test/Transforms/LoopUnroll/scevunroll.ll

; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
; RUN: opt < %s -S -passes='loop(indvars),loop-unroll' -verify-loop-info | FileCheck %s
;
; Unit tests for loop unrolling using ScalarEvolution to compute trip counts.
;
; Indvars is run first to generate an "old" SCEV result. Some unit
; tests may check that SCEV is properly invalidated between passes.

; Completely unroll loops without a canonical IV.
define i32 @sansCanonical(ptr %base) nounwind {
; CHECK-LABEL: @sansCanonical(
; CHECK-NEXT:  entry:
; CHECK-NEXT:    br label [[WHILE_BODY:%.*]]
; CHECK:       while.body:
; CHECK-NEXT:    [[ADR:%.*]] = getelementptr inbounds i32, ptr [[BASE:%.*]], i64 9
; CHECK-NEXT:    [[TMP:%.*]] = load i32, ptr [[ADR]], align 8
; CHECK-NEXT:    [[ADR_1:%.*]] = getelementptr inbounds i32, ptr [[BASE]], i64 8
; CHECK-NEXT:    [[TMP_1:%.*]] = load i32, ptr [[ADR_1]], align 8
; CHECK-NEXT:    [[SUM_NEXT_1:%.*]] = add i32 [[TMP]], [[TMP_1]]
; CHECK-NEXT:    [[ADR_2:%.*]] = getelementptr inbounds i32, ptr [[BASE]], i64 7
; CHECK-NEXT:    [[TMP_2:%.*]] = load i32, ptr [[ADR_2]], align 8
; CHECK-NEXT:    [[SUM_NEXT_2:%.*]] = add i32 [[SUM_NEXT_1]], [[TMP_2]]
; CHECK-NEXT:    [[ADR_3:%.*]] = getelementptr inbounds i32, ptr [[BASE]], i64 6
; CHECK-NEXT:    [[TMP_3:%.*]] = load i32, ptr [[ADR_3]], align 8
; CHECK-NEXT:    [[SUM_NEXT_3:%.*]] = add i32 [[SUM_NEXT_2]], [[TMP_3]]
; CHECK-NEXT:    [[ADR_4:%.*]] = getelementptr inbounds i32, ptr [[BASE]], i64 5
; CHECK-NEXT:    [[TMP_4:%.*]] = load i32, ptr [[ADR_4]], align 8
; CHECK-NEXT:    [[SUM_NEXT_4:%.*]] = add i32 [[SUM_NEXT_3]], [[TMP_4]]
; CHECK-NEXT:    [[ADR_5:%.*]] = getelementptr inbounds i32, ptr [[BASE]], i64 4
; CHECK-NEXT:    [[TMP_5:%.*]] = load i32, ptr [[ADR_5]], align 8
; CHECK-NEXT:    [[SUM_NEXT_5:%.*]] = add i32 [[SUM_NEXT_4]], [[TMP_5]]
; CHECK-NEXT:    [[ADR_6:%.*]] = getelementptr inbounds i32, ptr [[BASE]], i64 3
; CHECK-NEXT:    [[TMP_6:%.*]] = load i32, ptr [[ADR_6]], align 8
; CHECK-NEXT:    [[SUM_NEXT_6:%.*]] = add i32 [[SUM_NEXT_5]], [[TMP_6]]
; CHECK-NEXT:    [[ADR_7:%.*]] = getelementptr inbounds i32, ptr [[BASE]], i64 2
; CHECK-NEXT:    [[TMP_7:%.*]] = load i32, ptr [[ADR_7]], align 8
; CHECK-NEXT:    [[SUM_NEXT_7:%.*]] = add i32 [[SUM_NEXT_6]], [[TMP_7]]
; CHECK-NEXT:    [[ADR_8:%.*]] = getelementptr inbounds i32, ptr [[BASE]], i64 1
; CHECK-NEXT:    [[TMP_8:%.*]] = load i32, ptr [[ADR_8]], align 8
; CHECK-NEXT:    [[SUM_NEXT_8:%.*]] = add i32 [[SUM_NEXT_7]], [[TMP_8]]
; CHECK-NEXT:    ret i32 [[SUM_NEXT_8]]
;
entry:
  br label %while.body

while.body:
  %iv = phi i64 [ 10, %entry ], [ %iv.next, %while.body ]
  %sum = phi i32 [ 0, %entry ], [ %sum.next, %while.body ]
  %iv.next = add i64 %iv, -1
  %adr = getelementptr inbounds i32, ptr %base, i64 %iv.next
  %tmp = load i32, ptr %adr, align 8
  %sum.next = add i32 %sum, %tmp
  %iv.narrow = trunc i64 %iv.next to i32
  %cmp.i65 = icmp sgt i32 %iv.narrow, 0
  br i1 %cmp.i65, label %while.body, label %exit

exit:
  ret i32 %sum
}

; SCEV unrolling properly handles loops with multiple exits. In this
; case, the computed trip count based on a canonical IV is *not* for a
; latch block.
define i64 @earlyLoopTest(ptr %base) nounwind {
; CHECK-LABEL: @earlyLoopTest(
; CHECK-NEXT:  entry:
; CHECK-NEXT:    br label [[LOOP:%.*]]
; CHECK:       loop:
; CHECK-NEXT:    [[VAL:%.*]] = load i64, ptr [[BASE:%.*]], align 4
; CHECK-NEXT:    br label [[TAIL:%.*]]
; CHECK:       tail:
; CHECK-NEXT:    [[CMP2:%.*]] = icmp ne i64 [[VAL]], 0
; CHECK-NEXT:    br i1 [[CMP2]], label [[LOOP_1:%.*]], label [[EXIT2:%.*]]
; CHECK:       loop.1:
; CHECK-NEXT:    [[ADR_1:%.*]] = getelementptr i64, ptr [[BASE]], i64 1
; CHECK-NEXT:    [[VAL_1:%.*]] = load i64, ptr [[ADR_1]], align 4
; CHECK-NEXT:    [[S_NEXT_1:%.*]] = add i64 [[VAL]], [[VAL_1]]
; CHECK-NEXT:    br label [[TAIL_1:%.*]]
; CHECK:       tail.1:
; CHECK-NEXT:    [[CMP2_1:%.*]] = icmp ne i64 [[VAL_1]], 0
; CHECK-NEXT:    br i1 [[CMP2_1]], label [[LOOP_2:%.*]], label [[EXIT2]]
; CHECK:       loop.2:
; CHECK-NEXT:    [[ADR_2:%.*]] = getelementptr i64, ptr [[BASE]], i64 2
; CHECK-NEXT:    [[VAL_2:%.*]] = load i64, ptr [[ADR_2]], align 4
; CHECK-NEXT:    [[S_NEXT_2:%.*]] = add i64 [[S_NEXT_1]], [[VAL_2]]
; CHECK-NEXT:    br label [[TAIL_2:%.*]]
; CHECK:       tail.2:
; CHECK-NEXT:    [[CMP2_2:%.*]] = icmp ne i64 [[VAL_2]], 0
; CHECK-NEXT:    br i1 [[CMP2_2]], label [[LOOP_3:%.*]], label [[EXIT2]]
; CHECK:       loop.3:
; CHECK-NEXT:    [[ADR_3:%.*]] = getelementptr i64, ptr [[BASE]], i64 3
; CHECK-NEXT:    [[VAL_3:%.*]] = load i64, ptr [[ADR_3]], align 4
; CHECK-NEXT:    [[S_NEXT_3:%.*]] = add i64 [[S_NEXT_2]], [[VAL_3]]
; CHECK-NEXT:    br i1 false, label [[TAIL_3:%.*]], label [[EXIT1:%.*]]
; CHECK:       tail.3:
; CHECK-NEXT:    br label [[EXIT2]]
; CHECK:       exit1:
; CHECK-NEXT:    [[S_LCSSA:%.*]] = phi i64 [ [[S_NEXT_2]], [[LOOP_3]] ]
; CHECK-NEXT:    ret i64 [[S_LCSSA]]
; CHECK:       exit2:
; CHECK-NEXT:    [[S_NEXT_LCSSA1:%.*]] = phi i64 [ [[VAL]], [[TAIL]] ], [ [[S_NEXT_1]], [[TAIL_1]] ], [ [[S_NEXT_2]], [[TAIL_2]] ], [ [[S_NEXT_3]], [[TAIL_3]] ]
; CHECK-NEXT:    ret i64 [[S_NEXT_LCSSA1]]
;
entry:
  br label %loop

loop:
  %iv = phi i64 [ 0, %entry ], [ %inc, %tail ]
  %s = phi i64 [ 0, %entry ], [ %s.next, %tail ]
  %adr = getelementptr i64, ptr %base, i64 %iv
  %val = load i64, ptr %adr
  %s.next = add i64 %s, %val
  %inc = add i64 %iv, 1
  %cmp = icmp ne i64 %inc, 4
  br i1 %cmp, label %tail, label %exit1

tail:
  %cmp2 = icmp ne i64 %val, 0
  br i1 %cmp2, label %loop, label %exit2

exit1:
  ret i64 %s

exit2:
  ret i64 %s.next
}

; SCEV properly unrolls multi-exit loops.
define i32 @multiExit(ptr %base) nounwind {
; CHECK-LABEL: @multiExit(
; CHECK-NEXT:  entry:
; CHECK-NEXT:    br label [[L1:%.*]]
; CHECK:       l1:
; CHECK-NEXT:    [[VAL:%.*]] = load i32, ptr [[BASE:%.*]], align 4
; CHECK-NEXT:    br i1 false, label [[L2:%.*]], label [[EXIT1:%.*]]
; CHECK:       l2:
; CHECK-NEXT:    ret i32 [[VAL]]
; CHECK:       exit1:
; CHECK-NEXT:    ret i32 1
;
entry:
  br label %l1
l1:
  %iv1 = phi i32 [ 0, %entry ], [ %inc1, %l2 ]
  %iv2 = phi i32 [ 0, %entry ], [ %inc2, %l2 ]
  %inc1 = add i32 %iv1, 1
  %inc2 = add i32 %iv2, 1
  %adr = getelementptr i32, ptr %base, i32 %iv1
  %val = load i32, ptr %adr
  %cmp1 = icmp slt i32 %iv1, 5
  br i1 %cmp1, label %l2, label %exit1
l2:
  %cmp2 = icmp slt i32 %iv2, 10
  br i1 %cmp2, label %l1, label %exit2
exit1:
  ret i32 1
exit2:
  ret i32 %val
}


; SCEV can unroll a multi-exit loops even if the latch block has no
; known trip count, but an early exit has a known trip count. In this
; case we must be careful not to optimize the latch branch away.
define i32 @multiExitIncomplete(ptr %base) nounwind {
; CHECK-LABEL: @multiExitIncomplete(
; CHECK-NEXT:  entry:
; CHECK-NEXT:    br label [[L1:%.*]]
; CHECK:       l1:
; CHECK-NEXT:    [[VAL:%.*]] = load i32, ptr [[BASE:%.*]], align 4
; CHECK-NEXT:    br label [[L2:%.*]]
; CHECK:       l2:
; CHECK-NEXT:    br label [[L3:%.*]]
; CHECK:       l3:
; CHECK-NEXT:    [[CMP3:%.*]] = icmp ne i32 [[VAL]], 0
; CHECK-NEXT:    br i1 [[CMP3]], label [[L1_1:%.*]], label [[EXIT3:%.*]]
; CHECK:       l1.1:
; CHECK-NEXT:    [[ADR_1:%.*]] = getelementptr i32, ptr [[BASE]], i32 1
; CHECK-NEXT:    [[VAL_1:%.*]] = load i32, ptr [[ADR_1]], align 4
; CHECK-NEXT:    br label [[L2_1:%.*]]
; CHECK:       l2.1:
; CHECK-NEXT:    br label [[L3_1:%.*]]
; CHECK:       l3.1:
; CHECK-NEXT:    [[CMP3_1:%.*]] = icmp ne i32 [[VAL_1]], 0
; CHECK-NEXT:    br i1 [[CMP3_1]], label [[L1_2:%.*]], label [[EXIT3]]
; CHECK:       l1.2:
; CHECK-NEXT:    [[ADR_2:%.*]] = getelementptr i32, ptr [[BASE]], i32 2
; CHECK-NEXT:    [[VAL_2:%.*]] = load i32, ptr [[ADR_2]], align 4
; CHECK-NEXT:    br label [[L2_2:%.*]]
; CHECK:       l2.2:
; CHECK-NEXT:    br label [[L3_2:%.*]]
; CHECK:       l3.2:
; CHECK-NEXT:    [[CMP3_2:%.*]] = icmp ne i32 [[VAL_2]], 0
; CHECK-NEXT:    br i1 [[CMP3_2]], label [[L1_3:%.*]], label [[EXIT3]]
; CHECK:       l1.3:
; CHECK-NEXT:    [[ADR_3:%.*]] = getelementptr i32, ptr [[BASE]], i32 3
; CHECK-NEXT:    [[VAL_3:%.*]] = load i32, ptr [[ADR_3]], align 4
; CHECK-NEXT:    br label [[L2_3:%.*]]
; CHECK:       l2.3:
; CHECK-NEXT:    br label [[L3_3:%.*]]
; CHECK:       l3.3:
; CHECK-NEXT:    [[CMP3_3:%.*]] = icmp ne i32 [[VAL_3]], 0
; CHECK-NEXT:    br i1 [[CMP3_3]], label [[L1_4:%.*]], label [[EXIT3]]
; CHECK:       l1.4:
; CHECK-NEXT:    [[ADR_4:%.*]] = getelementptr i32, ptr [[BASE]], i32 4
; CHECK-NEXT:    [[VAL_4:%.*]] = load i32, ptr [[ADR_4]], align 4
; CHECK-NEXT:    br label [[L2_4:%.*]]
; CHECK:       l2.4:
; CHECK-NEXT:    br label [[L3_4:%.*]]
; CHECK:       l3.4:
; CHECK-NEXT:    [[CMP3_4:%.*]] = icmp ne i32 [[VAL_4]], 0
; CHECK-NEXT:    br i1 [[CMP3_4]], label [[L1_5:%.*]], label [[EXIT3]]
; CHECK:       l1.5:
; CHECK-NEXT:    br i1 false, label [[L2_5:%.*]], label [[EXIT1:%.*]]
; CHECK:       l2.5:
; CHECK-NEXT:    br i1 true, label [[L3_5:%.*]], label [[EXIT2:%.*]]
; CHECK:       l3.5:
; CHECK-NEXT:    br label [[EXIT3]]
; CHECK:       exit1:
; CHECK-NEXT:    ret i32 1
; CHECK:       exit2:
; CHECK-NEXT:    ret i32 2
; CHECK:       exit3:
; CHECK-NEXT:    ret i32 3
;
entry:
  br label %l1
l1:
  %iv1 = phi i32 [ 0, %entry ], [ %inc1, %l3 ]
  %iv2 = phi i32 [ 0, %entry ], [ %inc2, %l3 ]
  %inc1 = add i32 %iv1, 1
  %inc2 = add i32 %iv2, 1
  %adr = getelementptr i32, ptr %base, i32 %iv1
  %val = load i32, ptr %adr
  %cmp1 = icmp slt i32 %iv1, 5
  br i1 %cmp1, label %l2, label %exit1
l2:
  %cmp2 = icmp slt i32 %iv2, 10
  br i1 %cmp2, label %l3, label %exit2
l3:
  %cmp3 = icmp ne i32 %val, 0
  br i1 %cmp3, label %l1, label %exit3

exit1:
  ret i32 1
exit2:
  ret i32 2
exit3:
  ret i32 3
}

; When loop unroll merges a loop exit with one of its parent loop's
; exits, SCEV must forget its ExitNotTaken info.
define void @nestedUnroll() nounwind {
; CHECK-LABEL: @nestedUnroll(
; CHECK-NEXT:  entry:
; CHECK-NEXT:    br label [[FOR_INC:%.*]]
; CHECK:       for.inc:
; CHECK-NEXT:    br label [[FOR_BODY38:%.*]]
; CHECK:       for.body38:
; CHECK-NEXT:    br label [[FOR_BODY43:%.*]]
; CHECK:       for.body43:
; CHECK-NEXT:    br label [[FOR_BODY87:%.*]]
; CHECK:       for.body87:
; CHECK-NEXT:    br label [[FOR_BODY87]]
;
entry:
  br label %for.inc

for.inc:
  br i1 false, label %for.inc, label %for.body38.preheader

for.body38.preheader:
  br label %for.body38

for.body38:
  %i.113 = phi i32 [ %inc76, %for.inc74 ], [ 0, %for.body38.preheader ]
  %mul48 = mul nsw i32 %i.113, 6
  br label %for.body43

for.body43:
  %j.011 = phi i32 [ 0, %for.body38 ], [ %inc72, %for.body43 ]
  %add49 = add nsw i32 %j.011, %mul48
  %sh_prom50 = zext i32 %add49 to i64
  %inc72 = add nsw i32 %j.011, 1
  br i1 false, label %for.body43, label %for.inc74

for.inc74:
  %inc76 = add nsw i32 %i.113, 1
  br i1 false, label %for.body38, label %for.body87.preheader

for.body87.preheader:
  br label %for.body87

for.body87:
  br label %for.body87
}

; PR16130: clang produces incorrect code with loop/expression at -O2
; rdar:14036816 loop-unroll makes assumptions about undefined behavior
;
; The loop latch is assumed to exit after the first iteration because
; of the induction variable's NSW flag. However, the loop latch's
; equality test is skipped and the loop exits after the second
; iteration via the early exit. So loop unrolling cannot assume that
; the loop latch's exit count of zero is an upper bound on the number
; of iterations.
define void @nsw_latch(ptr %a) nounwind {
; CHECK-LABEL: @nsw_latch(
; CHECK-NEXT:  entry:
; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
; CHECK:       for.body:
; CHECK-NEXT:    br label [[FOR_COND:%.*]]
; CHECK:       for.cond:
; CHECK-NEXT:    br i1 false, label [[RETURN:%.*]], label [[FOR_BODY_1:%.*]]
; CHECK:       for.body.1:
; CHECK-NEXT:    br i1 false, label [[FOR_COND_1:%.*]], label [[RETURN]]
; CHECK:       for.cond.1:
; CHECK-NEXT:    br label [[RETURN]]
; CHECK:       return:
; CHECK-NEXT:    [[B_03_LCSSA:%.*]] = phi i32 [ 0, [[FOR_COND]] ], [ 8, [[FOR_BODY_1]] ], [ 0, [[FOR_COND_1]] ]
; CHECK-NEXT:    [[RETVAL_0:%.*]] = phi i32 [ 0, [[FOR_COND]] ], [ 1, [[FOR_BODY_1]] ], [ 0, [[FOR_COND_1]] ]
; CHECK-NEXT:    store i32 [[B_03_LCSSA]], ptr [[A:%.*]], align 4
; CHECK-NEXT:    ret void
;
entry:
  br label %for.body

for.body:                                         ; preds = %for.cond, %entry
  %b.03 = phi i32 [ 0, %entry ], [ %add, %for.cond ]
  %tobool = icmp eq i32 %b.03, 0
  %add = add nsw i32 %b.03, 8
  br i1 %tobool, label %for.cond, label %return

for.cond:                                         ; preds = %for.body
  %cmp = icmp eq i32 %add, 13
  br i1 %cmp, label %return, label %for.body

return:                                           ; preds = %for.body, %for.cond
  %b.03.lcssa = phi i32 [ %b.03, %for.body ], [ %b.03, %for.cond ]
  %retval.0 = phi i32 [ 1, %for.body ], [ 0, %for.cond ]
  store i32 %b.03.lcssa, ptr %a, align 4
  ret void
}

; Test case for PR56044. Check that SCEVs for exit phis are properly invalidated.
define i32 @test_pr56044(ptr %src, i32 %a) {
; CHECK-LABEL: @test_pr56044(
; CHECK-NEXT:  entry:
; CHECK-NEXT:    br label [[LOOP_1_PEEL_BEGIN:%.*]]
; CHECK:       loop.1.peel.begin:
; CHECK-NEXT:    br label [[LOOP_1_PEEL:%.*]]
; CHECK:       loop.1.peel:
; CHECK-NEXT:    call void @fn(i32 5)
; CHECK-NEXT:    [[L_PEEL:%.*]] = load i64, ptr [[SRC:%.*]], align 8
; CHECK-NEXT:    [[ADD_PEEL:%.*]] = add i64 [[L_PEEL]], [[L_PEEL]]
; CHECK-NEXT:    [[EC_1_PEEL:%.*]] = icmp sgt i32 [[A:%.*]], 4
; CHECK-NEXT:    br i1 [[EC_1_PEEL]], label [[MID:%.*]], label [[LOOP_1_PEEL_NEXT:%.*]]
; CHECK:       loop.1.peel.next:
; CHECK-NEXT:    br label [[LOOP_1_PEEL_NEXT1:%.*]]
; CHECK:       loop.1.peel.next1:
; CHECK-NEXT:    br label [[ENTRY_PEEL_NEWPH:%.*]]
; CHECK:       entry.peel.newph:
; CHECK-NEXT:    br label [[LOOP_1:%.*]]
; CHECK:       loop.1:
; CHECK-NEXT:    call void @fn(i32 18)
; CHECK-NEXT:    [[L:%.*]] = load i64, ptr [[SRC]], align 8
; CHECK-NEXT:    [[ADD:%.*]] = add i64 [[L]], [[L]]
; CHECK-NEXT:    [[EC_1:%.*]] = icmp sgt i32 [[A]], 4
; CHECK-NEXT:    br i1 [[EC_1]], label [[MID_LOOPEXIT:%.*]], label [[LOOP_1]], !llvm.loop [[LOOP0:![0-9]+]]
; CHECK:       mid.loopexit:
; CHECK-NEXT:    [[LCSSA_1_PH:%.*]] = phi i64 [ [[ADD]], [[LOOP_1]] ]
; CHECK-NEXT:    br label [[MID]]
; CHECK:       mid:
; CHECK-NEXT:    [[LCSSA_1:%.*]] = phi i64 [ [[ADD_PEEL]], [[LOOP_1_PEEL]] ], [ [[LCSSA_1_PH]], [[MID_LOOPEXIT]] ]
; CHECK-NEXT:    [[TRUNC:%.*]] = trunc i64 [[LCSSA_1]] to i32
; CHECK-NEXT:    [[ADD_2:%.*]] = sub i32 [[A]], [[TRUNC]]
; CHECK-NEXT:    br label [[LOOP_2_PEEL_BEGIN:%.*]]
; CHECK:       loop.2.peel.begin:
; CHECK-NEXT:    br label [[LOOP_2_PEEL:%.*]]
; CHECK:       loop.2.peel:
; CHECK-NEXT:    [[IV_2_NEXT_PEEL:%.*]] = add i32 0, [[ADD_2]]
; CHECK-NEXT:    [[IV_1_NEXT_PEEL:%.*]] = add nuw nsw i32 0, 1
; CHECK-NEXT:    [[EC_2_PEEL:%.*]] = icmp ult i32 [[IV_1_NEXT_PEEL]], 12345
; CHECK-NEXT:    br i1 [[EC_2_PEEL]], label [[LOOP_2_PEEL_NEXT:%.*]], label [[EXIT:%.*]]
; CHECK:       loop.2.peel.next:
; CHECK-NEXT:    br label [[LOOP_2_PEEL_NEXT2:%.*]]
; CHECK:       loop.2.peel.next2:
; CHECK-NEXT:    br label [[MID_PEEL_NEWPH:%.*]]
; CHECK:       mid.peel.newph:
; CHECK-NEXT:    br label [[LOOP_2:%.*]]
; CHECK:       loop.2:
; CHECK-NEXT:    [[IV_1:%.*]] = phi i32 [ [[IV_1_NEXT_PEEL]], [[MID_PEEL_NEWPH]] ], [ [[IV_1_NEXT:%.*]], [[LOOP_2]] ]
; CHECK-NEXT:    [[IV_2:%.*]] = phi i32 [ [[IV_2_NEXT_PEEL]], [[MID_PEEL_NEWPH]] ], [ [[IV_2_NEXT:%.*]], [[LOOP_2]] ]
; CHECK-NEXT:    [[IV_2_NEXT]] = add i32 2, [[IV_2]]
; CHECK-NEXT:    [[IV_1_NEXT]] = add nuw nsw i32 [[IV_1]], 1
; CHECK-NEXT:    [[EC_2:%.*]] = icmp ult i32 [[IV_1_NEXT]], 12345
; CHECK-NEXT:    br i1 [[EC_2]], label [[LOOP_2]], label [[EXIT_LOOPEXIT:%.*]], !llvm.loop [[LOOP2:![0-9]+]]
; CHECK:       exit.loopexit:
; CHECK-NEXT:    [[LCSSA_2_PH:%.*]] = phi i32 [ [[IV_2_NEXT]], [[LOOP_2]] ]
; CHECK-NEXT:    br label [[EXIT]]
; CHECK:       exit:
; CHECK-NEXT:    [[LCSSA_2:%.*]] = phi i32 [ [[IV_2_NEXT_PEEL]], [[LOOP_2_PEEL]] ], [ [[LCSSA_2_PH]], [[EXIT_LOOPEXIT]] ]
; CHECK-NEXT:    ret i32 [[LCSSA_2]]
;
entry:
  br label %loop.1

loop.1:
  %p.1 = phi i32 [ 5, %entry ], [ 18, %loop.1 ]
  call void @fn(i32 %p.1)
  %l = load i64, ptr %src, align 8
  %add = add i64 %l, %l
  %ec.1 = icmp sgt i32 %a, 4
  br i1 %ec.1, label %mid, label %loop.1

mid:
  %lcssa.1 = phi i64 [ %add, %loop.1 ]
  %trunc = trunc i64 %lcssa.1 to i32
  %add.2 = sub i32 %a, %trunc
  br label %loop.2

loop.2:
  %iv.1 = phi i32 [ 0, %mid ], [ %iv.1.next, %loop.2 ]
  %iv.2 = phi i32 [ %add.2, %mid ], [ %iv.2.next, %loop.2 ]
  %p.2 = phi i32 [ 0, %mid ], [ 2, %loop.2 ]
  %iv.2.next = add i32 %p.2, %iv.2
  %iv.1.next = add nuw nsw i32 %iv.1, 1
  %ec.2 = icmp ult i32 %iv.1.next, 12345
  br i1 %ec.2, label %loop.2, label %exit

exit:
  %lcssa.2 = phi i32 [ %iv.2.next, %loop.2 ]
  ret i32 %lcssa.2
}

declare void @fn(i32)