicBlock &MBB) {
LRU.addLiveOuts(MBB);
for (auto &II : instructionsWithoutDebug(MBB.rbegin(), MBB.rend())) {
- bool NZCVDead = LRU.available(AArch64::NZCV);
+ bool NZCVDead = !LRU.contains(AArch64::NZCV);
if (NZCVDead && II.definesRegister(AArch64::NZCV)) {
// The instruction defines NZCV, but NZCV is dead.
unsigned NewOpc = getNonFlagSettingVariant(II.getOpcode());
diff --git a/llvm/lib/Target/ARM/ARMBaseInstrInfo.cpp b/llvm/lib/Target/ARM/ARMBaseInstrInfo.cpp
index 1ffdde0360cf623..f087ce71f8ba73c 100644
--- a/llvm/lib/Target/ARM/ARMBaseInstrInfo.cpp
+++ b/llvm/lib/Target/ARM/ARMBaseInstrInfo.cpp
@@ -6259,8 +6259,8 @@ bool ARMBaseInstrInfo::isMBBSafeToOutlineFrom(MachineBasicBlock ...
|
Member
|
@llvm/pr-subscribers-llvm-regalloc ChangesCurrently there's only a provision to check whether a register is contained in the LiveRegUnits BitVector. Introducing the method which is available in LivePhysRegs to check whether the register is not reserved as well. The naming convention has been retained like in LivePhysRegs. The motivation for this PR is to separate the commits in [AMDGPU] [SIFrameLowering] Use LiveRegUnits instead of LivePhysRegs into separate PRs. This will be used to replace the use of LivePhysRegs with LiveRegUnits in AMDGPU back-end's SIFrameLowering in a coming PRPatch is 23.02 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/66061.diff 16 Files Affected:
diff --git a/llvm/include/llvm/CodeGen/LiveRegUnits.h b/llvm/include/llvm/CodeGen/LiveRegUnits.h
index e96165d6b3bbf3a..e38aab6d514750d 100644
--- a/llvm/include/llvm/CodeGen/LiveRegUnits.h
+++ b/llvm/include/llvm/CodeGen/LiveRegUnits.h
@@ -112,13 +112,13 @@ class LiveRegUnits {
/// The regmask has the same format as the one in the RegMask machine operand.
void addRegsInMask(const uint32_t *RegMask);
- /// Returns true if no part of physical register \p Reg is live.
- bool available(MCPhysReg Reg) const {
- for (MCRegUnit Unit : TRI->regunits(Reg)) {
- if (Units.test(Unit))
- return false;
- }
- return true;
+ /// Returns true if no part of physical register \p Reg is live or reserved.
+ bool available(const MachineRegisterInfo &MRI, MCPhysReg Reg) const;
+
+ /// Returns true if any part of physical register \p Reg is live
+ bool contains(MCPhysReg Reg) const {
+ return llvm::any_of(TRI->regunits(Reg),
+ [&](MCRegUnit Unit) { return Units.test(Unit); });
}
/// Updates liveness when stepping backwards over the instruction \p MI.
diff --git a/llvm/include/llvm/CodeGen/MachineOutliner.h b/llvm/include/llvm/CodeGen/MachineOutliner.h
index d0ff02fea4ff9b7..f6ecb890accd1f9 100644
--- a/llvm/include/llvm/CodeGen/MachineOutliner.h
+++ b/llvm/include/llvm/CodeGen/MachineOutliner.h
@@ -156,7 +156,7 @@ struct Candidate {
const TargetRegisterInfo &TRI) {
if (!FromEndOfBlockToStartOfSeqWasSet)
initFromEndOfBlockToStartOfSeq(TRI);
- return FromEndOfBlockToStartOfSeq.available(Reg);
+ return !FromEndOfBlockToStartOfSeq.contains(Reg);
}
/// \returns True if `isAvailableAcrossAndOutOfSeq` fails for any register
@@ -166,7 +166,7 @@ struct Candidate {
if (!FromEndOfBlockToStartOfSeqWasSet)
initFromEndOfBlockToStartOfSeq(TRI);
return any_of(Regs, [&](Register Reg) {
- return !FromEndOfBlockToStartOfSeq.available(Reg);
+ return FromEndOfBlockToStartOfSeq.contains(Reg);
});
}
@@ -181,7 +181,7 @@ struct Candidate {
bool isAvailableInsideSeq(Register Reg, const TargetRegisterInfo &TRI) {
if (!InSeqWasSet)
initInSeq(TRI);
- return InSeq.available(Reg);
+ return !InSeq.contains(Reg);
}
/// The number of instructions that would be saved by outlining every
diff --git a/llvm/lib/CodeGen/DeadMachineInstructionElim.cpp b/llvm/lib/CodeGen/DeadMachineInstructionElim.cpp
index 6a7de3b241feeea..e3c5c57ec068c94 100644
--- a/llvm/lib/CodeGen/DeadMachineInstructionElim.cpp
+++ b/llvm/lib/CodeGen/DeadMachineInstructionElim.cpp
@@ -79,7 +79,7 @@ bool DeadMachineInstructionElim::isDead(const MachineInstr *MI) const {
Register Reg = MO.getReg();
if (Reg.isPhysical()) {
// Don't delete live physreg defs, or any reserved register defs.
- if (!LivePhysRegs.available(Reg) || MRI->isReserved(Reg))
+ if (!LivePhysRegs.available(*MRI, Reg))
return false;
} else {
if (MO.isDead()) {
diff --git a/llvm/lib/CodeGen/LiveRegUnits.cpp b/llvm/lib/CodeGen/LiveRegUnits.cpp
index 34de09dd2944bc7..d711cd44f260da4 100644
--- a/llvm/lib/CodeGen/LiveRegUnits.cpp
+++ b/llvm/lib/CodeGen/LiveRegUnits.cpp
@@ -41,6 +41,11 @@ void LiveRegUnits::addRegsInMask(const uint32_t *RegMask) {
}
}
+bool LiveRegUnits::available(const MachineRegisterInfo &MRI,
+ MCPhysReg Reg) const {
+ return !MRI.isReserved(Reg) && !contains(Reg);
+}
+
void LiveRegUnits::stepBackward(const MachineInstr &MI) {
// Remove defined registers and regmask kills from the set.
for (const MachineOperand &MOP : MI.operands()) {
diff --git a/llvm/lib/CodeGen/MachineSink.cpp b/llvm/lib/CodeGen/MachineSink.cpp
index b4cbb93d758ef2f..c73d17dc2f1a669 100644
--- a/llvm/lib/CodeGen/MachineSink.cpp
+++ b/llvm/lib/CodeGen/MachineSink.cpp
@@ -1627,7 +1627,7 @@ static bool aliasWithRegsInLiveIn(MachineBasicBlock &MBB, unsigned Reg,
const TargetRegisterInfo *TRI) {
LiveRegUnits LiveInRegUnits(*TRI);
LiveInRegUnits.addLiveIns(MBB);
- return !LiveInRegUnits.available(Reg);
+ return LiveInRegUnits.contains(Reg);
}
static MachineBasicBlock *
@@ -1680,7 +1680,7 @@ static void clearKillFlags(MachineInstr *MI, MachineBasicBlock &CurBB,
for (auto U : UsedOpsInCopy) {
MachineOperand &MO = MI->getOperand(U);
Register SrcReg = MO.getReg();
- if (!UsedRegUnits.available(SrcReg)) {
+ if (UsedRegUnits.contains(SrcReg)) {
MachineBasicBlock::iterator NI = std::next(MI->getIterator());
for (MachineInstr &UI : make_range(NI, CurBB.end())) {
if (UI.killsRegister(SrcReg, TRI)) {
@@ -1725,7 +1725,7 @@ static bool hasRegisterDependency(MachineInstr *MI,
if (!Reg)
continue;
if (MO.isDef()) {
- if (!ModifiedRegUnits.available(Reg) || !UsedRegUnits.available(Reg)) {
+ if (ModifiedRegUnits.contains(Reg) || UsedRegUnits.contains(Reg)) {
HasRegDependency = true;
break;
}
@@ -1736,7 +1736,7 @@ static bool hasRegisterDependency(MachineInstr *MI,
// it's not perfectly clear if skipping the internal read is safe in all
// other targets.
} else if (MO.isUse()) {
- if (!ModifiedRegUnits.available(Reg)) {
+ if (ModifiedRegUnits.contains(Reg)) {
HasRegDependency = true;
break;
}
diff --git a/llvm/lib/CodeGen/RegisterScavenging.cpp b/llvm/lib/CodeGen/RegisterScavenging.cpp
index e6ff5701bc4bd20..938fc9ba215c74e 100644
--- a/llvm/lib/CodeGen/RegisterScavenging.cpp
+++ b/llvm/lib/CodeGen/RegisterScavenging.cpp
@@ -109,7 +109,7 @@ void RegScavenger::backward() {
bool RegScavenger::isRegUsed(Register Reg, bool includeReserved) const {
if (isReserved(Reg))
return includeReserved;
- return !LiveUnits.available(Reg);
+ return LiveUnits.contains(Reg);
}
Register RegScavenger::FindUnusedReg(const TargetRegisterClass *RC) const {
@@ -164,8 +164,8 @@ findSurvivorBackwards(const MachineRegisterInfo &MRI,
if (I == To) {
// See if one of the registers in RC wasn't used so far.
for (MCPhysReg Reg : AllocationOrder) {
- if (!MRI.isReserved(Reg) && Used.available(Reg) &&
- LiveOut.available(Reg))
+ if (!MRI.isReserved(Reg) && !Used.contains(Reg) &&
+ !LiveOut.contains(Reg))
return std::make_pair(Reg, MBB.end());
}
// Otherwise we will continue up to InstrLimit instructions to find
@@ -186,10 +186,10 @@ findSurvivorBackwards(const MachineRegisterInfo &MRI,
MI.getFlag(MachineInstr::FrameSetup))
break;
- if (Survivor == 0 || !Used.available(Survivor)) {
+ if (Survivor == 0 || Used.contains(Survivor)) {
MCPhysReg AvilableReg = 0;
for (MCPhysReg Reg : AllocationOrder) {
- if (!MRI.isReserved(Reg) && Used.available(Reg)) {
+ if (Used.available(MRI, Reg)) {
AvilableReg = Reg;
break;
}
diff --git a/llvm/lib/Target/AArch64/AArch64A57FPLoadBalancing.cpp b/llvm/lib/Target/AArch64/AArch64A57FPLoadBalancing.cpp
index ea8e30269ece25f..17e5a60438f903a 100644
--- a/llvm/lib/Target/AArch64/AArch64A57FPLoadBalancing.cpp
+++ b/llvm/lib/Target/AArch64/AArch64A57FPLoadBalancing.cpp
@@ -518,7 +518,7 @@ int AArch64A57FPLoadBalancing::scavengeRegister(Chain *G, Color C,
unsigned RegClassID = ChainBegin->getDesc().operands()[0].RegClass;
auto Ord = RCI.getOrder(TRI->getRegClass(RegClassID));
for (auto Reg : Ord) {
- if (!Units.available(Reg))
+ if (Units.contains(Reg))
continue;
if (C == getColor(Reg))
return Reg;
diff --git a/llvm/lib/Target/AArch64/AArch64FalkorHWPFFix.cpp b/llvm/lib/Target/AArch64/AArch64FalkorHWPFFix.cpp
index 32686b25f21184b..7d481efb0136e7b 100644
--- a/llvm/lib/Target/AArch64/AArch64FalkorHWPFFix.cpp
+++ b/llvm/lib/Target/AArch64/AArch64FalkorHWPFFix.cpp
@@ -748,7 +748,7 @@ void FalkorHWPFFix::runOnLoop(MachineLoop &L, MachineFunction &Fn) {
}
for (unsigned ScratchReg : AArch64::GPR64RegClass) {
- if (!LR.available(ScratchReg) || MRI.isReserved(ScratchReg))
+ if (!LR.available(MRI, ScratchReg))
continue;
LoadInfo NewLdI(LdI);
diff --git a/llvm/lib/Target/AArch64/AArch64InstrInfo.cpp b/llvm/lib/Target/AArch64/AArch64InstrInfo.cpp
index a41ac0e44a7700b..61294bb2043bc94 100644
--- a/llvm/lib/Target/AArch64/AArch64InstrInfo.cpp
+++ b/llvm/lib/Target/AArch64/AArch64InstrInfo.cpp
@@ -7838,8 +7838,8 @@ AArch64InstrInfo::getOutlinableRanges(MachineBasicBlock &MBB,
// where these registers are dead. We will only outline from those ranges.
LiveRegUnits LRU(getRegisterInfo());
auto AreAllUnsafeRegsDead = [&LRU]() {
- return LRU.available(AArch64::W16) && LRU.available(AArch64::W17) &&
- LRU.available(AArch64::NZCV);
+ return !LRU.contains(AArch64::W16) && !LRU.contains(AArch64::W17) &&
+ !LRU.contains(AArch64::NZCV);
};
// We need to know if LR is live across an outlining boundary later on in
@@ -7909,7 +7909,7 @@ AArch64InstrInfo::getOutlinableRanges(MachineBasicBlock &MBB,
CreateNewRangeStartingAt(MI.getIterator());
continue;
}
- LRAvailableEverywhere &= LRU.available(AArch64::LR);
+ LRAvailableEverywhere &= !LRU.contains(AArch64::LR);
RangeBegin = MI.getIterator();
++RangeLen;
}
diff --git a/llvm/lib/Target/AArch64/AArch64LoadStoreOptimizer.cpp b/llvm/lib/Target/AArch64/AArch64LoadStoreOptimizer.cpp
index 41af5522d967dbf..14d1e873dbac2dd 100644
--- a/llvm/lib/Target/AArch64/AArch64LoadStoreOptimizer.cpp
+++ b/llvm/lib/Target/AArch64/AArch64LoadStoreOptimizer.cpp
@@ -1265,7 +1265,7 @@ bool AArch64LoadStoreOpt::findMatchingStore(
BaseReg == AArch64InstrInfo::getLdStBaseOp(MI).getReg() &&
AArch64InstrInfo::getLdStOffsetOp(MI).isImm() &&
isLdOffsetInRangeOfSt(LoadMI, MI, TII) &&
- ModifiedRegUnits.available(getLdStRegOp(MI).getReg())) {
+ !ModifiedRegUnits.contains(getLdStRegOp(MI).getReg())) {
StoreI = MBBI;
return true;
}
@@ -1278,7 +1278,7 @@ bool AArch64LoadStoreOpt::findMatchingStore(
// Otherwise, if the base register is modified, we have no match, so
// return early.
- if (!ModifiedRegUnits.available(BaseReg))
+ if (ModifiedRegUnits.contains(BaseReg))
return false;
// If we encounter a store aliased with the load, return early.
@@ -1510,7 +1510,7 @@ static std::optional tryToFindRegisterToRename(
auto *RegClass = TRI->getMinimalPhysRegClass(Reg);
for (const MCPhysReg &PR : *RegClass) {
- if (DefinedInBB.available(PR) && UsedInBetween.available(PR) &&
+ if (!DefinedInBB.contains(PR) && !UsedInBetween.contains(PR) &&
!RegInfo.isReserved(PR) && !AnySubOrSuperRegCalleePreserved(PR) &&
CanBeUsedForAllClasses(PR)) {
DefinedInBB.addReg(PR);
@@ -1615,9 +1615,9 @@ AArch64LoadStoreOpt::findMatchingInsn(MachineBasicBlock::iterator I,
// can't be paired: bail and keep looking.
if (IsPreLdSt) {
bool IsOutOfBounds = MIOffset != TII->getMemScale(MI);
- bool IsBaseRegUsed = !UsedRegUnits.available(
+ bool IsBaseRegUsed = UsedRegUnits.contains(
AArch64InstrInfo::getLdStBaseOp(MI).getReg());
- bool IsBaseRegModified = !ModifiedRegUnits.available(
+ bool IsBaseRegModified = ModifiedRegUnits.contains(
AArch64InstrInfo::getLdStBaseOp(MI).getReg());
// If the stored value and the address of the second instruction is
// the same, it needs to be using the updated register and therefore
@@ -1694,16 +1694,16 @@ AArch64LoadStoreOpt::findMatchingInsn(MachineBasicBlock::iterator I,
// ldr x2 [x3]
// ldr x4 [x2, #8],
// the first and third ldr cannot be converted to ldp x1, x4, [x2]
- if (!ModifiedRegUnits.available(BaseReg))
+ if (ModifiedRegUnits.contains(BaseReg))
return E;
// If the Rt of the second instruction was not modified or used between
// the two instructions and none of the instructions between the second
// and first alias with the second, we can combine the second into the
// first.
- if (ModifiedRegUnits.available(getLdStRegOp(MI).getReg()) &&
+ if (!ModifiedRegUnits.contains(getLdStRegOp(MI).getReg()) &&
!(MI.mayLoad() &&
- !UsedRegUnits.available(getLdStRegOp(MI).getReg())) &&
+ UsedRegUnits.contains(getLdStRegOp(MI).getReg())) &&
!mayAlias(MI, MemInsns, AA)) {
Flags.setMergeForward(false);
@@ -1716,10 +1716,10 @@ AArch64LoadStoreOpt::findMatchingInsn(MachineBasicBlock::iterator I,
// first and the second alias with the first, we can combine the first
// into the second.
if (!(MayLoad &&
- !UsedRegUnits.available(getLdStRegOp(FirstMI).getReg())) &&
+ UsedRegUnits.contains(getLdStRegOp(FirstMI).getReg())) &&
!mayAlias(FirstMI, MemInsns, AA)) {
- if (ModifiedRegUnits.available(getLdStRegOp(FirstMI).getReg())) {
+ if (!ModifiedRegUnits.contains(getLdStRegOp(FirstMI).getReg())) {
Flags.setMergeForward(true);
Flags.clearRenameReg();
return MBBI;
@@ -1761,7 +1761,7 @@ AArch64LoadStoreOpt::findMatchingInsn(MachineBasicBlock::iterator I,
// Otherwise, if the base register is modified, we have no match, so
// return early.
- if (!ModifiedRegUnits.available(BaseReg))
+ if (ModifiedRegUnits.contains(BaseReg))
return E;
// Update list of instructions that read/write memory.
@@ -1987,8 +1987,8 @@ MachineBasicBlock::iterator AArch64LoadStoreOpt::findMatchingUpdateInsnForward(
// return early.
// If we are optimizing SP, do not allow instructions that may load or store
// in between the load and the optimized value update.
- if (!ModifiedRegUnits.available(BaseReg) ||
- !UsedRegUnits.available(BaseReg) ||
+ if (ModifiedRegUnits.contains(BaseReg) ||
+ UsedRegUnits.contains(BaseReg) ||
(BaseRegSP && MBBI->mayLoadOrStore()))
return E;
}
@@ -2062,8 +2062,8 @@ MachineBasicBlock::iterator AArch64LoadStoreOpt::findMatchingUpdateInsnBackward(
// Otherwise, if the base register is used or modified, we have no match, so
// return early.
- if (!ModifiedRegUnits.available(BaseReg) ||
- !UsedRegUnits.available(BaseReg))
+ if (ModifiedRegUnits.contains(BaseReg) ||
+ UsedRegUnits.contains(BaseReg))
return E;
// Keep track if we have a memory access before an SP pre-increment, in this
// case we need to validate later that the update amount respects the red
diff --git a/llvm/lib/Target/AArch64/AArch64RedundantCopyElimination.cpp b/llvm/lib/Target/AArch64/AArch64RedundantCopyElimination.cpp
index 1494312886a40de..a4fc06cce5bd450 100644
--- a/llvm/lib/Target/AArch64/AArch64RedundantCopyElimination.cpp
+++ b/llvm/lib/Target/AArch64/AArch64RedundantCopyElimination.cpp
@@ -190,7 +190,7 @@ bool AArch64RedundantCopyElimination::knownRegValInBlock(
// register of the compare is not modified (including a self-clobbering
// compare) between the compare and conditional branch we known the value
// of the 1st source operand.
- if (PredI.getOperand(2).isImm() && DomBBClobberedRegs.available(SrcReg) &&
+ if (PredI.getOperand(2).isImm() && !DomBBClobberedRegs.contains(SrcReg) &&
SrcReg != DstReg) {
// We've found the instruction that sets NZCV.
int32_t KnownImm = PredI.getOperand(2).getImm();
@@ -210,7 +210,7 @@ bool AArch64RedundantCopyElimination::knownRegValInBlock(
// The destination register must not be modified between the NZCV setting
// instruction and the conditional branch.
- if (!DomBBClobberedRegs.available(DstReg))
+ if (DomBBClobberedRegs.contains(DstReg))
return Res;
FirstUse = PredI;
@@ -254,7 +254,7 @@ bool AArch64RedundantCopyElimination::knownRegValInBlock(
// The destination register of the NZCV setting instruction must not be
// modified before the conditional branch.
- if (!DomBBClobberedRegs.available(DstReg))
+ if (DomBBClobberedRegs.contains(DstReg))
return false;
// We've found the instruction that sets NZCV whose DstReg == 0.
@@ -323,12 +323,12 @@ bool AArch64RedundantCopyElimination::optimizeBlock(MachineBasicBlock *MBB) {
MCPhysReg CopyDstReg = PredI->getOperand(0).getReg();
MCPhysReg CopySrcReg = PredI->getOperand(1).getReg();
for (auto &KnownReg : KnownRegs) {
- if (!OptBBClobberedRegs.available(KnownReg.Reg))
+ if (OptBBClobberedRegs.contains(KnownReg.Reg))
continue;
// If we have X = COPY Y, and Y is known to be zero, then now X is
// known to be zero.
if (CopySrcReg == KnownReg.Reg &&
- OptBBClobberedRegs.available(CopyDstReg)) {
+ !OptBBClobberedRegs.contains(CopyDstReg)) {
KnownRegs.push_back(RegImm(CopyDstReg, KnownReg.Imm));
if (SeenFirstUse)
FirstUse = PredI;
@@ -337,7 +337,7 @@ bool AArch64RedundantCopyElimination::optimizeBlock(MachineBasicBlock *MBB) {
// If we have X = COPY Y, and X is known to be zero, then now Y is
// known to be zero.
if (CopyDstReg == KnownReg.Reg &&
- OptBBClobberedRegs.available(CopySrcReg)) {
+ !OptBBClobberedRegs.contains(CopySrcReg)) {
KnownRegs.push_back(RegImm(CopySrcReg, KnownReg.Imm));
if (SeenFirstUse)
FirstUse = PredI;
@@ -354,7 +354,7 @@ bool AArch64RedundantCopyElimination::optimizeBlock(MachineBasicBlock *MBB) {
OptBBUsedRegs, TRI);
// Stop if all of the known-zero regs have been clobbered.
if (all_of(KnownRegs, [&](RegImm KnownReg) {
- return !OptBBClobberedRegs.available(KnownReg.Reg);
+ return OptBBClobberedRegs.contains(KnownReg.Reg);
}))
break;
}
diff --git a/llvm/lib/Target/AArch64/GISel/AArch64PostSelectOptimize.cpp b/llvm/lib/Target/AArch64/GISel/AArch64PostSelectOptimize.cpp
index 94584e20f5ab3fb..ff3404b6584acc5 100644
--- a/llvm/lib/Target/AArch64/GISel/AArch64PostSelectOptimize.cpp
+++ b/llvm/lib/Target/AArch64/GISel/AArch64PostSelectOptimize.cpp
@@ -198,7 +198,7 @@ bool AArch64PostSelectOptimize::optimizeNZCVDefs(MachineBasicBlock &MBB) {
LRU.addLiveOuts(MBB);
for (auto &II : instructionsWithoutDebug(MBB.rbegin(), MBB.rend())) {
- bool NZCVDead = LRU.available(AArch64::NZCV);
+ bool NZCVDead = !LRU.contains(AArch64::NZCV);
if (NZCVDead && II.definesRegister(AArch64::NZCV)) {
// The instruction defines NZCV, but NZCV is dead.
unsigned NewOpc = getNonFlagSettingVariant(II.getOpcode());
diff --git a/llvm/lib/Target/ARM/ARMBaseInstrInfo.cpp b/llvm/lib/Target/ARM/ARMBaseInstrInfo.cpp
index 1ffdde0360cf623..f087ce71f8ba73c 100644
--- a/llvm/lib/Target/ARM/ARMBaseInstrInfo.cpp
+++ b/llvm/lib/Target/ARM/ARMBaseInstrInfo.cpp
@@ -6259,8 +6259,8 @@ bool ARMBaseInstrInfo::isMBBSafeToOutlineFrom(MachineBasicBlock ...
|
arsenm
reviewed
Sep 12, 2023
| /// Returns true if any part of physical register \p Reg is live | ||
| bool contains(MCPhysReg Reg) const { | ||
| return llvm::any_of(TRI->regunits(Reg), | ||
| [&](MCRegUnit Unit) { return Units.test(Unit); }); |
Contributor
There was a problem hiding this comment.
I think this just needs to capture this
| void addRegsInMask(const uint32_t *RegMask); | ||
|
|
||
| /// Returns true if no part of physical register \p Reg is live. | ||
| bool available(MCPhysReg Reg) const { |
Contributor
There was a problem hiding this comment.
I think available should stay as it is here (with rename)? A separate helper can combine the isReserved check
Contributor
Author
There was a problem hiding this comment.
The reason to keep this naming was to keep it analogous to the functions in LivePhysRegs (where "available" has the isReserved check and "contains" is also there) so that porting from LivePhysRegs to LiveRegUnits in other files can be smoother.
Would you have suggestions on a better name for these two functions (available and contains)?
Contributor
There was a problem hiding this comment.
Just keep the available implementation in the header, it's trivia
Collaborator
There was a problem hiding this comment.
+1 on keeping available
|
|
||
| bool LiveRegUnits::available(const MachineRegisterInfo &MRI, | ||
| MCPhysReg Reg) const { | ||
| return !MRI.isReserved(Reg) && !contains(Reg); |
Contributor
There was a problem hiding this comment.
swap the order of the checks
| CreateNewRangeStartingAt(MI.getIterator()); | ||
| continue; | ||
| } | ||
| LRAvailableEverywhere &= LRU.available(AArch64::LR); |
Contributor
There was a problem hiding this comment.
I think the positive is available check is more readable than having to negate most of the uses
arsenm
requested changes
Feb 6, 2024
Contributor
arsenm
left a comment
arsenm
left a comment
There was a problem hiding this comment.
Needs update to main and addressed comments
Contributor
|
Ping? |
Collaborator
|
This should be abandoned. Not sure why it is still open. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
8 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Currently there's only a provision to check whether a register is contained in the LiveRegUnits BitVector. Introducing the method which is available in LivePhysRegs to check whether the register is not reserved as well. The naming convention has been retained like in LivePhysRegs.
The motivation for this PR is [AMDGPU] [SIFrameLowering] Use LiveRegUnits instead of LivePhysRegs