[SCEV] Use umin_seq for BECount of multi-exit loops

[nikic]

Summary

When computing the BECount for multi-exit loops, we need to combine individual exit counts using umin_seq rather than umin. This is because an earlier exit may exit on the first iteration, in which case later exit expressions will not be evaluated and could be poisonous. We cannot propagate potential poison values from later exits.

In particular, this avoids the introduction of "branch on poison" UB when optimizing multi-exit loops.

[nikic]

Rebase

[nikic]

llvm/test/Transforms/IndVarSimplify/ARM/code-size.ll

140

Worth noting that this gets InstCombined to length == 0 || tmp0 >= length -- we could further fold this to freeze(tmp0) >= length if we wanted.

[nikic]

Rebase over regenerated checks.

[nikic]

llvm/test/Transforms/PhaseOrdering/AArch64/peel-multiple-unreachable-exits-for-vectorization.ll

42

I wonder whether we wouldn't be better off lowering %x umin_seq %y to umin(%x, freeze %y) rather than %x == 0 ? 0 : umin(%x, %y). The latter is a more accurate representation (the former is a refinement), but it seems like in practice the former optimizes better. Especially if %x == 0 folds in some way, we have a hard time recovering from that.

For example, here's the diff between this patch and a umin(freeze) lowering for this test file: https://gist.github.com/nikic/a61a04e3ddf52108be721434f8f2228c

[nikic]

llvm/test/Transforms/PhaseOrdering/AArch64/peel-multiple-unreachable-exits-for-vectorization.ll

42

I've put up https://reviews.llvm.org/D125372 to implement this. Ultimately I don't know how important this is either way.

[reames]

Please rebase when you get a chance. Want to see what the revised test diffs look like.

Assuming they're not terrible, I'm planning to LGTM this. This is a active miscompile, and while we might need to revert if practical perf impacts are too terrible, my default is to fix the bug and only closely examine the perf if we have evidence we need to.

llvm/lib/Analysis/ScalarEvolution.cpp

8063

This definitely deserves a comment. :)

[nikic]

Rebase over expansion change.

[fhahn]

LGTM, thanks! Might be good to wait another day with committing in case @reames has additional thoughts.

[reames]

LGTM for me as well. Sorry, I crossed this off my mental list as done, but apparently forgot to actually give the LGTM.

llvm/lib/Analysis/ScalarEvolution.cpp

for (auto &ENT : ExitNotTaken) {
     if (Preds)
       for (auto *P : ENT.Predicates)
         Preds->push_back(P);
 
     assert((Preds || ENT.hasAlwaysTruePredicate()) &&
            "Predicate should be always true!");
   }
 
-  return SE->getUMinFromMismatchedTypes(Ops);
+  return SE->getUMinFromMismatchedTypes(Ops, /* Sequential */ true);
 }
 
 /// Get the exact not taken count for this loop exit.
 const SCEV *
 ScalarEvolution::BackedgeTakenInfo::getExact(const BasicBlock *ExitingBlock,
                                              ScalarEvolution *SE) const {
   for (auto &ENT : ExitNotTaken)
     if (ENT.ExitingBlock == ExitingBlock && ENT.hasAlwaysTruePredicate())

llvm/test/Analysis/ScalarEvolution/overflow-intrinsics-trip-count.ll

 
 for.end:                                          ; preds = %for.body, %entry
   ret void
 }
 
 define void @sadd_symbolic_non_latch(i16 %start) {
 ; CHECK-LABEL: 'sadd_symbolic_non_latch'
 ; CHECK-NEXT:  Determining loop execution counts for: @sadd_symbolic_non_latch
-; CHECK-NEXT:  Loop %for.body: <multiple exits> backedge-taken count is ((230 + (-1 * %start)) umin (32767 + (-1 * %start)))
+; CHECK-NEXT:  Loop %for.body: <multiple exits> backedge-taken count is ((32767 + (-1 * %start)) umin_seq (230 + (-1 * %start)))
 ; CHECK-NEXT:    exit count for for.body: (32767 + (-1 * %start))
 ; CHECK-NEXT:    exit count for for.latch: (230 + (-1 * %start))
 ; CHECK-NEXT:  Loop %for.body: max backedge-taken count is -1
-; CHECK-NEXT:  Loop %for.body: Predicated backedge-taken count is ((230 + (-1 * %start)) umin (32767 + (-1 * %start)))
+; CHECK-NEXT:  Loop %for.body: Predicated backedge-taken count is ((32767 + (-1 * %start)) umin_seq (230 + (-1 * %start)))
 ; CHECK-NEXT:   Predicates:
 ; CHECK:       Loop %for.body: Trip multiple is 1
 ;
 entry:
   br i1 undef, label %for.end, label %for.body.preheader
 
 for.body.preheader:                               ; preds = %entry
   br label %for.body

llvm/test/Analysis/ScalarEvolution/trip-count14.ll

 
 do.end:
   ret void
 }
 
 define void @s32_max2_unpredictable_exit(i32 %n, i32 %x, i32* %p) {
 ; CHECK-LABEL: 's32_max2_unpredictable_exit'
 ; CHECK-NEXT:  Determining loop execution counts for: @s32_max2_unpredictable_exit
-; CHECK-NEXT:  Loop %do.body: <multiple exits> backedge-taken count is (((-1 * %n) + ((2 + %n) smax %n)) umin ((-1 * %n) + %x))
+; CHECK-NEXT:  Loop %do.body: <multiple exits> backedge-taken count is (((-1 * %n) + %x) umin_seq ((-1 * %n) + ((2 + %n) smax %n)))
 ; CHECK-NEXT:    exit count for do.body: ((-1 * %n) + %x)
 ; CHECK-NEXT:    exit count for if.end: ((-1 * %n) + ((2 + %n) smax %n))
 ; CHECK-NEXT:  Loop %do.body: max backedge-taken count is 2
-; CHECK-NEXT:  Loop %do.body: Predicated backedge-taken count is (((-1 * %n) + ((2 + %n) smax %n)) umin ((-1 * %n) + %x))
+; CHECK-NEXT:  Loop %do.body: Predicated backedge-taken count is (((-1 * %n) + %x) umin_seq ((-1 * %n) + ((2 + %n) smax %n)))
 ; CHECK-NEXT:   Predicates:
 ; CHECK:       Loop %do.body: Trip multiple is 1
 ;
 entry:
   %add = add i32 %n, 2
   br label %do.body
 
 do.body:

 
 do.end:
   ret void
 }
 
 define void @u32_max2_unpredictable_exit(i32 %n, i32 %x, i32* %p) {
 ; CHECK-LABEL: 'u32_max2_unpredictable_exit'
 ; CHECK-NEXT:  Determining loop execution counts for: @u32_max2_unpredictable_exit
-; CHECK-NEXT:  Loop %do.body: <multiple exits> backedge-taken count is (((-1 * %n) + ((2 + %n) umax %n)) umin ((-1 * %n) + %x))
+; CHECK-NEXT:  Loop %do.body: <multiple exits> backedge-taken count is (((-1 * %n) + %x) umin_seq ((-1 * %n) + ((2 + %n) umax %n)))
 ; CHECK-NEXT:    exit count for do.body: ((-1 * %n) + %x)
 ; CHECK-NEXT:    exit count for if.end: ((-1 * %n) + ((2 + %n) umax %n))
 ; CHECK-NEXT:  Loop %do.body: max backedge-taken count is 2
-; CHECK-NEXT:  Loop %do.body: Predicated backedge-taken count is (((-1 * %n) + ((2 + %n) umax %n)) umin ((-1 * %n) + %x))
+; CHECK-NEXT:  Loop %do.body: Predicated backedge-taken count is (((-1 * %n) + %x) umin_seq ((-1 * %n) + ((2 + %n) umax %n)))
 ; CHECK-NEXT:   Predicates:
 ; CHECK:       Loop %do.body: Trip multiple is 1
 ;
 entry:
   %add = add i32 %n, 2
   br label %do.body
 
 do.body:

llvm/test/Transforms/IndVarSimplify/ARM/code-size.ll

exit:
   ret void
 }
 
 define i32 @test1(i32* %array, i32 %length, i32 %n) #0 {
 ; CHECK-V8M-LABEL: @test1(
 ; CHECK-V8M-NEXT:  loop.preheader:
 ; CHECK-V8M-NEXT:    [[UMAX:%.*]] = call i32 @llvm.umax.i32(i32 [[N:%.*]], i32 1)
 ; CHECK-V8M-NEXT:    [[TMP0:%.*]] = add i32 [[UMAX]], -1
-; CHECK-V8M-NEXT:    [[UMIN:%.*]] = call i32 @llvm.umin.i32(i32 [[LENGTH:%.*]], i32 [[TMP0]])
-; CHECK-V8M-NEXT:    [[TMP1:%.*]] = icmp ne i32 [[LENGTH]], [[UMIN]]
+; CHECK-V8M-NEXT:    [[TMP1:%.*]] = icmp eq i32 [[LENGTH:%.*]], 0
+; CHECK-V8M-NEXT:    [[UMIN:%.*]] = call i32 @llvm.umin.i32(i32 [[TMP0]], i32 [[LENGTH]])
+; CHECK-V8M-NEXT:    [[TMP2:%.*]] = select i1 [[TMP1]], i32 0, i32 [[UMIN]]
+; CHECK-V8M-NEXT:    [[TMP3:%.*]] = icmp ne i32 [[LENGTH]], [[TMP2]]
 ; CHECK-V8M-NEXT:    br label [[LOOP:%.*]]
 ; CHECK-V8M:       loop:
 ; CHECK-V8M-NEXT:    [[LOOP_ACC:%.*]] = phi i32 [ [[LOOP_ACC_NEXT:%.*]], [[GUARDED:%.*]] ], [ 0, [[LOOP_PREHEADER:%.*]] ]
 ; CHECK-V8M-NEXT:    [[I:%.*]] = phi i32 [ [[I_NEXT:%.*]], [[GUARDED]] ], [ 0, [[LOOP_PREHEADER]] ]
-; CHECK-V8M-NEXT:    br i1 [[TMP1]], label [[GUARDED]], label [[DEOPT:%.*]], !prof !0
+; CHECK-V8M-NEXT:    br i1 [[TMP3]], label [[GUARDED]], label [[DEOPT:%.*]], !prof [[PROF0:![0-9]+]]
 ; CHECK-V8M:       deopt:
 ; CHECK-V8M-NEXT:    call void @prevent_merging()
 ; CHECK-V8M-NEXT:    ret i32 -1
 ; CHECK-V8M:       guarded:
 ; CHECK-V8M-NEXT:    [[I_I64:%.*]] = zext i32 [[I]] to i64
 ; CHECK-V8M-NEXT:    [[ARRAY_I_PTR:%.*]] = getelementptr inbounds i32, i32* [[ARRAY:%.*]], i64 [[I_I64]]
 ; CHECK-V8M-NEXT:    [[ARRAY_I:%.*]] = load i32, i32* [[ARRAY_I_PTR]], align 4
 ; CHECK-V8M-NEXT:    [[LOOP_ACC_NEXT]] = add i32 [[LOOP_ACC]], [[ARRAY_I]]
 ; CHECK-V8M-NEXT:    [[I_NEXT]] = add nuw i32 [[I]], 1
 ; CHECK-V8M-NEXT:    [[CONTINUE:%.*]] = icmp ult i32 [[I_NEXT]], [[N]]
 ; CHECK-V8M-NEXT:    br i1 [[CONTINUE]], label [[LOOP]], label [[EXIT:%.*]]
 ; CHECK-V8M:       exit:
 ; CHECK-V8M-NEXT:    [[RESULT:%.*]] = phi i32 [ [[LOOP_ACC_NEXT]], [[GUARDED]] ]
 ; CHECK-V8M-NEXT:    ret i32 [[RESULT]]
 ;
 ; CHECK-V8A-LABEL: @test1(
 ; CHECK-V8A-NEXT:  loop.preheader:
 ; CHECK-V8A-NEXT:    [[UMAX:%.*]] = call i32 @llvm.umax.i32(i32 [[N:%.*]], i32 1)
 ; CHECK-V8A-NEXT:    [[TMP0:%.*]] = add i32 [[UMAX]], -1
-; CHECK-V8A-NEXT:    [[UMIN:%.*]] = call i32 @llvm.umin.i32(i32 [[LENGTH:%.*]], i32 [[TMP0]])
-; CHECK-V8A-NEXT:    [[TMP1:%.*]] = icmp ne i32 [[LENGTH]], [[UMIN]]
+; CHECK-V8A-NEXT:    [[TMP1:%.*]] = icmp eq i32 [[LENGTH:%.*]], 0
+; CHECK-V8A-NEXT:    [[UMIN:%.*]] = call i32 @llvm.umin.i32(i32 [[TMP0]], i32 [[LENGTH]])
+; CHECK-V8A-NEXT:    [[TMP2:%.*]] = select i1 [[TMP1]], i32 0, i32 [[UMIN]]
+; CHECK-V8A-NEXT:    [[TMP3:%.*]] = icmp ne i32 [[LENGTH]], [[TMP2]]
 ; CHECK-V8A-NEXT:    br label [[LOOP:%.*]]
 ; CHECK-V8A:       loop:
 ; CHECK-V8A-NEXT:    [[LOOP_ACC:%.*]] = phi i32 [ [[LOOP_ACC_NEXT:%.*]], [[GUARDED:%.*]] ], [ 0, [[LOOP_PREHEADER:%.*]] ]
 ; CHECK-V8A-NEXT:    [[I:%.*]] = phi i32 [ [[I_NEXT:%.*]], [[GUARDED]] ], [ 0, [[LOOP_PREHEADER]] ]
-; CHECK-V8A-NEXT:    br i1 [[TMP1]], label [[GUARDED]], label [[DEOPT:%.*]], !prof !0
+; CHECK-V8A-NEXT:    br i1 [[TMP3]], label [[GUARDED]], label [[DEOPT:%.*]], !prof [[PROF0:![0-9]+]]
 ; CHECK-V8A:       deopt:
 ; CHECK-V8A-NEXT:    call void @prevent_merging()
 ; CHECK-V8A-NEXT:    ret i32 -1
 ; CHECK-V8A:       guarded:
 ; CHECK-V8A-NEXT:    [[I_I64:%.*]] = zext i32 [[I]] to i64
 ; CHECK-V8A-NEXT:    [[ARRAY_I_PTR:%.*]] = getelementptr inbounds i32, i32* [[ARRAY:%.*]], i64 [[I_I64]]
 ; CHECK-V8A-NEXT:    [[ARRAY_I:%.*]] = load i32, i32* [[ARRAY_I_PTR]], align 4
 ; CHECK-V8A-NEXT:    [[LOOP_ACC_NEXT]] = add i32 [[LOOP_ACC]], [[ARRAY_I]]

 }
 
 declare void @maythrow()
 
 define i32 @test2(i32* %array, i32 %length, i32 %n) #0 {
 ; CHECK-V8M-LABEL: @test2(
 ; CHECK-V8M-NEXT:  loop.preheader:
 ; CHECK-V8M-NEXT:    [[TMP0:%.*]] = add i32 [[N:%.*]], -1
-; CHECK-V8M-NEXT:    [[UMIN:%.*]] = call i32 @llvm.umin.i32(i32 [[LENGTH:%.*]], i32 [[TMP0]])
-; CHECK-V8M-NEXT:    [[TMP1:%.*]] = icmp ne i32 [[LENGTH]], [[UMIN]]
+; CHECK-V8M-NEXT:    [[TMP1:%.*]] = icmp eq i32 [[LENGTH:%.*]], 0
+; CHECK-V8M-NEXT:    [[UMIN:%.*]] = call i32 @llvm.umin.i32(i32 [[TMP0]], i32 [[LENGTH]])
+; CHECK-V8M-NEXT:    [[TMP2:%.*]] = select i1 [[TMP1]], i32 0, i32 [[UMIN]]
+; CHECK-V8M-NEXT:    [[TMP3:%.*]] = icmp ne i32 [[LENGTH]], [[TMP2]]
 ; CHECK-V8M-NEXT:    br label [[LOOP:%.*]]
 ; CHECK-V8M:       loop:
 ; CHECK-V8M-NEXT:    [[LOOP_ACC:%.*]] = phi i32 [ [[LOOP_ACC_NEXT:%.*]], [[GUARDED:%.*]] ], [ 0, [[LOOP_PREHEADER:%.*]] ]
 ; CHECK-V8M-NEXT:    [[I:%.*]] = phi i32 [ [[I_NEXT:%.*]], [[GUARDED]] ], [ 0, [[LOOP_PREHEADER]] ]
-; CHECK-V8M-NEXT:    br i1 [[TMP1]], label [[GUARDED]], label [[DEOPT:%.*]], !prof !0
+; CHECK-V8M-NEXT:    br i1 [[TMP3]], label [[GUARDED]], label [[DEOPT:%.*]], !prof [[PROF0]]
 ; CHECK-V8M:       deopt:
 ; CHECK-V8M-NEXT:    call void @prevent_merging()
 ; CHECK-V8M-NEXT:    ret i32 -1
 ; CHECK-V8M:       guarded:
 ; CHECK-V8M-NEXT:    [[I_I64:%.*]] = zext i32 [[I]] to i64
 ; CHECK-V8M-NEXT:    [[ARRAY_I_PTR:%.*]] = getelementptr inbounds i32, i32* [[ARRAY:%.*]], i64 [[I_I64]]
 ; CHECK-V8M-NEXT:    [[ARRAY_I:%.*]] = load i32, i32* [[ARRAY_I_PTR]], align 4
 ; CHECK-V8M-NEXT:    [[LOOP_ACC_NEXT]] = add i32 [[LOOP_ACC]], [[ARRAY_I]]
 ; CHECK-V8M-NEXT:    [[I_NEXT]] = add nuw i32 [[I]], 1
 ; CHECK-V8M-NEXT:    [[CONTINUE:%.*]] = icmp ne i32 [[I_NEXT]], [[N]]
 ; CHECK-V8M-NEXT:    br i1 [[CONTINUE]], label [[LOOP]], label [[EXIT:%.*]]
 ; CHECK-V8M:       exit:
 ; CHECK-V8M-NEXT:    [[RESULT:%.*]] = phi i32 [ [[LOOP_ACC_NEXT]], [[GUARDED]] ]
 ; CHECK-V8M-NEXT:    ret i32 [[RESULT]]
 ;
 ; CHECK-V8A-LABEL: @test2(
 ; CHECK-V8A-NEXT:  loop.preheader:
 ; CHECK-V8A-NEXT:    [[TMP0:%.*]] = add i32 [[N:%.*]], -1
-; CHECK-V8A-NEXT:    [[UMIN:%.*]] = call i32 @llvm.umin.i32(i32 [[LENGTH:%.*]], i32 [[TMP0]])
-; CHECK-V8A-NEXT:    [[TMP1:%.*]] = icmp ne i32 [[LENGTH]], [[UMIN]]
+; CHECK-V8A-NEXT:    [[TMP1:%.*]] = icmp eq i32 [[LENGTH:%.*]], 0
+; CHECK-V8A-NEXT:    [[UMIN:%.*]] = call i32 @llvm.umin.i32(i32 [[TMP0]], i32 [[LENGTH]])
+; CHECK-V8A-NEXT:    [[TMP2:%.*]] = select i1 [[TMP1]], i32 0, i32 [[UMIN]]
+; CHECK-V8A-NEXT:    [[TMP3:%.*]] = icmp ne i32 [[LENGTH]], [[TMP2]]
 ; CHECK-V8A-NEXT:    br label [[LOOP:%.*]]
 ; CHECK-V8A:       loop:
 ; CHECK-V8A-NEXT:    [[LOOP_ACC:%.*]] = phi i32 [ [[LOOP_ACC_NEXT:%.*]], [[GUARDED:%.*]] ], [ 0, [[LOOP_PREHEADER:%.*]] ]
 ; CHECK-V8A-NEXT:    [[I:%.*]] = phi i32 [ [[I_NEXT:%.*]], [[GUARDED]] ], [ 0, [[LOOP_PREHEADER]] ]
-; CHECK-V8A-NEXT:    br i1 [[TMP1]], label [[GUARDED]], label [[DEOPT:%.*]], !prof !0
+; CHECK-V8A-NEXT:    br i1 [[TMP3]], label [[GUARDED]], label [[DEOPT:%.*]], !prof [[PROF0]]
 ; CHECK-V8A:       deopt:
 ; CHECK-V8A-NEXT:    call void @prevent_merging()
 ; CHECK-V8A-NEXT:    ret i32 -1
 ; CHECK-V8A:       guarded:
 ; CHECK-V8A-NEXT:    [[I_I64:%.*]] = zext i32 [[I]] to i64
 ; CHECK-V8A-NEXT:    [[ARRAY_I_PTR:%.*]] = getelementptr inbounds i32, i32* [[ARRAY:%.*]], i64 [[I_I64]]
 ; CHECK-V8A-NEXT:    [[ARRAY_I:%.*]] = load i32, i32* [[ARRAY_I_PTR]], align 4
 ; CHECK-V8A-NEXT:    [[LOOP_ACC_NEXT]] = add i32 [[LOOP_ACC]], [[ARRAY_I]]
exit:                                             ; preds = %guarded, %entry
   %result = phi i32 [ %loop.acc.next, %guarded ]
   ret i32 %result
 }
 
 define i32 @two_range_checks(i32* %array.1, i32 %length.1, i32* %array.2, i32 %length.2, i32 %n) #0 {
 ; CHECK-V8M-LABEL: @two_range_checks(
 ; CHECK-V8M-NEXT:  loop.preheader:
 ; CHECK-V8M-NEXT:    [[UMIN:%.*]] = call i32 @llvm.umin.i32(i32 [[LENGTH_2:%.*]], i32 [[LENGTH_1:%.*]])
-; CHECK-V8M-NEXT:    [[UMIN1:%.*]] = call i32 @llvm.umin.i32(i32 [[LENGTH_2]], i32 [[LENGTH_1]])
 ; CHECK-V8M-NEXT:    [[UMAX:%.*]] = call i32 @llvm.umax.i32(i32 [[N:%.*]], i32 1)
 ; CHECK-V8M-NEXT:    [[TMP0:%.*]] = add i32 [[UMAX]], -1
-; CHECK-V8M-NEXT:    [[UMIN2:%.*]] = call i32 @llvm.umin.i32(i32 [[UMIN1]], i32 [[TMP0]])
-; CHECK-V8M-NEXT:    [[TMP1:%.*]] = icmp ne i32 [[UMIN]], [[UMIN2]]
+; CHECK-V8M-NEXT:    [[TMP1:%.*]] = icmp eq i32 [[UMIN]], 0
+; CHECK-V8M-NEXT:    [[UMIN1:%.*]] = call i32 @llvm.umin.i32(i32 [[TMP0]], i32 [[UMIN]])
+; CHECK-V8M-NEXT:    [[TMP2:%.*]] = select i1 [[TMP1]], i32 0, i32 [[UMIN1]]
+; CHECK-V8M-NEXT:    [[TMP3:%.*]] = icmp ne i32 [[UMIN]], [[TMP2]]
 ; CHECK-V8M-NEXT:    br label [[LOOP:%.*]]
 ; CHECK-V8M:       loop:
 ; CHECK-V8M-NEXT:    [[LOOP_ACC:%.*]] = phi i32 [ [[LOOP_ACC_NEXT:%.*]], [[GUARDED:%.*]] ], [ 0, [[LOOP_PREHEADER:%.*]] ]
 ; CHECK-V8M-NEXT:    [[I:%.*]] = phi i32 [ [[I_NEXT:%.*]], [[GUARDED]] ], [ 0, [[LOOP_PREHEADER]] ]
-; CHECK-V8M-NEXT:    br i1 [[TMP1]], label [[GUARDED]], label [[DEOPT:%.*]], !prof !0
+; CHECK-V8M-NEXT:    br i1 [[TMP3]], label [[GUARDED]], label [[DEOPT:%.*]], !prof [[PROF0]]
 ; CHECK-V8M:       deopt:
 ; CHECK-V8M-NEXT:    call void @prevent_merging()
 ; CHECK-V8M-NEXT:    ret i32 -1
 ; CHECK-V8M:       guarded:
 ; CHECK-V8M-NEXT:    [[I_I64:%.*]] = zext i32 [[I]] to i64
 ; CHECK-V8M-NEXT:    [[ARRAY_1_I_PTR:%.*]] = getelementptr inbounds i32, i32* [[ARRAY_1:%.*]], i64 [[I_I64]]
 ; CHECK-V8M-NEXT:    [[ARRAY_1_I:%.*]] = load i32, i32* [[ARRAY_1_I_PTR]], align 4
 ; CHECK-V8M-NEXT:    [[LOOP_ACC_1:%.*]] = add i32 [[LOOP_ACC]], [[ARRAY_1_I]]
 ; CHECK-V8M-NEXT:    [[ARRAY_2_I_PTR:%.*]] = getelementptr inbounds i32, i32* [[ARRAY_2:%.*]], i64 [[I_I64]]
 ; CHECK-V8M-NEXT:    [[ARRAY_2_I:%.*]] = load i32, i32* [[ARRAY_2_I_PTR]], align 4
 ; CHECK-V8M-NEXT:    [[LOOP_ACC_NEXT]] = add i32 [[LOOP_ACC_1]], [[ARRAY_2_I]]
 ; CHECK-V8M-NEXT:    [[I_NEXT]] = add nuw i32 [[I]], 1
 ; CHECK-V8M-NEXT:    [[CONTINUE:%.*]] = icmp ult i32 [[I_NEXT]], [[N]]
 ; CHECK-V8M-NEXT:    br i1 [[CONTINUE]], label [[LOOP]], label [[EXIT:%.*]]
 ; CHECK-V8M:       exit:
 ; CHECK-V8M-NEXT:    [[RESULT:%.*]] = phi i32 [ [[LOOP_ACC_NEXT]], [[GUARDED]] ]
 ; CHECK-V8M-NEXT:    ret i32 [[RESULT]]
 ;
 ; CHECK-V8A-LABEL: @two_range_checks(
 ; CHECK-V8A-NEXT:  loop.preheader:
 ; CHECK-V8A-NEXT:    [[UMIN:%.*]] = call i32 @llvm.umin.i32(i32 [[LENGTH_2:%.*]], i32 [[LENGTH_1:%.*]])
-; CHECK-V8A-NEXT:    [[UMIN1:%.*]] = call i32 @llvm.umin.i32(i32 [[LENGTH_2]], i32 [[LENGTH_1]])
 ; CHECK-V8A-NEXT:    [[UMAX:%.*]] = call i32 @llvm.umax.i32(i32 [[N:%.*]], i32 1)
 ; CHECK-V8A-NEXT:    [[TMP0:%.*]] = add i32 [[UMAX]], -1
-; CHECK-V8A-NEXT:    [[UMIN2:%.*]] = call i32 @llvm.umin.i32(i32 [[UMIN1]], i32 [[TMP0]])
-; CHECK-V8A-NEXT:    [[TMP1:%.*]] = icmp ne i32 [[UMIN]], [[UMIN2]]
+; CHECK-V8A-NEXT:    [[TMP1:%.*]] = icmp eq i32 [[UMIN]], 0
+; CHECK-V8A-NEXT:    [[UMIN1:%.*]] = call i32 @llvm.umin.i32(i32 [[TMP0]], i32 [[UMIN]])
+; CHECK-V8A-NEXT:    [[TMP2:%.*]] = select i1 [[TMP1]], i32 0, i32 [[UMIN1]]
+; CHECK-V8A-NEXT:    [[TMP3:%.*]] = icmp ne i32 [[UMIN]], [[TMP2]]
 ; CHECK-V8A-NEXT:    br label [[LOOP:%.*]]
 ; CHECK-V8A:       loop:
 ; CHECK-V8A-NEXT:    [[LOOP_ACC:%.*]] = phi i32 [ [[LOOP_ACC_NEXT:%.*]], [[GUARDED:%.*]] ], [ 0, [[LOOP_PREHEADER:%.*]] ]
 ; CHECK-V8A-NEXT:    [[I:%.*]] = phi i32 [ [[I_NEXT:%.*]], [[GUARDED]] ], [ 0, [[LOOP_PREHEADER]] ]
-; CHECK-V8A-NEXT:    br i1 [[TMP1]], label [[GUARDED]], label [[DEOPT:%.*]], !prof !0
+; CHECK-V8A-NEXT:    br i1 [[TMP3]], label [[GUARDED]], label [[DEOPT:%.*]], !prof [[PROF0]]
 ; CHECK-V8A:       deopt:
 ; CHECK-V8A-NEXT:    call void @prevent_merging()
 ; CHECK-V8A-NEXT:    ret i32 -1
 ; CHECK-V8A:       guarded:
 ; CHECK-V8A-NEXT:    [[I_I64:%.*]] = zext i32 [[I]] to i64
 ; CHECK-V8A-NEXT:    [[ARRAY_1_I_PTR:%.*]] = getelementptr inbounds i32, i32* [[ARRAY_1:%.*]], i64 [[I_I64]]
 ; CHECK-V8A-NEXT:    [[ARRAY_1_I:%.*]] = load i32, i32* [[ARRAY_1_I_PTR]], align 4
 ; CHECK-V8A-NEXT:    [[LOOP_ACC_1:%.*]] = add i32 [[LOOP_ACC]], [[ARRAY_1_I]]
exit:                                             ; preds = %guarded, %entry
   ret i32 %result
 }
 
 define i32 @three_range_checks(i32* %array.1, i32 %length.1, i32* %array.2, i32 %length.2, i32* %array.3, i32 %length.3, i32 %n) #0 {
 ; CHECK-V8M-LABEL: @three_range_checks(
 ; CHECK-V8M-NEXT:  loop.preheader:
 ; CHECK-V8M-NEXT:    [[UMIN:%.*]] = call i32 @llvm.umin.i32(i32 [[LENGTH_3:%.*]], i32 [[LENGTH_2:%.*]])
 ; CHECK-V8M-NEXT:    [[UMIN1:%.*]] = call i32 @llvm.umin.i32(i32 [[UMIN]], i32 [[LENGTH_1:%.*]])
-; CHECK-V8M-NEXT:    [[UMIN2:%.*]] = call i32 @llvm.umin.i32(i32 [[LENGTH_3]], i32 [[LENGTH_2]])
-; CHECK-V8M-NEXT:    [[UMIN3:%.*]] = call i32 @llvm.umin.i32(i32 [[UMIN2]], i32 [[LENGTH_1]])
 ; CHECK-V8M-NEXT:    [[UMAX:%.*]] = call i32 @llvm.umax.i32(i32 [[N:%.*]], i32 1)
 ; CHECK-V8M-NEXT:    [[TMP0:%.*]] = add i32 [[UMAX]], -1
-; CHECK-V8M-NEXT:    [[UMIN4:%.*]] = call i32 @llvm.umin.i32(i32 [[UMIN3]], i32 [[TMP0]])
-; CHECK-V8M-NEXT:    [[TMP1:%.*]] = icmp ne i32 [[UMIN1]], [[UMIN4]]
+; CHECK-V8M-NEXT:    [[TMP1:%.*]] = icmp eq i32 [[UMIN1]], 0
+; CHECK-V8M-NEXT:    [[UMIN2:%.*]] = call i32 @llvm.umin.i32(i32 [[TMP0]], i32 [[UMIN1]])
+; CHECK-V8M-NEXT:    [[TMP2:%.*]] = select i1 [[TMP1]], i32 0, i32 [[UMIN2]]
+; CHECK-V8M-NEXT:    [[TMP3:%.*]] = icmp ne i32 [[UMIN1]], [[TMP2]]
 ; CHECK-V8M-NEXT:    br label [[LOOP:%.*]]
 ; CHECK-V8M:       loop:
 ; CHECK-V8M-NEXT:    [[LOOP_ACC:%.*]] = phi i32 [ [[LOOP_ACC_NEXT:%.*]], [[GUARDED:%.*]] ], [ 0, [[LOOP_PREHEADER:%.*]] ]
 ; CHECK-V8M-NEXT:    [[I:%.*]] = phi i32 [ [[I_NEXT:%.*]], [[GUARDED]] ], [ 0, [[LOOP_PREHEADER]] ]
-; CHECK-V8M-NEXT:    br i1 [[TMP1]], label [[GUARDED]], label [[DEOPT:%.*]], !prof !0
+; CHECK-V8M-NEXT:    br i1 [[TMP3]], label [[GUARDED]], label [[DEOPT:%.*]], !prof [[PROF0]]
 ; CHECK-V8M:       deopt:
 ; CHECK-V8M-NEXT:    call void @prevent_merging()
 ; CHECK-V8M-NEXT:    ret i32 -1
 ; CHECK-V8M:       guarded:
 ; CHECK-V8M-NEXT:    [[I_I64:%.*]] = zext i32 [[I]] to i64
 ; CHECK-V8M-NEXT:    [[ARRAY_1_I_PTR:%.*]] = getelementptr inbounds i32, i32* [[ARRAY_1:%.*]], i64 [[I_I64]]
 ; CHECK-V8M-NEXT:    [[ARRAY_1_I:%.*]] = load i32, i32* [[ARRAY_1_I_PTR]], align 4
 ; CHECK-V8M-NEXT:    [[LOOP_ACC_1:%.*]] = add i32 [[LOOP_ACC]], [[ARRAY_1_I]]

 ; CHECK-V8M:       exit:
 ; CHECK-V8M-NEXT:    [[RESULT:%.*]] = phi i32 [ [[LOOP_ACC_NEXT]], [[GUARDED]] ]
 ; CHECK-V8M-NEXT:    ret i32 [[RESULT]]
 ;
 ; CHECK-V8A-LABEL: @three_range_checks(
 ; CHECK-V8A-NEXT:  loop.preheader:
 ; CHECK-V8A-NEXT:    [[UMIN:%.*]] = call i32 @llvm.umin.i32(i32 [[LENGTH_3:%.*]], i32 [[LENGTH_2:%.*]])
 ; CHECK-V8A-NEXT:    [[UMIN1:%.*]] = call i32 @llvm.umin.i32(i32 [[UMIN]], i32 [[LENGTH_1:%.*]])
-; CHECK-V8A-NEXT:    [[UMIN2:%.*]] = call i32 @llvm.umin.i32(i32 [[LENGTH_3]], i32 [[LENGTH_2]])
-; CHECK-V8A-NEXT:    [[UMIN3:%.*]] = call i32 @llvm.umin.i32(i32 [[UMIN2]], i32 [[LENGTH_1]])
 ; CHECK-V8A-NEXT:    [[UMAX:%.*]] = call i32 @llvm.umax.i32(i32 [[N:%.*]], i32 1)
 ; CHECK-V8A-NEXT:    [[TMP0:%.*]] = add i32 [[UMAX]], -1
-; CHECK-V8A-NEXT:    [[UMIN4:%.*]] = call i32 @llvm.umin.i32(i32 [[UMIN3]], i32 [[TMP0]])
-; CHECK-V8A-NEXT:    [[TMP1:%.*]] = icmp ne i32 [[UMIN1]], [[UMIN4]]
+; CHECK-V8A-NEXT:    [[TMP1:%.*]] = icmp eq i32 [[UMIN1]], 0
+; CHECK-V8A-NEXT:    [[UMIN2:%.*]] = call i32 @llvm.umin.i32(i32 [[TMP0]], i32 [[UMIN1]])
+; CHECK-V8A-NEXT:    [[TMP2:%.*]] = select i1 [[TMP1]], i32 0, i32 [[UMIN2]]
+; CHECK-V8A-NEXT:    [[TMP3:%.*]] = icmp ne i32 [[UMIN1]], [[TMP2]]
 ; CHECK-V8A-NEXT:    br label [[LOOP:%.*]]
 ; CHECK-V8A:       loop:
 ; CHECK-V8A-NEXT:    [[LOOP_ACC:%.*]] = phi i32 [ [[LOOP_ACC_NEXT:%.*]], [[GUARDED:%.*]] ], [ 0, [[LOOP_PREHEADER:%.*]] ]
 ; CHECK-V8A-NEXT:    [[I:%.*]] = phi i32 [ [[I_NEXT:%.*]], [[GUARDED]] ], [ 0, [[LOOP_PREHEADER]] ]
-; CHECK-V8A-NEXT:    br i1 [[TMP1]], label [[GUARDED]], label [[DEOPT:%.*]], !prof !0
+; CHECK-V8A-NEXT:    br i1 [[TMP3]], label [[GUARDED]], label [[DEOPT:%.*]], !prof [[PROF0]]
 ; CHECK-V8A:       deopt:
 ; CHECK-V8A-NEXT:    call void @prevent_merging()
 ; CHECK-V8A-NEXT:    ret i32 -1
 ; CHECK-V8A:       guarded:
 ; CHECK-V8A-NEXT:    [[I_I64:%.*]] = zext i32 [[I]] to i64
 ; CHECK-V8A-NEXT:    [[ARRAY_1_I_PTR:%.*]] = getelementptr inbounds i32, i32* [[ARRAY_1:%.*]], i64 [[I_I64]]
 ; CHECK-V8A-NEXT:    [[ARRAY_1_I:%.*]] = load i32, i32* [[ARRAY_1_I_PTR]], align 4
 ; CHECK-V8A-NEXT:    [[LOOP_ACC_1:%.*]] = add i32 [[LOOP_ACC]], [[ARRAY_1_I]]
exit:                                             ; preds = %guarded, %entry
   %result = phi i32 [ %loop.acc.next, %guarded ]
   ret i32 %result
 }
 
 ; Analogous to the above, but with two distinct branches (on different conditions)
 define i32 @distinct_checks(i32* %array.1, i32 %length.1, i32* %array.2, i32 %length.2, i32* %array.3, i32 %length.3, i32 %n) #0 {
 ; CHECK-V8M-LABEL: @distinct_checks(
 ; CHECK-V8M-NEXT:  loop.preheader:
-; CHECK-V8M-NEXT:    [[UMIN:%.*]] = call i32 @llvm.umin.i32(i32 [[LENGTH_2:%.*]], i32 [[LENGTH_1:%.*]])
 ; CHECK-V8M-NEXT:    [[UMAX:%.*]] = call i32 @llvm.umax.i32(i32 [[N:%.*]], i32 1)
 ; CHECK-V8M-NEXT:    [[TMP0:%.*]] = add i32 [[UMAX]], -1
-; CHECK-V8M-NEXT:    [[UMIN1:%.*]] = call i32 @llvm.umin.i32(i32 [[UMIN]], i32 [[TMP0]])
-; CHECK-V8M-NEXT:    [[TMP1:%.*]] = icmp ne i32 [[LENGTH_1]], [[UMIN1]]
-; CHECK-V8M-NEXT:    [[TMP2:%.*]] = icmp ne i32 [[LENGTH_2]], [[UMIN1]]
+; CHECK-V8M-NEXT:    [[TMP1:%.*]] = icmp eq i32 [[LENGTH_1:%.*]], 0
+; CHECK-V8M-NEXT:    [[TMP2:%.*]] = icmp eq i32 [[LENGTH_2:%.*]], 0
+; CHECK-V8M-NEXT:    [[TMP3:%.*]] = select i1 [[TMP1]], i1 true, i1 [[TMP2]]
+; CHECK-V8M-NEXT:    [[UMIN:%.*]] = call i32 @llvm.umin.i32(i32 [[TMP0]], i32 [[LENGTH_2]])
+; CHECK-V8M-NEXT:    [[UMIN1:%.*]] = call i32 @llvm.umin.i32(i32 [[UMIN]], i32 [[LENGTH_1]])
+; CHECK-V8M-NEXT:    [[TMP4:%.*]] = select i1 [[TMP3]], i32 0, i32 [[UMIN1]]
+; CHECK-V8M-NEXT:    [[TMP5:%.*]] = icmp ne i32 [[LENGTH_1]], [[TMP4]]
+; CHECK-V8M-NEXT:    [[TMP6:%.*]] = icmp ne i32 [[LENGTH_2]], [[TMP4]]
 ; CHECK-V8M-NEXT:    br label [[LOOP:%.*]]
 ; CHECK-V8M:       loop:
 ; CHECK-V8M-NEXT:    [[LOOP_ACC:%.*]] = phi i32 [ [[LOOP_ACC_NEXT:%.*]], [[GUARDED1:%.*]] ], [ 0, [[LOOP_PREHEADER:%.*]] ]
 ; CHECK-V8M-NEXT:    [[I:%.*]] = phi i32 [ [[I_NEXT:%.*]], [[GUARDED1]] ], [ 0, [[LOOP_PREHEADER]] ]
-; CHECK-V8M-NEXT:    br i1 [[TMP1]], label [[GUARDED:%.*]], label [[DEOPT:%.*]], !prof !0
+; CHECK-V8M-NEXT:    br i1 [[TMP5]], label [[GUARDED:%.*]], label [[DEOPT:%.*]], !prof [[PROF0]]
 ; CHECK-V8M:       deopt:
 ; CHECK-V8M-NEXT:    call void @prevent_merging()
 ; CHECK-V8M-NEXT:    ret i32 -1
 ; CHECK-V8M:       guarded:
 ; CHECK-V8M-NEXT:    [[I_I64:%.*]] = zext i32 [[I]] to i64
 ; CHECK-V8M-NEXT:    [[ARRAY_1_I_PTR:%.*]] = getelementptr inbounds i32, i32* [[ARRAY_1:%.*]], i64 [[I_I64]]
 ; CHECK-V8M-NEXT:    [[ARRAY_1_I:%.*]] = load i32, i32* [[ARRAY_1_I_PTR]], align 4
 ; CHECK-V8M-NEXT:    [[LOOP_ACC_1:%.*]] = add i32 [[LOOP_ACC]], [[ARRAY_1_I]]
-; CHECK-V8M-NEXT:    br i1 [[TMP2]], label [[GUARDED1]], label [[DEOPT2:%.*]], !prof !0
+; CHECK-V8M-NEXT:    br i1 [[TMP6]], label [[GUARDED1]], label [[DEOPT2:%.*]], !prof [[PROF0]]
 ; CHECK-V8M:       deopt2:
 ; CHECK-V8M-NEXT:    call void @prevent_merging()
 ; CHECK-V8M-NEXT:    ret i32 -1
 ; CHECK-V8M:       guarded1:
 ; CHECK-V8M-NEXT:    [[ARRAY_3_I_PTR:%.*]] = getelementptr inbounds i32, i32* [[ARRAY_3:%.*]], i64 [[I_I64]]
 ; CHECK-V8M-NEXT:    [[ARRAY_3_I:%.*]] = load i32, i32* [[ARRAY_3_I_PTR]], align 4
 ; CHECK-V8M-NEXT:    [[LOOP_ACC_NEXT]] = add i32 [[LOOP_ACC_1]], [[ARRAY_3_I]]
 ; CHECK-V8M-NEXT:    [[I_NEXT]] = add nuw i32 [[I]], 1
 ; CHECK-V8M-NEXT:    [[CONTINUE:%.*]] = icmp ult i32 [[I_NEXT]], [[N]]
 ; CHECK-V8M-NEXT:    br i1 [[CONTINUE]], label [[LOOP]], label [[EXIT:%.*]]
 ; CHECK-V8M:       exit:
 ; CHECK-V8M-NEXT:    [[RESULT:%.*]] = phi i32 [ [[LOOP_ACC_NEXT]], [[GUARDED1]] ]
 ; CHECK-V8M-NEXT:    ret i32 [[RESULT]]
 ;
 ; CHECK-V8A-LABEL: @distinct_checks(
 ; CHECK-V8A-NEXT:  loop.preheader:
-; CHECK-V8A-NEXT:    [[UMIN:%.*]] = call i32 @llvm.umin.i32(i32 [[LENGTH_2:%.*]], i32 [[LENGTH_1:%.*]])
 ; CHECK-V8A-NEXT:    [[UMAX:%.*]] = call i32 @llvm.umax.i32(i32 [[N:%.*]], i32 1)
 ; CHECK-V8A-NEXT:    [[TMP0:%.*]] = add i32 [[UMAX]], -1
-; CHECK-V8A-NEXT:    [[UMIN1:%.*]] = call i32 @llvm.umin.i32(i32 [[UMIN]], i32 [[TMP0]])
-; CHECK-V8A-NEXT:    [[TMP1:%.*]] = icmp ne i32 [[LENGTH_1]], [[UMIN1]]
-; CHECK-V8A-NEXT:    [[TMP2:%.*]] = icmp ne i32 [[LENGTH_2]], [[UMIN1]]
+; CHECK-V8A-NEXT:    [[TMP1:%.*]] = icmp eq i32 [[LENGTH_1:%.*]], 0
+; CHECK-V8A-NEXT:    [[TMP2:%.*]] = icmp eq i32 [[LENGTH_2:%.*]], 0
+; CHECK-V8A-NEXT:    [[TMP3:%.*]] = select i1 [[TMP1]], i1 true, i1 [[TMP2]]
+; CHECK-V8A-NEXT:    [[UMIN:%.*]] = call i32 @llvm.umin.i32(i32 [[TMP0]], i32 [[LENGTH_2]])
+; CHECK-V8A-NEXT:    [[UMIN1:%.*]] = call i32 @llvm.umin.i32(i32 [[UMIN]], i32 [[LENGTH_1]])
+; CHECK-V8A-NEXT:    [[TMP4:%.*]] = select i1 [[TMP3]], i32 0, i32 [[UMIN1]]
+; CHECK-V8A-NEXT:    [[TMP5:%.*]] = icmp ne i32 [[LENGTH_1]], [[TMP4]]
+; CHECK-V8A-NEXT:    [[TMP6:%.*]] = icmp ne i32 [[LENGTH_2]], [[TMP4]]
 ; CHECK-V8A-NEXT:    br label [[LOOP:%.*]]
 ; CHECK-V8A:       loop:
 ; CHECK-V8A-NEXT:    [[LOOP_ACC:%.*]] = phi i32 [ [[LOOP_ACC_NEXT:%.*]], [[GUARDED1:%.*]] ], [ 0, [[LOOP_PREHEADER:%.*]] ]
 ; CHECK-V8A-NEXT:    [[I:%.*]] = phi i32 [ [[I_NEXT:%.*]], [[GUARDED1]] ], [ 0, [[LOOP_PREHEADER]] ]
-; CHECK-V8A-NEXT:    br i1 [[TMP1]], label [[GUARDED:%.*]], label [[DEOPT:%.*]], !prof !0
+; CHECK-V8A-NEXT:    br i1 [[TMP5]], label [[GUARDED:%.*]], label [[DEOPT:%.*]], !prof [[PROF0]]
 ; CHECK-V8A:       deopt:
 ; CHECK-V8A-NEXT:    call void @prevent_merging()
 ; CHECK-V8A-NEXT:    ret i32 -1
 ; CHECK-V8A:       guarded:
 ; CHECK-V8A-NEXT:    [[I_I64:%.*]] = zext i32 [[I]] to i64
 ; CHECK-V8A-NEXT:    [[ARRAY_1_I_PTR:%.*]] = getelementptr inbounds i32, i32* [[ARRAY_1:%.*]], i64 [[I_I64]]
 ; CHECK-V8A-NEXT:    [[ARRAY_1_I:%.*]] = load i32, i32* [[ARRAY_1_I_PTR]], align 4
 ; CHECK-V8A-NEXT:    [[LOOP_ACC_1:%.*]] = add i32 [[LOOP_ACC]], [[ARRAY_1_I]]
-; CHECK-V8A-NEXT:    br i1 [[TMP2]], label [[GUARDED1]], label [[DEOPT2:%.*]], !prof !0
+; CHECK-V8A-NEXT:    br i1 [[TMP6]], label [[GUARDED1]], label [[DEOPT2:%.*]], !prof [[PROF0]]
 ; CHECK-V8A:       deopt2:
 ; CHECK-V8A-NEXT:    call void @prevent_merging()
 ; CHECK-V8A-NEXT:    ret i32 -1
 ; CHECK-V8A:       guarded1:
 ; CHECK-V8A-NEXT:    [[ARRAY_3_I_PTR:%.*]] = getelementptr inbounds i32, i32* [[ARRAY_3:%.*]], i64 [[I_I64]]
 ; CHECK-V8A-NEXT:    [[ARRAY_3_I:%.*]] = load i32, i32* [[ARRAY_3_I_PTR]], align 4
 ; CHECK-V8A-NEXT:    [[LOOP_ACC_NEXT]] = add i32 [[LOOP_ACC_1]], [[ARRAY_3_I]]
 ; CHECK-V8A-NEXT:    [[I_NEXT]] = add nuw i32 [[I]], 1
exit:
   ret i32 %result
 }
 
 define i32 @duplicate_checks(i32* %array.1, i32* %array.2, i32* %array.3, i32 %length, i32 %n) #0 {
 ; CHECK-V8M-LABEL: @duplicate_checks(
 ; CHECK-V8M-NEXT:  loop.preheader:
 ; CHECK-V8M-NEXT:    [[UMAX:%.*]] = call i32 @llvm.umax.i32(i32 [[N:%.*]], i32 1)
 ; CHECK-V8M-NEXT:    [[TMP0:%.*]] = add i32 [[UMAX]], -1
-; CHECK-V8M-NEXT:    [[UMIN:%.*]] = call i32 @llvm.umin.i32(i32 [[LENGTH:%.*]], i32 [[TMP0]])
-; CHECK-V8M-NEXT:    [[TMP1:%.*]] = icmp ne i32 [[LENGTH]], [[UMIN]]
+; CHECK-V8M-NEXT:    [[TMP1:%.*]] = icmp eq i32 [[LENGTH:%.*]], 0
+; CHECK-V8M-NEXT:    [[UMIN:%.*]] = call i32 @llvm.umin.i32(i32 [[TMP0]], i32 [[LENGTH]])
+; CHECK-V8M-NEXT:    [[TMP2:%.*]] = select i1 [[TMP1]], i32 0, i32 [[UMIN]]
+; CHECK-V8M-NEXT:    [[TMP3:%.*]] = icmp ne i32 [[LENGTH]], [[TMP2]]
 ; CHECK-V8M-NEXT:    br label [[LOOP:%.*]]
 ; CHECK-V8M:       loop:
 ; CHECK-V8M-NEXT:    [[LOOP_ACC:%.*]] = phi i32 [ [[LOOP_ACC_NEXT:%.*]], [[GUARDED1:%.*]] ], [ 0, [[LOOP_PREHEADER:%.*]] ]
 ; CHECK-V8M-NEXT:    [[I:%.*]] = phi i32 [ [[I_NEXT:%.*]], [[GUARDED1]] ], [ 0, [[LOOP_PREHEADER]] ]
-; CHECK-V8M-NEXT:    br i1 [[TMP1]], label [[GUARDED:%.*]], label [[DEOPT:%.*]], !prof !0
+; CHECK-V8M-NEXT:    br i1 [[TMP3]], label [[GUARDED:%.*]], label [[DEOPT:%.*]], !prof [[PROF0]]
 ; CHECK-V8M:       deopt:
 ; CHECK-V8M-NEXT:    call void @prevent_merging()
 ; CHECK-V8M-NEXT:    ret i32 -1
 ; CHECK-V8M:       guarded:
 ; CHECK-V8M-NEXT:    [[I_I64:%.*]] = zext i32 [[I]] to i64
 ; CHECK-V8M-NEXT:    [[ARRAY_1_I_PTR:%.*]] = getelementptr inbounds i32, i32* [[ARRAY_1:%.*]], i64 [[I_I64]]
 ; CHECK-V8M-NEXT:    [[ARRAY_1_I:%.*]] = load i32, i32* [[ARRAY_1_I_PTR]], align 4
 ; CHECK-V8M-NEXT:    [[LOOP_ACC_1:%.*]] = add i32 [[LOOP_ACC]], [[ARRAY_1_I]]
-; CHECK-V8M-NEXT:    br i1 true, label [[GUARDED1]], label [[DEOPT2:%.*]], !prof !0
+; CHECK-V8M-NEXT:    br i1 true, label [[GUARDED1]], label [[DEOPT2:%.*]], !prof [[PROF0]]
 ; CHECK-V8M:       deopt2:
 ; CHECK-V8M-NEXT:    call void @prevent_merging()
 ; CHECK-V8M-NEXT:    ret i32 -1
 ; CHECK-V8M:       guarded1:
 ; CHECK-V8M-NEXT:    [[ARRAY_3_I_PTR:%.*]] = getelementptr inbounds i32, i32* [[ARRAY_3:%.*]], i64 [[I_I64]]
 ; CHECK-V8M-NEXT:    [[ARRAY_3_I:%.*]] = load i32, i32* [[ARRAY_3_I_PTR]], align 4
 ; CHECK-V8M-NEXT:    [[LOOP_ACC_NEXT]] = add i32 [[LOOP_ACC_1]], [[ARRAY_3_I]]
 ; CHECK-V8M-NEXT:    [[I_NEXT]] = add nuw i32 [[I]], 1
 ; CHECK-V8M-NEXT:    [[CONTINUE:%.*]] = icmp ult i32 [[I_NEXT]], [[N]]
 ; CHECK-V8M-NEXT:    br i1 [[CONTINUE]], label [[LOOP]], label [[EXIT:%.*]]
 ; CHECK-V8M:       exit:
 ; CHECK-V8M-NEXT:    [[RESULT:%.*]] = phi i32 [ [[LOOP_ACC_NEXT]], [[GUARDED1]] ]
 ; CHECK-V8M-NEXT:    ret i32 [[RESULT]]
 ;
 ; CHECK-V8A-LABEL: @duplicate_checks(
 ; CHECK-V8A-NEXT:  loop.preheader:
 ; CHECK-V8A-NEXT:    [[UMAX:%.*]] = call i32 @llvm.umax.i32(i32 [[N:%.*]], i32 1)
 ; CHECK-V8A-NEXT:    [[TMP0:%.*]] = add i32 [[UMAX]], -1
-; CHECK-V8A-NEXT:    [[UMIN:%.*]] = call i32 @llvm.umin.i32(i32 [[LENGTH:%.*]], i32 [[TMP0]])
-; CHECK-V8A-NEXT:    [[TMP1:%.*]] = icmp ne i32 [[LENGTH]], [[UMIN]]
+; CHECK-V8A-NEXT:    [[TMP1:%.*]] = icmp eq i32 [[LENGTH:%.*]], 0
+; CHECK-V8A-NEXT:    [[UMIN:%.*]] = call i32 @llvm.umin.i32(i32 [[TMP0]], i32 [[LENGTH]])
+; CHECK-V8A-NEXT:    [[TMP2:%.*]] = select i1 [[TMP1]], i32 0, i32 [[UMIN]]
+; CHECK-V8A-NEXT:    [[TMP3:%.*]] = icmp ne i32 [[LENGTH]], [[TMP2]]
 ; CHECK-V8A-NEXT:    br label [[LOOP:%.*]]
 ; CHECK-V8A:       loop:
 ; CHECK-V8A-NEXT:    [[LOOP_ACC:%.*]] = phi i32 [ [[LOOP_ACC_NEXT:%.*]], [[GUARDED1:%.*]] ], [ 0, [[LOOP_PREHEADER:%.*]] ]
 ; CHECK-V8A-NEXT:    [[I:%.*]] = phi i32 [ [[I_NEXT:%.*]], [[GUARDED1]] ], [ 0, [[LOOP_PREHEADER]] ]
-; CHECK-V8A-NEXT:    br i1 [[TMP1]], label [[GUARDED:%.*]], label [[DEOPT:%.*]], !prof !0
+; CHECK-V8A-NEXT:    br i1 [[TMP3]], label [[GUARDED:%.*]], label [[DEOPT:%.*]], !prof [[PROF0]]
 ; CHECK-V8A:       deopt:
 ; CHECK-V8A-NEXT:    call void @prevent_merging()
 ; CHECK-V8A-NEXT:    ret i32 -1
 ; CHECK-V8A:       guarded:
 ; CHECK-V8A-NEXT:    [[I_I64:%.*]] = zext i32 [[I]] to i64
 ; CHECK-V8A-NEXT:    [[ARRAY_1_I_PTR:%.*]] = getelementptr inbounds i32, i32* [[ARRAY_1:%.*]], i64 [[I_I64]]
 ; CHECK-V8A-NEXT:    [[ARRAY_1_I:%.*]] = load i32, i32* [[ARRAY_1_I_PTR]], align 4
 ; CHECK-V8A-NEXT:    [[LOOP_ACC_1:%.*]] = add i32 [[LOOP_ACC]], [[ARRAY_1_I]]
-; CHECK-V8A-NEXT:    br i1 true, label [[GUARDED1]], label [[DEOPT2:%.*]], !prof !0
+; CHECK-V8A-NEXT:    br i1 true, label [[GUARDED1]], label [[DEOPT2:%.*]], !prof [[PROF0]]
 ; CHECK-V8A:       deopt2:
 ; CHECK-V8A-NEXT:    call void @prevent_merging()
 ; CHECK-V8A-NEXT:    ret i32 -1
 ; CHECK-V8A:       guarded1:
 ; CHECK-V8A-NEXT:    [[ARRAY_3_I_PTR:%.*]] = getelementptr inbounds i32, i32* [[ARRAY_3:%.*]], i64 [[I_I64]]
 ; CHECK-V8A-NEXT:    [[ARRAY_3_I:%.*]] = load i32, i32* [[ARRAY_3_I_PTR]], align 4
 ; CHECK-V8A-NEXT:    [[LOOP_ACC_NEXT]] = add i32 [[LOOP_ACC_1]], [[ARRAY_3_I]]
 ; CHECK-V8A-NEXT:    [[I_NEXT]] = add nuw i32 [[I]], 1

 
 ; Demonstrate that this approach works with IVs of different steps, and types
 ; This version uses a manually lftred exit condition to work around an issue described
 ; in detail on next test.
 define i32 @different_ivs(i32* %array, i32 %length, i32 %n) #0 {
 ; CHECK-V8M-LABEL: @different_ivs(
 ; CHECK-V8M-NEXT:  loop.preheader:
 ; CHECK-V8M-NEXT:    [[N64:%.*]] = zext i32 [[N:%.*]] to i64
+; CHECK-V8M-NEXT:    [[TMP0:%.*]] = zext i32 [[LENGTH:%.*]] to i64
 ; CHECK-V8M-NEXT:    [[UMAX:%.*]] = call i64 @llvm.umax.i64(i64 [[N64]], i64 1)
-; CHECK-V8M-NEXT:    [[TMP0:%.*]] = add nsw i64 [[UMAX]], -1
-; CHECK-V8M-NEXT:    [[TMP1:%.*]] = zext i32 [[LENGTH:%.*]] to i64
-; CHECK-V8M-NEXT:    [[UMIN:%.*]] = call i64 @llvm.umin.i64(i64 [[TMP0]], i64 [[TMP1]])
-; CHECK-V8M-NEXT:    [[TMP2:%.*]] = zext i32 [[LENGTH]] to i64
-; CHECK-V8M-NEXT:    [[TMP3:%.*]] = icmp ne i64 [[TMP2]], [[UMIN]]
+; CHECK-V8M-NEXT:    [[TMP1:%.*]] = add nsw i64 [[UMAX]], -1
+; CHECK-V8M-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP0]], 0
+; CHECK-V8M-NEXT:    [[UMIN:%.*]] = call i64 @llvm.umin.i64(i64 [[TMP1]], i64 [[TMP0]])
+; CHECK-V8M-NEXT:    [[TMP3:%.*]] = select i1 [[TMP2]], i64 0, i64 [[UMIN]]
+; CHECK-V8M-NEXT:    [[TMP4:%.*]] = zext i32 [[LENGTH]] to i64
+; CHECK-V8M-NEXT:    [[TMP5:%.*]] = icmp ne i64 [[TMP4]], [[TMP3]]
 ; CHECK-V8M-NEXT:    br label [[LOOP:%.*]]
 ; CHECK-V8M:       loop:
 ; CHECK-V8M-NEXT:    [[LOOP_ACC:%.*]] = phi i32 [ [[LOOP_ACC_NEXT:%.*]], [[GUARDED:%.*]] ], [ 0, [[LOOP_PREHEADER:%.*]] ]
 ; CHECK-V8M-NEXT:    [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[GUARDED]] ], [ 0, [[LOOP_PREHEADER]] ]
-; CHECK-V8M-NEXT:    br i1 [[TMP3]], label [[GUARDED]], label [[DEOPT:%.*]], !prof !0
+; CHECK-V8M-NEXT:    br i1 [[TMP5]], label [[GUARDED]], label [[DEOPT:%.*]], !prof [[PROF0]]
 ; CHECK-V8M:       deopt:
 ; CHECK-V8M-NEXT:    call void @prevent_merging()
 ; CHECK-V8M-NEXT:    ret i32 -1
 ; CHECK-V8M:       guarded:
 ; CHECK-V8M-NEXT:    [[ARRAY_I_PTR:%.*]] = getelementptr inbounds i32, i32* [[ARRAY:%.*]], i64 [[I]]
 ; CHECK-V8M-NEXT:    [[ARRAY_I:%.*]] = load i32, i32* [[ARRAY_I_PTR]], align 4
 ; CHECK-V8M-NEXT:    [[LOOP_ACC_NEXT]] = add i32 [[LOOP_ACC]], [[ARRAY_I]]
 ; CHECK-V8M-NEXT:    [[I_NEXT]] = add nuw nsw i64 [[I]], 1
 ; CHECK-V8M-NEXT:    [[CONTINUE:%.*]] = icmp ult i64 [[I_NEXT]], [[N64]]
 ; CHECK-V8M-NEXT:    br i1 [[CONTINUE]], label [[LOOP]], label [[EXIT:%.*]]
 ; CHECK-V8M:       exit:
 ; CHECK-V8M-NEXT:    [[RESULT:%.*]] = phi i32 [ [[LOOP_ACC_NEXT]], [[GUARDED]] ]
 ; CHECK-V8M-NEXT:    ret i32 [[RESULT]]
 ;
 ; CHECK-V8A-LABEL: @different_ivs(
 ; CHECK-V8A-NEXT:  loop.preheader:
 ; CHECK-V8A-NEXT:    [[N64:%.*]] = zext i32 [[N:%.*]] to i64
+; CHECK-V8A-NEXT:    [[TMP0:%.*]] = zext i32 [[LENGTH:%.*]] to i64
 ; CHECK-V8A-NEXT:    [[UMAX:%.*]] = call i64 @llvm.umax.i64(i64 [[N64]], i64 1)
-; CHECK-V8A-NEXT:    [[TMP0:%.*]] = add nsw i64 [[UMAX]], -1
-; CHECK-V8A-NEXT:    [[TMP1:%.*]] = zext i32 [[LENGTH:%.*]] to i64
-; CHECK-V8A-NEXT:    [[UMIN:%.*]] = call i64 @llvm.umin.i64(i64 [[TMP0]], i64 [[TMP1]])
-; CHECK-V8A-NEXT:    [[TMP2:%.*]] = zext i32 [[LENGTH]] to i64
-; CHECK-V8A-NEXT:    [[TMP3:%.*]] = icmp ne i64 [[TMP2]], [[UMIN]]
+; CHECK-V8A-NEXT:    [[TMP1:%.*]] = add nsw i64 [[UMAX]], -1
+; CHECK-V8A-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP0]], 0
+; CHECK-V8A-NEXT:    [[UMIN:%.*]] = call i64 @llvm.umin.i64(i64 [[TMP1]], i64 [[TMP0]])
+; CHECK-V8A-NEXT:    [[TMP3:%.*]] = select i1 [[TMP2]], i64 0, i64 [[UMIN]]
+; CHECK-V8A-NEXT:    [[TMP4:%.*]] = zext i32 [[LENGTH]] to i64
+; CHECK-V8A-NEXT:    [[TMP5:%.*]] = icmp ne i64 [[TMP4]], [[TMP3]]
 ; CHECK-V8A-NEXT:    br label [[LOOP:%.*]]
 ; CHECK-V8A:       loop:
 ; CHECK-V8A-NEXT:    [[LOOP_ACC:%.*]] = phi i32 [ [[LOOP_ACC_NEXT:%.*]], [[GUARDED:%.*]] ], [ 0, [[LOOP_PREHEADER:%.*]] ]
 ; CHECK-V8A-NEXT:    [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[GUARDED]] ], [ 0, [[LOOP_PREHEADER]] ]
-; CHECK-V8A-NEXT:    br i1 [[TMP3]], label [[GUARDED]], label [[DEOPT:%.*]], !prof !0
+; CHECK-V8A-NEXT:    br i1 [[TMP5]], label [[GUARDED]], label [[DEOPT:%.*]], !prof [[PROF0]]
 ; CHECK-V8A:       deopt:
 ; CHECK-V8A-NEXT:    call void @prevent_merging()
 ; CHECK-V8A-NEXT:    ret i32 -1
 ; CHECK-V8A:       guarded:
 ; CHECK-V8A-NEXT:    [[ARRAY_I_PTR:%.*]] = getelementptr inbounds i32, i32* [[ARRAY:%.*]], i64 [[I]]
 ; CHECK-V8A-NEXT:    [[ARRAY_I:%.*]] = load i32, i32* [[ARRAY_I_PTR]], align 4
 ; CHECK-V8A-NEXT:    [[LOOP_ACC_NEXT]] = add i32 [[LOOP_ACC]], [[ARRAY_I]]
 ; CHECK-V8A-NEXT:    [[I_NEXT]] = add nuw nsw i64 [[I]], 1

llvm/test/Transforms/IndVarSimplify/loop-predication.ll

 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
 ; RUN: opt < %s -indvars -indvars-predicate-loops=1 -S | FileCheck %s
 
 declare void @prevent_merging()
 
 ; Base case
 define i32 @test1(i32* %array, i32 %length, i32 %n) {
 ; CHECK-LABEL: @test1(
 ; CHECK-NEXT:  loop.preheader:
 ; CHECK-NEXT:    [[UMAX:%.*]] = call i32 @llvm.umax.i32(i32 [[N:%.*]], i32 1)
 ; CHECK-NEXT:    [[TMP0:%.*]] = add i32 [[UMAX]], -1
-; CHECK-NEXT:    [[UMIN:%.*]] = call i32 @llvm.umin.i32(i32 [[LENGTH:%.*]], i32 [[TMP0]])
-; CHECK-NEXT:    [[TMP1:%.*]] = icmp ne i32 [[LENGTH]], [[UMIN]]
+; CHECK-NEXT:    [[TMP1:%.*]] = icmp eq i32 [[LENGTH:%.*]], 0
+; CHECK-NEXT:    [[UMIN:%.*]] = call i32 @llvm.umin.i32(i32 [[TMP0]], i32 [[LENGTH]])
+; CHECK-NEXT:    [[TMP2:%.*]] = select i1 [[TMP1]], i32 0, i32 [[UMIN]]
+; CHECK-NEXT:    [[TMP3:%.*]] = icmp ne i32 [[LENGTH]], [[TMP2]]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[LOOP_ACC:%.*]] = phi i32 [ [[LOOP_ACC_NEXT:%.*]], [[GUARDED:%.*]] ], [ 0, [[LOOP_PREHEADER:%.*]] ]
 ; CHECK-NEXT:    [[I:%.*]] = phi i32 [ [[I_NEXT:%.*]], [[GUARDED]] ], [ 0, [[LOOP_PREHEADER]] ]
-; CHECK-NEXT:    br i1 [[TMP1]], label [[GUARDED]], label [[DEOPT:%.*]], !prof !0
+; CHECK-NEXT:    br i1 [[TMP3]], label [[GUARDED]], label [[DEOPT:%.*]], !prof [[PROF0:![0-9]+]]
 ; CHECK:       deopt:
 ; CHECK-NEXT:    call void @prevent_merging()
 ; CHECK-NEXT:    ret i32 -1
 ; CHECK:       guarded:
 ; CHECK-NEXT:    [[I_I64:%.*]] = zext i32 [[I]] to i64
 ; CHECK-NEXT:    [[ARRAY_I_PTR:%.*]] = getelementptr inbounds i32, i32* [[ARRAY:%.*]], i64 [[I_I64]]
 ; CHECK-NEXT:    [[ARRAY_I:%.*]] = load i32, i32* [[ARRAY_I_PTR]], align 4
 ; CHECK-NEXT:    [[LOOP_ACC_NEXT]] = add i32 [[LOOP_ACC]], [[ARRAY_I]]

 define i32 @neg_store(i32* %array, i32 %length, i32 %n) {
 ; CHECK-LABEL: @neg_store(
 ; CHECK-NEXT:  loop.preheader:
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[LOOP_ACC:%.*]] = phi i32 [ [[LOOP_ACC_NEXT:%.*]], [[GUARDED:%.*]] ], [ 0, [[LOOP_PREHEADER:%.*]] ]
 ; CHECK-NEXT:    [[I:%.*]] = phi i32 [ [[I_NEXT:%.*]], [[GUARDED]] ], [ 0, [[LOOP_PREHEADER]] ]
 ; CHECK-NEXT:    [[WITHIN_BOUNDS:%.*]] = icmp ult i32 [[I]], [[LENGTH:%.*]]
-; CHECK-NEXT:    br i1 [[WITHIN_BOUNDS]], label [[GUARDED]], label [[DEOPT:%.*]], !prof !0
+; CHECK-NEXT:    br i1 [[WITHIN_BOUNDS]], label [[GUARDED]], label [[DEOPT:%.*]], !prof [[PROF0]]
 ; CHECK:       deopt:
 ; CHECK-NEXT:    call void @prevent_merging()
 ; CHECK-NEXT:    ret i32 -1
 ; CHECK:       guarded:
 ; CHECK-NEXT:    [[I_I64:%.*]] = zext i32 [[I]] to i64
 ; CHECK-NEXT:    [[ARRAY_I_PTR:%.*]] = getelementptr inbounds i32, i32* [[ARRAY:%.*]], i64 [[I_I64]]
 ; CHECK-NEXT:    [[ARRAY_I:%.*]] = load i32, i32* [[ARRAY_I_PTR]], align 4
 ; CHECK-NEXT:    [[LOOP_ACC_NEXT]] = add i32 [[LOOP_ACC]], [[ARRAY_I]]

 ; CHECK-LABEL: @neg_implicit_exit(
 ; CHECK-NEXT:  loop.preheader:
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[LOOP_ACC:%.*]] = phi i32 [ [[LOOP_ACC_NEXT:%.*]], [[GUARDED:%.*]] ], [ 0, [[LOOP_PREHEADER:%.*]] ]
 ; CHECK-NEXT:    [[I:%.*]] = phi i32 [ [[I_NEXT:%.*]], [[GUARDED]] ], [ 0, [[LOOP_PREHEADER]] ]
 ; CHECK-NEXT:    call void @maythrow()
 ; CHECK-NEXT:    [[WITHIN_BOUNDS:%.*]] = icmp ult i32 [[I]], [[LENGTH:%.*]]
-; CHECK-NEXT:    br i1 [[WITHIN_BOUNDS]], label [[GUARDED]], label [[DEOPT:%.*]], !prof !0
+; CHECK-NEXT:    br i1 [[WITHIN_BOUNDS]], label [[GUARDED]], label [[DEOPT:%.*]], !prof [[PROF0]]
 ; CHECK:       deopt:
 ; CHECK-NEXT:    call void @prevent_merging()
 ; CHECK-NEXT:    ret i32 -1
 ; CHECK:       guarded:
 ; CHECK-NEXT:    [[I_I64:%.*]] = zext i32 [[I]] to i64
 ; CHECK-NEXT:    [[ARRAY_I_PTR:%.*]] = getelementptr inbounds i32, i32* [[ARRAY:%.*]], i64 [[I_I64]]
 ; CHECK-NEXT:    [[ARRAY_I:%.*]] = load i32, i32* [[ARRAY_I_PTR]], align 4
 ; CHECK-NEXT:    [[LOOP_ACC_NEXT]] = add i32 [[LOOP_ACC]], [[ARRAY_I]]

 
 
 
 ; Base case, but in LFTR form (just for basic correctness checking)
 define i32 @test2(i32* %array, i32 %length, i32 %n) {
 ; CHECK-LABEL: @test2(
 ; CHECK-NEXT:  loop.preheader:
 ; CHECK-NEXT:    [[TMP0:%.*]] = add i32 [[N:%.*]], -1
-; CHECK-NEXT:    [[UMIN:%.*]] = call i32 @llvm.umin.i32(i32 [[LENGTH:%.*]], i32 [[TMP0]])
-; CHECK-NEXT:    [[TMP1:%.*]] = icmp ne i32 [[LENGTH]], [[UMIN]]
+; CHECK-NEXT:    [[TMP1:%.*]] = icmp eq i32 [[LENGTH:%.*]], 0
+; CHECK-NEXT:    [[UMIN:%.*]] = call i32 @llvm.umin.i32(i32 [[TMP0]], i32 [[LENGTH]])
+; CHECK-NEXT:    [[TMP2:%.*]] = select i1 [[TMP1]], i32 0, i32 [[UMIN]]
+; CHECK-NEXT:    [[TMP3:%.*]] = icmp ne i32 [[LENGTH]], [[TMP2]]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[LOOP_ACC:%.*]] = phi i32 [ [[LOOP_ACC_NEXT:%.*]], [[GUARDED:%.*]] ], [ 0, [[LOOP_PREHEADER:%.*]] ]
 ; CHECK-NEXT:    [[I:%.*]] = phi i32 [ [[I_NEXT:%.*]], [[GUARDED]] ], [ 0, [[LOOP_PREHEADER]] ]
-; CHECK-NEXT:    br i1 [[TMP1]], label [[GUARDED]], label [[DEOPT:%.*]], !prof !0
+; CHECK-NEXT:    br i1 [[TMP3]], label [[GUARDED]], label [[DEOPT:%.*]], !prof [[PROF0]]
 ; CHECK:       deopt:
 ; CHECK-NEXT:    call void @prevent_merging()
 ; CHECK-NEXT:    ret i32 -1
 ; CHECK:       guarded:
 ; CHECK-NEXT:    [[I_I64:%.*]] = zext i32 [[I]] to i64
 ; CHECK-NEXT:    [[ARRAY_I_PTR:%.*]] = getelementptr inbounds i32, i32* [[ARRAY:%.*]], i64 [[I_I64]]
 ; CHECK-NEXT:    [[ARRAY_I:%.*]] = load i32, i32* [[ARRAY_I_PTR]], align 4
 ; CHECK-NEXT:    [[LOOP_ACC_NEXT]] = add i32 [[LOOP_ACC]], [[ARRAY_I]]
exit:                                             ; preds = %guarded, %entry
   ret i32 %result
 }
 
 ; br (and rcheck1, rcheck2)
 define i32 @two_range_checks(i32* %array.1, i32 %length.1, i32* %array.2, i32 %length.2, i32 %n) {
 ; CHECK-LABEL: @two_range_checks(
 ; CHECK-NEXT:  loop.preheader:
 ; CHECK-NEXT:    [[UMIN:%.*]] = call i32 @llvm.umin.i32(i32 [[LENGTH_2:%.*]], i32 [[LENGTH_1:%.*]])
-; CHECK-NEXT:    [[UMIN1:%.*]] = call i32 @llvm.umin.i32(i32 [[LENGTH_2]], i32 [[LENGTH_1]])
 ; CHECK-NEXT:    [[UMAX:%.*]] = call i32 @llvm.umax.i32(i32 [[N:%.*]], i32 1)
 ; CHECK-NEXT:    [[TMP0:%.*]] = add i32 [[UMAX]], -1
-; CHECK-NEXT:    [[UMIN2:%.*]] = call i32 @llvm.umin.i32(i32 [[UMIN1]], i32 [[TMP0]])
-; CHECK-NEXT:    [[TMP1:%.*]] = icmp ne i32 [[UMIN]], [[UMIN2]]
+; CHECK-NEXT:    [[TMP1:%.*]] = icmp eq i32 [[UMIN]], 0
+; CHECK-NEXT:    [[UMIN1:%.*]] = call i32 @llvm.umin.i32(i32 [[TMP0]], i32 [[UMIN]])
+; CHECK-NEXT:    [[TMP2:%.*]] = select i1 [[TMP1]], i32 0, i32 [[UMIN1]]
+; CHECK-NEXT:    [[TMP3:%.*]] = icmp ne i32 [[UMIN]], [[TMP2]]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[LOOP_ACC:%.*]] = phi i32 [ [[LOOP_ACC_NEXT:%.*]], [[GUARDED:%.*]] ], [ 0, [[LOOP_PREHEADER:%.*]] ]
 ; CHECK-NEXT:    [[I:%.*]] = phi i32 [ [[I_NEXT:%.*]], [[GUARDED]] ], [ 0, [[LOOP_PREHEADER]] ]
-; CHECK-NEXT:    br i1 [[TMP1]], label [[GUARDED]], label [[DEOPT:%.*]], !prof !0
+; CHECK-NEXT:    br i1 [[TMP3]], label [[GUARDED]], label [[DEOPT:%.*]], !prof [[PROF0]]
 ; CHECK:       deopt:
 ; CHECK-NEXT:    call void @prevent_merging()
 ; CHECK-NEXT:    ret i32 -1
 ; CHECK:       guarded:
 ; CHECK-NEXT:    [[I_I64:%.*]] = zext i32 [[I]] to i64
 ; CHECK-NEXT:    [[ARRAY_1_I_PTR:%.*]] = getelementptr inbounds i32, i32* [[ARRAY_1:%.*]], i64 [[I_I64]]
 ; CHECK-NEXT:    [[ARRAY_1_I:%.*]] = load i32, i32* [[ARRAY_1_I_PTR]], align 4
 ; CHECK-NEXT:    [[LOOP_ACC_1:%.*]] = add i32 [[LOOP_ACC]], [[ARRAY_1_I]]
exit:                                             ; preds = %guarded, %entry
   ret i32 %result
 }
 
 define i32 @three_range_checks(i32* %array.1, i32 %length.1, i32* %array.2, i32 %length.2, i32* %array.3, i32 %length.3, i32 %n) {
 ; CHECK-LABEL: @three_range_checks(
 ; CHECK-NEXT:  loop.preheader:
 ; CHECK-NEXT:    [[UMIN:%.*]] = call i32 @llvm.umin.i32(i32 [[LENGTH_3:%.*]], i32 [[LENGTH_2:%.*]])
 ; CHECK-NEXT:    [[UMIN1:%.*]] = call i32 @llvm.umin.i32(i32 [[UMIN]], i32 [[LENGTH_1:%.*]])
-; CHECK-NEXT:    [[UMIN2:%.*]] = call i32 @llvm.umin.i32(i32 [[LENGTH_3]], i32 [[LENGTH_2]])
-; CHECK-NEXT:    [[UMIN3:%.*]] = call i32 @llvm.umin.i32(i32 [[UMIN2]], i32 [[LENGTH_1]])
 ; CHECK-NEXT:    [[UMAX:%.*]] = call i32 @llvm.umax.i32(i32 [[N:%.*]], i32 1)
 ; CHECK-NEXT:    [[TMP0:%.*]] = add i32 [[UMAX]], -1
-; CHECK-NEXT:    [[UMIN4:%.*]] = call i32 @llvm.umin.i32(i32 [[UMIN3]], i32 [[TMP0]])
-; CHECK-NEXT:    [[TMP1:%.*]] = icmp ne i32 [[UMIN1]], [[UMIN4]]
+; CHECK-NEXT:    [[TMP1:%.*]] = icmp eq i32 [[UMIN1]], 0
+; CHECK-NEXT:    [[UMIN2:%.*]] = call i32 @llvm.umin.i32(i32 [[TMP0]], i32 [[UMIN1]])
+; CHECK-NEXT:    [[TMP2:%.*]] = select i1 [[TMP1]], i32 0, i32 [[UMIN2]]
+; CHECK-NEXT:    [[TMP3:%.*]] = icmp ne i32 [[UMIN1]], [[TMP2]]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[LOOP_ACC:%.*]] = phi i32 [ [[LOOP_ACC_NEXT:%.*]], [[GUARDED:%.*]] ], [ 0, [[LOOP_PREHEADER:%.*]] ]
 ; CHECK-NEXT:    [[I:%.*]] = phi i32 [ [[I_NEXT:%.*]], [[GUARDED]] ], [ 0, [[LOOP_PREHEADER]] ]
-; CHECK-NEXT:    br i1 [[TMP1]], label [[GUARDED]], label [[DEOPT:%.*]], !prof !0
+; CHECK-NEXT:    br i1 [[TMP3]], label [[GUARDED]], label [[DEOPT:%.*]], !prof [[PROF0]]
 ; CHECK:       deopt:
 ; CHECK-NEXT:    call void @prevent_merging()
 ; CHECK-NEXT:    ret i32 -1
 ; CHECK:       guarded:
 ; CHECK-NEXT:    [[I_I64:%.*]] = zext i32 [[I]] to i64
 ; CHECK-NEXT:    [[ARRAY_1_I_PTR:%.*]] = getelementptr inbounds i32, i32* [[ARRAY_1:%.*]], i64 [[I_I64]]
 ; CHECK-NEXT:    [[ARRAY_1_I:%.*]] = load i32, i32* [[ARRAY_1_I_PTR]], align 4
 ; CHECK-NEXT:    [[LOOP_ACC_1:%.*]] = add i32 [[LOOP_ACC]], [[ARRAY_1_I]]
exit:                                             ; preds = %guarded, %entry
   %result = phi i32 [ %loop.acc.next, %guarded ]
   ret i32 %result
 }
 
 ; Analogous to the above, but with two distinct branches (on different conditions)
 define i32 @distinct_checks(i32* %array.1, i32 %length.1, i32* %array.2, i32 %length.2, i32* %array.3, i32 %length.3, i32 %n) {
 ; CHECK-LABEL: @distinct_checks(
 ; CHECK-NEXT:  loop.preheader:
-; CHECK-NEXT:    [[UMIN:%.*]] = call i32 @llvm.umin.i32(i32 [[LENGTH_2:%.*]], i32 [[LENGTH_1:%.*]])
 ; CHECK-NEXT:    [[UMAX:%.*]] = call i32 @llvm.umax.i32(i32 [[N:%.*]], i32 1)
 ; CHECK-NEXT:    [[TMP0:%.*]] = add i32 [[UMAX]], -1
-; CHECK-NEXT:    [[UMIN1:%.*]] = call i32 @llvm.umin.i32(i32 [[UMIN]], i32 [[TMP0]])
-; CHECK-NEXT:    [[TMP1:%.*]] = icmp ne i32 [[LENGTH_1]], [[UMIN1]]
-; CHECK-NEXT:    [[TMP2:%.*]] = icmp ne i32 [[LENGTH_2]], [[UMIN1]]
+; CHECK-NEXT:    [[TMP1:%.*]] = icmp eq i32 [[LENGTH_1:%.*]], 0
+; CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i32 [[LENGTH_2:%.*]], 0
+; CHECK-NEXT:    [[TMP3:%.*]] = select i1 [[TMP1]], i1 true, i1 [[TMP2]]
+; CHECK-NEXT:    [[UMIN:%.*]] = call i32 @llvm.umin.i32(i32 [[TMP0]], i32 [[LENGTH_2]])
+; CHECK-NEXT:    [[UMIN1:%.*]] = call i32 @llvm.umin.i32(i32 [[UMIN]], i32 [[LENGTH_1]])
+; CHECK-NEXT:    [[TMP4:%.*]] = select i1 [[TMP3]], i32 0, i32 [[UMIN1]]
+; CHECK-NEXT:    [[TMP5:%.*]] = icmp ne i32 [[LENGTH_1]], [[TMP4]]
+; CHECK-NEXT:    [[TMP6:%.*]] = icmp ne i32 [[LENGTH_2]], [[TMP4]]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[LOOP_ACC:%.*]] = phi i32 [ [[LOOP_ACC_NEXT:%.*]], [[GUARDED1:%.*]] ], [ 0, [[LOOP_PREHEADER:%.*]] ]
 ; CHECK-NEXT:    [[I:%.*]] = phi i32 [ [[I_NEXT:%.*]], [[GUARDED1]] ], [ 0, [[LOOP_PREHEADER]] ]
-; CHECK-NEXT:    br i1 [[TMP1]], label [[GUARDED:%.*]], label [[DEOPT:%.*]], !prof !0
+; CHECK-NEXT:    br i1 [[TMP5]], label [[GUARDED:%.*]], label [[DEOPT:%.*]], !prof [[PROF0]]
 ; CHECK:       deopt:
 ; CHECK-NEXT:    call void @prevent_merging()
 ; CHECK-NEXT:    ret i32 -1
 ; CHECK:       guarded:
 ; CHECK-NEXT:    [[I_I64:%.*]] = zext i32 [[I]] to i64
 ; CHECK-NEXT:    [[ARRAY_1_I_PTR:%.*]] = getelementptr inbounds i32, i32* [[ARRAY_1:%.*]], i64 [[I_I64]]
 ; CHECK-NEXT:    [[ARRAY_1_I:%.*]] = load i32, i32* [[ARRAY_1_I_PTR]], align 4
 ; CHECK-NEXT:    [[LOOP_ACC_1:%.*]] = add i32 [[LOOP_ACC]], [[ARRAY_1_I]]
-; CHECK-NEXT:    br i1 [[TMP2]], label [[GUARDED1]], label [[DEOPT2:%.*]], !prof !0
+; CHECK-NEXT:    br i1 [[TMP6]], label [[GUARDED1]], label [[DEOPT2:%.*]], !prof [[PROF0]]
 ; CHECK:       deopt2:
 ; CHECK-NEXT:    call void @prevent_merging()
 ; CHECK-NEXT:    ret i32 -1
 ; CHECK:       guarded1:
 ; CHECK-NEXT:    [[ARRAY_3_I_PTR:%.*]] = getelementptr inbounds i32, i32* [[ARRAY_3:%.*]], i64 [[I_I64]]
 ; CHECK-NEXT:    [[ARRAY_3_I:%.*]] = load i32, i32* [[ARRAY_3_I_PTR]], align 4
 ; CHECK-NEXT:    [[LOOP_ACC_NEXT]] = add i32 [[LOOP_ACC_1]], [[ARRAY_3_I]]
 ; CHECK-NEXT:    [[I_NEXT]] = add nuw i32 [[I]], 1
exit:
   ret i32 %result
 }
 
 define i32 @duplicate_checks(i32* %array.1, i32* %array.2, i32* %array.3, i32 %length, i32 %n) {
 ; CHECK-LABEL: @duplicate_checks(
 ; CHECK-NEXT:  loop.preheader:
 ; CHECK-NEXT:    [[UMAX:%.*]] = call i32 @llvm.umax.i32(i32 [[N:%.*]], i32 1)
 ; CHECK-NEXT:    [[TMP0:%.*]] = add i32 [[UMAX]], -1
-; CHECK-NEXT:    [[UMIN:%.*]] = call i32 @llvm.umin.i32(i32 [[LENGTH:%.*]], i32 [[TMP0]])
-; CHECK-NEXT:    [[TMP1:%.*]] = icmp ne i32 [[LENGTH]], [[UMIN]]
+; CHECK-NEXT:    [[TMP1:%.*]] = icmp eq i32 [[LENGTH:%.*]], 0
+; CHECK-NEXT:    [[UMIN:%.*]] = call i32 @llvm.umin.i32(i32 [[TMP0]], i32 [[LENGTH]])
+; CHECK-NEXT:    [[TMP2:%.*]] = select i1 [[TMP1]], i32 0, i32 [[UMIN]]
+; CHECK-NEXT:    [[TMP3:%.*]] = icmp ne i32 [[LENGTH]], [[TMP2]]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[LOOP_ACC:%.*]] = phi i32 [ [[LOOP_ACC_NEXT:%.*]], [[GUARDED1:%.*]] ], [ 0, [[LOOP_PREHEADER:%.*]] ]
 ; CHECK-NEXT:    [[I:%.*]] = phi i32 [ [[I_NEXT:%.*]], [[GUARDED1]] ], [ 0, [[LOOP_PREHEADER]] ]
-; CHECK-NEXT:    br i1 [[TMP1]], label [[GUARDED:%.*]], label [[DEOPT:%.*]], !prof !0
+; CHECK-NEXT:    br i1 [[TMP3]], label [[GUARDED:%.*]], label [[DEOPT:%.*]], !prof [[PROF0]]
 ; CHECK:       deopt:
 ; CHECK-NEXT:    call void @prevent_merging()
 ; CHECK-NEXT:    ret i32 -1
 ; CHECK:       guarded:
 ; CHECK-NEXT:    [[I_I64:%.*]] = zext i32 [[I]] to i64
 ; CHECK-NEXT:    [[ARRAY_1_I_PTR:%.*]] = getelementptr inbounds i32, i32* [[ARRAY_1:%.*]], i64 [[I_I64]]
 ; CHECK-NEXT:    [[ARRAY_1_I:%.*]] = load i32, i32* [[ARRAY_1_I_PTR]], align 4
 ; CHECK-NEXT:    [[LOOP_ACC_1:%.*]] = add i32 [[LOOP_ACC]], [[ARRAY_1_I]]
-; CHECK-NEXT:    br i1 true, label [[GUARDED1]], label [[DEOPT2:%.*]], !prof !0
+; CHECK-NEXT:    br i1 true, label [[GUARDED1]], label [[DEOPT2:%.*]], !prof [[PROF0]]
 ; CHECK:       deopt2:
 ; CHECK-NEXT:    call void @prevent_merging()
 ; CHECK-NEXT:    ret i32 -1
 ; CHECK:       guarded1:
 ; CHECK-NEXT:    [[ARRAY_3_I_PTR:%.*]] = getelementptr inbounds i32, i32* [[ARRAY_3:%.*]], i64 [[I_I64]]
 ; CHECK-NEXT:    [[ARRAY_3_I:%.*]] = load i32, i32* [[ARRAY_3_I_PTR]], align 4
 ; CHECK-NEXT:    [[LOOP_ACC_NEXT]] = add i32 [[LOOP_ACC_1]], [[ARRAY_3_I]]
 ; CHECK-NEXT:    [[I_NEXT]] = add nuw i32 [[I]], 1

 
 define i32 @provably_taken(i32* %array, i32* %length.ptr) {
 ; CHECK-LABEL: @provably_taken(
 ; CHECK-NEXT:  loop.preheader:
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[LOOP_ACC:%.*]] = phi i32 [ [[LOOP_ACC_NEXT:%.*]], [[GUARDED:%.*]] ], [ 0, [[LOOP_PREHEADER:%.*]] ]
 ; CHECK-NEXT:    [[I:%.*]] = phi i32 [ [[I_NEXT:%.*]], [[GUARDED]] ], [ 0, [[LOOP_PREHEADER]] ]
-; CHECK-NEXT:    br i1 false, label [[GUARDED]], label [[DEOPT:%.*]], !prof !0
+; CHECK-NEXT:    br i1 false, label [[GUARDED]], label [[DEOPT:%.*]], !prof [[PROF0]]
 ; CHECK:       deopt:
 ; CHECK-NEXT:    call void @prevent_merging()
 ; CHECK-NEXT:    ret i32 -1
 ; CHECK:       guarded:
 ; CHECK-NEXT:    [[I_I64:%.*]] = zext i32 [[I]] to i64
 ; CHECK-NEXT:    [[ARRAY_I_PTR:%.*]] = getelementptr inbounds i32, i32* [[ARRAY:%.*]], i64 [[I_I64]]
 ; CHECK-NEXT:    [[ARRAY_I:%.*]] = load i32, i32* [[ARRAY_I_PTR]], align 4
 ; CHECK-NEXT:    [[LOOP_ACC_NEXT]] = add i32 [[LOOP_ACC]], [[ARRAY_I]]

 }
 
 ; Non-latch exits can still be predicated
 define i32 @unconditional_latch(i32* %a, i32 %length) {
 ; CHECK-LABEL: @unconditional_latch(
 ; CHECK-NEXT:  loop.preheader:
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
-; CHECK-NEXT:    br i1 false, label [[GUARDED:%.*]], label [[DEOPT:%.*]], !prof !0
+; CHECK-NEXT:    br i1 false, label [[GUARDED:%.*]], label [[DEOPT:%.*]], !prof [[PROF0]]
 ; CHECK:       deopt:
 ; CHECK-NEXT:    call void @prevent_merging()
 ; CHECK-NEXT:    ret i32 -1
 ; CHECK:       guarded:
 ; CHECK-NEXT:    br label [[LOOP]]
 ;
 loop.preheader:
   br label %loop

 ; Side effect in loop must run proper number of times
 define i32 @unconditional_latch_with_side_effect(i32* %a, i32 %length) {
 ; CHECK-LABEL: @unconditional_latch_with_side_effect(
 ; CHECK-NEXT:  loop.preheader:
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[I:%.*]] = phi i32 [ [[I_NEXT:%.*]], [[GUARDED:%.*]] ], [ 400, [[LOOP_PREHEADER:%.*]] ]
 ; CHECK-NEXT:    [[WITHIN_BOUNDS:%.*]] = icmp ult i32 [[I]], [[LENGTH:%.*]]
-; CHECK-NEXT:    br i1 [[WITHIN_BOUNDS]], label [[GUARDED]], label [[DEOPT:%.*]], !prof !0
+; CHECK-NEXT:    br i1 [[WITHIN_BOUNDS]], label [[GUARDED]], label [[DEOPT:%.*]], !prof [[PROF0]]
 ; CHECK:       deopt:
 ; CHECK-NEXT:    call void @prevent_merging()
 ; CHECK-NEXT:    ret i32 -1
 ; CHECK:       guarded:
 ; CHECK-NEXT:    store volatile i32 0, i32* [[A:%.*]], align 4
 ; CHECK-NEXT:    [[I_NEXT]] = add i32 [[I]], 1
 ; CHECK-NEXT:    br label [[LOOP]]
 ;

 
 ; Demonstrate that this approach works with IVs of different steps, and types
 ; This version uses a manually lftred exit condition to work around an issue described
 ; in detail on next test.
 define i32 @different_ivs(i32* %array, i32 %length, i32 %n) {
 ; CHECK-LABEL: @different_ivs(
 ; CHECK-NEXT:  loop.preheader:
 ; CHECK-NEXT:    [[N64:%.*]] = zext i32 [[N:%.*]] to i64
+; CHECK-NEXT:    [[TMP0:%.*]] = zext i32 [[LENGTH:%.*]] to i64
 ; CHECK-NEXT:    [[UMAX:%.*]] = call i64 @llvm.umax.i64(i64 [[N64]], i64 1)
-; CHECK-NEXT:    [[TMP0:%.*]] = add nsw i64 [[UMAX]], -1
-; CHECK-NEXT:    [[TMP1:%.*]] = zext i32 [[LENGTH:%.*]] to i64
-; CHECK-NEXT:    [[UMIN:%.*]] = call i64 @llvm.umin.i64(i64 [[TMP0]], i64 [[TMP1]])
-; CHECK-NEXT:    [[TMP2:%.*]] = zext i32 [[LENGTH]] to i64
-; CHECK-NEXT:    [[TMP3:%.*]] = icmp ne i64 [[TMP2]], [[UMIN]]
+; CHECK-NEXT:    [[TMP1:%.*]] = add nsw i64 [[UMAX]], -1
+; CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[TMP0]], 0
+; CHECK-NEXT:    [[UMIN:%.*]] = call i64 @llvm.umin.i64(i64 [[TMP1]], i64 [[TMP0]])
+; CHECK-NEXT:    [[TMP3:%.*]] = select i1 [[TMP2]], i64 0, i64 [[UMIN]]
+; CHECK-NEXT:    [[TMP4:%.*]] = zext i32 [[LENGTH]] to i64
+; CHECK-NEXT:    [[TMP5:%.*]] = icmp ne i64 [[TMP4]], [[TMP3]]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[LOOP_ACC:%.*]] = phi i32 [ [[LOOP_ACC_NEXT:%.*]], [[GUARDED:%.*]] ], [ 0, [[LOOP_PREHEADER:%.*]] ]
 ; CHECK-NEXT:    [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[GUARDED]] ], [ 0, [[LOOP_PREHEADER]] ]
-; CHECK-NEXT:    br i1 [[TMP3]], label [[GUARDED]], label [[DEOPT:%.*]], !prof !0
+; CHECK-NEXT:    br i1 [[TMP5]], label [[GUARDED]], label [[DEOPT:%.*]], !prof [[PROF0]]
 ; CHECK:       deopt:
 ; CHECK-NEXT:    call void @prevent_merging()
 ; CHECK-NEXT:    ret i32 -1
 ; CHECK:       guarded:
 ; CHECK-NEXT:    [[ARRAY_I_PTR:%.*]] = getelementptr inbounds i32, i32* [[ARRAY:%.*]], i64 [[I]]
 ; CHECK-NEXT:    [[ARRAY_I:%.*]] = load i32, i32* [[ARRAY_I_PTR]], align 4
 ; CHECK-NEXT:    [[LOOP_ACC_NEXT]] = add i32 [[LOOP_ACC]], [[ARRAY_I]]
 ; CHECK-NEXT:    [[I_NEXT]] = add nuw nsw i64 [[I]], 1

 ; CHECK-NEXT:    [[J_START:%.*]] = sub nuw nsw i32 [[LENGTH]], 1
 ; CHECK-NEXT:    [[N64:%.*]] = zext i32 [[N:%.*]] to i64
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[LOOP_ACC:%.*]] = phi i32 [ [[LOOP_ACC_NEXT:%.*]], [[GUARDED:%.*]] ], [ 0, [[LOOP_PREHEADER]] ]
 ; CHECK-NEXT:    [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[GUARDED]] ], [ 0, [[LOOP_PREHEADER]] ]
 ; CHECK-NEXT:    [[J:%.*]] = phi i32 [ [[J_NEXT:%.*]], [[GUARDED]] ], [ [[J_START]], [[LOOP_PREHEADER]] ]
 ; CHECK-NEXT:    [[WITHIN_BOUNDS:%.*]] = icmp ult i32 [[J]], [[LENGTH]]
-; CHECK-NEXT:    br i1 [[WITHIN_BOUNDS]], label [[GUARDED]], label [[DEOPT:%.*]], !prof !0
+; CHECK-NEXT:    br i1 [[WITHIN_BOUNDS]], label [[GUARDED]], label [[DEOPT:%.*]], !prof [[PROF0]]
 ; CHECK:       deopt:
 ; CHECK-NEXT:    call void @prevent_merging()
 ; CHECK-NEXT:    ret i32 -1
 ; CHECK:       guarded:
 ; CHECK-NEXT:    [[ARRAY_I_PTR:%.*]] = getelementptr inbounds i32, i32* [[ARRAY:%.*]], i64 [[I]]
 ; CHECK-NEXT:    [[ARRAY_I:%.*]] = load i32, i32* [[ARRAY_I_PTR]], align 4
 ; CHECK-NEXT:    [[LOOP_ACC_NEXT]] = add i32 [[LOOP_ACC]], [[ARRAY_I]]
 ; CHECK-NEXT:    [[I_NEXT]] = add nuw nsw i64 [[I]], 1

 define i32 @neg_dominating_exit(i32* %array, i32 %length, i32 %length2, i32 %n) {
 ; CHECK-LABEL: @neg_dominating_exit(
 ; CHECK-NEXT:  loop.preheader:
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[LOOP_ACC:%.*]] = phi i32 [ [[LOOP_ACC_NEXT:%.*]], [[GUARDED2:%.*]] ], [ 0, [[LOOP_PREHEADER:%.*]] ]
 ; CHECK-NEXT:    [[I:%.*]] = phi i32 [ [[I_NEXT:%.*]], [[GUARDED2]] ], [ 0, [[LOOP_PREHEADER]] ]
 ; CHECK-NEXT:    [[WITHIN_BOUNDS:%.*]] = icmp ult i32 [[I]], [[LENGTH:%.*]]
-; CHECK-NEXT:    br i1 [[WITHIN_BOUNDS]], label [[GUARDED:%.*]], label [[DEOPT:%.*]], !prof !0
+; CHECK-NEXT:    br i1 [[WITHIN_BOUNDS]], label [[GUARDED:%.*]], label [[DEOPT:%.*]], !prof [[PROF0]]
 ; CHECK:       deopt:
 ; CHECK-NEXT:    [[RESULT:%.*]] = phi i32 [ [[LOOP_ACC]], [[LOOP]] ]
 ; CHECK-NEXT:    call void @prevent_merging()
 ; CHECK-NEXT:    ret i32 [[RESULT]]
 ; CHECK:       guarded:
 ; CHECK-NEXT:    [[WITHIN_BOUNDS2:%.*]] = icmp ult i32 [[I]], [[LENGTH2:%.*]]
-; CHECK-NEXT:    br i1 [[WITHIN_BOUNDS2]], label [[GUARDED2]], label [[DEOPT2:%.*]], !prof !0
+; CHECK-NEXT:    br i1 [[WITHIN_BOUNDS2]], label [[GUARDED2]], label [[DEOPT2:%.*]], !prof [[PROF0]]
 ; CHECK:       deopt2:
 ; CHECK-NEXT:    call void @prevent_merging()
 ; CHECK-NEXT:    ret i32 -1
 ; CHECK:       guarded2:
 ; CHECK-NEXT:    [[I_I64:%.*]] = zext i32 [[I]] to i64
 ; CHECK-NEXT:    [[ARRAY_I_PTR:%.*]] = getelementptr inbounds i32, i32* [[ARRAY:%.*]], i64 [[I_I64]]
 ; CHECK-NEXT:    [[ARRAY_I:%.*]] = load i32, i32* [[ARRAY_I_PTR]], align 4
 ; CHECK-NEXT:    [[LOOP_ACC_NEXT]] = add i32 [[LOOP_ACC]], [[ARRAY_I]]

llvm/test/Transforms/PhaseOrdering/AArch64/peel-multiple-unreachable-exits-for-vectorization.ll

 ; CHECK-NEXT:    [[END_INT_I6_PEEL:%.*]] = ptrtoint i64* [[END_I4_PEEL]] to i64
 ; CHECK-NEXT:    [[SUB_I7_PEEL:%.*]] = sub i64 [[END_INT_I6_PEEL]], [[START_INT_I5_PEEL]]
 ; CHECK-NEXT:    [[LV_I_PEEL:%.*]] = load i64, i64* [[START_I]], align 4
 ; CHECK-NEXT:    [[LV_I10_PEEL:%.*]] = load i64, i64* [[START_I2_PEEL]], align 4
 ; CHECK-NEXT:    [[SUM_NEXT_PEEL:%.*]] = add i64 [[LV_I_PEEL]], [[LV_I10_PEEL]]
 ; CHECK-NEXT:    [[C_PEEL:%.*]] = icmp sgt i64 [[N:%.*]], 0
 ; CHECK-NEXT:    br i1 [[C_PEEL]], label [[LOOP_PREHEADER:%.*]], label [[EXIT:%.*]]
 ; CHECK:       loop.preheader:
-; CHECK-NEXT:    [[UMIN:%.*]] = call i64 @llvm.umin.i64(i64 [[SUB_I7_PEEL]], i64 [[SUB_I]])
 ; CHECK-NEXT:    [[TMP0:%.*]] = add i64 [[N]], -1
-; CHECK-NEXT:    [[UMIN16:%.*]] = call i64 @llvm.umin.i64(i64 [[UMIN]], i64 [[TMP0]])
-; CHECK-NEXT:    [[TMP1:%.*]] = add i64 [[UMIN16]], 1
-; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP1]], 5
-; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[LOOP_PREHEADER22:%.*]], label [[VECTOR_PH:%.*]]
+; CHECK-NEXT:    [[TMP1:%.*]] = icmp eq i64* [[END_I]], [[START_I]]
+; CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64* [[END_I4_PEEL]], [[START_I2_PEEL]]
+; CHECK-NEXT:    [[TMP3:%.*]] = select i1 [[TMP1]], i1 true, i1 [[TMP2]]
+; CHECK-NEXT:    [[UMIN:%.*]] = call i64 @llvm.umin.i64(i64 [[TMP0]], i64 [[SUB_I7_PEEL]])
+; CHECK-NEXT:    [[UMIN16:%.*]] = call i64 @llvm.umin.i64(i64 [[UMIN]], i64 [[SUB_I]])
+; CHECK-NEXT:    [[UMIN16_OP:%.*]] = add i64 [[UMIN16]], 1
+; CHECK-NEXT:    [[TMP4:%.*]] = select i1 [[TMP3]], i64 1, i64 [[UMIN16_OP]]
+; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP4]], 5
+; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[LOOP_PREHEADER21:%.*]], label [[VECTOR_PH:%.*]]
 ; CHECK:       vector.ph:
-; CHECK-NEXT:    [[N_MOD_VF:%.*]] = and i64 [[TMP1]], 3
-; 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:    [[N_MOD_VF:%.*]] = and i64 [[TMP4]], 3
+; CHECK-NEXT:    [[TMP5:%.*]] = icmp eq i64 [[N_MOD_VF]], 0
+; CHECK-NEXT:    [[TMP6:%.*]] = select i1 [[TMP5]], i64 4, i64 [[N_MOD_VF]]
+; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 [[TMP4]], [[TMP6]]
 ; CHECK-NEXT:    [[IND_END:%.*]] = add i64 [[N_VEC]], 1
-; CHECK-NEXT:    [[TMP4:%.*]] = insertelement <2 x i64> <i64 poison, i64 0>, i64 [[SUM_NEXT_PEEL]], i64 0
+; CHECK-NEXT:    [[TMP7:%.*]] = insertelement <2 x i64> <i64 poison, i64 0>, i64 [[SUM_NEXT_PEEL]], i64 0
 ; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <2 x i64> [ [[TMP4]], [[VECTOR_PH]] ], [ [[TMP15:%.*]], [[VECTOR_BODY]] ]
-; CHECK-NEXT:    [[VEC_PHI18:%.*]] = phi <2 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP16:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <2 x i64> [ [[TMP7]], [[VECTOR_PH]] ], [ [[TMP18:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[VEC_PHI17:%.*]] = phi <2 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP19:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = or i64 [[INDEX]], 1
-; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr i64, i64* [[START_I]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP6:%.*]] = bitcast i64* [[TMP5]] to <2 x i64>*
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i64>, <2 x i64>* [[TMP6]], align 4
-; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr i64, i64* [[TMP5]], i64 2
-; CHECK-NEXT:    [[TMP8:%.*]] = bitcast i64* [[TMP7]] to <2 x i64>*
-; CHECK-NEXT:    [[WIDE_LOAD19:%.*]] = load <2 x i64>, <2 x i64>* [[TMP8]], align 4
-; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr i64, i64* [[START_I2_PEEL]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP10:%.*]] = bitcast i64* [[TMP9]] to <2 x i64>*
-; CHECK-NEXT:    [[WIDE_LOAD20:%.*]] = load <2 x i64>, <2 x i64>* [[TMP10]], align 4
-; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr i64, i64* [[TMP9]], i64 2
-; CHECK-NEXT:    [[TMP12:%.*]] = bitcast i64* [[TMP11]] to <2 x i64>*
-; CHECK-NEXT:    [[WIDE_LOAD21:%.*]] = load <2 x i64>, <2 x i64>* [[TMP12]], align 4
-; CHECK-NEXT:    [[TMP13:%.*]] = add <2 x i64> [[WIDE_LOAD]], [[VEC_PHI]]
-; CHECK-NEXT:    [[TMP14:%.*]] = add <2 x i64> [[WIDE_LOAD19]], [[VEC_PHI18]]
-; CHECK-NEXT:    [[TMP15]] = add <2 x i64> [[TMP13]], [[WIDE_LOAD20]]
-; CHECK-NEXT:    [[TMP16]] = add <2 x i64> [[TMP14]], [[WIDE_LOAD21]]
+; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr i64, i64* [[START_I]], i64 [[OFFSET_IDX]]
+; CHECK-NEXT:    [[TMP9:%.*]] = bitcast i64* [[TMP8]] to <2 x i64>*
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i64>, <2 x i64>* [[TMP9]], align 4
+; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr i64, i64* [[TMP8]], i64 2
+; CHECK-NEXT:    [[TMP11:%.*]] = bitcast i64* [[TMP10]] to <2 x i64>*
+; CHECK-NEXT:    [[WIDE_LOAD18:%.*]] = load <2 x i64>, <2 x i64>* [[TMP11]], align 4
+; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr i64, i64* [[START_I2_PEEL]], i64 [[OFFSET_IDX]]
+; CHECK-NEXT:    [[TMP13:%.*]] = bitcast i64* [[TMP12]] to <2 x i64>*
+; CHECK-NEXT:    [[WIDE_LOAD19:%.*]] = load <2 x i64>, <2 x i64>* [[TMP13]], align 4
+; CHECK-NEXT:    [[TMP14:%.*]] = getelementptr i64, i64* [[TMP12]], i64 2
+; CHECK-NEXT:    [[TMP15:%.*]] = bitcast i64* [[TMP14]] to <2 x i64>*
+; CHECK-NEXT:    [[WIDE_LOAD20:%.*]] = load <2 x i64>, <2 x i64>* [[TMP15]], align 4
+; CHECK-NEXT:    [[TMP16:%.*]] = add <2 x i64> [[WIDE_LOAD]], [[VEC_PHI]]
+; CHECK-NEXT:    [[TMP17:%.*]] = add <2 x i64> [[WIDE_LOAD18]], [[VEC_PHI17]]
+; CHECK-NEXT:    [[TMP18]] = add <2 x i64> [[TMP16]], [[WIDE_LOAD19]]
+; CHECK-NEXT:    [[TMP19]] = add <2 x i64> [[TMP17]], [[WIDE_LOAD20]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
-; CHECK-NEXT:    [[TMP17:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
-; CHECK-NEXT:    br i1 [[TMP17]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
+; CHECK-NEXT:    [[TMP20:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[TMP20]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    [[BIN_RDX:%.*]] = add <2 x i64> [[TMP16]], [[TMP15]]
-; CHECK-NEXT:    [[TMP18:%.*]] = call i64 @llvm.vector.reduce.add.v2i64(<2 x i64> [[BIN_RDX]])
-; CHECK-NEXT:    br label [[LOOP_PREHEADER22]]
+; CHECK-NEXT:    [[BIN_RDX:%.*]] = add <2 x i64> [[TMP19]], [[TMP18]]
+; CHECK-NEXT:    [[TMP21:%.*]] = call i64 @llvm.vector.reduce.add.v2i64(<2 x i64> [[BIN_RDX]])
+; CHECK-NEXT:    br label [[LOOP_PREHEADER21]]
 ; CHECK:       loop.preheader21:
 ; CHECK-NEXT:    [[IV_PH:%.*]] = phi i64 [ 1, [[LOOP_PREHEADER]] ], [ [[IND_END]], [[MIDDLE_BLOCK]] ]
-; CHECK-NEXT:    [[SUM_PH:%.*]] = phi i64 [ [[SUM_NEXT_PEEL]], [[LOOP_PREHEADER]] ], [ [[TMP18]], [[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    [[SUM_PH:%.*]] = phi i64 [ [[SUM_NEXT_PEEL]], [[LOOP_PREHEADER]] ], [ [[TMP21]], [[MIDDLE_BLOCK]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
-; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[IV_NEXT:%.*]], [[AT_WITH_INT_CONVERSION_EXIT12:%.*]] ], [ [[IV_PH]], [[LOOP_PREHEADER22]] ]
-; CHECK-NEXT:    [[SUM:%.*]] = phi i64 [ [[SUM_NEXT:%.*]], [[AT_WITH_INT_CONVERSION_EXIT12]] ], [ [[SUM_PH]], [[LOOP_PREHEADER22]] ]
+; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[IV_NEXT:%.*]], [[AT_WITH_INT_CONVERSION_EXIT12:%.*]] ], [ [[IV_PH]], [[LOOP_PREHEADER21]] ]
+; CHECK-NEXT:    [[SUM:%.*]] = phi i64 [ [[SUM_NEXT:%.*]], [[AT_WITH_INT_CONVERSION_EXIT12]] ], [ [[SUM_PH]], [[LOOP_PREHEADER21]] ]
 ; CHECK-NEXT:    [[INRANGE_I:%.*]] = icmp ult i64 [[SUB_I]], [[IV]]
 ; CHECK-NEXT:    br i1 [[INRANGE_I]], label [[ERROR_I:%.*]], label [[AT_WITH_INT_CONVERSION_EXIT:%.*]]
 ; CHECK:       error.i:
 ; CHECK-NEXT:    tail call void @error()
 ; CHECK-NEXT:    unreachable
 ; CHECK:       at_with_int_conversion.exit:
 ; CHECK-NEXT:    [[INRANGE_I8:%.*]] = icmp ult i64 [[SUB_I7_PEEL]], [[IV]]
 ; CHECK-NEXT:    br i1 [[INRANGE_I8]], label [[ERROR_I11:%.*]], label [[AT_WITH_INT_CONVERSION_EXIT12]]

 ; CHECK-NEXT:    [[SUB_I19_PEEL:%.*]] = sub i64 [[END_INT_I18_PEEL]], [[START_INT_I17_PEEL]]
 ; CHECK-NEXT:    [[LV_I10_PEEL:%.*]] = load i64, i64* [[START_I2_PEEL]], align 4
 ; CHECK-NEXT:    [[LV_I22_PEEL:%.*]] = load i64, i64* [[START_I14_PEEL]], align 4
 ; CHECK-NEXT:    [[ADD_2_PEEL:%.*]] = add i64 [[LV_I_PEEL]], [[LV_I10_PEEL]]
 ; CHECK-NEXT:    [[SUM_NEXT_PEEL:%.*]] = add i64 [[ADD_2_PEEL]], [[LV_I22_PEEL]]
 ; CHECK-NEXT:    [[COND_PEEL:%.*]] = icmp sgt i64 [[N:%.*]], 0
 ; CHECK-NEXT:    br i1 [[COND_PEEL]], label [[LOOP_PREHEADER:%.*]], label [[EXIT:%.*]]
 ; CHECK:       loop.preheader:
-; CHECK-NEXT:    [[UMIN:%.*]] = call i64 @llvm.umin.i64(i64 [[SUB_I19_PEEL]], i64 [[SUB_I7_PEEL]])
-; CHECK-NEXT:    [[UMIN28:%.*]] = call i64 @llvm.umin.i64(i64 [[UMIN]], i64 [[SUB_I]])
 ; CHECK-NEXT:    [[TMP0:%.*]] = add i64 [[N]], -1
-; CHECK-NEXT:    [[UMIN29:%.*]] = call i64 @llvm.umin.i64(i64 [[UMIN28]], i64 [[TMP0]])
-; CHECK-NEXT:    [[TMP1:%.*]] = add i64 [[UMIN29]], 1
-; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP1]], 5
-; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[LOOP_PREHEADER37:%.*]], label [[VECTOR_PH:%.*]]
+; CHECK-NEXT:    [[TMP1:%.*]] = icmp eq i64* [[END_I]], [[START_I]]
+; CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64* [[END_I4_PEEL]], [[START_I2_PEEL]]
+; CHECK-NEXT:    [[TMP3:%.*]] = icmp eq i64* [[END_I16_PEEL]], [[START_I14_PEEL]]
+; CHECK-NEXT:    [[TMP4:%.*]] = select i1 [[TMP1]], i1 true, i1 [[TMP2]]
+; CHECK-NEXT:    [[TMP5:%.*]] = select i1 [[TMP4]], i1 true, i1 [[TMP3]]
+; CHECK-NEXT:    [[UMIN:%.*]] = call i64 @llvm.umin.i64(i64 [[TMP0]], i64 [[SUB_I19_PEEL]])
+; CHECK-NEXT:    [[UMIN28:%.*]] = call i64 @llvm.umin.i64(i64 [[UMIN]], i64 [[SUB_I7_PEEL]])
+; CHECK-NEXT:    [[UMIN29:%.*]] = call i64 @llvm.umin.i64(i64 [[UMIN28]], i64 [[SUB_I]])
+; CHECK-NEXT:    [[UMIN29_OP:%.*]] = add i64 [[UMIN29]], 1
+; CHECK-NEXT:    [[TMP6:%.*]] = select i1 [[TMP5]], i64 1, i64 [[UMIN29_OP]]
+; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP6]], 5
+; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[LOOP_PREHEADER36:%.*]], label [[VECTOR_PH:%.*]]
 ; CHECK:       vector.ph:
-; CHECK-NEXT:    [[N_MOD_VF:%.*]] = and i64 [[TMP1]], 3
-; 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:    [[N_MOD_VF:%.*]] = and i64 [[TMP6]], 3
+; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[N_MOD_VF]], 0
+; CHECK-NEXT:    [[TMP8:%.*]] = select i1 [[TMP7]], i64 4, i64 [[N_MOD_VF]]
+; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 [[TMP6]], [[TMP8]]
 ; CHECK-NEXT:    [[IND_END:%.*]] = add i64 [[N_VEC]], 1
-; CHECK-NEXT:    [[TMP4:%.*]] = insertelement <2 x i64> <i64 poison, i64 0>, i64 [[SUM_NEXT_PEEL]], i64 0
+; CHECK-NEXT:    [[TMP9:%.*]] = insertelement <2 x i64> <i64 poison, i64 0>, i64 [[SUM_NEXT_PEEL]], i64 0
 ; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <2 x i64> [ [[TMP4]], [[VECTOR_PH]] ], [ [[TMP21:%.*]], [[VECTOR_BODY]] ]
-; CHECK-NEXT:    [[VEC_PHI31:%.*]] = phi <2 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP22:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <2 x i64> [ [[TMP9]], [[VECTOR_PH]] ], [ [[TMP26:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[VEC_PHI30:%.*]] = phi <2 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP27:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = or i64 [[INDEX]], 1
-; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr i64, i64* [[START_I]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP6:%.*]] = bitcast i64* [[TMP5]] to <2 x i64>*
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i64>, <2 x i64>* [[TMP6]], align 4
-; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr i64, i64* [[TMP5]], i64 2
-; CHECK-NEXT:    [[TMP8:%.*]] = bitcast i64* [[TMP7]] to <2 x i64>*
-; CHECK-NEXT:    [[WIDE_LOAD32:%.*]] = load <2 x i64>, <2 x i64>* [[TMP8]], align 4
-; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr i64, i64* [[START_I2_PEEL]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP10:%.*]] = bitcast i64* [[TMP9]] to <2 x i64>*
-; CHECK-NEXT:    [[WIDE_LOAD33:%.*]] = load <2 x i64>, <2 x i64>* [[TMP10]], align 4
-; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr i64, i64* [[TMP9]], i64 2
-; CHECK-NEXT:    [[TMP12:%.*]] = bitcast i64* [[TMP11]] to <2 x i64>*
-; CHECK-NEXT:    [[WIDE_LOAD34:%.*]] = load <2 x i64>, <2 x i64>* [[TMP12]], align 4
-; CHECK-NEXT:    [[TMP13:%.*]] = getelementptr i64, i64* [[START_I14_PEEL]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP14:%.*]] = bitcast i64* [[TMP13]] to <2 x i64>*
-; CHECK-NEXT:    [[WIDE_LOAD35:%.*]] = load <2 x i64>, <2 x i64>* [[TMP14]], align 4
-; CHECK-NEXT:    [[TMP15:%.*]] = getelementptr i64, i64* [[TMP13]], i64 2
-; CHECK-NEXT:    [[TMP16:%.*]] = bitcast i64* [[TMP15]] to <2 x i64>*
-; CHECK-NEXT:    [[WIDE_LOAD36:%.*]] = load <2 x i64>, <2 x i64>* [[TMP16]], align 4
-; CHECK-NEXT:    [[TMP17:%.*]] = add <2 x i64> [[WIDE_LOAD]], [[VEC_PHI]]
-; CHECK-NEXT:    [[TMP18:%.*]] = add <2 x i64> [[WIDE_LOAD32]], [[VEC_PHI31]]
-; CHECK-NEXT:    [[TMP19:%.*]] = add <2 x i64> [[TMP17]], [[WIDE_LOAD33]]
-; CHECK-NEXT:    [[TMP20:%.*]] = add <2 x i64> [[TMP18]], [[WIDE_LOAD34]]
-; CHECK-NEXT:    [[TMP21]] = add <2 x i64> [[TMP19]], [[WIDE_LOAD35]]
-; CHECK-NEXT:    [[TMP22]] = add <2 x i64> [[TMP20]], [[WIDE_LOAD36]]
+; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr i64, i64* [[START_I]], i64 [[OFFSET_IDX]]
+; CHECK-NEXT:    [[TMP11:%.*]] = bitcast i64* [[TMP10]] to <2 x i64>*
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i64>, <2 x i64>* [[TMP11]], align 4
+; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr i64, i64* [[TMP10]], i64 2
+; CHECK-NEXT:    [[TMP13:%.*]] = bitcast i64* [[TMP12]] to <2 x i64>*
+; CHECK-NEXT:    [[WIDE_LOAD31:%.*]] = load <2 x i64>, <2 x i64>* [[TMP13]], align 4
+; CHECK-NEXT:    [[TMP14:%.*]] = getelementptr i64, i64* [[START_I2_PEEL]], i64 [[OFFSET_IDX]]
+; CHECK-NEXT:    [[TMP15:%.*]] = bitcast i64* [[TMP14]] to <2 x i64>*
+; CHECK-NEXT:    [[WIDE_LOAD32:%.*]] = load <2 x i64>, <2 x i64>* [[TMP15]], align 4
+; CHECK-NEXT:    [[TMP16:%.*]] = getelementptr i64, i64* [[TMP14]], i64 2
+; CHECK-NEXT:    [[TMP17:%.*]] = bitcast i64* [[TMP16]] to <2 x i64>*
+; CHECK-NEXT:    [[WIDE_LOAD33:%.*]] = load <2 x i64>, <2 x i64>* [[TMP17]], align 4
+; CHECK-NEXT:    [[TMP18:%.*]] = getelementptr i64, i64* [[START_I14_PEEL]], i64 [[OFFSET_IDX]]
+; CHECK-NEXT:    [[TMP19:%.*]] = bitcast i64* [[TMP18]] to <2 x i64>*
+; CHECK-NEXT:    [[WIDE_LOAD34:%.*]] = load <2 x i64>, <2 x i64>* [[TMP19]], align 4
+; CHECK-NEXT:    [[TMP20:%.*]] = getelementptr i64, i64* [[TMP18]], i64 2
+; CHECK-NEXT:    [[TMP21:%.*]] = bitcast i64* [[TMP20]] to <2 x i64>*
+; CHECK-NEXT:    [[WIDE_LOAD35:%.*]] = load <2 x i64>, <2 x i64>* [[TMP21]], align 4
+; CHECK-NEXT:    [[TMP22:%.*]] = add <2 x i64> [[WIDE_LOAD]], [[VEC_PHI]]
+; CHECK-NEXT:    [[TMP23:%.*]] = add <2 x i64> [[WIDE_LOAD31]], [[VEC_PHI30]]
+; CHECK-NEXT:    [[TMP24:%.*]] = add <2 x i64> [[TMP22]], [[WIDE_LOAD32]]
+; CHECK-NEXT:    [[TMP25:%.*]] = add <2 x i64> [[TMP23]], [[WIDE_LOAD33]]
+; CHECK-NEXT:    [[TMP26]] = add <2 x i64> [[TMP24]], [[WIDE_LOAD34]]
+; CHECK-NEXT:    [[TMP27]] = add <2 x i64> [[TMP25]], [[WIDE_LOAD35]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
-; CHECK-NEXT:    [[TMP23:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
-; CHECK-NEXT:    br i1 [[TMP23]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
+; CHECK-NEXT:    [[TMP28:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[TMP28]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    [[BIN_RDX:%.*]] = add <2 x i64> [[TMP22]], [[TMP21]]
-; CHECK-NEXT:    [[TMP24:%.*]] = call i64 @llvm.vector.reduce.add.v2i64(<2 x i64> [[BIN_RDX]])
-; CHECK-NEXT:    br label [[LOOP_PREHEADER37]]
+; CHECK-NEXT:    [[BIN_RDX:%.*]] = add <2 x i64> [[TMP27]], [[TMP26]]
+; CHECK-NEXT:    [[TMP29:%.*]] = call i64 @llvm.vector.reduce.add.v2i64(<2 x i64> [[BIN_RDX]])
+; CHECK-NEXT:    br label [[LOOP_PREHEADER36]]
 ; CHECK:       loop.preheader36:
 ; CHECK-NEXT:    [[IV_PH:%.*]] = phi i64 [ 1, [[LOOP_PREHEADER]] ], [ [[IND_END]], [[MIDDLE_BLOCK]] ]
-; CHECK-NEXT:    [[SUM_PH:%.*]] = phi i64 [ [[SUM_NEXT_PEEL]], [[LOOP_PREHEADER]] ], [ [[TMP24]], [[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    [[SUM_PH:%.*]] = phi i64 [ [[SUM_NEXT_PEEL]], [[LOOP_PREHEADER]] ], [ [[TMP29]], [[MIDDLE_BLOCK]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
-; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[IV_NEXT:%.*]], [[AT_WITH_INT_CONVERSION_EXIT24:%.*]] ], [ [[IV_PH]], [[LOOP_PREHEADER37]] ]
-; CHECK-NEXT:    [[SUM:%.*]] = phi i64 [ [[SUM_NEXT:%.*]], [[AT_WITH_INT_CONVERSION_EXIT24]] ], [ [[SUM_PH]], [[LOOP_PREHEADER37]] ]
+; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[IV_NEXT:%.*]], [[AT_WITH_INT_CONVERSION_EXIT24:%.*]] ], [ [[IV_PH]], [[LOOP_PREHEADER36]] ]
+; CHECK-NEXT:    [[SUM:%.*]] = phi i64 [ [[SUM_NEXT:%.*]], [[AT_WITH_INT_CONVERSION_EXIT24]] ], [ [[SUM_PH]], [[LOOP_PREHEADER36]] ]
 ; CHECK-NEXT:    [[INRANGE_I:%.*]] = icmp ult i64 [[SUB_I]], [[IV]]
 ; CHECK-NEXT:    br i1 [[INRANGE_I]], label [[ERROR_I:%.*]], label [[AT_WITH_INT_CONVERSION_EXIT:%.*]]
 ; CHECK:       error.i:
 ; CHECK-NEXT:    tail call void @error()
 ; CHECK-NEXT:    unreachable
 ; CHECK:       at_with_int_conversion.exit:
 ; CHECK-NEXT:    [[GEP_IDX_I:%.*]] = getelementptr i64, i64* [[START_I]], i64 [[IV]]
 ; CHECK-NEXT:    [[LV_I:%.*]] = load i64, i64* [[GEP_IDX_I]], align 4