1 // RUN: %clang_cc1 -verify -fopenmp -x c++ -triple x86_64-apple-darwin10 -emit-llvm %s -o - | FileCheck %s
2 // RUN: %clang_cc1 -fopenmp -x c++ -std=c++11 -triple x86_64-apple-darwin10 -emit-pch -o %t %s
3 // RUN: %clang_cc1 -fopenmp -x c++ -triple x86_64-apple-darwin10 -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s
4 // expected-no-diagnostics
5 // REQUIRES: x86-registered-target
6 #ifndef HEADER
7 #define HEADER
8
9 volatile double g, g_orig;
10 volatile double &g1 = g_orig;
11
12 struct BaseS {
13 int x;
14 };
15 struct BaseS1 {
16 float y;
17 };
18
19 template <class T>
20 struct S : public BaseS, public BaseS1 {
21 T f;
SS22 S(T a) : f(a + g) {}
SS23 S() : f(g) {}
~SS24 ~S() {}
25 };
26 void red(BaseS1&, const BaseS1&);
27 void red_plus(BaseS1&, const BaseS1&);
28 void init(BaseS1&, const BaseS1&);
29 void init1(BaseS1&, const BaseS1&);
30 void init2(BaseS1&, const BaseS1&);
31 void init_plus(BaseS1&, const BaseS1&);
32 #pragma omp declare reduction(operator& : BaseS1 : red(omp_out, omp_in)) initializer(init(omp_priv, omp_orig))
33 #pragma omp declare reduction(+ : BaseS1 : red_plus(omp_out, omp_in)) initializer(init_plus(omp_priv, omp_orig))
34 #pragma omp declare reduction(&& : S<float>, S<int> : omp_out.f *= omp_in.f) initializer(init1(omp_priv, omp_orig))
35
36 // CHECK-DAG: [[S_FLOAT_TY:%.+]] = type { %{{[^,]+}}, %{{[^,]+}}, float }
37 // CHECK-DAG: [[S_INT_TY:%.+]] = type { %{{[^,]+}}, %{{[^,]+}}, i{{[0-9]+}} }
38 // CHECK-DAG: [[ATOMIC_REDUCE_BARRIER_LOC:@.+]] = private unnamed_addr constant %{{.+}} { i32 0, i32 18, i32 0, i32 0, i8*
39 // CHECK-DAG: [[IMPLICIT_BARRIER_LOC:@.+]] = private unnamed_addr constant %{{.+}} { i32 0, i32 66, i32 0, i32 0, i8*
40 // CHECK-DAG: [[REDUCTION_LOC:@.+]] = private unnamed_addr constant %{{.+}} { i32 0, i32 18, i32 0, i32 0, i8*
41 // CHECK-DAG: [[REDUCTION_LOCK:@.+]] = common global [8 x i32] zeroinitializer
42
43 #pragma omp declare reduction(operator&& : int : omp_out = 111 & omp_in)
44 template <typename T>
tmain()45 T tmain() {
46 T t;
47 S<T> test;
48 T t_var = T(), t_var1;
49 T vec[] = {1, 2};
50 S<T> s_arr[] = {1, 2};
51 S<T> &var = test;
52 S<T> var1;
53 #pragma omp declare reduction(operator& : T : omp_out = 15 + omp_in)
54 #pragma omp declare reduction(operator+ : T : omp_out = 1513 + omp_in) initializer(omp_priv = 321)
55 #pragma omp declare reduction(min : T : omp_out = 47 - omp_in) initializer(omp_priv = 432 / omp_orig)
56 #pragma omp declare reduction(operator&& : S<T> : omp_out.f = 17 * omp_in.f) initializer(init2(omp_priv, omp_orig))
57 #pragma omp declare reduction(operator&& : T : omp_out = 17 * omp_in)
58 #pragma omp parallel
59 #pragma omp for reduction(+ : t_var) reduction(& : var) reduction(&& : var1) reduction(min : t_var1) nowait
60 for (int i = 0; i < 2; ++i) {
61 vec[i] = t_var;
62 s_arr[i] = var;
63 }
64 #pragma omp parallel
65 #pragma omp for reduction(&& : t_var)
66 for (int i = 0; i < 2; ++i) {
67 vec[i] = t_var;
68 s_arr[i] = var;
69 }
70 return T();
71 }
72
73 extern S<float> **foo();
74
75 #pragma omp declare reduction(operator- : float, double : omp_out = 333 + omp_in)
76 #pragma omp declare reduction(min : float, double : omp_out = 555 + omp_in)
main()77 int main() {
78 #pragma omp declare reduction(operator+ : float, double : omp_out = 222 - omp_in) initializer(omp_priv = -1)
79 S<float> test;
80 float t_var = 0, t_var1;
81 int vec[] = {1, 2};
82 S<float> s_arr[] = {1, 2};
83 S<float> &var = test;
84 S<float> var1, arrs[10][4];
85 S<float> **var2 = foo();
86 S<float> vvar2[2];
87 S<float>(&var3)[2] = s_arr;
88 #pragma omp declare reduction(operator+ : int : omp_out = 555 * omp_in) initializer(omp_priv = 888)
89 #pragma omp parallel
90 #pragma omp for reduction(+ : t_var) reduction(& : var) reduction(&& : var1) reduction(min : t_var1)
91 for (int i = 0; i < 2; ++i) {
92 vec[i] = t_var;
93 s_arr[i] = var;
94 }
95 int arr[10][vec[1]];
96 #pragma omp parallel for reduction(+ : arr[1][ : vec[1]]) reduction(& : arrs[1 : vec[1]][1 : 2])
97 for (int i = 0; i < 10; ++i)
98 ++arr[1][i];
99 #pragma omp parallel
100 #pragma omp for reduction(+ : arr) reduction(& : arrs)
101 for (int i = 0; i < 10; ++i)
102 ++arr[1][i];
103 #pragma omp parallel
104 #pragma omp for reduction(& : var2[0 : 5][1 : 6])
105 for (int i = 0; i < 10; ++i)
106 ;
107 #pragma omp parallel
108 #pragma omp for reduction(& : vvar2[0 : 5])
109 for (int i = 0; i < 10; ++i)
110 ;
111 #pragma omp parallel
112 #pragma omp for reduction(& : var3[1 : 2])
113 for (int i = 0; i < 10; ++i)
114 ;
115 #pragma omp parallel
116 #pragma omp for reduction(& : var3)
117 for (int i = 0; i < 10; ++i)
118 ;
119 return tmain<int>();
120 }
121
122 // CHECK: define {{.*}}i{{[0-9]+}} @main()
123 // CHECK: [[TEST:%.+]] = alloca [[S_FLOAT_TY]],
124 // CHECK: call {{.*}} [[S_FLOAT_TY_CONSTR:@.+]]([[S_FLOAT_TY]]* [[TEST]])
125 // CHECK: call void (%{{.+}}*, i{{[0-9]+}}, void (i{{[0-9]+}}*, i{{[0-9]+}}*, ...)*, ...) @__kmpc_fork_call(%{{.+}}* @{{.+}}, i{{[0-9]+}} 6, void (i{{[0-9]+}}*, i{{[0-9]+}}*, ...)* bitcast (void (i{{[0-9]+}}*, i{{[0-9]+}}*, float*, [[S_FLOAT_TY]]*, [[S_FLOAT_TY]]*, float*, [2 x i32]*, [2 x [[S_FLOAT_TY]]]*)* [[MAIN_MICROTASK:@.+]] to void
126 // CHECK: call void (%{{.+}}*, i{{[0-9]+}}, void (i{{[0-9]+}}*, i{{[0-9]+}}*, ...)*, ...) @__kmpc_fork_call(%{{.+}}* @{{.+}}, i{{[0-9]+}} 5, void (i{{[0-9]+}}*, i{{[0-9]+}}*, ...)* bitcast (void (i{{[0-9]+}}*, i{{[0-9]+}}*, i64, i64, i32*, [2 x i32]*, [10 x [4 x [[S_FLOAT_TY]]]]*)* [[MAIN_MICROTASK1:@.+]] to void
127 // CHECK: call void (%{{.+}}*, i{{[0-9]+}}, void (i{{[0-9]+}}*, i{{[0-9]+}}*, ...)*, ...) @__kmpc_fork_call(%{{.+}}* @{{.+}}, i{{[0-9]+}} 4, void (i{{[0-9]+}}*, i{{[0-9]+}}*, ...)* bitcast (void (i{{[0-9]+}}*, i{{[0-9]+}}*, i64, i64, i32*, [10 x [4 x [[S_FLOAT_TY]]]]*)* [[MAIN_MICROTASK2:@.+]] to void
128 // CHECK: call void (%{{.+}}*, i{{[0-9]+}}, void (i{{[0-9]+}}*, i{{[0-9]+}}*, ...)*, ...) @__kmpc_fork_call(%{{.+}}* @{{.+}}, i{{[0-9]+}} 1, void (i{{[0-9]+}}*, i{{[0-9]+}}*, ...)* bitcast (void (i{{[0-9]+}}*, i{{[0-9]+}}*, [[S_FLOAT_TY]]***)* [[MAIN_MICROTASK3:@.+]] to void
129 // CHECK: call void (%{{.+}}*, i{{[0-9]+}}, void (i{{[0-9]+}}*, i{{[0-9]+}}*, ...)*, ...) @__kmpc_fork_call(%{{.+}}* @{{.+}}, i{{[0-9]+}} 1, void (i{{[0-9]+}}*, i{{[0-9]+}}*, ...)* bitcast (void (i{{[0-9]+}}*, i{{[0-9]+}}*, [2 x [[S_FLOAT_TY]]]*)* [[MAIN_MICROTASK4:@.+]] to void
130 // CHECK: call void (%{{.+}}*, i{{[0-9]+}}, void (i{{[0-9]+}}*, i{{[0-9]+}}*, ...)*, ...) @__kmpc_fork_call(%{{.+}}* @{{.+}}, i{{[0-9]+}} 1, void (i{{[0-9]+}}*, i{{[0-9]+}}*, ...)* bitcast (void (i{{[0-9]+}}*, i{{[0-9]+}}*, [2 x [[S_FLOAT_TY]]]*)* [[MAIN_MICROTASK5:@.+]] to void
131 // CHECK: call void (%{{.+}}*, i{{[0-9]+}}, void (i{{[0-9]+}}*, i{{[0-9]+}}*, ...)*, ...) @__kmpc_fork_call(%{{.+}}* @{{.+}}, i{{[0-9]+}} 1, void (i{{[0-9]+}}*, i{{[0-9]+}}*, ...)* bitcast (void (i{{[0-9]+}}*, i{{[0-9]+}}*, [2 x [[S_FLOAT_TY]]]*)* [[MAIN_MICROTASK6:@.+]] to void
132 // CHECK: = call {{.*}}i{{.+}} [[TMAIN_INT:@.+]]()
133 // CHECK: call {{.*}} [[S_FLOAT_TY_DESTR:@.+]]([[S_FLOAT_TY]]*
134 // CHECK: ret
135 //
136 // CHECK: define internal void [[MAIN_MICROTASK]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, float* dereferenceable(4) %{{.+}}, [[S_FLOAT_TY]]* dereferenceable(12) %{{.+}}, [[S_FLOAT_TY]]* dereferenceable(12) %{{.+}}, float* dereferenceable(4) %{{.+}}, [2 x i32]* dereferenceable(8) %vec, [2 x [[S_FLOAT_TY]]]* dereferenceable(24) %{{.+}})
137 // CHECK: [[T_VAR_PRIV:%.+]] = alloca float,
138 // CHECK: [[VAR_PRIV:%.+]] = alloca [[S_FLOAT_TY]],
139 // CHECK: [[VAR1_PRIV:%.+]] = alloca [[S_FLOAT_TY]],
140 // CHECK: [[T_VAR1_PRIV:%.+]] = alloca float,
141
142 // Reduction list for runtime.
143 // CHECK: [[RED_LIST:%.+]] = alloca [4 x i8*],
144
145 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]],
146
147 // CHECK: [[T_VAR_REF:%.+]] = load float*, float** %
148 // CHECK: [[VAR1_REF:%.+]] = load [[S_FLOAT_TY]]*, [[S_FLOAT_TY]]** %
149 // CHECK: [[T_VAR1_REF:%.+]] = load float*, float** %
150
151 // For + reduction operation initial value of private variable is -1.
152 // CHECK: store float -1.0{{.+}}, float*
153
154 // For & reduction operation initial value of private variable is defined by call of 'init()' function.
155 // CHECK: call {{.*}}void @_Z4initR6BaseS1RKS_(
156
157 // For && reduction operation initial value of private variable is 1.0.
158 // CHECK: call {{.*}}void @_Z5init1R6BaseS1RKS_(
159
160 // For min reduction operation initial value of private variable is largest repesentable value.
161 // CHECK: [[INIT:%.+]] = load float, float* @
162 // CHECK: store float [[INIT]], float* [[T_VAR1_PRIV]],
163
164 // CHECK: [[GTID_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[GTID_ADDR_ADDR]]
165 // CHECK: [[GTID:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[GTID_REF]]
166 // CHECK: call void @__kmpc_for_static_init_4(
167 // Skip checks for internal operations.
168 // CHECK: call void @__kmpc_for_static_fini(
169
170 // void *RedList[<n>] = {<ReductionVars>[0], ..., <ReductionVars>[<n>-1]};
171
172 // CHECK: [[T_VAR_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 0
173 // CHECK: [[BITCAST:%.+]] = bitcast float* [[T_VAR_PRIV]] to i8*
174 // CHECK: store i8* [[BITCAST]], i8** [[T_VAR_PRIV_REF]],
175 // CHECK: [[VAR_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 1
176 // CHECK: [[BITCAST:%.+]] = bitcast [[S_FLOAT_TY]]* [[VAR_PRIV]] to i8*
177 // CHECK: store i8* [[BITCAST]], i8** [[VAR_PRIV_REF]],
178 // CHECK: [[VAR1_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 2
179 // CHECK: [[BITCAST:%.+]] = bitcast [[S_FLOAT_TY]]* [[VAR1_PRIV]] to i8*
180 // CHECK: store i8* [[BITCAST]], i8** [[VAR1_PRIV_REF]],
181 // CHECK: [[T_VAR1_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 3
182 // CHECK: [[BITCAST:%.+]] = bitcast float* [[T_VAR1_PRIV]] to i8*
183 // CHECK: store i8* [[BITCAST]], i8** [[T_VAR1_PRIV_REF]],
184
185 // res = __kmpc_reduce(<loc>, <gtid>, <n>, sizeof(RedList), RedList, reduce_func, &<lock>);
186
187 // CHECK: [[BITCAST:%.+]] = bitcast [4 x i8*]* [[RED_LIST]] to i8*
188 // CHECK: [[RES:%.+]] = call i32 @__kmpc_reduce(%{{.+}}* [[REDUCTION_LOC]], i32 [[GTID]], i32 4, i64 32, i8* [[BITCAST]], void (i8*, i8*)* [[REDUCTION_FUNC:@.+]], [8 x i32]* [[REDUCTION_LOCK]])
189
190 // switch(res)
191 // CHECK: switch i32 [[RES]], label %[[RED_DONE:.+]] [
192 // CHECK: i32 1, label %[[CASE1:.+]]
193 // CHECK: i32 2, label %[[CASE2:.+]]
194 // CHECK: ]
195
196 // case 1:
197 // t_var += t_var_reduction;
198 // CHECK: fsub float 2.220000e+02, %
199
200 // var = var.operator &(var_reduction);
201 // CHECK: call {{.*}}void @_Z3redR6BaseS1RKS_(
202
203 // var1 = var1.operator &&(var1_reduction);
204 // CHECK: fmul float
205
206 // t_var1 = min(t_var1, t_var1_reduction);
207 // CHECK: fadd float 5.550000e+02, %
208
209 // __kmpc_end_reduce(<loc>, <gtid>, &<lock>);
210 // CHECK: call void @__kmpc_end_reduce(%{{.+}}* [[REDUCTION_LOC]], i32 [[GTID]], [8 x i32]* [[REDUCTION_LOCK]])
211
212 // break;
213 // CHECK: br label %[[RED_DONE]]
214
215 // case 2:
216 // t_var += t_var_reduction;
217 // CHECK: call void @__kmpc_critical(
218 // CHECK: fsub float 2.220000e+02, %
219 // CHECK: call void @__kmpc_end_critical(
220
221 // var = var.operator &(var_reduction);
222 // CHECK: call void @__kmpc_critical(
223 // CHECK: call {{.*}}void @_Z3redR6BaseS1RKS_(
224 // CHECK: call void @__kmpc_end_critical(
225
226 // var1 = var1.operator &&(var1_reduction);
227 // CHECK: call void @__kmpc_critical(
228 // CHECK: fmul float
229 // CHECK: call void @__kmpc_end_critical(
230
231 // t_var1 = min(t_var1, t_var1_reduction);
232 // CHECK: call void @__kmpc_critical(
233 // CHECK: fadd float 5.550000e+02, %
234 // CHECK: call void @__kmpc_end_critical(
235
236 // __kmpc_end_reduce(<loc>, <gtid>, &<lock>);
237 // CHECK: call void @__kmpc_end_reduce(%{{.+}}* [[REDUCTION_LOC]], i32 [[GTID]], [8 x i32]* [[REDUCTION_LOCK]])
238
239 // break;
240 // CHECK: br label %[[RED_DONE]]
241 // CHECK: [[RED_DONE]]
242 // CHECK-DAG: call {{.*}} [[S_FLOAT_TY_DESTR]]([[S_FLOAT_TY]]* [[VAR_PRIV]])
243 // CHECK-DAG: call {{.*}} [[S_FLOAT_TY_DESTR]]([[S_FLOAT_TY]]*
244 // CHECK: call void @__kmpc_barrier(%{{.+}}* [[IMPLICIT_BARRIER_LOC]], i{{[0-9]+}} [[GTID]])
245
246 // CHECK: ret void
247
248 // void reduce_func(void *lhs[<n>], void *rhs[<n>]) {
249 // *(Type0*)lhs[0] = ReductionOperation0(*(Type0*)lhs[0], *(Type0*)rhs[0]);
250 // ...
251 // *(Type<n>-1*)lhs[<n>-1] = ReductionOperation<n>-1(*(Type<n>-1*)lhs[<n>-1],
252 // *(Type<n>-1*)rhs[<n>-1]);
253 // }
254 // CHECK: define internal void [[REDUCTION_FUNC]](i8*, i8*)
255 // t_var_lhs = (float*)lhs[0];
256 // CHECK: [[T_VAR_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS:%.+]], i64 0, i64 0
257 // CHECK: [[T_VAR_RHS_VOID:%.+]] = load i8*, i8** [[T_VAR_RHS_REF]],
258 // CHECK: [[T_VAR_RHS:%.+]] = bitcast i8* [[T_VAR_RHS_VOID]] to float*
259 // t_var_rhs = (float*)rhs[0];
260 // CHECK: [[T_VAR_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS:%.+]], i64 0, i64 0
261 // CHECK: [[T_VAR_LHS_VOID:%.+]] = load i8*, i8** [[T_VAR_LHS_REF]],
262 // CHECK: [[T_VAR_LHS:%.+]] = bitcast i8* [[T_VAR_LHS_VOID]] to float*
263
264 // var_lhs = (S<float>*)lhs[1];
265 // CHECK: [[VAR_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS]], i64 0, i64 1
266 // CHECK: [[VAR_RHS_VOID:%.+]] = load i8*, i8** [[VAR_RHS_REF]],
267 // CHECK: [[VAR_RHS:%.+]] = bitcast i8* [[VAR_RHS_VOID]] to [[S_FLOAT_TY]]*
268 // var_rhs = (S<float>*)rhs[1];
269 // CHECK: [[VAR_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 1
270 // CHECK: [[VAR_LHS_VOID:%.+]] = load i8*, i8** [[VAR_LHS_REF]],
271 // CHECK: [[VAR_LHS:%.+]] = bitcast i8* [[VAR_LHS_VOID]] to [[S_FLOAT_TY]]*
272
273 // var1_lhs = (S<float>*)lhs[2];
274 // CHECK: [[VAR1_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS]], i64 0, i64 2
275 // CHECK: [[VAR1_RHS_VOID:%.+]] = load i8*, i8** [[VAR1_RHS_REF]],
276 // CHECK: [[VAR1_RHS:%.+]] = bitcast i8* [[VAR1_RHS_VOID]] to [[S_FLOAT_TY]]*
277 // var1_rhs = (S<float>*)rhs[2];
278 // CHECK: [[VAR1_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 2
279 // CHECK: [[VAR1_LHS_VOID:%.+]] = load i8*, i8** [[VAR1_LHS_REF]],
280 // CHECK: [[VAR1_LHS:%.+]] = bitcast i8* [[VAR1_LHS_VOID]] to [[S_FLOAT_TY]]*
281
282 // t_var1_lhs = (float*)lhs[3];
283 // CHECK: [[T_VAR1_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS]], i64 0, i64 3
284 // CHECK: [[T_VAR1_RHS_VOID:%.+]] = load i8*, i8** [[T_VAR1_RHS_REF]],
285 // CHECK: [[T_VAR1_RHS:%.+]] = bitcast i8* [[T_VAR1_RHS_VOID]] to float*
286 // t_var1_rhs = (float*)rhs[3];
287 // CHECK: [[T_VAR1_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 3
288 // CHECK: [[T_VAR1_LHS_VOID:%.+]] = load i8*, i8** [[T_VAR1_LHS_REF]],
289 // CHECK: [[T_VAR1_LHS:%.+]] = bitcast i8* [[T_VAR1_LHS_VOID]] to float*
290
291 // t_var_lhs += t_var_rhs;
292 // CHECK: fsub float 2.220000e+02, %
293
294 // var_lhs = var_lhs.operator &(var_rhs);
295 // CHECK: call {{.*}}void @_Z3redR6BaseS1RKS_(
296
297 // var1_lhs = var1_lhs.operator &&(var1_rhs);
298 // CHECK: fmul float
299
300 // t_var1_lhs = min(t_var1_lhs, t_var1_rhs);
301 // CHECK: fadd float 5.550000e+02, %
302 // CHECK: ret void
303
304 // CHECK: define internal void [[MAIN_MICROTASK1]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, i64 %{{.+}}, i64 %{{.+}}, i32* nonnull %{{.+}}, [2 x i32]* dereferenceable(8) %{{.+}}, [10 x [4 x [[S_FLOAT_TY]]]]* dereferenceable(480) %{{.+}})
305
306 // Reduction list for runtime.
307 // CHECK: [[RED_LIST:%.+]] = alloca [4 x i8*],
308
309 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]],
310
311 // CHECK: [[IDX1:%.+]] = mul nsw i64 1, %{{.+}}
312 // CHECK: [[LB1:%.+]] = getelementptr inbounds i32, i32* %{{.+}}, i64 [[IDX1]]
313 // CHECK: [[LB1_0:%.+]] = getelementptr inbounds i32, i32* [[LB1]], i64 0
314 // CHECK: [[IDX1:%.+]] = mul nsw i64 1, %{{.+}}
315 // CHECK: [[UB1:%.+]] = getelementptr inbounds i32, i32* %{{.+}}, i64 [[IDX1]]
316 // CHECK: [[UB1_UP:%.+]] = getelementptr inbounds i32, i32* [[UB1]], i64 %
317 // CHECK: [[UB_CAST:%.+]] = ptrtoint i32* [[UB1_UP]] to i64
318 // CHECK: [[LB_CAST:%.+]] = ptrtoint i32* [[LB1_0]] to i64
319 // CHECK: [[DIFF:%.+]] = sub i64 [[UB_CAST]], [[LB_CAST]]
320 // CHECK: [[SIZE_1:%.+]] = sdiv exact i64 [[DIFF]], ptrtoint (i32* getelementptr (i32, i32* null, i32 1) to i64)
321 // CHECK: [[ARR_SIZE:%.+]] = add nuw i64 [[SIZE_1]], 1
322 // CHECK: call i8* @llvm.stacksave()
323 // CHECK: [[ARR_PRIV:%.+]] = alloca i32, i64 [[ARR_SIZE]],
324
325 // Check initialization of private copy.
326 // CHECK: [[END:%.+]] = getelementptr i32, i32* [[ARR_PRIV]], i64 [[ARR_SIZE]]
327 // CHECK: [[ISEMPTY:%.+]] = icmp eq i32* [[ARR_PRIV]], [[END]]
328 // CHECK: br i1 [[ISEMPTY]],
329 // CHECK: phi i32*
330 // CHECK: store i32 888, i32* %
331 // CHECK: [[DONE:%.+]] = icmp eq i32* %{{.+}}, [[END]]
332 // CHECK: br i1 [[DONE]],
333
334 // CHECK: [[ARRS_PRIV:%.+]] = alloca [[S_FLOAT_TY]], i64 [[ARRS_SIZE:%.+]],
335
336 // Check initialization of private copy.
337 // CHECK: [[END:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[ARRS_PRIV]], i64 [[ARRS_SIZE]]
338 // CHECK: [[ISEMPTY:%.+]] = icmp eq [[S_FLOAT_TY]]* [[ARRS_PRIV]], [[END]]
339 // CHECK: br i1 [[ISEMPTY]],
340 // CHECK: phi [[S_FLOAT_TY]]*
341 // CHECK: call void @_Z4initR6BaseS1RKS_(%
342 // CHECK: [[DONE:%.+]] = icmp eq [[S_FLOAT_TY]]* %{{.+}}, [[END]]
343 // CHECK: br i1 [[DONE]],
344
345 // CHECK: [[GTID_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[GTID_ADDR_ADDR]]
346 // CHECK: [[GTID:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[GTID_REF]]
347 // CHECK: call void @__kmpc_for_static_init_4(
348 // Skip checks for internal operations.
349 // CHECK: call void @__kmpc_for_static_fini(
350
351 // void *RedList[<n>] = {<ReductionVars>[0], ..., <ReductionVars>[<n>-1]};
352
353 // CHECK: [[ARR_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 0
354 // CHECK: [[BITCAST:%.+]] = bitcast i32* [[ARR_PRIV]] to i8*
355 // CHECK: store i8* [[BITCAST]], i8** [[ARR_PRIV_REF]],
356 // CHECK: [[ARR_SIZE_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 1
357 // CHECK: [[BITCAST:%.+]] = inttoptr i64 [[ARR_SIZE]] to i8*
358 // CHECK: store i8* [[BITCAST]], i8** [[ARR_SIZE_REF]],
359 // CHECK: [[ARRS_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 2
360 // CHECK: [[BITCAST:%.+]] = bitcast [[S_FLOAT_TY]]* [[ARRS_PRIV]] to i8*
361 // CHECK: store i8* [[BITCAST]], i8** [[ARRS_PRIV_REF]],
362 // CHECK: [[ARRS_SIZE_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 3
363 // CHECK: [[BITCAST:%.+]] = inttoptr i64 [[ARRS_SIZE]] to i8*
364 // CHECK: store i8* [[BITCAST]], i8** [[ARRS_SIZE_REF]],
365
366 // res = __kmpc_reduce(<loc>, <gtid>, <n>, sizeof(RedList), RedList, reduce_func, &<lock>);
367
368 // CHECK: [[GTID_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[GTID_ADDR_ADDR]]
369 // CHECK: [[GTID:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[GTID_REF]]
370 // CHECK: [[BITCAST:%.+]] = bitcast [4 x i8*]* [[RED_LIST]] to i8*
371 // CHECK: [[RES:%.+]] = call i32 @__kmpc_reduce_nowait(%{{.+}}* [[REDUCTION_LOC]], i32 [[GTID]], i32 2, i64 32, i8* [[BITCAST]], void (i8*, i8*)* [[REDUCTION_FUNC:@.+]], [8 x i32]* [[REDUCTION_LOCK]])
372
373 // switch(res)
374 // CHECK: switch i32 [[RES]], label %[[RED_DONE:.+]] [
375 // CHECK: i32 1, label %[[CASE1:.+]]
376 // CHECK: i32 2, label %[[CASE2:.+]]
377 // CHECK: ]
378
379 // case 1:
380 // CHECK: [[CASE1]]
381
382 // arr[:] += arr_reduction[:];
383 // CHECK: [[END:%.+]] = getelementptr i32, i32* [[LB1_0]], i64 [[ARR_SIZE]]
384 // CHECK: [[ISEMPTY:%.+]] = icmp eq i32* [[LB1_0]], [[END]]
385 // CHECK: br i1 [[ISEMPTY]],
386 // CHECK: phi i32*
387 // CHECK: [[ADD:%.+]] = mul nsw i32 555, %
388 // CHECK: store i32 [[ADD]], i32* %
389 // CHECK: [[DONE:%.+]] = icmp eq i32* %{{.+}}, [[END]]
390 // CHECK: br i1 [[DONE]],
391
392 // arrs[:] = var.operator &(arrs_reduction[:]);
393 // CHECK: [[END:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[ARRS_LB:%.+]], i64 [[ARRS_SIZE]]
394 // CHECK: [[ISEMPTY:%.+]] = icmp eq [[S_FLOAT_TY]]* [[ARRS_LB]], [[END]]
395 // CHECK: br i1 [[ISEMPTY]],
396 // CHECK: phi [[S_FLOAT_TY]]*
397 // CHECK: call void @_Z3redR6BaseS1RKS_(%
398 // CHECK: [[DONE:%.+]] = icmp eq [[S_FLOAT_TY]]* %{{.+}}, [[END]]
399 // CHECK: br i1 [[DONE]],
400
401 // __kmpc_end_reduce(<loc>, <gtid>, &<lock>);
402 // CHECK: call void @__kmpc_end_reduce_nowait(%{{.+}}* [[REDUCTION_LOC]], i32 [[GTID]], [8 x i32]* [[REDUCTION_LOCK]])
403
404 // break;
405 // CHECK: br label %[[RED_DONE]]
406
407 // case 2:
408 // CHECK: [[CASE2]]
409
410 // arr[:] += arr_reduction[:];
411 // CHECK: [[END:%.+]] = getelementptr i32, i32* [[LB1_0]], i64 [[ARR_SIZE]]
412 // CHECK: [[ISEMPTY:%.+]] = icmp eq i32* [[LB1_0]], [[END]]
413 // CHECK: br i1 [[ISEMPTY]],
414 // CHECK: phi i32*
415 // CHECK: call void @__kmpc_critical(
416 // CHECK: [[ADD:%.+]] = mul nsw i32 555, %
417 // CHECK: call void @__kmpc_end_critical(
418 // CHECK: [[DONE:%.+]] = icmp eq i32* %{{.+}}, [[END]]
419 // CHECK: br i1 [[DONE]],
420
421 // arrs[:] = var.operator &(arrs_reduction[:]);
422 // CHECK: [[END:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[ARRS_LB:%.+]], i64 [[ARRS_SIZE]]
423 // CHECK: [[ISEMPTY:%.+]] = icmp eq [[S_FLOAT_TY]]* [[ARRS_LB]], [[END]]
424 // CHECK: br i1 [[ISEMPTY]],
425 // CHECK: phi [[S_FLOAT_TY]]*
426 // CHECK: call void @__kmpc_critical(
427 // CHECK: call void @_Z3redR6BaseS1RKS_(%
428 // CHECK: call void @__kmpc_end_critical(
429 // CHECK: [[DONE:%.+]] = icmp eq [[S_FLOAT_TY]]* %{{.+}}, [[END]]
430 // CHECK: br i1 [[DONE]],
431
432 // break;
433 // CHECK: br label %[[RED_DONE]]
434 // CHECK: [[RED_DONE]]
435
436 // Check destruction of private copy.
437 // CHECK: [[END:%.+]] = getelementptr inbounds [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[ARRS_PRIV]], i64 [[ARRS_SIZE]]
438 // CHECK: [[ISEMPTY:%.+]] = icmp eq [[S_FLOAT_TY]]* [[ARRS_PRIV]], [[END]]
439 // CHECK: br i1 [[ISEMPTY]],
440 // CHECK: phi [[S_FLOAT_TY]]*
441 // CHECK: call void @_ZN1SIfED1Ev([[S_FLOAT_TY]]* %
442 // CHECK: [[DONE:%.+]] = icmp eq [[S_FLOAT_TY]]* %{{.+}}, [[ARRS_PRIV]]
443 // CHECK: br i1 [[DONE]],
444 // CHECK: call void @llvm.stackrestore(i8*
445
446 // CHECK: ret void
447
448 // void reduce_func(void *lhs[<n>], void *rhs[<n>]) {
449 // *(Type0*)lhs[0] = ReductionOperation0(*(Type0*)lhs[0], *(Type0*)rhs[0]);
450 // ...
451 // *(Type<n>-1*)lhs[<n>-1] = ReductionOperation<n>-1(*(Type<n>-1*)lhs[<n>-1],
452 // *(Type<n>-1*)rhs[<n>-1]);
453 // }
454 // CHECK: define internal void [[REDUCTION_FUNC]](i8*, i8*)
455 // arr_rhs = (int*)rhs[0];
456 // CHECK: [[ARR_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS:%.+]], i64 0, i64 0
457 // CHECK: [[ARR_RHS_VOID:%.+]] = load i8*, i8** [[ARR_RHS_REF]],
458 // CHECK: [[ARR_RHS:%.+]] = bitcast i8* [[ARR_RHS_VOID]] to i32*
459 // arr_lhs = (int*)lhs[0];
460 // CHECK: [[ARR_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS:%.+]], i64 0, i64 0
461 // CHECK: [[ARR_LHS_VOID:%.+]] = load i8*, i8** [[ARR_LHS_REF]],
462 // CHECK: [[ARR_LHS:%.+]] = bitcast i8* [[ARR_LHS_VOID]] to i32*
463
464 // arr_size = (size_t)lhs[1];
465 // CHECK: [[ARR_SIZE_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 1
466 // CHECK: [[ARR_SIZE_VOID:%.+]] = load i8*, i8** [[ARR_SIZE_REF]],
467 // CHECK: [[ARR_SIZE:%.+]] = ptrtoint i8* [[ARR_SIZE_VOID]] to i64
468
469 // arrs_rhs = (S<float>*)rhs[2];
470 // CHECK: [[ARRS_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS]], i64 0, i64 2
471 // CHECK: [[ARRS_RHS_VOID:%.+]] = load i8*, i8** [[ARRS_RHS_REF]],
472 // CHECK: [[ARRS_RHS:%.+]] = bitcast i8* [[ARRS_RHS_VOID]] to [[S_FLOAT_TY]]*
473 // arrs_lhs = (S<float>*)lhs[2];
474 // CHECK: [[ARRS_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 2
475 // CHECK: [[ARRS_LHS_VOID:%.+]] = load i8*, i8** [[ARRS_LHS_REF]],
476 // CHECK: [[ARRS_LHS:%.+]] = bitcast i8* [[ARRS_LHS_VOID]] to [[S_FLOAT_TY]]*
477
478 // arrs_size = (size_t)lhs[3];
479 // CHECK: [[ARRS_SIZE_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 3
480 // CHECK: [[ARRS_SIZE_VOID:%.+]] = load i8*, i8** [[ARRS_SIZE_REF]],
481 // CHECK: [[ARRS_SIZE:%.+]] = ptrtoint i8* [[ARRS_SIZE_VOID]] to i64
482
483 // arr_lhs[:] += arr_rhs[:];
484 // CHECK: [[END:%.+]] = getelementptr i32, i32* [[ARR_LHS]], i64 [[ARR_SIZE]]
485 // CHECK: [[ISEMPTY:%.+]] = icmp eq i32* [[ARR_LHS]], [[END]]
486 // CHECK: br i1 [[ISEMPTY]],
487 // CHECK: phi i32*
488 // CHECK: [[ADD:%.+]] = mul nsw i32 555, %
489 // CHECK: [[DONE:%.+]] = icmp eq i32* %{{.+}}, [[END]]
490 // CHECK: br i1 [[DONE]],
491
492 // arrs_lhs = arrs_lhs.operator &(arrs_rhs);
493 // CHECK: [[END:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[ARRS_LB:%.+]], i64 [[ARRS_SIZE]]
494 // CHECK: [[ISEMPTY:%.+]] = icmp eq [[S_FLOAT_TY]]* [[ARRS_LB]], [[END]]
495 // CHECK: br i1 [[ISEMPTY]],
496 // CHECK: phi [[S_FLOAT_TY]]*
497 // CHECK: call void @_Z3redR6BaseS1RKS_(%
498 // CHECK: [[DONE:%.+]] = icmp eq [[S_FLOAT_TY]]* %{{.+}}, [[END]]
499 // CHECK: br i1 [[DONE]],
500
501 // CHECK: ret void
502
503 // CHECK: define internal void [[MAIN_MICROTASK2]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, i64 %{{.+}}, i64 %{{.+}}, i32* nonnull %{{.+}}, [10 x [4 x [[S_FLOAT_TY]]]]* dereferenceable(480) %{{.+}})
504
505 // CHECK: [[ARRS_PRIV:%.+]] = alloca [10 x [4 x [[S_FLOAT_TY]]]],
506
507 // Reduction list for runtime.
508 // CHECK: [[RED_LIST:%.+]] = alloca [3 x i8*],
509
510 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]],
511
512 // CHECK: [[ARR_SIZE:%.+]] = mul nuw i64 %{{.+}}, 4
513 // CHECK: call i8* @llvm.stacksave()
514 // CHECK: [[ARR_PRIV:%.+]] = alloca i32, i64 [[ARR_SIZE]],
515
516 // Check initialization of private copy.
517 // CHECK: [[END:%.+]] = getelementptr i32, i32* [[ARR_PRIV]], i64 [[ARR_SIZE]]
518 // CHECK: [[ISEMPTY:%.+]] = icmp eq i32* [[ARR_PRIV]], [[END]]
519 // CHECK: br i1 [[ISEMPTY]],
520 // CHECK: phi i32*
521 // CHECK: store i32 888, i32* %
522 // CHECK: [[DONE:%.+]] = icmp eq i32* %{{.+}}, [[END]]
523 // CHECK: br i1 [[DONE]],
524
525 // Check initialization of private copy.
526 // CHECK: [[LHS_BEGIN:%.+]] = bitcast [10 x [4 x [[S_FLOAT_TY]]]]* %{{.+}} to [[S_FLOAT_TY]]*
527 // CHECK: [[BEGIN:%.+]] = getelementptr inbounds [10 x [4 x [[S_FLOAT_TY]]]], [10 x [4 x [[S_FLOAT_TY]]]]* [[ARRS_PRIV]], i32 0, i32 0, i32 0
528 // CHECK: [[END:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[BEGIN]], i64 40
529 // CHECK: [[ISEMPTY:%.+]] = icmp eq [[S_FLOAT_TY]]* [[BEGIN]], [[END]]
530 // CHECK: br i1 [[ISEMPTY]],
531 // CHECK: phi [[S_FLOAT_TY]]*
532 // CHECK: call void @_Z4initR6BaseS1RKS_(%
533 // CHECK: [[DONE:%.+]] = icmp eq [[S_FLOAT_TY]]* %{{.+}}, [[END]]
534 // CHECK: br i1 [[DONE]],
535 // CHECK: [[ARRS_PRIV_BEGIN:%.+]] = bitcast [10 x [4 x [[S_FLOAT_TY]]]]* [[ARRS_PRIV]] to [[S_FLOAT_TY]]*
536
537 // CHECK: [[GTID_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[GTID_ADDR_ADDR]]
538 // CHECK: [[GTID:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[GTID_REF]]
539 // CHECK: call void @__kmpc_for_static_init_4(
540 // Skip checks for internal operations.
541 // CHECK: call void @__kmpc_for_static_fini(
542
543 // void *RedList[<n>] = {<ReductionVars>[0], ..., <ReductionVars>[<n>-1]};
544
545 // CHECK: [[ARR_PRIV_REF:%.+]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[RED_LIST]], i64 0, i64 0
546 // CHECK: [[BITCAST:%.+]] = bitcast i32* [[ARR_PRIV]] to i8*
547 // CHECK: store i8* [[BITCAST]], i8** [[ARR_PRIV_REF]],
548 // CHECK: [[ARR_SIZE_REF:%.+]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[RED_LIST]], i64 0, i64 1
549 // CHECK: [[BITCAST:%.+]] = inttoptr i64 [[ARR_SIZE]] to i8*
550 // CHECK: store i8* [[BITCAST]], i8** [[ARR_SIZE_REF]],
551 // CHECK: [[ARRS_PRIV_REF:%.+]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[RED_LIST]], i64 0, i64 2
552 // CHECK: [[BITCAST:%.+]] = bitcast [[S_FLOAT_TY]]* [[ARRS_PRIV_BEGIN]] to i8*
553 // CHECK: store i8* [[BITCAST]], i8** [[ARRS_PRIV_REF]],
554
555 // res = __kmpc_reduce(<loc>, <gtid>, <n>, sizeof(RedList), RedList, reduce_func, &<lock>);
556
557 // CHECK: [[GTID_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[GTID_ADDR_ADDR]]
558 // CHECK: [[GTID:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[GTID_REF]]
559 // CHECK: [[BITCAST:%.+]] = bitcast [3 x i8*]* [[RED_LIST]] to i8*
560 // CHECK: [[RES:%.+]] = call i32 @__kmpc_reduce(%{{.+}}* [[REDUCTION_LOC]], i32 [[GTID]], i32 2, i64 24, i8* [[BITCAST]], void (i8*, i8*)* [[REDUCTION_FUNC:@.+]], [8 x i32]* [[REDUCTION_LOCK]])
561
562 // switch(res)
563 // CHECK: switch i32 [[RES]], label %[[RED_DONE:.+]] [
564 // CHECK: i32 1, label %[[CASE1:.+]]
565 // CHECK: i32 2, label %[[CASE2:.+]]
566 // CHECK: ]
567
568 // case 1:
569 // CHECK: [[CASE1]]
570
571 // arr[:] += arr_reduction[:];
572 // CHECK: [[END:%.+]] = getelementptr i32, i32* [[LB1_0:%.+]], i64 [[ARR_SIZE]]
573 // CHECK: [[ISEMPTY:%.+]] = icmp eq i32* [[LB1_0]], [[END]]
574 // CHECK: br i1 [[ISEMPTY]],
575 // CHECK: phi i32*
576 // CHECK: [[ADD:%[^ ]+]] = mul nsw i32 555, %
577 // CHECK: store i32 [[ADD]], i32* %
578 // CHECK: [[DONE:%.+]] = icmp eq i32* %{{.+}}, [[END]]
579 // CHECK: br i1 [[DONE]],
580
581 // arrs[:] = var.operator &(arrs_reduction[:]);
582 // CHECK: [[END:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[LHS_BEGIN]], i64 40
583 // CHECK: [[ISEMPTY:%.+]] = icmp eq [[S_FLOAT_TY]]* [[LHS_BEGIN]], [[END]]
584 // CHECK: br i1 [[ISEMPTY]],
585 // CHECK: phi [[S_FLOAT_TY]]*
586 // CHECK: call void @_Z3redR6BaseS1RKS_(%
587 // CHECK: [[DONE:%.+]] = icmp eq [[S_FLOAT_TY]]* %{{.+}}, [[END]]
588 // CHECK: br i1 [[DONE]],
589
590 // __kmpc_end_reduce(<loc>, <gtid>, &<lock>);
591 // CHECK: call void @__kmpc_end_reduce(%{{.+}}* [[REDUCTION_LOC]], i32 [[GTID]], [8 x i32]* [[REDUCTION_LOCK]])
592
593 // break;
594 // CHECK: br label %[[RED_DONE]]
595
596 // case 2:
597 // CHECK: [[CASE2]]
598
599 // arr[:] += arr_reduction[:];
600 // CHECK: [[END:%.+]] = getelementptr i32, i32* [[LB1_0]], i64 [[ARR_SIZE]]
601 // CHECK: [[ISEMPTY:%.+]] = icmp eq i32* [[LB1_0]], [[END]]
602 // CHECK: br i1 [[ISEMPTY]],
603 // CHECK: phi i32*
604 // CHECK: call void @__kmpc_critical(
605 // CHECK: [[ADD:%.+]] = mul nsw i32 555, %
606 // CHECK: call void @__kmpc_end_critical(
607 // CHECK: [[DONE:%.+]] = icmp eq i32* %{{.+}}, [[END]]
608 // CHECK: br i1 [[DONE]],
609
610 // arrs[:] = var.operator &(arrs_reduction[:]);
611 // CHECK: [[END:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[LHS_BEGIN]], i64 40
612 // CHECK: [[ISEMPTY:%.+]] = icmp eq [[S_FLOAT_TY]]* [[LHS_BEGIN]], [[END]]
613 // CHECK: br i1 [[ISEMPTY]],
614 // CHECK: phi [[S_FLOAT_TY]]*
615 // CHECK: call void @__kmpc_critical(
616 // CHECK: call void @_Z3redR6BaseS1RKS_(%
617 // CHECK: call void @__kmpc_end_critical(
618 // CHECK: [[DONE:%.+]] = icmp eq [[S_FLOAT_TY]]* %{{.+}}, [[END]]
619 // CHECK: br i1 [[DONE]],
620
621 // break;
622 // CHECK: br label %[[RED_DONE]]
623 // CHECK: [[RED_DONE]]
624
625 // Check destruction of private copy.
626 // CHECK: [[BEGIN:%.+]] = getelementptr inbounds [10 x [4 x [[S_FLOAT_TY]]]], [10 x [4 x [[S_FLOAT_TY]]]]* [[ARRS_PRIV]], i32 0, i32 0, i32 0
627 // CHECK: [[END:%.+]] = getelementptr inbounds [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[BEGIN]], i64 40
628 // CHECK: br
629 // CHECK: phi [[S_FLOAT_TY]]*
630 // CHECK: call void @_ZN1SIfED1Ev([[S_FLOAT_TY]]* %
631 // CHECK: [[DONE:%.+]] = icmp eq [[S_FLOAT_TY]]* %{{.+}}, [[BEGIN]]
632 // CHECK: br i1 [[DONE]],
633 // CHECK: call void @llvm.stackrestore(i8*
634 // CHECK: call void @__kmpc_barrier(
635
636 // CHECK: ret void
637
638 // void reduce_func(void *lhs[<n>], void *rhs[<n>]) {
639 // *(Type0*)lhs[0] = ReductionOperation0(*(Type0*)lhs[0], *(Type0*)rhs[0]);
640 // ...
641 // *(Type<n>-1*)lhs[<n>-1] = ReductionOperation<n>-1(*(Type<n>-1*)lhs[<n>-1],
642 // *(Type<n>-1*)rhs[<n>-1]);
643 // }
644 // CHECK: define internal void [[REDUCTION_FUNC]](i8*, i8*)
645 // arr_rhs = (int*)rhs[0];
646 // CHECK: [[ARR_RHS_REF:%.+]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[RED_LIST_RHS:%.+]], i64 0, i64 0
647 // CHECK: [[ARR_RHS_VOID:%.+]] = load i8*, i8** [[ARR_RHS_REF]],
648 // CHECK: [[ARR_RHS:%.+]] = bitcast i8* [[ARR_RHS_VOID]] to i32*
649 // arr_lhs = (int*)lhs[0];
650 // CHECK: [[ARR_LHS_REF:%.+]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[RED_LIST_LHS:%.+]], i64 0, i64 0
651 // CHECK: [[ARR_LHS_VOID:%.+]] = load i8*, i8** [[ARR_LHS_REF]],
652 // CHECK: [[ARR_LHS:%.+]] = bitcast i8* [[ARR_LHS_VOID]] to i32*
653
654 // arr_size = (size_t)lhs[1];
655 // CHECK: [[ARR_SIZE_REF:%.+]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[RED_LIST_LHS]], i64 0, i64 1
656 // CHECK: [[ARR_SIZE_VOID:%.+]] = load i8*, i8** [[ARR_SIZE_REF]],
657 // CHECK: [[ARR_SIZE:%.+]] = ptrtoint i8* [[ARR_SIZE_VOID]] to i64
658
659 // arrs_rhs = (S<float>*)rhs[2];
660 // CHECK: [[ARRS_RHS_REF:%.+]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[RED_LIST_RHS]], i64 0, i64 2
661 // CHECK: [[ARRS_RHS_VOID:%.+]] = load i8*, i8** [[ARRS_RHS_REF]],
662 // CHECK: [[ARRS_RHS:%.+]] = bitcast i8* [[ARRS_RHS_VOID]] to [[S_FLOAT_TY]]*
663 // arrs_lhs = (S<float>*)lhs[2];
664 // CHECK: [[ARRS_LHS_REF:%.+]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[RED_LIST_LHS]], i64 0, i64 2
665 // CHECK: [[ARRS_LHS_VOID:%.+]] = load i8*, i8** [[ARRS_LHS_REF]],
666 // CHECK: [[ARRS_LHS:%.+]] = bitcast i8* [[ARRS_LHS_VOID]] to [[S_FLOAT_TY]]*
667
668 // arr_lhs[:] += arr_rhs[:];
669 // CHECK: [[END:%.+]] = getelementptr i32, i32* [[ARR_LHS]], i64 [[ARR_SIZE]]
670 // CHECK: [[ISEMPTY:%.+]] = icmp eq i32* [[ARR_LHS]], [[END]]
671 // CHECK: br i1 [[ISEMPTY]],
672 // CHECK: phi i32*
673 // CHECK: [[ADD:%.+]] = mul nsw i32 555, %
674 // CHECK: store i32 [[ADD]], i32* %
675 // CHECK: [[DONE:%.+]] = icmp eq i32* %{{.+}}, [[END]]
676 // CHECK: br i1 [[DONE]],
677
678 // arrs_lhs = arrs_lhs.operator &(arrs_rhs);
679 // CHECK: [[END:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[ARRS_LB:%.+]], i64 40
680 // CHECK: [[ISEMPTY:%.+]] = icmp eq [[S_FLOAT_TY]]* [[ARRS_LB]], [[END]]
681 // CHECK: br i1 [[ISEMPTY]],
682 // CHECK: phi [[S_FLOAT_TY]]*
683 // CHECK: call void @_Z3redR6BaseS1RKS_(%
684 // CHECK: [[DONE:%.+]] = icmp eq [[S_FLOAT_TY]]* %{{.+}}, [[END]]
685 // CHECK: br i1 [[DONE]],
686
687 // CHECK: ret void
688
689 // CHECK: define internal void [[MAIN_MICROTASK3]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, [[S_FLOAT_TY]]*** dereferenceable(8) %{{.+}})
690
691 // CHECK: [[VAR2_ORIG_ADDR:%.+]] = alloca [[S_FLOAT_TY]]***,
692
693 // Reduction list for runtime.
694 // CHECK: [[RED_LIST:%.+]] = alloca [2 x i8*],
695
696 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]],
697 // CHECK: [[VAR2_ORIG:%.+]] = load [[S_FLOAT_TY]]***, [[S_FLOAT_TY]]**** [[VAR2_ORIG_ADDR]],
698
699 // CHECK: load [[S_FLOAT_TY]]**, [[S_FLOAT_TY]]*** [[VAR2_ORIG]],
700 // CHECK: getelementptr inbounds [[S_FLOAT_TY]]*, [[S_FLOAT_TY]]** %{{.+}}, i64 0
701 // CHECK: load [[S_FLOAT_TY]]*, [[S_FLOAT_TY]]** %
702 // CHECK: [[LOW:%.+]] = getelementptr inbounds [[S_FLOAT_TY]], [[S_FLOAT_TY]]* %{{.+}}, i64 1
703 // CHECK: load [[S_FLOAT_TY]]**, [[S_FLOAT_TY]]*** [[VAR2_ORIG]],
704 // CHECK: getelementptr inbounds [[S_FLOAT_TY]]*, [[S_FLOAT_TY]]** %{{.+}}, i64 4
705 // CHECK: load [[S_FLOAT_TY]]*, [[S_FLOAT_TY]]** %
706 // CHECK: getelementptr inbounds [[S_FLOAT_TY]], [[S_FLOAT_TY]]* %{{.+}}, i64 6
707 // CHECK: [[LD:%.+]] = load [[S_FLOAT_TY]]**, [[S_FLOAT_TY]]*** [[VAR2_ORIG]],
708 // CHECK: [[ORIG_START:%.+]] = load [[S_FLOAT_TY]]*, [[S_FLOAT_TY]]** [[LD]],
709 // CHECK: [[LAST:%.+]] = ptrtoint [[S_FLOAT_TY]]* %{{.+}} to i64
710 // CHECK: [[FIRST:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[LOW]] to i64
711 // CHECK: [[BYTE_DIF:%.+]] = sub i64 [[LAST]], [[FIRST]]
712 // CHECK: [[DIF:%.+]] = sdiv exact i64 [[BYTE_DIF]], ptrtoint ([[S_FLOAT_TY]]* getelementptr ([[S_FLOAT_TY]], [[S_FLOAT_TY]]* null, i32 1) to i64)
713 // CHECK: [[SIZE:%.+]] = add nuw i64 [[DIF]], 1
714 // CHECK: call i8* @llvm.stacksave()
715 // CHECK: [[VAR2_PRIV:%.+]] = alloca [[S_FLOAT_TY]], i64 [[SIZE]],
716 // CHECK: [[START:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[ORIG_START]] to i64
717 // CHECK: [[LOW_BOUND:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[LOW]] to i64
718 // CHECK: [[OFFSET_BYTES:%.+]] = sub i64 [[START]], [[LOW_BOUND]]
719 // CHECK: [[OFFSET:%.+]] = sdiv exact i64 [[OFFSET_BYTES]], ptrtoint ([[S_FLOAT_TY]]* getelementptr ([[S_FLOAT_TY]], [[S_FLOAT_TY]]* null, i32 1) to i64)
720 // CHECK: [[PSEUDO_VAR2_PRIV:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[VAR2_PRIV]], i64 [[OFFSET]]
721 // CHECK: store [[S_FLOAT_TY]]** [[REF:.+]], [[S_FLOAT_TY]]*** %
722 // CHECK: store [[S_FLOAT_TY]]* [[PSEUDO_VAR2_PRIV]], [[S_FLOAT_TY]]** [[REF]]
723 // CHECK: ret void
724
725 // CHECK: define internal void [[MAIN_MICROTASK4]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, [2 x [[S_FLOAT_TY]]]* dereferenceable(24) %{{.+}})
726
727 // CHECK: [[VVAR2_ORIG_ADDR:%.+]] = alloca [2 x [[S_FLOAT_TY]]]*,
728
729 // Reduction list for runtime.
730 // CHECK: [[RED_LIST:%.+]] = alloca [2 x i8*],
731
732 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]],
733 // CHECK: [[VVAR2_ORIG:%.+]] = load [2 x [[S_FLOAT_TY]]]*, [2 x [[S_FLOAT_TY]]]** [[VVAR2_ORIG_ADDR]],
734
735 // CHECK: [[LOW:%.+]] = getelementptr inbounds [2 x [[S_FLOAT_TY]]], [2 x [[S_FLOAT_TY]]]* [[VVAR2_ORIG]], i64 0, i64 0
736 // CHECK: getelementptr inbounds [2 x [[S_FLOAT_TY]]], [2 x [[S_FLOAT_TY]]]* [[VVAR2_ORIG]], i64 0, i64 4
737 // CHECK: [[ORIG_START:%.+]] = bitcast [2 x [[S_FLOAT_TY]]]* [[VVAR2_ORIG]] to [[S_FLOAT_TY]]*
738 // CHECK: [[LAST:%.+]] = ptrtoint [[S_FLOAT_TY]]* %{{.+}} to i64
739 // CHECK: [[FIRST:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[LOW]] to i64
740 // CHECK: [[BYTE_DIF:%.+]] = sub i64 [[LAST]], [[FIRST]]
741 // CHECK: [[DIF:%.+]] = sdiv exact i64 [[BYTE_DIF]], ptrtoint ([[S_FLOAT_TY]]* getelementptr ([[S_FLOAT_TY]], [[S_FLOAT_TY]]* null, i32 1) to i64)
742 // CHECK: [[SIZE:%.+]] = add nuw i64 [[DIF]], 1
743 // CHECK: call i8* @llvm.stacksave()
744 // CHECK: [[VVAR2_PRIV:%.+]] = alloca [[S_FLOAT_TY]], i64 [[SIZE]],
745 // CHECK: [[START:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[ORIG_START]] to i64
746 // CHECK: [[LOW_BOUND:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[LOW]] to i64
747 // CHECK: [[OFFSET_BYTES:%.+]] = sub i64 [[START]], [[LOW_BOUND]]
748 // CHECK: [[OFFSET:%.+]] = sdiv exact i64 [[OFFSET_BYTES]], ptrtoint ([[S_FLOAT_TY]]* getelementptr ([[S_FLOAT_TY]], [[S_FLOAT_TY]]* null, i32 1) to i64)
749 // CHECK: [[PSEUDO_VVAR2_PRIV:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[VVAR2_PRIV]], i64 [[OFFSET]]
750 // CHECK: [[VVAR2_PRIV:%.+]] = bitcast [[S_FLOAT_TY]]* [[PSEUDO_VVAR2_PRIV]] to [2 x [[S_FLOAT_TY]]]*
751 // CHECK: ret void
752
753 // CHECK: define internal void [[MAIN_MICROTASK5]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, [2 x [[S_FLOAT_TY]]]* dereferenceable(24) %{{.+}})
754
755 // CHECK: [[VAR3_ORIG_ADDR:%.+]] = alloca [2 x [[S_FLOAT_TY]]]*,
756
757 // Reduction list for runtime.
758 // CHECK: [[RED_LIST:%.+]] = alloca [2 x i8*],
759
760 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]],
761
762 // CHECK: [[VAR3_ORIG:%.+]] = load [2 x [[S_FLOAT_TY]]]*, [2 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR]],
763 // CHECK: store [2 x [[S_FLOAT_TY]]]* [[VAR3_ORIG]], [2 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR:%.+]],
764 // CHECK: [[VAR3_ORIG:%.+]] = load [2 x [[S_FLOAT_TY]]]*, [2 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR]],
765 // CHECK: [[LOW:%.+]] = getelementptr inbounds [2 x [[S_FLOAT_TY]]], [2 x [[S_FLOAT_TY]]]* [[VAR3_ORIG]], i64 0, i64 1
766 // CHECK: [[VAR3_ORIG:%.+]] = load [2 x [[S_FLOAT_TY]]]*, [2 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR]],
767 // CHECK: getelementptr inbounds [2 x [[S_FLOAT_TY]]], [2 x [[S_FLOAT_TY]]]* [[VAR3_ORIG]], i64 0, i64 2
768 // CHECK: [[VAR3_ORIG:%.+]] = load [2 x [[S_FLOAT_TY]]]*, [2 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR]],
769 // CHECK: [[ORIG_START:%.+]] = bitcast [2 x [[S_FLOAT_TY]]]* [[VAR3_ORIG]] to [[S_FLOAT_TY]]*
770 // CHECK: [[LAST:%.+]] = ptrtoint [[S_FLOAT_TY]]* %{{.+}} to i64
771 // CHECK: [[FIRST:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[LOW]] to i64
772 // CHECK: [[BYTE_DIF:%.+]] = sub i64 [[LAST]], [[FIRST]]
773 // CHECK: [[DIF:%.+]] = sdiv exact i64 [[BYTE_DIF]], ptrtoint ([[S_FLOAT_TY]]* getelementptr ([[S_FLOAT_TY]], [[S_FLOAT_TY]]* null, i32 1) to i64)
774 // CHECK: [[SIZE:%.+]] = add nuw i64 [[DIF]], 1
775 // CHECK: call i8* @llvm.stacksave()
776 // CHECK: [[VAR3_PRIV:%.+]] = alloca [[S_FLOAT_TY]], i64 [[SIZE]],
777 // CHECK: [[START:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[ORIG_START]] to i64
778 // CHECK: [[LOW_BOUND:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[LOW]] to i64
779 // CHECK: [[OFFSET_BYTES:%.+]] = sub i64 [[START]], [[LOW_BOUND]]
780 // CHECK: [[OFFSET:%.+]] = sdiv exact i64 [[OFFSET_BYTES]], ptrtoint ([[S_FLOAT_TY]]* getelementptr ([[S_FLOAT_TY]], [[S_FLOAT_TY]]* null, i32 1) to i64)
781 // CHECK: [[PSEUDO_VAR3_PRIV:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[VAR3_PRIV]], i64 [[OFFSET]]
782 // CHECK: [[VAR3_PRIV:%.+]] = bitcast [[S_FLOAT_TY]]* [[PSEUDO_VAR3_PRIV]] to [2 x [[S_FLOAT_TY]]]*
783
784 // CHECK: store [2 x [[S_FLOAT_TY]]]* [[VAR3_PRIV]], [2 x [[S_FLOAT_TY]]]** %
785
786 // CHECK: ret void
787
788 // CHECK: define internal void [[MAIN_MICROTASK6]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, [2 x [[S_FLOAT_TY]]]* dereferenceable(24) %{{.+}})
789
790 // CHECK: [[VAR3_ORIG_ADDR:%.+]] = alloca [2 x [[S_FLOAT_TY]]]*,
791 // CHECK: [[VAR3_PRIV:%.+]] = alloca [2 x [[S_FLOAT_TY]]],
792
793 // Reduction list for runtime.
794 // CHECK: [[RED_LIST:%.+]] = alloca [1 x i8*],
795
796 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]],
797
798 // CHECK: [[VAR3_ORIG:%.+]] = load [2 x [[S_FLOAT_TY]]]*, [2 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR]],
799 // CHECK: store [2 x [[S_FLOAT_TY]]]* [[VAR3_ORIG]], [2 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR:%.+]],
800 // CHECK: [[VAR3_ORIG:%.+]] = load [2 x [[S_FLOAT_TY]]]*, [2 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR]],
801 // CHECK: bitcast [2 x [[S_FLOAT_TY]]]* [[VAR3_ORIG]] to [[S_FLOAT_TY]]*
802 // CHECK: getelementptr inbounds [2 x [[S_FLOAT_TY]]], [2 x [[S_FLOAT_TY]]]* [[VAR3_PRIV]], i32 0, i32 0
803 // CHECK: getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* %{{.+}}, i64 2
804
805 // CHECK: store [2 x [[S_FLOAT_TY]]]* [[VAR3_PRIV]], [2 x [[S_FLOAT_TY]]]** %
806
807 // CHECK: ret void
808
809 // CHECK: define {{.*}} i{{[0-9]+}} [[TMAIN_INT]]()
810 // CHECK: [[TEST:%.+]] = alloca [[S_INT_TY]],
811 // CHECK: call {{.*}} [[S_INT_TY_CONSTR:@.+]]([[S_INT_TY]]* [[TEST]])
812 // CHECK: call void (%{{.+}}*, i{{[0-9]+}}, void (i{{[0-9]+}}*, i{{[0-9]+}}*, ...)*, ...) @__kmpc_fork_call(%{{.+}}* @{{.+}}, i{{[0-9]+}} 6, void (i{{[0-9]+}}*, i{{[0-9]+}}*, ...)* bitcast (void (i{{[0-9]+}}*, i{{[0-9]+}}*, i32*, [[S_INT_TY]]*, [[S_INT_TY]]*, i32*, [2 x i32]*, [2 x [[S_INT_TY]]]*)* [[TMAIN_MICROTASK:@.+]] to void
813 // CHECK: call {{.*}} [[S_INT_TY_DESTR:@.+]]([[S_INT_TY]]*
814 // CHECK: ret
815 //
816 // CHECK: define internal void [[TMAIN_MICROTASK]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, i32* dereferenceable(4) %{{.+}}, [[S_INT_TY]]* dereferenceable(12) %{{.+}}, [[S_INT_TY]]* dereferenceable(12) %{{.+}}, i32* dereferenceable(4) %{{.+}}, [2 x i32]* dereferenceable(8) %{{.+}}, [2 x [[S_INT_TY]]]* dereferenceable(24) %{{.+}})
817 // CHECK: alloca i{{[0-9]+}},
818 // CHECK: alloca i{{[0-9]+}},
819 // CHECK: alloca i{{[0-9]+}},
820 // CHECK: alloca i{{[0-9]+}},
821 // CHECK: alloca i{{[0-9]+}},
822 // CHECK: [[T_VAR_PRIV:%.+]] = alloca i{{[0-9]+}},
823 // CHECK: [[VAR_PRIV:%.+]] = alloca [[S_INT_TY]],
824 // CHECK: [[VAR1_PRIV:%.+]] = alloca [[S_INT_TY]],
825 // CHECK: [[T_VAR1_PRIV:%.+]] = alloca i{{[0-9]+}},
826
827 // Reduction list for runtime.
828 // CHECK: [[RED_LIST:%.+]] = alloca [4 x i8*],
829
830 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]],
831
832 // CHECK: [[T_VAR_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** %
833 // CHECK: [[VAR1_REF:%.+]] = load [[S_INT_TY]]*, [[S_INT_TY]]** %
834 // CHECK: [[T_VAR1_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** %
835
836 // For + reduction operation initial value of private variable is 0.
837 // CHECK: store i32 321, i32* %
838
839 // For & reduction operation initial value of private variable is ones in all bits.
840 // CHECK: call void @_Z4initR6BaseS1RKS_(
841
842 // For && reduction operation initial value of private variable is 1.0.
843 // CHECK: call void @_Z5init2R6BaseS1RKS_(
844
845 // For min reduction operation initial value of private variable is largest repesentable value.
846 // CHECK: sdiv i32 432, %
847
848 // CHECK: [[GTID_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[GTID_ADDR_ADDR]]
849 // CHECK: [[GTID:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[GTID_REF]]
850 // CHECK: call void @__kmpc_for_static_init_4(
851 // Skip checks for internal operations.
852 // CHECK: call void @__kmpc_for_static_fini(
853
854 // void *RedList[<n>] = {<ReductionVars>[0], ..., <ReductionVars>[<n>-1]};
855
856 // CHECK: [[T_VAR_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 0
857 // CHECK: [[BITCAST:%.+]] = bitcast i{{[0-9]+}}* [[T_VAR_PRIV]] to i8*
858 // CHECK: store i8* [[BITCAST]], i8** [[T_VAR_PRIV_REF]],
859 // CHECK: [[VAR_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 1
860 // CHECK: [[BITCAST:%.+]] = bitcast [[S_INT_TY]]* [[VAR_PRIV]] to i8*
861 // CHECK: store i8* [[BITCAST]], i8** [[VAR_PRIV_REF]],
862 // CHECK: [[VAR1_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 2
863 // CHECK: [[BITCAST:%.+]] = bitcast [[S_INT_TY]]* [[VAR1_PRIV]] to i8*
864 // CHECK: store i8* [[BITCAST]], i8** [[VAR1_PRIV_REF]],
865 // CHECK: [[T_VAR1_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 3
866 // CHECK: [[BITCAST:%.+]] = bitcast i{{[0-9]+}}* [[T_VAR1_PRIV]] to i8*
867 // CHECK: store i8* [[BITCAST]], i8** [[T_VAR1_PRIV_REF]],
868
869 // res = __kmpc_reduce_nowait(<loc>, <gtid>, <n>, sizeof(RedList), RedList, reduce_func, &<lock>);
870
871 // CHECK: [[BITCAST:%.+]] = bitcast [4 x i8*]* [[RED_LIST]] to i8*
872 // CHECK: [[RES:%.+]] = call i32 @__kmpc_reduce_nowait(%{{.+}}* [[REDUCTION_LOC]], i32 [[GTID]], i32 4, i64 32, i8* [[BITCAST]], void (i8*, i8*)* [[REDUCTION_FUNC:@.+]], [8 x i32]* [[REDUCTION_LOCK]])
873
874 // switch(res)
875 // CHECK: switch i32 [[RES]], label %[[RED_DONE:.+]] [
876 // CHECK: i32 1, label %[[CASE1:.+]]
877 // CHECK: i32 2, label %[[CASE2:.+]]
878 // CHECK: ]
879
880 // case 1:
881 // t_var += t_var_reduction;
882 // CHECK: add nsw i32 1513, %
883
884 // var = var.operator &(var_reduction);
885 // CHECK: call void @_Z3redR6BaseS1RKS_(%
886
887 // var1 = var1.operator &&(var1_reduction);
888 // CHECK: mul nsw i32 17, %
889
890 // t_var1 = min(t_var1, t_var1_reduction);
891 // CHECK: sub nsw i32 47, %
892
893 // __kmpc_end_reduce_nowait(<loc>, <gtid>, &<lock>);
894 // CHECK: call void @__kmpc_end_reduce_nowait(%{{.+}}* [[REDUCTION_LOC]], i32 [[GTID]], [8 x i32]* [[REDUCTION_LOCK]])
895
896 // break;
897 // CHECK: br label %[[RED_DONE]]
898
899 // case 2:
900 // t_var += t_var_reduction;
901 // CHECK: call void @__kmpc_critical(
902 // CHECK: add nsw i32 1513, %
903 // CHECK: call void @__kmpc_end_critical(
904
905 // var = var.operator &(var_reduction);
906 // CHECK: call void @__kmpc_critical(
907 // CHECK: call void @_Z3redR6BaseS1RKS_(%
908 // CHECK: call void @__kmpc_end_critical(
909
910 // var1 = var1.operator &&(var1_reduction);
911 // CHECK: call void @__kmpc_critical(
912 // CHECK: mul nsw i32 17, %
913 // CHECK: call void @__kmpc_end_critical(
914
915 // t_var1 = min(t_var1, t_var1_reduction);
916 // CHECK: call void @__kmpc_critical(
917 // CHECK: sub nsw i32 47, %
918 // CHECK: call void @__kmpc_end_critical(
919
920 // break;
921 // CHECK: br label %[[RED_DONE]]
922 // CHECK: [[RED_DONE]]
923 // CHECK-DAG: call {{.*}} [[S_INT_TY_DESTR]]([[S_INT_TY]]* [[VAR_PRIV]])
924 // CHECK-DAG: call {{.*}} [[S_INT_TY_DESTR]]([[S_INT_TY]]*
925 // CHECK: ret void
926
927 // void reduce_func(void *lhs[<n>], void *rhs[<n>]) {
928 // *(Type0*)lhs[0] = ReductionOperation0(*(Type0*)lhs[0], *(Type0*)rhs[0]);
929 // ...
930 // *(Type<n>-1*)lhs[<n>-1] = ReductionOperation<n>-1(*(Type<n>-1*)lhs[<n>-1],
931 // *(Type<n>-1*)rhs[<n>-1]);
932 // }
933 // CHECK: define internal void [[REDUCTION_FUNC]](i8*, i8*)
934 // t_var_lhs = (i{{[0-9]+}}*)lhs[0];
935 // CHECK: [[T_VAR_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS:%.+]], i64 0, i64 0
936 // CHECK: [[T_VAR_RHS_VOID:%.+]] = load i8*, i8** [[T_VAR_RHS_REF]],
937 // CHECK: [[T_VAR_RHS:%.+]] = bitcast i8* [[T_VAR_RHS_VOID]] to i{{[0-9]+}}*
938 // t_var_rhs = (i{{[0-9]+}}*)rhs[0];
939 // CHECK: [[T_VAR_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS:%.+]], i64 0, i64 0
940 // CHECK: [[T_VAR_LHS_VOID:%.+]] = load i8*, i8** [[T_VAR_LHS_REF]],
941 // CHECK: [[T_VAR_LHS:%.+]] = bitcast i8* [[T_VAR_LHS_VOID]] to i{{[0-9]+}}*
942
943 // var_lhs = (S<i{{[0-9]+}}>*)lhs[1];
944 // CHECK: [[VAR_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS]], i64 0, i64 1
945 // CHECK: [[VAR_RHS_VOID:%.+]] = load i8*, i8** [[VAR_RHS_REF]],
946 // CHECK: [[VAR_RHS:%.+]] = bitcast i8* [[VAR_RHS_VOID]] to [[S_INT_TY]]*
947 // var_rhs = (S<i{{[0-9]+}}>*)rhs[1];
948 // CHECK: [[VAR_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 1
949 // CHECK: [[VAR_LHS_VOID:%.+]] = load i8*, i8** [[VAR_LHS_REF]],
950 // CHECK: [[VAR_LHS:%.+]] = bitcast i8* [[VAR_LHS_VOID]] to [[S_INT_TY]]*
951
952 // var1_lhs = (S<i{{[0-9]+}}>*)lhs[2];
953 // CHECK: [[VAR1_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS]], i64 0, i64 2
954 // CHECK: [[VAR1_RHS_VOID:%.+]] = load i8*, i8** [[VAR1_RHS_REF]],
955 // CHECK: [[VAR1_RHS:%.+]] = bitcast i8* [[VAR1_RHS_VOID]] to [[S_INT_TY]]*
956 // var1_rhs = (S<i{{[0-9]+}}>*)rhs[2];
957 // CHECK: [[VAR1_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 2
958 // CHECK: [[VAR1_LHS_VOID:%.+]] = load i8*, i8** [[VAR1_LHS_REF]],
959 // CHECK: [[VAR1_LHS:%.+]] = bitcast i8* [[VAR1_LHS_VOID]] to [[S_INT_TY]]*
960
961 // t_var1_lhs = (i{{[0-9]+}}*)lhs[3];
962 // CHECK: [[T_VAR1_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS]], i64 0, i64 3
963 // CHECK: [[T_VAR1_RHS_VOID:%.+]] = load i8*, i8** [[T_VAR1_RHS_REF]],
964 // CHECK: [[T_VAR1_RHS:%.+]] = bitcast i8* [[T_VAR1_RHS_VOID]] to i{{[0-9]+}}*
965 // t_var1_rhs = (i{{[0-9]+}}*)rhs[3];
966 // CHECK: [[T_VAR1_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 3
967 // CHECK: [[T_VAR1_LHS_VOID:%.+]] = load i8*, i8** [[T_VAR1_LHS_REF]],
968 // CHECK: [[T_VAR1_LHS:%.+]] = bitcast i8* [[T_VAR1_LHS_VOID]] to i{{[0-9]+}}*
969
970 // t_var_lhs += t_var_rhs;
971 // CHECK: add nsw i32 1513, %
972
973 // var_lhs = var_lhs.operator &(var_rhs);
974 // CHECK: call void @_Z3redR6BaseS1RKS_(%
975
976 // var1_lhs = var1_lhs.operator &&(var1_rhs);
977 // CHECK: mul nsw i32 17, %
978
979 // t_var1_lhs = min(t_var1_lhs, t_var1_rhs);
980 // CHECK: sub nsw i32 47, %
981 // CHECK: ret void
982
983 #endif
984
985