[llvm] [InstCombine] Remove ProfcheckDisableMetadataFixes checks (PR #222409)
Aiden Grossman via llvm-commits
llvm-commits at lists.llvm.org
Wed Sep 9 12:54:16 PDT 2026
https://github.com/boomanaiden154 updated https://github.com/llvm/llvm-project/pull/222409
>From ce111e5e4cd10040d9bac2c94533ecbfe7735910 Mon Sep 17 00:00:00 2001
From: Aiden Grossman <aidengrossman at google.com>
Date: Wed, 9 Sep 2026 17:50:28 +0000
Subject: [PATCH] =?UTF-8?q?[=F0=9D=98=80=F0=9D=97=BD=F0=9D=97=BF]=20change?=
=?UTF-8?q?s=20to=20main=20this=20commit=20is=20based=20on?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
Created using spr 1.3.7
[skip ci]
---
.../AggressiveInstCombine.cpp | 24 +++----
.../Transforms/IPO/FunctionSpecialization.cpp | 6 +-
.../lib/Transforms/IPO/WholeProgramDevirt.cpp | 4 +-
.../Transforms/Scalar/DFAJumpThreading.cpp | 12 ++--
.../Transforms/Scalar/JumpTableToSwitch.cpp | 7 +--
llvm/lib/Transforms/Scalar/LICM.cpp | 7 +--
.../Transforms/Scalar/LoopIdiomRecognize.cpp | 4 --
llvm/lib/Transforms/Scalar/MergeICmps.cpp | 5 --
.../Scalar/PartiallyInlineLibCalls.cpp | 7 +--
llvm/lib/Transforms/Scalar/SROA.cpp | 7 +--
.../Transforms/Scalar/SimpleLoopUnswitch.cpp | 17 +++--
llvm/lib/Transforms/Utils/LoopPeel.cpp | 4 +-
llvm/lib/Transforms/Utils/LoopUtils.cpp | 5 +-
.../Transforms/Utils/LowerMemIntrinsics.cpp | 6 --
llvm/lib/Transforms/Utils/SimplifyCFG.cpp | 62 ++++++++-----------
.../peel-last-iteration.ll | 61 +++++++-----------
16 files changed, 77 insertions(+), 161 deletions(-)
diff --git a/llvm/lib/Transforms/AggressiveInstCombine/AggressiveInstCombine.cpp b/llvm/lib/Transforms/AggressiveInstCombine/AggressiveInstCombine.cpp
index f72ff61f028db..9b0ed2bd0c54d 100644
--- a/llvm/lib/Transforms/AggressiveInstCombine/AggressiveInstCombine.cpp
+++ b/llvm/lib/Transforms/AggressiveInstCombine/AggressiveInstCombine.cpp
@@ -43,10 +43,6 @@ using namespace PatternMatch;
#define DEBUG_TYPE "aggressive-instcombine"
-namespace llvm {
-extern cl::opt<bool> ProfcheckDisableMetadataFixes;
-}
-
STATISTIC(NumAnyOrAllBitsSet, "Number of any/all-bits-set patterns folded");
STATISTIC(NumGuardedRotates,
"Number of guarded rotates transformed into funnel shifts");
@@ -1005,12 +1001,10 @@ static bool tryToRecognizeTableBasedCttz(LoadInst *LI, Type *AccessType,
Res = B.CreateSelect(Cmp, ZeroTableElem, Res);
// The true branch of select handles the cttz(0) case, which is rare.
- if (!ProfcheckDisableMetadataFixes) {
- if (Instruction *SelectI = dyn_cast<Instruction>(Res))
- SelectI->setMetadata(
- LLVMContext::MD_prof,
- MDBuilder(SelectI->getContext()).createUnlikelyBranchWeights());
- }
+ if (Instruction *SelectI = dyn_cast<Instruction>(Res))
+ SelectI->setMetadata(
+ LLVMContext::MD_prof,
+ MDBuilder(SelectI->getContext()).createUnlikelyBranchWeights());
// NOTE: If the table[0] is 0, but the cttz(0) is defined by the Target
// it should be handled as: `cttz(x) & (typeSize - 1)`.
@@ -1222,12 +1216,10 @@ static bool tryToRecognizeTableBasedLog2(LoadInst *LI, Type *AccessType,
B.CreateSelect(Cmp, B.CreateZExt(ZeroTableElem, XType), Sub);
// The true branch of select handles the log2(0) case, which is rare.
- if (!ProfcheckDisableMetadataFixes) {
- if (Instruction *SelectI = dyn_cast<Instruction>(Select))
- SelectI->setMetadata(
- LLVMContext::MD_prof,
- MDBuilder(SelectI->getContext()).createUnlikelyBranchWeights());
- }
+ if (Instruction *SelectI = dyn_cast<Instruction>(Select))
+ SelectI->setMetadata(
+ LLVMContext::MD_prof,
+ MDBuilder(SelectI->getContext()).createUnlikelyBranchWeights());
Result = Select;
}
diff --git a/llvm/lib/Transforms/IPO/FunctionSpecialization.cpp b/llvm/lib/Transforms/IPO/FunctionSpecialization.cpp
index e45c098c34ce8..38c31c4cc2499 100644
--- a/llvm/lib/Transforms/IPO/FunctionSpecialization.cpp
+++ b/llvm/lib/Transforms/IPO/FunctionSpecialization.cpp
@@ -91,8 +91,6 @@ static cl::opt<bool> SpecializeLiteralConstant(
"Enable specialization of functions that take a literal constant as an "
"argument"));
-extern cl::opt<bool> ProfcheckDisableMetadataFixes;
-
} // end namespace llvm
bool InstCostVisitor::canEliminateSuccessor(BasicBlock *BB,
@@ -793,7 +791,7 @@ bool FunctionSpecializer::run() {
auto &BFI = GetBFI(*Call->getFunction());
std::optional<uint64_t> Count =
BFI.getBlockProfileCount(Call->getParent());
- if (Count && !ProfcheckDisableMetadataFixes) {
+ if (Count) {
std::optional<uint64_t> MaybeCloneCount = Clone->getEntryCount();
if (MaybeCloneCount) {
uint64_t CallCount = *Count + *MaybeCloneCount;
@@ -1073,7 +1071,7 @@ Function *FunctionSpecializer::createSpecialization(Function *F,
// clone must.
Clone->setLinkage(GlobalValue::InternalLinkage);
- if (F->getEntryCount() && !ProfcheckDisableMetadataFixes)
+ if (F->getEntryCount())
Clone->setEntryCount(0);
// Initialize the lattice state of the arguments of the function clone,
diff --git a/llvm/lib/Transforms/IPO/WholeProgramDevirt.cpp b/llvm/lib/Transforms/IPO/WholeProgramDevirt.cpp
index 4098fa8c60c7c..e9d2036e5fdbb 100644
--- a/llvm/lib/Transforms/IPO/WholeProgramDevirt.cpp
+++ b/llvm/lib/Transforms/IPO/WholeProgramDevirt.cpp
@@ -196,8 +196,6 @@ static cl::list<std::string>
cl::desc("Prevent function(s) from being devirtualized"),
cl::Hidden, cl::CommaSeparated);
-extern cl::opt<bool> ProfcheckDisableMetadataFixes;
-
} // end namespace llvm
/// With Clang, a pure virtual class's deleting destructor is emitted as a
@@ -1577,7 +1575,7 @@ void DevirtModule::applyICallBranchFunnel(VTableSlotInfo &SlotInfo,
llvm::append_range(Args, CB.args());
CallBase *NewCS = nullptr;
- if (!JT.isDeclaration() && !ProfcheckDisableMetadataFixes) {
+ if (!JT.isDeclaration()) {
// Accumulate the call frequencies of the original call site, and use
// that as total entry count for the funnel function.
auto &F = *CB.getCaller();
diff --git a/llvm/lib/Transforms/Scalar/DFAJumpThreading.cpp b/llvm/lib/Transforms/Scalar/DFAJumpThreading.cpp
index 585c6e237be84..1949d52302e1e 100644
--- a/llvm/lib/Transforms/Scalar/DFAJumpThreading.cpp
+++ b/llvm/lib/Transforms/Scalar/DFAJumpThreading.cpp
@@ -136,8 +136,6 @@ static cl::opt<unsigned>
"accepted for the transformation"),
cl::Hidden, cl::init(40));
-extern cl::opt<bool> ProfcheckDisableMetadataFixes;
-
} // namespace llvm
namespace {
@@ -277,9 +275,8 @@ void DFAJumpThreading::unfold(DomTreeUpdater *DTU, LoopInfo *LI,
auto *BI =
CondBrInst::Create(SI->getCondition(), EndBlock, NewBlock, StartBlock);
BI->setDebugLoc(SelectBranchLoc);
- if (!ProfcheckDisableMetadataFixes)
- BI->setMetadata(LLVMContext::MD_prof,
- SI->getMetadata(LLVMContext::MD_prof));
+ BI->setMetadata(LLVMContext::MD_prof,
+ SI->getMetadata(LLVMContext::MD_prof));
DTU->applyUpdates({{DominatorTree::Insert, StartBlock, NewBlock}});
} else {
BasicBlock *EndBlock = SIUse->getParent();
@@ -320,9 +317,8 @@ void DFAJumpThreading::unfold(DomTreeUpdater *DTU, LoopInfo *LI,
DebugLoc SelectLoc = SI->getDebugLoc();
NewFToEnd->setDebugLoc(SelectLoc);
BI->setDebugLoc(SelectLoc);
- if (!ProfcheckDisableMetadataFixes)
- BI->setMetadata(LLVMContext::MD_prof,
- SI->getMetadata(LLVMContext::MD_prof));
+ BI->setMetadata(LLVMContext::MD_prof,
+ SI->getMetadata(LLVMContext::MD_prof));
DTU->applyUpdates({{DominatorTree::Insert, NewBlockT, NewBlockF},
{DominatorTree::Insert, NewBlockT, EndBlock},
{DominatorTree::Insert, NewBlockF, EndBlock}});
diff --git a/llvm/lib/Transforms/Scalar/JumpTableToSwitch.cpp b/llvm/lib/Transforms/Scalar/JumpTableToSwitch.cpp
index 5c8af03e8f2b0..a9247e5107b5b 100644
--- a/llvm/lib/Transforms/Scalar/JumpTableToSwitch.cpp
+++ b/llvm/lib/Transforms/Scalar/JumpTableToSwitch.cpp
@@ -39,10 +39,6 @@ static cl::opt<unsigned> FunctionSizeThreshold(
"or equal than this threshold."),
cl::init(50));
-namespace llvm {
-extern cl::opt<bool> ProfcheckDisableMetadataFixes;
-} // end namespace llvm
-
#define DEBUG_TYPE "jump-table-to-switch"
STATISTIC(NumEligibleJumpTables, "The number of jump tables seen by the pass "
@@ -193,8 +189,7 @@ expandToSwitch(CallBase *CB, const JumpTableTy &JT, DomTreeUpdater &DTU,
// Only set branch weights on the switch if we have non-zero branch weights.
// We can have no non-zero branch weights while having VP metadata if for
// example, all of the functions are external and not instrumented.
- if (HadProfile && !ProfcheckDisableMetadataFixes &&
- llvm::any_of(BranchWeights, not_equal_to(0))) {
+ if (HadProfile && llvm::any_of(BranchWeights, not_equal_to(0))) {
setBranchWeights(*Switch, downscaleWeights(BranchWeights),
/*IsExpected=*/false);
} else
diff --git a/llvm/lib/Transforms/Scalar/LICM.cpp b/llvm/lib/Transforms/Scalar/LICM.cpp
index 37adb72d06828..8a5210e75d41a 100644
--- a/llvm/lib/Transforms/Scalar/LICM.cpp
+++ b/llvm/lib/Transforms/Scalar/LICM.cpp
@@ -169,10 +169,6 @@ cl::opt<unsigned> llvm::SetLicmMssaNoAccForPromotionCap(
"number of accesses allowed to be present in a loop in order to "
"enable memory promotion."));
-namespace llvm {
-extern cl::opt<bool> ProfcheckDisableMetadataFixes;
-} // end namespace llvm
-
static bool inSubLoop(BasicBlock *BB, Loop *CurLoop, LoopInfo *LI);
static bool isNotUsedOrFoldableInLoop(const Instruction &I, const Loop *CurLoop,
const LoopSafetyInfo *SafetyInfo,
@@ -872,8 +868,7 @@ class ControlFlowHoister {
HoistTarget->getTerminator()->eraseFromParent();
// md_prof should also come from the original branch - since the
// condition was hoisted, the branch probabilities shouldn't change.
- if (!ProfcheckDisableMetadataFixes)
- NewBI->copyMetadata(*BI, {LLVMContext::MD_prof});
+ NewBI->copyMetadata(*BI, {LLVMContext::MD_prof});
// FIXME: Issue #152767: debug info should also be the same as the
// original branch, **if** the user explicitly indicated that.
NewBI->setDebugLoc(HoistTarget->getTerminator()->getDebugLoc());
diff --git a/llvm/lib/Transforms/Scalar/LoopIdiomRecognize.cpp b/llvm/lib/Transforms/Scalar/LoopIdiomRecognize.cpp
index dbe114f7c220a..acecc8f746839 100644
--- a/llvm/lib/Transforms/Scalar/LoopIdiomRecognize.cpp
+++ b/llvm/lib/Transforms/Scalar/LoopIdiomRecognize.cpp
@@ -186,8 +186,6 @@ static cl::opt<CRCStrategyKind> CRCStrategy(
clEnumValN(CRCStrategyKind::Clmul, "clmul",
"Use carry-less multiplication when possible")));
-extern cl::opt<bool> ProfcheckDisableMetadataFixes;
-
} // namespace llvm
namespace {
@@ -3500,7 +3498,6 @@ bool LoopIdiomRecognize::recognizeShiftUntilBitTest() {
CurLoop->getName() + ".ivcheck");
SmallVector<uint32_t> BranchWeights;
const bool HasBranchWeights =
- !ProfcheckDisableMetadataFixes &&
extractBranchWeights(*LoopHeaderBB->getTerminator(), BranchWeights);
auto *BI = Builder.CreateCondBr(IVCheck, SuccessorBB, LoopHeaderBB);
@@ -3848,7 +3845,6 @@ bool LoopIdiomRecognize::recognizeShiftUntilZero() {
Builder.SetInsertPoint(LoopHeaderBB->getTerminator());
SmallVector<uint32_t> BranchWeights;
const bool HasBranchWeights =
- !ProfcheckDisableMetadataFixes &&
extractBranchWeights(*LoopHeaderBB->getTerminator(), BranchWeights);
auto *BI = Builder.CreateCondBr(CIVCheck, SuccessorBB, LoopHeaderBB);
diff --git a/llvm/lib/Transforms/Scalar/MergeICmps.cpp b/llvm/lib/Transforms/Scalar/MergeICmps.cpp
index 9b4285f03b33a..6ac3045c6e12f 100644
--- a/llvm/lib/Transforms/Scalar/MergeICmps.cpp
+++ b/llvm/lib/Transforms/Scalar/MergeICmps.cpp
@@ -64,9 +64,6 @@ using namespace llvm;
#define DEBUG_TYPE "mergeicmps"
-namespace llvm {
-extern cl::opt<bool> ProfcheckDisableMetadataFixes;
-} // namespace llvm
namespace {
// A BCE atom "Binary Compare Expression Atom" represents an integer load
@@ -625,8 +622,6 @@ class MergedBlockName {
static std::optional<SmallVector<uint32_t, 2>>
computeMergedBranchWeights(ArrayRef<BCECmpBlock> Comparisons) {
assert(!Comparisons.empty());
- if (ProfcheckDisableMetadataFixes)
- return std::nullopt;
if (Comparisons.size() == 1) {
SmallVector<uint32_t, 2> Weights;
if (!extractBranchWeights(*Comparisons[0].BB->getTerminator(), Weights))
diff --git a/llvm/lib/Transforms/Scalar/PartiallyInlineLibCalls.cpp b/llvm/lib/Transforms/Scalar/PartiallyInlineLibCalls.cpp
index ec4051a7c43e4..6a1b992c42d73 100644
--- a/llvm/lib/Transforms/Scalar/PartiallyInlineLibCalls.cpp
+++ b/llvm/lib/Transforms/Scalar/PartiallyInlineLibCalls.cpp
@@ -28,10 +28,6 @@
using namespace llvm;
-namespace llvm {
-extern cl::opt<bool> ProfcheckDisableMetadataFixes;
-} // namespace llvm
-
#define DEBUG_TYPE "partially-inline-libcalls"
DEBUG_COUNTER(PILCounter, "partially-inline-libcalls-transform",
@@ -99,8 +95,7 @@ static bool optimizeSQRT(CallInst *Call, Function *CalledFunc,
: Builder.CreateFCmpOGE(Call->getOperand(0),
ConstantFP::get(Ty, 0.0));
CurrBBTerm->setCondition(FCmp);
- if (!ProfcheckDisableMetadataFixes &&
- CurrBBTerm->getFunction()->getEntryCount()) {
+ if (CurrBBTerm->getFunction()->getEntryCount()) {
// Presume the quick path - where we don't call the library call - is the
// frequent one
MDBuilder MDB(CurrBBTerm->getContext());
diff --git a/llvm/lib/Transforms/Scalar/SROA.cpp b/llvm/lib/Transforms/Scalar/SROA.cpp
index a0d7a0c921796..c933225c2ad21 100644
--- a/llvm/lib/Transforms/Scalar/SROA.cpp
+++ b/llvm/lib/Transforms/Scalar/SROA.cpp
@@ -121,7 +121,6 @@ namespace llvm {
/// Disable running mem2reg during SROA in order to test or debug SROA.
static cl::opt<bool> SROASkipMem2Reg("sroa-skip-mem2reg", cl::init(false),
cl::Hidden);
-extern cl::opt<bool> ProfcheckDisableMetadataFixes;
} // namespace llvm
namespace {
@@ -1820,8 +1819,7 @@ static void speculateSelectInstLoads(SelectInst &SI, LoadInst &LI,
}
Value *V = IRB.CreateSelect(SI.getCondition(), TL, FL,
- LI.getName() + ".sroa.speculated",
- ProfcheckDisableMetadataFixes ? nullptr : &SI);
+ LI.getName() + ".sroa.speculated", &SI);
LLVM_DEBUG(dbgs() << " speculated to: " << *V << "\n");
LI.replaceAllUsesWith(V);
@@ -4522,8 +4520,7 @@ class AggLoadStoreRewriter : public InstVisitor<AggLoadStoreRewriter, bool> {
Cond = SI->getCondition();
True = SI->getTrueValue();
False = SI->getFalseValue();
- if (!ProfcheckDisableMetadataFixes)
- MDFrom = SI;
+ MDFrom = SI;
} else {
Cond = Sel->getOperand(0);
True = ConstantInt::get(Sel->getType(), 1);
diff --git a/llvm/lib/Transforms/Scalar/SimpleLoopUnswitch.cpp b/llvm/lib/Transforms/Scalar/SimpleLoopUnswitch.cpp
index 7f6a08454b749..abe8d19b1a583 100644
--- a/llvm/lib/Transforms/Scalar/SimpleLoopUnswitch.cpp
+++ b/llvm/lib/Transforms/Scalar/SimpleLoopUnswitch.cpp
@@ -141,7 +141,6 @@ static cl::opt<unsigned> InjectInvariantConditionHotnesThreshold(
static cl::opt<bool> EstimateProfile("simple-loop-unswitch-estimate-profile",
cl::Hidden, cl::init(true));
-extern cl::opt<bool> ProfcheckDisableMetadataFixes;
} // namespace llvm
AnalysisKey ShouldRunExtraSimpleLoopUnswitch::Key;
@@ -293,8 +292,8 @@ static void buildPartialUnswitchConditionalBranch(
const CondBrInst &ComputeProfFrom) {
SmallVector<uint32_t> BranchWeights;
- bool HasBranchWeights = EstimateProfile && !ProfcheckDisableMetadataFixes &&
- extractBranchWeights(ComputeProfFrom, BranchWeights);
+ bool HasBranchWeights =
+ EstimateProfile && extractBranchWeights(ComputeProfFrom, BranchWeights);
// If Direction is true, that means we had a disjunction and that the "true"
// case exits. The probability of the disjunction of the subset of terms is at
// most as high as the original one. So, if the probability is higher than the
@@ -380,8 +379,7 @@ static void buildPartialInvariantUnswitchConditionalBranch(
// The expectation is that ToDuplicate[0] is the condition used by the
// OriginalBranch, case in which we can clone the profile metadata from there.
auto *ProfData =
- !ProfcheckDisableMetadataFixes &&
- ToDuplicate[0] == skipTrivialSelect(OriginalBranch.getCondition())
+ ToDuplicate[0] == skipTrivialSelect(OriginalBranch.getCondition())
? OriginalBranch.getMetadata(LLVMContext::MD_prof)
: nullptr;
auto *BR =
@@ -2580,7 +2578,7 @@ static CondBrInst *turnGuardIntoBranch(IntrinsicInst *GI, Loop &L,
// however, that the deopt path is unlikely.
Instruction *DeoptBlockTerm = SplitBlockAndInsertIfThen(
GI->getArgOperand(0), GI, true,
- !ProfcheckDisableMetadataFixes && EstimateProfile
+ EstimateProfile
? MDBuilder(GI->getContext()).createUnlikelyBranchWeights()
: nullptr,
&DTU, &LI);
@@ -2950,10 +2948,9 @@ injectPendingInvariantConditions(NonTrivialUnswitchCandidate Candidate, Loop &L,
setExplicitlyUnknownBranchWeightsIfProfiled(*InvariantBr, DEBUG_TYPE);
Builder.SetInsertPoint(CheckBlock);
- Builder.CreateCondBr(
- TI->getCondition(), TI->getSuccessor(0), TI->getSuccessor(1),
- !ProfcheckDisableMetadataFixes ? TI->getMetadata(LLVMContext::MD_prof)
- : nullptr);
+ Builder.CreateCondBr(TI->getCondition(), TI->getSuccessor(0),
+ TI->getSuccessor(1),
+ TI->getMetadata(LLVMContext::MD_prof));
TI->eraseFromParent();
// Fixup phis.
diff --git a/llvm/lib/Transforms/Utils/LoopPeel.cpp b/llvm/lib/Transforms/Utils/LoopPeel.cpp
index cc45a3e09bd88..315763d442786 100644
--- a/llvm/lib/Transforms/Utils/LoopPeel.cpp
+++ b/llvm/lib/Transforms/Utils/LoopPeel.cpp
@@ -90,7 +90,6 @@ static cl::opt<bool> EnablePeelingForIV(
static const char *PeeledCountMetaData = "llvm.loop.peeled.count";
-extern cl::opt<bool> ProfcheckDisableMetadataFixes;
} // namespace llvm
// Check whether we are capable of peeling this loop.
@@ -1231,8 +1230,7 @@ void llvm::peelLoop(Loop *L, unsigned PeelCount, bool PeelLast, LoopInfo *LI,
auto *BI = B.CreateCondBr(Cond, NewPreHeader, InsertTop);
SmallVector<uint32_t> Weights;
auto *OrigLatchBr = Latch->getTerminator();
- auto HasBranchWeights = !ProfcheckDisableMetadataFixes &&
- extractBranchWeights(*OrigLatchBr, Weights);
+ auto HasBranchWeights = extractBranchWeights(*OrigLatchBr, Weights);
if (HasBranchWeights) {
// The probability that the new guard skips the loop to execute just one
// iteration is the original loop's probability of exiting at the latch
diff --git a/llvm/lib/Transforms/Utils/LoopUtils.cpp b/llvm/lib/Transforms/Utils/LoopUtils.cpp
index d3f2f0beacc6a..a2e544801b9c4 100644
--- a/llvm/lib/Transforms/Utils/LoopUtils.cpp
+++ b/llvm/lib/Transforms/Utils/LoopUtils.cpp
@@ -54,9 +54,6 @@ using namespace llvm::PatternMatch;
static const char *LLVMLoopDisableNonforced = "llvm.loop.disable_nonforced";
static const char *LLVMLoopDisableLICM = "llvm.licm.disable";
-namespace llvm {
-extern cl::opt<bool> ProfcheckDisableMetadataFixes;
-} // namespace llvm
bool llvm::formDedicatedExitBlocks(Loop *L, DominatorTree *DT, LoopInfo *LI,
MemorySSAUpdater *MSSAU,
@@ -994,7 +991,7 @@ bool llvm::setLoopEstimatedTripCount(
return true;
// Calculate taken and exit weights.
- unsigned LatchExitWeight = ProfcheckDisableMetadataFixes ? 0 : 1;
+ unsigned LatchExitWeight = 1;
unsigned BackedgeTakenWeight = 0;
if (EstimatedTripCount != 0) {
diff --git a/llvm/lib/Transforms/Utils/LowerMemIntrinsics.cpp b/llvm/lib/Transforms/Utils/LowerMemIntrinsics.cpp
index c48e173b05479..84131f6592e50 100644
--- a/llvm/lib/Transforms/Utils/LowerMemIntrinsics.cpp
+++ b/llvm/lib/Transforms/Utils/LowerMemIntrinsics.cpp
@@ -26,10 +26,6 @@
using namespace llvm;
-namespace llvm {
-extern cl::opt<bool> ProfcheckDisableMetadataFixes;
-}
-
/// \returns \p Len urem \p OpSize, checking for optimization opportunities.
/// \p OpSizeVal must be the integer value of the \c ConstantInt \p OpSize.
static Value *getRuntimeLoopRemainder(IRBuilderBase &B, Value *Len,
@@ -71,8 +67,6 @@ struct LoopExpansionInfo {
};
std::optional<uint64_t> getAverageMemOpLoopTripCount(const MemIntrinsic &I) {
- if (ProfcheckDisableMetadataFixes)
- return std::nullopt;
if (std::optional<uint64_t> EC = I.getFunction()->getEntryCount();
!EC || *EC == 0)
return std::nullopt;
diff --git a/llvm/lib/Transforms/Utils/SimplifyCFG.cpp b/llvm/lib/Transforms/Utils/SimplifyCFG.cpp
index ca96f2e70d810..f58fc8166ffd6 100644
--- a/llvm/lib/Transforms/Utils/SimplifyCFG.cpp
+++ b/llvm/lib/Transforms/Utils/SimplifyCFG.cpp
@@ -4206,13 +4206,12 @@ static bool performBranchToCommonDestFolding(CondBrInst *BI, CondBrInst *PBI,
Value *BICond = VMap[BI->getCondition()];
PBI->setCondition(
createLogicalOp(Builder, Opc, PBI->getCondition(), BICond, "or.cond"));
- if (!ProfcheckDisableMetadataFixes)
- if (auto *SI = dyn_cast<SelectInst>(PBI->getCondition()))
- if (!MDWeights.empty()) {
- assert(isSelectInRoleOfConjunctionOrDisjunction(SI));
- setFittedBranchWeights(*SI, {MDWeights[0], MDWeights[1]},
- /*IsExpected=*/false, /*ElideAllZero=*/true);
- }
+ if (auto *SI = dyn_cast<SelectInst>(PBI->getCondition()))
+ if (!MDWeights.empty()) {
+ assert(isSelectInRoleOfConjunctionOrDisjunction(SI));
+ setFittedBranchWeights(*SI, {MDWeights[0], MDWeights[1]},
+ /*IsExpected=*/false, /*ElideAllZero=*/true);
+ }
++NumFoldBranchToCommonDest;
return true;
@@ -4558,8 +4557,7 @@ static bool mergeConditionalStoreToAddress(
auto *T = SplitBlockAndInsertIfThen(CombinedPred, InsertPt,
/*Unreachable=*/false,
/*BranchWeights=*/nullptr, DTU);
- if (hasBranchWeightMD(*PBranch) && hasBranchWeightMD(*QBranch) &&
- !ProfcheckDisableMetadataFixes) {
+ if (hasBranchWeightMD(*PBranch) && hasBranchWeightMD(*QBranch)) {
SmallVector<uint32_t, 2> PWeights, QWeights;
extractBranchWeights(*PBranch, PWeights);
extractBranchWeights(*QBranch, QWeights);
@@ -4945,16 +4943,15 @@ static bool SimplifyCondBranchToCondBranch(CondBrInst *PBI, CondBrInst *BI,
/*ElideAllZero=*/true);
// Cond may be a select instruction with the first operand set to "true", or
// the second to "false" (see how createLogicalOp works for `and` and `or`)
- if (!ProfcheckDisableMetadataFixes)
- if (auto *SI = dyn_cast<SelectInst>(Cond)) {
- assert(isSelectInRoleOfConjunctionOrDisjunction(SI));
- // The select is predicated on PBICond
- assert(SI->getCondition() == PBICond);
- // The corresponding probabilities are what was referred to above as
- // PredCommon and PredOther.
- setFittedBranchWeights(*SI, {PredCommon, PredOther},
- /*IsExpected=*/false, /*ElideAllZero=*/true);
- }
+ if (auto *SI = dyn_cast<SelectInst>(Cond)) {
+ assert(isSelectInRoleOfConjunctionOrDisjunction(SI));
+ // The select is predicated on PBICond
+ assert(SI->getCondition() == PBICond);
+ // The corresponding probabilities are what was referred to above as
+ // PredCommon and PredOther.
+ setFittedBranchWeights(*SI, {PredCommon, PredOther},
+ /*IsExpected=*/false, /*ElideAllZero=*/true);
+ }
}
// OtherDest may have phi nodes. If so, add an entry from PBI's
@@ -5197,8 +5194,7 @@ bool SimplifyCFGOpt::simplifyIndirectBrOnSelect(IndirectBrInst *IBI,
// The select's profile becomes the profile of the conditional branch that
// replaces the indirect branch.
SmallVector<uint32_t> SelectBranchWeights(2);
- if (!ProfcheckDisableMetadataFixes)
- extractBranchWeights(*SI, SelectBranchWeights);
+ extractBranchWeights(*SI, SelectBranchWeights);
// Perform the actual simplification.
return simplifyTerminatorOnSelect(IBI, SI->getCondition(), TrueBB, FalseBB,
SelectBranchWeights[0],
@@ -5452,8 +5448,7 @@ bool SimplifyCFGOpt::simplifyBranchOnICmpChain(CondBrInst *BI,
return false;
SmallVector<uint32_t> BranchWeights;
- const bool HasProfile = !ProfcheckDisableMetadataFixes &&
- extractBranchWeights(*BI, BranchWeights);
+ const bool HasProfile = extractBranchWeights(*BI, BranchWeights);
// Figure out which block is which destination.
BasicBlock *DefaultBB = BI->getSuccessor(1);
@@ -6743,8 +6738,7 @@ static Value *foldSwitchToSelect(const SwitchCaseResultVectorTy &ResultVector,
// default: return 4; %3 = select i1 %2, i32 2, i32 %1
// }
- const bool HasBranchWeights =
- !BranchWeights.empty() && !ProfcheckDisableMetadataFixes;
+ const bool HasBranchWeights = !BranchWeights.empty();
if (ResultVector.size() == 2 && ResultVector[0].second.size() == 1 &&
ResultVector[1].second.size() == 1) {
@@ -6939,11 +6933,9 @@ static bool trySwitchToSelect(SwitchInst *SI, IRBuilder<> &Builder,
assert(PHI != nullptr && "PHI for value select not found");
Builder.SetInsertPoint(SI);
SmallVector<uint32_t, 4> BranchWeights;
- if (!ProfcheckDisableMetadataFixes) {
- [[maybe_unused]] auto HasWeights =
- extractBranchWeights(getBranchWeightMDNode(*SI), BranchWeights);
- assert(!HasWeights == (BranchWeights.empty()));
- }
+ [[maybe_unused]] auto HasWeights =
+ extractBranchWeights(getBranchWeightMDNode(*SI), BranchWeights);
+ assert(!HasWeights == (BranchWeights.empty()));
assert(BranchWeights.empty() ||
(BranchWeights.size() >=
UniqueResults.size() + (DefaultResult != nullptr)));
@@ -7820,8 +7812,8 @@ static bool simplifySwitchLookup(SwitchInst *SI, IRBuilder<> &Builder,
Updates.push_back({DominatorTree::Insert, LookupBB, CommonDest});
SmallVector<uint32_t> BranchWeights;
- const bool HasBranchWeights = CondBranch && !ProfcheckDisableMetadataFixes &&
- extractBranchWeights(*SI, BranchWeights);
+ const bool HasBranchWeights =
+ CondBranch && extractBranchWeights(*SI, BranchWeights);
uint64_t ToLookupWeight = 0;
uint64_t ToDefaultWeight = 0;
@@ -8121,8 +8113,7 @@ static bool simplifySwitchOfPowersOfTwo(SwitchInst *SI, IRBuilder<> &Builder,
BasicBlock *SplitBB = SplitBlock(OrigBB, SI, DTU);
auto It = OrigBB->getTerminator()->getIterator();
SmallVector<uint32_t> Weights;
- auto HasWeights =
- !ProfcheckDisableMetadataFixes && extractBranchWeights(*SI, Weights);
+ auto HasWeights = extractBranchWeights(*SI, Weights);
auto *BI = CondBrInst::Create(IsPow2, SplitBB, DefaultCaseBB, It);
if (HasWeights && any_of(Weights, not_equal_to(0))) {
// IsPow2 covers a subset of the cases in which we'd go to the default
@@ -8605,8 +8596,7 @@ bool SimplifyCFGOpt::simplifyIndirectBr(IndirectBrInst *IBI) {
BasicBlock *BB = IBI->getParent();
bool Changed = false;
SmallVector<uint32_t> BranchWeights;
- const bool HasBranchWeights = !ProfcheckDisableMetadataFixes &&
- extractBranchWeights(*IBI, BranchWeights);
+ const bool HasBranchWeights = extractBranchWeights(*IBI, BranchWeights);
DenseMap<const BasicBlock *, uint64_t> TargetWeight;
if (HasBranchWeights)
diff --git a/llvm/test/Transforms/LoopUnroll/branch-weights-freq/peel-last-iteration.ll b/llvm/test/Transforms/LoopUnroll/branch-weights-freq/peel-last-iteration.ll
index 43e2cd8dcd89c..8c9f7ce3be403 100644
--- a/llvm/test/Transforms/LoopUnroll/branch-weights-freq/peel-last-iteration.ll
+++ b/llvm/test/Transforms/LoopUnroll/branch-weights-freq/peel-last-iteration.ll
@@ -1,7 +1,4 @@
-; Disable this test in profcheck because the first run would cause profcheck to fail.
-; REQUIRES: !profcheck
-; RUN: opt -p "print<block-freq>,loop-unroll,print<block-freq>" -scev-cheap-expansion-budget=3 -S %s -profcheck-disable-metadata-fixes 2>&1 | FileCheck %s --check-prefixes=COMMON,BAD
-; RUN: opt -p "print<block-freq>,loop-unroll,print<block-freq>" -scev-cheap-expansion-budget=3 -S %s 2>&1 | FileCheck %s --check-prefixes=COMMON,GOOD
+; RUN: opt -p "print<block-freq>,loop-unroll,print<block-freq>" -scev-cheap-expansion-budget=3 -S %s 2>&1 | FileCheck %s
define i32 @test_expansion_cost_2(i32 %start, i32 %end) !prof !0 {
entry:
@@ -29,38 +26,24 @@ exit:
!1 = !{!"branch_weights", i32 2, i32 3}
!2 = !{!"branch_weights", i32 1, i32 50}
-; COMMON: block-frequency-info: test_expansion_cost_2
-; COMMON-NEXT: entry: float = 1.0
-; COMMON-NEXT: loop.header: float = 51.0
-; COMMON-NEXT: then: float = 20.4
-; COMMON-NEXT: loop.latch: float = 51.0
-; COMMON-NEXT: exit: float = 1.0
-
-; COMMON: block-frequency-info: test_expansion_cost_2
-; GOOD-NEXT: entry: float = 1.0
-; GOOD-NEXT: entry.split: float = 0.98039
-; GOOD-NEXT: loop.header: float = 50.0
-; GOOD-NEXT: then: float = 20.0
-; GOOD-NEXT: loop.latch: float = 50.0
-; GOOD-NEXT: exit.peel.begin.loopexit: float = 0.98039
-; GOOD-NEXT: exit.peel.begin: float = 1.0
-; GOOD-NEXT: loop.header.peel: float = 1.0
-; GOOD-NEXT: then.peel: float = 0.4
-; GOOD-NEXT: loop.latch.peel: float = 1.0
-; GOOD-NEXT: exit.peel.next: float = 1.0
-; GOOD-NEXT: loop.header.peel.next: float = 1.0
-; GOOD-NEXT: exit: float = 1.0
-
-; BAD-NEXT: entry: float = 1.0
-; BAD-NEXT: entry.split: float = 0.625
-; BAD-NEXT: loop.header: float = 31.875
-; BAD-NEXT: then: float = 12.75
-; BAD-NEXT: loop.latch: float = 31.875
-; BAD-NEXT: exit.peel.begin.loopexit: float = 0.625
-; BAD-NEXT: exit.peel.begin: float = 1.0
-; BAD-NEXT: loop.header.peel: float = 1.0
-; BAD-NEXT: then.peel: float = 0.4
-; BAD-NEXT: loop.latch.peel: float = 1.0
-; BAD-NEXT: exit.peel.next: float = 1.0
-; BAD-NEXT: loop.header.peel.next: float = 1.0
-; BAD-NEXT: exit: float = 1.0
\ No newline at end of file
+; CHECK: block-frequency-info: test_expansion_cost_2
+; CHECK-NEXT: entry: float = 1.0
+; CHECK-NEXT: loop.header: float = 51.0
+; CHECK-NEXT: then: float = 20.4
+; CHECK-NEXT: loop.latch: float = 51.0
+; CHECK-NEXT: exit: float = 1.0
+
+; CHECK: block-frequency-info: test_expansion_cost_2
+; CHECK-NEXT: entry: float = 1.0
+; CHECK-NEXT: entry.split: float = 0.98039
+; CHECK-NEXT: loop.header: float = 50.0
+; CHECK-NEXT: then: float = 20.0
+; CHECK-NEXT: loop.latch: float = 50.0
+; CHECK-NEXT: exit.peel.begin.loopexit: float = 0.98039
+; CHECK-NEXT: exit.peel.begin: float = 1.0
+; CHECK-NEXT: loop.header.peel: float = 1.0
+; CHECK-NEXT: then.peel: float = 0.4
+; CHECK-NEXT: loop.latch.peel: float = 1.0
+; CHECK-NEXT: exit.peel.next: float = 1.0
+; CHECK-NEXT: loop.header.peel.next: float = 1.0
+; CHECK-NEXT: exit: float = 1.0
\ No newline at end of file
More information about the llvm-commits
mailing list