[llvm] [LV] Don't vectorize if trip count expansion may introduce UB. (PR #92177)

Florian Hahn via llvm-commits llvm-commits at lists.llvm.org
Wed Jun 26 07:39:57 PDT 2024


https://github.com/fhahn updated https://github.com/llvm/llvm-project/pull/92177

>From 1862db9dcb3f3c98f5924a5ce1b145905676e29f Mon Sep 17 00:00:00 2001
From: Florian Hahn <flo at fhahn.com>
Date: Tue, 14 May 2024 20:36:16 +0100
Subject: [PATCH] [LV] Don't vectorize if trip count expansion may introduce
 UB.

Introduce a utility to check if a SCEV expansion may introduce UB
(couldn't find a similar utility after a quick glance) and use to the
avoid vectorizing when expanding the trip count introduces UB.

Fixes https://github.com/llvm/llvm-project/issues/89958.
---
 .../Utils/ScalarEvolutionExpander.h           |   7 +
 .../Utils/ScalarEvolutionExpander.cpp         |  48 +++-
 .../Transforms/Vectorize/LoopVectorize.cpp    |   3 +
 .../trip-count-expansion-may-introduce-ub.ll  | 255 +++++++++++-------
 4 files changed, 220 insertions(+), 93 deletions(-)

diff --git a/llvm/include/llvm/Transforms/Utils/ScalarEvolutionExpander.h b/llvm/include/llvm/Transforms/Utils/ScalarEvolutionExpander.h
index 62c1e15a9a60e..51a3153f5a7d8 100644
--- a/llvm/include/llvm/Transforms/Utils/ScalarEvolutionExpander.h
+++ b/llvm/include/llvm/Transforms/Utils/ScalarEvolutionExpander.h
@@ -418,6 +418,13 @@ class SCEVExpander : public SCEVVisitor<SCEVExpander, Value *> {
   BasicBlock::iterator findInsertPointAfter(Instruction *I,
                                             Instruction *MustDominate) const;
 
+  /// If \p L contains exits which may execute conditionally and contain UDiv
+  /// expressions with divisors that can be 0, expanding \p BTC may introduce
+  /// new UB. In that case, rewrite UDiv(A, B) to  UDiv(A, UMAX(1, B)). If B is
+  /// 0, that exit cannot be taken.
+  static const SCEV *rewriteExpressionToRemoveUB(const SCEV *BTC, Loop *L,
+                                                 ScalarEvolution &SE);
+
 private:
   LLVMContext &getContext() const { return SE.getContext(); }
 
diff --git a/llvm/lib/Transforms/Utils/ScalarEvolutionExpander.cpp b/llvm/lib/Transforms/Utils/ScalarEvolutionExpander.cpp
index c7d758aa575e6..527dc8e448efe 100644
--- a/llvm/lib/Transforms/Utils/ScalarEvolutionExpander.cpp
+++ b/llvm/lib/Transforms/Utils/ScalarEvolutionExpander.cpp
@@ -2341,12 +2341,17 @@ struct SCEVFindUnsafe {
 };
 } // namespace
 
-bool SCEVExpander::isSafeToExpand(const SCEV *S) const {
+static bool isSafeToExpand(const SCEV *S, bool CanonicalMode,
+                           ScalarEvolution &SE) {
   SCEVFindUnsafe Search(SE, CanonicalMode);
   visitAll(S, Search);
   return !Search.IsUnsafe;
 }
 
+bool SCEVExpander::isSafeToExpand(const SCEV *S) const {
+  return ::isSafeToExpand(S, CanonicalMode, SE);
+}
+
 bool SCEVExpander::isSafeToExpandAt(const SCEV *S,
                                     const Instruction *InsertionPoint) const {
   if (!isSafeToExpand(S))
@@ -2369,6 +2374,47 @@ bool SCEVExpander::isSafeToExpandAt(const SCEV *S,
   return false;
 }
 
+namespace {
+struct SCEVUDivRewriter : public SCEVRewriteVisitor<SCEVUDivRewriter> {
+  SCEVUDivRewriter(ScalarEvolution &SE) : SCEVRewriteVisitor(SE) {}
+
+  static const SCEV *rewrite(const SCEV *Scev, ScalarEvolution &SE) {
+    SCEVUDivRewriter Rewriter(SE);
+    return Rewriter.visit(Scev);
+  }
+
+  const SCEV *visitUDivExpr(const SCEVUDivExpr *Expr) {
+    auto *LHS = visit(Expr->getLHS());
+    auto *RHS = visit(Expr->getRHS());
+    return SE.getUDivExpr(LHS, SE.getUMaxExpr(SE.getOne(RHS->getType()), RHS));
+  }
+};
+} // namespace
+
+const SCEV *SCEVExpander::rewriteExpressionToRemoveUB(const SCEV *Expr, Loop *L,
+                                                      ScalarEvolution &SE) {
+  SmallVector<BasicBlock *> Exiting;
+  L->getExitingBlocks(Exiting);
+  // Check if exit count for any exit that may execute unconditionally may in
+  // introduce UB. Note that we can skip checks in the header or if there's a
+  // single exit, as in those cases we know that the exit count will be
+  // evaluated in each loop iteration. There are other cases where the exiting
+  // block executes on each loop iteration, but we don't have a cheap way to
+  // check at the moment.
+
+  if (Exiting.size() == 1 || all_of(Exiting, [L, &SE](BasicBlock *E) {
+        if (L->getHeader() == E)
+          return true;
+        const SCEV *EC = SE.getExitCount(L, E);
+        if (isa<SCEVCouldNotCompute>(EC))
+          return true;
+        return ::isSafeToExpand(EC, true, SE);
+      }))
+    return Expr;
+
+  return SCEVUDivRewriter::rewrite(Expr, SE);
+}
+
 void SCEVExpanderCleaner::cleanup() {
   // Result is used, nothing to remove.
   if (ResultUsed)
diff --git a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
index 91cdd8e774c12..1ddf1df0e2f69 100644
--- a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
+++ b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
@@ -984,6 +984,9 @@ const SCEV *createTripCountSCEV(Type *IdxTy, PredicatedScalarEvolution &PSE,
   assert(!isa<SCEVCouldNotCompute>(BackedgeTakenCount) && "Invalid loop count");
 
   ScalarEvolution &SE = *PSE.getSE();
+  if (OrigLoop)
+    BackedgeTakenCount = SCEVExpander::rewriteExpressionToRemoveUB(
+        BackedgeTakenCount, OrigLoop, SE);
   return SE.getTripCountFromExitCount(BackedgeTakenCount, IdxTy, OrigLoop);
 }
 
diff --git a/llvm/test/Transforms/LoopVectorize/trip-count-expansion-may-introduce-ub.ll b/llvm/test/Transforms/LoopVectorize/trip-count-expansion-may-introduce-ub.ll
index bf687a31d8d17..c82797c5d808f 100644
--- a/llvm/test/Transforms/LoopVectorize/trip-count-expansion-may-introduce-ub.ll
+++ b/llvm/test/Transforms/LoopVectorize/trip-count-expansion-may-introduce-ub.ll
@@ -141,20 +141,21 @@ define i64 @multi_exit_2_exit_count_with_udiv_by_value_in_block_executed_uncondi
 ; CHECK-SAME: ptr [[A:%.*]], i64 [[N:%.*]]) {
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    [[SMAX:%.*]] = call i64 @llvm.smax.i64(i64 [[N]], i64 0)
-; CHECK-NEXT:    [[TMP0:%.*]] = udiv i64 42, [[N]]
+; CHECK-NEXT:    [[UMAX:%.*]] = call i64 @llvm.umax.i64(i64 [[N]], i64 1)
+; CHECK-NEXT:    [[TMP0:%.*]] = udiv i64 42, [[UMAX]]
 ; CHECK-NEXT:    [[UMIN:%.*]] = call i64 @llvm.umin.i64(i64 [[SMAX]], i64 [[TMP0]])
 ; CHECK-NEXT:    [[TMP1:%.*]] = add nuw nsw i64 [[UMIN]], 1
 ; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ule i64 [[TMP1]], 4
-; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[ENTRY:%.*]]
 ; CHECK:       vector.ph:
 ; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[TMP1]], 4
 ; CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[N_MOD_VF]], 0
 ; CHECK-NEXT:    [[TMP3:%.*]] = select i1 [[TMP2]], i64 4, i64 [[N_MOD_VF]]
 ; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 [[TMP1]], [[TMP3]]
-; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
+; CHECK-NEXT:    br label [[LOOP_HEADER:%.*]]
 ; CHECK:       vector.body:
-; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_STORE_CONTINUE6:%.*]] ]
-; CHECK-NEXT:    [[TMP4:%.*]] = add i64 [[INDEX]], 0
+; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ 0, [[ENTRY]] ], [ [[IV_NEXT:%.*]], [[LOOP_LATCH:%.*]] ]
+; CHECK-NEXT:    [[TMP4:%.*]] = add i64 [[IV]], 0
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP4]]
 ; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[TMP5]], i32 0
 ; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP6]], align 4
@@ -169,7 +170,7 @@ define i64 @multi_exit_2_exit_count_with_udiv_by_value_in_block_executed_uncondi
 ; CHECK-NEXT:    [[TMP10:%.*]] = extractelement <4 x i1> [[TMP7]], i32 1
 ; CHECK-NEXT:    br i1 [[TMP10]], label [[PRED_STORE_IF1:%.*]], label [[PRED_STORE_CONTINUE2:%.*]]
 ; CHECK:       pred.store.if1:
-; CHECK-NEXT:    [[TMP11:%.*]] = add i64 [[INDEX]], 1
+; CHECK-NEXT:    [[TMP11:%.*]] = add i64 [[IV]], 1
 ; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP11]]
 ; CHECK-NEXT:    store i32 1, ptr [[TMP12]], align 4
 ; CHECK-NEXT:    br label [[PRED_STORE_CONTINUE2]]
@@ -177,30 +178,30 @@ define i64 @multi_exit_2_exit_count_with_udiv_by_value_in_block_executed_uncondi
 ; CHECK-NEXT:    [[TMP13:%.*]] = extractelement <4 x i1> [[TMP7]], i32 2
 ; CHECK-NEXT:    br i1 [[TMP13]], label [[PRED_STORE_IF3:%.*]], label [[PRED_STORE_CONTINUE4:%.*]]
 ; CHECK:       pred.store.if3:
-; CHECK-NEXT:    [[TMP14:%.*]] = add i64 [[INDEX]], 2
+; CHECK-NEXT:    [[TMP14:%.*]] = add i64 [[IV]], 2
 ; CHECK-NEXT:    [[TMP15:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP14]]
 ; CHECK-NEXT:    store i32 1, ptr [[TMP15]], align 4
 ; CHECK-NEXT:    br label [[PRED_STORE_CONTINUE4]]
 ; CHECK:       pred.store.continue4:
 ; CHECK-NEXT:    [[TMP16:%.*]] = extractelement <4 x i1> [[TMP7]], i32 3
-; CHECK-NEXT:    br i1 [[TMP16]], label [[PRED_STORE_IF5:%.*]], label [[PRED_STORE_CONTINUE6]]
+; CHECK-NEXT:    br i1 [[TMP16]], label [[PRED_STORE_IF5:%.*]], label [[LOOP_LATCH]]
 ; CHECK:       pred.store.if5:
-; CHECK-NEXT:    [[TMP17:%.*]] = add i64 [[INDEX]], 3
+; CHECK-NEXT:    [[TMP17:%.*]] = add i64 [[IV]], 3
 ; CHECK-NEXT:    [[TMP18:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP17]]
 ; CHECK-NEXT:    store i32 1, ptr [[TMP18]], align 4
-; CHECK-NEXT:    br label [[PRED_STORE_CONTINUE6]]
+; CHECK-NEXT:    br label [[LOOP_LATCH]]
 ; CHECK:       pred.store.continue6:
-; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
-; CHECK-NEXT:    [[TMP19:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
-; CHECK-NEXT:    br i1 [[TMP19]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
+; CHECK-NEXT:    [[IV_NEXT]] = add nuw i64 [[IV]], 4
+; CHECK-NEXT:    [[TMP19:%.*]] = icmp eq i64 [[IV_NEXT]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[TMP19]], label [[MIDDLE_BLOCK:%.*]], label [[LOOP_HEADER]], !llvm.loop [[LOOP6:![0-9]+]]
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    br label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
-; CHECK-NEXT:    br label [[LOOP_HEADER:%.*]]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY1:%.*]] ]
+; CHECK-NEXT:    br label [[LOOP_HEADER1:%.*]]
 ; CHECK:       loop.header:
-; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP_LATCH:%.*]] ]
-; CHECK-NEXT:    [[GEP:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IV]]
+; CHECK-NEXT:    [[IV1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT1:%.*]], [[LOOP_LATCH1:%.*]] ]
+; CHECK-NEXT:    [[GEP:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IV1]]
 ; CHECK-NEXT:    [[L:%.*]] = load i32, ptr [[GEP]], align 4
 ; CHECK-NEXT:    [[C_2:%.*]] = icmp eq i32 [[L]], 10
 ; CHECK-NEXT:    br i1 [[C_2]], label [[THEN:%.*]], label [[CONTINUE:%.*]]
@@ -209,14 +210,14 @@ define i64 @multi_exit_2_exit_count_with_udiv_by_value_in_block_executed_uncondi
 ; CHECK-NEXT:    br label [[CONTINUE]]
 ; CHECK:       continue:
 ; CHECK-NEXT:    [[D:%.*]] = udiv i64 42, [[N]]
-; CHECK-NEXT:    [[C_1:%.*]] = icmp slt i64 [[IV]], [[D]]
-; CHECK-NEXT:    br i1 [[C_1]], label [[LOOP_LATCH]], label [[EXIT:%.*]]
+; CHECK-NEXT:    [[C_1:%.*]] = icmp slt i64 [[IV1]], [[D]]
+; CHECK-NEXT:    br i1 [[C_1]], label [[LOOP_LATCH1]], label [[EXIT:%.*]]
 ; CHECK:       loop.latch:
-; CHECK-NEXT:    [[IV_NEXT]] = add i64 [[IV]], 1
-; CHECK-NEXT:    [[C_0:%.*]] = icmp slt i64 [[IV]], [[N]]
-; CHECK-NEXT:    br i1 [[C_0]], label [[LOOP_HEADER]], label [[EXIT]], !llvm.loop [[LOOP7:![0-9]+]]
+; CHECK-NEXT:    [[IV_NEXT1]] = add i64 [[IV1]], 1
+; CHECK-NEXT:    [[C_0:%.*]] = icmp slt i64 [[IV1]], [[N]]
+; CHECK-NEXT:    br i1 [[C_0]], label [[LOOP_HEADER1]], label [[EXIT]], !llvm.loop [[LOOP7:![0-9]+]]
 ; CHECK:       exit:
-; CHECK-NEXT:    [[P:%.*]] = phi i64 [ 1, [[CONTINUE]] ], [ 0, [[LOOP_LATCH]] ]
+; CHECK-NEXT:    [[P:%.*]] = phi i64 [ 1, [[CONTINUE]] ], [ 0, [[LOOP_LATCH1]] ]
 ; CHECK-NEXT:    ret i64 [[P]]
 ;
 entry:
@@ -257,16 +258,16 @@ define i64 @multi_exit_2_exit_count_with_udiv_by_constant_in_block_executed_unco
 ; CHECK-NEXT:    [[UMIN:%.*]] = call i64 @llvm.umin.i64(i64 [[SMAX]], i64 [[TMP0]])
 ; CHECK-NEXT:    [[TMP1:%.*]] = add nuw nsw i64 [[UMIN]], 1
 ; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ule i64 [[TMP1]], 4
-; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[ENTRY:%.*]]
 ; CHECK:       vector.ph:
 ; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[TMP1]], 4
 ; CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[N_MOD_VF]], 0
 ; CHECK-NEXT:    [[TMP3:%.*]] = select i1 [[TMP2]], i64 4, i64 [[N_MOD_VF]]
 ; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 [[TMP1]], [[TMP3]]
-; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
+; CHECK-NEXT:    br label [[LOOP_HEADER:%.*]]
 ; CHECK:       vector.body:
-; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_STORE_CONTINUE6:%.*]] ]
-; CHECK-NEXT:    [[TMP4:%.*]] = add i64 [[INDEX]], 0
+; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ 0, [[ENTRY]] ], [ [[IV_NEXT:%.*]], [[LOOP_LATCH:%.*]] ]
+; CHECK-NEXT:    [[TMP4:%.*]] = add i64 [[IV]], 0
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP4]]
 ; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[TMP5]], i32 0
 ; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP6]], align 4
@@ -281,7 +282,7 @@ define i64 @multi_exit_2_exit_count_with_udiv_by_constant_in_block_executed_unco
 ; CHECK-NEXT:    [[TMP10:%.*]] = extractelement <4 x i1> [[TMP7]], i32 1
 ; CHECK-NEXT:    br i1 [[TMP10]], label [[PRED_STORE_IF1:%.*]], label [[PRED_STORE_CONTINUE2:%.*]]
 ; CHECK:       pred.store.if1:
-; CHECK-NEXT:    [[TMP11:%.*]] = add i64 [[INDEX]], 1
+; CHECK-NEXT:    [[TMP11:%.*]] = add i64 [[IV]], 1
 ; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP11]]
 ; CHECK-NEXT:    store i32 1, ptr [[TMP12]], align 4
 ; CHECK-NEXT:    br label [[PRED_STORE_CONTINUE2]]
@@ -289,30 +290,30 @@ define i64 @multi_exit_2_exit_count_with_udiv_by_constant_in_block_executed_unco
 ; CHECK-NEXT:    [[TMP13:%.*]] = extractelement <4 x i1> [[TMP7]], i32 2
 ; CHECK-NEXT:    br i1 [[TMP13]], label [[PRED_STORE_IF3:%.*]], label [[PRED_STORE_CONTINUE4:%.*]]
 ; CHECK:       pred.store.if3:
-; CHECK-NEXT:    [[TMP14:%.*]] = add i64 [[INDEX]], 2
+; CHECK-NEXT:    [[TMP14:%.*]] = add i64 [[IV]], 2
 ; CHECK-NEXT:    [[TMP15:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP14]]
 ; CHECK-NEXT:    store i32 1, ptr [[TMP15]], align 4
 ; CHECK-NEXT:    br label [[PRED_STORE_CONTINUE4]]
 ; CHECK:       pred.store.continue4:
 ; CHECK-NEXT:    [[TMP16:%.*]] = extractelement <4 x i1> [[TMP7]], i32 3
-; CHECK-NEXT:    br i1 [[TMP16]], label [[PRED_STORE_IF5:%.*]], label [[PRED_STORE_CONTINUE6]]
+; CHECK-NEXT:    br i1 [[TMP16]], label [[PRED_STORE_IF5:%.*]], label [[LOOP_LATCH]]
 ; CHECK:       pred.store.if5:
-; CHECK-NEXT:    [[TMP17:%.*]] = add i64 [[INDEX]], 3
+; CHECK-NEXT:    [[TMP17:%.*]] = add i64 [[IV]], 3
 ; CHECK-NEXT:    [[TMP18:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP17]]
 ; CHECK-NEXT:    store i32 1, ptr [[TMP18]], align 4
-; CHECK-NEXT:    br label [[PRED_STORE_CONTINUE6]]
+; CHECK-NEXT:    br label [[LOOP_LATCH]]
 ; CHECK:       pred.store.continue6:
-; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
-; CHECK-NEXT:    [[TMP19:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
-; CHECK-NEXT:    br i1 [[TMP19]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
+; CHECK-NEXT:    [[IV_NEXT]] = add nuw i64 [[IV]], 4
+; CHECK-NEXT:    [[TMP19:%.*]] = icmp eq i64 [[IV_NEXT]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[TMP19]], label [[MIDDLE_BLOCK:%.*]], label [[LOOP_HEADER]], !llvm.loop [[LOOP8:![0-9]+]]
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    br label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
-; CHECK-NEXT:    br label [[LOOP_HEADER:%.*]]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY1:%.*]] ]
+; CHECK-NEXT:    br label [[LOOP_HEADER1:%.*]]
 ; CHECK:       loop.header:
-; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP_LATCH:%.*]] ]
-; CHECK-NEXT:    [[GEP:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IV]]
+; CHECK-NEXT:    [[IV1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT1:%.*]], [[LOOP_LATCH1:%.*]] ]
+; CHECK-NEXT:    [[GEP:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IV1]]
 ; CHECK-NEXT:    [[L:%.*]] = load i32, ptr [[GEP]], align 4
 ; CHECK-NEXT:    [[C_2:%.*]] = icmp eq i32 [[L]], 10
 ; CHECK-NEXT:    br i1 [[C_2]], label [[THEN:%.*]], label [[CONTINUE:%.*]]
@@ -321,14 +322,14 @@ define i64 @multi_exit_2_exit_count_with_udiv_by_constant_in_block_executed_unco
 ; CHECK-NEXT:    br label [[CONTINUE]]
 ; CHECK:       continue:
 ; CHECK-NEXT:    [[D:%.*]] = udiv i64 [[N]], 42
-; CHECK-NEXT:    [[C_1:%.*]] = icmp slt i64 [[IV]], [[D]]
-; CHECK-NEXT:    br i1 [[C_1]], label [[LOOP_LATCH]], label [[EXIT:%.*]]
+; CHECK-NEXT:    [[C_1:%.*]] = icmp slt i64 [[IV1]], [[D]]
+; CHECK-NEXT:    br i1 [[C_1]], label [[LOOP_LATCH1]], label [[EXIT:%.*]]
 ; CHECK:       loop.latch:
-; CHECK-NEXT:    [[IV_NEXT]] = add i64 [[IV]], 1
-; CHECK-NEXT:    [[C_0:%.*]] = icmp slt i64 [[IV]], [[N]]
-; CHECK-NEXT:    br i1 [[C_0]], label [[LOOP_HEADER]], label [[EXIT]], !llvm.loop [[LOOP9:![0-9]+]]
+; CHECK-NEXT:    [[IV_NEXT1]] = add i64 [[IV1]], 1
+; CHECK-NEXT:    [[C_0:%.*]] = icmp slt i64 [[IV1]], [[N]]
+; CHECK-NEXT:    br i1 [[C_0]], label [[LOOP_HEADER1]], label [[EXIT]], !llvm.loop [[LOOP9:![0-9]+]]
 ; CHECK:       exit:
-; CHECK-NEXT:    [[P:%.*]] = phi i64 [ 1, [[CONTINUE]] ], [ 0, [[LOOP_LATCH]] ]
+; CHECK-NEXT:    [[P:%.*]] = phi i64 [ 1, [[CONTINUE]] ], [ 0, [[LOOP_LATCH1]] ]
 ; CHECK-NEXT:    ret i64 [[P]]
 ;
 entry:
@@ -459,14 +460,13 @@ exit:
   ret i64 %p
 }
 
-; FIXME: Currently miscompiled as we unconditionally execute udiv after
-; vectorization.
 define i64 @multi_exit_4_exit_count_with_udiv_by_value_in_latch(ptr %dst, i64 %N) {
 ; CHECK-LABEL: define i64 @multi_exit_4_exit_count_with_udiv_by_value_in_latch(
 ; CHECK-SAME: ptr [[DST:%.*]], i64 [[N:%.*]]) {
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    [[SMAX:%.*]] = call i64 @llvm.smax.i64(i64 [[N]], i64 0)
-; CHECK-NEXT:    [[TMP0:%.*]] = udiv i64 42, [[N]]
+; CHECK-NEXT:    [[UMAX:%.*]] = call i64 @llvm.umax.i64(i64 [[N]], i64 1)
+; CHECK-NEXT:    [[TMP0:%.*]] = udiv i64 42, [[UMAX]]
 ; CHECK-NEXT:    [[UMIN:%.*]] = call i64 @llvm.umin.i64(i64 [[SMAX]], i64 [[TMP0]])
 ; CHECK-NEXT:    [[TMP1:%.*]] = add nuw nsw i64 [[UMIN]], 1
 ; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ule i64 [[TMP1]], 4
@@ -542,34 +542,34 @@ define i64 @multi_exit_4_exit_count_with_udiv_by_constant_in_latch(ptr %dst, i64
 ; CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[N_MOD_VF]], 0
 ; CHECK-NEXT:    [[TMP3:%.*]] = select i1 [[TMP2]], i64 4, i64 [[N_MOD_VF]]
 ; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 [[TMP1]], [[TMP3]]
-; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
+; CHECK-NEXT:    br label [[LOOP_HEADER:%.*]]
 ; CHECK:       vector.body:
-; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; CHECK-NEXT:    [[TMP4:%.*]] = add i64 [[INDEX]], 0
-; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 [[TMP4]]
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[TMP5]], i32 0
+; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[LOOP_HEADER]] ]
+; CHECK-NEXT:    [[IV:%.*]] = add i64 [[INDEX]], 0
+; CHECK-NEXT:    [[GEP:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 [[IV]]
+; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[GEP]], i32 0
 ; CHECK-NEXT:    store <4 x i32> <i32 1, i32 1, i32 1, i32 1>, ptr [[TMP6]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
-; CHECK-NEXT:    br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP12:![0-9]+]]
+; CHECK-NEXT:    br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[LOOP_HEADER]], !llvm.loop [[LOOP12:![0-9]+]]
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    br label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
-; CHECK-NEXT:    br label [[LOOP_HEADER:%.*]]
+; CHECK-NEXT:    br label [[LOOP_HEADER1:%.*]]
 ; CHECK:       loop.header:
-; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP_LATCH:%.*]] ]
-; CHECK-NEXT:    [[GEP:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 [[IV]]
-; CHECK-NEXT:    store i32 1, ptr [[GEP]], align 4
-; CHECK-NEXT:    [[C_0:%.*]] = icmp slt i64 [[IV]], [[N]]
+; CHECK-NEXT:    [[IV1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP_LATCH:%.*]] ]
+; CHECK-NEXT:    [[GEP1:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 [[IV1]]
+; CHECK-NEXT:    store i32 1, ptr [[GEP1]], align 4
+; CHECK-NEXT:    [[C_0:%.*]] = icmp slt i64 [[IV1]], [[N]]
 ; CHECK-NEXT:    br i1 [[C_0]], label [[LOOP_LATCH]], label [[EXIT:%.*]]
 ; CHECK:       loop.latch:
-; CHECK-NEXT:    [[IV_NEXT]] = add i64 [[IV]], 1
+; CHECK-NEXT:    [[IV_NEXT]] = add i64 [[IV1]], 1
 ; CHECK-NEXT:    [[D:%.*]] = udiv i64 [[N]], 42
-; CHECK-NEXT:    [[C_1:%.*]] = icmp slt i64 [[IV]], [[D]]
-; CHECK-NEXT:    br i1 [[C_1]], label [[LOOP_HEADER]], label [[EXIT]], !llvm.loop [[LOOP13:![0-9]+]]
+; CHECK-NEXT:    [[C_1:%.*]] = icmp slt i64 [[IV1]], [[D]]
+; CHECK-NEXT:    br i1 [[C_1]], label [[LOOP_HEADER1]], label [[EXIT]], !llvm.loop [[LOOP13:![0-9]+]]
 ; CHECK:       exit:
-; CHECK-NEXT:    [[P:%.*]] = phi i64 [ 1, [[LOOP_HEADER]] ], [ 0, [[LOOP_LATCH]] ]
+; CHECK-NEXT:    [[P:%.*]] = phi i64 [ 1, [[LOOP_HEADER1]] ], [ 0, [[LOOP_LATCH]] ]
 ; CHECK-NEXT:    ret i64 [[P]]
 ;
 entry:
@@ -647,15 +647,14 @@ exit:
   ret void
 }
 
-; FIXME: Currently miscompiled as we unconditionally execute udiv after
-; vectorization.
 define i64 @multi_exit_4_exit_count_with_urem_by_value_in_latch(ptr %dst, i64 %N) {
 ; CHECK-LABEL: define i64 @multi_exit_4_exit_count_with_urem_by_value_in_latch(
 ; CHECK-SAME: ptr [[DST:%.*]], i64 [[N:%.*]]) {
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    [[SMAX:%.*]] = call i64 @llvm.smax.i64(i64 [[N]], i64 0)
-; CHECK-NEXT:    [[TMP0:%.*]] = udiv i64 42, [[N]]
-; CHECK-NEXT:    [[TMP1:%.*]] = mul nuw i64 [[N]], [[TMP0]]
+; CHECK-NEXT:    [[UMAX:%.*]] = call i64 @llvm.umax.i64(i64 [[N]], i64 1)
+; CHECK-NEXT:    [[TMP0:%.*]] = udiv i64 42, [[UMAX]]
+; CHECK-NEXT:    [[TMP1:%.*]] = mul i64 [[N]], [[TMP0]]
 ; CHECK-NEXT:    [[TMP2:%.*]] = sub i64 42, [[TMP1]]
 ; CHECK-NEXT:    [[SMAX1:%.*]] = call i64 @llvm.smax.i64(i64 [[TMP2]], i64 0)
 ; CHECK-NEXT:    [[UMIN:%.*]] = call i64 @llvm.umin.i64(i64 [[SMAX]], i64 [[SMAX1]])
@@ -723,45 +722,45 @@ define i64 @multi_exit_4_exit_count_with_urem_by_constant_in_latch(ptr %dst, i64
 ; CHECK-SAME: ptr [[DST:%.*]], i64 [[N:%.*]]) {
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    [[SMAX:%.*]] = call i64 @llvm.smax.i64(i64 [[N]], i64 0)
-; CHECK-NEXT:    [[TMP2:%.*]] = urem i64 [[N]], 42
-; CHECK-NEXT:    [[SMAX1:%.*]] = call i64 @llvm.smax.i64(i64 [[TMP2]], i64 0)
+; CHECK-NEXT:    [[TMP0:%.*]] = urem i64 [[N]], 42
+; CHECK-NEXT:    [[SMAX1:%.*]] = call i64 @llvm.smax.i64(i64 [[TMP0]], i64 0)
 ; CHECK-NEXT:    [[UMIN:%.*]] = call i64 @llvm.umin.i64(i64 [[SMAX]], i64 [[SMAX1]])
-; CHECK-NEXT:    [[TMP3:%.*]] = add nuw i64 [[UMIN]], 1
-; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ule i64 [[TMP3]], 4
+; CHECK-NEXT:    [[TMP1:%.*]] = add nuw i64 [[UMIN]], 1
+; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ule i64 [[TMP1]], 4
 ; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
 ; CHECK:       vector.ph:
-; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[TMP3]], 4
-; CHECK-NEXT:    [[TMP4:%.*]] = icmp eq i64 [[N_MOD_VF]], 0
-; CHECK-NEXT:    [[TMP5:%.*]] = select i1 [[TMP4]], i64 4, i64 [[N_MOD_VF]]
-; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 [[TMP3]], [[TMP5]]
-; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
+; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[TMP1]], 4
+; CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[N_MOD_VF]], 0
+; CHECK-NEXT:    [[TMP3:%.*]] = select i1 [[TMP2]], i64 4, i64 [[N_MOD_VF]]
+; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 [[TMP1]], [[TMP3]]
+; CHECK-NEXT:    br label [[LOOP_HEADER:%.*]]
 ; CHECK:       vector.body:
-; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; CHECK-NEXT:    [[TMP6:%.*]] = add i64 [[INDEX]], 0
-; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 [[TMP6]]
-; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[TMP7]], i32 0
-; CHECK-NEXT:    store <4 x i32> <i32 1, i32 1, i32 1, i32 1>, ptr [[TMP8]], align 4
+; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[LOOP_HEADER]] ]
+; CHECK-NEXT:    [[IV:%.*]] = add i64 [[INDEX]], 0
+; CHECK-NEXT:    [[GEP:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 [[IV]]
+; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[GEP]], i32 0
+; CHECK-NEXT:    store <4 x i32> <i32 1, i32 1, i32 1, i32 1>, ptr [[TMP6]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
-; CHECK-NEXT:    [[TMP9:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
-; CHECK-NEXT:    br i1 [[TMP9]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP18:![0-9]+]]
+; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[LOOP_HEADER]], !llvm.loop [[LOOP18:![0-9]+]]
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    br label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
-; CHECK-NEXT:    br label [[LOOP_HEADER:%.*]]
+; CHECK-NEXT:    br label [[LOOP_HEADER1:%.*]]
 ; CHECK:       loop.header:
-; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP_LATCH:%.*]] ]
-; CHECK-NEXT:    [[GEP:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 [[IV]]
-; CHECK-NEXT:    store i32 1, ptr [[GEP]], align 4
-; CHECK-NEXT:    [[C_0:%.*]] = icmp slt i64 [[IV]], [[N]]
+; CHECK-NEXT:    [[IV1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP_LATCH:%.*]] ]
+; CHECK-NEXT:    [[GEP1:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 [[IV1]]
+; CHECK-NEXT:    store i32 1, ptr [[GEP1]], align 4
+; CHECK-NEXT:    [[C_0:%.*]] = icmp slt i64 [[IV1]], [[N]]
 ; CHECK-NEXT:    br i1 [[C_0]], label [[LOOP_LATCH]], label [[EXIT:%.*]]
 ; CHECK:       loop.latch:
-; CHECK-NEXT:    [[IV_NEXT]] = add i64 [[IV]], 1
+; CHECK-NEXT:    [[IV_NEXT]] = add i64 [[IV1]], 1
 ; CHECK-NEXT:    [[D:%.*]] = urem i64 [[N]], 42
-; CHECK-NEXT:    [[C_1:%.*]] = icmp slt i64 [[IV]], [[D]]
-; CHECK-NEXT:    br i1 [[C_1]], label [[LOOP_HEADER]], label [[EXIT]], !llvm.loop [[LOOP19:![0-9]+]]
+; CHECK-NEXT:    [[C_1:%.*]] = icmp slt i64 [[IV1]], [[D]]
+; CHECK-NEXT:    br i1 [[C_1]], label [[LOOP_HEADER1]], label [[EXIT]], !llvm.loop [[LOOP19:![0-9]+]]
 ; CHECK:       exit:
-; CHECK-NEXT:    [[P:%.*]] = phi i64 [ 1, [[LOOP_HEADER]] ], [ 0, [[LOOP_LATCH]] ]
+; CHECK-NEXT:    [[P:%.*]] = phi i64 [ 1, [[LOOP_HEADER1]] ], [ 0, [[LOOP_LATCH]] ]
 ; CHECK-NEXT:    ret i64 [[P]]
 ;
 entry:
@@ -867,6 +866,76 @@ exit:
   ret i64 %p
 }
 
+define i64 @multi_exit_4_exit_count_with_udiv_by_value_in_latch1(ptr %dst, i64 %N) {
+; CHECK-LABEL: define i64 @multi_exit_4_exit_count_with_udiv_by_value_in_latch1(
+; CHECK-SAME: ptr [[DST:%.*]], i64 [[N:%.*]]) {
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[SMAX:%.*]] = call i64 @llvm.smax.i64(i64 [[N]], i64 0)
+; CHECK-NEXT:    [[UMAX:%.*]] = call i64 @llvm.umax.i64(i64 [[N]], i64 1)
+; CHECK-NEXT:    [[TMP0:%.*]] = udiv i64 42, [[UMAX]]
+; CHECK-NEXT:    [[UMIN:%.*]] = call i64 @llvm.umin.i64(i64 [[SMAX]], i64 [[TMP0]])
+; CHECK-NEXT:    [[TMP1:%.*]] = add nuw nsw i64 [[UMIN]], 1
+; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ule i64 [[TMP1]], 4
+; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; CHECK:       vector.ph:
+; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[TMP1]], 4
+; CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[N_MOD_VF]], 0
+; CHECK-NEXT:    [[TMP3:%.*]] = select i1 [[TMP2]], i64 4, i64 [[N_MOD_VF]]
+; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 [[TMP1]], [[TMP3]]
+; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
+; CHECK:       vector.body:
+; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[TMP4:%.*]] = add i64 [[INDEX]], 0
+; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 [[TMP4]]
+; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[TMP5]], i32 0
+; CHECK-NEXT:    store <4 x i32> <i32 1, i32 1, i32 1, i32 1>, ptr [[TMP6]], align 4
+; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
+; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP20:![0-9]+]]
+; CHECK:       middle.block:
+; CHECK-NEXT:    br label [[SCALAR_PH]]
+; CHECK:       scalar.ph:
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    br label [[LOOP_HEADER:%.*]]
+; CHECK:       loop.header:
+; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP_LATCH:%.*]] ]
+; CHECK-NEXT:    [[GEP:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 [[IV]]
+; CHECK-NEXT:    store i32 1, ptr [[GEP]], align 4
+; CHECK-NEXT:    [[C_0:%.*]] = icmp slt i64 [[IV]], [[N]]
+; CHECK-NEXT:    br i1 [[C_0]], label [[LOOP_LATCH]], label [[EXIT:%.*]]
+; CHECK:       loop.latch:
+; CHECK-NEXT:    [[IV_NEXT]] = add i64 [[IV]], 1
+; CHECK-NEXT:    [[D:%.*]] = udiv i64 42, [[N]]
+; CHECK-NEXT:    [[X:%.*]] = sub i64 100, [[D]]
+; CHECK-NEXT:    [[C_1:%.*]] = icmp slt i64 [[IV]], [[D]]
+; CHECK-NEXT:    br i1 [[C_1]], label [[LOOP_HEADER]], label [[EXIT]], !llvm.loop [[LOOP21:![0-9]+]]
+; CHECK:       exit:
+; CHECK-NEXT:    [[P:%.*]] = phi i64 [ 1, [[LOOP_HEADER]] ], [ 0, [[LOOP_LATCH]] ]
+; CHECK-NEXT:    ret i64 [[P]]
+;
+entry:
+  br label %loop.header
+
+loop.header:
+  %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop.latch ]
+  %gep = getelementptr inbounds i32, ptr %dst, i64 %iv
+  store i32 1, ptr %gep
+  %c.0 = icmp slt i64 %iv, %N
+  br i1 %c.0, label %loop.latch, label %exit
+
+loop.latch:
+  %iv.next = add i64 %iv, 1
+  %d = udiv i64 42, %N
+  %x = sub i64 100, %d
+  %c.1 = icmp slt i64 %iv, %d
+  br i1 %c.1, label %loop.header, label %exit
+
+exit:
+  %p = phi i64 [ 1, %loop.header ], [ 0, %loop.latch]
+  ret i64 %p
+}
+
+
 ;.
 ; CHECK: [[LOOP0]] = distinct !{[[LOOP0]], [[META1:![0-9]+]], [[META2:![0-9]+]]}
 ; CHECK: [[META1]] = !{!"llvm.loop.isvectorized", i32 1}
@@ -888,4 +957,6 @@ exit:
 ; CHECK: [[LOOP17]] = distinct !{[[LOOP17]], [[META2]], [[META1]]}
 ; CHECK: [[LOOP18]] = distinct !{[[LOOP18]], [[META1]], [[META2]]}
 ; CHECK: [[LOOP19]] = distinct !{[[LOOP19]], [[META2]], [[META1]]}
+; CHECK: [[LOOP20]] = distinct !{[[LOOP20]], [[META1]], [[META2]]}
+; CHECK: [[LOOP21]] = distinct !{[[LOOP21]], [[META2]], [[META1]]}
 ;.



More information about the llvm-commits mailing list