Skip to content

[SCEVExpander] Preserve gep nuw during expansion #102133

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 1 commit into from
Oct 2, 2024
Merged

[SCEVExpander] Preserve gep nuw during expansion #102133

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion llvm/include/llvm/Transforms/Utils/ScalarEvolutionExpander.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -451,7 +451,7 @@ class SCEVExpander : public SCEVVisitor<SCEVExpander, Value *> {

/// Expand a SCEVAddExpr with a pointer type into a GEP instead of using
/// ptrtoint+arithmetic+inttoptr.
Value *expandAddToGEP(const SCEV *Op, Value *V);
Value *expandAddToGEP(const SCEV *Op, Value *V, SCEV::NoWrapFlags Flags);

/// Find a previous Value in ExprValueMap for expand.
/// DropPoisonGeneratingInsts is populated with instructions for which
Expand Down
16 changes: 10 additions & 6 deletions llvm/lib/Transforms/Utils/ScalarEvolutionExpander.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -352,16 +352,19 @@ Value *SCEVExpander::InsertBinop(Instruction::BinaryOps Opcode,
/// loop-invariant portions of expressions, after considering what
/// can be folded using target addressing modes.
///
Value *SCEVExpander::expandAddToGEP(const SCEV *Offset, Value *V) {
Value *SCEVExpander::expandAddToGEP(const SCEV *Offset, Value *V,
SCEV::NoWrapFlags Flags) {
assert(!isa<Instruction>(V) ||
SE.DT.dominates(cast<Instruction>(V), &*Builder.GetInsertPoint()));

Value *Idx = expand(Offset);
GEPNoWrapFlags NW = (Flags & SCEV::FlagNUW) ? GEPNoWrapFlags::noUnsignedWrap()
: GEPNoWrapFlags::none();

// Fold a GEP with constant operands.
if (Constant *CLHS = dyn_cast<Constant>(V))
if (Constant *CRHS = dyn_cast<Constant>(Idx))
return Builder.CreatePtrAdd(CLHS, CRHS);
return Builder.CreatePtrAdd(CLHS, CRHS, "", NW);
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Don't we need to merge the flags if we find an existing GEP in the loop below? In particular, is it safe to reuse a GEP with flags when expanding one without?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Good point. This looks like a pre-existing issue, as we could also be reusing a gep inbounds.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Finally got around to this, the fix is here: #109293


// Do a quick scan to see if we have this GEP nearby. If so, reuse it.
unsigned ScanLimit = 6;
Expand All @@ -380,7 +383,7 @@ Value *SCEVExpander::expandAddToGEP(const SCEV *Offset, Value *V) {
GEP->getSourceElementType() == Builder.getInt8Ty() &&
GEP->getOperand(1) == Idx) {
rememberFlags(GEP);
GEP->setNoWrapFlags(GEPNoWrapFlags::none());
GEP->setNoWrapFlags(GEP->getNoWrapFlags() & NW);
return &*IP;
}
}
Expand All @@ -402,7 +405,7 @@ Value *SCEVExpander::expandAddToGEP(const SCEV *Offset, Value *V) {
}

// Emit a GEP.
return Builder.CreatePtrAdd(V, Idx, "scevgep");
return Builder.CreatePtrAdd(V, Idx, "scevgep", NW);
}

/// PickMostRelevantLoop - Given two loops pick the one that's most relevant for
Expand Down Expand Up @@ -549,7 +552,7 @@ Value *SCEVExpander::visitAddExpr(const SCEVAddExpr *S) {
X = SE.getSCEV(U->getValue());
NewOps.push_back(X);
}
Sum = expandAddToGEP(SE.getAddExpr(NewOps), Sum);
Sum = expandAddToGEP(SE.getAddExpr(NewOps), Sum, S->getNoWrapFlags());
} else if (Op->isNonConstantNegative()) {
// Instead of doing a negate and add, just do a subtract.
Value *W = expand(SE.getNegativeSCEV(Op));
Expand Down Expand Up @@ -1251,7 +1254,8 @@ Value *SCEVExpander::visitAddRecExpr(const SCEVAddRecExpr *S) {
if (!S->getStart()->isZero()) {
if (isa<PointerType>(S->getType())) {
Value *StartV = expand(SE.getPointerBase(S));
return expandAddToGEP(SE.removePointerBase(S), StartV);
return expandAddToGEP(SE.removePointerBase(S), StartV,
S->getNoWrapFlags(SCEV::FlagNUW));
}

SmallVector<const SCEV *, 4> NewOps(S->operands());
Expand Down
2 changes: 1 addition & 1 deletion llvm/test/Analysis/ScalarEvolution/scev-expander-reuse-gep.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -3,7 +3,7 @@
; RUN: opt -mtriple=i386-apple-macosx10.12.0 < %s -loop-reduce -S | FileCheck %s

; CHECK: %ptr4.ptr1 = select i1 %cmp.i, ptr %ptr4, ptr %ptr1
; CHECK-NEXT: %scevgep = getelementptr i8, ptr %ptr4.ptr1, i32 1
; CHECK-NEXT: %scevgep = getelementptr nuw i8, ptr %ptr4.ptr1, i32 1
; CHECK-NEXT: br label %while.cond.i

target datalayout = "e-m:o-p:32:32-f64:32:64-f80:128-n8:16:32-S128"
Expand Down
2 changes: 1 addition & 1 deletion llvm/test/Transforms/IndVarSimplify/D108043.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -9,7 +9,7 @@ define internal fastcc void @func_2() unnamed_addr {
; CHECK-NEXT: lbl_2898.preheader:
; CHECK-NEXT: br label [[LBL_2898:%.*]]
; CHECK: lbl_2898.loopexit:
; CHECK-NEXT: store ptr getelementptr inbounds (i8, ptr @g_2168, i64 76), ptr @g_1150, align 8
; CHECK-NEXT: store ptr getelementptr inbounds nuw (i8, ptr @g_2168, i64 76), ptr @g_1150, align 8
; CHECK-NEXT: br label [[LBL_2898]]
; CHECK: lbl_2898:
; CHECK-NEXT: br label [[FOR_COND884:%.*]]
Expand Down
6 changes: 3 additions & 3 deletions llvm/test/Transforms/IndVarSimplify/lftr-dead-ivs.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -112,7 +112,7 @@ define void @dom_store_preinc() #0 {
; CHECK-NEXT: [[P_0:%.*]] = phi ptr [ @data, [[ENTRY:%.*]] ], [ [[TMP3:%.*]], [[LOOP]] ]
; CHECK-NEXT: store volatile i8 0, ptr [[P_0]], align 1
; CHECK-NEXT: [[TMP3]] = getelementptr inbounds i8, ptr [[P_0]], i64 1
; CHECK-NEXT: [[EXITCOND:%.*]] = icmp ne ptr [[P_0]], getelementptr (i8, ptr @data, i64 245)
; CHECK-NEXT: [[EXITCOND:%.*]] = icmp ne ptr [[P_0]], getelementptr nuw (i8, ptr @data, i64 245)
; CHECK-NEXT: br i1 [[EXITCOND]], label [[LOOP]], label [[EXIT:%.*]]
; CHECK: exit:
; CHECK-NEXT: ret void
Expand Down Expand Up @@ -141,7 +141,7 @@ define void @dom_store_postinc() #0 {
; CHECK-NEXT: [[P_0:%.*]] = phi ptr [ @data, [[ENTRY:%.*]] ], [ [[TMP3:%.*]], [[LOOP]] ]
; CHECK-NEXT: [[TMP3]] = getelementptr inbounds i8, ptr [[P_0]], i64 1
; CHECK-NEXT: store volatile i8 0, ptr [[TMP3]], align 1
; CHECK-NEXT: [[EXITCOND:%.*]] = icmp ne ptr [[TMP3]], getelementptr (i8, ptr @data, i64 246)
; CHECK-NEXT: [[EXITCOND:%.*]] = icmp ne ptr [[TMP3]], getelementptr nuw (i8, ptr @data, i64 246)
; CHECK-NEXT: br i1 [[EXITCOND]], label [[LOOP]], label [[EXIT:%.*]]
; CHECK: exit:
; CHECK-NEXT: ret void
Expand Down Expand Up @@ -170,7 +170,7 @@ define i8 @dom_load() #0 {
; CHECK-NEXT: [[P_0:%.*]] = phi ptr [ @data, [[ENTRY:%.*]] ], [ [[TMP3:%.*]], [[LOOP]] ]
; CHECK-NEXT: [[TMP3]] = getelementptr inbounds i8, ptr [[P_0]], i64 1
; CHECK-NEXT: [[V:%.*]] = load i8, ptr [[TMP3]], align 1
; CHECK-NEXT: [[EXITCOND:%.*]] = icmp ne ptr [[TMP3]], getelementptr (i8, ptr @data, i64 246)
; CHECK-NEXT: [[EXITCOND:%.*]] = icmp ne ptr [[TMP3]], getelementptr nuw (i8, ptr @data, i64 246)
; CHECK-NEXT: br i1 [[EXITCOND]], label [[LOOP]], label [[EXIT:%.*]]
; CHECK: exit:
; CHECK-NEXT: [[V_LCSSA:%.*]] = phi i8 [ [[V]], [[LOOP]] ]
Expand Down
2 changes: 1 addition & 1 deletion llvm/test/Transforms/IndVarSimplify/lftr.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -196,7 +196,7 @@ define void @test_zext(ptr %a) #0 {
; CHECK-NEXT: [[T2:%.*]] = load i8, ptr [[DOT0]], align 1
; CHECK-NEXT: [[T3]] = getelementptr inbounds i8, ptr [[P_0]], i64 1
; CHECK-NEXT: store i8 [[T2]], ptr [[P_0]], align 1
; CHECK-NEXT: [[EXITCOND:%.*]] = icmp ne ptr [[P_0]], getelementptr inbounds (i8, ptr @data, i64 239)
; CHECK-NEXT: [[EXITCOND:%.*]] = icmp ne ptr [[P_0]], getelementptr inbounds nuw (i8, ptr @data, i64 239)
; CHECK-NEXT: br i1 [[EXITCOND]], label [[LOOP]], label [[EXIT:%.*]]
; CHECK: exit:
; CHECK-NEXT: ret void
Expand Down
2 changes: 1 addition & 1 deletion llvm/test/Transforms/LoopIdiom/basic-address-space.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -13,7 +13,7 @@ define void @test10(ptr addrspace(2) %X) nounwind ssp {
; CHECK: bb.nph:
; CHECK-NEXT: [[I_04:%.*]] = phi i16 [ 0, [[ENTRY:%.*]] ], [ [[INC12:%.*]], [[FOR_INC10:%.*]] ]
; CHECK-NEXT: [[TMP0:%.*]] = mul nuw nsw i16 [[I_04]], 100
; CHECK-NEXT: [[SCEVGEP:%.*]] = getelementptr i8, ptr addrspace(2) [[X]], i16 [[TMP0]]
; CHECK-NEXT: [[SCEVGEP:%.*]] = getelementptr nuw i8, ptr addrspace(2) [[X]], i16 [[TMP0]]
; CHECK-NEXT: br label [[FOR_BODY5:%.*]]
; CHECK: for.body5:
; CHECK-NEXT: [[J_02:%.*]] = phi i16 [ 0, [[BB_NPH]] ], [ [[INC:%.*]], [[FOR_BODY5]] ]
Expand Down
4 changes: 2 additions & 2 deletions llvm/test/Transforms/LoopIdiom/basic.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -481,7 +481,7 @@ define void @test10(ptr %X) nounwind ssp {
; CHECK-NEXT: [[INDVAR:%.*]] = phi i64 [ [[INDVAR_NEXT:%.*]], [[FOR_INC10:%.*]] ], [ 0, [[ENTRY:%.*]] ]
; CHECK-NEXT: [[I_04:%.*]] = phi i32 [ 0, [[ENTRY]] ], [ [[INC12:%.*]], [[FOR_INC10]] ]
; CHECK-NEXT: [[TMP0:%.*]] = mul nuw nsw i64 [[INDVAR]], 100
; CHECK-NEXT: [[SCEVGEP:%.*]] = getelementptr i8, ptr [[X]], i64 [[TMP0]]
; CHECK-NEXT: [[SCEVGEP:%.*]] = getelementptr nuw i8, ptr [[X]], i64 [[TMP0]]
; CHECK-NEXT: br label [[FOR_BODY5:%.*]]
; CHECK: for.body5:
; CHECK-NEXT: [[J_02:%.*]] = phi i32 [ 0, [[BB_NPH]] ], [ [[INC:%.*]], [[FOR_BODY5]] ]
Expand Down Expand Up @@ -682,7 +682,7 @@ define void @PR14241(ptr %s, i64 %size) {
; CHECK-NEXT: entry:
; CHECK-NEXT: [[END_IDX:%.*]] = add i64 [[SIZE:%.*]], -1
; CHECK-NEXT: [[END_PTR:%.*]] = getelementptr inbounds i32, ptr [[S:%.*]], i64 [[END_IDX]]
; CHECK-NEXT: [[SCEVGEP:%.*]] = getelementptr i8, ptr [[S]], i64 4
; CHECK-NEXT: [[SCEVGEP:%.*]] = getelementptr nuw i8, ptr [[S]], i64 4
; CHECK-NEXT: [[TMP0:%.*]] = shl i64 [[SIZE]], 2
; CHECK-NEXT: [[TMP1:%.*]] = add i64 [[TMP0]], -8
; CHECK-NEXT: [[TMP2:%.*]] = lshr i64 [[TMP1]], 2
Expand Down
4 changes: 2 additions & 2 deletions llvm/test/Transforms/LoopIdiom/memset-pr52104.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -6,12 +6,12 @@ define void @f(ptr nocapture nonnull align 4 dereferenceable(20) %0, i32 %1) loc
; CHECK-NEXT: [[TMP3:%.*]] = trunc i32 [[TMP1:%.*]] to i2
; CHECK-NEXT: [[TMP4:%.*]] = zext i2 [[TMP3]] to i64
; CHECK-NEXT: [[TMP5:%.*]] = shl nuw nsw i64 [[TMP4]], 2
; CHECK-NEXT: [[UGLYGEP:%.*]] = getelementptr i8, ptr [[TMP0:%.*]], i64 [[TMP5]]
; CHECK-NEXT: [[SCEVGEP:%.*]] = getelementptr nuw i8, ptr [[TMP0:%.*]], i64 [[TMP5]]
; CHECK-NEXT: [[TMP6:%.*]] = sub i2 -1, [[TMP3]]
; CHECK-NEXT: [[TMP7:%.*]] = zext i2 [[TMP6]] to i64
; CHECK-NEXT: [[TMP8:%.*]] = shl nuw nsw i64 [[TMP7]], 2
; CHECK-NEXT: [[TMP9:%.*]] = add nuw nsw i64 [[TMP8]], 4
; CHECK-NEXT: call void @llvm.memset.p0.i64(ptr align 4 [[UGLYGEP]], i8 0, i64 [[TMP9]], i1 false)
; CHECK-NEXT: call void @llvm.memset.p0.i64(ptr align 4 [[SCEVGEP]], i8 0, i64 [[TMP9]], i1 false)
; CHECK-NEXT: br label [[TMP10:%.*]]
; CHECK: 10:
; CHECK-NEXT: [[TMP11:%.*]] = phi i32 [ [[TMP15:%.*]], [[TMP10]] ], [ [[TMP1]], [[TMP2:%.*]] ]
Expand Down
2 changes: 1 addition & 1 deletion llvm/test/Transforms/LoopStrengthReduce/2011-12-19-PostincQuadratic.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -16,7 +16,7 @@ define void @vb() nounwind {
; CHECK-NEXT: for.cond.preheader:
; CHECK-NEXT: br label [[FOR_BODY7:%.*]]
; CHECK: for.body7:
; CHECK-NEXT: [[LSR_IV1:%.*]] = phi ptr [ [[SCEVGEP:%.*]], [[FOR_BODY7]] ], [ getelementptr inbounds (i8, ptr @b, i32 4), [[FOR_COND_PREHEADER:%.*]] ]
; CHECK-NEXT: [[LSR_IV1:%.*]] = phi ptr [ [[SCEVGEP:%.*]], [[FOR_BODY7]] ], [ getelementptr inbounds nuw (i8, ptr @b, i32 4), [[FOR_COND_PREHEADER:%.*]] ]
; CHECK-NEXT: [[LSR_IV:%.*]] = phi i32 [ [[LSR_IV_NEXT:%.*]], [[FOR_BODY7]] ], [ 8, [[FOR_COND_PREHEADER]] ]
; CHECK-NEXT: [[INDVARS_IV77:%.*]] = phi i32 [ [[INDVARS_IV_NEXT78:%.*]], [[FOR_BODY7]] ], [ 1, [[FOR_COND_PREHEADER]] ]
; CHECK-NEXT: [[INDVARS_IV_NEXT78]] = add i32 [[INDVARS_IV77]], 1
Expand Down
2 changes: 1 addition & 1 deletion llvm/test/Transforms/LoopStrengthReduce/X86/2012-01-13-phielim.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -96,7 +96,7 @@ define void @test2(i32 %n) nounwind uwtable {
; CHECK-NEXT: br label [[FOR_COND468:%.*]]
; CHECK: for.cond468:
; CHECK-NEXT: [[LSR_IV1:%.*]] = phi i32 [ 1, [[FOR_COND468_PREHEADER]] ], [ [[LSR_IV_NEXT:%.*]], [[IF_THEN477:%.*]] ]
; CHECK-NEXT: [[LSR_IV:%.*]] = phi ptr [ getelementptr inbounds (i8, ptr @tags, i64 8), [[FOR_COND468_PREHEADER]] ], [ [[SCEVGEP:%.*]], [[IF_THEN477]] ]
; CHECK-NEXT: [[LSR_IV:%.*]] = phi ptr [ getelementptr inbounds nuw (i8, ptr @tags, i64 8), [[FOR_COND468_PREHEADER]] ], [ [[SCEVGEP:%.*]], [[IF_THEN477]] ]
; CHECK-NEXT: [[K_0:%.*]] = load i32, ptr [[LSR_IV]], align 4
; CHECK-NEXT: [[CMP469:%.*]] = icmp slt i32 [[LSR_IV1]], [[N:%.*]]
; CHECK-NEXT: br i1 [[CMP469]], label [[FOR_BODY471:%.*]], label [[FOR_INC498_PREHEADER:%.*]]
Expand Down
2 changes: 1 addition & 1 deletion llvm/test/Transforms/LoopStrengthReduce/post-inc-icmpzero.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -18,7 +18,7 @@ define void @_Z15IntegerToStringjjR7Vector2(i32 %i, i32 %radix, ptr nocapture %r
; CHECK-NEXT: [[ADD_PTR:%.*]] = getelementptr [33 x i16], ptr [[BUFFER]], i64 0, i64 33
; CHECK-NEXT: [[SUB_PTR_LHS_CAST:%.*]] = ptrtoint ptr [[ADD_PTR]] to i64
; CHECK-NEXT: [[SUB_PTR_RHS_CAST:%.*]] = ptrtoint ptr [[ADD_PTR]] to i64
; CHECK-NEXT: [[SCEVGEP3:%.*]] = getelementptr i8, ptr [[BUFFER]], i64 64
; CHECK-NEXT: [[SCEVGEP3:%.*]] = getelementptr nuw i8, ptr [[BUFFER]], i64 64
; CHECK-NEXT: br label [[DO_BODY:%.*]]
; CHECK: do.body:
; CHECK-NEXT: [[LSR_IV10:%.*]] = phi i64 [ [[LSR_IV_NEXT11:%.*]], [[DO_BODY]] ], [ -1, [[ENTRY:%.*]] ]
Expand Down
2 changes: 1 addition & 1 deletion llvm/test/Transforms/LoopVectorize/AArch64/sve-interleaved-accesses.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1446,7 +1446,7 @@ define void @PR34743(ptr %a, ptr %b, i64 %n) #1 {
; CHECK-NEXT: [[TMP5:%.*]] = and i64 [[TMP4]], -4
; CHECK-NEXT: [[TMP6:%.*]] = getelementptr i8, ptr [[B:%.*]], i64 [[TMP5]]
; CHECK-NEXT: [[SCEVGEP:%.*]] = getelementptr i8, ptr [[TMP6]], i64 4
; CHECK-NEXT: [[SCEVGEP1:%.*]] = getelementptr i8, ptr [[A]], i64 2
; CHECK-NEXT: [[SCEVGEP1:%.*]] = getelementptr nuw i8, ptr [[A]], i64 2
; CHECK-NEXT: [[TMP7:%.*]] = getelementptr i8, ptr [[A]], i64 [[TMP5]]
; CHECK-NEXT: [[SCEVGEP2:%.*]] = getelementptr i8, ptr [[TMP7]], i64 6
; CHECK-NEXT: [[BOUND0:%.*]] = icmp ult ptr [[B]], [[SCEVGEP2]]
Expand Down
10 changes: 5 additions & 5 deletions llvm/test/Transforms/LoopVectorize/induction.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1552,7 +1552,7 @@ define void @scalarize_induction_variable_04(ptr %a, ptr %p, i32 %n) {
; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP2]], 2
; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_MEMCHECK:%.*]]
; CHECK: vector.memcheck:
; CHECK-NEXT: [[SCEVGEP:%.*]] = getelementptr i8, ptr [[P:%.*]], i64 4
; CHECK-NEXT: [[SCEVGEP:%.*]] = getelementptr nuw i8, ptr [[P:%.*]], i64 4
; CHECK-NEXT: [[TMP3:%.*]] = add i32 [[N]], -1
; CHECK-NEXT: [[TMP4:%.*]] = zext i32 [[TMP3]] to i64
; CHECK-NEXT: [[TMP5:%.*]] = shl nuw nsw i64 [[TMP4]], 3
Expand Down Expand Up @@ -1617,7 +1617,7 @@ define void @scalarize_induction_variable_04(ptr %a, ptr %p, i32 %n) {
; IND-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp eq i32 [[TMP0]], 0
; IND-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_MEMCHECK:%.*]]
; IND: vector.memcheck:
; IND-NEXT: [[SCEVGEP:%.*]] = getelementptr i8, ptr [[P:%.*]], i64 4
; IND-NEXT: [[SCEVGEP:%.*]] = getelementptr nuw i8, ptr [[P:%.*]], i64 4
; IND-NEXT: [[TMP3:%.*]] = add i32 [[N]], -1
; IND-NEXT: [[TMP4:%.*]] = zext i32 [[TMP3]] to i64
; IND-NEXT: [[TMP5:%.*]] = shl nuw nsw i64 [[TMP4]], 3
Expand Down Expand Up @@ -1680,7 +1680,7 @@ define void @scalarize_induction_variable_04(ptr %a, ptr %p, i32 %n) {
; UNROLL-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP0]], 3
; UNROLL-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_MEMCHECK:%.*]]
; UNROLL: vector.memcheck:
; UNROLL-NEXT: [[SCEVGEP:%.*]] = getelementptr i8, ptr [[P:%.*]], i64 4
; UNROLL-NEXT: [[SCEVGEP:%.*]] = getelementptr nuw i8, ptr [[P:%.*]], i64 4
; UNROLL-NEXT: [[TMP3:%.*]] = add i32 [[N]], -1
; UNROLL-NEXT: [[TMP4:%.*]] = zext i32 [[TMP3]] to i64
; UNROLL-NEXT: [[TMP5:%.*]] = shl nuw nsw i64 [[TMP4]], 3
Expand Down Expand Up @@ -1757,7 +1757,7 @@ define void @scalarize_induction_variable_04(ptr %a, ptr %p, i32 %n) {
; UNROLL-NO-IC-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP2]], 4
; UNROLL-NO-IC-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_MEMCHECK:%.*]]
; UNROLL-NO-IC: vector.memcheck:
; UNROLL-NO-IC-NEXT: [[SCEVGEP:%.*]] = getelementptr i8, ptr [[P:%.*]], i64 4
; UNROLL-NO-IC-NEXT: [[SCEVGEP:%.*]] = getelementptr nuw i8, ptr [[P:%.*]], i64 4
; UNROLL-NO-IC-NEXT: [[TMP3:%.*]] = add i32 [[N]], -1
; UNROLL-NO-IC-NEXT: [[TMP4:%.*]] = zext i32 [[TMP3]] to i64
; UNROLL-NO-IC-NEXT: [[TMP5:%.*]] = shl nuw nsw i64 [[TMP4]], 3
Expand Down Expand Up @@ -1836,7 +1836,7 @@ define void @scalarize_induction_variable_04(ptr %a, ptr %p, i32 %n) {
; INTERLEAVE-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP0]], 8
; INTERLEAVE-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_MEMCHECK:%.*]]
; INTERLEAVE: vector.memcheck:
; INTERLEAVE-NEXT: [[SCEVGEP:%.*]] = getelementptr i8, ptr [[P:%.*]], i64 4
; INTERLEAVE-NEXT: [[SCEVGEP:%.*]] = getelementptr nuw i8, ptr [[P:%.*]], i64 4
; INTERLEAVE-NEXT: [[TMP3:%.*]] = add i32 [[N]], -1
; INTERLEAVE-NEXT: [[TMP4:%.*]] = zext i32 [[TMP3]] to i64
; INTERLEAVE-NEXT: [[TMP5:%.*]] = shl nuw nsw i64 [[TMP4]], 3
Expand Down
2 changes: 1 addition & 1 deletion llvm/test/Transforms/LoopVectorize/interleaved-accesses.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1478,7 +1478,7 @@ define void @PR34743(ptr %a, ptr %b, i64 %n) {
; CHECK-NEXT: [[TMP3:%.*]] = and i64 [[TMP2]], -4
; CHECK-NEXT: [[TMP4:%.*]] = getelementptr i8, ptr [[B:%.*]], i64 [[TMP3]]
; CHECK-NEXT: [[SCEVGEP:%.*]] = getelementptr i8, ptr [[TMP4]], i64 4
; CHECK-NEXT: [[SCEVGEP1:%.*]] = getelementptr i8, ptr [[A]], i64 2
; CHECK-NEXT: [[SCEVGEP1:%.*]] = getelementptr nuw i8, ptr [[A]], i64 2
; CHECK-NEXT: [[TMP5:%.*]] = getelementptr i8, ptr [[A]], i64 [[TMP3]]
; CHECK-NEXT: [[SCEVGEP2:%.*]] = getelementptr i8, ptr [[TMP5]], i64 6
; CHECK-NEXT: [[BOUND0:%.*]] = icmp ult ptr [[B]], [[SCEVGEP2]]
Expand Down
8 changes: 4 additions & 4 deletions llvm/test/Transforms/LoopVectorize/multiple-strides-vectorization.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -38,12 +38,12 @@ define void @Test(ptr nocapture %obj, i64 %z) #0 {
; CHECK-NEXT: [[I:%.*]] = phi i64 [ 0, [[TMP0:%.*]] ], [ [[I_NEXT:%.*]], [[DOTOUTER:%.*]] ]
; CHECK-NEXT: [[TMP3:%.*]] = shl nuw nsw i64 [[I]], 7
; CHECK-NEXT: [[TMP4:%.*]] = add i64 [[TMP3]], 256
; CHECK-NEXT: [[SCEVGEP:%.*]] = getelementptr i8, ptr [[OBJ]], i64 [[TMP4]]
; CHECK-NEXT: [[SCEVGEP:%.*]] = getelementptr nuw i8, ptr [[OBJ]], i64 [[TMP4]]
; CHECK-NEXT: [[TMP5:%.*]] = add i64 [[TMP2]], [[TMP3]]
; CHECK-NEXT: [[SCEVGEP1:%.*]] = getelementptr i8, ptr [[OBJ]], i64 [[TMP5]]
; CHECK-NEXT: [[TMP6:%.*]] = shl nuw nsw i64 [[I]], 2
; CHECK-NEXT: [[TMP7:%.*]] = add i64 [[TMP6]], 128
; CHECK-NEXT: [[SCEVGEP3:%.*]] = getelementptr i8, ptr [[OBJ]], i64 [[TMP7]]
; CHECK-NEXT: [[SCEVGEP3:%.*]] = getelementptr nuw i8, ptr [[OBJ]], i64 [[TMP7]]
; CHECK-NEXT: [[TMP8:%.*]] = add i64 [[TMP6]], 132
; CHECK-NEXT: [[SCEVGEP4:%.*]] = getelementptr i8, ptr [[OBJ]], i64 [[TMP8]]
; CHECK-NEXT: [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_S:%.*]], ptr [[OBJ]], i64 0, i32 1, i64 [[I]]
Expand Down Expand Up @@ -107,12 +107,12 @@ define void @Test(ptr nocapture %obj, i64 %z) #0 {
; CHECK-NEXT: br i1 [[EXITCOND_INNER]], label [[DOTOUTER]], label [[DOTINNER]], !llvm.loop [[LOOP11:![0-9]+]]
;
; CHECK-HOIST-LABEL: @Test(
; CHECK-HOIST-NEXT: [[SCEVGEP:%.*]] = getelementptr i8, ptr [[OBJ:%.*]], i64 256
; CHECK-HOIST-NEXT: [[SCEVGEP:%.*]] = getelementptr nuw i8, ptr [[OBJ:%.*]], i64 256
; CHECK-HOIST-NEXT: [[TMP1:%.*]] = shl i64 [[Z:%.*]], 2
; CHECK-HOIST-NEXT: [[TMP2:%.*]] = add i64 [[TMP1]], 4224
; CHECK-HOIST-NEXT: [[SCEVGEP1:%.*]] = getelementptr i8, ptr [[OBJ]], i64 [[TMP2]]
; CHECK-HOIST-NEXT: [[SCEVGEP2:%.*]] = getelementptr i8, ptr [[OBJ]], i64 [[TMP1]]
; CHECK-HOIST-NEXT: [[SCEVGEP3:%.*]] = getelementptr i8, ptr [[OBJ]], i64 128
; CHECK-HOIST-NEXT: [[SCEVGEP3:%.*]] = getelementptr nuw i8, ptr [[OBJ]], i64 128
; CHECK-HOIST-NEXT: br label [[DOTOUTER_PREHEADER:%.*]]
; CHECK-HOIST: .outer.preheader:
; CHECK-HOIST-NEXT: [[I:%.*]] = phi i64 [ 0, [[TMP0:%.*]] ], [ [[I_NEXT:%.*]], [[DOTOUTER:%.*]] ]
Expand Down
2 changes: 1 addition & 1 deletion llvm/test/Transforms/LoopVectorize/pr47343-expander-lcssa-after-cfg-update.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -39,7 +39,7 @@ define void @f() {
; CHECK: vector.memcheck:
; CHECK-NEXT: [[SCEVGEP:%.*]] = getelementptr i8, ptr [[TMP1]], i64 1
; CHECK-NEXT: [[BOUND0:%.*]] = icmp ult ptr @f.e, [[SCEVGEP]]
; CHECK-NEXT: [[BOUND1:%.*]] = icmp ult ptr [[TMP1]], getelementptr inbounds (i8, ptr @f.e, i64 4)
; CHECK-NEXT: [[BOUND1:%.*]] = icmp ult ptr [[TMP1]], getelementptr inbounds nuw (i8, ptr @f.e, i64 4)
; CHECK-NEXT: [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]
; CHECK-NEXT: br i1 [[FOUND_CONFLICT]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
; CHECK: vector.ph:
Expand Down
4 changes: 2 additions & 2 deletions llvm/test/Transforms/LoopVectorize/runtime-checks-difference.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -183,7 +183,7 @@ define void @nested_loop_outer_iv_addrec_invariant_in_inner1(ptr %a, ptr %b, i64

; CHECK: outer.header:
; CHECK: [[OUTER_IV_SHL_2:%.]] = shl i64 %outer.iv, 2
; CHECK-NEXT: [[A_GEP_UPPER:%.*]] = getelementptr i8, ptr %a, i64 [[OUTER_IV_SHL_2]]
; CHECK-NEXT: [[A_GEP_UPPER:%.*]] = getelementptr nuw i8, ptr %a, i64 [[OUTER_IV_SHL_2]]
; CHECK-NEXT: [[OUTER_IV_4:%.]] = add i64 [[OUTER_IV_SHL_2]], 4
; CHECK-NEXT: [[A_GEP_UPPER_4:%.*]] = getelementptr i8, ptr %a, i64 [[OUTER_IV_4]]
; CHECK: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], 4
Expand Down Expand Up @@ -233,7 +233,7 @@ define void @nested_loop_outer_iv_addrec_invariant_in_inner2(ptr %a, ptr %b, i64

; CHECK: outer.header:
; CHECK: [[OUTER_IV_SHL_2:%.]] = shl i64 %outer.iv, 2
; CHECK-NEXT: [[A_GEP_UPPER:%.*]] = getelementptr i8, ptr %a, i64 [[OUTER_IV_SHL_2]]
; CHECK-NEXT: [[A_GEP_UPPER:%.*]] = getelementptr nuw i8, ptr %a, i64 [[OUTER_IV_SHL_2]]
; CHECK-NEXT: [[OUTER_IV_4:%.]] = add i64 [[OUTER_IV_SHL_2]], 4
; CHECK-NEXT: [[A_GEP_UPPER_4:%.*]] = getelementptr i8, ptr %a, i64 [[OUTER_IV_4]]
; CHECK: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], 4
Expand Down
Loading
Loading