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
5 // RUN: %clang_cc1 -verify -fopenmp-simd -x c++ -triple x86_64-apple-darwin10 -emit-llvm %s -o - | FileCheck --check-prefix SIMD-ONLY0 %s
6 // RUN: %clang_cc1 -fopenmp-simd -x c++ -std=c++11 -triple x86_64-apple-darwin10 -emit-pch -o %t %s
7 // RUN: %clang_cc1 -fopenmp-simd -x c++ -triple x86_64-apple-darwin10 -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck --check-prefix SIMD-ONLY0 %s
8 // SIMD-ONLY0-NOT: {{__kmpc|__tgt}}
9 // expected-no-diagnostics
10 #ifndef HEADER
11 #define HEADER
12
13 enum omp_allocator_handle_t {
14 omp_null_allocator = 0,
15 omp_default_mem_alloc = 1,
16 omp_large_cap_mem_alloc = 2,
17 omp_const_mem_alloc = 3,
18 omp_high_bw_mem_alloc = 4,
19 omp_low_lat_mem_alloc = 5,
20 omp_cgroup_mem_alloc = 6,
21 omp_pteam_mem_alloc = 7,
22 omp_thread_mem_alloc = 8,
23 KMP_ALLOCATOR_MAX_HANDLE = __UINTPTR_MAX__
24 };
25
26 volatile double g, g_orig;
27 volatile double &g1 = g_orig;
28
29 struct BaseS {
30 int x;
31 };
32 struct BaseS1 {
33 float y;
34 };
35
36 template <class T>
37 struct S : public BaseS, public BaseS1 {
38 T f;
SS39 S(T a) : f(a + g) {}
SS40 S() : f(g) {}
41 S& operator=(const S&);
~SS42 ~S() {}
43 };
44 void red(BaseS1&, const BaseS1&);
45 void red_plus(BaseS1&, const BaseS1&);
46 void init(BaseS1&, const BaseS1&);
47 void init1(BaseS1&, const BaseS1&);
48 void init2(BaseS1&, const BaseS1&);
49 void init_plus(BaseS1&, const BaseS1&);
50 #pragma omp declare reduction(operator& : BaseS1 : red(omp_out, omp_in)) initializer(init(omp_priv, omp_orig))
51 #pragma omp declare reduction(+ : BaseS1 : red_plus(omp_out, omp_in)) initializer(init_plus(omp_priv, omp_orig))
52 #pragma omp declare reduction(&& : S<float>, S<int> : omp_out.f *= omp_in.f) initializer(init1(omp_priv, omp_orig))
53
54 // CHECK-DAG: [[S_FLOAT_TY:%.+]] = type { %{{[^,]+}}, %{{[^,]+}}, float }
55 // CHECK-DAG: [[S_INT_TY:%.+]] = type { %{{[^,]+}}, %{{[^,]+}}, i{{[0-9]+}} }
56 // CHECK-DAG: [[IMPLICIT_BARRIER_LOC:@.+]] = private unnamed_addr global %struct.ident_t { i32 0, i32 66, i32 0, i32 0, i8*
57 // CHECK-DAG: [[REDUCTION_LOC:@.+]] = private unnamed_addr global %struct.ident_t { i32 0, i32 18, i32 0, i32 0, i8*
58 // CHECK-DAG: [[REDUCTION_LOCK:@.+]] = common global [8 x i32] zeroinitializer
59
60 #pragma omp declare reduction(operator* : S<int> : omp_out.f = 17 * omp_in.f) initializer(omp_priv = S<int>())
61 // CHECK-LABEL: bazz
bazz()62 void bazz() {
63 S<int> s;
64 // CHECK: [[S_ADDR:%.+]] = alloca [[S_INT_TY]],
65 // CHECK: call {{.*}} [[S_INT_TY_CONSTR:@.+]]([[S_INT_TY]]* [[S_ADDR]])
66 // CHECK: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @{{.+}}, i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [[S_INT_TY]]*)* [[BAZZ_OUTLINE:@.+]] to void (i32*, i32*, ...)*), [[S_INT_TY]]* [[S_ADDR]])
67 // CHECK: call {{.*}} [[S_INT_TY_DESTR:@.+]]([[S_INT_TY]]*
68 // CHECK: ret void
69 #pragma omp parallel
70 #pragma omp simd reduction(*: s)
71 for (int I = 0; I < 10; ++I)
72 ;
73 }
74
75 // CHECK: define internal void [[BAZZ_OUTLINE]](i32* {{.+}}, i32* {{.+}}, [[S_INT_TY]]* {{.+}})
76 // CHECK: [[S_PRIV_ADDR:%.+]] = alloca [[S_INT_TY]],
77 // CHECK: call void [[BAZZ_INIT:@.+]]([[S_INT_TY]]* [[S_PRIV_ADDR]], [[S_INT_TY]]* [[S_ORIG_ADDR:%.+]])
78 // CHECK: call void @{{.+}}([[S_INT_TY]]* [[S_ORIG_ADDR]], [[S_INT_TY]]* [[S_PRIV_ADDR]])
79 // CHECK-NEXT: call void [[S_INT_TY_DESTR]]([[S_INT_TY]]* [[S_PRIV_ADDR]])
80 // CHECK-NEXT: ret void
81
82 // CHECK: define internal void [[BAZZ_INIT]]([[S_INT_TY]]* {{.*}}[[S_PRIV_ADDR:%.+]], [[S_INT_TY]]* {{.*}}[[S_ORIG_ADDR:%.+]])
83 // CHECK: store [[S_INT_TY]]* [[S_PRIV_ADDR]], [[S_INT_TY]]** [[S_PRIV_ADDR_REF:%.+]],
84 // CHECK: [[S_PRIV_ADDR:%.+]] = load [[S_INT_TY]]*, [[S_INT_TY]]** [[S_PRIV_ADDR_REF]],
85 // CHECK: call void [[S_INT_TY_CONSTR]]([[S_INT_TY]]* [[S_PRIV_ADDR]])
86 // CHECK-NEXT: ret void
87
88 #pragma omp declare reduction(operator&& : int : omp_out = 111 & omp_in)
89 template <typename T, int length>
90 T tmain() {
91 T t;
92 S<T> test;
93 T t_var = T(), t_var1;
94 T vec[] = {1, 2};
95 S<T> s_arr[] = {1, 2};
96 S<T> &var = test;
97 S<T> var1;
98 S<T> arr[length];
99 #pragma omp declare reduction(operator& : T : omp_out = 15 + omp_in)
100 #pragma omp declare reduction(operator+ : T : omp_out = 1513 + omp_in) initializer(omp_priv = 321)
101 #pragma omp declare reduction(min : T : omp_out = 47 - omp_in) initializer(omp_priv = 432 / omp_orig)
102 #pragma omp declare reduction(operator&& : S<T> : omp_out.f = 17 * omp_in.f) initializer(init2(omp_priv, omp_orig))
103 #pragma omp declare reduction(operator&& : T : omp_out = 17 * omp_in)
104 #pragma omp parallel
105 #pragma omp for reduction(+ : t_var) reduction(& : var) reduction(&& : var1) reduction(min : t_var1) nowait
106 for (int i = 0; i < 2; ++i) {
107 vec[i] = t_var;
108 s_arr[i] = var;
109 }
110 #pragma omp parallel
111 #pragma omp for reduction(&& : t_var)
112 for (int i = 0; i < 2; ++i) {
113 vec[i] = t_var;
114 s_arr[i] = var;
115 }
116 #pragma omp parallel
117 #pragma omp for reduction(+ : arr[1:length-2])
118 for (int i = 0; i < 2; ++i) {
119 vec[i] = t_var;
120 s_arr[i] = var;
121 }
122 return T();
123 }
124
125 extern S<float> **foo();
126
127 #pragma omp declare reduction(operator- : float, double : omp_out = 333 + omp_in)
128 #pragma omp declare reduction(min : float, double : omp_out = 555 + omp_in)
main()129 int main() {
130 #pragma omp declare reduction(operator+ : float, double : omp_out = 222 - omp_in) initializer(omp_priv = -1)
131 S<float> test;
132 float t_var = 0, t_var1;
133 int vec[] = {1, 2};
134 S<float> s_arr[] = {1, 2, 3, 4};
135 S<float> &var = test;
136 S<float> var1, arrs[10][4];
137 S<float> **var2 = foo();
138 S<float> vvar2[5];
139 S<float>(&var3)[4] = s_arr;
140 #pragma omp declare reduction(operator+ : int : omp_out = 555 * omp_in) initializer(omp_priv = 888)
141 #pragma omp parallel
142 #pragma omp for reduction(+ : t_var) reduction(& : var) reduction(&& : var1) reduction(min : t_var1)
143 for (int i = 0; i < 2; ++i) {
144 vec[i] = t_var;
145 s_arr[i] = var;
146 }
147 int arr[10][vec[1]];
148 #pragma omp parallel for reduction(+ : arr[1][ : vec[1]]) reduction(& : arrs[1 : vec[1]][1 : 2])
149 for (int i = 0; i < 10; ++i)
150 ++arr[1][i];
151 #pragma omp parallel
152 #pragma omp for reduction(+ : arr) reduction(& : arrs)
153 for (int i = 0; i < 10; ++i)
154 ++arr[1][i];
155 #pragma omp parallel
156 #pragma omp for reduction(& : var2[0 : 5][1 : 6])
157 for (int i = 0; i < 10; ++i)
158 ;
159 #pragma omp parallel
160 #pragma omp for reduction(& : vvar2[0 : 5])
161 for (int i = 0; i < 10; ++i)
162 ;
163 #pragma omp parallel
164 #pragma omp for reduction(& : var3[1 : 2])
165 for (int i = 0; i < 10; ++i)
166 ;
167 #pragma omp parallel
168 #pragma omp for reduction(& : var3) allocate(omp_cgroup_mem_alloc: var3)
169 for (int i = 0; i < 10; ++i)
170 ;
171 return tmain<int, 42>();
172 }
173
174 // CHECK: define {{.*}}i{{[0-9]+}} @main()
175 // CHECK: [[TEST:%.+]] = alloca [[S_FLOAT_TY]],
176 // CHECK: call {{.*}} [[S_FLOAT_TY_CONSTR:@.+]]([[S_FLOAT_TY]]* [[TEST]])
177 // 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]*, [4 x [[S_FLOAT_TY]]]*)* [[MAIN_MICROTASK:@.+]] to void
178 // 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
179 // 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
180 // 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
181 // 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]+}}*, [5 x [[S_FLOAT_TY]]]*)* [[MAIN_MICROTASK4:@.+]] to void
182 // 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]+}}*, [4 x [[S_FLOAT_TY]]]*)* [[MAIN_MICROTASK5:@.+]] to void
183 // 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]+}}*, [4 x [[S_FLOAT_TY]]]*)* [[MAIN_MICROTASK6:@.+]] to void
184 // CHECK: = call {{.*}}i{{.+}} [[TMAIN_INT_42:@.+]]()
185 // CHECK: call {{.*}} [[S_FLOAT_TY_DESTR:@.+]]([[S_FLOAT_TY]]*
186 // CHECK: ret
187 //
188 // 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, [4 x [[S_FLOAT_TY]]]* dereferenceable(48) %{{.+}})
189 // CHECK: [[T_VAR_PRIV:%.+]] = alloca float,
190 // CHECK: [[VAR_PRIV:%.+]] = alloca [[S_FLOAT_TY]],
191 // CHECK: [[VAR1_PRIV:%.+]] = alloca [[S_FLOAT_TY]],
192 // CHECK: [[T_VAR1_PRIV:%.+]] = alloca float,
193
194 // Reduction list for runtime.
195 // CHECK: [[RED_LIST:%.+]] = alloca [4 x i8*],
196
197 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]],
198
199 // CHECK: [[T_VAR_REF:%.+]] = load float*, float** %
200 // CHECK: [[VAR1_REF:%.+]] = load [[S_FLOAT_TY]]*, [[S_FLOAT_TY]]** %
201 // CHECK: [[T_VAR1_REF:%.+]] = load float*, float** %
202
203 // For + reduction operation initial value of private variable is -1.
204 // CHECK: call void [[RED_INIT1:@.+]](float* %{{.+}}, float* %{{.+}})
205
206 // For & reduction operation initial value of private variable is defined by call of 'init()' function.
207 // CHECK: call void [[RED_INIT2:@.+]](
208
209 // For && reduction operation initial value of private variable is 1.0.
210 // CHECK: call void [[RED_INIT3:@.+]](
211
212 // For min reduction operation initial value of private variable is largest repesentable value.
213 // CHECK: [[INIT:%.+]] = load float, float* @
214 // CHECK: store float [[INIT]], float* [[T_VAR1_PRIV]],
215
216 // CHECK: [[GTID_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[GTID_ADDR_ADDR]]
217 // CHECK: [[GTID:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[GTID_REF]]
218 // CHECK: call void @__kmpc_for_static_init_4(
219 // Skip checks for internal operations.
220 // CHECK: call void @__kmpc_for_static_fini(
221
222 // void *RedList[<n>] = {<ReductionVars>[0], ..., <ReductionVars>[<n>-1]};
223
224 // CHECK: [[T_VAR_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 0
225 // CHECK: [[BITCAST:%.+]] = bitcast float* [[T_VAR_PRIV]] to i8*
226 // CHECK: store i8* [[BITCAST]], i8** [[T_VAR_PRIV_REF]],
227 // CHECK: [[VAR_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 1
228 // CHECK: [[BITCAST:%.+]] = bitcast [[S_FLOAT_TY]]* [[VAR_PRIV]] to i8*
229 // CHECK: store i8* [[BITCAST]], i8** [[VAR_PRIV_REF]],
230 // CHECK: [[VAR1_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 2
231 // CHECK: [[BITCAST:%.+]] = bitcast [[S_FLOAT_TY]]* [[VAR1_PRIV]] to i8*
232 // CHECK: store i8* [[BITCAST]], i8** [[VAR1_PRIV_REF]],
233 // CHECK: [[T_VAR1_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 3
234 // CHECK: [[BITCAST:%.+]] = bitcast float* [[T_VAR1_PRIV]] to i8*
235 // CHECK: store i8* [[BITCAST]], i8** [[T_VAR1_PRIV_REF]],
236
237 // res = __kmpc_reduce(<loc>, <gtid>, <n>, sizeof(RedList), RedList, reduce_func, &<lock>);
238
239 // CHECK: [[BITCAST:%.+]] = bitcast [4 x i8*]* [[RED_LIST]] to i8*
240 // 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]])
241
242 // switch(res)
243 // CHECK: switch i32 [[RES]], label %[[RED_DONE:.+]] [
244 // CHECK: i32 1, label %[[CASE1:.+]]
245 // CHECK: i32 2, label %[[CASE2:.+]]
246 // CHECK: ]
247
248 // case 1:
249 // t_var += t_var_reduction;
250 // CHECK: call void [[RED_COMB1:@.+]](float* %{{.+}}, float* %{{.+}})
251
252 // var = var.operator &(var_reduction);
253 // CHECK: call void [[RED_COMB2:@.+]](
254
255 // var1 = var1.operator &&(var1_reduction);
256 // CHECK: call void [[RED_COMB3:@.+]](
257
258 // t_var1 = min(t_var1, t_var1_reduction);
259 // CHECK: call void [[RED_COMB4:@.+]](
260
261 // __kmpc_end_reduce(<loc>, <gtid>, &<lock>);
262 // CHECK: call void @__kmpc_end_reduce(%{{.+}}* [[REDUCTION_LOC]], i32 [[GTID]], [8 x i32]* [[REDUCTION_LOCK]])
263
264 // break;
265 // CHECK: br label %[[RED_DONE]]
266
267 // case 2:
268 // t_var += t_var_reduction;
269 // CHECK: call void @__kmpc_critical(
270 // CHECK: call void [[RED_COMB1]](float* %{{.+}}, float* %{{.+}})
271 // CHECK: call void @__kmpc_end_critical(
272
273 // var = var.operator &(var_reduction);
274 // CHECK: call void @__kmpc_critical(
275 // CHECK: call void [[RED_COMB2]]
276 // CHECK: call void @__kmpc_end_critical(
277
278 // var1 = var1.operator &&(var1_reduction);
279 // CHECK: call void @__kmpc_critical(
280 // CHECK: call void [[RED_COMB3]]
281 // CHECK: call void @__kmpc_end_critical(
282
283 // t_var1 = min(t_var1, t_var1_reduction);
284 // CHECK: call void @__kmpc_critical(
285 // CHECK: call void [[RED_COMB4]]
286 // CHECK: call void @__kmpc_end_critical(
287
288 // __kmpc_end_reduce(<loc>, <gtid>, &<lock>);
289 // CHECK: call void @__kmpc_end_reduce(%{{.+}}* [[REDUCTION_LOC]], i32 [[GTID]], [8 x i32]* [[REDUCTION_LOCK]])
290
291 // break;
292 // CHECK: br label %[[RED_DONE]]
293 // CHECK: [[RED_DONE]]
294 // CHECK-DAG: call {{.*}} [[S_FLOAT_TY_DESTR]]([[S_FLOAT_TY]]* [[VAR_PRIV]])
295 // CHECK-DAG: call {{.*}} [[S_FLOAT_TY_DESTR]]([[S_FLOAT_TY]]*
296 // CHECK: call void @__kmpc_barrier(%{{.+}}* [[IMPLICIT_BARRIER_LOC]], i{{[0-9]+}} [[GTID]])
297
298 // CHECK: ret void
299
300 // CHECK: define internal void [[RED_COMB1]](float* noalias %0, float* noalias %1)
301 // CHECK: fsub float 2.220000e+02, %
302
303 // CHECK: define internal void [[RED_INIT1]](float* noalias %0, float* noalias %1)
304 // CHECK: store float -1.0{{.+}}, float*
305
306 // CHECK: define internal void [[RED_COMB2]](
307 // CHECK: call {{.*}}void @_Z3redR6BaseS1RKS_(
308
309 // CHECK: define internal void [[RED_INIT2]](
310 // CHECK: call {{.*}}void @_Z4initR6BaseS1RKS_(
311
312 // CHECK: define internal void [[RED_COMB3]](
313 // CHECK: fmul float
314
315 // CHECK: define internal void [[RED_INIT3]](
316 // CHECK: call {{.*}}void @_Z5init1R6BaseS1RKS_(
317
318 // void reduce_func(void *lhs[<n>], void *rhs[<n>]) {
319 // *(Type0*)lhs[0] = ReductionOperation0(*(Type0*)lhs[0], *(Type0*)rhs[0]);
320 // ...
321 // *(Type<n>-1*)lhs[<n>-1] = ReductionOperation<n>-1(*(Type<n>-1*)lhs[<n>-1],
322 // *(Type<n>-1*)rhs[<n>-1]);
323 // }
324 // CHECK: define internal void [[REDUCTION_FUNC]](i8* %0, i8* %1)
325 // t_var_lhs = (float*)lhs[0];
326 // CHECK: [[T_VAR_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS:%.+]], i64 0, i64 0
327 // CHECK: [[T_VAR_RHS_VOID:%.+]] = load i8*, i8** [[T_VAR_RHS_REF]],
328 // CHECK: [[T_VAR_RHS:%.+]] = bitcast i8* [[T_VAR_RHS_VOID]] to float*
329 // t_var_rhs = (float*)rhs[0];
330 // CHECK: [[T_VAR_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS:%.+]], i64 0, i64 0
331 // CHECK: [[T_VAR_LHS_VOID:%.+]] = load i8*, i8** [[T_VAR_LHS_REF]],
332 // CHECK: [[T_VAR_LHS:%.+]] = bitcast i8* [[T_VAR_LHS_VOID]] to float*
333
334 // var_lhs = (S<float>*)lhs[1];
335 // CHECK: [[VAR_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS]], i64 0, i64 1
336 // CHECK: [[VAR_RHS_VOID:%.+]] = load i8*, i8** [[VAR_RHS_REF]],
337 // CHECK: [[VAR_RHS:%.+]] = bitcast i8* [[VAR_RHS_VOID]] to [[S_FLOAT_TY]]*
338 // var_rhs = (S<float>*)rhs[1];
339 // CHECK: [[VAR_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 1
340 // CHECK: [[VAR_LHS_VOID:%.+]] = load i8*, i8** [[VAR_LHS_REF]],
341 // CHECK: [[VAR_LHS:%.+]] = bitcast i8* [[VAR_LHS_VOID]] to [[S_FLOAT_TY]]*
342
343 // var1_lhs = (S<float>*)lhs[2];
344 // CHECK: [[VAR1_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS]], i64 0, i64 2
345 // CHECK: [[VAR1_RHS_VOID:%.+]] = load i8*, i8** [[VAR1_RHS_REF]],
346 // CHECK: [[VAR1_RHS:%.+]] = bitcast i8* [[VAR1_RHS_VOID]] to [[S_FLOAT_TY]]*
347 // var1_rhs = (S<float>*)rhs[2];
348 // CHECK: [[VAR1_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 2
349 // CHECK: [[VAR1_LHS_VOID:%.+]] = load i8*, i8** [[VAR1_LHS_REF]],
350 // CHECK: [[VAR1_LHS:%.+]] = bitcast i8* [[VAR1_LHS_VOID]] to [[S_FLOAT_TY]]*
351
352 // t_var1_lhs = (float*)lhs[3];
353 // CHECK: [[T_VAR1_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS]], i64 0, i64 3
354 // CHECK: [[T_VAR1_RHS_VOID:%.+]] = load i8*, i8** [[T_VAR1_RHS_REF]],
355 // CHECK: [[T_VAR1_RHS:%.+]] = bitcast i8* [[T_VAR1_RHS_VOID]] to float*
356 // t_var1_rhs = (float*)rhs[3];
357 // CHECK: [[T_VAR1_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 3
358 // CHECK: [[T_VAR1_LHS_VOID:%.+]] = load i8*, i8** [[T_VAR1_LHS_REF]],
359 // CHECK: [[T_VAR1_LHS:%.+]] = bitcast i8* [[T_VAR1_LHS_VOID]] to float*
360
361 // t_var_lhs += t_var_rhs;
362 // CHECK: call void [[RED_COMB1]](float* %{{.+}}, float* %{{.+}})
363
364 // var_lhs = var_lhs.operator &(var_rhs);
365 // CHECK: call void [[RED_COMB2]](
366
367 // var1_lhs = var1_lhs.operator &&(var1_rhs);
368 // CHECK: call void [[RED_COMB3]](
369
370 // t_var1_lhs = min(t_var1_lhs, t_var1_rhs);
371 // CHECK: call void [[RED_COMB4]](
372 // CHECK: ret void
373
374 // CHECK: define internal void [[RED_COMB4]](
375 // CHECK: fadd float 5.550000e+02, %
376
377 // CHECK: define internal void [[MAIN_MICROTASK1]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, i64 %{{.+}}, i64 %{{.+}}, i32* {{.+}} %{{.+}}, [2 x i32]* dereferenceable(8) %{{.+}}, [10 x [4 x [[S_FLOAT_TY]]]]* dereferenceable(480) %{{.+}})
378
379 // Reduction list for runtime.
380 // CHECK: [[RED_LIST:%.+]] = alloca [4 x i8*],
381
382 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]],
383 // CHECK: store i{{[0-9]+}}* %{{.+}}, i{{[0-9]+}}**
384 // CHECK: store i{{[0-9]+}}* %{{.+}}, i{{[0-9]+}}** [[ARR_ADDR:%.+]],
385 // CHECK: [[ARR:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[ARR_ADDR]],
386
387 // CHECK: [[UB_CAST:%.+]] = ptrtoint i32* [[UB1_UP:%.+]] to i64
388 // CHECK: [[LB_CAST:%.+]] = ptrtoint i32* [[LB1_0:%.+]] to i64
389 // CHECK: [[DIFF:%.+]] = sub i64 [[UB_CAST]], [[LB_CAST]]
390 // CHECK: [[SIZE_1:%.+]] = sdiv exact i64 [[DIFF]], ptrtoint (i32* getelementptr (i32, i32* null, i32 1) to i64)
391 // CHECK: [[ARR_SIZE:%.+]] = add nuw i64 [[SIZE_1]], 1
392 // CHECK: call i8* @llvm.stacksave()
393 // CHECK: [[ARR_PRIV:%.+]] = alloca i32, i64 [[ARR_SIZE]],
394
395 // Check initialization of private copy.
396 // CHECK: [[END:%.+]] = getelementptr i32, i32* [[ARR_PRIV]], i64 [[ARR_SIZE]]
397 // CHECK: [[ISEMPTY:%.+]] = icmp eq i32* [[ARR_PRIV]], [[END]]
398 // CHECK: br i1 [[ISEMPTY]],
399 // CHECK: phi i32*
400 // CHECK: call void [[RED_INIT5:@.+]](i32* %{{.+}}, i32* %{{.+}})
401 // CHECK: [[DONE:%.+]] = icmp eq i32* %{{.+}}, [[END]]
402 // CHECK: br i1 [[DONE]],
403
404 // CHECK: [[ARRS_PRIV:%.+]] = alloca [[S_FLOAT_TY]], i64 [[ARRS_SIZE:%.+]],
405
406 // Check initialization of private copy.
407 // CHECK: [[END:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[ARRS_PRIV]], i64 [[ARRS_SIZE]]
408 // CHECK: [[ISEMPTY:%.+]] = icmp eq [[S_FLOAT_TY]]* [[ARRS_PRIV]], [[END]]
409 // CHECK: br i1 [[ISEMPTY]],
410 // CHECK: phi [[S_FLOAT_TY]]*
411 // CHECK: call void [[RED_INIT2]](
412 // CHECK: [[DONE:%.+]] = icmp eq [[S_FLOAT_TY]]* %{{.+}}, [[END]]
413 // CHECK: br i1 [[DONE]],
414
415 // CHECK: [[GTID_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[GTID_ADDR_ADDR]]
416 // CHECK: [[GTID:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[GTID_REF]]
417 // CHECK: call void @__kmpc_for_static_init_4(
418 // Skip checks for internal operations.
419 // CHECK: call void @__kmpc_for_static_fini(
420
421 // void *RedList[<n>] = {<ReductionVars>[0], ..., <ReductionVars>[<n>-1]};
422
423 // CHECK: [[ARR_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 0
424 // CHECK: [[BITCAST:%.+]] = bitcast i32* [[ARR_PRIV]] to i8*
425 // CHECK: store i8* [[BITCAST]], i8** [[ARR_PRIV_REF]],
426 // CHECK: [[ARR_SIZE_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 1
427 // CHECK: [[BITCAST:%.+]] = inttoptr i64 [[ARR_SIZE]] to i8*
428 // CHECK: store i8* [[BITCAST]], i8** [[ARR_SIZE_REF]],
429 // CHECK: [[ARRS_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 2
430 // CHECK: [[BITCAST:%.+]] = bitcast [[S_FLOAT_TY]]* [[ARRS_PRIV]] to i8*
431 // CHECK: store i8* [[BITCAST]], i8** [[ARRS_PRIV_REF]],
432 // CHECK: [[ARRS_SIZE_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 3
433 // CHECK: [[BITCAST:%.+]] = inttoptr i64 [[ARRS_SIZE]] to i8*
434 // CHECK: store i8* [[BITCAST]], i8** [[ARRS_SIZE_REF]],
435
436 // res = __kmpc_reduce(<loc>, <gtid>, <n>, sizeof(RedList), RedList, reduce_func, &<lock>);
437
438 // CHECK: [[GTID_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[GTID_ADDR_ADDR]]
439 // CHECK: [[GTID:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[GTID_REF]]
440 // CHECK: [[BITCAST:%.+]] = bitcast [4 x i8*]* [[RED_LIST]] to i8*
441 // 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]])
442
443 // switch(res)
444 // CHECK: switch i32 [[RES]], label %[[RED_DONE:.+]] [
445 // CHECK: i32 1, label %[[CASE1:.+]]
446 // CHECK: i32 2, label %[[CASE2:.+]]
447 // CHECK: ]
448
449 // case 1:
450 // CHECK: [[CASE1]]
451
452 // arr[:] += arr_reduction[:];
453 // CHECK: [[END:%.+]] = getelementptr i32, i32* [[LB1_0]], i64 [[ARR_SIZE]]
454 // CHECK: [[ISEMPTY:%.+]] = icmp eq i32* [[LB1_0]], [[END]]
455 // CHECK: br i1 [[ISEMPTY]],
456 // CHECK: phi i32*
457 // CHECK: call void [[RED_COMB5:@.+]](i32* %{{.+}}, i32* %{{.+}})
458 // CHECK: [[DONE:%.+]] = icmp eq i32* %{{.+}}, [[END]]
459 // CHECK: br i1 [[DONE]],
460
461 // arrs[:] = var.operator &(arrs_reduction[:]);
462 // CHECK: [[END:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[ARRS_LB:%.+]], i64 [[ARRS_SIZE]]
463 // CHECK: [[ISEMPTY:%.+]] = icmp eq [[S_FLOAT_TY]]* [[ARRS_LB]], [[END]]
464 // CHECK: br i1 [[ISEMPTY]],
465 // CHECK: phi [[S_FLOAT_TY]]*
466 // CHECK: call void [[RED_COMB2]](
467 // CHECK: [[DONE:%.+]] = icmp eq [[S_FLOAT_TY]]* %{{.+}}, [[END]]
468 // CHECK: br i1 [[DONE]],
469
470 // __kmpc_end_reduce(<loc>, <gtid>, &<lock>);
471 // CHECK: call void @__kmpc_end_reduce_nowait(%{{.+}}* [[REDUCTION_LOC]], i32 [[GTID]], [8 x i32]* [[REDUCTION_LOCK]])
472
473 // break;
474 // CHECK: br label %[[RED_DONE]]
475
476 // case 2:
477 // CHECK: [[CASE2]]
478
479 // arr[:] += arr_reduction[:];
480 // CHECK: [[END:%.+]] = getelementptr i32, i32* [[LB1_0]], i64 [[ARR_SIZE]]
481 // CHECK: [[ISEMPTY:%.+]] = icmp eq i32* [[LB1_0]], [[END]]
482 // CHECK: br i1 [[ISEMPTY]],
483 // CHECK: phi i32*
484 // CHECK: call void @__kmpc_critical(
485 // CHECK: call void [[RED_COMB5]](
486 // CHECK: call void @__kmpc_end_critical(
487 // CHECK: [[DONE:%.+]] = icmp eq i32* %{{.+}}, [[END]]
488 // CHECK: br i1 [[DONE]],
489
490 // arrs[:] = var.operator &(arrs_reduction[:]);
491 // CHECK: [[END:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[ARRS_LB:%.+]], i64 [[ARRS_SIZE]]
492 // CHECK: [[ISEMPTY:%.+]] = icmp eq [[S_FLOAT_TY]]* [[ARRS_LB]], [[END]]
493 // CHECK: br i1 [[ISEMPTY]],
494 // CHECK: phi [[S_FLOAT_TY]]*
495 // CHECK: call void @__kmpc_critical(
496 // CHECK: call void [[RED_COMB2]](
497 // CHECK: call void @__kmpc_end_critical(
498 // CHECK: [[DONE:%.+]] = icmp eq [[S_FLOAT_TY]]* %{{.+}}, [[END]]
499 // CHECK: br i1 [[DONE]],
500
501 // break;
502 // CHECK: br label %[[RED_DONE]]
503 // CHECK: [[RED_DONE]]
504
505 // Check destruction of private copy.
506 // CHECK: [[END:%.+]] = getelementptr inbounds [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[ARRS_PRIV]], i64 [[ARRS_SIZE]]
507 // CHECK: [[ISEMPTY:%.+]] = icmp eq [[S_FLOAT_TY]]* [[ARRS_PRIV]], [[END]]
508 // CHECK: br i1 [[ISEMPTY]],
509 // CHECK: phi [[S_FLOAT_TY]]*
510 // CHECK: call void @_ZN1SIfED1Ev([[S_FLOAT_TY]]* %
511 // CHECK: [[DONE:%.+]] = icmp eq [[S_FLOAT_TY]]* %{{.+}}, [[ARRS_PRIV]]
512 // CHECK: br i1 [[DONE]],
513 // CHECK: call void @llvm.stackrestore(i8*
514
515 // CHECK: ret void
516
517 // CHECK: define internal void [[RED_COMB5]](i32* noalias %0, i32* noalias %1)
518 // CHECK: mul nsw i32 555, %
519
520 // CHECK: define internal void [[RED_INIT5]](i32* noalias %0, i32* noalias %1)
521 // CHECK: store i32 888, i32* %
522
523 // void reduce_func(void *lhs[<n>], void *rhs[<n>]) {
524 // *(Type0*)lhs[0] = ReductionOperation0(*(Type0*)lhs[0], *(Type0*)rhs[0]);
525 // ...
526 // *(Type<n>-1*)lhs[<n>-1] = ReductionOperation<n>-1(*(Type<n>-1*)lhs[<n>-1],
527 // *(Type<n>-1*)rhs[<n>-1]);
528 // }
529 // CHECK: define internal void [[REDUCTION_FUNC]](i8* %0, i8* %1)
530 // arr_rhs = (int*)rhs[0];
531 // CHECK: [[ARR_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS:%.+]], i64 0, i64 0
532 // CHECK: [[ARR_RHS_VOID:%.+]] = load i8*, i8** [[ARR_RHS_REF]],
533 // CHECK: [[ARR_RHS:%.+]] = bitcast i8* [[ARR_RHS_VOID]] to i32*
534 // arr_lhs = (int*)lhs[0];
535 // CHECK: [[ARR_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS:%.+]], i64 0, i64 0
536 // CHECK: [[ARR_LHS_VOID:%.+]] = load i8*, i8** [[ARR_LHS_REF]],
537 // CHECK: [[ARR_LHS:%.+]] = bitcast i8* [[ARR_LHS_VOID]] to i32*
538
539 // arr_size = (size_t)lhs[1];
540 // CHECK: [[ARR_SIZE_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 1
541 // CHECK: [[ARR_SIZE_VOID:%.+]] = load i8*, i8** [[ARR_SIZE_REF]],
542 // CHECK: [[ARR_SIZE:%.+]] = ptrtoint i8* [[ARR_SIZE_VOID]] to i64
543
544 // arrs_rhs = (S<float>*)rhs[2];
545 // CHECK: [[ARRS_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS]], i64 0, i64 2
546 // CHECK: [[ARRS_RHS_VOID:%.+]] = load i8*, i8** [[ARRS_RHS_REF]],
547 // CHECK: [[ARRS_RHS:%.+]] = bitcast i8* [[ARRS_RHS_VOID]] to [[S_FLOAT_TY]]*
548 // arrs_lhs = (S<float>*)lhs[2];
549 // CHECK: [[ARRS_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 2
550 // CHECK: [[ARRS_LHS_VOID:%.+]] = load i8*, i8** [[ARRS_LHS_REF]],
551 // CHECK: [[ARRS_LHS:%.+]] = bitcast i8* [[ARRS_LHS_VOID]] to [[S_FLOAT_TY]]*
552
553 // arrs_size = (size_t)lhs[3];
554 // CHECK: [[ARRS_SIZE_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 3
555 // CHECK: [[ARRS_SIZE_VOID:%.+]] = load i8*, i8** [[ARRS_SIZE_REF]],
556 // CHECK: [[ARRS_SIZE:%.+]] = ptrtoint i8* [[ARRS_SIZE_VOID]] to i64
557
558 // arr_lhs[:] += arr_rhs[:];
559 // CHECK: [[END:%.+]] = getelementptr i32, i32* [[ARR_LHS]], i64 [[ARR_SIZE]]
560 // CHECK: [[ISEMPTY:%.+]] = icmp eq i32* [[ARR_LHS]], [[END]]
561 // CHECK: br i1 [[ISEMPTY]],
562 // CHECK: phi i32*
563 // CHECK: call void [[RED_COMB5]](
564 // CHECK: [[DONE:%.+]] = icmp eq i32* %{{.+}}, [[END]]
565 // CHECK: br i1 [[DONE]],
566
567 // arrs_lhs = arrs_lhs.operator &(arrs_rhs);
568 // CHECK: [[END:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[ARRS_LB:%.+]], i64 [[ARRS_SIZE]]
569 // CHECK: [[ISEMPTY:%.+]] = icmp eq [[S_FLOAT_TY]]* [[ARRS_LB]], [[END]]
570 // CHECK: br i1 [[ISEMPTY]],
571 // CHECK: phi [[S_FLOAT_TY]]*
572 // CHECK: call void [[RED_COMB2]](
573 // CHECK: [[DONE:%.+]] = icmp eq [[S_FLOAT_TY]]* %{{.+}}, [[END]]
574 // CHECK: br i1 [[DONE]],
575
576 // CHECK: ret void
577
578 // CHECK: define internal void [[MAIN_MICROTASK2]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, i64 %{{.+}}, i64 %{{.+}}, i32* {{.+}} %{{.+}}, [10 x [4 x [[S_FLOAT_TY]]]]* dereferenceable(480) %{{.+}})
579
580 // CHECK: [[ARRS_PRIV:%.+]] = alloca [10 x [4 x [[S_FLOAT_TY]]]],
581
582 // Reduction list for runtime.
583 // CHECK: [[RED_LIST:%.+]] = alloca [3 x i8*],
584
585 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]],
586
587 // CHECK: [[ARR_SIZE:%.+]] = udiv exact i64
588 // CHECK: call i8* @llvm.stacksave()
589 // CHECK: [[ARR_PRIV:%.+]] = alloca i32, i64 [[ARR_SIZE]],
590
591 // Check initialization of private copy.
592 // CHECK: [[END:%.+]] = getelementptr i32, i32* [[ARR_PRIV]], i64 [[ARR_SIZE]]
593 // CHECK: [[ISEMPTY:%.+]] = icmp eq i32* [[ARR_PRIV]], [[END]]
594 // CHECK: br i1 [[ISEMPTY]],
595 // CHECK: phi i32*
596 // CHECK: call void [[RED_INIT5]](
597 // CHECK: [[DONE:%.+]] = icmp eq i32* %{{.+}}, [[END]]
598 // CHECK: br i1 [[DONE]],
599
600 // Check initialization of private copy.
601 // 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
602 // CHECK: [[LHS_BEGIN:%.+]] = bitcast [10 x [4 x [[S_FLOAT_TY]]]]* %{{.+}} to [[S_FLOAT_TY]]*
603 // CHECK: [[END:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[BEGIN]], i64 40
604 // CHECK: [[ISEMPTY:%.+]] = icmp eq [[S_FLOAT_TY]]* [[BEGIN]], [[END]]
605 // CHECK: br i1 [[ISEMPTY]],
606 // CHECK: phi [[S_FLOAT_TY]]*
607 // CHECK: call void [[RED_INIT2]](
608 // CHECK: [[DONE:%.+]] = icmp eq [[S_FLOAT_TY]]* %{{.+}}, [[END]]
609 // CHECK: br i1 [[DONE]],
610 // CHECK: [[LHS_BEGIN:%.+]] = bitcast [10 x [4 x [[S_FLOAT_TY]]]]* %{{.+}} to [[S_FLOAT_TY]]*
611 // CHECK: [[ARRS_PRIV_BEGIN:%.+]] = bitcast [10 x [4 x [[S_FLOAT_TY]]]]* [[ARRS_PRIV]] to [[S_FLOAT_TY]]*
612
613 // CHECK: [[GTID_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[GTID_ADDR_ADDR]]
614 // CHECK: [[GTID:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[GTID_REF]]
615 // CHECK: call void @__kmpc_for_static_init_4(
616 // Skip checks for internal operations.
617 // CHECK: call void @__kmpc_for_static_fini(
618
619 // void *RedList[<n>] = {<ReductionVars>[0], ..., <ReductionVars>[<n>-1]};
620
621 // CHECK: [[ARR_PRIV_REF:%.+]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[RED_LIST]], i64 0, i64 0
622 // CHECK: [[BITCAST:%.+]] = bitcast i32* [[ARR_PRIV]] to i8*
623 // CHECK: store i8* [[BITCAST]], i8** [[ARR_PRIV_REF]],
624 // CHECK: [[ARR_SIZE_REF:%.+]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[RED_LIST]], i64 0, i64 1
625 // CHECK: [[BITCAST:%.+]] = inttoptr i64 [[ARR_SIZE]] to i8*
626 // CHECK: store i8* [[BITCAST]], i8** [[ARR_SIZE_REF]],
627 // CHECK: [[ARRS_PRIV_REF:%.+]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[RED_LIST]], i64 0, i64 2
628 // CHECK: [[BITCAST:%.+]] = bitcast [[S_FLOAT_TY]]* [[ARRS_PRIV_BEGIN]] to i8*
629 // CHECK: store i8* [[BITCAST]], i8** [[ARRS_PRIV_REF]],
630
631 // res = __kmpc_reduce(<loc>, <gtid>, <n>, sizeof(RedList), RedList, reduce_func, &<lock>);
632
633 // CHECK: [[GTID_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[GTID_ADDR_ADDR]]
634 // CHECK: [[GTID:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[GTID_REF]]
635 // CHECK: [[BITCAST:%.+]] = bitcast [3 x i8*]* [[RED_LIST]] to i8*
636 // 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]])
637
638 // switch(res)
639 // CHECK: switch i32 [[RES]], label %[[RED_DONE:.+]] [
640 // CHECK: i32 1, label %[[CASE1:.+]]
641 // CHECK: i32 2, label %[[CASE2:.+]]
642 // CHECK: ]
643
644 // case 1:
645 // CHECK: [[CASE1]]
646
647 // arr[:] += arr_reduction[:];
648 // CHECK: [[END:%.+]] = getelementptr i32, i32* [[LB1_0:%.+]], i64 [[ARR_SIZE]]
649 // CHECK: [[ISEMPTY:%.+]] = icmp eq i32* [[LB1_0]], [[END]]
650 // CHECK: br i1 [[ISEMPTY]],
651 // CHECK: phi i32*
652 // CHECK: call void [[RED_COMB5]](
653 // CHECK: [[DONE:%.+]] = icmp eq i32* %{{.+}}, [[END]]
654 // CHECK: br i1 [[DONE]],
655
656 // arrs[:] = var.operator &(arrs_reduction[:]);
657 // CHECK: [[END:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[LHS_BEGIN]], i64 40
658 // CHECK: [[ISEMPTY:%.+]] = icmp eq [[S_FLOAT_TY]]* [[LHS_BEGIN]], [[END]]
659 // CHECK: br i1 [[ISEMPTY]],
660 // CHECK: phi [[S_FLOAT_TY]]*
661 // CHECK: call void [[RED_COMB2]](
662 // CHECK: [[DONE:%.+]] = icmp eq [[S_FLOAT_TY]]* %{{.+}}, [[END]]
663 // CHECK: br i1 [[DONE]],
664
665 // __kmpc_end_reduce(<loc>, <gtid>, &<lock>);
666 // CHECK: call void @__kmpc_end_reduce(%{{.+}}* [[REDUCTION_LOC]], i32 [[GTID]], [8 x i32]* [[REDUCTION_LOCK]])
667
668 // break;
669 // CHECK: br label %[[RED_DONE]]
670
671 // case 2:
672 // CHECK: [[CASE2]]
673
674 // arr[:] += arr_reduction[:];
675 // CHECK: [[END:%.+]] = getelementptr i32, i32* [[LB1_0]], i64 [[ARR_SIZE]]
676 // CHECK: [[ISEMPTY:%.+]] = icmp eq i32* [[LB1_0]], [[END]]
677 // CHECK: br i1 [[ISEMPTY]],
678 // CHECK: phi i32*
679 // CHECK: call void @__kmpc_critical(
680 // CHECK: call void [[RED_COMB5]](
681 // CHECK: call void @__kmpc_end_critical(
682 // CHECK: [[DONE:%.+]] = icmp eq i32* %{{.+}}, [[END]]
683 // CHECK: br i1 [[DONE]],
684
685 // arrs[:] = var.operator &(arrs_reduction[:]);
686 // CHECK: [[END:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[LHS_BEGIN]], i64 40
687 // CHECK: [[ISEMPTY:%.+]] = icmp eq [[S_FLOAT_TY]]* [[LHS_BEGIN]], [[END]]
688 // CHECK: br i1 [[ISEMPTY]],
689 // CHECK: phi [[S_FLOAT_TY]]*
690 // CHECK: call void @__kmpc_critical(
691 // CHECK: call void [[RED_COMB2]](
692 // CHECK: call void @__kmpc_end_critical(
693 // CHECK: [[DONE:%.+]] = icmp eq [[S_FLOAT_TY]]* %{{.+}}, [[END]]
694 // CHECK: br i1 [[DONE]],
695
696 // break;
697 // CHECK: br label %[[RED_DONE]]
698 // CHECK: [[RED_DONE]]
699
700 // Check destruction of private copy.
701 // 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
702 // CHECK: [[END:%.+]] = getelementptr inbounds [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[BEGIN]], i64 40
703 // CHECK: br
704 // CHECK: phi [[S_FLOAT_TY]]*
705 // CHECK: call void @_ZN1SIfED1Ev([[S_FLOAT_TY]]* %
706 // CHECK: [[DONE:%.+]] = icmp eq [[S_FLOAT_TY]]* %{{.+}}, [[BEGIN]]
707 // CHECK: br i1 [[DONE]],
708 // CHECK: call void @llvm.stackrestore(i8*
709 // CHECK: call void @__kmpc_barrier(
710
711 // CHECK: ret void
712
713 // void reduce_func(void *lhs[<n>], void *rhs[<n>]) {
714 // *(Type0*)lhs[0] = ReductionOperation0(*(Type0*)lhs[0], *(Type0*)rhs[0]);
715 // ...
716 // *(Type<n>-1*)lhs[<n>-1] = ReductionOperation<n>-1(*(Type<n>-1*)lhs[<n>-1],
717 // *(Type<n>-1*)rhs[<n>-1]);
718 // }
719 // CHECK: define internal void [[REDUCTION_FUNC]](i8* %0, i8* %1)
720 // arr_rhs = (int*)rhs[0];
721 // CHECK: [[ARR_RHS_REF:%.+]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[RED_LIST_RHS:%.+]], i64 0, i64 0
722 // CHECK: [[ARR_RHS_VOID:%.+]] = load i8*, i8** [[ARR_RHS_REF]],
723 // CHECK: [[ARR_RHS:%.+]] = bitcast i8* [[ARR_RHS_VOID]] to i32*
724 // arr_lhs = (int*)lhs[0];
725 // CHECK: [[ARR_LHS_REF:%.+]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[RED_LIST_LHS:%.+]], i64 0, i64 0
726 // CHECK: [[ARR_LHS_VOID:%.+]] = load i8*, i8** [[ARR_LHS_REF]],
727 // CHECK: [[ARR_LHS:%.+]] = bitcast i8* [[ARR_LHS_VOID]] to i32*
728
729 // arr_size = (size_t)lhs[1];
730 // CHECK: [[ARR_SIZE_REF:%.+]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[RED_LIST_LHS]], i64 0, i64 1
731 // CHECK: [[ARR_SIZE_VOID:%.+]] = load i8*, i8** [[ARR_SIZE_REF]],
732 // CHECK: [[ARR_SIZE:%.+]] = ptrtoint i8* [[ARR_SIZE_VOID]] to i64
733
734 // arrs_rhs = (S<float>*)rhs[2];
735 // CHECK: [[ARRS_RHS_REF:%.+]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[RED_LIST_RHS]], i64 0, i64 2
736 // CHECK: [[ARRS_RHS_VOID:%.+]] = load i8*, i8** [[ARRS_RHS_REF]],
737 // CHECK: [[ARRS_RHS:%.+]] = bitcast i8* [[ARRS_RHS_VOID]] to [[S_FLOAT_TY]]*
738 // arrs_lhs = (S<float>*)lhs[2];
739 // CHECK: [[ARRS_LHS_REF:%.+]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[RED_LIST_LHS]], i64 0, i64 2
740 // CHECK: [[ARRS_LHS_VOID:%.+]] = load i8*, i8** [[ARRS_LHS_REF]],
741 // CHECK: [[ARRS_LHS:%.+]] = bitcast i8* [[ARRS_LHS_VOID]] to [[S_FLOAT_TY]]*
742
743 // arr_lhs[:] += arr_rhs[:];
744 // CHECK: [[END:%.+]] = getelementptr i32, i32* [[ARR_LHS]], i64 [[ARR_SIZE]]
745 // CHECK: [[ISEMPTY:%.+]] = icmp eq i32* [[ARR_LHS]], [[END]]
746 // CHECK: br i1 [[ISEMPTY]],
747 // CHECK: phi i32*
748 // CHECK: call void [[RED_COMB5]](
749 // CHECK: [[DONE:%.+]] = icmp eq i32* %{{.+}}, [[END]]
750 // CHECK: br i1 [[DONE]],
751
752 // arrs_lhs = arrs_lhs.operator &(arrs_rhs);
753 // CHECK: [[END:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[ARRS_LB:%.+]], i64 40
754 // CHECK: [[ISEMPTY:%.+]] = icmp eq [[S_FLOAT_TY]]* [[ARRS_LB]], [[END]]
755 // CHECK: br i1 [[ISEMPTY]],
756 // CHECK: phi [[S_FLOAT_TY]]*
757 // CHECK: call void [[RED_COMB2]](
758 // CHECK: [[DONE:%.+]] = icmp eq [[S_FLOAT_TY]]* %{{.+}}, [[END]]
759 // CHECK: br i1 [[DONE]],
760
761 // CHECK: ret void
762
763 // CHECK: define internal void [[MAIN_MICROTASK3]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, [[S_FLOAT_TY]]*** dereferenceable(8) %{{.+}})
764
765 // CHECK: [[VAR2_ORIG_ADDR:%.+]] = alloca [[S_FLOAT_TY]]***,
766
767 // Reduction list for runtime.
768 // CHECK: [[RED_LIST:%.+]] = alloca [2 x i8*],
769
770 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]],
771 // CHECK: [[VAR2_ORIG:%.+]] = load [[S_FLOAT_TY]]***, [[S_FLOAT_TY]]**** [[VAR2_ORIG_ADDR]],
772
773 // CHECK: [[LAST:%.+]] = ptrtoint [[S_FLOAT_TY]]* %{{.+}} to i64
774 // CHECK: [[FIRST:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[LOW:%.+]] to i64
775 // CHECK: [[BYTE_DIF:%.+]] = sub i64 [[LAST]], [[FIRST]]
776 // CHECK: [[DIF:%.+]] = sdiv exact i64 [[BYTE_DIF]], ptrtoint ([[S_FLOAT_TY]]* getelementptr ([[S_FLOAT_TY]], [[S_FLOAT_TY]]* null, i32 1) to i64)
777 // CHECK: [[SIZE:%.+]] = add nuw i64 [[DIF]], 1
778 // CHECK: call i8* @llvm.stacksave()
779 // CHECK: [[VAR2_PRIV:%.+]] = alloca [[S_FLOAT_TY]], i64 [[SIZE]],
780 // CHECK: [[LD:%.+]] = load [[S_FLOAT_TY]]**, [[S_FLOAT_TY]]*** [[VAR2_ORIG]],
781 // CHECK: [[ORIG_START:%.+]] = load [[S_FLOAT_TY]]*, [[S_FLOAT_TY]]** [[LD]],
782 // CHECK: [[START:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[ORIG_START]] to i64
783 // CHECK: [[LOW_BOUND:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[LOW]] to i64
784 // CHECK: [[OFFSET_BYTES:%.+]] = sub i64 [[START]], [[LOW_BOUND]]
785 // CHECK: [[OFFSET:%.+]] = sdiv exact i64 [[OFFSET_BYTES]], ptrtoint ([[S_FLOAT_TY]]* getelementptr ([[S_FLOAT_TY]], [[S_FLOAT_TY]]* null, i32 1) to i64)
786 // CHECK: [[PSEUDO_VAR2_PRIV:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[VAR2_PRIV]], i64 [[OFFSET]]
787 // CHECK: store [[S_FLOAT_TY]]** [[REF:.+]], [[S_FLOAT_TY]]*** %
788 // CHECK: store [[S_FLOAT_TY]]* [[PSEUDO_VAR2_PRIV]], [[S_FLOAT_TY]]** [[REF]]
789 // CHECK: ret void
790
791 // CHECK: define internal void [[MAIN_MICROTASK4]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, [5 x [[S_FLOAT_TY]]]* dereferenceable(60) %{{.+}})
792
793 // CHECK: [[VVAR2_ORIG_ADDR:%.+]] = alloca [5 x [[S_FLOAT_TY]]]*,
794 // CHECK: [[VVAR2_PRIV:%.+]] = alloca [5 x [[S_FLOAT_TY]]],
795
796 // Reduction list for runtime.
797 // CHECK: [[RED_LIST:%.+]] = alloca [1 x i8*],
798
799 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]],
800 // CHECK: [[VVAR2_ORIG:%.+]] = load [5 x [[S_FLOAT_TY]]]*, [5 x [[S_FLOAT_TY]]]** [[VVAR2_ORIG_ADDR]],
801
802 // CHECK: [[LOW:%.+]] = getelementptr inbounds [5 x [[S_FLOAT_TY]]], [5 x [[S_FLOAT_TY]]]* [[VVAR2_ORIG]], i64 0, i64 0
803 // CHECK: [[ORIG_START:%.+]] = bitcast [5 x [[S_FLOAT_TY]]]* [[VVAR2_ORIG]] to [[S_FLOAT_TY]]*
804 // CHECK: [[START:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[ORIG_START]] to i64
805 // CHECK: [[LOW_BOUND:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[LOW]] to i64
806 // CHECK: [[OFFSET_BYTES:%.+]] = sub i64 [[START]], [[LOW_BOUND]]
807 // CHECK: [[OFFSET:%.+]] = sdiv exact i64 [[OFFSET_BYTES]], ptrtoint ([[S_FLOAT_TY]]* getelementptr ([[S_FLOAT_TY]], [[S_FLOAT_TY]]* null, i32 1) to i64)
808 // CHECK: [[VVAR2_PRIV_PTR:%.+]] = bitcast [5 x [[S_FLOAT_TY]]]* [[VVAR2_PRIV]] to [[S_FLOAT_TY]]*
809 // CHECK: [[VVAR2_PRIV:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[VVAR2_PRIV_PTR]], i64 [[OFFSET]]
810 // CHECK: ret void
811
812 // CHECK: define internal void [[MAIN_MICROTASK5]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, [4 x [[S_FLOAT_TY]]]* dereferenceable(48) %{{.+}})
813
814 // CHECK: [[VAR3_ORIG_ADDR:%.+]] = alloca [4 x [[S_FLOAT_TY]]]*,
815 // CHECK: [[VAR3_PRIV:%.+]] = alloca [2 x [[S_FLOAT_TY]]],
816
817 // Reduction list for runtime.
818 // CHECK: [[RED_LIST:%.+]] = alloca [1 x i8*],
819
820 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]],
821
822 // CHECK: [[VAR3_ORIG:%.+]] = load [4 x [[S_FLOAT_TY]]]*, [4 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR]],
823 // CHECK: store [4 x [[S_FLOAT_TY]]]* [[VAR3_ORIG]], [4 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR:%.+]],
824 // CHECK: [[VAR3_ORIG:%.+]] = load [4 x [[S_FLOAT_TY]]]*, [4 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR]],
825 // CHECK: store [4 x [[S_FLOAT_TY]]]* [[VAR3_ORIG]], [4 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR:%.+]],
826 // CHECK: [[VAR3_ORIG:%.+]] = load [4 x [[S_FLOAT_TY]]]*, [4 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR]],
827
828 // CHECK: [[LOW:%.+]] = getelementptr inbounds [4 x [[S_FLOAT_TY]]], [4 x [[S_FLOAT_TY]]]* [[VAR3_ORIG]], i64 0, i64 1
829 // CHECK: [[VAR3_ORIG:%.+]] = load [4 x [[S_FLOAT_TY]]]*, [4 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR]],
830
831 // CHECK: [[VAR3_ORIG:%.+]] = load [4 x [[S_FLOAT_TY]]]*, [4 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR]],
832 // CHECK: [[ORIG_START:%.+]] = bitcast [4 x [[S_FLOAT_TY]]]* [[VAR3_ORIG]] to [[S_FLOAT_TY]]*
833 // CHECK: [[START:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[ORIG_START]] to i64
834 // CHECK: [[LOW_BOUND:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[LOW]] to i64
835 // CHECK: [[OFFSET_BYTES:%.+]] = sub i64 [[START]], [[LOW_BOUND]]
836 // CHECK: [[OFFSET:%.+]] = sdiv exact i64 [[OFFSET_BYTES]], ptrtoint ([[S_FLOAT_TY]]* getelementptr ([[S_FLOAT_TY]], [[S_FLOAT_TY]]* null, i32 1) to i64)
837 // CHECK: [[VAR3_PRIV_PTR:%.+]] = bitcast [2 x [[S_FLOAT_TY]]]* [[VAR3_PRIV]] to [[S_FLOAT_TY]]*
838 // CHECK: [[PSEUDO_VAR3_PRIV:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[VAR3_PRIV_PTR]], i64 [[OFFSET]]
839 // CHECK: [[VAR3_PRIV:%.+]] = bitcast [[S_FLOAT_TY]]* [[PSEUDO_VAR3_PRIV]] to [4 x [[S_FLOAT_TY]]]*
840
841 // CHECK: store [4 x [[S_FLOAT_TY]]]* [[VAR3_PRIV]], [4 x [[S_FLOAT_TY]]]** %
842
843 // CHECK: ret void
844
845 // CHECK: define internal void [[MAIN_MICROTASK6]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, [4 x [[S_FLOAT_TY]]]* dereferenceable(48) %{{.+}})
846
847 // CHECK: [[VAR3_ORIG_ADDR:%.+]] = alloca [4 x [[S_FLOAT_TY]]]*,
848
849 // Reduction list for runtime.
850 // CHECK: [[RED_LIST:%.+]] = alloca [1 x i8*],
851
852 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]],
853
854 // CHECK: [[VAR3_ORIG:%.+]] = load [4 x [[S_FLOAT_TY]]]*, [4 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR]],
855 // CHECK: store [4 x [[S_FLOAT_TY]]]* [[VAR3_ORIG]], [4 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR:%.+]],
856 // CHECK: [[VAR3_ORIG:%.+]] = load [4 x [[S_FLOAT_TY]]]*, [4 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR]],
857 // CHECK: store [4 x [[S_FLOAT_TY]]]* [[VAR3_ORIG]], [4 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR:%.+]],
858 // CHECK: [[VAR3_ORIG:%.+]] = load [4 x [[S_FLOAT_TY]]]*, [4 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR]],
859 // CHECK: [[VAR3_VOID_PTR:%.+]] = call i8* @__kmpc_alloc(i32 [[GTID:%.+]], i64 48, i8* inttoptr (i64 6 to i8*))
860 // CHECK: [[VAR3_PRIV:%.+]] = bitcast i8* [[VAR3_VOID_PTR]] to [4 x [[S_FLOAT_TY]]]*
861 // CHECK: getelementptr inbounds [4 x [[S_FLOAT_TY]]], [4 x [[S_FLOAT_TY]]]* [[VAR3_PRIV]], i32 0, i32 0
862 // CHECK: bitcast [4 x [[S_FLOAT_TY]]]* [[VAR3_ORIG]] to [[S_FLOAT_TY]]*
863 // CHECK: getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* %{{.+}}, i64 4
864
865 // CHECK: store [4 x [[S_FLOAT_TY]]]* [[VAR3_PRIV]], [4 x [[S_FLOAT_TY]]]** %
866 // CHECK: call void @__kmpc_free(i32 [[GTID]], i8* [[VAR3_VOID_PTR]], i8* inttoptr (i64 6 to i8*))
867 // CHECK: ret void
868
869 // CHECK: define {{.*}} i{{[0-9]+}} [[TMAIN_INT_42]]()
870 // CHECK: [[TEST:%.+]] = alloca [[S_INT_TY]],
871 // CHECK: call {{.*}} [[S_INT_TY_CONSTR]]([[S_INT_TY]]* [[TEST]])
872 // 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
873 // Not interested in this one:
874 // CHECK: call void (%{{.+}}*, i{{[0-9]+}}, void (i{{[0-9]+}}*, i{{[0-9]+}}*, ...)*, ...) @__kmpc_fork_call(%{{.+}}* @{{.+}}, i{{[0-9]+}} 4,
875 // 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]+}}*, [42 x [[S_INT_TY]]]*, [2 x i32]*, i32*, [2 x [[S_INT_TY]]]*, [[S_INT_TY]]*)* [[TMAIN_MICROTASK2:@.+]] to void
876 // CHECK: call {{.*}} [[S_INT_TY_DESTR]]([[S_INT_TY]]*
877 // CHECK: call {{.*}} [[S_INT_TY_DESTR]]([[S_INT_TY]]*
878 // CHECK: ret
879 //
880 // 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) %{{.+}})
881 // CHECK: alloca i{{[0-9]+}},
882 // CHECK: alloca i{{[0-9]+}},
883 // CHECK: alloca i{{[0-9]+}},
884 // CHECK: alloca i{{[0-9]+}},
885 // CHECK: alloca i{{[0-9]+}},
886 // CHECK: alloca i{{[0-9]+}},
887 // CHECK: [[T_VAR_PRIV:%.+]] = alloca i{{[0-9]+}},
888 // CHECK: [[VAR_PRIV:%.+]] = alloca [[S_INT_TY]],
889 // CHECK: [[VAR1_PRIV:%.+]] = alloca [[S_INT_TY]],
890 // CHECK: [[T_VAR1_PRIV:%.+]] = alloca i{{[0-9]+}},
891
892 // Reduction list for runtime.
893 // CHECK: [[RED_LIST:%.+]] = alloca [4 x i8*],
894
895 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]],
896
897 // CHECK: [[T_VAR_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** %
898 // CHECK: [[VAR1_REF:%.+]] = load [[S_INT_TY]]*, [[S_INT_TY]]** %
899 // CHECK: [[T_VAR1_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** %
900
901 // For + reduction operation initial value of private variable is 0.
902 // CHECK: call void [[RED_INIT6:@.+]](
903
904 // For & reduction operation initial value of private variable is ones in all bits.
905 // CHECK: call void [[RED_INIT2:@.+]](
906
907 // For && reduction operation initial value of private variable is 1.0.
908 // CHECK: call void [[RED_INIT7:@.+]](
909
910 // For min reduction operation initial value of private variable is largest repesentable value.
911 // CHECK: call void [[RED_INIT8:@.+]](
912
913 // CHECK: [[GTID_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[GTID_ADDR_ADDR]]
914 // CHECK: [[GTID:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[GTID_REF]]
915 // CHECK: call void @__kmpc_for_static_init_4(
916 // Skip checks for internal operations.
917 // CHECK: call void @__kmpc_for_static_fini(
918
919 // void *RedList[<n>] = {<ReductionVars>[0], ..., <ReductionVars>[<n>-1]};
920
921 // CHECK: [[T_VAR_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 0
922 // CHECK: [[BITCAST:%.+]] = bitcast i{{[0-9]+}}* [[T_VAR_PRIV]] to i8*
923 // CHECK: store i8* [[BITCAST]], i8** [[T_VAR_PRIV_REF]],
924 // CHECK: [[VAR_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 1
925 // CHECK: [[BITCAST:%.+]] = bitcast [[S_INT_TY]]* [[VAR_PRIV]] to i8*
926 // CHECK: store i8* [[BITCAST]], i8** [[VAR_PRIV_REF]],
927 // CHECK: [[VAR1_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 2
928 // CHECK: [[BITCAST:%.+]] = bitcast [[S_INT_TY]]* [[VAR1_PRIV]] to i8*
929 // CHECK: store i8* [[BITCAST]], i8** [[VAR1_PRIV_REF]],
930 // CHECK: [[T_VAR1_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 3
931 // CHECK: [[BITCAST:%.+]] = bitcast i{{[0-9]+}}* [[T_VAR1_PRIV]] to i8*
932 // CHECK: store i8* [[BITCAST]], i8** [[T_VAR1_PRIV_REF]],
933
934 // res = __kmpc_reduce_nowait(<loc>, <gtid>, <n>, sizeof(RedList), RedList, reduce_func, &<lock>);
935
936 // CHECK: [[BITCAST:%.+]] = bitcast [4 x i8*]* [[RED_LIST]] to i8*
937 // 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]])
938
939 // switch(res)
940 // CHECK: switch i32 [[RES]], label %[[RED_DONE:.+]] [
941 // CHECK: i32 1, label %[[CASE1:.+]]
942 // CHECK: i32 2, label %[[CASE2:.+]]
943 // CHECK: ]
944
945 // case 1:
946 // t_var += t_var_reduction;
947 // CHECK: call void [[RED_COMB6:@.+]](
948
949 // var = var.operator &(var_reduction);
950 // CHECK: call void [[RED_COMB2]](
951
952 // var1 = var1.operator &&(var1_reduction);
953 // CHECK: call void [[RED_COMB7:@.+]](
954
955 // t_var1 = min(t_var1, t_var1_reduction);
956 // CHECK: call void [[RED_COMB8:@.+]](
957
958 // __kmpc_end_reduce_nowait(<loc>, <gtid>, &<lock>);
959 // CHECK: call void @__kmpc_end_reduce_nowait(%{{.+}}* [[REDUCTION_LOC]], i32 [[GTID]], [8 x i32]* [[REDUCTION_LOCK]])
960
961 // break;
962 // CHECK: br label %[[RED_DONE]]
963
964 // case 2:
965 // t_var += t_var_reduction;
966 // CHECK: call void @__kmpc_critical(
967 // CHECK: call void [[RED_COMB6]](
968 // CHECK: call void @__kmpc_end_critical(
969
970 // var = var.operator &(var_reduction);
971 // CHECK: call void @__kmpc_critical(
972 // CHECK: call void [[RED_COMB2]](
973 // CHECK: call void @__kmpc_end_critical(
974
975 // var1 = var1.operator &&(var1_reduction);
976 // CHECK: call void @__kmpc_critical(
977 // CHECK: call void [[RED_COMB7]](
978 // CHECK: call void @__kmpc_end_critical(
979
980 // t_var1 = min(t_var1, t_var1_reduction);
981 // CHECK: call void @__kmpc_critical(
982 // CHECK: call void [[RED_COMB8]](
983 // CHECK: call void @__kmpc_end_critical(
984
985 // break;
986 // CHECK: br label %[[RED_DONE]]
987 // CHECK: [[RED_DONE]]
988 // CHECK-DAG: call {{.*}} [[S_INT_TY_DESTR]]([[S_INT_TY]]* [[VAR_PRIV]])
989 // CHECK-DAG: call {{.*}} [[S_INT_TY_DESTR]]([[S_INT_TY]]*
990 // CHECK: ret void
991
992 // CHECK: define internal void [[RED_COMB6]](i32* noalias %0, i32* noalias %1)
993 // CHECK: add nsw i32 1513, %
994
995 // CHECK: define internal void [[RED_INIT6]](i32* noalias %0, i32* noalias %1)
996 // CHECK: store i32 321, i32* %
997
998 // CHECK: define internal void [[RED_COMB7]](
999 // CHECK: mul nsw i32 17, %
1000
1001 // CHECK: define internal void [[RED_INIT7]](
1002 // CHECK: call void @_Z5init2R6BaseS1RKS_(
1003
1004 // CHECK: define internal void [[RED_COMB8]](i32* noalias %0, i32* noalias %1)
1005 // CHECK: sub nsw i32 47, %
1006
1007 // CHECK: define internal void [[RED_INIT8]](i32* noalias %0, i32* noalias %1)
1008 // CHECK: sdiv i32 432, %
1009
1010 // void reduce_func(void *lhs[<n>], void *rhs[<n>]) {
1011 // *(Type0*)lhs[0] = ReductionOperation0(*(Type0*)lhs[0], *(Type0*)rhs[0]);
1012 // ...
1013 // *(Type<n>-1*)lhs[<n>-1] = ReductionOperation<n>-1(*(Type<n>-1*)lhs[<n>-1],
1014 // *(Type<n>-1*)rhs[<n>-1]);
1015 // }
1016 // CHECK: define internal void [[REDUCTION_FUNC]](i8* %0, i8* %1)
1017 // t_var_lhs = (i{{[0-9]+}}*)lhs[0];
1018 // CHECK: [[T_VAR_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS:%.+]], i64 0, i64 0
1019 // CHECK: [[T_VAR_RHS_VOID:%.+]] = load i8*, i8** [[T_VAR_RHS_REF]],
1020 // CHECK: [[T_VAR_RHS:%.+]] = bitcast i8* [[T_VAR_RHS_VOID]] to i{{[0-9]+}}*
1021 // t_var_rhs = (i{{[0-9]+}}*)rhs[0];
1022 // CHECK: [[T_VAR_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS:%.+]], i64 0, i64 0
1023 // CHECK: [[T_VAR_LHS_VOID:%.+]] = load i8*, i8** [[T_VAR_LHS_REF]],
1024 // CHECK: [[T_VAR_LHS:%.+]] = bitcast i8* [[T_VAR_LHS_VOID]] to i{{[0-9]+}}*
1025
1026 // var_lhs = (S<i{{[0-9]+}}>*)lhs[1];
1027 // CHECK: [[VAR_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS]], i64 0, i64 1
1028 // CHECK: [[VAR_RHS_VOID:%.+]] = load i8*, i8** [[VAR_RHS_REF]],
1029 // CHECK: [[VAR_RHS:%.+]] = bitcast i8* [[VAR_RHS_VOID]] to [[S_INT_TY]]*
1030 // var_rhs = (S<i{{[0-9]+}}>*)rhs[1];
1031 // CHECK: [[VAR_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 1
1032 // CHECK: [[VAR_LHS_VOID:%.+]] = load i8*, i8** [[VAR_LHS_REF]],
1033 // CHECK: [[VAR_LHS:%.+]] = bitcast i8* [[VAR_LHS_VOID]] to [[S_INT_TY]]*
1034
1035 // var1_lhs = (S<i{{[0-9]+}}>*)lhs[2];
1036 // CHECK: [[VAR1_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS]], i64 0, i64 2
1037 // CHECK: [[VAR1_RHS_VOID:%.+]] = load i8*, i8** [[VAR1_RHS_REF]],
1038 // CHECK: [[VAR1_RHS:%.+]] = bitcast i8* [[VAR1_RHS_VOID]] to [[S_INT_TY]]*
1039 // var1_rhs = (S<i{{[0-9]+}}>*)rhs[2];
1040 // CHECK: [[VAR1_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 2
1041 // CHECK: [[VAR1_LHS_VOID:%.+]] = load i8*, i8** [[VAR1_LHS_REF]],
1042 // CHECK: [[VAR1_LHS:%.+]] = bitcast i8* [[VAR1_LHS_VOID]] to [[S_INT_TY]]*
1043
1044 // t_var1_lhs = (i{{[0-9]+}}*)lhs[3];
1045 // CHECK: [[T_VAR1_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS]], i64 0, i64 3
1046 // CHECK: [[T_VAR1_RHS_VOID:%.+]] = load i8*, i8** [[T_VAR1_RHS_REF]],
1047 // CHECK: [[T_VAR1_RHS:%.+]] = bitcast i8* [[T_VAR1_RHS_VOID]] to i{{[0-9]+}}*
1048 // t_var1_rhs = (i{{[0-9]+}}*)rhs[3];
1049 // CHECK: [[T_VAR1_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 3
1050 // CHECK: [[T_VAR1_LHS_VOID:%.+]] = load i8*, i8** [[T_VAR1_LHS_REF]],
1051 // CHECK: [[T_VAR1_LHS:%.+]] = bitcast i8* [[T_VAR1_LHS_VOID]] to i{{[0-9]+}}*
1052
1053 // t_var_lhs += t_var_rhs;
1054 // CHECK: call void [[RED_COMB6]](
1055
1056 // var_lhs = var_lhs.operator &(var_rhs);
1057 // CHECK: call void [[RED_COMB2]](
1058
1059 // var1_lhs = var1_lhs.operator &&(var1_rhs);
1060 // CHECK: call void [[RED_COMB7]](
1061
1062 // t_var1_lhs = min(t_var1_lhs, t_var1_rhs);
1063 // CHECK: call void [[RED_COMB8]](
1064 // CHECK: ret void
1065
1066 // CHECK: define internal void [[TMAIN_MICROTASK2]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, [42 x [[S_INT_TY]]]* dereferenceable(504) %{{.*}}, [2 x i32]* dereferenceable(8) %{{.*}}, i32* dereferenceable(4) %{{.*}}, [2 x [[S_INT_TY]]]* dereferenceable(24) %{{.*}}, [[S_INT_TY]]* dereferenceable(12) %{{.*}})
1067
1068 // CHECK: [[ARR_ORIG_ADDR:%.+]] = alloca [42 x [[S_INT_TY]]]*,
1069 // CHECK: [[ARR_PRIV:%.+]] = alloca [40 x [[S_INT_TY]]],
1070
1071 // Reduction list for runtime.
1072 // CHECK: [[RED_LIST:%.+]] = alloca [1 x i8*],
1073
1074 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]],
1075
1076 // CHECK: [[ARR_ORIG:%.+]] = load [42 x [[S_INT_TY]]]*, [42 x [[S_INT_TY]]]** [[ARR_ORIG_ADDR]],
1077 // CHECK: [[LOW:%.+]] = getelementptr inbounds [42 x [[S_INT_TY]]], [42 x [[S_INT_TY]]]* [[ARR_ORIG]], i64 0, i64 1
1078 // CHECK: [[ORIG_START:%.+]] = bitcast [42 x [[S_INT_TY]]]* [[ARR_ORIG]] to [[S_INT_TY]]*
1079 // CHECK: [[START:%.+]] = ptrtoint [[S_INT_TY]]* [[ORIG_START]] to i64
1080 // CHECK: [[LOW_BOUND:%.+]] = ptrtoint [[S_INT_TY]]* [[LOW]] to i64
1081 // CHECK: [[OFFSET_BYTES:%.+]] = sub i64 [[START]], [[LOW_BOUND]]
1082 // CHECK: [[OFFSET:%.+]] = sdiv exact i64 [[OFFSET_BYTES]], ptrtoint ([[S_INT_TY]]* getelementptr ([[S_INT_TY]], [[S_INT_TY]]* null, i32 1) to i64)
1083 // CHECK: [[ARR_PRIV_PTR:%.+]] = bitcast [40 x [[S_INT_TY]]]* [[ARR_PRIV]] to [[S_INT_TY]]*
1084 // CHECK: [[PSEUDO_ARR_PRIV:%.+]] = getelementptr [[S_INT_TY]], [[S_INT_TY]]* [[ARR_PRIV_PTR]], i64 [[OFFSET]]
1085 // CHECK: [[ARR_PRIV:%.+]] = bitcast [[S_INT_TY]]* [[PSEUDO_ARR_PRIV]] to [42 x [[S_INT_TY]]]*
1086
1087 // CHECK: ret void
1088
1089 #endif
1090
1091