1 // RUN: %clang_cc1 -verify -triple x86_64-apple-darwin10 -fopenmp -x c++ -emit-llvm %s -o - -femit-all-decls | FileCheck %s 2 // RUN: %clang_cc1 -fopenmp -x c++ -triple x86_64-apple-darwin10 -emit-pch -o %t %s 3 // RUN: %clang_cc1 -fopenmp -x c++ -triple x86_64-apple-darwin10 -include-pch %t -verify %s -emit-llvm -o - -femit-all-decls | FileCheck %s 4 // expected-no-diagnostics 5 #ifndef HEADER 6 #define HEADER 7 8 // CHECK-LABEL: @main 9 int main(int argc, char **argv) { 10 // CHECK: [[GTID:%.+]] = call i32 @__kmpc_global_thread_num(%ident_t* [[DEFLOC:@.+]]) 11 // CHECK: call void @__kmpc_taskgroup(%ident_t* [[DEFLOC]], i32 [[GTID]]) 12 // CHECK: [[TASKV:%.+]] = call i8* @__kmpc_omp_task_alloc(%ident_t* [[DEFLOC]], i32 [[GTID]], i32 33, i64 80, i64 1, i32 (i32, i8*)* bitcast (i32 (i32, [[TDP_TY:%.+]]*)* [[TASK1:@.+]] to i32 (i32, i8*)*)) 13 // CHECK: [[TASK:%.+]] = bitcast i8* [[TASKV]] to [[TDP_TY]]* 14 // CHECK: [[TASK_DATA:%.+]] = getelementptr inbounds [[TDP_TY]], [[TDP_TY]]* [[TASK]], i32 0, i32 0 15 // CHECK: [[DOWN:%.+]] = getelementptr inbounds [[TD_TY:%.+]], [[TD_TY]]* [[TASK_DATA]], i32 0, i32 5 16 // CHECK: store i64 0, i64* [[DOWN]], 17 // CHECK: [[UP:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* [[TASK_DATA]], i32 0, i32 6 18 // CHECK: store i64 9, i64* [[UP]], 19 // CHECK: [[ST:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* [[TASK_DATA]], i32 0, i32 7 20 // CHECK: store i64 1, i64* [[ST]], 21 // CHECK: [[ST_VAL:%.+]] = load i64, i64* [[ST]], 22 // CHECK: call void @__kmpc_taskloop(%ident_t* [[DEFLOC]], i32 [[GTID]], i8* [[TASKV]], i32 1, i64* [[DOWN]], i64* [[UP]], i64 [[ST_VAL]], i32 0, i32 0, i64 0, i8* null) 23 // CHECK: call void @__kmpc_end_taskgroup(%ident_t* [[DEFLOC]], i32 [[GTID]]) 24 #pragma omp taskloop simd priority(argc) 25 for (int i = 0; i < 10; ++i) 26 ; 27 // CHECK: [[TASKV:%.+]] = call i8* @__kmpc_omp_task_alloc(%ident_t* [[DEFLOC]], i32 [[GTID]], i32 1, i64 80, i64 1, i32 (i32, i8*)* bitcast (i32 (i32, [[TDP_TY:%.+]]*)* [[TASK2:@.+]] to i32 (i32, i8*)*)) 28 // CHECK: [[TASK:%.+]] = bitcast i8* [[TASKV]] to [[TDP_TY]]* 29 // CHECK: [[TASK_DATA:%.+]] = getelementptr inbounds [[TDP_TY]], [[TDP_TY]]* [[TASK]], i32 0, i32 0 30 // CHECK: [[DOWN:%.+]] = getelementptr inbounds [[TD_TY:%.+]], [[TD_TY]]* [[TASK_DATA]], i32 0, i32 5 31 // CHECK: store i64 0, i64* [[DOWN]], 32 // CHECK: [[UP:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* [[TASK_DATA]], i32 0, i32 6 33 // CHECK: store i64 9, i64* [[UP]], 34 // CHECK: [[ST:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* [[TASK_DATA]], i32 0, i32 7 35 // CHECK: store i64 1, i64* [[ST]], 36 // CHECK: [[ST_VAL:%.+]] = load i64, i64* [[ST]], 37 // CHECK: [[GRAINSIZE:%.+]] = zext i32 %{{.+}} to i64 38 // CHECK: call void @__kmpc_taskloop(%ident_t* [[DEFLOC]], i32 [[GTID]], i8* [[TASKV]], i32 1, i64* [[DOWN]], i64* [[UP]], i64 [[ST_VAL]], i32 0, i32 1, i64 [[GRAINSIZE]], i8* null) 39 #pragma omp taskloop simd nogroup grainsize(argc) simdlen(4) 40 for (int i = 0; i < 10; ++i) 41 ; 42 // CHECK: call void @__kmpc_taskgroup(%ident_t* [[DEFLOC]], i32 [[GTID]]) 43 // CHECK: [[TASKV:%.+]] = call i8* @__kmpc_omp_task_alloc(%ident_t* [[DEFLOC]], i32 [[GTID]], i32 1, i64 80, i64 24, i32 (i32, i8*)* bitcast (i32 (i32, [[TDP_TY:%.+]]*)* [[TASK3:@.+]] to i32 (i32, i8*)*)) 44 // CHECK: [[TASK:%.+]] = bitcast i8* [[TASKV]] to [[TDP_TY]]* 45 // CHECK: [[TASK_DATA:%.+]] = getelementptr inbounds [[TDP_TY]], [[TDP_TY]]* [[TASK]], i32 0, i32 0 46 // CHECK: [[IF:%.+]] = icmp ne i32 %{{.+}}, 0 47 // CHECK: [[IF_INT:%.+]] = sext i1 [[IF]] to i32 48 // CHECK: [[DOWN:%.+]] = getelementptr inbounds [[TD_TY:%.+]], [[TD_TY]]* [[TASK_DATA]], i32 0, i32 5 49 // CHECK: store i64 0, i64* [[DOWN]], 50 // CHECK: [[UP:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* [[TASK_DATA]], i32 0, i32 6 51 // CHECK: store i64 %{{.+}}, i64* [[UP]], 52 // CHECK: [[ST:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* [[TASK_DATA]], i32 0, i32 7 53 // CHECK: store i64 1, i64* [[ST]], 54 // CHECK: [[ST_VAL:%.+]] = load i64, i64* [[ST]], 55 // CHECK: call void @__kmpc_taskloop(%ident_t* [[DEFLOC]], i32 [[GTID]], i8* [[TASKV]], i32 [[IF_INT]], i64* [[DOWN]], i64* [[UP]], i64 [[ST_VAL]], i32 0, i32 2, i64 4, i8* null) 56 // CHECK: call void @__kmpc_end_taskgroup(%ident_t* [[DEFLOC]], i32 [[GTID]]) 57 int i; 58 #pragma omp taskloop simd if(argc) shared(argc, argv) collapse(2) num_tasks(4) safelen(32) 59 for (i = 0; i < argc; ++i) 60 for (int j = argc; j < argv[argc][argc]; ++j) 61 ; 62 } 63 64 // CHECK: define internal i32 [[TASK1]]( 65 // CHECK: [[DOWN:%.+]] = getelementptr inbounds [[TD_TY:%.+]], [[TD_TY]]* %{{.+}}, i32 0, i32 5 66 // CHECK: [[DOWN_VAL:%.+]] = load i64, i64* [[DOWN]], 67 // CHECK: [[UP:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* %{{.+}}, i32 0, i32 6 68 // CHECK: [[UP_VAL:%.+]] = load i64, i64* [[UP]], 69 // CHECK: [[ST:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* %{{.+}}, i32 0, i32 7 70 // CHECK: [[ST_VAL:%.+]] = load i64, i64* [[ST]], 71 // CHECK: [[LITER:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* %{{.+}}, i32 0, i32 8 72 // CHECK: [[LITER_VAL:%.+]] = load i32, i32* [[LITER]], 73 // CHECK: store i64 [[DOWN_VAL]], i64* [[LB:%[^,]+]], 74 // CHECK: store i64 [[UP_VAL]], i64* [[UB:%[^,]+]], 75 // CHECK: store i64 [[ST_VAL]], i64* [[ST:%[^,]+]], 76 // CHECK: store i32 [[LITER_VAL]], i32* [[LITER:%[^,]+]], 77 // CHECK: [[LB_VAL:%.+]] = load i64, i64* [[LB]], 78 // CHECK: [[LB_I32:%.+]] = trunc i64 [[LB_VAL]] to i32 79 // CHECK: store i32 [[LB_I32]], i32* [[CNT:%.+]], 80 // CHECK: br label 81 // CHECK: [[VAL:%.+]] = load i32, i32* [[CNT]],{{.*}}!llvm.mem.parallel_loop_access [[LOOP1:!.+]] 82 // CHECK: [[VAL_I64:%.+]] = sext i32 [[VAL]] to i64 83 // CHECK: [[UB_VAL:%.+]] = load i64, i64* [[UB]],{{.*}}!llvm.mem.parallel_loop_access [[LOOP1]] 84 // CHECK: [[CMP:%.+]] = icmp ule i64 [[VAL_I64]], [[UB_VAL]] 85 // CHECK: br i1 [[CMP]], label %{{.+}}, label %{{.+}} 86 // CHECK: load i32, i32* %{{.*}}!llvm.mem.parallel_loop_access [[LOOP1]] 87 // CHECK: store i32 %{{.*}}!llvm.mem.parallel_loop_access [[LOOP1]] 88 // CHECK: load i32, i32* %{{.*}}!llvm.mem.parallel_loop_access [[LOOP1]] 89 // CHECK: add nsw i32 %{{.+}}, 1 90 // CHECK: store i32 %{{.+}}, i32* %{{.*}}!llvm.mem.parallel_loop_access [[LOOP1]] 91 // CHECK: br label %{{.*}}!llvm.loop [[LOOP1]] 92 // CHECK: ret i32 0 93 94 // CHECK: define internal i32 [[TASK2]]( 95 // CHECK: [[DOWN:%.+]] = getelementptr inbounds [[TD_TY:%.+]], [[TD_TY]]* %{{.+}}, i32 0, i32 5 96 // CHECK: [[DOWN_VAL:%.+]] = load i64, i64* [[DOWN]], 97 // CHECK: [[UP:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* %{{.+}}, i32 0, i32 6 98 // CHECK: [[UP_VAL:%.+]] = load i64, i64* [[UP]], 99 // CHECK: [[ST:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* %{{.+}}, i32 0, i32 7 100 // CHECK: [[ST_VAL:%.+]] = load i64, i64* [[ST]], 101 // CHECK: [[LITER:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* %{{.+}}, i32 0, i32 8 102 // CHECK: [[LITER_VAL:%.+]] = load i32, i32* [[LITER]], 103 // CHECK: store i64 [[DOWN_VAL]], i64* [[LB:%[^,]+]], 104 // CHECK: store i64 [[UP_VAL]], i64* [[UB:%[^,]+]], 105 // CHECK: store i64 [[ST_VAL]], i64* [[ST:%[^,]+]], 106 // CHECK: store i32 [[LITER_VAL]], i32* [[LITER:%[^,]+]], 107 // CHECK: [[LB_VAL:%.+]] = load i64, i64* [[LB]], 108 // CHECK: [[LB_I32:%.+]] = trunc i64 [[LB_VAL]] to i32 109 // CHECK: store i32 [[LB_I32]], i32* [[CNT:%.+]], 110 // CHECK: br label 111 // CHECK: [[VAL:%.+]] = load i32, i32* [[CNT]],{{.*}}!llvm.mem.parallel_loop_access [[LOOP2:!.+]] 112 // CHECK: [[VAL_I64:%.+]] = sext i32 [[VAL]] to i64 113 // CHECK: [[UB_VAL:%.+]] = load i64, i64* [[UB]],{{.*}}!llvm.mem.parallel_loop_access [[LOOP2]] 114 // CHECK: [[CMP:%.+]] = icmp ule i64 [[VAL_I64]], [[UB_VAL]] 115 // CHECK: br i1 [[CMP]], label %{{.+}}, label %{{.+}} 116 // CHECK: load i32, i32* %{{.*}}!llvm.mem.parallel_loop_access [[LOOP2]] 117 // CHECK: store i32 %{{.*}}!llvm.mem.parallel_loop_access [[LOOP2]] 118 // CHECK: load i32, i32* %{{.*}}!llvm.mem.parallel_loop_access [[LOOP2]] 119 // CHECK: add nsw i32 %{{.+}}, 1 120 // CHECK: store i32 %{{.+}}, i32* %{{.*}}!llvm.mem.parallel_loop_access [[LOOP2]] 121 // CHECK: br label %{{.*}}!llvm.loop [[LOOP2]] 122 // CHECK: ret i32 0 123 124 // CHECK: define internal i32 [[TASK3]]( 125 // CHECK: [[DOWN:%.+]] = getelementptr inbounds [[TD_TY:%.+]], [[TD_TY]]* %{{.+}}, i32 0, i32 5 126 // CHECK: [[DOWN_VAL:%.+]] = load i64, i64* [[DOWN]], 127 // CHECK: [[UP:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* %{{.+}}, i32 0, i32 6 128 // CHECK: [[UP_VAL:%.+]] = load i64, i64* [[UP]], 129 // CHECK: [[ST:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* %{{.+}}, i32 0, i32 7 130 // CHECK: [[ST_VAL:%.+]] = load i64, i64* [[ST]], 131 // CHECK: [[LITER:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* %{{.+}}, i32 0, i32 8 132 // CHECK: [[LITER_VAL:%.+]] = load i32, i32* [[LITER]], 133 // CHECK: store i64 [[DOWN_VAL]], i64* [[LB:%[^,]+]], 134 // CHECK: store i64 [[UP_VAL]], i64* [[UB:%[^,]+]], 135 // CHECK: store i64 [[ST_VAL]], i64* [[ST:%[^,]+]], 136 // CHECK: store i32 [[LITER_VAL]], i32* [[LITER:%[^,]+]], 137 // CHECK: [[LB_VAL:%.+]] = load i64, i64* [[LB]], 138 // CHECK: store i64 [[LB_VAL]], i64* [[CNT:%.+]], 139 // CHECK: br label 140 // CHECK-NOT: !llvm.mem.parallel_loop_access 141 // CHECK: br label %{{.*}}!llvm.loop 142 // CHECK: ret i32 0 143 144 // CHECK-LABEL: @_ZN1SC2Ei 145 struct S { 146 int a; 147 S(int c) { 148 // CHECK: [[GTID:%.+]] = call i32 @__kmpc_global_thread_num(%ident_t* [[DEFLOC:@.+]]) 149 // CHECK: [[TASKV:%.+]] = call i8* @__kmpc_omp_task_alloc(%ident_t* [[DEFLOC]], i32 [[GTID]], i32 1, i64 80, i64 16, i32 (i32, i8*)* bitcast (i32 (i32, [[TDP_TY:%.+]]*)* [[TASK4:@.+]] to i32 (i32, i8*)*)) 150 // CHECK: [[TASK:%.+]] = bitcast i8* [[TASKV]] to [[TDP_TY]]* 151 // CHECK: [[TASK_DATA:%.+]] = getelementptr inbounds [[TDP_TY]], [[TDP_TY]]* [[TASK]], i32 0, i32 0 152 // CHECK: [[DOWN:%.+]] = getelementptr inbounds [[TD_TY:%.+]], [[TD_TY]]* [[TASK_DATA]], i32 0, i32 5 153 // CHECK: store i64 0, i64* [[DOWN]], 154 // CHECK: [[UP:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* [[TASK_DATA]], i32 0, i32 6 155 // CHECK: store i64 %{{.+}}, i64* [[UP]], 156 // CHECK: [[ST:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* [[TASK_DATA]], i32 0, i32 7 157 // CHECK: store i64 1, i64* [[ST]], 158 // CHECK: [[ST_VAL:%.+]] = load i64, i64* [[ST]], 159 // CHECK: [[NUM_TASKS:%.+]] = zext i32 %{{.+}} to i64 160 // CHECK: call void @__kmpc_taskloop(%ident_t* [[DEFLOC]], i32 [[GTID]], i8* [[TASKV]], i32 1, i64* [[DOWN]], i64* [[UP]], i64 [[ST_VAL]], i32 0, i32 2, i64 [[NUM_TASKS]], i8* null) 161 #pragma omp taskloop simd shared(c) num_tasks(a) simdlen(64) safelen(8) 162 for (a = 0; a < c; ++a) 163 ; 164 } 165 } s(1); 166 167 // CHECK: define internal i32 [[TASK4]]( 168 // CHECK: [[DOWN:%.+]] = getelementptr inbounds [[TD_TY:%.+]], [[TD_TY]]* %{{.+}}, i32 0, i32 5 169 // CHECK: [[DOWN_VAL:%.+]] = load i64, i64* [[DOWN]], 170 // CHECK: [[UP:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* %{{.+}}, i32 0, i32 6 171 // CHECK: [[UP_VAL:%.+]] = load i64, i64* [[UP]], 172 // CHECK: [[ST:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* %{{.+}}, i32 0, i32 7 173 // CHECK: [[ST_VAL:%.+]] = load i64, i64* [[ST]], 174 // CHECK: [[LITER:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* %{{.+}}, i32 0, i32 8 175 // CHECK: [[LITER_VAL:%.+]] = load i32, i32* [[LITER]], 176 // CHECK: store i64 [[DOWN_VAL]], i64* [[LB:%[^,]+]], 177 // CHECK: store i64 [[UP_VAL]], i64* [[UB:%[^,]+]], 178 // CHECK: store i64 [[ST_VAL]], i64* [[ST:%[^,]+]], 179 // CHECK: store i32 [[LITER_VAL]], i32* [[LITER:%[^,]+]], 180 // CHECK: [[LB_VAL:%.+]] = load i64, i64* [[LB]], 181 // CHECK: [[LB_I32:%.+]] = trunc i64 [[LB_VAL]] to i32 182 // CHECK: store i32 [[LB_I32]], i32* [[CNT:%.+]], 183 // CHECK: br label 184 // CHECK: [[VAL:%.+]] = load i32, i32* [[CNT]], 185 // CHECK: [[VAL_I64:%.+]] = sext i32 [[VAL]] to i64 186 // CHECK: [[UB_VAL:%.+]] = load i64, i64* [[UB]], 187 // CHECK: [[CMP:%.+]] = icmp ule i64 [[VAL_I64]], [[UB_VAL]] 188 // CHECK: br i1 [[CMP]], label %{{.+}}, label %{{.+}} 189 // CHECK: load i32, i32* % 190 // CHECK-NOT: !llvm.mem.parallel_loop_access 191 // CHECK: store i32 % 192 // CHECK-NOT: !llvm.mem.parallel_loop_access 193 // CHECK: load i32, i32* % 194 // CHECK-NOT: !llvm.mem.parallel_loop_access 195 // CHECK: add nsw i32 %{{.+}}, 1 196 // CHECK: store i32 %{{.+}}, i32* % 197 // CHECK-NOT: !llvm.mem.parallel_loop_access 198 // CHECK: br label %{{.*}}!llvm.loop 199 // CHECK: ret i32 0 200 201 // CHECK: !{!"llvm.loop.vectorize.enable", i1 true} 202 // CHECK: !{!"llvm.loop.vectorize.width", i32 4} 203 // CHECK: !{!"llvm.loop.vectorize.width", i32 32} 204 // CHECK: !{!"llvm.loop.vectorize.width", i32 64} 205 206 #endif 207