1 ; RUN: opt -loop-unroll-and-jam -allow-unroll-and-jam -verify-loop-info < %s -S | FileCheck %s
2 ; RUN: opt -passes='loop(loop-unroll-and-jam),verify<loops>' -allow-unroll-and-jam < %s -S | FileCheck %s
4 ; Check that the newly created loops to not fail to be added to LI
5 ; This test deliberately disables UnJ on the middle loop, performing it instead on the
6 ; outer of 3 nested loops. The (new) inner loops need to be added to LI.
8 target datalayout = "e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-f80:128-n8:16:32:64-S128"
12 ; CHECK: for.cond17.preheader:
13 ; CHECK: br label %for.cond20.preheader
14 ; CHECK: for.cond20.preheader:
15 ; CHECK: br label %for.cond23.preheader
16 ; CHECK: for.cond23.preheader:
17 ; CHECK: br label %for.body25
19 ; CHECK: br i1 [[CMP24:%.*]], label %for.body25, label %for.inc45
21 ; CHECK: br label %for.body25.1
23 ; CHECK: br i1 [[CMP18_3:%.*]], label %for.cond20.preheader, label %for.end50
26 ; CHECK: for.body25.1:
27 ; CHECK: br i1 [[CMP24_1:%.*]], label %for.body25.1, label %for.inc45.1
29 ; CHECK: br label %for.body25.2
30 ; CHECK: for.body25.2:
31 ; CHECK: br i1 [[CMP24_2:%.*]], label %for.body25.2, label %for.inc45.2
33 ; CHECK: br label %for.body25.3
34 ; CHECK: for.body25.3:
35 ; CHECK: br i1 [[CMP24_3:%.*]], label %for.body25.3, label %for.inc45.3
37 ; CHECK: br i1 [[CMP21_3:%.*]], label %for.cond23.preheader, label %for.inc48
40 %A = alloca [8 x [8 x i32]], align 16
41 %B = alloca [8 x [8 x i32]], align 16
42 %C = alloca [8 x [8 x i32]], align 16
43 br label %for.cond17.preheader
45 for.cond17.preheader: ; preds = %for.inc14
46 br label %for.cond20.preheader
48 for.cond20.preheader: ; preds = %for.cond17.preheader, %for.inc48
49 %i.13 = phi i32 [ 0, %for.cond17.preheader ], [ %inc49, %for.inc48 ]
50 br label %for.cond23.preheader
52 for.cond23.preheader: ; preds = %for.cond20.preheader, %for.inc45
53 %j.12 = phi i32 [ 0, %for.cond20.preheader ], [ %inc46, %for.inc45 ]
56 for.body25: ; preds = %for.cond23.preheader, %for.body25
57 %k.01 = phi i32 [ 0, %for.cond23.preheader ], [ %inc43, %for.body25 ]
58 %idxprom26 = zext i32 %i.13 to i64
59 %idxprom28 = zext i32 %j.12 to i64
60 %arrayidx29 = getelementptr inbounds [8 x [8 x i32]], [8 x [8 x i32]]* %C, i64 0, i64 %idxprom26, i64 %idxprom28
61 %0 = load i32, i32* %arrayidx29, align 4
62 %idxprom30 = zext i32 %i.13 to i64
63 %idxprom32 = zext i32 %k.01 to i64
64 %arrayidx33 = getelementptr inbounds [8 x [8 x i32]], [8 x [8 x i32]]* %A, i64 0, i64 %idxprom30, i64 %idxprom32
65 %1 = load i32, i32* %arrayidx33, align 4
66 %idxprom34 = zext i32 %k.01 to i64
67 %idxprom36 = zext i32 %j.12 to i64
68 %arrayidx37 = getelementptr inbounds [8 x [8 x i32]], [8 x [8 x i32]]* %B, i64 0, i64 %idxprom34, i64 %idxprom36
69 %2 = load i32, i32* %arrayidx37, align 4
70 %mul = mul nsw i32 %1, %2
71 %add = add nsw i32 %0, %mul
72 %idxprom38 = zext i32 %i.13 to i64
73 %idxprom40 = zext i32 %j.12 to i64
74 %arrayidx41 = getelementptr inbounds [8 x [8 x i32]], [8 x [8 x i32]]* %C, i64 0, i64 %idxprom38, i64 %idxprom40
75 store i32 %add, i32* %arrayidx41, align 4
76 %inc43 = add nuw nsw i32 %k.01, 1
77 %cmp24 = icmp ult i32 %k.01, 7
78 br i1 %cmp24, label %for.body25, label %for.inc45
80 for.inc45: ; preds = %for.body25
81 %inc46 = add nuw nsw i32 %j.12, 1
82 %cmp21 = icmp ult i32 %j.12, 7
83 br i1 %cmp21, label %for.cond23.preheader, label %for.inc48, !llvm.loop !7
85 for.inc48: ; preds = %for.inc45
86 %inc49 = add nuw nsw i32 %i.13, 1
87 %cmp18 = icmp ult i32 %i.13, 7
88 br i1 %cmp18, label %for.cond20.preheader, label %for.end50, !llvm.loop !5
90 for.end50: ; preds = %for.inc48
94 !5 = distinct !{!5, !6}
95 !6 = !{!"llvm.loop.unroll_and_jam.count", i32 4}
96 !7 = distinct !{!7, !8}
97 !8 = !{!"llvm.loop.unroll_and_jam.disable"}