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