1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
|
; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 5
; RUN: opt -passes=loop-interchange -cache-line-size=64 -S < %s | FileCheck %s
; Check that nsw/nuw flags are dropped when interchanging loops.
;
; int sum = 0;
; for (int i = 0; i < 2; i++)
; for (int j = 0; j < 2; j++)
; sum += A[j][i];
;
define void @reduction_add(ptr %A) {
; CHECK-LABEL: define void @reduction_add(
; CHECK-SAME: ptr [[A:%.*]]) {
; CHECK-NEXT: [[ENTRY:.*:]]
; CHECK-NEXT: br label %[[FOR_J_PREHEADER:.*]]
; CHECK: [[FOR_I_HEADER_PREHEADER:.*]]:
; CHECK-NEXT: br label %[[FOR_I_HEADER:.*]]
; CHECK: [[FOR_I_HEADER]]:
; CHECK-NEXT: [[I:%.*]] = phi i32 [ [[I_INC:%.*]], %[[FOR_I_LATCH:.*]] ], [ 0, %[[FOR_I_HEADER_PREHEADER]] ]
; CHECK-NEXT: [[SUM_J:%.*]] = phi i32 [ [[SUM_J_NEXT:%.*]], %[[FOR_I_LATCH]] ], [ [[SUM_I:%.*]], %[[FOR_I_HEADER_PREHEADER]] ]
; CHECK-NEXT: br label %[[FOR_J_SPLIT1:.*]]
; CHECK: [[FOR_J_PREHEADER]]:
; CHECK-NEXT: br label %[[FOR_J:.*]]
; CHECK: [[FOR_J]]:
; CHECK-NEXT: [[J:%.*]] = phi i32 [ [[TMP0:%.*]], %[[FOR_J_SPLIT:.*]] ], [ 0, %[[FOR_J_PREHEADER]] ]
; CHECK-NEXT: [[SUM_I]] = phi i32 [ [[SUM_I_LCSSA:%.*]], %[[FOR_J_SPLIT]] ], [ 0, %[[FOR_J_PREHEADER]] ]
; CHECK-NEXT: br label %[[FOR_I_HEADER_PREHEADER]]
; CHECK: [[FOR_J_SPLIT1]]:
; CHECK-NEXT: [[IDX:%.*]] = getelementptr inbounds [2 x [2 x i32]], ptr [[A]], i32 0, i32 [[J]], i32 [[I]]
; CHECK-NEXT: [[A:%.*]] = load i32, ptr [[IDX]], align 4
; CHECK-NEXT: [[SUM_J_NEXT]] = add i32 [[SUM_J]], [[A]]
; CHECK-NEXT: [[J_INC:%.*]] = add i32 [[J]], 1
; CHECK-NEXT: [[CMP_J:%.*]] = icmp slt i32 [[J_INC]], 2
; CHECK-NEXT: br label %[[FOR_I_LATCH]]
; CHECK: [[FOR_J_SPLIT]]:
; CHECK-NEXT: [[SUM_I_LCSSA]] = phi i32 [ [[SUM_J_NEXT]], %[[FOR_I_LATCH]] ]
; CHECK-NEXT: [[TMP0]] = add i32 [[J]], 1
; CHECK-NEXT: [[TMP1:%.*]] = icmp slt i32 [[TMP0]], 2
; CHECK-NEXT: br i1 [[TMP1]], label %[[FOR_J]], label %[[EXIT:.*]]
; CHECK: [[FOR_I_LATCH]]:
; CHECK-NEXT: [[I_INC]] = add i32 [[I]], 1
; CHECK-NEXT: [[CMP_I:%.*]] = icmp slt i32 [[I_INC]], 2
; CHECK-NEXT: br i1 [[CMP_I]], label %[[FOR_I_HEADER]], label %[[FOR_J_SPLIT]]
; CHECK: [[EXIT]]:
; CHECK-NEXT: ret void
;
entry:
br label %for.i.header
for.i.header:
%i = phi i32 [ 0, %entry ], [ %i.inc, %for.i.latch ]
%sum.i = phi i32 [ 0, %entry ], [ %sum.i.lcssa, %for.i.latch ]
br label %for.j
for.j:
%j = phi i32 [ 0, %for.i.header ], [ %j.inc, %for.j ]
%sum.j = phi i32 [ %sum.i, %for.i.header ], [ %sum.j.next, %for.j ]
%idx = getelementptr inbounds [2 x [2 x i32]], ptr %A, i32 0, i32 %j, i32 %i
%a = load i32, ptr %idx, align 4
%sum.j.next = add nuw nsw i32 %sum.j, %a
%j.inc = add i32 %j, 1
%cmp.j = icmp slt i32 %j.inc, 2
br i1 %cmp.j, label %for.j, label %for.i.latch
for.i.latch:
%sum.i.lcssa = phi i32 [ %sum.j.next, %for.j ]
%i.inc = add i32 %i, 1
%cmp.i = icmp slt i32 %i.inc, 2
br i1 %cmp.i, label %for.i.header, label %exit
exit:
ret void
}
; Check that nsw/nuw flags are dropped when interchanging loops.
;
; int prod = 1;
; for (int i = 0; i < 2; i++)
; for (int j = 0; j < 2; j++)
; prod *= A[j][i];
;
define void @reduction_mul(ptr %A) {
; CHECK-LABEL: define void @reduction_mul(
; CHECK-SAME: ptr [[A:%.*]]) {
; CHECK-NEXT: [[ENTRY:.*:]]
; CHECK-NEXT: br label %[[FOR_J_PREHEADER:.*]]
; CHECK: [[FOR_I_HEADER_PREHEADER:.*]]:
; CHECK-NEXT: br label %[[FOR_I_HEADER:.*]]
; CHECK: [[FOR_I_HEADER]]:
; CHECK-NEXT: [[I:%.*]] = phi i32 [ [[I_INC:%.*]], %[[FOR_I_LATCH:.*]] ], [ 0, %[[FOR_I_HEADER_PREHEADER]] ]
; CHECK-NEXT: [[PROD_J:%.*]] = phi i32 [ [[PROD_J_NEXT:%.*]], %[[FOR_I_LATCH]] ], [ [[PROD_I:%.*]], %[[FOR_I_HEADER_PREHEADER]] ]
; CHECK-NEXT: br label %[[FOR_J_SPLIT1:.*]]
; CHECK: [[FOR_J_PREHEADER]]:
; CHECK-NEXT: br label %[[FOR_J:.*]]
; CHECK: [[FOR_J]]:
; CHECK-NEXT: [[J:%.*]] = phi i32 [ [[TMP0:%.*]], %[[FOR_J_SPLIT:.*]] ], [ 0, %[[FOR_J_PREHEADER]] ]
; CHECK-NEXT: [[PROD_I]] = phi i32 [ [[PROD_I_LCSSA:%.*]], %[[FOR_J_SPLIT]] ], [ 1, %[[FOR_J_PREHEADER]] ]
; CHECK-NEXT: br label %[[FOR_I_HEADER_PREHEADER]]
; CHECK: [[FOR_J_SPLIT1]]:
; CHECK-NEXT: [[IDX:%.*]] = getelementptr inbounds [2 x [2 x i32]], ptr [[A]], i32 0, i32 [[J]], i32 [[I]]
; CHECK-NEXT: [[A:%.*]] = load i32, ptr [[IDX]], align 4
; CHECK-NEXT: [[PROD_J_NEXT]] = mul i32 [[PROD_J]], [[A]]
; CHECK-NEXT: [[J_INC:%.*]] = add i32 [[J]], 1
; CHECK-NEXT: [[CMP_J:%.*]] = icmp slt i32 [[J_INC]], 2
; CHECK-NEXT: br label %[[FOR_I_LATCH]]
; CHECK: [[FOR_J_SPLIT]]:
; CHECK-NEXT: [[PROD_I_LCSSA]] = phi i32 [ [[PROD_J_NEXT]], %[[FOR_I_LATCH]] ]
; CHECK-NEXT: [[TMP0]] = add i32 [[J]], 1
; CHECK-NEXT: [[TMP1:%.*]] = icmp slt i32 [[TMP0]], 2
; CHECK-NEXT: br i1 [[TMP1]], label %[[FOR_J]], label %[[EXIT:.*]]
; CHECK: [[FOR_I_LATCH]]:
; CHECK-NEXT: [[I_INC]] = add i32 [[I]], 1
; CHECK-NEXT: [[CMP_I:%.*]] = icmp slt i32 [[I_INC]], 2
; CHECK-NEXT: br i1 [[CMP_I]], label %[[FOR_I_HEADER]], label %[[FOR_J_SPLIT]]
; CHECK: [[EXIT]]:
; CHECK-NEXT: ret void
;
entry:
br label %for.i.header
for.i.header:
%i = phi i32 [ 0, %entry ], [ %i.inc, %for.i.latch ]
%prod.i = phi i32 [ 1, %entry ], [ %prod.i.lcssa, %for.i.latch ]
br label %for.j
for.j:
%j = phi i32 [ 0, %for.i.header ], [ %j.inc, %for.j ]
%prod.j = phi i32 [ %prod.i, %for.i.header ], [ %prod.j.next, %for.j ]
%idx = getelementptr inbounds [2 x [2 x i32]], ptr %A, i32 0, i32 %j, i32 %i
%a = load i32, ptr %idx, align 4
%prod.j.next = mul nsw nuw i32 %prod.j, %a
%j.inc = add i32 %j, 1
%cmp.j = icmp slt i32 %j.inc, 2
br i1 %cmp.j, label %for.j, label %for.i.latch
for.i.latch:
%prod.i.lcssa = phi i32 [ %prod.j.next, %for.j ]
%i.inc = add i32 %i, 1
%cmp.i = icmp slt i32 %i.inc, 2
br i1 %cmp.i, label %for.i.header, label %exit
exit:
ret void
}
|
