1 ; Test optimization remarks generated by the LoopInterchange pass.
3 ; RUN: opt < %s -basicaa -loop-interchange -verify-dom-info -verify-loop-info \
4 ; RUN: -pass-remarks-output=%t -pass-remarks-missed='loop-interchange' \
5 ; RUN: -pass-remarks='loop-interchange' -S
6 ; RUN: cat %t | FileCheck %s
8 @A = common global [100 x [100 x i32]] zeroinitializer
9 @B = common global [100 x [100 x i32]] zeroinitializer
10 @C = common global [100 x i32] zeroinitializer
12 ;;---------------------------------------Test case 01---------------------------------
13 ;; Loops interchange is not profitable.
14 ;; for(int i=1;i<N;i++)
15 ;; for(int j=1;j<N;j++)
16 ;; A[i-1][j-1] = A[i - 1][j-1] + B[i][j];
18 define void @test01(i32 %N){
20 %cmp31 = icmp sgt i32 %N, 1
21 br i1 %cmp31, label %for.cond1.preheader.lr.ph, label %for.end19
23 for.cond1.preheader.lr.ph:
25 br label %for.body3.lr.ph
28 %indvars.iv34 = phi i64 [ 1, %for.cond1.preheader.lr.ph ], [ %indvars.iv.next35, %for.inc17 ]
29 %1 = add nsw i64 %indvars.iv34, -1
33 %indvars.iv = phi i64 [ 1, %for.body3.lr.ph ], [ %indvars.iv.next, %for.body3 ]
34 %2 = add nsw i64 %indvars.iv, -1
35 %arrayidx6 = getelementptr inbounds [100 x [100 x i32]], [100 x [100 x i32]]* @A, i64 0, i64 %1, i64 %2
36 %3 = load i32, i32* %arrayidx6
37 %arrayidx10 = getelementptr inbounds [100 x [100 x i32]], [100 x [100 x i32]]* @B, i64 0, i64 %indvars.iv34, i64 %indvars.iv
38 %4 = load i32, i32* %arrayidx10
39 %add = add nsw i32 %4, %3
40 store i32 %add, i32* %arrayidx6
41 %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
42 %lftr.wideiv = trunc i64 %indvars.iv to i32
43 %exitcond = icmp eq i32 %lftr.wideiv, %0
44 br i1 %exitcond, label %for.inc17, label %for.body3
47 %indvars.iv.next35 = add nuw nsw i64 %indvars.iv34, 1
48 %lftr.wideiv37 = trunc i64 %indvars.iv34 to i32
49 %exitcond38 = icmp eq i32 %lftr.wideiv37, %0
50 br i1 %exitcond38, label %for.end19, label %for.body3.lr.ph
57 ; CHECK-NEXT: Pass: loop-interchange
58 ; CHECK-NEXT: Name: Dependence
59 ; CHECK-NEXT: Function: test01
61 ; CHECK-NEXT: - String: Cannot interchange loops due to dependences.
64 ;;--------------------------------------Test case 02------------------------------------
65 ;; [FIXME] This loop though valid is currently not interchanged due to the
66 ;; limitation that we cannot split the inner loop latch due to multiple use of inner induction
67 ;; variable.(used to increment the loop counter and to access A[j+1][i+1]
68 ;; for(int i=0;i<N-1;i++)
69 ;; for(int j=1;j<N-1;j++)
70 ;; A[j+1][i+1] = A[j+1][i+1] + k;
72 define void @test02(i32 %k, i32 %N) {
74 %sub = add nsw i32 %N, -1
75 %cmp26 = icmp sgt i32 %N, 1
76 br i1 %cmp26, label %for.cond1.preheader.lr.ph, label %for.end17
78 for.cond1.preheader.lr.ph:
79 %cmp324 = icmp sgt i32 %sub, 1
81 %1 = sext i32 %sub to i64
82 br label %for.cond1.preheader
85 %cmp = icmp slt i64 %indvars.iv.next29, %1
86 br i1 %cmp, label %for.cond1.preheader, label %for.end17
89 %indvars.iv28 = phi i64 [ 0, %for.cond1.preheader.lr.ph ], [ %indvars.iv.next29, %for.cond.loopexit ]
90 %indvars.iv.next29 = add nuw nsw i64 %indvars.iv28, 1
91 br i1 %cmp324, label %for.body4, label %for.cond.loopexit
94 %indvars.iv = phi i64 [ %indvars.iv.next, %for.body4 ], [ 1, %for.cond1.preheader ]
95 %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
96 %arrayidx7 = getelementptr inbounds [100 x [100 x i32]], [100 x [100 x i32]]* @A, i64 0, i64 %indvars.iv.next, i64 %indvars.iv.next29
97 %2 = load i32, i32* %arrayidx7
98 %add8 = add nsw i32 %2, %k
99 store i32 %add8, i32* %arrayidx7
100 %lftr.wideiv = trunc i64 %indvars.iv to i32
101 %exitcond = icmp eq i32 %lftr.wideiv, %0
102 br i1 %exitcond, label %for.cond.loopexit, label %for.body4
109 ; CHECK-NEXT: Pass: loop-interchange
110 ; CHECK-NEXT: Name: Dependence
111 ; CHECK-NEXT: Function: test02
113 ; CHECK-NEXT: - String: Cannot interchange loops due to dependences.
116 ;;-----------------------------------Test case 03-------------------------------
117 ;; Test to make sure we can handle output dependencies.
119 ;; for (int i = 0; i < 2; ++i)
120 ;; for(int j = 0; j < 3; ++j) {
125 @A10 = local_unnamed_addr global [3 x [3 x i32]] zeroinitializer, align 16
127 define void @test03() {
129 br label %for.cond1.preheader
131 for.cond.loopexit: ; preds = %for.body4
132 %exitcond28 = icmp ne i64 %indvars.iv.next27, 2
133 br i1 %exitcond28, label %for.cond1.preheader, label %for.cond.cleanup
135 for.cond1.preheader: ; preds = %for.cond.loopexit, %entry
136 %indvars.iv26 = phi i64 [ 0, %entry ], [ %indvars.iv.next27, %for.cond.loopexit ]
137 %indvars.iv.next27 = add nuw nsw i64 %indvars.iv26, 1
140 for.cond.cleanup: ; preds = %for.cond.loopexit
143 for.body4: ; preds = %for.body4, %for.cond1.preheader
144 %indvars.iv = phi i64 [ 0, %for.cond1.preheader ], [ %indvars.iv.next, %for.body4 ]
145 %arrayidx6 = getelementptr inbounds [3 x [3 x i32]], [3 x [3 x i32]]* @A10, i64 0, i64 %indvars.iv, i64 %indvars.iv26
146 %tmp = trunc i64 %indvars.iv26 to i32
147 store i32 %tmp, i32* %arrayidx6, align 4
148 %arrayidx10 = getelementptr inbounds [3 x [3 x i32]], [3 x [3 x i32]]* @A10, i64 0, i64 %indvars.iv, i64 %indvars.iv.next27
149 %tmp1 = trunc i64 %indvars.iv to i32
150 store i32 %tmp1, i32* %arrayidx10, align 4
151 %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
152 %exitcond = icmp ne i64 %indvars.iv.next, 3
153 br i1 %exitcond, label %for.body4, label %for.cond.loopexit
157 ; CHECK-NEXT: Pass: loop-interchange
158 ; CHECK-NEXT: Name: Dependence
159 ; CHECK-NEXT: Function: test03
161 ; CHECK-NEXT: - String: Cannot interchange loops due to dependences.
164 ;;--------------------------------------Test case 04-------------------------------------
165 ;; Loops not tightly nested are not interchanged
166 ;; for(int j=0;j<N;j++) {
168 ;; for(int i=0;i<N;i++)
169 ;; A[j][i] = A[j][i]+B[j];
172 define void @test04(i32 %k, i32 %N){
174 %cmp30 = icmp sgt i32 %N, 0
175 br i1 %cmp30, label %for.body.lr.ph, label %for.end17
179 %1 = zext i32 %k to i64
183 %indvars.iv32 = phi i64 [ 0, %for.body.lr.ph ], [ %indvars.iv.next33, %for.inc15 ]
184 %2 = add nsw i64 %indvars.iv32, %1
185 %arrayidx = getelementptr inbounds [100 x i32], [100 x i32]* @C, i64 0, i64 %indvars.iv32
186 %3 = trunc i64 %2 to i32
187 store i32 %3, i32* %arrayidx
191 %indvars.iv = phi i64 [ 0, %for.body ], [ %indvars.iv.next, %for.body3 ]
192 %arrayidx7 = getelementptr inbounds [100 x [100 x i32]], [100 x [100 x i32]]* @A, i64 0, i64 %indvars.iv32, i64 %indvars.iv
193 %4 = load i32, i32* %arrayidx7
194 %add10 = add nsw i32 %3, %4
195 store i32 %add10, i32* %arrayidx7
196 %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
197 %lftr.wideiv = trunc i64 %indvars.iv to i32
198 %exitcond = icmp eq i32 %lftr.wideiv, %0
199 br i1 %exitcond, label %for.inc15, label %for.body3
202 %indvars.iv.next33 = add nuw nsw i64 %indvars.iv32, 1
203 %lftr.wideiv35 = trunc i64 %indvars.iv32 to i32
204 %exitcond36 = icmp eq i32 %lftr.wideiv35, %0
205 br i1 %exitcond36, label %for.end17, label %for.body
212 ; CHECK-NEXT: Pass: loop-interchange
213 ; CHECK-NEXT: Name: Dependence
214 ; CHECK-NEXT: Function: test04
216 ; CHECK-NEXT: - String: Cannot interchange loops due to dependences.