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 // RUN: %clang_cc1 -verify -fopenmp -x c++ -std=c++11 -DLAMBDA -triple x86_64-apple-darwin10 -emit-llvm %s -o - | FileCheck -check-prefix=LAMBDA %s
5 // RUN: %clang_cc1 -verify -fopenmp -x c++ -fblocks -DBLOCKS -triple x86_64-apple-darwin10 -emit-llvm %s -o - | FileCheck -check-prefix=BLOCKS %s
6 // expected-no-diagnostics
7 // REQUIRES: x86-registered-target
8 #ifndef HEADER
9 #define HEADER
10 
11 volatile double g;
12 
13 template <class T>
14 struct S {
15   T f;
SS16   S(T a) : f(a + g) {}
SS17   S() : f(g) {}
operator TS18   operator T() { return T(); }
operator &S19   S &operator&(const S &) { return *this; }
~SS20   ~S() {}
21 };
22 
23 // CHECK-DAG: [[S_FLOAT_TY:%.+]] = type { float }
24 // CHECK-DAG: [[S_INT_TY:%.+]] = type { i{{[0-9]+}} }
25 // CHECK-DAG: [[ATOMIC_REDUCE_BARRIER_LOC:@.+]] = private unnamed_addr constant %{{.+}} { i32 0, i32 18, i32 0, i32 0, i8*
26 // CHECK-DAG: [[SINGLE_BARRIER_LOC:@.+]] = private unnamed_addr constant %{{.+}} { i32 0, i32 322, i32 0, i32 0, i8*
27 // CHECK-DAG: [[REDUCTION_LOC:@.+]] = private unnamed_addr constant %{{.+}} { i32 0, i32 18, i32 0, i32 0, i8*
28 // CHECK-DAG: [[REDUCTION_LOCK:@.+]] = common global [8 x i32] zeroinitializer
29 
30 template <typename T>
tmain()31 T tmain() {
32   T t;
33   S<T> test;
34   T t_var = T(), t_var1;
35   T vec[] = {1, 2};
36   S<T> s_arr[] = {1, 2};
37   S<T> var(3), var1;
38 #pragma omp parallel
39 #pragma omp sections reduction(+:t_var) reduction(&:var) reduction(&& : var1) reduction(min: t_var1) nowait
40   {
41     vec[0] = t_var;
42 #pragma omp section
43     s_arr[0] = var;
44   }
45   return T();
46 }
47 
main()48 int main() {
49 #ifdef LAMBDA
50   // LAMBDA: [[G:@.+]] = global double
51   // LAMBDA-LABEL: @main
52   // LAMBDA: call void [[OUTER_LAMBDA:@.+]](
53   [&]() {
54   // LAMBDA: define{{.*}} internal{{.*}} void [[OUTER_LAMBDA]](
55   // LAMBDA: call void {{.+}} @__kmpc_fork_call({{.+}}, i32 0, {{.+}}* [[OMP_REGION:@.+]] to {{.+}})
56 #pragma omp parallel
57 #pragma omp sections reduction(+:g)
58     {
59     // LAMBDA: define{{.*}} internal{{.*}} void [[OMP_REGION]](i32* noalias %{{.+}}, i32* noalias %{{.+}})
60     // LAMBDA: [[G_PRIVATE_ADDR:%.+]] = alloca double,
61 
62     // Reduction list for runtime.
63     // LAMBDA: [[RED_LIST:%.+]] = alloca [1 x i8*],
64 
65     // LAMBDA: store double 0.0{{.+}}, double* [[G_PRIVATE_ADDR]]
66     // LAMBDA: call void @__kmpc_for_static_init_4(
67     g = 1;
68     // LAMBDA: store double 1.0{{.+}}, double* [[G_PRIVATE_ADDR]],
69     // LAMBDA: [[G_PRIVATE_ADDR_REF:%.+]] = getelementptr inbounds %{{.+}}, %{{.+}}* [[ARG:%.+]], i{{[0-9]+}} 0, i{{[0-9]+}} 0
70     // LAMBDA: store double* [[G_PRIVATE_ADDR]], double** [[G_PRIVATE_ADDR_REF]]
71     // LAMBDA: call void [[INNER_LAMBDA:@.+]](%{{.+}}* [[ARG]])
72     // LAMBDA: call void @__kmpc_for_static_fini(
73 
74     // LAMBDA: [[G_PRIV_REF:%.+]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[RED_LIST]], i64 0, i64 0
75     // LAMBDA: [[BITCAST:%.+]] = bitcast double* [[G_PRIVATE_ADDR]] to i8*
76     // LAMBDA: store i8* [[BITCAST]], i8** [[G_PRIV_REF]],
77     // LAMBDA: call i32 @__kmpc_reduce(
78     // LAMBDA: switch i32 %{{.+}}, label %[[REDUCTION_DONE:.+]] [
79     // LAMBDA: i32 1, label %[[CASE1:.+]]
80     // LAMBDA: i32 2, label %[[CASE2:.+]]
81     // LAMBDA: [[CASE1]]
82     // LAMBDA: [[G_VAL:%.+]] = load double, double* [[G]]
83     // LAMBDA: [[G_PRIV_VAL:%.+]] = load double, double* [[G_PRIVATE_ADDR]]
84     // LAMBDA: [[ADD:%.+]] = fadd double [[G_VAL]], [[G_PRIV_VAL]]
85     // LAMBDA: store double [[ADD]], double* [[G]]
86     // LAMBDA: call void @__kmpc_end_reduce(
87     // LAMBDA: br label %[[REDUCTION_DONE]]
88     // LAMBDA: [[CASE2]]
89     // LAMBDA: [[G_PRIV_VAL:%.+]] = load double, double* [[G_PRIVATE_ADDR]]
90     // LAMBDA: fadd double
91     // LAMBDA: cmpxchg i64*
92     // LAMBDA: call void @__kmpc_end_reduce(
93     // LAMBDA: br label %[[REDUCTION_DONE]]
94     // LAMBDA: [[REDUCTION_DONE]]
95     // LAMBDA: ret void
96 #pragma omp section
97     [&]() {
98       // LAMBDA: define {{.+}} void [[INNER_LAMBDA]](%{{.+}}* [[ARG_PTR:%.+]])
99       // LAMBDA: store %{{.+}}* [[ARG_PTR]], %{{.+}}** [[ARG_PTR_REF:%.+]],
100       g = 2;
101       // LAMBDA: [[ARG_PTR:%.+]] = load %{{.+}}*, %{{.+}}** [[ARG_PTR_REF]]
102       // LAMBDA: [[G_PTR_REF:%.+]] = getelementptr inbounds %{{.+}}, %{{.+}}* [[ARG_PTR]], i{{[0-9]+}} 0, i{{[0-9]+}} 0
103       // LAMBDA: [[G_REF:%.+]] = load double*, double** [[G_PTR_REF]]
104       // LAMBDA: store double 2.0{{.+}}, double* [[G_REF]]
105     }();
106   }
107   }();
108   return 0;
109 #elif defined(BLOCKS)
110   // BLOCKS: [[G:@.+]] = global double
111   // BLOCKS-LABEL: @main
112   // BLOCKS: call void {{%.+}}(i8
113   ^{
114   // BLOCKS: define{{.*}} internal{{.*}} void {{.+}}(i8*
115   // BLOCKS: call void {{.+}} @__kmpc_fork_call({{.+}}, i32 0, {{.+}}* [[OMP_REGION:@.+]] to {{.+}})
116 #pragma omp parallel
117 #pragma omp sections reduction(-:g)
118     {
119     // BLOCKS: define{{.*}} internal{{.*}} void [[OMP_REGION]](i32* noalias %{{.+}}, i32* noalias %{{.+}})
120     // BLOCKS: [[G_PRIVATE_ADDR:%.+]] = alloca double,
121 
122     // Reduction list for runtime.
123     // BLOCKS: [[RED_LIST:%.+]] = alloca [1 x i8*],
124 
125     // BLOCKS: store double 0.0{{.+}}, double* [[G_PRIVATE_ADDR]]
126     g = 1;
127     // BLOCKS: call void @__kmpc_for_static_init_4(
128     // BLOCKS: store double 1.0{{.+}}, double* [[G_PRIVATE_ADDR]],
129     // BLOCKS-NOT: [[G]]{{[[^:word:]]}}
130     // BLOCKS: double* [[G_PRIVATE_ADDR]]
131     // BLOCKS-NOT: [[G]]{{[[^:word:]]}}
132     // BLOCKS: call void {{%.+}}(i8
133     // BLOCKS: call void @__kmpc_for_static_fini(
134 
135     // BLOCKS: [[G_PRIV_REF:%.+]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[RED_LIST]], i64 0, i64 0
136     // BLOCKS: [[BITCAST:%.+]] = bitcast double* [[G_PRIVATE_ADDR]] to i8*
137     // BLOCKS: store i8* [[BITCAST]], i8** [[G_PRIV_REF]],
138     // BLOCKS: call i32 @__kmpc_reduce(
139     // BLOCKS: switch i32 %{{.+}}, label %[[REDUCTION_DONE:.+]] [
140     // BLOCKS: i32 1, label %[[CASE1:.+]]
141     // BLOCKS: i32 2, label %[[CASE2:.+]]
142     // BLOCKS: [[CASE1]]
143     // BLOCKS: [[G_VAL:%.+]] = load double, double* [[G]]
144     // BLOCKS: [[G_PRIV_VAL:%.+]] = load double, double* [[G_PRIVATE_ADDR]]
145     // BLOCKS: [[ADD:%.+]] = fadd double [[G_VAL]], [[G_PRIV_VAL]]
146     // BLOCKS: store double [[ADD]], double* [[G]]
147     // BLOCKS: call void @__kmpc_end_reduce(
148     // BLOCKS: br label %[[REDUCTION_DONE]]
149     // BLOCKS: [[CASE2]]
150     // BLOCKS: [[G_PRIV_VAL:%.+]] = load double, double* [[G_PRIVATE_ADDR]]
151     // BLOCKS: fadd double
152     // BLOCKS: cmpxchg i64*
153     // BLOCKS: call void @__kmpc_end_reduce(
154     // BLOCKS: br label %[[REDUCTION_DONE]]
155     // BLOCKS: [[REDUCTION_DONE]]
156     // BLOCKS: ret void
157 #pragma omp section
158     ^{
159       // BLOCKS: define {{.+}} void {{@.+}}(i8*
160       g = 2;
161       // BLOCKS-NOT: [[G]]{{[[^:word:]]}}
162       // BLOCKS: store double 2.0{{.+}}, double*
163       // BLOCKS-NOT: [[G]]{{[[^:word:]]}}
164       // BLOCKS: ret
165     }();
166   }
167   }();
168   return 0;
169 #else
170   S<float> test;
171   float t_var = 0, t_var1;
172   int vec[] = {1, 2};
173   S<float> s_arr[] = {1, 2};
174   S<float> var(3), var1;
175 #pragma omp parallel
176 #pragma omp sections reduction(+:t_var) reduction(&:var) reduction(&& : var1) reduction(min: t_var1)
177   {
178     {
179     vec[0] = t_var;
180     s_arr[0] = var;
181     vec[1] = t_var1;
182     s_arr[1] = var1;
183     }
184   }
185   return tmain<int>();
186 #endif
187 }
188 
189 // CHECK: define {{.*}}i{{[0-9]+}} @main()
190 // CHECK: [[TEST:%.+]] = alloca [[S_FLOAT_TY]],
191 // CHECK: call {{.*}} [[S_FLOAT_TY_CONSTR:@.+]]([[S_FLOAT_TY]]* [[TEST]])
192 // 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
193 // CHECK: = call {{.*}}i{{.+}} [[TMAIN_INT:@.+]]()
194 // CHECK: call {{.*}} [[S_FLOAT_TY_DESTR:@.+]]([[S_FLOAT_TY]]*
195 // CHECK: ret
196 //
197 // CHECK: define internal void [[MAIN_MICROTASK]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}},
198 // CHECK-NOT: alloca float,
199 // CHECK-NOT: alloca [[S_FLOAT_TY]],
200 // CHECK-NOT: alloca [[S_FLOAT_TY]],
201 // CHECK-NOT: alloca float,
202 
203 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]],
204 
205 // CHECK: [[GTID_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[GTID_ADDR_ADDR]]
206 // CHECK: [[GTID:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[GTID_REF]]
207 // CHECK: call i32 @__kmpc_single(
208 
209 // CHECK-NOT: call {{.*}} [[S_FLOAT_TY_DESTR]]([[S_FLOAT_TY]]* [[VAR_PRIV]])
210 // CHECK-NOT: call {{.*}} [[S_FLOAT_TY_DESTR]]([[S_FLOAT_TY]]*
211 
212 // CHECK: call void @__kmpc_end_single(
213 
214 // CHECK: call void @__kmpc_barrier(%{{.+}}* [[SINGLE_BARRIER_LOC]], i{{[0-9]+}} [[GTID]])
215 
216 // CHECK: ret void
217 
218 // CHECK: define {{.*}} i{{[0-9]+}} [[TMAIN_INT]]()
219 // CHECK: [[TEST:%.+]] = alloca [[S_INT_TY]],
220 // CHECK: call {{.*}} [[S_INT_TY_CONSTR:@.+]]([[S_INT_TY]]* [[TEST]])
221 // 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
222 // CHECK: call {{.*}} [[S_INT_TY_DESTR:@.+]]([[S_INT_TY]]*
223 // CHECK: ret
224 //
225 // CHECK: define internal void [[TMAIN_MICROTASK]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}},
226 // CHECK: alloca i{{[0-9]+}},
227 // CHECK: alloca i{{[0-9]+}},
228 // CHECK: alloca i{{[0-9]+}},
229 // CHECK: alloca i{{[0-9]+}},
230 // CHECK: alloca i{{[0-9]+}},
231 // CHECK: [[T_VAR_PRIV:%.+]] = alloca i{{[0-9]+}},
232 // CHECK: [[VAR_PRIV:%.+]] = alloca [[S_INT_TY]],
233 // CHECK: [[VAR1_PRIV:%.+]] = alloca [[S_INT_TY]],
234 // CHECK: [[T_VAR1_PRIV:%.+]] = alloca i{{[0-9]+}},
235 
236 // Reduction list for runtime.
237 // CHECK: [[RED_LIST:%.+]] = alloca [4 x i8*],
238 
239 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]],
240 
241 // CHECK: [[T_VAR_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** %
242 // CHECK: [[VAR_REF:%.+]] = load [[S_INT_TY]]*, [[S_INT_TY]]** %
243 // CHECK: [[VAR1_REF:%.+]] = load [[S_INT_TY]]*, [[S_INT_TY]]** %
244 // CHECK: [[T_VAR1_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** %
245 
246 // For + reduction operation initial value of private variable is 0.
247 // CHECK: store i{{[0-9]+}} 0, i{{[0-9]+}}* [[T_VAR_PRIV]],
248 
249 // For & reduction operation initial value of private variable is ones in all bits.
250 // CHECK: call {{.*}} [[S_INT_TY_CONSTR:@.+]]([[S_INT_TY]]* [[VAR_PRIV]])
251 
252 // For && reduction operation initial value of private variable is 1.0.
253 // CHECK: call {{.*}} [[S_INT_TY_CONSTR:@.+]]([[S_INT_TY]]* [[VAR1_PRIV]])
254 
255 // For min reduction operation initial value of private variable is largest repesentable value.
256 // CHECK: store i{{[0-9]+}} 2147483647, i{{[0-9]+}}* [[T_VAR1_PRIV]],
257 
258 // CHECK: [[GTID_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[GTID_ADDR_ADDR]]
259 // CHECK: [[GTID:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[GTID_REF]]
260 // CHECK: call void @__kmpc_for_static_init_4(
261 // Skip checks for internal operations.
262 // CHECK: call void @__kmpc_for_static_fini(
263 
264 // void *RedList[<n>] = {<ReductionVars>[0], ..., <ReductionVars>[<n>-1]};
265 
266 // CHECK: [[T_VAR_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 0
267 // CHECK: [[BITCAST:%.+]] = bitcast i{{[0-9]+}}* [[T_VAR_PRIV]] to i8*
268 // CHECK: store i8* [[BITCAST]], i8** [[T_VAR_PRIV_REF]],
269 // CHECK: [[VAR_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 1
270 // CHECK: [[BITCAST:%.+]] = bitcast [[S_INT_TY]]* [[VAR_PRIV]] to i8*
271 // CHECK: store i8* [[BITCAST]], i8** [[VAR_PRIV_REF]],
272 // CHECK: [[VAR1_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 2
273 // CHECK: [[BITCAST:%.+]] = bitcast [[S_INT_TY]]* [[VAR1_PRIV]] to i8*
274 // CHECK: store i8* [[BITCAST]], i8** [[VAR1_PRIV_REF]],
275 // CHECK: [[T_VAR1_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i64 0, i64 3
276 // CHECK: [[BITCAST:%.+]] = bitcast i{{[0-9]+}}* [[T_VAR1_PRIV]] to i8*
277 // CHECK: store i8* [[BITCAST]], i8** [[T_VAR1_PRIV_REF]],
278 
279 // res = __kmpc_reduce_nowait(<loc>, <gtid>, <n>, sizeof(RedList), RedList, reduce_func, &<lock>);
280 
281 // CHECK: [[BITCAST:%.+]] = bitcast [4 x i8*]* [[RED_LIST]] to i8*
282 // 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]])
283 
284 // switch(res)
285 // CHECK: switch i32 [[RES]], label %[[RED_DONE:.+]] [
286 // CHECK: i32 1, label %[[CASE1:.+]]
287 // CHECK: i32 2, label %[[CASE2:.+]]
288 // CHECK: ]
289 
290 // case 1:
291 // t_var += t_var_reduction;
292 // CHECK: [[T_VAR_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR_REF]],
293 // CHECK: [[T_VAR_PRIV_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR_PRIV]],
294 // CHECK: [[UP:%.+]] = add nsw i{{[0-9]+}} [[T_VAR_VAL]], [[T_VAR_PRIV_VAL]]
295 // CHECK: store i{{[0-9]+}} [[UP]], i{{[0-9]+}}* [[T_VAR_REF]],
296 
297 // var = var.operator &(var_reduction);
298 // CHECK: [[UP:%.+]] = call dereferenceable(4) [[S_INT_TY]]* @{{.+}}([[S_INT_TY]]* [[VAR_REF]], [[S_INT_TY]]* dereferenceable(4) [[VAR_PRIV]])
299 // CHECK: [[BC1:%.+]] = bitcast [[S_INT_TY]]* [[VAR_REF]] to i8*
300 // CHECK: [[BC2:%.+]] = bitcast [[S_INT_TY]]* [[UP]] to i8*
301 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* [[BC1]], i8* [[BC2]], i64 4, i32 4, i1 false)
302 
303 // var1 = var1.operator &&(var1_reduction);
304 // CHECK: [[TO_INT:%.+]] = call i{{[0-9]+}} @{{.+}}([[S_INT_TY]]* [[VAR1_REF]])
305 // CHECK: [[VAR1_BOOL:%.+]] = icmp ne i{{[0-9]+}} [[TO_INT]], 0
306 // CHECK: br i1 [[VAR1_BOOL]], label %[[TRUE:.+]], label %[[END2:.+]]
307 // CHECK: [[TRUE]]
308 // CHECK: [[TO_INT:%.+]] = call i{{[0-9]+}} @{{.+}}([[S_INT_TY]]* [[VAR1_PRIV]])
309 // CHECK: [[VAR1_REDUCTION_BOOL:%.+]] = icmp ne i{{[0-9]+}} [[TO_INT]], 0
310 // CHECK: br label %[[END2]]
311 // CHECK: [[END2]]
312 // CHECK: [[COND_LVALUE:%.+]] = phi i1 [ false, %{{.+}} ], [ [[VAR1_REDUCTION_BOOL]], %[[TRUE]] ]
313 // CHECK: [[CONV:%.+]] = zext i1 [[COND_LVALUE]] to i32
314 // CHECK:  call void @{{.+}}([[S_INT_TY]]* [[COND_LVALUE:%.+]], i32 [[CONV]])
315 // CHECK: [[BC1:%.+]] = bitcast [[S_INT_TY]]* [[VAR1_REF]] to i8*
316 // CHECK: [[BC2:%.+]] = bitcast [[S_INT_TY]]* [[COND_LVALUE]] to i8*
317 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* [[BC1]], i8* [[BC2]], i64 4, i32 4, i1 false)
318 
319 // t_var1 = min(t_var1, t_var1_reduction);
320 // CHECK: [[T_VAR1_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR1_REF]],
321 // CHECK: [[T_VAR1_PRIV_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR1_PRIV]],
322 // CHECK: [[CMP:%.+]] = icmp slt i{{[0-9]+}} [[T_VAR1_VAL]], [[T_VAR1_PRIV_VAL]]
323 // CHECK: br i1 [[CMP]]
324 // CHECK: [[UP:%.+]] = phi i32
325 // CHECK: store i{{[0-9]+}} [[UP]], i{{[0-9]+}}* [[T_VAR1_REF]],
326 
327 // __kmpc_end_reduce_nowait(<loc>, <gtid>, &<lock>);
328 // CHECK: call void @__kmpc_end_reduce_nowait(%{{.+}}* [[REDUCTION_LOC]], i32 [[GTID]], [8 x i32]* [[REDUCTION_LOCK]])
329 
330 // break;
331 // CHECK: br label %[[RED_DONE]]
332 
333 // case 2:
334 // t_var += t_var_reduction;
335 // CHECK: [[T_VAR_PRIV_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR_PRIV]]
336 // CHECK: atomicrmw add i32* [[T_VAR_REF]], i32 [[T_VAR_PRIV_VAL]] monotonic
337 
338 // var = var.operator &(var_reduction);
339 // CHECK: call void @__kmpc_critical(
340 // CHECK: [[UP:%.+]] = call dereferenceable(4) [[S_INT_TY]]* @{{.+}}([[S_INT_TY]]* [[VAR_REF]], [[S_INT_TY]]* dereferenceable(4) [[VAR_PRIV]])
341 // CHECK: [[BC1:%.+]] = bitcast [[S_INT_TY]]* [[VAR_REF]] to i8*
342 // CHECK: [[BC2:%.+]] = bitcast [[S_INT_TY]]* [[UP]] to i8*
343 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* [[BC1]], i8* [[BC2]], i64 4, i32 4, i1 false)
344 // CHECK: call void @__kmpc_end_critical(
345 
346 // var1 = var1.operator &&(var1_reduction);
347 // CHECK: call void @__kmpc_critical(
348 // CHECK: [[TO_INT:%.+]] = call i{{[0-9]+}} @{{.+}}([[S_INT_TY]]* [[VAR1_REF]])
349 // CHECK: [[VAR1_BOOL:%.+]] = icmp ne i{{[0-9]+}} [[TO_INT]], 0
350 // CHECK: br i1 [[VAR1_BOOL]], label %[[TRUE:.+]], label %[[END2:.+]]
351 // CHECK: [[TRUE]]
352 // CHECK: [[TO_INT:%.+]] = call i{{[0-9]+}} @{{.+}}([[S_INT_TY]]* [[VAR1_PRIV]])
353 // CHECK: [[VAR1_REDUCTION_BOOL:%.+]] = icmp ne i{{[0-9]+}} [[TO_INT]], 0
354 // CHECK: br label %[[END2]]
355 // CHECK: [[END2]]
356 // CHECK: [[COND_LVALUE:%.+]] = phi i1 [ false, %{{.+}} ], [ [[VAR1_REDUCTION_BOOL]], %[[TRUE]] ]
357 // CHECK: [[CONV:%.+]] = zext i1 [[COND_LVALUE]] to i32
358 // CHECK:  call void @{{.+}}([[S_INT_TY]]* [[COND_LVALUE:%.+]], i32 [[CONV]])
359 // CHECK: [[BC1:%.+]] = bitcast [[S_INT_TY]]* [[VAR1_REF]] to i8*
360 // CHECK: [[BC2:%.+]] = bitcast [[S_INT_TY]]* [[COND_LVALUE]] to i8*
361 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* [[BC1]], i8* [[BC2]], i64 4, i32 4, i1 false)
362 // CHECK: call void @__kmpc_end_critical(
363 
364 // t_var1 = min(t_var1, t_var1_reduction);
365 // CHECK: [[T_VAR1_PRIV_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR1_PRIV]]
366 // CHECK: atomicrmw min i32* [[T_VAR1_REF]], i32 [[T_VAR1_PRIV_VAL]] monotonic
367 
368 // break;
369 // CHECK: br label %[[RED_DONE]]
370 // CHECK: [[RED_DONE]]
371 // CHECK-DAG: call {{.*}} [[S_INT_TY_DESTR]]([[S_INT_TY]]* [[VAR_PRIV]])
372 // CHECK-DAG: call {{.*}} [[S_INT_TY_DESTR]]([[S_INT_TY]]*
373 // CHECK: ret void
374 
375 // void reduce_func(void *lhs[<n>], void *rhs[<n>]) {
376 //  *(Type0*)lhs[0] = ReductionOperation0(*(Type0*)lhs[0], *(Type0*)rhs[0]);
377 //  ...
378 //  *(Type<n>-1*)lhs[<n>-1] = ReductionOperation<n>-1(*(Type<n>-1*)lhs[<n>-1],
379 //  *(Type<n>-1*)rhs[<n>-1]);
380 // }
381 // CHECK: define internal void [[REDUCTION_FUNC]](i8*, i8*)
382 // t_var_lhs = (i{{[0-9]+}}*)lhs[0];
383 // CHECK: [[T_VAR_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS:%.+]], i64 0, i64 0
384 // CHECK: [[T_VAR_RHS_VOID:%.+]] = load i8*, i8** [[T_VAR_RHS_REF]],
385 // CHECK: [[T_VAR_RHS:%.+]] = bitcast i8* [[T_VAR_RHS_VOID]] to i{{[0-9]+}}*
386 // t_var_rhs = (i{{[0-9]+}}*)rhs[0];
387 // CHECK: [[T_VAR_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS:%.+]], i64 0, i64 0
388 // CHECK: [[T_VAR_LHS_VOID:%.+]] = load i8*, i8** [[T_VAR_LHS_REF]],
389 // CHECK: [[T_VAR_LHS:%.+]] = bitcast i8* [[T_VAR_LHS_VOID]] to i{{[0-9]+}}*
390 
391 // var_lhs = (S<i{{[0-9]+}}>*)lhs[1];
392 // CHECK: [[VAR_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS]], i64 0, i64 1
393 // CHECK: [[VAR_RHS_VOID:%.+]] = load i8*, i8** [[VAR_RHS_REF]],
394 // CHECK: [[VAR_RHS:%.+]] = bitcast i8* [[VAR_RHS_VOID]] to [[S_INT_TY]]*
395 // var_rhs = (S<i{{[0-9]+}}>*)rhs[1];
396 // CHECK: [[VAR_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 1
397 // CHECK: [[VAR_LHS_VOID:%.+]] = load i8*, i8** [[VAR_LHS_REF]],
398 // CHECK: [[VAR_LHS:%.+]] = bitcast i8* [[VAR_LHS_VOID]] to [[S_INT_TY]]*
399 
400 // var1_lhs = (S<i{{[0-9]+}}>*)lhs[2];
401 // CHECK: [[VAR1_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS]], i64 0, i64 2
402 // CHECK: [[VAR1_RHS_VOID:%.+]] = load i8*, i8** [[VAR1_RHS_REF]],
403 // CHECK: [[VAR1_RHS:%.+]] = bitcast i8* [[VAR1_RHS_VOID]] to [[S_INT_TY]]*
404 // var1_rhs = (S<i{{[0-9]+}}>*)rhs[2];
405 // CHECK: [[VAR1_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 2
406 // CHECK: [[VAR1_LHS_VOID:%.+]] = load i8*, i8** [[VAR1_LHS_REF]],
407 // CHECK: [[VAR1_LHS:%.+]] = bitcast i8* [[VAR1_LHS_VOID]] to [[S_INT_TY]]*
408 
409 // t_var1_lhs = (i{{[0-9]+}}*)lhs[3];
410 // CHECK: [[T_VAR1_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS]], i64 0, i64 3
411 // CHECK: [[T_VAR1_RHS_VOID:%.+]] = load i8*, i8** [[T_VAR1_RHS_REF]],
412 // CHECK: [[T_VAR1_RHS:%.+]] = bitcast i8* [[T_VAR1_RHS_VOID]] to i{{[0-9]+}}*
413 // t_var1_rhs = (i{{[0-9]+}}*)rhs[3];
414 // CHECK: [[T_VAR1_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i64 0, i64 3
415 // CHECK: [[T_VAR1_LHS_VOID:%.+]] = load i8*, i8** [[T_VAR1_LHS_REF]],
416 // CHECK: [[T_VAR1_LHS:%.+]] = bitcast i8* [[T_VAR1_LHS_VOID]] to i{{[0-9]+}}*
417 
418 // t_var_lhs += t_var_rhs;
419 // CHECK: [[T_VAR_LHS_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR_LHS]],
420 // CHECK: [[T_VAR_RHS_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR_RHS]],
421 // CHECK: [[UP:%.+]] = add nsw i{{[0-9]+}} [[T_VAR_LHS_VAL]], [[T_VAR_RHS_VAL]]
422 // CHECK: store i{{[0-9]+}} [[UP]], i{{[0-9]+}}* [[T_VAR_LHS]],
423 
424 // var_lhs = var_lhs.operator &(var_rhs);
425 // CHECK: [[UP:%.+]] = call dereferenceable(4) [[S_INT_TY]]* @{{.+}}([[S_INT_TY]]* [[VAR_LHS]], [[S_INT_TY]]* dereferenceable(4) [[VAR_RHS]])
426 // CHECK: [[BC1:%.+]] = bitcast [[S_INT_TY]]* [[VAR_LHS]] to i8*
427 // CHECK: [[BC2:%.+]] = bitcast [[S_INT_TY]]* [[UP]] to i8*
428 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* [[BC1]], i8* [[BC2]], i64 4, i32 4, i1 false)
429 
430 // var1_lhs = var1_lhs.operator &&(var1_rhs);
431 // CHECK: [[TO_INT:%.+]] = call i{{[0-9]+}} @{{.+}}([[S_INT_TY]]* [[VAR1_LHS]])
432 // CHECK: [[VAR1_BOOL:%.+]] = icmp ne i{{[0-9]+}} [[TO_INT]], 0
433 // CHECK: br i1 [[VAR1_BOOL]], label %[[TRUE:.+]], label %[[END2:.+]]
434 // CHECK: [[TRUE]]
435 // CHECK: [[TO_INT:%.+]] = call i{{[0-9]+}} @{{.+}}([[S_INT_TY]]* [[VAR1_RHS]])
436 // CHECK: [[VAR1_REDUCTION_BOOL:%.+]] = icmp ne i{{[0-9]+}} [[TO_INT]], 0
437 // CHECK: br label %[[END2]]
438 // CHECK: [[END2]]
439 // CHECK: [[COND_LVALUE:%.+]] = phi i1 [ false, %{{.+}} ], [ [[VAR1_REDUCTION_BOOL]], %[[TRUE]] ]
440 // CHECK: [[CONV:%.+]] = zext i1 [[COND_LVALUE]] to i32
441 // CHECK:  call void @{{.+}}([[S_INT_TY]]* [[COND_LVALUE:%.+]], i32 [[CONV]])
442 // CHECK: [[BC1:%.+]] = bitcast [[S_INT_TY]]* [[VAR1_LHS]] to i8*
443 // CHECK: [[BC2:%.+]] = bitcast [[S_INT_TY]]* [[COND_LVALUE]] to i8*
444 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* [[BC1]], i8* [[BC2]], i64 4, i32 4, i1 false)
445 
446 // t_var1_lhs = min(t_var1_lhs, t_var1_rhs);
447 // CHECK: [[T_VAR1_LHS_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR1_LHS]],
448 // CHECK: [[T_VAR1_RHS_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR1_RHS]],
449 // CHECK: [[CMP:%.+]] = icmp slt i{{[0-9]+}} [[T_VAR1_LHS_VAL]], [[T_VAR1_RHS_VAL]]
450 // CHECK: br i1 [[CMP]]
451 // CHECK: [[UP:%.+]] = phi i32
452 // CHECK: store i{{[0-9]+}} [[UP]], i{{[0-9]+}}* [[T_VAR1_LHS]],
453 // CHECK: ret void
454 
455 #endif
456 
457