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
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
|
// RUN: mlir-translate -mlir-to-llvmir %s | FileCheck %s
// nonesense minimised code simulating the control flow graph generated by flang
// for array reductions. The important thing here is that we are testing a byref
// reduction with a cleanup region, and the various regions contain multiple
// blocks
omp.declare_reduction @add_reduction_byref_box_Uxf32 : !llvm.ptr init {
^bb0(%arg0: !llvm.ptr):
%0 = llvm.mlir.constant(1 : i64) : i64
%1 = llvm.alloca %0 x !llvm.struct<(ptr, i64, i32, i8, i8, i8, i8, array<1 x array<3 x i64>>)> : (i64) -> !llvm.ptr
omp.yield(%1 : !llvm.ptr)
} combiner {
^bb0(%arg0: !llvm.ptr, %arg1: !llvm.ptr):
%0 = llvm.mlir.constant(0 : i64) : i64
%1 = llvm.mlir.constant(0 : index) : i64
%2 = llvm.mlir.constant(1 : index) : i64
llvm.br ^bb1(%0 : i64)
^bb1(%3: i64): // 2 preds: ^bb0, ^bb2
%4 = llvm.icmp "sgt" %3, %1 : i64
llvm.cond_br %4, ^bb2, ^bb3
^bb2: // pred: ^bb1
%5 = llvm.sub %3, %2 : i64
llvm.br ^bb1(%5 : i64)
^bb3: // pred: ^bb1
omp.yield(%arg0 : !llvm.ptr)
} cleanup {
^bb0(%arg0: !llvm.ptr):
%0 = llvm.mlir.constant(0 : i64) : i64
%1 = llvm.ptrtoint %arg0 : !llvm.ptr to i64
%2 = llvm.icmp "ne" %1, %0 : i64
llvm.cond_br %2, ^bb1, ^bb2
^bb1: // pred: ^bb0
llvm.br ^bb2
^bb2: // 2 preds: ^bb0, ^bb1
omp.yield
}
llvm.func @sectionsreduction_(%arg0: !llvm.ptr {fir.bindc_name = "x"}) attributes {fir.internal_name = "_QPsectionsreduction"} {
%0 = llvm.mlir.constant(1 : i64) : i64
%1 = llvm.mlir.constant(0 : index) : i64
%2 = llvm.mlir.constant(1 : index) : i64
omp.parallel {
%3 = llvm.alloca %0 x !llvm.struct<(ptr, i64, i32, i8, i8, i8, i8, array<1 x array<3 x i64>>)> : (i64) -> !llvm.ptr
omp.sections reduction(byref @add_reduction_byref_box_Uxf32 -> %3 : !llvm.ptr) {
^bb0(%arg1: !llvm.ptr):
omp.section {
^bb0(%arg2: !llvm.ptr):
llvm.br ^bb1(%0 : i64)
^bb1(%4: i64): // 2 preds: ^bb0, ^bb2
%5 = llvm.icmp "sgt" %4, %1 : i64
llvm.cond_br %5, ^bb2, ^bb3
^bb2: // pred: ^bb1
%6 = llvm.sub %4, %2 : i64
llvm.br ^bb1(%6 : i64)
^bb3: // pred: ^bb1
omp.terminator
}
omp.section {
^bb0(%arg2: !llvm.ptr):
llvm.br ^bb1(%0 : i64)
^bb1(%4: i64): // 2 preds: ^bb0, ^bb2
%5 = llvm.icmp "sgt" %4, %1 : i64
llvm.cond_br %5, ^bb2, ^bb3
^bb2: // pred: ^bb1
%6 = llvm.sub %4, %2 : i64
llvm.br ^bb1(%6 : i64)
^bb3: // pred: ^bb1
omp.terminator
}
omp.terminator
}
omp.terminator
}
llvm.return
}
// CHECK-LABEL: define internal void @sectionsreduction_..omp_par
// CHECK: omp.par.entry:
// CHECK: %[[VAL_6:.*]] = alloca i32, align 4
// CHECK: %[[VAL_7:.*]] = alloca i32, align 4
// CHECK: %[[VAL_8:.*]] = alloca i32, align 4
// CHECK: %[[VAL_9:.*]] = alloca i32, align 4
// CHECK: %[[VAL_10:.*]] = alloca i32, align 4
// CHECK: %[[VAL_11:.*]] = load i32, ptr %[[VAL_12:.*]], align 4
// CHECK: store i32 %[[VAL_11]], ptr %[[VAL_10]], align 4
// CHECK: %[[VAL_13:.*]] = load i32, ptr %[[VAL_10]], align 4
// CHECK: %[[VAL_14:.*]] = alloca [1 x ptr], align 8
// CHECK: br label %[[VAL_15:.*]]
// CHECK: omp.reduction.init: ; preds = %[[VAL_16:.*]]
// CHECK: br label %[[VAL_17:.*]]
// CHECK: omp.par.region: ; preds = %[[VAL_15]]
// CHECK: br label %[[VAL_18:.*]]
// CHECK: omp.par.region1: ; preds = %[[VAL_17]]
// CHECK: %[[VAL_19:.*]] = alloca { ptr, i64, i32, i8, i8, i8, i8, [1 x [3 x i64]] }, i64 1, align 8
// CHECK: %[[VAL_20:.*]] = alloca { ptr, i64, i32, i8, i8, i8, i8, [1 x [3 x i64]] }, i64 1, align 8
// CHECK: %[[VAL_21:.*]] = alloca ptr, align 8
// CHECK: store ptr %[[VAL_20]], ptr %[[VAL_21]], align 8
// CHECK: br label %[[VAL_22:.*]]
// CHECK: omp_section_loop.preheader: ; preds = %[[VAL_18]]
// CHECK: store i32 0, ptr %[[VAL_7]], align 4
// CHECK: store i32 1, ptr %[[VAL_8]], align 4
// CHECK: store i32 1, ptr %[[VAL_9]], align 4
// CHECK: %[[VAL_23:.*]] = call i32 @__kmpc_global_thread_num(ptr @1)
// CHECK: call void @__kmpc_for_static_init_4u(ptr @1, i32 %[[VAL_23]], i32 34, ptr %[[VAL_6]], ptr %[[VAL_7]], ptr %[[VAL_8]], ptr %[[VAL_9]], i32 1, i32 0)
// CHECK: %[[VAL_24:.*]] = load i32, ptr %[[VAL_7]], align 4
// CHECK: %[[VAL_25:.*]] = load i32, ptr %[[VAL_8]], align 4
// CHECK: %[[VAL_26:.*]] = sub i32 %[[VAL_25]], %[[VAL_24]]
// CHECK: %[[VAL_27:.*]] = add i32 %[[VAL_26]], 1
// CHECK: br label %[[VAL_28:.*]]
// CHECK: omp_section_loop.header: ; preds = %[[VAL_29:.*]], %[[VAL_22]]
// CHECK: %[[VAL_30:.*]] = phi i32 [ 0, %[[VAL_22]] ], [ %[[VAL_31:.*]], %[[VAL_29]] ]
// CHECK: br label %[[VAL_32:.*]]
// CHECK: omp_section_loop.cond: ; preds = %[[VAL_28]]
// CHECK: %[[VAL_33:.*]] = icmp ult i32 %[[VAL_30]], %[[VAL_27]]
// CHECK: br i1 %[[VAL_33]], label %[[VAL_34:.*]], label %[[VAL_35:.*]]
// CHECK: omp_section_loop.exit: ; preds = %[[VAL_32]]
// CHECK: call void @__kmpc_for_static_fini(ptr @1, i32 %[[VAL_23]])
// CHECK: %[[VAL_36:.*]] = call i32 @__kmpc_global_thread_num(ptr @1)
// CHECK: call void @__kmpc_barrier(ptr @2, i32 %[[VAL_36]])
// CHECK: br label %[[VAL_37:.*]]
// CHECK: omp_section_loop.after: ; preds = %[[VAL_35]]
// CHECK: br label %[[VAL_38:.*]]
// CHECK: omp_section_loop.aftersections.fini: ; preds = %[[VAL_37]]
// CHECK: %[[VAL_39:.*]] = getelementptr inbounds [1 x ptr], ptr %[[VAL_14]], i64 0, i64 0
// CHECK: store ptr %[[VAL_21]], ptr %[[VAL_39]], align 8
// CHECK: %[[VAL_40:.*]] = call i32 @__kmpc_global_thread_num(ptr @1)
// CHECK: %[[VAL_41:.*]] = call i32 @__kmpc_reduce(ptr @1, i32 %[[VAL_40]], i32 1, i64 8, ptr %[[VAL_14]], ptr @.omp.reduction.func, ptr @.gomp_critical_user_.reduction.var)
// CHECK: switch i32 %[[VAL_41]], label %[[VAL_42:.*]] [
// CHECK: i32 1, label %[[VAL_43:.*]]
// CHECK: i32 2, label %[[VAL_44:.*]]
// CHECK: ]
// CHECK: reduce.switch.atomic: ; preds = %[[VAL_38]]
// CHECK: unreachable
// CHECK: reduce.switch.nonatomic: ; preds = %[[VAL_38]]
// CHECK: %[[VAL_45:.*]] = load ptr, ptr %[[VAL_21]], align 8
// CHECK: br label %[[VAL_46:.*]]
// CHECK: omp.reduction.nonatomic.body: ; preds = %[[VAL_43]]
// CHECK: br label %[[VAL_47:.*]]
// CHECK: omp.reduction.nonatomic.body16: ; preds = %[[VAL_48:.*]], %[[VAL_46]]
// CHECK: %[[VAL_49:.*]] = phi i64 [ %[[VAL_50:.*]], %[[VAL_48]] ], [ 0, %[[VAL_46]] ]
// CHECK: %[[VAL_51:.*]] = icmp sgt i64 %[[VAL_49]], 0
// CHECK: br i1 %[[VAL_51]], label %[[VAL_48]], label %[[VAL_52:.*]]
// CHECK: omp.reduction.nonatomic.body18: ; preds = %[[VAL_47]]
// CHECK: br label %[[VAL_53:.*]]
// CHECK: omp.region.cont15: ; preds = %[[VAL_52]]
// CHECK: %[[VAL_54:.*]] = phi ptr [ %[[VAL_19]], %[[VAL_52]] ]
// CHECK: call void @__kmpc_end_reduce(ptr @1, i32 %[[VAL_40]], ptr @.gomp_critical_user_.reduction.var)
// CHECK: br label %[[VAL_42]]
// CHECK: omp.reduction.nonatomic.body17: ; preds = %[[VAL_47]]
// CHECK: %[[VAL_50]] = sub i64 %[[VAL_49]], 1
// CHECK: br label %[[VAL_47]]
// CHECK: reduce.finalize: ; preds = %[[VAL_53]], %[[VAL_38]]
// CHECK: %[[VAL_55:.*]] = call i32 @__kmpc_global_thread_num(ptr @1)
// CHECK: call void @__kmpc_barrier(ptr @2, i32 %[[VAL_55]])
// CHECK: %[[VAL_56:.*]] = load ptr, ptr %[[VAL_21]], align 8
// CHECK: br label %[[VAL_57:.*]]
// CHECK: omp.reduction.cleanup: ; preds = %[[VAL_42]]
// CHECK: %[[VAL_58:.*]] = ptrtoint ptr %[[VAL_56]] to i64
// CHECK: %[[VAL_59:.*]] = icmp ne i64 %[[VAL_58]], 0
// CHECK: br i1 %[[VAL_59]], label %[[VAL_60:.*]], label %[[VAL_61:.*]]
// CHECK: omp.reduction.cleanup22: ; preds = %[[VAL_60]], %[[VAL_57]]
// CHECK: br label %[[VAL_62:.*]]
// CHECK: omp.region.cont20: ; preds = %[[VAL_61]]
// CHECK: br label %[[VAL_63:.*]]
// CHECK: omp.region.cont: ; preds = %[[VAL_62]]
// CHECK: br label %[[VAL_64:.*]]
// CHECK: omp.par.pre_finalize: ; preds = %[[VAL_63]]
// CHECK: br label %[[VAL_65:.*]]
// CHECK: omp.reduction.cleanup21: ; preds = %[[VAL_57]]
// CHECK: br label %[[VAL_61]]
// CHECK: omp_section_loop.body: ; preds = %[[VAL_32]]
// CHECK: %[[VAL_66:.*]] = add i32 %[[VAL_30]], %[[VAL_24]]
// CHECK: %[[VAL_67:.*]] = mul i32 %[[VAL_66]], 1
// CHECK: %[[VAL_68:.*]] = add i32 %[[VAL_67]], 0
// CHECK: switch i32 %[[VAL_68]], label %[[VAL_69:.*]] [
// CHECK: i32 0, label %[[VAL_70:.*]]
// CHECK: i32 1, label %[[VAL_71:.*]]
// CHECK: ]
// CHECK: omp_section_loop.body.case6: ; preds = %[[VAL_34]]
// CHECK: br label %[[VAL_72:.*]]
// CHECK: omp.section.region8: ; preds = %[[VAL_71]]
// CHECK: br label %[[VAL_73:.*]]
// CHECK: omp.section.region9: ; preds = %[[VAL_74:.*]], %[[VAL_72]]
// CHECK: %[[VAL_75:.*]] = phi i64 [ %[[VAL_76:.*]], %[[VAL_74]] ], [ 1, %[[VAL_72]] ]
// CHECK: %[[VAL_77:.*]] = icmp sgt i64 %[[VAL_75]], 0
// CHECK: br i1 %[[VAL_77]], label %[[VAL_74]], label %[[VAL_78:.*]]
// CHECK: omp.section.region11: ; preds = %[[VAL_73]]
// CHECK: br label %[[VAL_79:.*]]
// CHECK: omp.region.cont7: ; preds = %[[VAL_78]]
// CHECK: br label %[[VAL_69]]
// CHECK: omp.section.region10: ; preds = %[[VAL_73]]
// CHECK: %[[VAL_76]] = sub i64 %[[VAL_75]], 1
// CHECK: br label %[[VAL_73]]
// CHECK: omp_section_loop.body.case: ; preds = %[[VAL_34]]
// CHECK: br label %[[VAL_80:.*]]
// CHECK: omp.section.region: ; preds = %[[VAL_70]]
// CHECK: br label %[[VAL_81:.*]]
// CHECK: omp.section.region3: ; preds = %[[VAL_82:.*]], %[[VAL_80]]
// CHECK: %[[VAL_83:.*]] = phi i64 [ %[[VAL_84:.*]], %[[VAL_82]] ], [ 1, %[[VAL_80]] ]
// CHECK: %[[VAL_85:.*]] = icmp sgt i64 %[[VAL_83]], 0
// CHECK: br i1 %[[VAL_85]], label %[[VAL_82]], label %[[VAL_86:.*]]
// CHECK: omp.section.region5: ; preds = %[[VAL_81]]
// CHECK: br label %[[VAL_87:.*]]
// CHECK: omp.region.cont2: ; preds = %[[VAL_86]]
// CHECK: br label %[[VAL_69]]
// CHECK: omp.section.region4: ; preds = %[[VAL_81]]
// CHECK: %[[VAL_84]] = sub i64 %[[VAL_83]], 1
// CHECK: br label %[[VAL_81]]
// CHECK: omp_section_loop.body.sections.after: ; preds = %[[VAL_79]], %[[VAL_87]], %[[VAL_34]]
// CHECK: br label %[[VAL_29]]
// CHECK: omp_section_loop.inc: ; preds = %[[VAL_69]]
// CHECK: %[[VAL_31]] = add nuw i32 %[[VAL_30]], 1
// CHECK: br label %[[VAL_28]]
// CHECK: omp.par.outlined.exit.exitStub: ; preds = %[[VAL_64]]
// CHECK: ret void
|
