[llvm] [SimpleLoopUnswitch] Fix exponential unswitch (PR #66883)
Nikita Popov via llvm-commits
llvm-commits at lists.llvm.org
Wed Sep 20 03:14:01 PDT 2023
https://github.com/nikic created https://github.com/llvm/llvm-project/pull/66883
When unswitching via invariant condition injection, we currently mark the condition in the old loop, so that it does not get unswitched again. However, if there are multiple branches for which conditions can be injected, then we can do that for both the old and new loops. This means that the number of unswitches increases exponentially.
Change the handling to be more similar to partial unswitching, where we instead mark the whole loop, rather than a single condition. This means that we will only generate a linear number of loops.
TBH I think even that is still highly undesirable, and we should probably be unswitching all candidates at the same time, so that we end up with only two loops. But at least this mitigates the worst case.
The test case is a reduced variant that generates 1700 lines of IR without this patch and 290 with it.
Fixes https://github.com/llvm/llvm-project/issues/66868.
>From 866577f13d99bf0ba903a81a5ef038a334a7fc2f Mon Sep 17 00:00:00 2001
From: Nikita Popov <npopov at redhat.com>
Date: Wed, 20 Sep 2023 11:41:42 +0200
Subject: [PATCH] [SimpleLoopUnswitch] Fix exponential unswitch
When unswitching via invariant condition injection, we currently
mark the condition in the old loop, so that it does not get
unswitched again. However, if there are multiple branches for
which conditions can be injected, then we can do that for both
the old and new loop. This means that the number of unswitches
increases exponentially.
Change the handling to be more similar to partial unswitching,
where we instead mark the whole loop, rather than a single
condition. This means that we will only generate a linear number
of loops.
TBH I think even that is still highly undesirable, and we should
probably be unswitching all candidates at the same time, so that
we end up with only two loops. But at least this mitigates the
worst case.
The test case is a reduced variant that generates 1700 lines of IR
without this patch and 290 with it.
Fixes https://github.com/llvm/llvm-project/issues/66868.
---
.../Transforms/Scalar/SimpleLoopUnswitch.cpp | 57 ++--
...inject-invariant-conditions-exponential.ll | 260 ++++++++++++++++++
.../inject-invariant-conditions.ll | 55 ++--
3 files changed, 327 insertions(+), 45 deletions(-)
create mode 100644 llvm/test/Transforms/SimpleLoopUnswitch/inject-invariant-conditions-exponential.ll
diff --git a/llvm/lib/Transforms/Scalar/SimpleLoopUnswitch.cpp b/llvm/lib/Transforms/Scalar/SimpleLoopUnswitch.cpp
index cad026ffbbd6b74..8ac974bdfecf84a 100644
--- a/llvm/lib/Transforms/Scalar/SimpleLoopUnswitch.cpp
+++ b/llvm/lib/Transforms/Scalar/SimpleLoopUnswitch.cpp
@@ -2127,9 +2127,10 @@ static void unswitchNontrivialInvariants(
Loop &L, Instruction &TI, ArrayRef<Value *> Invariants,
IVConditionInfo &PartialIVInfo, DominatorTree &DT, LoopInfo &LI,
AssumptionCache &AC,
- function_ref<void(bool, bool, ArrayRef<Loop *>)> UnswitchCB,
+ function_ref<void(bool, bool, bool, ArrayRef<Loop *>)> UnswitchCB,
ScalarEvolution *SE, MemorySSAUpdater *MSSAU,
- function_ref<void(Loop &, StringRef)> DestroyLoopCB, bool InsertFreeze) {
+ function_ref<void(Loop &, StringRef)> DestroyLoopCB, bool InsertFreeze,
+ bool InjectedCondition) {
auto *ParentBB = TI.getParent();
BranchInst *BI = dyn_cast<BranchInst>(&TI);
SwitchInst *SI = BI ? nullptr : cast<SwitchInst>(&TI);
@@ -2582,7 +2583,7 @@ static void unswitchNontrivialInvariants(
for (Loop *UpdatedL : llvm::concat<Loop *>(NonChildClonedLoops, HoistedLoops))
if (UpdatedL->getParentLoop() == ParentL)
SibLoops.push_back(UpdatedL);
- UnswitchCB(IsStillLoop, PartiallyInvariant, SibLoops);
+ UnswitchCB(IsStillLoop, PartiallyInvariant, InjectedCondition, SibLoops);
if (MSSAU && VerifyMemorySSA)
MSSAU->getMemorySSA()->verifyMemorySSA();
@@ -2980,13 +2981,6 @@ static bool shouldTryInjectInvariantCondition(
/// the metadata.
bool shouldTryInjectBasingOnMetadata(const BranchInst *BI,
const BasicBlock *TakenSucc) {
- // Skip branches that have already been unswithed this way. After successful
- // unswitching of injected condition, we will still have a copy of this loop
- // which looks exactly the same as original one. To prevent the 2nd attempt
- // of unswitching it in the same pass, mark this branch as "nothing to do
- // here".
- if (BI->hasMetadata("llvm.invariant.condition.injection.disabled"))
- return false;
SmallVector<uint32_t> Weights;
if (!extractBranchWeights(*BI, Weights))
return false;
@@ -3069,13 +3063,9 @@ injectPendingInvariantConditions(NonTrivialUnswitchCandidate Candidate, Loop &L,
Builder.CreateCondBr(InjectedCond, InLoopSucc, CheckBlock);
Builder.SetInsertPoint(CheckBlock);
- auto *NewTerm = Builder.CreateCondBr(TI->getCondition(), TI->getSuccessor(0),
- TI->getSuccessor(1));
-
+ Builder.CreateCondBr(TI->getCondition(), TI->getSuccessor(0),
+ TI->getSuccessor(1));
TI->eraseFromParent();
- // Prevent infinite unswitching.
- NewTerm->setMetadata("llvm.invariant.condition.injection.disabled",
- MDNode::get(BB->getContext(), {}));
// Fixup phis.
for (auto &I : *InLoopSucc) {
@@ -3443,7 +3433,7 @@ static bool shouldInsertFreeze(Loop &L, Instruction &TI, DominatorTree &DT,
static bool unswitchBestCondition(
Loop &L, DominatorTree &DT, LoopInfo &LI, AssumptionCache &AC,
AAResults &AA, TargetTransformInfo &TTI,
- function_ref<void(bool, bool, ArrayRef<Loop *>)> UnswitchCB,
+ function_ref<void(bool, bool, bool, ArrayRef<Loop *>)> UnswitchCB,
ScalarEvolution *SE, MemorySSAUpdater *MSSAU,
function_ref<void(Loop &, StringRef)> DestroyLoopCB) {
// Collect all invariant conditions within this loop (as opposed to an inner
@@ -3453,9 +3443,10 @@ static bool unswitchBestCondition(
Instruction *PartialIVCondBranch = nullptr;
collectUnswitchCandidates(UnswitchCandidates, PartialIVInfo,
PartialIVCondBranch, L, LI, AA, MSSAU);
- collectUnswitchCandidatesWithInjections(UnswitchCandidates, PartialIVInfo,
- PartialIVCondBranch, L, DT, LI, AA,
- MSSAU);
+ if (!findOptionMDForLoop(&L, "llvm.loop.unswitch.injection.disable"))
+ collectUnswitchCandidatesWithInjections(UnswitchCandidates, PartialIVInfo,
+ PartialIVCondBranch, L, DT, LI, AA,
+ MSSAU);
// If we didn't find any candidates, we're done.
if (UnswitchCandidates.empty())
return false;
@@ -3476,8 +3467,11 @@ static bool unswitchBestCondition(
return false;
}
- if (Best.hasPendingInjection())
+ bool InjectedCondition = false;
+ if (Best.hasPendingInjection()) {
Best = injectPendingInvariantConditions(Best, L, DT, LI, AC, MSSAU);
+ InjectedCondition = true;
+ }
assert(!Best.hasPendingInjection() &&
"All injections should have been done by now!");
@@ -3505,7 +3499,7 @@ static bool unswitchBestCondition(
<< ") terminator: " << *Best.TI << "\n");
unswitchNontrivialInvariants(L, *Best.TI, Best.Invariants, PartialIVInfo, DT,
LI, AC, UnswitchCB, SE, MSSAU, DestroyLoopCB,
- InsertFreeze);
+ InsertFreeze, InjectedCondition);
return true;
}
@@ -3534,7 +3528,7 @@ static bool
unswitchLoop(Loop &L, DominatorTree &DT, LoopInfo &LI, AssumptionCache &AC,
AAResults &AA, TargetTransformInfo &TTI, bool Trivial,
bool NonTrivial,
- function_ref<void(bool, bool, ArrayRef<Loop *>)> UnswitchCB,
+ function_ref<void(bool, bool, bool, ArrayRef<Loop *>)> UnswitchCB,
ScalarEvolution *SE, MemorySSAUpdater *MSSAU,
ProfileSummaryInfo *PSI, BlockFrequencyInfo *BFI,
function_ref<void(Loop &, StringRef)> DestroyLoopCB) {
@@ -3549,7 +3543,8 @@ unswitchLoop(Loop &L, DominatorTree &DT, LoopInfo &LI, AssumptionCache &AC,
if (Trivial && unswitchAllTrivialConditions(L, DT, LI, SE, MSSAU)) {
// If we unswitched successfully we will want to clean up the loop before
// processing it further so just mark it as unswitched and return.
- UnswitchCB(/*CurrentLoopValid*/ true, false, {});
+ UnswitchCB(/*CurrentLoopValid*/ true, /*PartiallyInvariant*/ false,
+ /*InjectedCondition*/ false, {});
return true;
}
@@ -3645,6 +3640,7 @@ PreservedAnalyses SimpleLoopUnswitchPass::run(Loop &L, LoopAnalysisManager &AM,
auto UnswitchCB = [&L, &U, &LoopName](bool CurrentLoopValid,
bool PartiallyInvariant,
+ bool InjectedCondition,
ArrayRef<Loop *> NewLoops) {
// If we did a non-trivial unswitch, we have added new (cloned) loops.
if (!NewLoops.empty())
@@ -3664,6 +3660,16 @@ PreservedAnalyses SimpleLoopUnswitchPass::run(Loop &L, LoopAnalysisManager &AM,
Context, L.getLoopID(), {"llvm.loop.unswitch.partial"},
{DisableUnswitchMD});
L.setLoopID(NewLoopID);
+ } else if (InjectedCondition) {
+ // Do the same for injection of invariant conditions.
+ auto &Context = L.getHeader()->getContext();
+ MDNode *DisableUnswitchMD = MDNode::get(
+ Context,
+ MDString::get(Context, "llvm.loop.unswitch.injection.disable"));
+ MDNode *NewLoopID = makePostTransformationMetadata(
+ Context, L.getLoopID(), {"llvm.loop.unswitch.injection"},
+ {DisableUnswitchMD});
+ L.setLoopID(NewLoopID);
} else
U.revisitCurrentLoop();
} else
@@ -3756,6 +3762,7 @@ bool SimpleLoopUnswitchLegacyPass::runOnLoop(Loop *L, LPPassManager &LPM) {
auto *SE = SEWP ? &SEWP->getSE() : nullptr;
auto UnswitchCB = [&L, &LPM](bool CurrentLoopValid, bool PartiallyInvariant,
+ bool InjectedCondition,
ArrayRef<Loop *> NewLoops) {
// If we did a non-trivial unswitch, we have added new (cloned) loops.
for (auto *NewL : NewLoops)
@@ -3768,7 +3775,7 @@ bool SimpleLoopUnswitchLegacyPass::runOnLoop(Loop *L, LPPassManager &LPM) {
// If the current loop has been unswitched using a partially invariant
// condition, we should not re-add the current loop to avoid unswitching
// on the same condition again.
- if (!PartiallyInvariant)
+ if (!PartiallyInvariant && !InjectedCondition)
LPM.addLoop(*L);
} else
LPM.markLoopAsDeleted(*L);
diff --git a/llvm/test/Transforms/SimpleLoopUnswitch/inject-invariant-conditions-exponential.ll b/llvm/test/Transforms/SimpleLoopUnswitch/inject-invariant-conditions-exponential.ll
new file mode 100644
index 000000000000000..bd88b4030d675e6
--- /dev/null
+++ b/llvm/test/Transforms/SimpleLoopUnswitch/inject-invariant-conditions-exponential.ll
@@ -0,0 +1,260 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 3
+; RUN: opt -S -passes='simple-loop-unswitch<nontrivial>' < %s | FileCheck %s
+
+; Make sure invariant condition injection does not result in exponential
+; size increase.
+
+; FIXME: It probably shouldn't result in linear size increase either.
+
+define void @ham(i64 %arg) {
+; CHECK-LABEL: define void @ham(
+; CHECK-SAME: i64 [[ARG:%.*]]) {
+; CHECK-NEXT: bb:
+; CHECK-NEXT: [[INJECTED_COND:%.*]] = icmp ule i64 [[ARG]], [[ARG]]
+; CHECK-NEXT: [[INJECTED_COND_FR:%.*]] = freeze i1 [[INJECTED_COND]]
+; CHECK-NEXT: br i1 [[INJECTED_COND_FR]], label [[BB_SPLIT_US:%.*]], label [[BB_SPLIT:%.*]]
+; CHECK: bb.split.us:
+; CHECK-NEXT: [[INJECTED_COND1:%.*]] = icmp ule i64 [[ARG]], [[ARG]]
+; CHECK-NEXT: [[INJECTED_COND1_FR:%.*]] = freeze i1 [[INJECTED_COND1]]
+; CHECK-NEXT: br i1 [[INJECTED_COND1_FR]], label [[BB_SPLIT_US_SPLIT_US:%.*]], label [[BB_SPLIT_US_SPLIT:%.*]]
+; CHECK: bb.split.us.split.us:
+; CHECK-NEXT: [[INJECTED_COND2:%.*]] = icmp ule i64 [[ARG]], [[ARG]]
+; CHECK-NEXT: [[INJECTED_COND2_FR:%.*]] = freeze i1 [[INJECTED_COND2]]
+; CHECK-NEXT: br i1 [[INJECTED_COND2_FR]], label [[BB_SPLIT_US_SPLIT_US_SPLIT_US:%.*]], label [[BB_SPLIT_US_SPLIT_US_SPLIT:%.*]]
+; CHECK: bb.split.us.split.us.split.us:
+; CHECK-NEXT: [[INJECTED_COND3:%.*]] = icmp ule i64 [[ARG]], [[ARG]]
+; CHECK-NEXT: [[INJECTED_COND3_FR:%.*]] = freeze i1 [[INJECTED_COND3]]
+; CHECK-NEXT: br i1 [[INJECTED_COND3_FR]], label [[BB_SPLIT_US_SPLIT_US_SPLIT_US_SPLIT_US:%.*]], label [[BB_SPLIT_US_SPLIT_US_SPLIT_US_SPLIT:%.*]]
+; CHECK: bb.split.us.split.us.split.us.split.us:
+; CHECK-NEXT: [[INJECTED_COND4:%.*]] = icmp ule i64 [[ARG]], [[ARG]]
+; CHECK-NEXT: [[INJECTED_COND4_FR:%.*]] = freeze i1 [[INJECTED_COND4]]
+; CHECK-NEXT: br i1 [[INJECTED_COND4_FR]], label [[BB_SPLIT_US_SPLIT_US_SPLIT_US_SPLIT_US_SPLIT_US:%.*]], label [[BB_SPLIT_US_SPLIT_US_SPLIT_US_SPLIT_US_SPLIT:%.*]]
+; CHECK: bb.split.us.split.us.split.us.split.us.split.us:
+; CHECK-NEXT: br label [[BB1_US_US_US_US_US:%.*]]
+; CHECK: bb1.us.us.us.us.us:
+; CHECK-NEXT: [[PHI_US_US_US_US_US:%.*]] = phi i64 [ 0, [[BB_SPLIT_US_SPLIT_US_SPLIT_US_SPLIT_US_SPLIT_US]] ], [ [[ADD_US_US_US_US_US:%.*]], [[BB20_US_US_US_US_US:%.*]] ]
+; CHECK-NEXT: [[ADD_US_US_US_US_US]] = add nuw i64 [[PHI_US_US_US_US_US]], 1
+; CHECK-NEXT: [[ICMP_US_US_US_US_US:%.*]] = icmp ult i64 [[PHI_US_US_US_US_US]], [[ARG]]
+; CHECK-NEXT: br i1 [[ICMP_US_US_US_US_US]], label [[BB2_US_US_US_US_US:%.*]], label [[BB21_SPLIT_US_SPLIT_US_SPLIT_US_SPLIT_US_SPLIT_US:%.*]], !prof [[PROF0:![0-9]+]]
+; CHECK: bb2.us.us.us.us.us:
+; CHECK-NEXT: [[ICMP3_US_US_US_US_US:%.*]] = icmp ult i64 [[PHI_US_US_US_US_US]], [[ARG]]
+; CHECK-NEXT: br label [[BB4_US_US_US_US_US:%.*]]
+; CHECK: bb4.us.us.us.us.us:
+; CHECK-NEXT: [[ICMP5_US_US_US_US_US:%.*]] = icmp ult i64 [[PHI_US_US_US_US_US]], [[ARG]]
+; CHECK-NEXT: br label [[BB6_US_US_US_US_US:%.*]]
+; CHECK: bb6.us.us.us.us.us:
+; CHECK-NEXT: [[ICMP7_US_US_US_US_US:%.*]] = icmp ult i64 [[PHI_US_US_US_US_US]], [[ARG]]
+; CHECK-NEXT: br label [[BB8_US_US_US_US_US:%.*]]
+; CHECK: bb8.us.us.us.us.us:
+; CHECK-NEXT: [[ICMP9_US_US_US_US_US:%.*]] = icmp ult i64 [[PHI_US_US_US_US_US]], [[ARG]]
+; CHECK-NEXT: br label [[BB10_US_US_US_US_US:%.*]]
+; CHECK: bb10.us.us.us.us.us:
+; CHECK-NEXT: [[ICMP11_US_US_US_US_US:%.*]] = icmp ult i64 [[PHI_US_US_US_US_US]], [[ARG]]
+; CHECK-NEXT: br label [[BB20_US_US_US_US_US]]
+; CHECK: bb20.us.us.us.us.us:
+; CHECK-NEXT: br label [[BB1_US_US_US_US_US]]
+; CHECK: bb21.split.us.split.us.split.us.split.us.split.us:
+; CHECK-NEXT: br label [[BB21_SPLIT_US_SPLIT_US_SPLIT_US_SPLIT_US:%.*]]
+; CHECK: bb.split.us.split.us.split.us.split.us.split:
+; CHECK-NEXT: br label [[BB1_US_US_US_US:%.*]]
+; CHECK: bb1.us.us.us.us:
+; CHECK-NEXT: [[PHI_US_US_US_US:%.*]] = phi i64 [ 0, [[BB_SPLIT_US_SPLIT_US_SPLIT_US_SPLIT_US_SPLIT]] ], [ [[ADD_US_US_US_US:%.*]], [[BB20_US_US_US_US:%.*]] ]
+; CHECK-NEXT: [[ADD_US_US_US_US]] = add nuw i64 [[PHI_US_US_US_US]], 1
+; CHECK-NEXT: [[ICMP_US_US_US_US:%.*]] = icmp ult i64 [[PHI_US_US_US_US]], [[ARG]]
+; CHECK-NEXT: br i1 [[ICMP_US_US_US_US]], label [[BB2_US_US_US_US:%.*]], label [[BB21_SPLIT_US_SPLIT_US_SPLIT_US_SPLIT_US_SPLIT:%.*]], !prof [[PROF0]]
+; CHECK: bb2.us.us.us.us:
+; CHECK-NEXT: [[ICMP3_US_US_US_US:%.*]] = icmp ult i64 [[PHI_US_US_US_US]], [[ARG]]
+; CHECK-NEXT: br label [[BB4_US_US_US_US:%.*]]
+; CHECK: bb4.us.us.us.us:
+; CHECK-NEXT: [[ICMP5_US_US_US_US:%.*]] = icmp ult i64 [[PHI_US_US_US_US]], [[ARG]]
+; CHECK-NEXT: br label [[BB6_US_US_US_US:%.*]]
+; CHECK: bb6.us.us.us.us:
+; CHECK-NEXT: [[ICMP7_US_US_US_US:%.*]] = icmp ult i64 [[PHI_US_US_US_US]], [[ARG]]
+; CHECK-NEXT: br label [[BB8_US_US_US_US:%.*]]
+; CHECK: bb8.us.us.us.us:
+; CHECK-NEXT: [[ICMP9_US_US_US_US:%.*]] = icmp ult i64 [[PHI_US_US_US_US]], [[ARG]]
+; CHECK-NEXT: br label [[BB10_US_US_US_US:%.*]]
+; CHECK: bb10.us.us.us.us:
+; CHECK-NEXT: [[ICMP11_US_US_US_US:%.*]] = icmp ult i64 [[PHI_US_US_US_US]], [[ARG]]
+; CHECK-NEXT: br label [[BB10_US_US_US_US_CHECK:%.*]]
+; CHECK: bb10.us.us.us.us.check:
+; CHECK-NEXT: br i1 [[ICMP11_US_US_US_US]], label [[BB20_US_US_US_US]], label [[BB21_SPLIT_US_SPLIT_US_SPLIT_US_SPLIT_US_SPLIT]]
+; CHECK: bb20.us.us.us.us:
+; CHECK-NEXT: br label [[BB1_US_US_US_US]], !llvm.loop [[LOOP1:![0-9]+]]
+; CHECK: bb21.split.us.split.us.split.us.split.us.split:
+; CHECK-NEXT: br label [[BB21_SPLIT_US_SPLIT_US_SPLIT_US_SPLIT_US]]
+; CHECK: bb21.split.us.split.us.split.us.split.us:
+; CHECK-NEXT: br label [[BB21_SPLIT_US_SPLIT_US_SPLIT_US:%.*]]
+; CHECK: bb.split.us.split.us.split.us.split:
+; CHECK-NEXT: br label [[BB1_US_US_US:%.*]]
+; CHECK: bb1.us.us.us:
+; CHECK-NEXT: [[PHI_US_US_US:%.*]] = phi i64 [ 0, [[BB_SPLIT_US_SPLIT_US_SPLIT_US_SPLIT]] ], [ [[ADD_US_US_US:%.*]], [[BB20_US_US_US:%.*]] ]
+; CHECK-NEXT: [[ADD_US_US_US]] = add nuw i64 [[PHI_US_US_US]], 1
+; CHECK-NEXT: [[ICMP_US_US_US:%.*]] = icmp ult i64 [[PHI_US_US_US]], [[ARG]]
+; CHECK-NEXT: br i1 [[ICMP_US_US_US]], label [[BB2_US_US_US:%.*]], label [[BB21_SPLIT_US_SPLIT_US_SPLIT_US_SPLIT:%.*]], !prof [[PROF0]]
+; CHECK: bb2.us.us.us:
+; CHECK-NEXT: [[ICMP3_US_US_US:%.*]] = icmp ult i64 [[PHI_US_US_US]], [[ARG]]
+; CHECK-NEXT: br label [[BB4_US_US_US:%.*]]
+; CHECK: bb4.us.us.us:
+; CHECK-NEXT: [[ICMP5_US_US_US:%.*]] = icmp ult i64 [[PHI_US_US_US]], [[ARG]]
+; CHECK-NEXT: br label [[BB6_US_US_US:%.*]]
+; CHECK: bb6.us.us.us:
+; CHECK-NEXT: [[ICMP7_US_US_US:%.*]] = icmp ult i64 [[PHI_US_US_US]], [[ARG]]
+; CHECK-NEXT: br label [[BB8_US_US_US:%.*]]
+; CHECK: bb8.us.us.us:
+; CHECK-NEXT: [[ICMP9_US_US_US:%.*]] = icmp ult i64 [[PHI_US_US_US]], [[ARG]]
+; CHECK-NEXT: br label [[BB8_US_US_US_CHECK:%.*]]
+; CHECK: bb8.us.us.us.check:
+; CHECK-NEXT: br i1 [[ICMP9_US_US_US]], label [[BB10_US_US_US:%.*]], label [[BB21_SPLIT_US_SPLIT_US_SPLIT_US_SPLIT]]
+; CHECK: bb10.us.us.us:
+; CHECK-NEXT: [[ICMP11_US_US_US:%.*]] = icmp ult i64 [[PHI_US_US_US]], [[ARG]]
+; CHECK-NEXT: br i1 [[ICMP11_US_US_US]], label [[BB20_US_US_US]], label [[BB21_SPLIT_US_SPLIT_US_SPLIT_US_SPLIT]], !prof [[PROF0]]
+; CHECK: bb20.us.us.us:
+; CHECK-NEXT: br label [[BB1_US_US_US]], !llvm.loop [[LOOP3:![0-9]+]]
+; CHECK: bb21.split.us.split.us.split.us.split:
+; CHECK-NEXT: br label [[BB21_SPLIT_US_SPLIT_US_SPLIT_US]]
+; CHECK: bb21.split.us.split.us.split.us:
+; CHECK-NEXT: br label [[BB21_SPLIT_US_SPLIT_US:%.*]]
+; CHECK: bb.split.us.split.us.split:
+; CHECK-NEXT: br label [[BB1_US_US:%.*]]
+; CHECK: bb1.us.us:
+; CHECK-NEXT: [[PHI_US_US:%.*]] = phi i64 [ 0, [[BB_SPLIT_US_SPLIT_US_SPLIT]] ], [ [[ADD_US_US:%.*]], [[BB20_US_US:%.*]] ]
+; CHECK-NEXT: [[ADD_US_US]] = add nuw i64 [[PHI_US_US]], 1
+; CHECK-NEXT: [[ICMP_US_US:%.*]] = icmp ult i64 [[PHI_US_US]], [[ARG]]
+; CHECK-NEXT: br i1 [[ICMP_US_US]], label [[BB2_US_US:%.*]], label [[BB21_SPLIT_US_SPLIT_US_SPLIT:%.*]], !prof [[PROF0]]
+; CHECK: bb2.us.us:
+; CHECK-NEXT: [[ICMP3_US_US:%.*]] = icmp ult i64 [[PHI_US_US]], [[ARG]]
+; CHECK-NEXT: br label [[BB4_US_US:%.*]]
+; CHECK: bb4.us.us:
+; CHECK-NEXT: [[ICMP5_US_US:%.*]] = icmp ult i64 [[PHI_US_US]], [[ARG]]
+; CHECK-NEXT: br label [[BB6_US_US:%.*]]
+; CHECK: bb6.us.us:
+; CHECK-NEXT: [[ICMP7_US_US:%.*]] = icmp ult i64 [[PHI_US_US]], [[ARG]]
+; CHECK-NEXT: br label [[BB6_US_US_CHECK:%.*]]
+; CHECK: bb6.us.us.check:
+; CHECK-NEXT: br i1 [[ICMP7_US_US]], label [[BB8_US_US:%.*]], label [[BB21_SPLIT_US_SPLIT_US_SPLIT]]
+; CHECK: bb8.us.us:
+; CHECK-NEXT: [[ICMP9_US_US:%.*]] = icmp ult i64 [[PHI_US_US]], [[ARG]]
+; CHECK-NEXT: br i1 [[ICMP9_US_US]], label [[BB10_US_US:%.*]], label [[BB21_SPLIT_US_SPLIT_US_SPLIT]], !prof [[PROF0]]
+; CHECK: bb10.us.us:
+; CHECK-NEXT: [[ICMP11_US_US:%.*]] = icmp ult i64 [[PHI_US_US]], [[ARG]]
+; CHECK-NEXT: br i1 [[ICMP11_US_US]], label [[BB20_US_US]], label [[BB21_SPLIT_US_SPLIT_US_SPLIT]], !prof [[PROF0]]
+; CHECK: bb20.us.us:
+; CHECK-NEXT: br label [[BB1_US_US]], !llvm.loop [[LOOP4:![0-9]+]]
+; CHECK: bb21.split.us.split.us.split:
+; CHECK-NEXT: br label [[BB21_SPLIT_US_SPLIT_US]]
+; CHECK: bb21.split.us.split.us:
+; CHECK-NEXT: br label [[BB21_SPLIT_US:%.*]]
+; CHECK: bb.split.us.split:
+; CHECK-NEXT: br label [[BB1_US:%.*]]
+; CHECK: bb1.us:
+; CHECK-NEXT: [[PHI_US:%.*]] = phi i64 [ 0, [[BB_SPLIT_US_SPLIT]] ], [ [[ADD_US:%.*]], [[BB20_US:%.*]] ]
+; CHECK-NEXT: [[ADD_US]] = add nuw i64 [[PHI_US]], 1
+; CHECK-NEXT: [[ICMP_US:%.*]] = icmp ult i64 [[PHI_US]], [[ARG]]
+; CHECK-NEXT: br i1 [[ICMP_US]], label [[BB2_US:%.*]], label [[BB21_SPLIT_US_SPLIT:%.*]], !prof [[PROF0]]
+; CHECK: bb2.us:
+; CHECK-NEXT: [[ICMP3_US:%.*]] = icmp ult i64 [[PHI_US]], [[ARG]]
+; CHECK-NEXT: br label [[BB4_US:%.*]]
+; CHECK: bb4.us:
+; CHECK-NEXT: [[ICMP5_US:%.*]] = icmp ult i64 [[PHI_US]], [[ARG]]
+; CHECK-NEXT: br label [[BB4_US_CHECK:%.*]]
+; CHECK: bb4.us.check:
+; CHECK-NEXT: br i1 [[ICMP5_US]], label [[BB6_US:%.*]], label [[BB22_SPLIT_US:%.*]]
+; CHECK: bb6.us:
+; CHECK-NEXT: [[ICMP7_US:%.*]] = icmp ult i64 [[PHI_US]], [[ARG]]
+; CHECK-NEXT: br i1 [[ICMP7_US]], label [[BB8_US:%.*]], label [[BB21_SPLIT_US_SPLIT]], !prof [[PROF0]]
+; CHECK: bb8.us:
+; CHECK-NEXT: [[ICMP9_US:%.*]] = icmp ult i64 [[PHI_US]], [[ARG]]
+; CHECK-NEXT: br i1 [[ICMP9_US]], label [[BB10_US:%.*]], label [[BB21_SPLIT_US_SPLIT]], !prof [[PROF0]]
+; CHECK: bb10.us:
+; CHECK-NEXT: [[ICMP11_US:%.*]] = icmp ult i64 [[PHI_US]], [[ARG]]
+; CHECK-NEXT: br i1 [[ICMP11_US]], label [[BB20_US]], label [[BB21_SPLIT_US_SPLIT]], !prof [[PROF0]]
+; CHECK: bb20.us:
+; CHECK-NEXT: br label [[BB1_US]], !llvm.loop [[LOOP5:![0-9]+]]
+; CHECK: bb21.split.us.split:
+; CHECK-NEXT: br label [[BB21_SPLIT_US]]
+; CHECK: bb21.split.us:
+; CHECK-NEXT: br label [[BB21:%.*]]
+; CHECK: bb22.split.us:
+; CHECK-NEXT: br label [[BB22:%.*]]
+; CHECK: bb.split:
+; CHECK-NEXT: br label [[BB1:%.*]]
+; CHECK: bb1:
+; CHECK-NEXT: [[PHI:%.*]] = phi i64 [ 0, [[BB_SPLIT]] ], [ [[ADD:%.*]], [[BB20:%.*]] ]
+; CHECK-NEXT: [[ADD]] = add nuw i64 [[PHI]], 1
+; CHECK-NEXT: [[ICMP:%.*]] = icmp ult i64 [[PHI]], [[ARG]]
+; CHECK-NEXT: br i1 [[ICMP]], label [[BB2:%.*]], label [[BB21_SPLIT:%.*]], !prof [[PROF0]]
+; CHECK: bb2:
+; CHECK-NEXT: [[ICMP3:%.*]] = icmp ult i64 [[PHI]], [[ARG]]
+; CHECK-NEXT: br label [[BB2_CHECK:%.*]]
+; CHECK: bb2.check:
+; CHECK-NEXT: br i1 [[ICMP3]], label [[BB4:%.*]], label [[BB21_SPLIT]]
+; CHECK: bb4:
+; CHECK-NEXT: [[ICMP5:%.*]] = icmp ult i64 [[PHI]], [[ARG]]
+; CHECK-NEXT: br i1 [[ICMP5]], label [[BB6:%.*]], label [[BB22_SPLIT:%.*]], !prof [[PROF0]]
+; CHECK: bb6:
+; CHECK-NEXT: [[ICMP7:%.*]] = icmp ult i64 [[PHI]], [[ARG]]
+; CHECK-NEXT: br i1 [[ICMP7]], label [[BB8:%.*]], label [[BB21_SPLIT]], !prof [[PROF0]]
+; CHECK: bb8:
+; CHECK-NEXT: [[ICMP9:%.*]] = icmp ult i64 [[PHI]], [[ARG]]
+; CHECK-NEXT: br i1 [[ICMP9]], label [[BB10:%.*]], label [[BB21_SPLIT]], !prof [[PROF0]]
+; CHECK: bb10:
+; CHECK-NEXT: [[ICMP11:%.*]] = icmp ult i64 [[PHI]], [[ARG]]
+; CHECK-NEXT: br i1 [[ICMP11]], label [[BB20]], label [[BB21_SPLIT]], !prof [[PROF0]]
+; CHECK: bb20:
+; CHECK-NEXT: br label [[BB1]], !llvm.loop [[LOOP6:![0-9]+]]
+; CHECK: bb21.split:
+; CHECK-NEXT: br label [[BB21]]
+; CHECK: bb21:
+; CHECK-NEXT: call void @zot()
+; CHECK-NEXT: ret void
+; CHECK: bb22.split:
+; CHECK-NEXT: br label [[BB22]]
+; CHECK: bb22:
+; CHECK-NEXT: call void @zot()
+; CHECK-NEXT: ret void
+;
+bb:
+ br label %bb1
+
+bb1: ; preds = %bb20, %bb
+ %phi = phi i64 [ 0, %bb ], [ %add, %bb20 ]
+ %add = add nuw i64 %phi, 1
+ %icmp = icmp ult i64 %phi, %arg
+ br i1 %icmp, label %bb2, label %bb21, !prof !0
+
+bb2: ; preds = %bb1
+ %icmp3 = icmp ult i64 %phi, %arg
+ br i1 %icmp3, label %bb4, label %bb21, !prof !0
+
+bb4: ; preds = %bb2
+ %icmp5 = icmp ult i64 %phi, %arg
+ br i1 %icmp5, label %bb6, label %bb22, !prof !0
+
+bb6: ; preds = %bb4
+ %icmp7 = icmp ult i64 %phi, %arg
+ br i1 %icmp7, label %bb8, label %bb21, !prof !0
+
+bb8: ; preds = %bb6
+ %icmp9 = icmp ult i64 %phi, %arg
+ br i1 %icmp9, label %bb10, label %bb21, !prof !0
+
+bb10: ; preds = %bb8
+ %icmp11 = icmp ult i64 %phi, %arg
+ br i1 %icmp11, label %bb20, label %bb21, !prof !0
+
+bb20: ; preds = %bb18
+ br label %bb1
+
+bb21: ; preds = %bb18, %bb16, %bb14, %bb12, %bb10, %bb8, %bb6, %bb2, %bb1
+ call void @zot()
+ ret void
+
+bb22: ; preds = %bb4
+ call void @zot()
+ ret void
+}
+
+declare void @zot()
+
+!0 = !{!"branch_weights", i32 2000, i32 1}
diff --git a/llvm/test/Transforms/SimpleLoopUnswitch/inject-invariant-conditions.ll b/llvm/test/Transforms/SimpleLoopUnswitch/inject-invariant-conditions.ll
index 615f36027450a0a..536e0c6a0e74adf 100644
--- a/llvm/test/Transforms/SimpleLoopUnswitch/inject-invariant-conditions.ll
+++ b/llvm/test/Transforms/SimpleLoopUnswitch/inject-invariant-conditions.ll
@@ -1,4 +1,4 @@
-; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --check-globals
; RUN: opt < %s -S -simple-loop-unswitch-inject-invariant-conditions=true -passes="loop(simple-loop-unswitch<nontrivial>),simplifycfg" | FileCheck %s
; RUN: opt < %s -S -simple-loop-unswitch-inject-invariant-conditions=true -passes="loop-mssa(simple-loop-unswitch<nontrivial>),simplifycfg" -verify-memoryssa | FileCheck %s
@@ -29,13 +29,13 @@ define i32 @test_01(ptr noundef %p, i32 noundef %n, i32 noundef %limit, ptr noun
; CHECK-NEXT: br i1 [[BOUND_CHECK]], label [[GUARDED:%.*]], label [[COMMON_RET]], !prof [[PROF1]]
; CHECK: guarded:
; CHECK-NEXT: [[RANGE_CHECK:%.*]] = icmp ult i32 [[EL]], [[X]]
-; CHECK-NEXT: br i1 [[RANGE_CHECK]], label [[BACKEDGE]], label [[COMMON_RET]], !llvm.invariant.condition.injection.disabled !0
+; CHECK-NEXT: br i1 [[RANGE_CHECK]], label [[BACKEDGE]], label [[COMMON_RET]]
; CHECK: backedge:
; CHECK-NEXT: [[ARR_PTR:%.*]] = getelementptr i32, ptr [[ARR]], i32 [[EL]]
; CHECK-NEXT: store i32 [[IV]], ptr [[ARR_PTR]], align 4
; CHECK-NEXT: [[IV_NEXT]] = add i32 [[IV]], 1
; CHECK-NEXT: [[LOOP_COND:%.*]] = icmp slt i32 [[IV_NEXT]], [[N]]
-; CHECK-NEXT: br i1 [[LOOP_COND]], label [[LOOP]], label [[COMMON_RET]]
+; CHECK-NEXT: br i1 [[LOOP_COND]], label [[LOOP]], label [[COMMON_RET]], !llvm.loop [[LOOP2:![0-9]+]]
; CHECK: common.ret:
; CHECK-NEXT: [[COMMON_RET_OP:%.*]] = phi i32 [ 0, [[BACKEDGE]] ], [ 0, [[GUARDED_US]] ], [ -1, [[LOOP]] ], [ -1, [[LOOP_US]] ], [ -2, [[GUARDED]] ]
; CHECK-NEXT: ret i32 [[COMMON_RET_OP]]
@@ -160,7 +160,7 @@ define i32 @test_01_constants(ptr noundef %p, ptr noundef %arr, ptr noundef %x_p
; CHECK-NEXT: store i32 [[IV]], ptr [[ARR_PTR]], align 4
; CHECK-NEXT: [[IV_NEXT]] = add i32 [[IV]], 1
; CHECK-NEXT: [[LOOP_COND:%.*]] = icmp slt i32 [[IV_NEXT]], 1000
-; CHECK-NEXT: br i1 [[LOOP_COND]], label [[LOOP]], label [[COMMON_RET]]
+; CHECK-NEXT: br i1 [[LOOP_COND]], label [[LOOP]], label [[COMMON_RET]], !llvm.loop [[LOOP4:![0-9]+]]
; CHECK: common.ret:
; CHECK-NEXT: [[COMMON_RET_OP:%.*]] = phi i32 [ 0, [[BACKEDGE]] ], [ 0, [[GUARDED_US]] ], [ -1, [[LOOP]] ], [ -1, [[LOOP_US]] ]
; CHECK-NEXT: ret i32 [[COMMON_RET_OP]]
@@ -210,7 +210,7 @@ define i32 @test_01_neg_degenerate_profile(ptr noundef %p, i32 noundef %n, i32 n
; CHECK-NEXT: br i1 [[BOUND_CHECK]], label [[GUARDED:%.*]], label [[COMMON_RET:%.*]], !prof [[PROF1]]
; CHECK: guarded:
; CHECK-NEXT: [[RANGE_CHECK:%.*]] = icmp ult i32 [[EL]], [[X]]
-; CHECK-NEXT: br i1 [[RANGE_CHECK]], label [[BACKEDGE]], label [[COMMON_RET]], !prof [[PROF2:![0-9]+]]
+; CHECK-NEXT: br i1 [[RANGE_CHECK]], label [[BACKEDGE]], label [[COMMON_RET]], !prof [[PROF5:![0-9]+]]
; CHECK: backedge:
; CHECK-NEXT: [[ARR_PTR:%.*]] = getelementptr i32, ptr [[ARR:%.*]], i32 [[EL]]
; CHECK-NEXT: store i32 [[IV]], ptr [[ARR_PTR]], align 4
@@ -267,7 +267,7 @@ define i32 @test_01_neg_cold(ptr noundef %p, i32 noundef %n, i32 noundef %limit,
; CHECK-NEXT: br i1 [[BOUND_CHECK]], label [[GUARDED:%.*]], label [[COMMON_RET:%.*]], !prof [[PROF1]]
; CHECK: guarded:
; CHECK-NEXT: [[RANGE_CHECK:%.*]] = icmp ult i32 [[EL]], [[X]]
-; CHECK-NEXT: br i1 [[RANGE_CHECK]], label [[BACKEDGE]], label [[COMMON_RET]], !prof [[PROF3:![0-9]+]]
+; CHECK-NEXT: br i1 [[RANGE_CHECK]], label [[BACKEDGE]], label [[COMMON_RET]], !prof [[PROF6:![0-9]+]]
; CHECK: backedge:
; CHECK-NEXT: [[ARR_PTR:%.*]] = getelementptr i32, ptr [[ARR:%.*]], i32 [[EL]]
; CHECK-NEXT: store i32 [[IV]], ptr [[ARR_PTR]], align 4
@@ -321,10 +321,10 @@ define i32 @test_01_neg_overflowing_metadata(ptr noundef %p, i32 noundef %n, i32
; CHECK-NEXT: [[EL_PTR:%.*]] = getelementptr i32, ptr [[P:%.*]], i32 [[IV]]
; CHECK-NEXT: [[EL:%.*]] = load i32, ptr [[EL_PTR]], align 4
; CHECK-NEXT: [[BOUND_CHECK:%.*]] = icmp ult i32 [[EL]], [[LIMIT:%.*]]
-; CHECK-NEXT: br i1 [[BOUND_CHECK]], label [[GUARDED:%.*]], label [[COMMON_RET:%.*]], !prof [[PROF4:![0-9]+]]
+; CHECK-NEXT: br i1 [[BOUND_CHECK]], label [[GUARDED:%.*]], label [[COMMON_RET:%.*]], !prof [[PROF7:![0-9]+]]
; CHECK: guarded:
; CHECK-NEXT: [[RANGE_CHECK:%.*]] = icmp ult i32 [[EL]], [[X]]
-; CHECK-NEXT: br i1 [[RANGE_CHECK]], label [[BACKEDGE]], label [[COMMON_RET]], !prof [[PROF4]]
+; CHECK-NEXT: br i1 [[RANGE_CHECK]], label [[BACKEDGE]], label [[COMMON_RET]], !prof [[PROF7]]
; CHECK: backedge:
; CHECK-NEXT: [[ARR_PTR:%.*]] = getelementptr i32, ptr [[ARR:%.*]], i32 [[EL]]
; CHECK-NEXT: store i32 [[IV]], ptr [[ARR_PTR]], align 4
@@ -395,13 +395,13 @@ define i32 @test_02(ptr noundef %p, i32 noundef %n, i32 noundef %limit, ptr noun
; CHECK-NEXT: br i1 [[BOUND_CHECK]], label [[GUARDED:%.*]], label [[COMMON_RET]], !prof [[PROF1]]
; CHECK: guarded:
; CHECK-NEXT: [[RANGE_CHECK:%.*]] = icmp ult i32 [[EL]], [[X]]
-; CHECK-NEXT: br i1 [[RANGE_CHECK]], label [[BACKEDGE]], label [[COMMON_RET]], !llvm.invariant.condition.injection.disabled !0
+; CHECK-NEXT: br i1 [[RANGE_CHECK]], label [[BACKEDGE]], label [[COMMON_RET]]
; CHECK: backedge:
; CHECK-NEXT: [[ARR_PTR:%.*]] = getelementptr i32, ptr [[ARR]], i32 [[EL]]
; CHECK-NEXT: store i32 [[IV]], ptr [[ARR_PTR]], align 4
; CHECK-NEXT: [[IV_NEXT]] = add i32 [[IV]], 1
; CHECK-NEXT: [[LOOP_COND:%.*]] = icmp slt i32 [[IV_NEXT]], [[N]]
-; CHECK-NEXT: br i1 [[LOOP_COND]], label [[LOOP]], label [[COMMON_RET]]
+; CHECK-NEXT: br i1 [[LOOP_COND]], label [[LOOP]], label [[COMMON_RET]], !llvm.loop [[LOOP8:![0-9]+]]
; CHECK: common.ret:
; CHECK-NEXT: [[COMMON_RET_OP:%.*]] = phi i32 [ 0, [[BACKEDGE]] ], [ 0, [[GUARDED_US]] ], [ -1, [[LOOP]] ], [ -1, [[LOOP_US]] ], [ -2, [[GUARDED]] ]
; CHECK-NEXT: ret i32 [[COMMON_RET_OP]]
@@ -465,13 +465,13 @@ define i32 @test_02_inverse(ptr noundef %p, i32 noundef %n, i32 noundef %limit,
; CHECK-NEXT: br i1 [[BOUND_CHECK]], label [[GUARDED:%.*]], label [[COMMON_RET]], !prof [[PROF1]]
; CHECK: guarded:
; CHECK-NEXT: [[RANGE_CHECK:%.*]] = icmp uge i32 [[EL]], [[X]]
-; CHECK-NEXT: br i1 [[RANGE_CHECK]], label [[COMMON_RET]], label [[BACKEDGE]], !llvm.invariant.condition.injection.disabled !0
+; CHECK-NEXT: br i1 [[RANGE_CHECK]], label [[COMMON_RET]], label [[BACKEDGE]]
; CHECK: backedge:
; CHECK-NEXT: [[ARR_PTR:%.*]] = getelementptr i32, ptr [[ARR]], i32 [[EL]]
; CHECK-NEXT: store i32 [[IV]], ptr [[ARR_PTR]], align 4
; CHECK-NEXT: [[IV_NEXT]] = add i32 [[IV]], 1
; CHECK-NEXT: [[LOOP_COND:%.*]] = icmp slt i32 [[IV_NEXT]], [[N]]
-; CHECK-NEXT: br i1 [[LOOP_COND]], label [[LOOP]], label [[COMMON_RET]]
+; CHECK-NEXT: br i1 [[LOOP_COND]], label [[LOOP]], label [[COMMON_RET]], !llvm.loop [[LOOP9:![0-9]+]]
; CHECK: common.ret:
; CHECK-NEXT: [[COMMON_RET_OP:%.*]] = phi i32 [ 0, [[BACKEDGE]] ], [ 0, [[GUARDED_US]] ], [ -1, [[LOOP]] ], [ -1, [[LOOP_US]] ], [ -2, [[GUARDED]] ]
; CHECK-NEXT: ret i32 [[COMMON_RET_OP]]
@@ -519,7 +519,7 @@ define i32 @test_03(ptr noundef %p, i32 noundef %n, i32 noundef %limit, ptr noun
; CHECK-NEXT: [[EL_PTR_US:%.*]] = getelementptr i32, ptr [[P:%.*]], i32 [[IV_US]]
; CHECK-NEXT: [[EL_US:%.*]] = load i32, ptr [[EL_PTR_US]], align 4
; CHECK-NEXT: [[BOUND_CHECK_US:%.*]] = icmp slt i32 [[EL_US]], 0
-; CHECK-NEXT: br i1 [[BOUND_CHECK_US]], label [[COMMON_RET:%.*]], label [[GUARDED_US]], !prof [[PROF5:![0-9]+]]
+; CHECK-NEXT: br i1 [[BOUND_CHECK_US]], label [[COMMON_RET:%.*]], label [[GUARDED_US]], !prof [[PROF10:![0-9]+]]
; CHECK: guarded.us:
; CHECK-NEXT: [[RANGE_CHECK_US:%.*]] = icmp ult i32 [[EL_US]], [[X]]
; CHECK-NEXT: [[ARR_PTR_US:%.*]] = getelementptr i32, ptr [[ARR:%.*]], i32 [[EL_US]]
@@ -532,16 +532,16 @@ define i32 @test_03(ptr noundef %p, i32 noundef %n, i32 noundef %limit, ptr noun
; CHECK-NEXT: [[EL_PTR:%.*]] = getelementptr i32, ptr [[P]], i32 [[IV]]
; CHECK-NEXT: [[EL:%.*]] = load i32, ptr [[EL_PTR]], align 4
; CHECK-NEXT: [[BOUND_CHECK:%.*]] = icmp slt i32 [[EL]], 0
-; CHECK-NEXT: br i1 [[BOUND_CHECK]], label [[COMMON_RET]], label [[GUARDED:%.*]], !prof [[PROF5]]
+; CHECK-NEXT: br i1 [[BOUND_CHECK]], label [[COMMON_RET]], label [[GUARDED:%.*]], !prof [[PROF10]]
; CHECK: guarded:
; CHECK-NEXT: [[RANGE_CHECK:%.*]] = icmp ult i32 [[EL]], [[X]]
-; CHECK-NEXT: br i1 [[RANGE_CHECK]], label [[BACKEDGE]], label [[COMMON_RET]], !llvm.invariant.condition.injection.disabled !0
+; CHECK-NEXT: br i1 [[RANGE_CHECK]], label [[BACKEDGE]], label [[COMMON_RET]]
; CHECK: backedge:
; CHECK-NEXT: [[ARR_PTR:%.*]] = getelementptr i32, ptr [[ARR]], i32 [[EL]]
; CHECK-NEXT: store i32 [[IV]], ptr [[ARR_PTR]], align 4
; CHECK-NEXT: [[IV_NEXT]] = add i32 [[IV]], 1
; CHECK-NEXT: [[LOOP_COND:%.*]] = icmp slt i32 [[IV_NEXT]], [[N]]
-; CHECK-NEXT: br i1 [[LOOP_COND]], label [[LOOP]], label [[COMMON_RET]]
+; CHECK-NEXT: br i1 [[LOOP_COND]], label [[LOOP]], label [[COMMON_RET]], !llvm.loop [[LOOP11:![0-9]+]]
; CHECK: common.ret:
; CHECK-NEXT: [[COMMON_RET_OP:%.*]] = phi i32 [ 0, [[BACKEDGE]] ], [ 0, [[GUARDED_US]] ], [ -1, [[LOOP]] ], [ -1, [[LOOP_US]] ], [ -2, [[GUARDED]] ]
; CHECK-NEXT: ret i32 [[COMMON_RET_OP]]
@@ -589,7 +589,7 @@ define i32 @test_04(ptr noundef %p, i32 noundef %n, i32 noundef %limit, ptr noun
; CHECK-NEXT: [[EL_PTR_US:%.*]] = getelementptr i8, ptr [[P:%.*]], i32 [[IV_US]]
; CHECK-NEXT: [[EL_US:%.*]] = load i8, ptr [[EL_PTR_US]], align 4
; CHECK-NEXT: [[BOUND_CHECK_US:%.*]] = icmp slt i8 [[EL_US]], 0
-; CHECK-NEXT: br i1 [[BOUND_CHECK_US]], label [[COMMON_RET:%.*]], label [[GUARDED_US]], !prof [[PROF5]]
+; CHECK-NEXT: br i1 [[BOUND_CHECK_US]], label [[COMMON_RET:%.*]], label [[GUARDED_US]], !prof [[PROF10]]
; CHECK: guarded.us:
; CHECK-NEXT: [[EL_WIDE_US:%.*]] = zext i8 [[EL_US]] to i32
; CHECK-NEXT: [[RANGE_CHECK_US:%.*]] = icmp ult i32 [[EL_WIDE_US]], [[X]]
@@ -603,17 +603,17 @@ define i32 @test_04(ptr noundef %p, i32 noundef %n, i32 noundef %limit, ptr noun
; CHECK-NEXT: [[EL_PTR:%.*]] = getelementptr i8, ptr [[P]], i32 [[IV]]
; CHECK-NEXT: [[EL:%.*]] = load i8, ptr [[EL_PTR]], align 4
; CHECK-NEXT: [[BOUND_CHECK:%.*]] = icmp slt i8 [[EL]], 0
-; CHECK-NEXT: br i1 [[BOUND_CHECK]], label [[COMMON_RET]], label [[GUARDED:%.*]], !prof [[PROF5]]
+; CHECK-NEXT: br i1 [[BOUND_CHECK]], label [[COMMON_RET]], label [[GUARDED:%.*]], !prof [[PROF10]]
; CHECK: guarded:
; CHECK-NEXT: [[EL_WIDE:%.*]] = zext i8 [[EL]] to i32
; CHECK-NEXT: [[RANGE_CHECK:%.*]] = icmp ult i32 [[EL_WIDE]], [[X]]
-; CHECK-NEXT: br i1 [[RANGE_CHECK]], label [[BACKEDGE]], label [[COMMON_RET]], !llvm.invariant.condition.injection.disabled !0
+; CHECK-NEXT: br i1 [[RANGE_CHECK]], label [[BACKEDGE]], label [[COMMON_RET]]
; CHECK: backedge:
; CHECK-NEXT: [[ARR_PTR:%.*]] = getelementptr i32, ptr [[ARR]], i32 [[EL_WIDE]]
; CHECK-NEXT: store i32 [[IV]], ptr [[ARR_PTR]], align 4
; CHECK-NEXT: [[IV_NEXT]] = add i32 [[IV]], 1
; CHECK-NEXT: [[LOOP_COND:%.*]] = icmp slt i32 [[IV_NEXT]], [[N]]
-; CHECK-NEXT: br i1 [[LOOP_COND]], label [[LOOP]], label [[COMMON_RET]]
+; CHECK-NEXT: br i1 [[LOOP_COND]], label [[LOOP]], label [[COMMON_RET]], !llvm.loop [[LOOP12:![0-9]+]]
; CHECK: common.ret:
; CHECK-NEXT: [[COMMON_RET_OP:%.*]] = phi i32 [ 0, [[BACKEDGE]] ], [ 0, [[GUARDED_US]] ], [ -1, [[LOOP]] ], [ -1, [[LOOP_US]] ], [ -2, [[GUARDED]] ]
; CHECK-NEXT: ret i32 [[COMMON_RET_OP]]
@@ -650,3 +650,18 @@ bound_check_failed: ; preds = %loop
range_check_failed: ; preds = %guarded
ret i32 -2
}
+;.
+; CHECK: [[META0:![0-9]+]] = !{}
+; CHECK: [[PROF1]] = !{!"branch_weights", i32 100, i32 1}
+; CHECK: [[LOOP2]] = distinct !{!2, !3}
+; CHECK: [[META3:![0-9]+]] = !{!"llvm.loop.unswitch.injection.disable"}
+; CHECK: [[LOOP4]] = distinct !{!4, !3}
+; CHECK: [[PROF5]] = !{!"branch_weights", i32 0, i32 0}
+; CHECK: [[PROF6]] = !{!"branch_weights", i32 2, i32 3}
+; CHECK: [[PROF7]] = !{!"branch_weights", i32 -1, i32 -1000}
+; CHECK: [[LOOP8]] = distinct !{!8, !3}
+; CHECK: [[LOOP9]] = distinct !{!9, !3}
+; CHECK: [[PROF10]] = !{!"branch_weights", i32 1, i32 100}
+; CHECK: [[LOOP11]] = distinct !{!11, !3}
+; CHECK: [[LOOP12]] = distinct !{!12, !3}
+;.
More information about the llvm-commits
mailing list