1 // RUN: %clang_cc1 -DCHECK -verify -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - -Wno-openmp-mapping | FileCheck %s --check-prefix CHECK --check-prefix CHECK-64
2 // RUN: %clang_cc1 -DCHECK -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s -Wno-openmp-mapping
3 // RUN: %clang_cc1 -DCHECK -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - -Wno-openmp-mapping | FileCheck %s --check-prefix CHECK --check-prefix CHECK-64
4 // RUN: %clang_cc1 -DCHECK -verify -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - -Wno-openmp-mapping | FileCheck %s --check-prefix CHECK --check-prefix CHECK-32
5 // RUN: %clang_cc1 -DCHECK -fopenmp -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s -Wno-openmp-mapping
6 // RUN: %clang_cc1 -DCHECK -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - -Wno-openmp-mapping | FileCheck %s --check-prefix CHECK --check-prefix CHECK-32
7
8 // RUN: %clang_cc1 -DCHECK -verify -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - -Wno-openmp-mapping | FileCheck --check-prefix SIMD-ONLY0 %s
9 // RUN: %clang_cc1 -DCHECK -fopenmp-simd -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s -Wno-openmp-mapping
10 // RUN: %clang_cc1 -DCHECK -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - -Wno-openmp-mapping | FileCheck --check-prefix SIMD-ONLY0 %s
11 // RUN: %clang_cc1 -DCHECK -verify -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - -Wno-openmp-mapping | FileCheck --check-prefix SIMD-ONLY0 %s
12 // RUN: %clang_cc1 -DCHECK -fopenmp-simd -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s -Wno-openmp-mapping
13 // RUN: %clang_cc1 -DCHECK -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - -Wno-openmp-mapping | FileCheck --check-prefix SIMD-ONLY0 %s
14 // SIMD-ONLY0-NOT: {{__kmpc|__tgt}}
15
16 // RUN: %clang_cc1 -DLAMBDA -verify -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - -Wno-openmp-mapping | FileCheck %s --check-prefix LAMBDA --check-prefix LAMBDA-64
17 // RUN: %clang_cc1 -DLAMBDA -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s -Wno-openmp-mapping
18 // RUN: %clang_cc1 -DLAMBDA -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - -Wno-openmp-mapping | FileCheck %s --check-prefix LAMBDA --check-prefix LAMBDA-64
19
20 // RUN: %clang_cc1 -DLAMBDA -verify -fopenmp-simd -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - -Wno-openmp-mapping | FileCheck --check-prefix SIMD-ONLY1 %s
21 // RUN: %clang_cc1 -DLAMBDA -fopenmp-simd -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s -Wno-openmp-mapping
22 // RUN: %clang_cc1 -DLAMBDA -fopenmp-simd -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - -Wno-openmp-mapping | FileCheck --check-prefix SIMD-ONLY1 %s
23 // SIMD-ONLY1-NOT: {{__kmpc|__tgt}}
24
25 // expected-no-diagnostics
26 #ifndef HEADER
27 #define HEADER
28
29 struct St {
30 int a, b;
StSt31 St() : a(0), b(0) {}
StSt32 St(const St &st) : a(st.a + st.b), b(0) {}
~StSt33 ~St() {}
34 };
35
36 volatile int g = 1212;
37 volatile int &g1 = g;
38
39 template <class T>
40 struct S {
41 T f;
SS42 S(T a) : f(a + g) {}
SS43 S() : f(g) {}
SS44 S(const S &s, St t = St()) : f(s.f + t.a) {}
operator TS45 operator T() { return T(); }
~SS46 ~S() {}
47 };
48
49 // CHECK-DAG: [[S_FLOAT_TY:%.+]] = type { float }
50 // CHECK-DAG: [[S_INT_TY:%.+]] = type { i{{[0-9]+}} }
51 // CHECK-DAG: [[ST_TY:%.+]] = type { i{{[0-9]+}}, i{{[0-9]+}} }
52
53 template <typename T>
tmain()54 T tmain() {
55 S<T> test;
56 T t_var = T();
57 T vec[] = {1, 2};
58 S<T> s_arr[] = {1, 2};
59 S<T> &var = test;
60 #pragma omp target
61 #pragma omp teams distribute parallel for simd firstprivate(t_var, vec, s_arr, var)
62 for (int i = 0; i < 2; ++i) {
63 vec[i] = t_var;
64 s_arr[i] = var;
65 }
66 return T();
67 }
68
69 // CHECK-DAG: [[TEST:@.+]] ={{.*}} global [[S_FLOAT_TY]] zeroinitializer,
70 S<float> test;
71 // CHECK-DAG: [[T_VAR:@.+]] ={{.*}} global i{{[0-9]+}} 333,
72 int t_var = 333;
73 // CHECK-DAG: [[VEC:@.+]] ={{.*}} global [2 x i{{[0-9]+}}] [i{{[0-9]+}} 1, i{{[0-9]+}} 2],
74 int vec[] = {1, 2};
75 // CHECK-DAG: [[S_ARR:@.+]] ={{.*}} global [2 x [[S_FLOAT_TY]]] zeroinitializer,
76 S<float> s_arr[] = {1, 2};
77 // CHECK-DAG: [[VAR:@.+]] ={{.*}} global [[S_FLOAT_TY]] zeroinitializer,
78 S<float> var(3);
79 // CHECK-DAG: [[SIVAR:@.+]] = internal global i{{[0-9]+}} 0,
80
main()81 int main() {
82 static int sivar;
83 #ifdef LAMBDA
84 // LAMBDA: [[G:@.+]] ={{.*}} global i{{[0-9]+}} 1212,
85 // LAMBDA-LABEL: @main
86 // LAMBDA: call void [[OUTER_LAMBDA:@.+]](
87 [&]() {
88 // LAMBDA: define{{.*}} internal{{.*}} void [[OUTER_LAMBDA]](
89 // LAMBDA: call i32 @__tgt_target_teams_mapper(%struct.ident_t* @{{.+}}, i64 -1, i8* @{{[^,]+}}, i32 3, i8** %{{[^,]+}}, i8** %{{[^,]+}}, i{{64|32}}* {{.+}}@{{[^,]+}}, i32 0, i32 0), i64* {{.+}}@{{[^,]+}}, i32 0, i32 0), i8** null
90 // LAMBDA: call void @[[LOFFL1:.+]](i{{64|32}} %{{.+}})
91 // LAMBDA: ret
92 #pragma omp target
93 #pragma omp teams distribute parallel for simd firstprivate(g, g1, sivar)
94 for (int i = 0; i < 2; ++i) {
95 // LAMBDA: define{{.*}} internal{{.*}} void @[[LOFFL1]](i{{64|32}} {{%.+}}, i{{64|32}} {{%.+}})
96 // LAMBDA: {{%.+}} = alloca i{{[0-9]+}},
97 // LAMBDA: {{%.+}} = alloca i{{[0-9]+}},
98 // LAMBDA: {{%.+}} = alloca i{{[0-9]+}},
99 // LAMBDA: [[G_CAST:%.+]] = alloca i{{[0-9]+}},
100 // LAMBDA: [[G1_CAST:%.+]] = alloca i{{[0-9]+}},
101 // LAMBDA: [[SIVAR_CAST:%.+]] = alloca i{{[0-9]+}},
102 // LAMBDA-DAG: [[G_CAST_VAL:%.+]] = load{{.+}} [[G_CAST]],
103 // LAMBDA-DAG: [[G1_CAST_VAL:%.+]] = load{{.+}} [[G1_CAST]],
104 // LAMBDA-DAG: [[SIVAR_CAST_VAL:%.+]] = load{{.+}} [[SIVAR_CAST]],
105 // LAMBDA: call void {{.+}} @__kmpc_fork_teams({{.+}}, i32 3, {{.+}} @[[LOUTL1:.+]] to {{.+}}, {{.+}} [[G_CAST_VAL]], {{.+}} [[G1_CAST_VAL]], {{.+}} [[SIVAR_CAST_VAL]])
106 // LAMBDA: ret void
107
108 // LAMBDA: define internal void @[[LOUTL1]]({{.+}})
109 // Skip global and bound tid vars
110 // LAMBDA: {{.+}} = alloca i32*,
111 // LAMBDA: {{.+}} = alloca i32*,
112 // LAMBDA: [[G_ADDR:%.+]] = alloca i{{[0-9]+}},
113 // LAMBDA: [[G1_ADDR:%.+]] = alloca i{{[0-9]+}},
114 // LAMBDA: [[SIVAR_ADDR:%.+]] = alloca i{{[0-9]+}},
115 // LAMBDA: [[G1_TMP:%.+]] = alloca i32*,
116 // skip loop vars
117 // LAMBDA-DAG: store {{.+}}, {{.+}} [[G_ADDR]],
118 // LAMBDA-DAG: store {{.+}}, {{.+}} [[G1_ADDR]],
119 // LAMBDA-DAG: store {{.+}}, {{.+}} [[SIVAR_ADDR]],
120 // LAMBDA-DAG: [[G_CONV:%.+]] = bitcast {{.+}} [[G_ADDR]] to
121 // LAMBDA-DAG: [[G1_CONV:%.+]] = bitcast {{.+}} [[G1_ADDR]] to
122 // LAMBDA-DAG: [[SIVAR_CONV:%.+]] = bitcast {{.+}} [[SIVAR_ADDR]] to
123 // LAMBDA-DAG: store{{.+}} [[G1_CONV]], {{.+}} [[G1_TMP]],
124 g = 1;
125 g1 = 1;
126 sivar = 2;
127 // LAMBDA: call void @__kmpc_for_static_init_4(
128 // LAMBDA: call void {{.*}} @__kmpc_fork_call({{.+}}, {{.+}}, {{.+}} @[[LPAR_OUTL:.+]] to
129 // LAMBDA: call void @__kmpc_for_static_fini(
130 // LAMBDA: ret void
131
132 // LAMBDA: define internal void @[[LPAR_OUTL]]({{.+}})
133 // Skip global and bound tid vars, and prev lb and ub vars
134 // LAMBDA: {{.+}} = alloca i32*,
135 // LAMBDA: {{.+}} = alloca i32*,
136 // LAMBDA: {{.+}} = alloca i{{[0-9]+}},
137 // LAMBDA: {{.+}} = alloca i{{[0-9]+}},
138 // LAMBDA: [[G_ADDR:%.+]] = alloca i{{[0-9]+}},
139 // LAMBDA: [[G1_ADDR:%.+]] = alloca i{{[0-9]+}},
140 // LAMBDA: [[SIVAR_ADDR:%.+]] = alloca i{{[0-9]+}},
141 // LAMBDA: [[G1_TMP:%.+]] = alloca i32*,
142 // skip loop vars
143 // LAMBDA-DAG: store {{.+}}, {{.+}} [[G_ADDR]],
144 // LAMBDA-DAG: store {{.+}}, {{.+}} [[G1_ADDR]],
145 // LAMBDA-DAG: store {{.+}}, {{.+}} [[SIVAR_ADDR]],
146 // LAMBDA-DAG: [[G_CONV:%.+]] = bitcast {{.+}} [[G_ADDR]] to
147 // LAMBDA-DAG: [[G1_CONV:%.+]] = bitcast {{.+}} [[G1_ADDR]] to
148 // LAMBDA-DAG: [[SIVAR_CONV:%.+]] = bitcast {{.+}} [[SIVAR_ADDR]] to
149 // LAMBDA-DAG: store{{.+}} [[G1_CONV]], {{.+}} [[G1_TMP]],
150
151 // use of private vars
152 // LAMBDA-DAG: store{{.+}} 1, {{.+}} [[G_CONV]],
153 // LAMBDA-DAG: [[G1:%.+]] = load{{.+}}, {{.+}}* [[G1_TMP]]
154 // LAMBDA-DAG: store{{.+}} 1, {{.+}} [[G1]],
155 // LAMBDA-DAG: store{{.+}} 2, {{.+}} [[SIVAR_CONV]],
156 // LAMBDA-DAG: [[G1_REF:%.+]] = load{{.+}}, {{.+}} [[G1_TMP]],
157 // LAMBDA: call void [[INNER_LAMBDA:@.+]](
158 // LAMBDA: call void @__kmpc_for_static_fini(
159 // LAMBDA: ret void
160 [&]() {
161 // LAMBDA: define {{.+}} void [[INNER_LAMBDA]]({{.+}} [[ARG_PTR:%.+]])
162 // LAMBDA: store %{{.+}}* [[ARG_PTR]], %{{.+}}** [[ARG_PTR_REF:%.+]],
163 g = 2;
164 g1 = 2;
165 sivar = 4;
166 // LAMBDA: [[ARG_PTR:%.+]] = load %{{.+}}*, %{{.+}}** [[ARG_PTR_REF]]
167
168 // LAMBDA: [[G_PTR_REF:%.+]] = getelementptr inbounds %{{.+}}, %{{.+}}* [[ARG_PTR]], i{{[0-9]+}} 0, i{{[0-9]+}} 0
169 // LAMBDA: [[G_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[G_PTR_REF]]
170 // LAMBDA: store i{{[0-9]+}} 2, i{{[0-9]+}}* [[G_REF]]
171 // LAMBDA: [[G1_PTR_REF:%.+]] = getelementptr inbounds %{{.+}}, %{{.+}}* [[ARG_PTR]], i{{[0-9]+}} 0, i{{[0-9]+}} 1
172 // LAMBDA: [[G1_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[G1_PTR_REF]]
173 // LAMBDA: store i{{[0-9]+}} 2, i{{[0-9]+}}* [[G1_REF]]
174 // LAMBDA: [[SIVAR_PTR_REF:%.+]] = getelementptr inbounds %{{.+}}, %{{.+}}* [[ARG_PTR]], i{{[0-9]+}} 0, i{{[0-9]+}} 2
175 // LAMBDA: [[SIVAR_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[SIVAR_PTR_REF]]
176 // LAMBDA: store i{{[0-9]+}} 4, i{{[0-9]+}}* [[SIVAR_REF]]
177 }();
178 }
179 }();
180 return 0;
181
182 // LAMBDA: !{!"llvm.loop.vectorize.enable", i1 true}
183
184 #else
185 #pragma omp target
186 #pragma omp teams distribute parallel for simd firstprivate(t_var, vec, s_arr, var, sivar)
187 for (int i = 0; i < 2; ++i) {
188 vec[i] = t_var;
189 s_arr[i] = var;
190 sivar += i;
191 }
192 return tmain<int>();
193 #endif
194 }
195
196 // CHECK: define {{.*}}i{{[0-9]+}} @main()
197 // CHECK: call i32 @__tgt_target_teams_mapper(%struct.ident_t* @{{.+}}, i64 -1, i8* @{{[^,]+}}, i32 5, i8** %{{[^,]+}}, i8** %{{[^,]+}}, i{{64|32}}* {{.+}}@{{[^,]+}}, i32 0, i32 0), i64* {{.+}}@{{[^,]+}}, i32 0, i32 0), i8** null
198 // CHECK: call void @[[OFFL1:.+]](i{{64|32}} %{{.+}})
199 // CHECK: {{%.+}} = call{{.*}} i32 @[[TMAIN_INT:.+]]()
200 // CHECK: ret
201
202 // CHECK: define{{.*}} void @[[OFFL1]]({{.+}})
203 // CHECK: [[T_VAR_PRIV:%.+]] = alloca i{{[0-9]+}},
204 // CHECK: [[VEC_PRIV:%.+]] = alloca [2 x i{{[0-9]+}}]*,
205 // CHECK: [[S_ARR_PRIV:%.+]] = alloca [2 x [[S_FLOAT_TY]]]*,
206 // CHECK: [[VAR_PRIV:%.+]] = alloca [[S_FLOAT_TY]]*,
207 // CHECK: [[SIVAR_PRIV:%.+]] = alloca i{{[0-9]+}},
208 // CHECK: [[T_VAR_CAST:%.+]] = alloca i{{[0-9]+}},
209 // CHECK: [[SIVAR_CAST:%.+]] = alloca i{{[0-9]+}},
210
211 // CHECK-DAG: [[VEC_TE_PAR:%.+]] = load [2 x i{{[0-9]+}}]*, [2 x i{{[0-9]+}}]** [[VEC_PRIV]],
212 // CHECK-DAG: [[T_VAR_TE_PAR:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR_CAST]],
213 // CHECK-DAG: [[S_ARR_TE_PAR:%.+]] = load [2 x [[S_FLOAT_TY]]]*, [2 x [[S_FLOAT_TY]]]** [[S_ARR_PRIV]],
214 // CHECK-DAG: [[VAR_TE_PAR:%.+]] = load [[S_FLOAT_TY]]*, [[S_FLOAT_TY]]** [[VAR_PRIV]],
215 // CHECK-DAG: [[SIVAR_TE_PAR:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[SIVAR_CAST]],
216
217 // CHECK: call void {{.+}} @__kmpc_fork_teams({{.+}}, i32 5, {{.+}} @[[OUTL1:.+]] to {{.+}}, [2 x i{{[0-9]+}}]* [[VEC_TE_PAR]], i{{[0-9]+}} [[T_VAR_TE_PAR]], [2 x [[S_FLOAT_TY]]]* [[S_ARR_TE_PAR]], [[S_FLOAT_TY]]* [[VAR_TE_PAR]], i{{[0-9]+}} [[SIVAR_TE_PAR]])
218 // CHECK: ret void
219
220 // CHECK: define internal void @[[OUTL1]]({{.+}})
221 // Skip global and bound tid vars
222 // CHECK: {{.+}} = alloca i32*,
223 // CHECK: {{.+}} = alloca i32*,
224 // CHECK: [[VEC_ADDR:%.+]] = alloca [2 x i{{[0-9]+}}]*,
225 // CHECK: [[T_VAR_ADDR:%.+]] = alloca i{{[0-9]+}},
226 // CHECK: [[S_ARR_ADDR:%.+]] = alloca [2 x [[S_FLOAT_TY]]]*,
227 // CHECK: [[VAR_ADDR:%.+]] = alloca [[S_FLOAT_TY]]*,
228 // CHECK: [[SIVAR_ADDR:%.+]] = alloca i{{[0-9]+}},
229 // Skip temp vars for loop
230 // CHECK: alloca i{{[0-9]+}},
231 // CHECK: alloca i{{[0-9]+}},
232 // CHECK: alloca i{{[0-9]+}},
233 // CHECK: alloca i{{[0-9]+}},
234 // CHECK: alloca i{{[0-9]+}},
235 // CHECK: [[VEC_PRIV:%.+]] = alloca [2 x i{{[0-9]+}}],
236 // CHECK: [[S_ARR_PRIV:%.+]] = alloca [2 x [[S_FLOAT_TY]]],
237 // CHECK: [[AGG_TMP1:%.+]] = alloca [[ST_TY]],
238 // CHECK: [[VAR_PRIV:%.+]] = alloca [[S_FLOAT_TY]],
239 // CHECK: [[AGG_TMP2:%.+]] = alloca [[ST_TY]],
240
241 // param copy
242 // CHECK: store [2 x i{{[0-9]+}}]* {{.+}}, [2 x i{{[0-9]+}}]** [[VEC_ADDR]],
243 // CHECK: store i{{[0-9]+}} {{.+}}, i{{[0-9]+}}* [[T_VAR_ADDR]],
244 // CHECK: store [2 x [[S_FLOAT_TY]]]* {{.+}}, [2 x [[S_FLOAT_TY]]]** [[S_ARR_ADDR]],
245 // CHECK: store [[S_FLOAT_TY]]* {{.+}}, [[S_FLOAT_TY]]** [[VAR_ADDR]],
246 // CHECK: store i{{[0-9]+}} {{.+}}, i{{[0-9]+}}* [[SIVAR_ADDR]],
247
248 // T_VAR and SIVAR
249 // CHECK-64-DAG: [[CONV_TVAR:%.+]] = bitcast i64* [[T_VAR_ADDR]] to i32*
250 // CHECK-64-DAG: [[CONV_SIVAR:%.+]] = bitcast i64* [[SIVAR_ADDR]] to i32*
251
252 // preparation vars
253 // CHECK-DAG: [[VEC_ADDR_VAL:%.+]] = load [2 x i{{[0-9]+}}]*, [2 x i{{[0-9]+}}]** [[VEC_ADDR]],
254 // CHECK-DAG: [[S_ARR_ADDR_REF:%.+]] = load [2 x [[S_FLOAT_TY]]]*, [2 x [[S_FLOAT_TY]]]** [[S_ARR_ADDR]],
255 // CHECK-DAG: [[VAR_ADDR_REF:%.+]] = load{{.+}} [[VAR_ADDR]],
256
257 // firstprivate vec(vec): copy from *_addr into priv1 and then from priv1 into priv2
258 // CHECK-DAG: [[VEC_DEST_PRIV:%.+]] = bitcast [2 x i{{[0-9]+}}]* [[VEC_PRIV]] to i8*
259 // CHECK-DAG: [[VEC_SRC:%.+]] = bitcast [2 x i{{[0-9]+}}]* [[VEC_ADDR_VAL]] to i8*
260 // CHECK: call void @llvm.memcpy.{{.+}}(i8* align {{[0-9]+}} [[VEC_DEST_PRIV]], i8* align {{[0-9]+}} [[VEC_SRC]], {{.+}})
261
262 // firstprivate(s_arr)
263 // CHECK-DAG: [[S_ARR_PRIV_BGN:%.+]] = getelementptr{{.*}} [2 x [[S_FLOAT_TY]]], [2 x [[S_FLOAT_TY]]]* [[S_ARR_PRIV]],
264 // CHECK-DAG: [[S_ARR_ADDR_BGN:%.+]] = bitcast [2 x [[S_FLOAT_TY]]]* [[S_ARR_ADDR_REF]] to
265 // CHECK-DAG: [[S_ARR_FIN:%.+]] = icmp{{.+}} [[S_ARR_PRIV_BGN]],
266 // CHECK-DAG: [[S_ARR_SRC_COPY:%.+]] = phi{{.+}} [ [[S_ARR_ADDR_BGN]], {{.+}} ], [ [[S_ARR_SRC:%.+]], {{.+}} ]
267 // CHECK-DAG: [[S_ARR_DST_COPY:%.+]] = phi{{.+}} [ [[S_ARR_PRIV_BGN]], {{.+}}], [ [[S_ARR_DST:%.+]], {{.+}} ]
268 // CHECK-DAG: call void @{{.+}}({{.+}} [[AGG_TMP1]])
269 // CHECK-DAG: call void @{{.+}}({{.+}} [[S_ARR_DST_COPY]], {{.+}} [[S_ARR_SRC_COPY]], {{.+}} [[AGG_TMP1]])
270 // CHECK-DAG: call void @{{.+}}({{.+}} [[AGG_TMP1]])
271 // CHECK-DAG: [[S_ARR_DST]] = getelementptr {{.+}} [[S_ARR_DST_COPY]],
272 // CHECK-DAG: [[S_ARR_SRC]] = getelementptr {{.+}} [[S_ARR_SRC_COPY]],
273
274 // firstprivate(var)
275 // CHECK-DAG: call void @{{.+}}({{.+}} [[AGG_TMP2]])
276 // CHECK-DAG: call void @{{.+}}({{.+}} [[VAR_PRIV]], {{.+}} [[VAR_ADDR_REF]], {{.+}} [[AGG_TMP2]])
277 // CHECK-DAG: call void @{{.+}}({{.+}} [[AGG_TMP2]])
278
279 // CHECK: call void @__kmpc_for_static_init_4(
280 // CHECK: call void {{.*}} @__kmpc_fork_call({{.+}}, {{.+}}, {{.+}} @[[PAR_OUTL:.+]] to
281 // CHECK: call void @__kmpc_for_static_fini(
282 // CHECK: ret void
283
284 // CHECK: define internal void @[[PAR_OUTL]]({{.+}})
285 // Skip global and bound tid vars, and prev lb ub vars
286 // CHECK: {{.+}} = alloca i32*,
287 // CHECK: {{.+}} = alloca i32*,
288 // CHECK: {{.+}} = alloca i{{[0-9]+}},
289 // CHECK: {{.+}} = alloca i{{[0-9]+}},
290 // CHECK: [[VEC_ADDR:%.+]] = alloca [2 x i{{[0-9]+}}]*,
291 // CHECK: [[T_VAR_ADDR:%.+]] = alloca i{{[0-9]+}},
292 // CHECK: [[S_ARR_ADDR:%.+]] = alloca [2 x [[S_FLOAT_TY]]]*,
293 // CHECK: [[VAR_ADDR:%.+]] = alloca [[S_FLOAT_TY]]*,
294 // CHECK: [[SIVAR_ADDR:%.+]] = alloca i{{[0-9]+}},
295 // Skip temp vars for loop
296 // CHECK: alloca i{{[0-9]+}},
297 // CHECK: alloca i{{[0-9]+}},
298 // CHECK: alloca i{{[0-9]+}},
299 // CHECK: alloca i{{[0-9]+}},
300 // CHECK: alloca i{{[0-9]+}},
301 // CHECK: [[VEC_PRIV:%.+]] = alloca [2 x i{{[0-9]+}}],
302 // CHECK: [[S_ARR_PRIV:%.+]] = alloca [2 x [[S_FLOAT_TY]]],
303 // CHECK: [[AGG_TMP1:%.+]] = alloca [[ST_TY]],
304 // CHECK: [[VAR_PRIV:%.+]] = alloca [[S_FLOAT_TY]],
305 // CHECK: [[AGG_TMP2:%.+]] = alloca [[ST_TY]],
306
307 // param copy
308 // CHECK: store [2 x i{{[0-9]+}}]* {{.+}}, [2 x i{{[0-9]+}}]** [[VEC_ADDR]],
309 // CHECK: store i{{[0-9]+}} {{.+}}, i{{[0-9]+}}* [[T_VAR_ADDR]],
310 // CHECK: store [2 x [[S_FLOAT_TY]]]* {{.+}}, [2 x [[S_FLOAT_TY]]]** [[S_ARR_ADDR]],
311 // CHECK: store [[S_FLOAT_TY]]* {{.+}}, [[S_FLOAT_TY]]** [[VAR_ADDR]],
312 // CHECK: store i{{[0-9]+}} {{.+}}, i{{[0-9]+}}* [[SIVAR_ADDR]],
313
314 // T_VAR and SIVAR
315 // CHECK-64-DAG: [[CONV_TVAR:%.+]] = bitcast i64* [[T_VAR_ADDR]] to i32*
316 // CHECK-64-DAG: [[CONV_SIVAR:%.+]] = bitcast i64* [[SIVAR_ADDR]] to i32*
317
318 // preparation vars
319 // CHECK-DAG: [[VEC_ADDR_VAL:%.+]] = load [2 x i{{[0-9]+}}]*, [2 x i{{[0-9]+}}]** [[VEC_ADDR]],
320 // CHECK-DAG: [[S_ARR_ADDR_REF:%.+]] = load [2 x [[S_FLOAT_TY]]]*, [2 x [[S_FLOAT_TY]]]** [[S_ARR_ADDR]],
321 // CHECK-DAG: [[VAR_ADDR_REF:%.+]] = load{{.+}} [[VAR_ADDR]],
322
323 // firstprivate vec(vec): copy from *_addr into priv1 and then from priv1 into priv2
324 // CHECK-DAG: [[VEC_DEST_PRIV:%.+]] = bitcast [2 x i{{[0-9]+}}]* [[VEC_PRIV]] to i8*
325 // CHECK-DAG: [[VEC_SRC:%.+]] = bitcast [2 x i{{[0-9]+}}]* [[VEC_ADDR_VAL]] to i8*
326 // CHECK: call void @llvm.memcpy.{{.+}}(i8* align {{[0-9]+}} [[VEC_DEST_PRIV]], i8* align {{[0-9]+}} [[VEC_SRC]], {{.+}})
327
328 // firstprivate(s_arr)
329 // CHECK-DAG: [[S_ARR_PRIV_BGN:%.+]] = getelementptr{{.*}} [2 x [[S_FLOAT_TY]]], [2 x [[S_FLOAT_TY]]]* [[S_ARR_PRIV]],
330 // CHECK-DAG: [[S_ARR_ADDR_BGN:%.+]] = bitcast [2 x [[S_FLOAT_TY]]]* [[S_ARR_ADDR_REF]] to
331 // CHECK-DAG: [[S_ARR_FIN:%.+]] = icmp{{.+}} [[S_ARR_PRIV_BGN]],
332 // CHECK-DAG: [[S_ARR_SRC_COPY:%.+]] = phi{{.+}} [ [[S_ARR_ADDR_BGN]], {{.+}} ], [ [[S_ARR_SRC:%.+]], {{.+}} ]
333 // CHECK-DAG: [[S_ARR_DST_COPY:%.+]] = phi{{.+}} [ [[S_ARR_PRIV_BGN]], {{.+}}], [ [[S_ARR_DST:%.+]], {{.+}} ]
334 // CHECK-DAG: call void @{{.+}}({{.+}} [[AGG_TMP1]])
335 // CHECK-DAG: call void @{{.+}}({{.+}} [[S_ARR_DST_COPY]], {{.+}} [[S_ARR_SRC_COPY]], {{.+}} [[AGG_TMP1]])
336 // CHECK-DAG: call void @{{.+}}({{.+}} [[AGG_TMP1]])
337 // CHECK-DAG: [[S_ARR_DST]] = getelementptr {{.+}} [[S_ARR_DST_COPY]],
338 // CHECK-DAG: [[S_ARR_SRC]] = getelementptr {{.+}} [[S_ARR_SRC_COPY]],
339
340 // firstprivate(var)
341 // CHECK-DAG: call void @{{.+}}({{.+}} [[AGG_TMP2]])
342 // CHECK-DAG: call void @{{.+}}({{.+}} [[VAR_PRIV]], {{.+}} [[VAR_ADDR_REF]], {{.+}} [[AGG_TMP2]])
343 // CHECK-DAG: call void @{{.+}}({{.+}} [[AGG_TMP2]])
344
345 // CHECK: call void @__kmpc_for_static_init_4(
346 // CHECK-32-DAG: {{.+}} = {{.+}} [[T_VAR_ADDR]]
347 // CHECK-64-DAG: {{.+}} = {{.+}} [[CONV_TVAR]]
348 // CHECK-DAG: {{.+}} = {{.+}} [[VEC_PRIV]]
349 // CHECK-DAG: {{.+}} = {{.+}} [[S_ARR_PRIV]]
350 // CHECK-DAG: {{.+}} = {{.+}} [[VAR_PRIV]]
351 // CHECK-32-DAG: {{.+}} = {{.+}} [[SIVAR_ADDR]]
352 // CHECK-64-DAG: {{.+}} = {{.+}} [[CONV_SIVAR]]
353 // CHECK: call void @__kmpc_for_static_fini(
354 // CHECK: ret void
355
356 // CHECK: define{{.*}} i{{[0-9]+}} @[[TMAIN_INT]]()
357 // CHECK: call i32 @__tgt_target_teams_mapper(%struct.ident_t* @{{.+}}, i64 -1, i8* @{{[^,]+}}, i32 4, i8** %{{[^,]+}}, i8** %{{[^,]+}}, i{{64|32}}* {{.+}}@{{[^,]+}}, i32 0, i32 0), i64* {{.+}}@{{[^,]+}}, i32 0, i32 0), i8** null
358 // CHECK: call void @[[TOFFL1:.+]](i{{64|32}} %{{.+}})
359 // CHECK: ret
360
361 // CHECK: define {{.*}}void @[[TOFFL1]]({{.+}})
362 // CHECK: [[TT_VAR_PRIV:%.+]] = alloca i{{[0-9]+}},
363 // CHECK: [[TVEC_PRIV:%.+]] = alloca [2 x i{{[0-9]+}}]*,
364 // CHECK: [[TS_ARR_PRIV:%.+]] = alloca [2 x [[S_INT_TY]]]*,
365 // CHECK: [[TVAR_PRIV:%.+]] = alloca [[S_INT_TY]]*,
366 // CHECK: [[TT_VAR_CAST:%.+]] = alloca i{{[0-9]+}},
367
368 // CHECK-DAG: [[TVEC_TE_PAR:%.+]] = load [2 x i{{[0-9]+}}]*, [2 x i{{[0-9]+}}]** [[TVEC_PRIV]],
369 // CHECK-DAG: [[TT_VAR_TE_PAR:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[TT_VAR_CAST]],
370 // CHECK-DAG: [[TS_ARR_TE_PAR:%.+]] = load [2 x [[S_INT_TY]]]*, [2 x [[S_INT_TY]]]** [[TS_ARR_PRIV]],
371 // CHECK-DAG: [[TVAR_TE_PAR:%.+]] = load [[S_INT_TY]]*, [[S_INT_TY]]** [[TVAR_PRIV]],
372
373 // CHECK: [[TVAR_TE_PAR:%.+]] = load [[S_INT_TY]]*, [[S_INT_TY]]** %
374 // CHECK: call void {{.+}} @__kmpc_fork_teams({{.+}}, i32 4, {{.+}} @[[TOUTL1:.+]] to {{.+}}, [2 x i{{[0-9]+}}]* [[TVEC_TE_PAR]], i{{[0-9]+}} [[TT_VAR_TE_PAR]], [2 x [[S_INT_TY]]]* [[TS_ARR_TE_PAR]], [[S_INT_TY]]* [[TVAR_TE_PAR]])
375 // CHECK: ret void
376
377 // CHECK: define internal void @[[TOUTL1]]({{.+}})
378 // Skip global and bound tid vars
379 // CHECK: {{.+}} = alloca i32*,
380 // CHECK: {{.+}} = alloca i32*,
381 // CHECK: [[VEC_ADDR:%.+]] = alloca [2 x i{{[0-9]+}}]*,
382 // CHECK: [[T_VAR_ADDR:%.+]] = alloca i{{[0-9]+}},
383 // CHECK: [[S_ARR_ADDR:%.+]] = alloca [2 x [[S_INT_TY]]]*,
384 // CHECK: [[VAR_ADDR:%.+]] = alloca [[S_INT_TY]]*,
385 // CHECK: [[TMP_VAR_ADDR:%.+]] = alloca [[S_INT_TY]]*,
386 // Skip temp vars for loop
387 // CHECK: alloca i{{[0-9]+}},
388 // CHECK: alloca i{{[0-9]+}},
389 // CHECK: alloca i{{[0-9]+}},
390 // CHECK: alloca i{{[0-9]+}},
391 // CHECK: alloca i{{[0-9]+}},
392 // CHECK: [[VEC_PRIV:%.+]] = alloca [2 x i{{[0-9]+}}],
393 // CHECK: [[S_ARR_PRIV:%.+]] = alloca [2 x [[S_INT_TY]]],
394 // CHECK: [[AGG_TMP1:%.+]] = alloca [[ST_TY]],
395 // CHECK: [[VAR_PRIV:%.+]] = alloca [[S_INT_TY]],
396 // CHECK: [[AGG_TMP2:%.+]] = alloca [[ST_TY]],
397 // CHECK: [[TMP:%.+]] = alloca [[S_INT_TY]]*,
398
399 // param copy
400 // CHECK: store [2 x i{{[0-9]+}}]* {{.+}}, [2 x i{{[0-9]+}}]** [[VEC_ADDR]],
401 // CHECK: store i{{[0-9]+}} {{.+}}, i{{[0-9]+}}* [[T_VAR_ADDR]],
402 // CHECK: store [2 x [[S_INT_TY]]]* {{.+}}, [2 x [[S_INT_TY]]]** [[S_ARR_ADDR]],
403 // CHECK: store [[S_INT_TY]]* {{.+}}, [[S_INT_TY]]** [[VAR_ADDR]],
404
405 // T_VAR and preparation variables
406 // CHECK: [[VEC_ADDR_VAL:%.+]] = load [2 x i{{[0-9]+}}]*, [2 x i{{[0-9]+}}]** [[VEC_ADDR]],
407 // CHECK-64: [[CONV_TVAR:%.+]] = bitcast i64* [[T_VAR_ADDR]] to i32*
408 // CHECK: [[S_ARR_ADDR_REF:%.+]] = load [2 x [[S_INT_TY]]]*, [2 x [[S_INT_TY]]]** [[S_ARR_ADDR]],
409
410 // firstprivate vec(vec): copy from *_addr into priv1 and then from priv1 into priv2
411 // CHECK-DAG: [[VEC_DEST_PRIV:%.+]] = bitcast [2 x i{{[0-9]+}}]* [[VEC_PRIV]] to i8*
412 // CHECK-DAG: [[VEC_SRC:%.+]] = bitcast [2 x i{{[0-9]+}}]* [[VEC_ADDR_VAL]] to i8*
413 // CHECK: call void @llvm.memcpy.{{.+}}(i8* align {{[0-9]+}} [[VEC_DEST_PRIV]], i8* align {{[0-9]+}} [[VEC_SRC]], {{.+}})
414
415 // firstprivate(s_arr)
416 // CHECK-DAG: [[S_ARR_PRIV_BGN:%.+]] = getelementptr{{.*}} [2 x [[S_INT_TY]]], [2 x [[S_INT_TY]]]* [[S_ARR_PRIV]],
417 // CHECK-DAG: [[S_ARR_ADDR_BGN:%.+]] = bitcast [2 x [[S_INT_TY]]]* [[S_ARR_ADDR_REF]] to
418 // CHECK-DAG: [[S_ARR_FIN:%.+]] = icmp{{.+}} [[S_ARR_PRIV_BGN]],
419 // CHECK-DAG: [[S_ARR_SRC_COPY:%.+]] = phi{{.+}} [ [[S_ARR_ADDR_BGN]], {{.+}} ], [ [[S_ARR_SRC:%.+]], {{.+}} ]
420 // CHECK-DAG: [[S_ARR_DST_COPY:%.+]] = phi{{.+}} [ [[S_ARR_PRIV_BGN]], {{.+}} ], [ [[S_ARR_DST:%.+]], {{.+}} ]
421 // CHECK-DAG: call void @{{.+}}({{.+}} [[AGG_TMP1]])
422 // CHECK-DAG: call void @{{.+}}({{.+}} [[S_ARR_DST_COPY]], {{.+}} [[S_ARR_SRC_COPY]], {{.+}} [[AGG_TMP1]])
423 // CHECK-DAG: call void @{{.+}}({{.+}} [[AGG_TMP1]])
424 // CHECK-DAG: [[S_ARR_DST]] = getelementptr {{.+}} [[S_ARR_DST_COPY]],
425 // CHECK-DAG: [[S_ARR_SRC]] = getelementptr {{.+}} [[S_ARR_SRC_COPY]],
426
427 // firstprivate(var)
428 // CHECK-DAG: [[VAR_ADDR_REF:%.+]] = load{{.+}} [[TMP_VAR_ADDR]],
429 // CHECK-DAG: call void @{{.+}}({{.+}} [[AGG_TMP2]])
430 // CHECK-DAG: call void @{{.+}}({{.+}} [[VAR_PRIV]], {{.+}} [[VAR_ADDR_REF]], {{.+}} [[AGG_TMP2]])
431 // CHECK-DAG: call void @{{.+}}({{.+}} [[AGG_TMP2]])
432 // CHECK-DAG: store [[S_INT_TY]]* [[VAR_PRIV]], [[S_INT_TY]]** [[TMP]],
433
434 // CHECK: call void @__kmpc_for_static_init_4(
435 // CHECK: call void {{.*}} @__kmpc_fork_call({{.+}}, {{.+}}, {{.+}} @[[TPAR_OUTL:.+]] to
436 // CHECK: call void @__kmpc_for_static_fini(
437 // CHECK: ret void
438
439 // CHECK: define internal void @[[TPAR_OUTL]]({{.+}})
440 // Skip global and bound tid vars
441 // CHECK: {{.+}} = alloca i32*,
442 // CHECK: {{.+}} = alloca i32*,
443 // CHECK: [[VEC_ADDR:%.+]] = alloca [2 x i{{[0-9]+}}]*,
444 // CHECK: [[T_VAR_ADDR:%.+]] = alloca i{{[0-9]+}},
445 // CHECK: [[S_ARR_ADDR:%.+]] = alloca [2 x [[S_INT_TY]]]*,
446 // CHECK: [[VAR_ADDR:%.+]] = alloca [[S_INT_TY]]*,
447 // Skip temp vars for loop
448 // CHECK: alloca i{{[0-9]+}},
449 // CHECK: alloca i{{[0-9]+}},
450 // CHECK: alloca i{{[0-9]+}},
451 // CHECK: alloca i{{[0-9]+}},
452 // CHECK: alloca i{{[0-9]+}},
453 // CHECK: [[VEC_PRIV:%.+]] = alloca [2 x i{{[0-9]+}}],
454 // CHECK: [[S_ARR_PRIV:%.+]] = alloca [2 x [[S_INT_TY]]],
455 // CHECK: [[AGG_TMP1:%.+]] = alloca [[ST_TY]],
456 // CHECK: [[VAR_PRIV:%.+]] = alloca [[S_INT_TY]],
457 // CHECK: [[AGG_TMP2:%.+]] = alloca [[ST_TY]],
458 // CHECK: [[TMP:%.+]] = alloca [[S_INT_TY]]*,
459
460 // param copy
461 // CHECK: store [2 x i{{[0-9]+}}]* {{.+}}, [2 x i{{[0-9]+}}]** [[VEC_ADDR]],
462 // CHECK: store i{{[0-9]+}} {{.+}}, i{{[0-9]+}}* [[T_VAR_ADDR]],
463 // CHECK: store [2 x [[S_INT_TY]]]* {{.+}}, [2 x [[S_INT_TY]]]** [[S_ARR_ADDR]],
464 // CHECK: store [[S_INT_TY]]* {{.+}}, [[S_INT_TY]]** [[VAR_ADDR]],
465
466 // T_VAR and preparation variables
467 // CHECK: [[VEC_ADDR_VAL:%.+]] = load [2 x i{{[0-9]+}}]*, [2 x i{{[0-9]+}}]** [[VEC_ADDR]],
468 // CHECK-64: [[CONV_TVAR:%.+]] = bitcast i64* [[T_VAR_ADDR]] to i32*
469 // CHECK: [[S_ARR_ADDR_REF:%.+]] = load [2 x [[S_INT_TY]]]*, [2 x [[S_INT_TY]]]** [[S_ARR_ADDR]],
470
471 // firstprivate vec(vec): copy from *_addr into priv1 and then from priv1 into priv2
472 // CHECK-DAG: [[VEC_DEST_PRIV:%.+]] = bitcast [2 x i{{[0-9]+}}]* [[VEC_PRIV]] to i8*
473 // CHECK-DAG: [[VEC_SRC:%.+]] = bitcast [2 x i{{[0-9]+}}]* [[VEC_ADDR_VAL]] to i8*
474 // CHECK: call void @llvm.memcpy.{{.+}}(i8* align {{[0-9]+}} [[VEC_DEST_PRIV]], i8* align {{[0-9]+}} [[VEC_SRC]], {{.+}})
475
476 // firstprivate(s_arr)
477 // CHECK-DAG: [[S_ARR_PRIV_BGN:%.+]] = getelementptr{{.*}} [2 x [[S_INT_TY]]], [2 x [[S_INT_TY]]]* [[S_ARR_PRIV]],
478 // CHECK-DAG: [[S_ARR_ADDR_BGN:%.+]] = bitcast [2 x [[S_INT_TY]]]* [[S_ARR_ADDR_REF]] to
479 // CHECK-DAG: [[S_ARR_FIN:%.+]] = icmp{{.+}} [[S_ARR_PRIV_BGN]],
480 // CHECK-DAG: [[S_ARR_SRC_COPY:%.+]] = phi{{.+}} [ [[S_ARR_ADDR_BGN]], {{.+}} ], [ [[S_ARR_SRC:%.+]], {{.+}} ]
481 // CHECK-DAG: [[S_ARR_DST_COPY:%.+]] = phi{{.+}} [ [[S_ARR_PRIV_BGN]], {{.+}} ], [ [[S_ARR_DST:%.+]], {{.+}} ]
482 // CHECK-DAG: call void @{{.+}}({{.+}} [[AGG_TMP1]])
483 // CHECK-DAG: call void @{{.+}}({{.+}} [[S_ARR_DST_COPY]], {{.+}} [[S_ARR_SRC_COPY]], {{.+}} [[AGG_TMP1]])
484 // CHECK-DAG: call void @{{.+}}({{.+}} [[AGG_TMP1]])
485 // CHECK-DAG: [[S_ARR_DST]] = getelementptr {{.+}} [[S_ARR_DST_COPY]],
486 // CHECK-DAG: [[S_ARR_SRC]] = getelementptr {{.+}} [[S_ARR_SRC_COPY]],
487
488 // firstprivate(var)
489 // CHECK-DAG: [[VAR_ADDR_REF:%.+]] = load{{.+}} [[TMP_VAR_ADDR]],
490 // CHECK-DAG: call void @{{.+}}({{.+}} [[AGG_TMP2]])
491 // CHECK-DAG: call void @{{.+}}({{.+}} [[VAR_PRIV]], {{.+}} [[VAR_ADDR_REF]], {{.+}} [[AGG_TMP2]])
492 // CHECK-DAG: call void @{{.+}}({{.+}} [[AGG_TMP2]])
493 // CHECK-DAG: store [[S_INT_TY]]* [[VAR_PRIV]], [[S_INT_TY]]** [[TMP]],
494
495 // CHECK: call void @__kmpc_for_static_init_4(
496 // CHECK-32-DAG: {{.+}} = {{.+}} [[T_VAR_ADDR]]
497 // CHECK-64-DAG: {{.+}} = {{.+}} [[CONV_TVAR]]
498 // CHECK-DAG: {{.+}} = {{.+}} [[VEC_PRIV]]
499 // CHECK-DAG: {{.+}} = {{.+}} [[TMP]]
500 // CHECK-DAG: {{.+}} = {{.+}} [[S_ARR_PRIV]]
501 // CHECK: call void @__kmpc_for_static_fini(
502 // CHECK: ret void
503
504 // CHECK: !{!"llvm.loop.vectorize.enable", i1 true}
505
506 #endif
507