1; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
2; RUN: opt < %s -passes=loop-vectorize -force-vector-interleave=3 -force-vector-width=2 -S | FileCheck %s
3
4target datalayout = "e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-f80:128-n8:16:32:64-S128"
5
6; Make sure the loop is vectorized and unrolled under -Os without folding its
7; tail based on its trip-count being provably divisible by chosen VFxIC.
8
9define dso_local void @constTC(i32* noalias nocapture %A) optsize {
10; CHECK-LABEL: @constTC(
11; CHECK-NEXT:  entry:
12; CHECK-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
13; CHECK:       vector.ph:
14; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
15; CHECK:       vector.body:
16; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
17; CHECK-NEXT:    [[TMP0:%.*]] = add i32 [[INDEX]], 0
18; CHECK-NEXT:    [[TMP1:%.*]] = add i32 [[INDEX]], 2
19; CHECK-NEXT:    [[TMP2:%.*]] = add i32 [[INDEX]], 4
20; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i32 [[TMP0]]
21; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, i32* [[A]], i32 [[TMP1]]
22; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, i32* [[A]], i32 [[TMP2]]
23; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, i32* [[TMP3]], i32 0
24; CHECK-NEXT:    [[TMP7:%.*]] = bitcast i32* [[TMP6]] to <2 x i32>*
25; CHECK-NEXT:    store <2 x i32> <i32 13, i32 13>, <2 x i32>* [[TMP7]], align 1
26; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, i32* [[TMP3]], i32 2
27; CHECK-NEXT:    [[TMP9:%.*]] = bitcast i32* [[TMP8]] to <2 x i32>*
28; CHECK-NEXT:    store <2 x i32> <i32 13, i32 13>, <2 x i32>* [[TMP9]], align 1
29; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i32, i32* [[TMP3]], i32 4
30; CHECK-NEXT:    [[TMP11:%.*]] = bitcast i32* [[TMP10]] to <2 x i32>*
31; CHECK-NEXT:    store <2 x i32> <i32 13, i32 13>, <2 x i32>* [[TMP11]], align 1
32; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 6
33; CHECK-NEXT:    [[TMP12:%.*]] = icmp eq i32 [[INDEX_NEXT]], 1800
34; CHECK-NEXT:    br i1 [[TMP12]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], [[LOOP0:!llvm.loop !.*]]
35; CHECK:       middle.block:
36; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i32 1800, 1800
37; CHECK-NEXT:    br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
38; CHECK:       scalar.ph:
39; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 1800, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
40; CHECK-NEXT:    br label [[LOOP:%.*]]
41; CHECK:       loop:
42; CHECK-NEXT:    [[RIV:%.*]] = phi i32 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[RIVPLUS1:%.*]], [[LOOP]] ]
43; CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[A]], i32 [[RIV]]
44; CHECK-NEXT:    store i32 13, i32* [[ARRAYIDX]], align 1
45; CHECK-NEXT:    [[RIVPLUS1]] = add nuw nsw i32 [[RIV]], 1
46; CHECK-NEXT:    [[COND:%.*]] = icmp eq i32 [[RIVPLUS1]], 1800
47; CHECK-NEXT:    br i1 [[COND]], label [[EXIT]], label [[LOOP]], [[LOOP2:!llvm.loop !.*]]
48; CHECK:       exit:
49; CHECK-NEXT:    ret void
50;
51entry:
52  br label %loop
53
54loop:
55  %riv = phi i32 [ 0, %entry ], [ %rivPlus1, %loop ]
56  %arrayidx = getelementptr inbounds i32, i32* %A, i32 %riv
57  store i32 13, i32* %arrayidx, align 1
58  %rivPlus1 = add nuw nsw i32 %riv, 1
59  %cond = icmp eq i32 %rivPlus1, 1800
60  br i1 %cond, label %exit, label %loop
61
62exit:
63  ret void
64}
65