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 // REQUIRES: x86-registered-target 6 #ifndef HEADER 7 #define HEADER 8 9 // CHECK-LABEL: @main 10 int main(int argc, char **argv) { 11 // CHECK: [[GTID:%.+]] = call i32 @__kmpc_global_thread_num(%ident_t* [[DEFLOC:@.+]]) 12 // CHECK: [[TASKV:%.+]] = call i8* @__kmpc_omp_task_alloc(%ident_t* [[DEFLOC]], i32 [[GTID]], i32 1, i64 64, 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: getelementptr inbounds [[TD_TY:%.+]], [[TD_TY]]* [[TASK_DATA]], i32 0, i32 3 16 // CHECK: store i32 (i32, i8*)* null, i32 (i32, i8*)** %{{.+}} 17 // CHECK: [[DOWN:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* [[TASK_DATA]], i32 0, i32 4 18 // CHECK: store i64 0, i64* [[DOWN]], 19 // CHECK: [[UP:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* [[TASK_DATA]], i32 0, i32 5 20 // CHECK: store i64 9, i64* [[UP]], 21 // CHECK: [[ST:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* [[TASK_DATA]], i32 0, i32 6 22 // CHECK: store i64 1, i64* [[ST]], 23 // CHECK: [[ST_VAL:%.+]] = load i64, i64* [[ST]], 24 // 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) 25 #pragma omp taskloop simd 26 for (int i = 0; i < 10; ++i) 27 ; 28 // CHECK: [[TASKV:%.+]] = call i8* @__kmpc_omp_task_alloc(%ident_t* [[DEFLOC]], i32 [[GTID]], i32 1, i64 64, i64 1, i32 (i32, i8*)* bitcast (i32 (i32, [[TDP_TY:%.+]]*)* [[TASK2:@.+]] to i32 (i32, i8*)*)) 29 // CHECK: [[TASK:%.+]] = bitcast i8* [[TASKV]] to [[TDP_TY]]* 30 // CHECK: [[TASK_DATA:%.+]] = getelementptr inbounds [[TDP_TY]], [[TDP_TY]]* [[TASK]], i32 0, i32 0 31 // CHECK: getelementptr inbounds [[TD_TY:%.+]], [[TD_TY]]* [[TASK_DATA]], i32 0, i32 3 32 // CHECK: store i32 (i32, i8*)* null, i32 (i32, i8*)** %{{.+}} 33 // CHECK: [[DOWN:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* [[TASK_DATA]], i32 0, i32 4 34 // CHECK: store i64 0, i64* [[DOWN]], 35 // CHECK: [[UP:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* [[TASK_DATA]], i32 0, i32 5 36 // CHECK: store i64 9, i64* [[UP]], 37 // CHECK: [[ST:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* [[TASK_DATA]], i32 0, i32 6 38 // CHECK: store i64 1, i64* [[ST]], 39 // CHECK: [[ST_VAL:%.+]] = load i64, i64* [[ST]], 40 // CHECK: [[GRAINSIZE:%.+]] = zext i32 %{{.+}} to i64 41 // CHECK: call void @__kmpc_taskloop(%ident_t* [[DEFLOC]], i32 [[GTID]], i8* [[TASKV]], i32 1, i64* [[DOWN]], i64* [[UP]], i64 [[ST_VAL]], i32 1, i32 1, i64 [[GRAINSIZE]], i8* null) 42 #pragma omp taskloop simd nogroup grainsize(argc) simdlen(4) 43 for (int i = 0; i < 10; ++i) 44 ; 45 // CHECK: [[TASKV:%.+]] = call i8* @__kmpc_omp_task_alloc(%ident_t* [[DEFLOC]], i32 [[GTID]], i32 1, i64 64, i64 24, i32 (i32, i8*)* bitcast (i32 (i32, [[TDP_TY:%.+]]*)* [[TASK3:@.+]] to i32 (i32, i8*)*)) 46 // CHECK: [[TASK:%.+]] = bitcast i8* [[TASKV]] to [[TDP_TY]]* 47 // CHECK: [[TASK_DATA:%.+]] = getelementptr inbounds [[TDP_TY]], [[TDP_TY]]* [[TASK]], i32 0, i32 0 48 // CHECK: getelementptr inbounds [[TD_TY:%.+]], [[TD_TY]]* [[TASK_DATA]], i32 0, i32 3 49 // CHECK: store i32 (i32, i8*)* null, i32 (i32, i8*)** %{{.+}} 50 // CHECK: [[IF:%.+]] = icmp ne i32 %{{.+}}, 0 51 // CHECK: [[IF_INT:%.+]] = sext i1 [[IF]] to i32 52 // CHECK: [[DOWN:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* [[TASK_DATA]], i32 0, i32 4 53 // CHECK: store i64 0, i64* [[DOWN]], 54 // CHECK: [[UP:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* [[TASK_DATA]], i32 0, i32 5 55 // CHECK: store i64 %{{.+}}, i64* [[UP]], 56 // CHECK: [[ST:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* [[TASK_DATA]], i32 0, i32 6 57 // CHECK: store i64 1, i64* [[ST]], 58 // CHECK: [[ST_VAL:%.+]] = load i64, i64* [[ST]], 59 // 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) 60 int i; 61 #pragma omp taskloop simd if(argc) shared(argc, argv) collapse(2) num_tasks(4) safelen(32) 62 for (i = 0; i < argc; ++i) 63 for (int j = argc; j < argv[argc][argc]; ++j) 64 ; 65 } 66 67 // CHECK: define internal i32 [[TASK1]]( 68 // CHECK: [[DOWN:%.+]] = getelementptr inbounds [[TD_TY:%.+]], [[TD_TY]]* %{{.+}}, i32 0, i32 4 69 // CHECK: [[DOWN_VAL:%.+]] = load i64, i64* [[DOWN]], 70 // CHECK: [[UP:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* %{{.+}}, i32 0, i32 5 71 // CHECK: [[UP_VAL:%.+]] = load i64, i64* [[UP]], 72 // CHECK: [[ST:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* %{{.+}}, i32 0, i32 6 73 // CHECK: [[ST_VAL:%.+]] = load i64, i64* [[ST]], 74 // CHECK: [[LITER:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* %{{.+}}, i32 0, i32 7 75 // CHECK: [[LITER_VAL:%.+]] = load i32, i32* [[LITER]], 76 // CHECK: store i64 [[DOWN_VAL]], i64* [[LB:%[^,]+]], 77 // CHECK: store i64 [[UP_VAL]], i64* [[UB:%[^,]+]], 78 // CHECK: store i64 [[ST_VAL]], i64* [[ST:%[^,]+]], 79 // CHECK: store i32 [[LITER_VAL]], i32* [[LITER:%[^,]+]], 80 // CHECK: [[LB_VAL:%.+]] = load i64, i64* [[LB]], 81 // CHECK: [[LB_I32:%.+]] = trunc i64 [[LB_VAL]] to i32 82 // CHECK: store i32 [[LB_I32]], i32* [[CNT:%.+]], 83 // CHECK: br label 84 // CHECK: [[VAL:%.+]] = load i32, i32* [[CNT]],{{.*}}!llvm.mem.parallel_loop_access [[LOOP1:!.+]] 85 // CHECK: [[VAL_I64:%.+]] = sext i32 [[VAL]] to i64 86 // CHECK: [[UB_VAL:%.+]] = load i64, i64* [[UB]],{{.*}}!llvm.mem.parallel_loop_access [[LOOP1]] 87 // CHECK: [[CMP:%.+]] = icmp ule i64 [[VAL_I64]], [[UB_VAL]] 88 // CHECK: br i1 [[CMP]], label %{{.+}}, label %{{.+}} 89 // CHECK: load i32, i32* %{{.*}}!llvm.mem.parallel_loop_access [[LOOP1]] 90 // CHECK: store i32 %{{.*}}!llvm.mem.parallel_loop_access [[LOOP1]] 91 // CHECK: load i32, i32* %{{.*}}!llvm.mem.parallel_loop_access [[LOOP1]] 92 // CHECK: add nsw i32 %{{.+}}, 1 93 // CHECK: store i32 %{{.+}}, i32* %{{.*}}!llvm.mem.parallel_loop_access [[LOOP1]] 94 // CHECK: br label %{{.*}}!llvm.loop [[LOOP1]] 95 // CHECK: ret i32 0 96 97 // CHECK: define internal i32 [[TASK2]]( 98 // CHECK: [[DOWN:%.+]] = getelementptr inbounds [[TD_TY:%.+]], [[TD_TY]]* %{{.+}}, i32 0, i32 4 99 // CHECK: [[DOWN_VAL:%.+]] = load i64, i64* [[DOWN]], 100 // CHECK: [[UP:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* %{{.+}}, i32 0, i32 5 101 // CHECK: [[UP_VAL:%.+]] = load i64, i64* [[UP]], 102 // CHECK: [[ST:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* %{{.+}}, i32 0, i32 6 103 // CHECK: [[ST_VAL:%.+]] = load i64, i64* [[ST]], 104 // CHECK: [[LITER:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* %{{.+}}, i32 0, i32 7 105 // CHECK: [[LITER_VAL:%.+]] = load i32, i32* [[LITER]], 106 // CHECK: store i64 [[DOWN_VAL]], i64* [[LB:%[^,]+]], 107 // CHECK: store i64 [[UP_VAL]], i64* [[UB:%[^,]+]], 108 // CHECK: store i64 [[ST_VAL]], i64* [[ST:%[^,]+]], 109 // CHECK: store i32 [[LITER_VAL]], i32* [[LITER:%[^,]+]], 110 // CHECK: [[LB_VAL:%.+]] = load i64, i64* [[LB]], 111 // CHECK: [[LB_I32:%.+]] = trunc i64 [[LB_VAL]] to i32 112 // CHECK: store i32 [[LB_I32]], i32* [[CNT:%.+]], 113 // CHECK: br label 114 // CHECK: [[VAL:%.+]] = load i32, i32* [[CNT]],{{.*}}!llvm.mem.parallel_loop_access [[LOOP2:!.+]] 115 // CHECK: [[VAL_I64:%.+]] = sext i32 [[VAL]] to i64 116 // CHECK: [[UB_VAL:%.+]] = load i64, i64* [[UB]],{{.*}}!llvm.mem.parallel_loop_access [[LOOP2]] 117 // CHECK: [[CMP:%.+]] = icmp ule i64 [[VAL_I64]], [[UB_VAL]] 118 // CHECK: br i1 [[CMP]], label %{{.+}}, label %{{.+}} 119 // CHECK: load i32, i32* %{{.*}}!llvm.mem.parallel_loop_access [[LOOP2]] 120 // CHECK: store i32 %{{.*}}!llvm.mem.parallel_loop_access [[LOOP2]] 121 // CHECK: load i32, i32* %{{.*}}!llvm.mem.parallel_loop_access [[LOOP2]] 122 // CHECK: add nsw i32 %{{.+}}, 1 123 // CHECK: store i32 %{{.+}}, i32* %{{.*}}!llvm.mem.parallel_loop_access [[LOOP2]] 124 // CHECK: br label %{{.*}}!llvm.loop [[LOOP2]] 125 // CHECK: ret i32 0 126 127 // CHECK: define internal i32 [[TASK3]]( 128 // CHECK: [[DOWN:%.+]] = getelementptr inbounds [[TD_TY:%.+]], [[TD_TY]]* %{{.+}}, i32 0, i32 4 129 // CHECK: [[DOWN_VAL:%.+]] = load i64, i64* [[DOWN]], 130 // CHECK: [[UP:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* %{{.+}}, i32 0, i32 5 131 // CHECK: [[UP_VAL:%.+]] = load i64, i64* [[UP]], 132 // CHECK: [[ST:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* %{{.+}}, i32 0, i32 6 133 // CHECK: [[ST_VAL:%.+]] = load i64, i64* [[ST]], 134 // CHECK: [[LITER:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* %{{.+}}, i32 0, i32 7 135 // CHECK: [[LITER_VAL:%.+]] = load i32, i32* [[LITER]], 136 // CHECK: store i64 [[DOWN_VAL]], i64* [[LB:%[^,]+]], 137 // CHECK: store i64 [[UP_VAL]], i64* [[UB:%[^,]+]], 138 // CHECK: store i64 [[ST_VAL]], i64* [[ST:%[^,]+]], 139 // CHECK: store i32 [[LITER_VAL]], i32* [[LITER:%[^,]+]], 140 // CHECK: [[LB_VAL:%.+]] = load i64, i64* [[LB]], 141 // CHECK: store i64 [[LB_VAL]], i64* [[CNT:%.+]], 142 // CHECK: br label 143 // CHECK-NOT: !llvm.mem.parallel_loop_access 144 // CHECK: br label %{{.*}}!llvm.loop 145 // CHECK: ret i32 0 146 147 // CHECK-LABEL: @_ZN1SC2Ei 148 struct S { 149 int a; 150 S(int c) { 151 // CHECK: [[GTID:%.+]] = call i32 @__kmpc_global_thread_num(%ident_t* [[DEFLOC:@.+]]) 152 // CHECK: [[TASKV:%.+]] = call i8* @__kmpc_omp_task_alloc(%ident_t* [[DEFLOC]], i32 [[GTID]], i32 1, i64 64, i64 16, i32 (i32, i8*)* bitcast (i32 (i32, [[TDP_TY:%.+]]*)* [[TASK4:@.+]] to i32 (i32, i8*)*)) 153 // CHECK: [[TASK:%.+]] = bitcast i8* [[TASKV]] to [[TDP_TY]]* 154 // CHECK: [[TASK_DATA:%.+]] = getelementptr inbounds [[TDP_TY]], [[TDP_TY]]* [[TASK]], i32 0, i32 0 155 // CHECK: getelementptr inbounds [[TD_TY:%.+]], [[TD_TY]]* [[TASK_DATA]], i32 0, i32 3 156 // CHECK: store i32 (i32, i8*)* null, i32 (i32, i8*)** %{{.+}} 157 // CHECK: [[DOWN:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* [[TASK_DATA]], i32 0, i32 4 158 // CHECK: store i64 0, i64* [[DOWN]], 159 // CHECK: [[UP:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* [[TASK_DATA]], i32 0, i32 5 160 // CHECK: store i64 %{{.+}}, i64* [[UP]], 161 // CHECK: [[ST:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* [[TASK_DATA]], i32 0, i32 6 162 // CHECK: store i64 1, i64* [[ST]], 163 // CHECK: [[ST_VAL:%.+]] = load i64, i64* [[ST]], 164 // CHECK: [[NUM_TASKS:%.+]] = zext i32 %{{.+}} to i64 165 // 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) 166 #pragma omp taskloop simd shared(c) num_tasks(a) simdlen(64) safelen(8) 167 for (a = 0; a < c; ++a) 168 ; 169 } 170 } s(1); 171 172 // CHECK: define internal i32 [[TASK4]]( 173 // CHECK: [[DOWN:%.+]] = getelementptr inbounds [[TD_TY:%.+]], [[TD_TY]]* %{{.+}}, i32 0, i32 4 174 // CHECK: [[DOWN_VAL:%.+]] = load i64, i64* [[DOWN]], 175 // CHECK: [[UP:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* %{{.+}}, i32 0, i32 5 176 // CHECK: [[UP_VAL:%.+]] = load i64, i64* [[UP]], 177 // CHECK: [[ST:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* %{{.+}}, i32 0, i32 6 178 // CHECK: [[ST_VAL:%.+]] = load i64, i64* [[ST]], 179 // CHECK: [[LITER:%.+]] = getelementptr inbounds [[TD_TY]], [[TD_TY]]* %{{.+}}, i32 0, i32 7 180 // CHECK: [[LITER_VAL:%.+]] = load i32, i32* [[LITER]], 181 // CHECK: store i64 [[DOWN_VAL]], i64* [[LB:%[^,]+]], 182 // CHECK: store i64 [[UP_VAL]], i64* [[UB:%[^,]+]], 183 // CHECK: store i64 [[ST_VAL]], i64* [[ST:%[^,]+]], 184 // CHECK: store i32 [[LITER_VAL]], i32* [[LITER:%[^,]+]], 185 // CHECK: [[LB_VAL:%.+]] = load i64, i64* [[LB]], 186 // CHECK: [[LB_I32:%.+]] = trunc i64 [[LB_VAL]] to i32 187 // CHECK: store i32 [[LB_I32]], i32* [[CNT:%.+]], 188 // CHECK: br label 189 // CHECK: [[VAL:%.+]] = load i32, i32* [[CNT]], 190 // CHECK: [[VAL_I64:%.+]] = sext i32 [[VAL]] to i64 191 // CHECK: [[UB_VAL:%.+]] = load i64, i64* [[UB]], 192 // CHECK: [[CMP:%.+]] = icmp ule i64 [[VAL_I64]], [[UB_VAL]] 193 // CHECK: br i1 [[CMP]], label %{{.+}}, label %{{.+}} 194 // CHECK: load i32, i32* % 195 // CHECK-NOT: !llvm.mem.parallel_loop_access 196 // CHECK: store i32 % 197 // CHECK-NOT: !llvm.mem.parallel_loop_access 198 // CHECK: load i32, i32* % 199 // CHECK-NOT: !llvm.mem.parallel_loop_access 200 // CHECK: add nsw i32 %{{.+}}, 1 201 // CHECK: store i32 %{{.+}}, i32* % 202 // CHECK-NOT: !llvm.mem.parallel_loop_access 203 // CHECK: br label %{{.*}}!llvm.loop 204 // CHECK: ret i32 0 205 206 // CHECK: !{!"llvm.loop.vectorize.enable", i1 true} 207 // CHECK: !{!"llvm.loop.vectorize.width", i32 4} 208 // CHECK: !{!"llvm.loop.vectorize.width", i32 32} 209 // CHECK: !{!"llvm.loop.vectorize.width", i32 64} 210 211 #endif 212