[ScalarEvolution] Apply loop guards against min/max for its arguments

[dmakogon]

Summary

This replaces several rewriting rules in ScalarEvolution::applyLoopGuards that are applied to min/max expressions with the equivalent ones but applied to its arguments.
So currently given we have a loop guard min(a, b) >= c, the min expression gets rewritten as max(c, min(a, b)).
With such approach, we're unable to apply the rewrite if min operands are zext for example (min(zext(a), zext(b))), however it's equivalent to the expression zext(min(a, b)) for which we can apply the rewrite.
We can rewrite the min operands instead with these expressions:

• a --> max(c, a) and
• b --> max(c, b)

and this would allow us to apply the loop guard in this and similar cases: min(zext(a), zext(b)) would get rewritten as min(zext(max(c, a)), zext(max(c, b))) instead of just being skipped.

The table of replaced rules (omitting predicates signedness for simplicity):

Old rule	New rules
min(a, b) >= c --> max(c, min(a, b))	a --> max(a, c) and b --> max(b, c)
min(a, b) > c --> max(c + 1, min(a, b))	a --> max(a, c + 1) and b --> max(b, c + 1)
max(a, b) <= c --> min(c, max(a, b))	a --> min(a, c) and b --> min(b, c)
max(a, b) < c --> min(c - 1, max(a, b))	a --> min(a, c - 1) and b --> min(b, c - 1)

[dmakogon]

There is one regression I didn't notice initially (trip-count-minmax.ll). Posted fix for review - https://reviews.llvm.org/D145231

[dmakogon]

Cleanup, got rid of copy-paste code

[mkazantsev]

llvm/lib/Analysis/ScalarEvolution.cpp

15077

This is not correct (c + 1 may overflow). Why didn't you just keep it as

//  'min(a, b) >  c'   ->   '(a > c) && (b > c)';

?

15079

Same

15097

Bug when RHS = SINT_MIN.

15121

Why do you think that eq/ne cannot come here?

[mkazantsev]

This comment was removed by mkazantsev.

[mkazantsev]

Okay, seems that same bug is in underlying code. I suggest to fix it first. It seems to be turning predicates that are trivially true into things that may be false .

llvm/lib/Analysis/ScalarEvolution.cpp

15121

Question withdrawn, I see.

[mkazantsev]

llvm/lib/Analysis/ScalarEvolution.cpp

15074–15081

This comment is out of place. You are not constructing c + 1 or c - 1 here, it is done in the caller code. Here enough to say that

min(a, b) > c     ->   (a > c) && (b > c); // same for >=
max(a, b) < c     ->   (a < c) && (b < c); // same for <=

[mkazantsev]

llvm/lib/Analysis/ScalarEvolution.cpp

15096–15099

I propose refactoring like:

case CmpInst::ICMP_UGT:
  RHS = RHS - 1;
  LLVM_FALLTHROUGH;
case CmpInst::ICMP_UGE:
  RewrittenRHS = getUMaxExpr(RewrittenLHS, RHS);

[mkazantsev]

Generally, you are just replicating the bug and not introducing a new one. I don't insist that fix is blocking this one, you can make them in any order you like more. Please address restructuring comments. Also fix the bug with max/min having another number of operands than 2.

[mkazantsev]

llvm/lib/Analysis/ScalarEvolution.cpp

15093–15096

There can be more (or less?) than 2 operands.

[mkazantsev]

llvm/lib/Analysis/ScalarEvolution.cpp

15123–15124

What if the operands are also min/max? Maybe use a DFS traversal instead to collect all leaves?

[mkazantsev]

llvm/lib/Analysis/ScalarEvolution.cpp

15123–15124

I mean, non-recursive traversal using stack.

[mkazantsev]

llvm/lib/Analysis/ScalarEvolution.cpp

15085

You should set RewrittenRHS if it happens.

15087

Can we just always call AddMinMaxRewrites and, if it's not min or max, just to the trivial thing?

[mkazantsev]

So far plans look follwing:

1. Refactor switch cases into fallthroughs;
2. Then, make an NFC that calls a function. Instead of

RewrittenRHS = getUMaxExpr(RewrittenLHS, RHS);

smth like

RewrittenRHS = AddRewrites(RewrittenLHS, RHS);

The handling of predicate can be moved inside this function. You don't need to have 2 predicate switches for calling, you just need one. Maybe 2nd (small) switch is only for checks and adding + - 1.

3. Into this function, add handling of min/max (non-recursive stack-based traversal). The returned value should be the rewrite of original RewrittenLHS no matter if it's a min or not.

As a side thing - add checks that, whenever you compute RHS + - 1, it is not a MAX/MIN int respectively. In this case we should not rewrite anything.

[dmakogon]

Thanks for the suggestions, Max.

1. Now the switch uses fallthroughs.
2. The expressions are now processed in a worklist, for min/max we enqueue their operands (this also means we support any number of operands).
3. Suggest to fix +/- 1 bug in a follow-up. The bug was there before, this patch doesn't expose it anyway.

And also there was no use in adding a new function, the traversal is done in the same function as before.

[mkazantsev]

Underlying code seems buggy to me. I'd prefer it fixed in follow-ups.

Your change mostly LG, if possible, do variable renaming as a separate patch.

Please wait couple of days in case if someone else has concerns before landing this.

llvm/lib/Analysis/ScalarEvolution.cpp

15079–15080

I know it's not your checks, but this can be changed further.

• Why pointers are checked only for ult? ugt will also make us computations with it, and why not check it for signed?
  • Does this whole thing even make any sense for pointers?
• Why we do umax(..., UMIN + 1) correction for unsigned but not smax(..., SMIN + 1) for signed? Why don't we do the same correction for umin(RHS, UMAX - 1) when we are about to add one? Looks like an attempt to hide some bugs.

15220

Can variable renaming be done separately?

llvm/lib/Analysis/ScalarEvolution.cpp

auto CollectCondition = [&](ICmpInst::Predicate Predicate, const SCEV *LHS,
     // Checks whether \p S has already been rewritten. In that case returns the
     // existing rewrite because we want to chain further rewrites onto the
     // already rewritten value. Otherwise returns \p S.
     auto GetMaybeRewritten = [&](const SCEV *S) {
       auto I = RewriteMap.find(S);
       return I != RewriteMap.end() ? I->second : S;
     };
 
     const SCEV *RewrittenLHS = GetMaybeRewritten(LHS);
 
     const SCEV *RewrittenRHS = nullptr;
     switch (Predicate) {
     case CmpInst::ICMP_ULT: {
       if (RHS->getType()->isPointerTy())
         break;
       const SCEV *One = getOne(RHS->getType());
-      RewrittenRHS =
-          getUMinExpr(RewrittenLHS, getMinusSCEV(getUMaxExpr(RHS, One), One));
+      const SCEV *RHSMinusOne = getMinusSCEV(getUMaxExpr(RHS, One), One);
+      // umax(a, b) <u c -> (a <=u c - 1) && (b <=u c - 1)
+      if (auto *UMax = dyn_cast<SCEVUMaxExpr>(RewrittenLHS)) {
+        auto *Op1 = UMax->getOperand(0);
+        auto *Op2 = UMax->getOperand(1);
+        auto *Op1Rewritten = GetMaybeRewritten(Op1);
+        auto *Op2Rewritten = GetMaybeRewritten(Op2);
+        AddRewrite(Op1, Op1Rewritten, getUMinExpr(Op1Rewritten, RHSMinusOne));
+        AddRewrite(Op2, Op2Rewritten, getUMinExpr(Op2Rewritten, RHSMinusOne));
+        break;
+      }
+      RewrittenRHS = getUMinExpr(RewrittenLHS, RHSMinusOne);
+      break;
+    }
+    case CmpInst::ICMP_SLT: {
+      const SCEV *RHSMinusOne = getMinusSCEV(RHS, getOne(RHS->getType()));
+      // smax(a, b) <s c -> (a <=s c - 1) && (b <=s c - 1)
+      if (auto *SMax = dyn_cast<SCEVSMaxExpr>(RewrittenLHS)) {
+        auto *Op1 = SMax->getOperand(0);
+        auto *Op2 = SMax->getOperand(1);
+        auto *Op1Rewritten = GetMaybeRewritten(Op1);
+        auto *Op2Rewritten = GetMaybeRewritten(Op2);
+        AddRewrite(Op1, Op1Rewritten, getSMinExpr(Op1Rewritten, RHSMinusOne));
+        AddRewrite(Op2, Op2Rewritten, getSMinExpr(Op2Rewritten, RHSMinusOne));
         break;
       }
-    case CmpInst::ICMP_SLT:
-      RewrittenRHS =
-          getSMinExpr(RewrittenLHS, getMinusSCEV(RHS, getOne(RHS->getType())));
+      RewrittenRHS = getSMinExpr(RewrittenLHS, RHSMinusOne);
       break;
+    }
     case CmpInst::ICMP_ULE:
+      // umax(a, b) <=u c -> a <=u c && b <=u c
+      if (auto *UMax = dyn_cast<SCEVUMaxExpr>(RewrittenLHS)) {
+        auto *Op1 = UMax->getOperand(0);
+        auto *Op2 = UMax->getOperand(1);
+        auto *Op1Rewritten = GetMaybeRewritten(Op1);
+        auto *Op2Rewritten = GetMaybeRewritten(Op2);
+        AddRewrite(Op1, Op1Rewritten, getUMinExpr(Op1Rewritten, RHS));
+        AddRewrite(Op2, Op2Rewritten, getUMinExpr(Op2Rewritten, RHS));
+        break;
+      }
       RewrittenRHS = getUMinExpr(RewrittenLHS, RHS);
       break;
     case CmpInst::ICMP_SLE:
+      // smax(a, b) <=s c -> a <=s c && b <=s c
+      if (auto *SMax = dyn_cast<SCEVSMaxExpr>(RewrittenLHS)) {
+        auto *Op1 = SMax->getOperand(0);
+        auto *Op2 = SMax->getOperand(1);
+        auto *Op1Rewritten = GetMaybeRewritten(Op1);
+        auto *Op2Rewritten = GetMaybeRewritten(Op2);
+        AddRewrite(Op1, Op1Rewritten, getSMinExpr(Op1Rewritten, RHS));
+        AddRewrite(Op2, Op2Rewritten, getSMinExpr(Op2Rewritten, RHS));
+        break;
+      }
       RewrittenRHS = getSMinExpr(RewrittenLHS, RHS);
       break;
-    case CmpInst::ICMP_UGT:
-      RewrittenRHS =
-          getUMaxExpr(RewrittenLHS, getAddExpr(RHS, getOne(RHS->getType())));
-      break;
-    case CmpInst::ICMP_SGT:
-      RewrittenRHS =
-          getSMaxExpr(RewrittenLHS, getAddExpr(RHS, getOne(RHS->getType())));
+    case CmpInst::ICMP_UGT: {
+      const SCEV *RHSPlusOne = getAddExpr(RHS, getOne(RHS->getType()));
+      // umin(a, b) >u c -> (a >=u c + 1) && (b >=u c + 1)
+      if (auto *UMin = dyn_cast<SCEVUMinExpr>(RewrittenLHS)) {
+        auto *Op1 = UMin->getOperand(0);
+        auto *Op2 = UMin->getOperand(1);
+        auto *Op1Rewritten = GetMaybeRewritten(Op1);
+        auto *Op2Rewritten = GetMaybeRewritten(Op2);
+        AddRewrite(Op1, Op1Rewritten, getUMaxExpr(Op1Rewritten, RHSPlusOne));
+        AddRewrite(Op2, Op2Rewritten, getUMaxExpr(Op2Rewritten, RHSPlusOne));
+        break;
+      }
+      RewrittenRHS = getUMaxExpr(RewrittenLHS, RHSPlusOne);
+      break;
+    }
+    case CmpInst::ICMP_SGT: {
+      const SCEV *RHSPlusOne = getAddExpr(RHS, getOne(RHS->getType()));
+      // smin(a, b) >s c -> (a >=s c + 1) && (b >=s c + 1)
+      if (auto *SMin = dyn_cast<SCEVSMinExpr>(RewrittenLHS)) {
+        auto *Op1 = SMin->getOperand(0);
+        auto *Op2 = SMin->getOperand(1);
+        auto *Op1Rewritten = GetMaybeRewritten(Op1);
+        auto *Op2Rewritten = GetMaybeRewritten(Op2);
+        AddRewrite(Op1, Op1Rewritten, getSMaxExpr(Op1Rewritten, RHSPlusOne));
+        AddRewrite(Op2, Op2Rewritten, getSMaxExpr(Op2Rewritten, RHSPlusOne));
+        break;
+      }
+      RewrittenRHS = getSMaxExpr(RewrittenLHS, RHSPlusOne);
       break;
+    }
     case CmpInst::ICMP_UGE:
+      // umin(a, b) >=u c -> a >=u c && b >=u c
+      if (auto *UMin = dyn_cast<SCEVUMinExpr>(RewrittenLHS)) {
+        auto *Op1 = UMin->getOperand(0);
+        auto *Op2 = UMin->getOperand(1);
+        auto *Op1Rewritten = GetMaybeRewritten(Op1);
+        auto *Op2Rewritten = GetMaybeRewritten(Op2);
+        AddRewrite(Op1, Op1Rewritten, getUMaxExpr(Op1Rewritten, RHS));
+        AddRewrite(Op2, Op2Rewritten, getUMaxExpr(Op2Rewritten, RHS));
+        break;
+      }
       RewrittenRHS = getUMaxExpr(RewrittenLHS, RHS);
       break;
     case CmpInst::ICMP_SGE:
+      // smin(a, b) >=s c -> a >=s c && b >=s c
+      if (auto *SMin = dyn_cast<SCEVSMinExpr>(RewrittenLHS)) {
+        auto *Op1 = SMin->getOperand(0);
+        auto *Op2 = SMin->getOperand(1);
+        auto *Op1Rewritten = GetMaybeRewritten(Op1);
+        auto *Op2Rewritten = GetMaybeRewritten(Op2);
+        AddRewrite(Op1, Op1Rewritten, getSMaxExpr(Op1Rewritten, RHS));
+        AddRewrite(Op2, Op2Rewritten, getSMaxExpr(Op2Rewritten, RHS));
+        break;
+      }
       RewrittenRHS = getSMaxExpr(RewrittenLHS, RHS);
       break;
     case CmpInst::ICMP_EQ:
       if (isa<SCEVConstant>(RHS))
         RewrittenRHS = RHS;
       break;
     case CmpInst::ICMP_NE:
       if (isa<SCEVConstant>(RHS) &&
const SCEV *ScalarEvolution::applyLoopGuards(const SCEV *Expr, const Loop *L) {
   // First, collect information from assumptions dominating the loop.
   for (auto &AssumeVH : AC.assumptions()) {
     if (!AssumeVH)
       continue;
     auto *AssumeI = cast<CallInst>(AssumeVH);
     if (!DT.dominates(AssumeI, Header))
       continue;
     Terms.emplace_back(AssumeI->getOperand(0), true);
   }
 
   // Second, collect information from llvm.experimental.guards dominating the loop.
   auto *GuardDecl = F.getParent()->getFunction(
       Intrinsic::getName(Intrinsic::experimental_guard));
   if (GuardDecl)
     for (const auto *GU : GuardDecl->users())
       if (const auto *Guard = dyn_cast<IntrinsicInst>(GU))
         if (Guard->getFunction() == Header->getParent() && DT.dominates(Guard, Header))

llvm/test/Analysis/ScalarEvolution/max-backedge-taken-count-guard-info-rewrite-expressions.ll

 ; CHECK-NEXT:  Classifying expressions for: @rewrite_sext_slt_narrow_check
 ; CHECK-NEXT:    %smin = call i32 @llvm.smax.i32(i32 %N, i32 4)
 ; CHECK-NEXT:    --> (4 smax %N) U: [4,-2147483648) S: [4,-2147483648)
 ; CHECK-NEXT:    %ext = sext i32 %smin to i64
 ; CHECK-NEXT:    --> (zext i32 (4 smax %N) to i64) U: [4,2147483648) S: [4,2147483648)
 ; CHECK-NEXT:    %n.vec = and i64 %ext, 28
 ; CHECK-NEXT:    --> (4 * (zext i3 (trunc i64 ((zext i32 (4 smax %N) to i64) /u 4) to i3) to i64))<nuw><nsw> U: [0,29) S: [0,29)
 ; CHECK-NEXT:    %index = phi i64 [ 0, %loop.ph ], [ %index.next, %loop ]
-; CHECK-NEXT:    --> {0,+,4}<nuw><%loop> U: [0,-3) S: [-9223372036854775808,9223372036854775805) Exits: (4 * ((-4 + (4 * (zext i3 (trunc i64 ((zext i32 (4 smax %N) to i64) /u 4) to i3) to i64))<nuw><nsw>)<nsw> /u 4))<nuw> LoopDispositions: { %loop: Computable }
+; CHECK-NEXT:    --> {0,+,4}<nuw><%loop> U: [0,13) S: [0,13) Exits: (4 * ((-4 + (4 * (zext i3 (trunc i64 ((zext i32 (4 smax %N) to i64) /u 4) to i3) to i64))<nuw><nsw>)<nsw> /u 4))<nuw> LoopDispositions: { %loop: Computable }
 ; CHECK-NEXT:    %gep = getelementptr inbounds i32, ptr %arr, i64 %index
-; CHECK-NEXT:    --> {%arr,+,16}<%loop> U: full-set S: full-set Exits: ((16 * ((-4 + (4 * (zext i3 (trunc i64 ((zext i32 (4 smax %N) to i64) /u 4) to i3) to i64))<nuw><nsw>)<nsw> /u 4)) + %arr) LoopDispositions: { %loop: Computable }
+; CHECK-NEXT:    --> {%arr,+,16}<nuw><%loop> U: full-set S: full-set Exits: ((16 * ((-4 + (4 * (zext i3 (trunc i64 ((zext i32 (4 smax %N) to i64) /u 4) to i3) to i64))<nuw><nsw>)<nsw> /u 4)) + %arr) LoopDispositions: { %loop: Computable }
 ; CHECK-NEXT:    %index.next = add nuw i64 %index, 4
-; CHECK-NEXT:    --> {4,+,4}<nuw><%loop> U: [4,-3) S: [-9223372036854775808,9223372036854775805) Exits: (4 + (4 * ((-4 + (4 * (zext i3 (trunc i64 ((zext i32 (4 smax %N) to i64) /u 4) to i3) to i64))<nuw><nsw>)<nsw> /u 4))<nuw>) LoopDispositions: { %loop: Computable }
+; CHECK-NEXT:    --> {4,+,4}<nuw><%loop> U: [4,17) S: [4,17) Exits: (4 + (4 * ((-4 + (4 * (zext i3 (trunc i64 ((zext i32 (4 smax %N) to i64) /u 4) to i3) to i64))<nuw><nsw>)<nsw> /u 4))<nuw>) LoopDispositions: { %loop: Computable }
 ; CHECK-NEXT:  Determining loop execution counts for: @rewrite_sext_slt_narrow_check
 ; CHECK-NEXT:  Loop %loop: backedge-taken count is ((-4 + (4 * (zext i3 (trunc i64 ((zext i32 (4 smax %N) to i64) /u 4) to i3) to i64))<nuw><nsw>)<nsw> /u 4)
-; CHECK-NEXT:  Loop %loop: constant max backedge-taken count is 4611686018427387903
+; CHECK-NEXT:  Loop %loop: constant max backedge-taken count is 3
 ; CHECK-NEXT:  Loop %loop: symbolic max backedge-taken count is ((-4 + (4 * (zext i3 (trunc i64 ((zext i32 (4 smax %N) to i64) /u 4) to i3) to i64))<nuw><nsw>)<nsw> /u 4)
 ; CHECK-NEXT:  Loop %loop: Predicated backedge-taken count is ((-4 + (4 * (zext i3 (trunc i64 ((zext i32 (4 smax %N) to i64) /u 4) to i3) to i64))<nuw><nsw>)<nsw> /u 4)
 ; CHECK-NEXT:   Predicates:
 ; CHECK:       Loop %loop: Trip multiple is 1
 ;
 entry:
   %smin = call i32 @llvm.smax.i32(i32 %N, i32 4)
   %ext = sext i32 %smin to i64

 ; CHECK-NEXT:  Classifying expressions for: @rewrite_zext_ult_narrow_check
 ; CHECK-NEXT:    %umin = call i32 @llvm.umax.i32(i32 %N, i32 4)
 ; CHECK-NEXT:    --> (4 umax %N) U: [4,0) S: [4,0)
 ; CHECK-NEXT:    %ext = zext i32 %umin to i64
 ; CHECK-NEXT:    --> (4 umax (zext i32 %N to i64)) U: [4,4294967296) S: [4,4294967296)
 ; CHECK-NEXT:    %n.vec = and i64 %ext, 28
 ; CHECK-NEXT:    --> (4 * (zext i3 (trunc i64 ((4 umax (zext i32 %N to i64)) /u 4) to i3) to i64))<nuw><nsw> U: [0,29) S: [0,29)
 ; CHECK-NEXT:    %index = phi i64 [ 0, %loop.ph ], [ %index.next, %loop ]
-; CHECK-NEXT:    --> {0,+,4}<nuw><%loop> U: [0,-3) S: [-9223372036854775808,9223372036854775805) Exits: (4 * ((-4 + (4 * (zext i3 (trunc i64 ((4 umax (zext i32 %N to i64)) /u 4) to i3) to i64))<nuw><nsw>)<nsw> /u 4))<nuw> LoopDispositions: { %loop: Computable }
+; CHECK-NEXT:    --> {0,+,4}<nuw><%loop> U: [0,13) S: [0,13) Exits: (4 * ((-4 + (4 * (zext i3 (trunc i64 ((4 umax (zext i32 %N to i64)) /u 4) to i3) to i64))<nuw><nsw>)<nsw> /u 4))<nuw> LoopDispositions: { %loop: Computable }
 ; CHECK-NEXT:    %gep = getelementptr inbounds i32, ptr %arr, i64 %index
-; CHECK-NEXT:    --> {%arr,+,16}<%loop> U: full-set S: full-set Exits: ((16 * ((-4 + (4 * (zext i3 (trunc i64 ((4 umax (zext i32 %N to i64)) /u 4) to i3) to i64))<nuw><nsw>)<nsw> /u 4)) + %arr) LoopDispositions: { %loop: Computable }
+; CHECK-NEXT:    --> {%arr,+,16}<nuw><%loop> U: full-set S: full-set Exits: ((16 * ((-4 + (4 * (zext i3 (trunc i64 ((4 umax (zext i32 %N to i64)) /u 4) to i3) to i64))<nuw><nsw>)<nsw> /u 4)) + %arr) LoopDispositions: { %loop: Computable }
 ; CHECK-NEXT:    %index.next = add nuw i64 %index, 4
-; CHECK-NEXT:    --> {4,+,4}<nuw><%loop> U: [4,-3) S: [-9223372036854775808,9223372036854775805) Exits: (4 + (4 * ((-4 + (4 * (zext i3 (trunc i64 ((4 umax (zext i32 %N to i64)) /u 4) to i3) to i64))<nuw><nsw>)<nsw> /u 4))<nuw>) LoopDispositions: { %loop: Computable }
+; CHECK-NEXT:    --> {4,+,4}<nuw><%loop> U: [4,17) S: [4,17) Exits: (4 + (4 * ((-4 + (4 * (zext i3 (trunc i64 ((4 umax (zext i32 %N to i64)) /u 4) to i3) to i64))<nuw><nsw>)<nsw> /u 4))<nuw>) LoopDispositions: { %loop: Computable }
 ; CHECK-NEXT:  Determining loop execution counts for: @rewrite_zext_ult_narrow_check
 ; CHECK-NEXT:  Loop %loop: backedge-taken count is ((-4 + (4 * (zext i3 (trunc i64 ((4 umax (zext i32 %N to i64)) /u 4) to i3) to i64))<nuw><nsw>)<nsw> /u 4)
-; CHECK-NEXT:  Loop %loop: constant max backedge-taken count is 4611686018427387903
+; CHECK-NEXT:  Loop %loop: constant max backedge-taken count is 3
 ; CHECK-NEXT:  Loop %loop: symbolic max backedge-taken count is ((-4 + (4 * (zext i3 (trunc i64 ((4 umax (zext i32 %N to i64)) /u 4) to i3) to i64))<nuw><nsw>)<nsw> /u 4)
 ; CHECK-NEXT:  Loop %loop: Predicated backedge-taken count is ((-4 + (4 * (zext i3 (trunc i64 ((4 umax (zext i32 %N to i64)) /u 4) to i3) to i64))<nuw><nsw>)<nsw> /u 4)
 ; CHECK-NEXT:   Predicates:
 ; CHECK:       Loop %loop: Trip multiple is 1
 ;
 entry:
   %umin = call i32 @llvm.umax.i32(i32 %N, i32 4)
   %ext = zext i32 %umin to i64

 ; CHECK-NEXT:  Classifying expressions for: @rewrite_zext_ule_narrow_check
 ; CHECK-NEXT:    %umin = call i32 @llvm.umax.i32(i32 %N, i32 4)
 ; CHECK-NEXT:    --> (4 umax %N) U: [4,0) S: [4,0)
 ; CHECK-NEXT:    %ext = zext i32 %umin to i64
 ; CHECK-NEXT:    --> (4 umax (zext i32 %N to i64)) U: [4,4294967296) S: [4,4294967296)
 ; CHECK-NEXT:    %n.vec = and i64 %ext, 28
 ; CHECK-NEXT:    --> (4 * (zext i3 (trunc i64 ((4 umax (zext i32 %N to i64)) /u 4) to i3) to i64))<nuw><nsw> U: [0,29) S: [0,29)
 ; CHECK-NEXT:    %index = phi i64 [ 0, %loop.ph ], [ %index.next, %loop ]
-; CHECK-NEXT:    --> {0,+,4}<nuw><%loop> U: [0,-3) S: [-9223372036854775808,9223372036854775805) Exits: (4 * ((-4 + (4 * (zext i3 (trunc i64 ((4 umax (zext i32 %N to i64)) /u 4) to i3) to i64))<nuw><nsw>)<nsw> /u 4))<nuw> LoopDispositions: { %loop: Computable }
+; CHECK-NEXT:    --> {0,+,4}<nuw><%loop> U: [0,13) S: [0,13) Exits: (4 * ((-4 + (4 * (zext i3 (trunc i64 ((4 umax (zext i32 %N to i64)) /u 4) to i3) to i64))<nuw><nsw>)<nsw> /u 4))<nuw> LoopDispositions: { %loop: Computable }
 ; CHECK-NEXT:    %gep = getelementptr inbounds i32, ptr %arr, i64 %index
-; CHECK-NEXT:    --> {%arr,+,16}<%loop> U: full-set S: full-set Exits: ((16 * ((-4 + (4 * (zext i3 (trunc i64 ((4 umax (zext i32 %N to i64)) /u 4) to i3) to i64))<nuw><nsw>)<nsw> /u 4)) + %arr) LoopDispositions: { %loop: Computable }
+; CHECK-NEXT:    --> {%arr,+,16}<nuw><%loop> U: full-set S: full-set Exits: ((16 * ((-4 + (4 * (zext i3 (trunc i64 ((4 umax (zext i32 %N to i64)) /u 4) to i3) to i64))<nuw><nsw>)<nsw> /u 4)) + %arr) LoopDispositions: { %loop: Computable }
 ; CHECK-NEXT:    %index.next = add nuw i64 %index, 4
-; CHECK-NEXT:    --> {4,+,4}<nuw><%loop> U: [4,-3) S: [-9223372036854775808,9223372036854775805) Exits: (4 + (4 * ((-4 + (4 * (zext i3 (trunc i64 ((4 umax (zext i32 %N to i64)) /u 4) to i3) to i64))<nuw><nsw>)<nsw> /u 4))<nuw>) LoopDispositions: { %loop: Computable }
+; CHECK-NEXT:    --> {4,+,4}<nuw><%loop> U: [4,17) S: [4,17) Exits: (4 + (4 * ((-4 + (4 * (zext i3 (trunc i64 ((4 umax (zext i32 %N to i64)) /u 4) to i3) to i64))<nuw><nsw>)<nsw> /u 4))<nuw>) LoopDispositions: { %loop: Computable }
 ; CHECK-NEXT:  Determining loop execution counts for: @rewrite_zext_ule_narrow_check
 ; CHECK-NEXT:  Loop %loop: backedge-taken count is ((-4 + (4 * (zext i3 (trunc i64 ((4 umax (zext i32 %N to i64)) /u 4) to i3) to i64))<nuw><nsw>)<nsw> /u 4)
-; CHECK-NEXT:  Loop %loop: constant max backedge-taken count is 4611686018427387903
+; CHECK-NEXT:  Loop %loop: constant max backedge-taken count is 3
 ; CHECK-NEXT:  Loop %loop: symbolic max backedge-taken count is ((-4 + (4 * (zext i3 (trunc i64 ((4 umax (zext i32 %N to i64)) /u 4) to i3) to i64))<nuw><nsw>)<nsw> /u 4)
 ; CHECK-NEXT:  Loop %loop: Predicated backedge-taken count is ((-4 + (4 * (zext i3 (trunc i64 ((4 umax (zext i32 %N to i64)) /u 4) to i3) to i64))<nuw><nsw>)<nsw> /u 4)
 ; CHECK-NEXT:   Predicates:
 ; CHECK:       Loop %loop: Trip multiple is 1
 ;
 entry:
   %umin = call i32 @llvm.umax.i32(i32 %N, i32 4)
   %ext = zext i32 %umin to i64

 ; CHECK-NEXT:  Classifying expressions for: @rewrite_zext_sle_narrow_check
 ; CHECK-NEXT:    %smin = call i32 @llvm.smax.i32(i32 %N, i32 4)
 ; CHECK-NEXT:    --> (4 smax %N) U: [4,-2147483648) S: [4,-2147483648)
 ; CHECK-NEXT:    %ext = sext i32 %smin to i64
 ; CHECK-NEXT:    --> (zext i32 (4 smax %N) to i64) U: [4,2147483648) S: [4,2147483648)
 ; CHECK-NEXT:    %n.vec = and i64 %ext, 28
 ; CHECK-NEXT:    --> (4 * (zext i3 (trunc i64 ((zext i32 (4 smax %N) to i64) /u 4) to i3) to i64))<nuw><nsw> U: [0,29) S: [0,29)
 ; CHECK-NEXT:    %index = phi i64 [ 0, %loop.ph ], [ %index.next, %loop ]
-; CHECK-NEXT:    --> {0,+,4}<nuw><%loop> U: [0,-3) S: [-9223372036854775808,9223372036854775805) Exits: (4 * ((-4 + (4 * (zext i3 (trunc i64 ((zext i32 (4 smax %N) to i64) /u 4) to i3) to i64))<nuw><nsw>)<nsw> /u 4))<nuw> LoopDispositions: { %loop: Computable }
+; CHECK-NEXT:    --> {0,+,4}<nuw><%loop> U: [0,13) S: [0,13) Exits: (4 * ((-4 + (4 * (zext i3 (trunc i64 ((zext i32 (4 smax %N) to i64) /u 4) to i3) to i64))<nuw><nsw>)<nsw> /u 4))<nuw> LoopDispositions: { %loop: Computable }
 ; CHECK-NEXT:    %gep = getelementptr inbounds i32, ptr %arr, i64 %index
-; CHECK-NEXT:    --> {%arr,+,16}<%loop> U: full-set S: full-set Exits: ((16 * ((-4 + (4 * (zext i3 (trunc i64 ((zext i32 (4 smax %N) to i64) /u 4) to i3) to i64))<nuw><nsw>)<nsw> /u 4)) + %arr) LoopDispositions: { %loop: Computable }
+; CHECK-NEXT:    --> {%arr,+,16}<nuw><%loop> U: full-set S: full-set Exits: ((16 * ((-4 + (4 * (zext i3 (trunc i64 ((zext i32 (4 smax %N) to i64) /u 4) to i3) to i64))<nuw><nsw>)<nsw> /u 4)) + %arr) LoopDispositions: { %loop: Computable }
 ; CHECK-NEXT:    %index.next = add nuw i64 %index, 4
-; CHECK-NEXT:    --> {4,+,4}<nuw><%loop> U: [4,-3) S: [-9223372036854775808,9223372036854775805) Exits: (4 + (4 * ((-4 + (4 * (zext i3 (trunc i64 ((zext i32 (4 smax %N) to i64) /u 4) to i3) to i64))<nuw><nsw>)<nsw> /u 4))<nuw>) LoopDispositions: { %loop: Computable }
+; CHECK-NEXT:    --> {4,+,4}<nuw><%loop> U: [4,17) S: [4,17) Exits: (4 + (4 * ((-4 + (4 * (zext i3 (trunc i64 ((zext i32 (4 smax %N) to i64) /u 4) to i3) to i64))<nuw><nsw>)<nsw> /u 4))<nuw>) LoopDispositions: { %loop: Computable }
 ; CHECK-NEXT:  Determining loop execution counts for: @rewrite_zext_sle_narrow_check
 ; CHECK-NEXT:  Loop %loop: backedge-taken count is ((-4 + (4 * (zext i3 (trunc i64 ((zext i32 (4 smax %N) to i64) /u 4) to i3) to i64))<nuw><nsw>)<nsw> /u 4)
-; CHECK-NEXT:  Loop %loop: constant max backedge-taken count is 4611686018427387903
+; CHECK-NEXT:  Loop %loop: constant max backedge-taken count is 3
 ; CHECK-NEXT:  Loop %loop: symbolic max backedge-taken count is ((-4 + (4 * (zext i3 (trunc i64 ((zext i32 (4 smax %N) to i64) /u 4) to i3) to i64))<nuw><nsw>)<nsw> /u 4)
 ; CHECK-NEXT:  Loop %loop: Predicated backedge-taken count is ((-4 + (4 * (zext i3 (trunc i64 ((zext i32 (4 smax %N) to i64) /u 4) to i3) to i64))<nuw><nsw>)<nsw> /u 4)
 ; CHECK-NEXT:   Predicates:
 ; CHECK:       Loop %loop: Trip multiple is 1
 ;
 entry:
   %smin = call i32 @llvm.smax.i32(i32 %N, i32 4)
   %ext = sext i32 %smin to i64

 ; CHECK-NEXT:  Classifying expressions for: @rewrite_zext_uge_narrow_check
 ; CHECK-NEXT:    %umin = call i32 @llvm.umin.i32(i32 %N, i32 16)
 ; CHECK-NEXT:    --> (16 umin %N) U: [0,17) S: [0,17)
 ; CHECK-NEXT:    %ext = zext i32 %umin to i64
 ; CHECK-NEXT:    --> (16 umin (zext i32 %N to i64)) U: [0,17) S: [0,17)
 ; CHECK-NEXT:    %n.vec = and i64 %ext, 28
 ; CHECK-NEXT:    --> (4 * ((16 umin (zext i32 %N to i64)) /u 4))<nuw><nsw> U: [0,17) S: [0,17)
 ; CHECK-NEXT:    %index = phi i64 [ 0, %loop.ph ], [ %index.next, %loop ]
-; CHECK-NEXT:    --> {0,+,4}<nuw><%loop> U: [0,-3) S: [-9223372036854775808,9223372036854775805) Exits: (4 * ((-4 + (4 * ((16 umin (zext i32 %N to i64)) /u 4))<nuw><nsw>)<nsw> /u 4))<nuw> LoopDispositions: { %loop: Computable }
+; CHECK-NEXT:    --> {0,+,4}<nuw><%loop> U: [0,13) S: [0,13) Exits: (4 * ((-4 + (4 * ((16 umin (zext i32 %N to i64)) /u 4))<nuw><nsw>)<nsw> /u 4))<nuw> LoopDispositions: { %loop: Computable }
 ; CHECK-NEXT:    %gep = getelementptr inbounds i32, ptr %arr, i64 %index
-; CHECK-NEXT:    --> {%arr,+,16}<%loop> U: full-set S: full-set Exits: ((16 * ((-4 + (4 * ((16 umin (zext i32 %N to i64)) /u 4))<nuw><nsw>)<nsw> /u 4)) + %arr) LoopDispositions: { %loop: Computable }
+; CHECK-NEXT:    --> {%arr,+,16}<nuw><%loop> U: full-set S: full-set Exits: ((16 * ((-4 + (4 * ((16 umin (zext i32 %N to i64)) /u 4))<nuw><nsw>)<nsw> /u 4)) + %arr) LoopDispositions: { %loop: Computable }
 ; CHECK-NEXT:    %index.next = add nuw i64 %index, 4
-; CHECK-NEXT:    --> {4,+,4}<nuw><%loop> U: [4,-3) S: [-9223372036854775808,9223372036854775805) Exits: (4 + (4 * ((-4 + (4 * ((16 umin (zext i32 %N to i64)) /u 4))<nuw><nsw>)<nsw> /u 4))<nuw>) LoopDispositions: { %loop: Computable }
+; CHECK-NEXT:    --> {4,+,4}<nuw><%loop> U: [4,17) S: [4,17) Exits: (4 + (4 * ((-4 + (4 * ((16 umin (zext i32 %N to i64)) /u 4))<nuw><nsw>)<nsw> /u 4))<nuw>) LoopDispositions: { %loop: Computable }
 ; CHECK-NEXT:  Determining loop execution counts for: @rewrite_zext_uge_narrow_check
 ; CHECK-NEXT:  Loop %loop: backedge-taken count is ((-4 + (4 * ((16 umin (zext i32 %N to i64)) /u 4))<nuw><nsw>)<nsw> /u 4)
-; CHECK-NEXT:  Loop %loop: constant max backedge-taken count is 4611686018427387903
+; CHECK-NEXT:  Loop %loop: constant max backedge-taken count is 3
 ; CHECK-NEXT:  Loop %loop: symbolic max backedge-taken count is ((-4 + (4 * ((16 umin (zext i32 %N to i64)) /u 4))<nuw><nsw>)<nsw> /u 4)
 ; CHECK-NEXT:  Loop %loop: Predicated backedge-taken count is ((-4 + (4 * ((16 umin (zext i32 %N to i64)) /u 4))<nuw><nsw>)<nsw> /u 4)
 ; CHECK-NEXT:   Predicates:
 ; CHECK:       Loop %loop: Trip multiple is 1
 ;
 entry:
   %umin = call i32 @llvm.umin.i32(i32 %N, i32 16)
   %ext = zext i32 %umin to i64

 ; CHECK-NEXT:  Classifying expressions for: @rewrite_sext_sge_narrow_check
 ; CHECK-NEXT:    %smin = call i32 @llvm.smin.i32(i32 %N, i32 16)
 ; CHECK-NEXT:    --> (16 smin %N) U: [-2147483648,17) S: [-2147483648,17)
 ; CHECK-NEXT:    %ext = sext i32 %smin to i64
 ; CHECK-NEXT:    --> (16 smin (sext i32 %N to i64)) U: [-2147483648,17) S: [-2147483648,17)
 ; CHECK-NEXT:    %n.vec = and i64 %ext, 28
 ; CHECK-NEXT:    --> (4 * (zext i3 (trunc i64 ((16 smin (sext i32 %N to i64)) /u 4) to i3) to i64))<nuw><nsw> U: [0,29) S: [0,29)
 ; CHECK-NEXT:    %index = phi i64 [ 0, %loop.ph ], [ %index.next, %loop ]
-; CHECK-NEXT:    --> {0,+,4}<nuw><%loop> U: [0,-3) S: [-9223372036854775808,9223372036854775805) Exits: (4 * ((-4 + (4 * (zext i3 (trunc i64 ((16 smin (sext i32 %N to i64)) /u 4) to i3) to i64))<nuw><nsw>)<nsw> /u 4))<nuw> LoopDispositions: { %loop: Computable }
+; CHECK-NEXT:    --> {0,+,4}<nuw><%loop> U: [0,13) S: [0,13) Exits: (4 * ((-4 + (4 * (zext i3 (trunc i64 ((16 smin (sext i32 %N to i64)) /u 4) to i3) to i64))<nuw><nsw>)<nsw> /u 4))<nuw> LoopDispositions: { %loop: Computable }
 ; CHECK-NEXT:    %gep = getelementptr inbounds i32, ptr %arr, i64 %index
-; CHECK-NEXT:    --> {%arr,+,16}<%loop> U: full-set S: full-set Exits: ((16 * ((-4 + (4 * (zext i3 (trunc i64 ((16 smin (sext i32 %N to i64)) /u 4) to i3) to i64))<nuw><nsw>)<nsw> /u 4)) + %arr) LoopDispositions: { %loop: Computable }
+; CHECK-NEXT:    --> {%arr,+,16}<nuw><%loop> U: full-set S: full-set Exits: ((16 * ((-4 + (4 * (zext i3 (trunc i64 ((16 smin (sext i32 %N to i64)) /u 4) to i3) to i64))<nuw><nsw>)<nsw> /u 4)) + %arr) LoopDispositions: { %loop: Computable }
 ; CHECK-NEXT:    %index.next = add nuw i64 %index, 4
-; CHECK-NEXT:    --> {4,+,4}<nuw><%loop> U: [4,-3) S: [-9223372036854775808,9223372036854775805) Exits: (4 + (4 * ((-4 + (4 * (zext i3 (trunc i64 ((16 smin (sext i32 %N to i64)) /u 4) to i3) to i64))<nuw><nsw>)<nsw> /u 4))<nuw>) LoopDispositions: { %loop: Computable }
+; CHECK-NEXT:    --> {4,+,4}<nuw><%loop> U: [4,17) S: [4,17) Exits: (4 + (4 * ((-4 + (4 * (zext i3 (trunc i64 ((16 smin (sext i32 %N to i64)) /u 4) to i3) to i64))<nuw><nsw>)<nsw> /u 4))<nuw>) LoopDispositions: { %loop: Computable }
 ; CHECK-NEXT:  Determining loop execution counts for: @rewrite_sext_sge_narrow_check
 ; CHECK-NEXT:  Loop %loop: backedge-taken count is ((-4 + (4 * (zext i3 (trunc i64 ((16 smin (sext i32 %N to i64)) /u 4) to i3) to i64))<nuw><nsw>)<nsw> /u 4)
-; CHECK-NEXT:  Loop %loop: constant max backedge-taken count is 4611686018427387903
+; CHECK-NEXT:  Loop %loop: constant max backedge-taken count is 3
 ; CHECK-NEXT:  Loop %loop: symbolic max backedge-taken count is ((-4 + (4 * (zext i3 (trunc i64 ((16 smin (sext i32 %N to i64)) /u 4) to i3) to i64))<nuw><nsw>)<nsw> /u 4)
 ; CHECK-NEXT:  Loop %loop: Predicated backedge-taken count is ((-4 + (4 * (zext i3 (trunc i64 ((16 smin (sext i32 %N to i64)) /u 4) to i3) to i64))<nuw><nsw>)<nsw> /u 4)
 ; CHECK-NEXT:   Predicates:
 ; CHECK:       Loop %loop: Trip multiple is 1
 ;
 entry:
   %smin = call i32 @llvm.smin.i32(i32 %N, i32 16)
   %ext = sext i32 %smin to i64

 ; CHECK-NEXT:  Classifying expressions for: @rewrite_zext_ugt_narrow_check
 ; CHECK-NEXT:    %umin = call i32 @llvm.umin.i32(i32 %N, i32 16)
 ; CHECK-NEXT:    --> (16 umin %N) U: [0,17) S: [0,17)
 ; CHECK-NEXT:    %ext = zext i32 %umin to i64
 ; CHECK-NEXT:    --> (16 umin (zext i32 %N to i64)) U: [0,17) S: [0,17)
 ; CHECK-NEXT:    %n.vec = and i64 %ext, 28
 ; CHECK-NEXT:    --> (4 * ((16 umin (zext i32 %N to i64)) /u 4))<nuw><nsw> U: [0,17) S: [0,17)
 ; CHECK-NEXT:    %index = phi i64 [ 0, %loop.ph ], [ %index.next, %loop ]
-; CHECK-NEXT:    --> {0,+,4}<nuw><%loop> U: [0,-3) S: [-9223372036854775808,9223372036854775805) Exits: (4 * ((-4 + (4 * ((16 umin (zext i32 %N to i64)) /u 4))<nuw><nsw>)<nsw> /u 4))<nuw> LoopDispositions: { %loop: Computable }
+; CHECK-NEXT:    --> {0,+,4}<nuw><%loop> U: [0,13) S: [0,13) Exits: (4 * ((-4 + (4 * ((16 umin (zext i32 %N to i64)) /u 4))<nuw><nsw>)<nsw> /u 4))<nuw> LoopDispositions: { %loop: Computable }
 ; CHECK-NEXT:    %gep = getelementptr inbounds i32, ptr %arr, i64 %index
-; CHECK-NEXT:    --> {%arr,+,16}<%loop> U: full-set S: full-set Exits: ((16 * ((-4 + (4 * ((16 umin (zext i32 %N to i64)) /u 4))<nuw><nsw>)<nsw> /u 4)) + %arr) LoopDispositions: { %loop: Computable }
+; CHECK-NEXT:    --> {%arr,+,16}<nuw><%loop> U: full-set S: full-set Exits: ((16 * ((-4 + (4 * ((16 umin (zext i32 %N to i64)) /u 4))<nuw><nsw>)<nsw> /u 4)) + %arr) LoopDispositions: { %loop: Computable }
 ; CHECK-NEXT:    %index.next = add nuw i64 %index, 4
-; CHECK-NEXT:    --> {4,+,4}<nuw><%loop> U: [4,-3) S: [-9223372036854775808,9223372036854775805) Exits: (4 + (4 * ((-4 + (4 * ((16 umin (zext i32 %N to i64)) /u 4))<nuw><nsw>)<nsw> /u 4))<nuw>) LoopDispositions: { %loop: Computable }
+; CHECK-NEXT:    --> {4,+,4}<nuw><%loop> U: [4,17) S: [4,17) Exits: (4 + (4 * ((-4 + (4 * ((16 umin (zext i32 %N to i64)) /u 4))<nuw><nsw>)<nsw> /u 4))<nuw>) LoopDispositions: { %loop: Computable }
 ; CHECK-NEXT:  Determining loop execution counts for: @rewrite_zext_ugt_narrow_check
 ; CHECK-NEXT:  Loop %loop: backedge-taken count is ((-4 + (4 * ((16 umin (zext i32 %N to i64)) /u 4))<nuw><nsw>)<nsw> /u 4)
-; CHECK-NEXT:  Loop %loop: constant max backedge-taken count is 4611686018427387903
+; CHECK-NEXT:  Loop %loop: constant max backedge-taken count is 3
 ; CHECK-NEXT:  Loop %loop: symbolic max backedge-taken count is ((-4 + (4 * ((16 umin (zext i32 %N to i64)) /u 4))<nuw><nsw>)<nsw> /u 4)
 ; CHECK-NEXT:  Loop %loop: Predicated backedge-taken count is ((-4 + (4 * ((16 umin (zext i32 %N to i64)) /u 4))<nuw><nsw>)<nsw> /u 4)
 ; CHECK-NEXT:   Predicates:
 ; CHECK:       Loop %loop: Trip multiple is 1
 ;
 entry:
   %umin = call i32 @llvm.umin.i32(i32 %N, i32 16)
   %ext = zext i32 %umin to i64

 ; CHECK-NEXT:  Classifying expressions for: @rewrite_sext_sgt_narrow_check
 ; CHECK-NEXT:    %smin = call i32 @llvm.smin.i32(i32 %N, i32 16)
 ; CHECK-NEXT:    --> (16 smin %N) U: [-2147483648,17) S: [-2147483648,17)
 ; CHECK-NEXT:    %ext = sext i32 %smin to i64
 ; CHECK-NEXT:    --> (16 smin (sext i32 %N to i64)) U: [-2147483648,17) S: [-2147483648,17)
 ; CHECK-NEXT:    %n.vec = and i64 %ext, 28
 ; CHECK-NEXT:    --> (4 * (zext i3 (trunc i64 ((16 smin (sext i32 %N to i64)) /u 4) to i3) to i64))<nuw><nsw> U: [0,29) S: [0,29)
 ; CHECK-NEXT:    %index = phi i64 [ 0, %loop.ph ], [ %index.next, %loop ]
-; CHECK-NEXT:    --> {0,+,4}<nuw><%loop> U: [0,-3) S: [-9223372036854775808,9223372036854775805) Exits: (4 * ((-4 + (4 * (zext i3 (trunc i64 ((16 smin (sext i32 %N to i64)) /u 4) to i3) to i64))<nuw><nsw>)<nsw> /u 4))<nuw> LoopDispositions: { %loop: Computable }
+; CHECK-NEXT:    --> {0,+,4}<nuw><%loop> U: [0,13) S: [0,13) Exits: (4 * ((-4 + (4 * (zext i3 (trunc i64 ((16 smin (sext i32 %N to i64)) /u 4) to i3) to i64))<nuw><nsw>)<nsw> /u 4))<nuw> LoopDispositions: { %loop: Computable }
 ; CHECK-NEXT:    %gep = getelementptr inbounds i32, ptr %arr, i64 %index
-; CHECK-NEXT:    --> {%arr,+,16}<%loop> U: full-set S: full-set Exits: ((16 * ((-4 + (4 * (zext i3 (trunc i64 ((16 smin (sext i32 %N to i64)) /u 4) to i3) to i64))<nuw><nsw>)<nsw> /u 4)) + %arr) LoopDispositions: { %loop: Computable }
+; CHECK-NEXT:    --> {%arr,+,16}<nuw><%loop> U: full-set S: full-set Exits: ((16 * ((-4 + (4 * (zext i3 (trunc i64 ((16 smin (sext i32 %N to i64)) /u 4) to i3) to i64))<nuw><nsw>)<nsw> /u 4)) + %arr) LoopDispositions: { %loop: Computable }
 ; CHECK-NEXT:    %index.next = add nuw i64 %index, 4
-; CHECK-NEXT:    --> {4,+,4}<nuw><%loop> U: [4,-3) S: [-9223372036854775808,9223372036854775805) Exits: (4 + (4 * ((-4 + (4 * (zext i3 (trunc i64 ((16 smin (sext i32 %N to i64)) /u 4) to i3) to i64))<nuw><nsw>)<nsw> /u 4))<nuw>) LoopDispositions: { %loop: Computable }
+; CHECK-NEXT:    --> {4,+,4}<nuw><%loop> U: [4,17) S: [4,17) Exits: (4 + (4 * ((-4 + (4 * (zext i3 (trunc i64 ((16 smin (sext i32 %N to i64)) /u 4) to i3) to i64))<nuw><nsw>)<nsw> /u 4))<nuw>) LoopDispositions: { %loop: Computable }
 ; CHECK-NEXT:  Determining loop execution counts for: @rewrite_sext_sgt_narrow_check
 ; CHECK-NEXT:  Loop %loop: backedge-taken count is ((-4 + (4 * (zext i3 (trunc i64 ((16 smin (sext i32 %N to i64)) /u 4) to i3) to i64))<nuw><nsw>)<nsw> /u 4)
-; CHECK-NEXT:  Loop %loop: constant max backedge-taken count is 4611686018427387903
+; CHECK-NEXT:  Loop %loop: constant max backedge-taken count is 3
 ; CHECK-NEXT:  Loop %loop: symbolic max backedge-taken count is ((-4 + (4 * (zext i3 (trunc i64 ((16 smin (sext i32 %N to i64)) /u 4) to i3) to i64))<nuw><nsw>)<nsw> /u 4)
 ; CHECK-NEXT:  Loop %loop: Predicated backedge-taken count is ((-4 + (4 * (zext i3 (trunc i64 ((16 smin (sext i32 %N to i64)) /u 4) to i3) to i64))<nuw><nsw>)<nsw> /u 4)
 ; CHECK-NEXT:   Predicates:
 ; CHECK:       Loop %loop: Trip multiple is 1
 ;
 entry:
   %smin = call i32 @llvm.smin.i32(i32 %N, i32 16)
   %ext = sext i32 %smin to i64