This is an archive of the discontinued LLVM Phabricator instance.

Differential D42604

PR36032 fix assert cause by not computed SCEV PredicatedBackEdgeCount
ClosedPublic

Authored by evstupac on Jan 26 2018, 3:00 PM.

Details

Reviewers
a.elovikov
silviu.baranga
dim
dorit
sbaranga
Commits
rGa732611ac822: Fix PR36032, PR35432
rL326154: Fix PR36032, PR35432
Summary

The change fix an assert fail at ScalarEvolutionExpander.cpp (generateOverflowCheck):

const SCEV *ExitCount =                              
    SE.getPredicatedBackedgeTakenCount(AR->getLoop(), Pred); 
                                                     
assert(ExitCount != SE.getCouldNotCompute() && "Invalid loop count");

(generateOverflowCheck is called only from expandWrapPredicate)

by inserting an early check on SCEV PredicateBackEdgeCount when createAddRecFromPHIWithCastsImpl

Diff Detail

Repository
rL LLVM

Event Timeline

evstupac created this revision.Jan 26 2018, 3:00 PM
Herald added a subscriber: sanjoy. · View Herald TranscriptJan 26 2018, 3:00 PM
sbaranga added a subscriber: sbaranga.Jan 29 2018, 1:12 AM

Thanks for submitting the fix! Please see the inline comments.

Silviu

lib/Analysis/ScalarEvolution.cpp
4657

This assumes that L is a sub-loop of the loop being processed by the rewriter.

To make this correct you would also have to add the predicates returned by getPredicatedBackedgeTakenCount to the set of returned predicates.

I think this function might also get called from getPredicatedBackedgeTakenCount and might result in an infinite loop.

I think there is an easier solution (see the comment below).

11561

I think at least for now the fix can be done here, and should check that all overflow predicates are done on L (the loop currently being processed by the rewriter).

egarcia added a subscriber: egarcia.Jan 29 2018, 9:47 AM
evstupac updated this revision to Diff 133710.Feb 9 2018, 3:58 PM
evstupac added inline comments.
lib/Analysis/ScalarEvolution.cpp
11561

Moved the check - trying to address your suggestion. Please take a look.

sbaranga added a comment.Feb 12 2018, 1:03 AM

LGTM with one minor nit.

lib/Analysis/ScalarEvolution.cpp
11548

Not related to this change, but it might be worth not processing Expr if it is invariant in L.

11558

This should be cast<> instead of dyn_cast since WP->getExpr() is always an AddRecExpr.

evstupac updated this revision to Diff 133946.Feb 12 2018, 3:04 PM
sbaranga accepted this revision.Feb 19 2018, 8:09 AM
This revision is now accepted and ready to land.Feb 19 2018, 8:09 AM
Closed by commit rL326154: Fix PR36032, PR35432 (authored by evstupac). · Explain WhyFeb 26 2018, 4:19 PM
This revision was automatically updated to reflect the committed changes.

Revision Contents

PathSize
lib/
Analysis/
5 lines
test/
Transforms/
LoopVectorize/
213 lines
154 lines

Diff 133710

lib/Analysis/ScalarEvolution.cpp

  • This file is larger than 256 KB, so syntax highlighting is disabled by default.
Show First 20 Lines • Show All 4,648 Lines • ▼ Show 20 LinesScalarEvolution::createAddRecFromPHIWithCastsImpl(const SCEVUnknown *SymbolicPHI) {
if (const auto *AR = dyn_cast<SCEVAddRecExpr>(PHISCEV)) { if (const auto *AR = dyn_cast<SCEVAddRecExpr>(PHISCEV)) {
SCEVWrapPredicate::IncrementWrapFlags AddedFlags = SCEVWrapPredicate::IncrementWrapFlags AddedFlags =
Signed ? SCEVWrapPredicate::IncrementNSSW Signed ? SCEVWrapPredicate::IncrementNSSW
: SCEVWrapPredicate::IncrementNUSW; : SCEVWrapPredicate::IncrementNUSW;
const SCEVPredicate *AddRecPred = getWrapPredicate(AR, AddedFlags); const SCEVPredicate *AddRecPred = getWrapPredicate(AR, AddedFlags);
Predicates.push_back(AddRecPred); Predicates.push_back(AddRecPred);
} }
// Create the Equal Predicates P2,P3: // Create the Equal Predicates P2,P3:
sbarangaUnsubmitted
Not Done
ReplyInline Actions

This assumes that L is a sub-loop of the loop being processed by the rewriter.

To make this correct you would also have to add the predicates returned by getPredicatedBackedgeTakenCount to the set of returned predicates.

I think this function might also get called from getPredicatedBackedgeTakenCount and might result in an infinite loop.

I think there is an easier solution (see the comment below).

sbaranga: This assumes that L is a sub-loop of the loop being processed by the rewriter. To make this…
// It is possible that the predicates P2 and/or P3 are computable at // It is possible that the predicates P2 and/or P3 are computable at
// compile time due to StartVal and/or Accum being constants. // compile time due to StartVal and/or Accum being constants.
// If either one is, then we can check that now and escape if either P2 // If either one is, then we can check that now and escape if either P2
// or P3 is false. // or P3 is false.
// Construct the extended SCEV: (Ext ix (Trunc iy (Expr) to ix) to iy) // Construct the extended SCEV: (Ext ix (Trunc iy (Expr) to ix) to iy)
// for each of StartVal and Accum // for each of StartVal and Accum
▲ Show 20 Lines • Show All 6,874 Lines • ▼ Show 20 Linesprivate:
} }
// If \p Expr represents a PHINode, we try to see if it can be represented // If \p Expr represents a PHINode, we try to see if it can be represented
// as an AddRec, possibly under a predicate (PHISCEVPred). If it is possible // as an AddRec, possibly under a predicate (PHISCEVPred). If it is possible
// to add this predicate as a runtime overflow check, we return the AddRec. // to add this predicate as a runtime overflow check, we return the AddRec.
// If \p Expr does not meet these conditions (is not a PHI node, or we // If \p Expr does not meet these conditions (is not a PHI node, or we
// couldn't create an AddRec for it, or couldn't add the predicate), we just // couldn't create an AddRec for it, or couldn't add the predicate), we just
// return \p Expr. // return \p Expr.
const SCEV *convertToAddRecWithPreds(const SCEVUnknown *Expr) { const SCEV *convertToAddRecWithPreds(const SCEVUnknown *Expr) {
sbarangaUnsubmitted
Not Done
ReplyInline Actions

Not related to this change, but it might be worth not processing Expr if it is invariant in L.

sbaranga: Not related to this change, but it might be worth not processing Expr if it is invariant in L.
if (!isa<PHINode>(Expr->getValue())) if (!isa<PHINode>(Expr->getValue()))
return Expr; return Expr;
Optional<std::pair<const SCEV *, SmallVector<const SCEVPredicate *, 3>>> Optional<std::pair<const SCEV *, SmallVector<const SCEVPredicate *, 3>>>
PredicatedRewrite = SE.createAddRecFromPHIWithCasts(Expr); PredicatedRewrite = SE.createAddRecFromPHIWithCasts(Expr);
if (!PredicatedRewrite) if (!PredicatedRewrite)
return Expr; return Expr;
for (auto *P : PredicatedRewrite->second){ for (auto *P : PredicatedRewrite->second){
// Wrap predicates from outer loops are not supported.
if (auto *WP = dyn_cast<const SCEVWrapPredicate>(P))
if (auto *AR = dyn_cast<const SCEVAddRecExpr>(WP->getExpr()))
sbarangaUnsubmitted
Not Done
ReplyInline Actions

This should be cast<> instead of dyn_cast since WP->getExpr() is always an AddRecExpr.

sbaranga: This should be cast<> instead of dyn_cast since WP->getExpr() is always an AddRecExpr.
if (L != AR->getLoop())
return Expr;
if (!addOverflowAssumption(P)) if (!addOverflowAssumption(P))
sbarangaUnsubmitted
Not Done
ReplyInline Actions

I think at least for now the fix can be done here, and should check that all overflow predicates are done on L (the loop currently being processed by the rewriter).

sbaranga: I think at least for now the fix can be done here, and should check that all overflow…
evstupacAuthorUnsubmitted
Not Done
ReplyInline Actions

Moved the check - trying to address your suggestion. Please take a look.

evstupac: Moved the check - trying to address your suggestion. Please take a look.
return Expr; return Expr;
} }
return PredicatedRewrite->first; return PredicatedRewrite->first;
} }
SmallPtrSetImpl<const SCEVPredicate *> *NewPreds; SmallPtrSetImpl<const SCEVPredicate *> *NewPreds;
SCEVUnionPredicate *Pred; SCEVUnionPredicate *Pred;
const Loop *L; const Loop *L;
▲ Show 20 Lines • Show All 296 Lines • Show Last 20 Lines

test/Transforms/LoopVectorize/pr35432.ll

; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
; RUN: opt -loop-vectorize -S < %s | FileCheck %s
; The test checks that there is no assert caused by issue described in PR35432
target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
target triple = "x86_64-unknown-linux-gnu"
@a = common local_unnamed_addr global [192 x [192 x i32]] zeroinitializer, align 16
; Function Attrs: nounwind uwtable
define i32 @main() local_unnamed_addr #0 {
; CHECK-LABEL: @main(
; CHECK-NEXT: entry:
; CHECK-NEXT: [[I:%.*]] = alloca i32, align 4
; CHECK-NEXT: [[S:%.*]] = alloca i16, align 2
; CHECK-NEXT: [[TMP0:%.*]] = bitcast i32* [[I]] to i8*
; CHECK-NEXT: call void @llvm.lifetime.start.p0i8(i64 4, i8* nonnull [[TMP0]])
; CHECK-NEXT: store i32 0, i32* [[I]], align 4
; CHECK-NEXT: [[TMP1:%.*]] = bitcast i16* [[S]] to i8*
; CHECK-NEXT: call void @llvm.lifetime.start.p0i8(i64 2, i8* nonnull [[TMP1]])
; CHECK-NEXT: [[CALL:%.*]] = call i32 (i32*, ...) bitcast (i32 (...)* @goo to i32 (i32*, ...)*)(i32* nonnull [[I]])
; CHECK-NEXT: [[TMP2:%.*]] = load i32, i32* [[I]], align 4
; CHECK-NEXT: [[STOREMERGE6:%.*]] = trunc i32 [[TMP2]] to i16
; CHECK-NEXT: store i16 [[STOREMERGE6]], i16* [[S]], align 2
; CHECK-NEXT: [[CONV17:%.*]] = and i32 [[TMP2]], 65472
; CHECK-NEXT: [[CMP8:%.*]] = icmp eq i32 [[CONV17]], 0
; CHECK-NEXT: br i1 [[CMP8]], label [[FOR_BODY_LR_PH:%.*]], label [[FOR_END12:%.*]]
; CHECK: for.body.lr.ph:
; CHECK-NEXT: [[TMP3:%.*]] = sub i32 -1, [[TMP2]]
; CHECK-NEXT: br label [[FOR_BODY:%.*]]
; CHECK: for.body:
; CHECK-NEXT: [[STOREMERGE_IN9:%.*]] = phi i32 [ [[TMP2]], [[FOR_BODY_LR_PH]] ], [ [[ADD:%.*]], [[FOR_INC9:%.*]] ]
; CHECK-NEXT: [[CONV52:%.*]] = and i32 [[STOREMERGE_IN9]], 255
; CHECK-NEXT: [[CMP63:%.*]] = icmp ult i32 [[TMP2]], [[CONV52]]
; CHECK-NEXT: br i1 [[CMP63]], label [[FOR_BODY8_LR_PH:%.*]], label [[FOR_INC9]]
; CHECK: for.body8.lr.ph:
; CHECK-NEXT: [[CONV3:%.*]] = trunc i32 [[STOREMERGE_IN9]] to i8
; CHECK-NEXT: [[DOTPROMOTED:%.*]] = load i32, i32* getelementptr inbounds ([192 x [192 x i32]], [192 x [192 x i32]]* @a, i64 0, i64 0, i64 0), align 16
; CHECK-NEXT: [[TMP4:%.*]] = add i8 [[CONV3]], -1
; CHECK-NEXT: [[TMP5:%.*]] = zext i8 [[TMP4]] to i32
; CHECK-NEXT: [[TMP6:%.*]] = sub i32 -1, [[TMP5]]
; CHECK-NEXT: [[TMP7:%.*]] = icmp ugt i32 [[TMP6]], [[TMP3]]
; CHECK-NEXT: [[UMAX:%.*]] = select i1 [[TMP7]], i32 [[TMP6]], i32 [[TMP3]]
; CHECK-NEXT: [[TMP8:%.*]] = add i32 [[UMAX]], 2
; CHECK-NEXT: [[TMP9:%.*]] = add i32 [[TMP8]], [[TMP5]]
; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP9]], 8
; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_SCEVCHECK:%.*]]
; CHECK: vector.scevcheck:
; CHECK-NEXT: [[TMP10:%.*]] = add i8 [[CONV3]], -1
; CHECK-NEXT: [[TMP11:%.*]] = zext i8 [[TMP10]] to i32
; CHECK-NEXT: [[TMP12:%.*]] = sub i32 -1, [[TMP11]]
; CHECK-NEXT: [[TMP13:%.*]] = icmp ugt i32 [[TMP12]], [[TMP3]]
; CHECK-NEXT: [[UMAX1:%.*]] = select i1 [[TMP13]], i32 [[TMP12]], i32 [[TMP3]]
; CHECK-NEXT: [[TMP14:%.*]] = add i32 [[UMAX1]], 1
; CHECK-NEXT: [[TMP15:%.*]] = add i32 [[TMP14]], [[TMP11]]
; CHECK-NEXT: [[TMP16:%.*]] = trunc i32 [[TMP15]] to i8
; CHECK-NEXT: [[MUL:%.*]] = call { i8, i1 } @llvm.umul.with.overflow.i8(i8 1, i8 [[TMP16]])
; CHECK-NEXT: [[MUL_RESULT:%.*]] = extractvalue { i8, i1 } [[MUL]], 0
; CHECK-NEXT: [[MUL_OVERFLOW:%.*]] = extractvalue { i8, i1 } [[MUL]], 1
; CHECK-NEXT: [[TMP17:%.*]] = add i8 [[TMP10]], [[MUL_RESULT]]
; CHECK-NEXT: [[TMP18:%.*]] = sub i8 [[TMP10]], [[MUL_RESULT]]
; CHECK-NEXT: [[TMP19:%.*]] = icmp ugt i8 [[TMP18]], [[TMP10]]
; CHECK-NEXT: [[TMP20:%.*]] = icmp ult i8 [[TMP17]], [[TMP10]]
; CHECK-NEXT: [[TMP21:%.*]] = select i1 true, i1 [[TMP19]], i1 [[TMP20]]
; CHECK-NEXT: [[TMP22:%.*]] = icmp ugt i32 [[TMP15]], 255
; CHECK-NEXT: [[TMP23:%.*]] = or i1 [[TMP21]], [[TMP22]]
; CHECK-NEXT: [[TMP24:%.*]] = or i1 [[TMP23]], [[MUL_OVERFLOW]]
; CHECK-NEXT: [[TMP25:%.*]] = or i1 false, [[TMP24]]
; CHECK-NEXT: br i1 [[TMP25]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
; CHECK: vector.ph:
; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i32 [[TMP9]], 8
; CHECK-NEXT: [[N_VEC:%.*]] = sub i32 [[TMP9]], [[N_MOD_VF]]
; CHECK-NEXT: [[CAST_CRD:%.*]] = trunc i32 [[N_VEC]] to i8
; CHECK-NEXT: [[IND_END:%.*]] = sub i8 [[CONV3]], [[CAST_CRD]]
; CHECK-NEXT: [[TMP26:%.*]] = insertelement <4 x i32> zeroinitializer, i32 [[DOTPROMOTED]], i32 0
; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]
; CHECK: vector.body:
; CHECK-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT: [[VEC_PHI:%.*]] = phi <4 x i32> [ [[TMP26]], [[VECTOR_PH]] ], [ [[TMP30:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT: [[VEC_PHI2:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP31:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT: [[TMP27:%.*]] = trunc i32 [[INDEX]] to i8
; CHECK-NEXT: [[OFFSET_IDX:%.*]] = sub i8 [[CONV3]], [[TMP27]]
; CHECK-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i8> undef, i8 [[OFFSET_IDX]], i32 0
; CHECK-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i8> [[BROADCAST_SPLATINSERT]], <4 x i8> undef, <4 x i32> zeroinitializer
; CHECK-NEXT: [[INDUCTION:%.*]] = add <4 x i8> [[BROADCAST_SPLAT]], <i8 0, i8 -1, i8 -2, i8 -3>
; CHECK-NEXT: [[INDUCTION3:%.*]] = add <4 x i8> [[BROADCAST_SPLAT]], <i8 -4, i8 -5, i8 -6, i8 -7>
; CHECK-NEXT: [[TMP28:%.*]] = add i8 [[OFFSET_IDX]], 0
; CHECK-NEXT: [[TMP29:%.*]] = add i8 [[OFFSET_IDX]], -4
; CHECK-NEXT: [[TMP30]] = add <4 x i32> [[VEC_PHI]], <i32 1, i32 1, i32 1, i32 1>
; CHECK-NEXT: [[TMP31]] = add <4 x i32> [[VEC_PHI2]], <i32 1, i32 1, i32 1, i32 1>
; CHECK-NEXT: [[TMP32:%.*]] = add i8 [[TMP28]], -1
; CHECK-NEXT: [[TMP33:%.*]] = add i8 [[TMP29]], -1
; CHECK-NEXT: [[TMP34:%.*]] = zext i8 [[TMP32]] to i32
; CHECK-NEXT: [[TMP35:%.*]] = zext i8 [[TMP33]] to i32
; CHECK-NEXT: [[INDEX_NEXT]] = add i32 [[INDEX]], 8
; CHECK-NEXT: [[TMP36:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-NEXT: br i1 [[TMP36]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop !0
; CHECK: middle.block:
; CHECK-NEXT: [[BIN_RDX:%.*]] = add <4 x i32> [[TMP31]], [[TMP30]]
; CHECK-NEXT: [[RDX_SHUF:%.*]] = shufflevector <4 x i32> [[BIN_RDX]], <4 x i32> undef, <4 x i32> <i32 2, i32 3, i32 undef, i32 undef>
; CHECK-NEXT: [[BIN_RDX4:%.*]] = add <4 x i32> [[BIN_RDX]], [[RDX_SHUF]]
; CHECK-NEXT: [[RDX_SHUF5:%.*]] = shufflevector <4 x i32> [[BIN_RDX4]], <4 x i32> undef, <4 x i32> <i32 1, i32 undef, i32 undef, i32 undef>
; CHECK-NEXT: [[BIN_RDX6:%.*]] = add <4 x i32> [[BIN_RDX4]], [[RDX_SHUF5]]
; CHECK-NEXT: [[TMP37:%.*]] = extractelement <4 x i32> [[BIN_RDX6]], i32 0
; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i32 [[TMP9]], [[N_VEC]]
; CHECK-NEXT: br i1 [[CMP_N]], label [[FOR_COND4_FOR_INC9_CRIT_EDGE:%.*]], label [[SCALAR_PH]]
; CHECK: scalar.ph:
; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i8 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[CONV3]], [[FOR_BODY8_LR_PH]] ], [ [[CONV3]], [[VECTOR_SCEVCHECK]] ]
; CHECK-NEXT: [[BC_MERGE_RDX:%.*]] = phi i32 [ [[DOTPROMOTED]], [[FOR_BODY8_LR_PH]] ], [ [[DOTPROMOTED]], [[VECTOR_SCEVCHECK]] ], [ [[TMP37]], [[MIDDLE_BLOCK]] ]
; CHECK-NEXT: br label [[FOR_BODY8:%.*]]
; CHECK: for.body8:
; CHECK-NEXT: [[INC5:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[INC:%.*]], [[FOR_BODY8]] ]
; CHECK-NEXT: [[C_04:%.*]] = phi i8 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[DEC:%.*]], [[FOR_BODY8]] ]
; CHECK-NEXT: [[INC]] = add i32 [[INC5]], 1
; CHECK-NEXT: [[DEC]] = add i8 [[C_04]], -1
; CHECK-NEXT: [[CONV5:%.*]] = zext i8 [[DEC]] to i32
; CHECK-NEXT: [[CMP6:%.*]] = icmp ult i32 [[TMP2]], [[CONV5]]
; CHECK-NEXT: br i1 [[CMP6]], label [[FOR_BODY8]], label [[FOR_COND4_FOR_INC9_CRIT_EDGE]], !llvm.loop !2
; CHECK: for.cond4.for.inc9_crit_edge:
; CHECK-NEXT: [[INC_LCSSA:%.*]] = phi i32 [ [[INC]], [[FOR_BODY8]] ], [ [[TMP37]], [[MIDDLE_BLOCK]] ]
; CHECK-NEXT: store i32 [[INC_LCSSA]], i32* getelementptr inbounds ([192 x [192 x i32]], [192 x [192 x i32]]* @a, i64 0, i64 0, i64 0), align 16
; CHECK-NEXT: br label [[FOR_INC9]]
; CHECK: for.inc9:
; CHECK-NEXT: [[CONV10:%.*]] = and i32 [[STOREMERGE_IN9]], 65535
; CHECK-NEXT: [[ADD]] = add nuw nsw i32 [[CONV10]], 1
; CHECK-NEXT: [[CONV1:%.*]] = and i32 [[ADD]], 65472
; CHECK-NEXT: [[CMP:%.*]] = icmp eq i32 [[CONV1]], 0
; CHECK-NEXT: br i1 [[CMP]], label [[FOR_BODY]], label [[FOR_COND_FOR_END12_CRIT_EDGE:%.*]]
; CHECK: for.cond.for.end12_crit_edge:
; CHECK-NEXT: [[ADD_LCSSA:%.*]] = phi i32 [ [[ADD]], [[FOR_INC9]] ]
; CHECK-NEXT: [[STOREMERGE:%.*]] = trunc i32 [[ADD_LCSSA]] to i16
; CHECK-NEXT: store i16 [[STOREMERGE]], i16* [[S]], align 2
; CHECK-NEXT: br label [[FOR_END12]]
; CHECK: for.end12:
; CHECK-NEXT: [[CALL13:%.*]] = call i32 (i16*, ...) bitcast (i32 (...)* @foo to i32 (i16*, ...)*)(i16* nonnull [[S]])
; CHECK-NEXT: call void @llvm.lifetime.end.p0i8(i64 2, i8* nonnull [[TMP1]])
; CHECK-NEXT: call void @llvm.lifetime.end.p0i8(i64 4, i8* nonnull [[TMP0]])
; CHECK-NEXT: ret i32 0
;
entry:
%i = alloca i32, align 4
%s = alloca i16, align 2
%0 = bitcast i32* %i to i8*
call void @llvm.lifetime.start.p0i8(i64 4, i8* nonnull %0) #3
store i32 0, i32* %i, align 4
%1 = bitcast i16* %s to i8*
call void @llvm.lifetime.start.p0i8(i64 2, i8* nonnull %1) #3
%call = call i32 (i32*, ...) bitcast (i32 (...)* @goo to i32 (i32*, ...)*)(i32* nonnull %i) #3
%2 = load i32, i32* %i, align 4
%storemerge6 = trunc i32 %2 to i16
store i16 %storemerge6, i16* %s, align 2
%conv17 = and i32 %2, 65472
%cmp8 = icmp eq i32 %conv17, 0
br i1 %cmp8, label %for.body.lr.ph, label %for.end12
for.body.lr.ph: ; preds = %entry
br label %for.body
for.body: ; preds = %for.body.lr.ph, %for.inc9
%storemerge.in9 = phi i32 [ %2, %for.body.lr.ph ], [ %add, %for.inc9 ]
%conv52 = and i32 %storemerge.in9, 255
%cmp63 = icmp ult i32 %2, %conv52
br i1 %cmp63, label %for.body8.lr.ph, label %for.inc9
for.body8.lr.ph: ; preds = %for.body
%conv3 = trunc i32 %storemerge.in9 to i8
%.promoted = load i32, i32* getelementptr inbounds ([192 x [192 x i32]], [192 x [192 x i32]]* @a, i64 0, i64 0, i64 0), align 16
br label %for.body8
for.body8: ; preds = %for.body8.lr.ph, %for.body8
%inc5 = phi i32 [ %.promoted, %for.body8.lr.ph ], [ %inc, %for.body8 ]
%c.04 = phi i8 [ %conv3, %for.body8.lr.ph ], [ %dec, %for.body8 ]
%inc = add i32 %inc5, 1
%dec = add i8 %c.04, -1
%conv5 = zext i8 %dec to i32
%cmp6 = icmp ult i32 %2, %conv5
br i1 %cmp6, label %for.body8, label %for.cond4.for.inc9_crit_edge
for.cond4.for.inc9_crit_edge: ; preds = %for.body8
%inc.lcssa = phi i32 [ %inc, %for.body8 ]
store i32 %inc.lcssa, i32* getelementptr inbounds ([192 x [192 x i32]], [192 x [192 x i32]]* @a, i64 0, i64 0, i64 0), align 16
br label %for.inc9
for.inc9: ; preds = %for.cond4.for.inc9_crit_edge, %for.body
%conv10 = and i32 %storemerge.in9, 65535
%add = add nuw nsw i32 %conv10, 1
%conv1 = and i32 %add, 65472
%cmp = icmp eq i32 %conv1, 0
br i1 %cmp, label %for.body, label %for.cond.for.end12_crit_edge
for.cond.for.end12_crit_edge: ; preds = %for.inc9
%add.lcssa = phi i32 [ %add, %for.inc9 ]
%storemerge = trunc i32 %add.lcssa to i16
store i16 %storemerge, i16* %s, align 2
br label %for.end12
for.end12: ; preds = %for.cond.for.end12_crit_edge, %entry
%call13 = call i32 (i16*, ...) bitcast (i32 (...)* @foo to i32 (i16*, ...)*)(i16* nonnull %s) #3
call void @llvm.lifetime.end.p0i8(i64 2, i8* nonnull %1) #3
call void @llvm.lifetime.end.p0i8(i64 4, i8* nonnull %0) #3
ret i32 0
}
; Function Attrs: argmemonly nounwind
declare void @llvm.lifetime.start.p0i8(i64, i8* nocapture) #1
declare i32 @goo(...) local_unnamed_addr #2
declare i32 @foo(...) local_unnamed_addr #2
; Function Attrs: argmemonly nounwind
declare void @llvm.lifetime.end.p0i8(i64, i8* nocapture) #1

test/Transforms/LoopVectorize/pr36032.ll

; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
; RUN: opt -loop-vectorize -S < %s | FileCheck %s
; The test checks that there is no assert caused by issue described in PR36032
target datalayout = "e-m:e-i8:8:32-i16:16:32-i64:64-i128:128-n32:64-S128"
target triple = "aarch64--"
%struct.anon = type { i8 }
@c = local_unnamed_addr global [6 x i8] zeroinitializer, align 1
@b = internal global %struct.anon zeroinitializer, align 1
; Function Attrs: noreturn nounwind
define void @_Z1dv() local_unnamed_addr #0 {
; CHECK-LABEL: @_Z1dv(
; CHECK-NEXT: entry:
; CHECK-NEXT: [[CALL:%.*]] = tail call i8* @"_ZN3$_01aEv"(%struct.anon* nonnull @b)
; CHECK-NEXT: [[SCEVGEP1:%.*]] = getelementptr i8, i8* [[CALL]], i64 4
; CHECK-NEXT: br label [[FOR_COND:%.*]]
; CHECK: for.cond:
; CHECK-NEXT: [[F_0:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[ADD5:%.*]], [[FOR_COND_CLEANUP:%.*]] ]
; CHECK-NEXT: [[G_0:%.*]] = phi i32 [ undef, [[ENTRY]] ], [ [[G_1_LCSSA:%.*]], [[FOR_COND_CLEANUP]] ]
; CHECK-NEXT: [[CMP12:%.*]] = icmp ult i32 [[G_0]], 4
; CHECK-NEXT: [[CONV:%.*]] = and i32 [[F_0]], 65535
; CHECK-NEXT: br i1 [[CMP12]], label [[FOR_BODY_LR_PH:%.*]], label [[FOR_COND_CLEANUP]]
; CHECK: for.body.lr.ph:
; CHECK-NEXT: [[TMP0:%.*]] = zext i32 [[G_0]] to i64
; CHECK-NEXT: [[TMP1:%.*]] = sub i64 4, [[TMP0]]
; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP1]], 4
; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_SCEVCHECK:%.*]]
; CHECK: vector.scevcheck:
; CHECK-NEXT: [[TMP2:%.*]] = sub i64 3, [[TMP0]]
; CHECK-NEXT: [[TMP3:%.*]] = add i32 [[G_0]], [[CONV]]
; CHECK-NEXT: [[TMP4:%.*]] = trunc i64 [[TMP2]] to i32
; CHECK-NEXT: [[MUL:%.*]] = call { i32, i1 } @llvm.umul.with.overflow.i32(i32 1, i32 [[TMP4]])
; CHECK-NEXT: [[MUL_RESULT:%.*]] = extractvalue { i32, i1 } [[MUL]], 0
; CHECK-NEXT: [[MUL_OVERFLOW:%.*]] = extractvalue { i32, i1 } [[MUL]], 1
; CHECK-NEXT: [[TMP5:%.*]] = add i32 [[TMP3]], [[MUL_RESULT]]
; CHECK-NEXT: [[TMP6:%.*]] = sub i32 [[TMP3]], [[MUL_RESULT]]
; CHECK-NEXT: [[TMP7:%.*]] = icmp ugt i32 [[TMP6]], [[TMP3]]
; CHECK-NEXT: [[TMP8:%.*]] = icmp ult i32 [[TMP5]], [[TMP3]]
; CHECK-NEXT: [[TMP9:%.*]] = select i1 false, i1 [[TMP7]], i1 [[TMP8]]
; CHECK-NEXT: [[TMP10:%.*]] = icmp ugt i64 [[TMP2]], 4294967295
; CHECK-NEXT: [[TMP11:%.*]] = or i1 [[TMP9]], [[TMP10]]
; CHECK-NEXT: [[TMP12:%.*]] = or i1 [[TMP11]], [[MUL_OVERFLOW]]
; CHECK-NEXT: [[TMP13:%.*]] = or i1 false, [[TMP12]]
; CHECK-NEXT: br i1 [[TMP13]], label [[SCALAR_PH]], label [[VECTOR_MEMCHECK:%.*]]
; CHECK: vector.memcheck:
; CHECK-NEXT: [[SCEVGEP:%.*]] = getelementptr i8, i8* [[CALL]], i64 [[TMP0]]
; CHECK-NEXT: [[TMP14:%.*]] = add i32 [[G_0]], [[CONV]]
; CHECK-NEXT: [[TMP15:%.*]] = zext i32 [[TMP14]] to i64
; CHECK-NEXT: [[SCEVGEP2:%.*]] = getelementptr [6 x i8], [6 x i8]* @c, i64 0, i64 [[TMP15]]
; CHECK-NEXT: [[TMP16:%.*]] = sub i64 [[TMP15]], [[TMP0]]
; CHECK-NEXT: [[SCEVGEP3:%.*]] = getelementptr i8, i8* getelementptr inbounds ([6 x i8], [6 x i8]* @c, i64 0, i64 4), i64 [[TMP16]]
; CHECK-NEXT: [[BOUND0:%.*]] = icmp ult i8* [[SCEVGEP]], [[SCEVGEP3]]
; CHECK-NEXT: [[BOUND1:%.*]] = icmp ult i8* [[SCEVGEP2]], [[SCEVGEP1]]
; CHECK-NEXT: [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]
; CHECK-NEXT: [[MEMCHECK_CONFLICT:%.*]] = and i1 [[FOUND_CONFLICT]], true
; CHECK-NEXT: br i1 [[MEMCHECK_CONFLICT]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
; CHECK: vector.ph:
; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[TMP1]], 4
; CHECK-NEXT: [[N_VEC:%.*]] = sub i64 [[TMP1]], [[N_MOD_VF]]
; CHECK-NEXT: [[IND_END:%.*]] = add i64 [[TMP0]], [[N_VEC]]
; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]
; CHECK: vector.body:
; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT: [[OFFSET_IDX:%.*]] = add i64 [[TMP0]], [[INDEX]]
; CHECK-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i64> undef, i64 [[OFFSET_IDX]], i32 0
; CHECK-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i64> [[BROADCAST_SPLATINSERT]], <4 x i64> undef, <4 x i32> zeroinitializer
; CHECK-NEXT: [[INDUCTION:%.*]] = add <4 x i64> [[BROADCAST_SPLAT]], <i64 0, i64 1, i64 2, i64 3>
; CHECK-NEXT: [[TMP17:%.*]] = add i64 [[OFFSET_IDX]], 0
; CHECK-NEXT: [[OFFSET_IDX4:%.*]] = add i64 [[TMP0]], [[INDEX]]
; CHECK-NEXT: [[TMP18:%.*]] = trunc i64 [[OFFSET_IDX4]] to i32
; CHECK-NEXT: [[BROADCAST_SPLATINSERT5:%.*]] = insertelement <4 x i32> undef, i32 [[TMP18]], i32 0
; CHECK-NEXT: [[BROADCAST_SPLAT6:%.*]] = shufflevector <4 x i32> [[BROADCAST_SPLATINSERT5]], <4 x i32> undef, <4 x i32> zeroinitializer
; CHECK-NEXT: [[INDUCTION7:%.*]] = add <4 x i32> [[BROADCAST_SPLAT6]], <i32 0, i32 1, i32 2, i32 3>
; CHECK-NEXT: [[TMP19:%.*]] = add i32 [[TMP18]], 0
; CHECK-NEXT: [[TMP20:%.*]] = add i32 [[CONV]], [[TMP19]]
; CHECK-NEXT: [[TMP21:%.*]] = zext i32 [[TMP20]] to i64
; CHECK-NEXT: [[TMP22:%.*]] = getelementptr inbounds [6 x i8], [6 x i8]* @c, i64 0, i64 [[TMP21]]
; CHECK-NEXT: [[TMP23:%.*]] = getelementptr i8, i8* [[TMP22]], i32 0
; CHECK-NEXT: [[TMP24:%.*]] = bitcast i8* [[TMP23]] to <4 x i8>*
; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <4 x i8>, <4 x i8>* [[TMP24]], align 1, !alias.scope !0
; CHECK-NEXT: [[TMP25:%.*]] = getelementptr inbounds i8, i8* [[CALL]], i64 [[TMP17]]
; CHECK-NEXT: [[TMP26:%.*]] = getelementptr i8, i8* [[TMP25]], i32 0
; CHECK-NEXT: [[TMP27:%.*]] = bitcast i8* [[TMP26]] to <4 x i8>*
; CHECK-NEXT: store <4 x i8> [[WIDE_LOAD]], <4 x i8>* [[TMP27]], align 1, !alias.scope !3, !noalias !0
; CHECK-NEXT: [[INDEX_NEXT]] = add i64 [[INDEX]], 4
; CHECK-NEXT: [[TMP28:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-NEXT: br i1 [[TMP28]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop !5
; CHECK: middle.block:
; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[TMP1]], [[N_VEC]]
; CHECK-NEXT: br i1 [[CMP_N]], label [[FOR_COND_CLEANUP_LOOPEXIT:%.*]], label [[SCALAR_PH]]
; CHECK: scalar.ph:
; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[TMP0]], [[FOR_BODY_LR_PH]] ], [ [[TMP0]], [[VECTOR_SCEVCHECK]] ], [ [[TMP0]], [[VECTOR_MEMCHECK]] ]
; CHECK-NEXT: br label [[FOR_BODY:%.*]]
; CHECK: for.cond.cleanup.loopexit:
; CHECK-NEXT: br label [[FOR_COND_CLEANUP]]
; CHECK: for.cond.cleanup:
; CHECK-NEXT: [[G_1_LCSSA]] = phi i32 [ [[G_0]], [[FOR_COND]] ], [ 4, [[FOR_COND_CLEANUP_LOOPEXIT]] ]
; CHECK-NEXT: [[ADD5]] = add nuw nsw i32 [[CONV]], 4
; CHECK-NEXT: br label [[FOR_COND]]
; CHECK: for.body:
; CHECK-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
; CHECK-NEXT: [[TMP29:%.*]] = trunc i64 [[INDVARS_IV]] to i32
; CHECK-NEXT: [[ADD:%.*]] = add i32 [[CONV]], [[TMP29]]
; CHECK-NEXT: [[IDXPROM:%.*]] = zext i32 [[ADD]] to i64
; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [6 x i8], [6 x i8]* @c, i64 0, i64 [[IDXPROM]]
; CHECK-NEXT: [[TMP30:%.*]] = load i8, i8* [[ARRAYIDX]], align 1
; CHECK-NEXT: [[ARRAYIDX3:%.*]] = getelementptr inbounds i8, i8* [[CALL]], i64 [[INDVARS_IV]]
; CHECK-NEXT: store i8 [[TMP30]], i8* [[ARRAYIDX3]], align 1
; CHECK-NEXT: [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
; CHECK-NEXT: [[EXITCOND:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], 4
; CHECK-NEXT: br i1 [[EXITCOND]], label [[FOR_COND_CLEANUP_LOOPEXIT]], label [[FOR_BODY]], !llvm.loop !7
;
entry:
%call = tail call i8* @"_ZN3$_01aEv"(%struct.anon* nonnull @b) #2
br label %for.cond
for.cond: ; preds = %for.cond.cleanup, %entry
%f.0 = phi i32 [ 0, %entry ], [ %add5, %for.cond.cleanup ]
%g.0 = phi i32 [ undef, %entry ], [ %g.1.lcssa, %for.cond.cleanup ]
%cmp12 = icmp ult i32 %g.0, 4
%conv = and i32 %f.0, 65535
br i1 %cmp12, label %for.body.lr.ph, label %for.cond.cleanup
for.body.lr.ph: ; preds = %for.cond
%0 = zext i32 %g.0 to i64
br label %for.body
for.cond.cleanup.loopexit: ; preds = %for.body
br label %for.cond.cleanup
for.cond.cleanup: ; preds = %for.cond.cleanup.loopexit, %for.cond
%g.1.lcssa = phi i32 [ %g.0, %for.cond ], [ 4, %for.cond.cleanup.loopexit ]
%add5 = add nuw nsw i32 %conv, 4
br label %for.cond
for.body: ; preds = %for.body, %for.body.lr.ph
%indvars.iv = phi i64 [ %0, %for.body.lr.ph ], [ %indvars.iv.next, %for.body ]
%1 = trunc i64 %indvars.iv to i32
%add = add i32 %conv, %1
%idxprom = zext i32 %add to i64
%arrayidx = getelementptr inbounds [6 x i8], [6 x i8]* @c, i64 0, i64 %idxprom
%2 = load i8, i8* %arrayidx, align 1
%arrayidx3 = getelementptr inbounds i8, i8* %call, i64 %indvars.iv
store i8 %2, i8* %arrayidx3, align 1
%indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
%exitcond = icmp eq i64 %indvars.iv.next, 4
br i1 %exitcond, label %for.cond.cleanup.loopexit, label %for.body
}
declare i8* @"_ZN3$_01aEv"(%struct.anon*) local_unnamed_addr #1