1 // RUN: %clang_cc1 -verify -fopenmp=libiomp5 -x c++ -triple x86_64-unknown-unknown -emit-llvm %s -fexceptions -fcxx-exceptions -o - | FileCheck %s 2 // RUN: %clang_cc1 -fopenmp=libiomp5 -x c++ -std=c++11 -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -emit-pch -o %t %s 3 // RUN: %clang_cc1 -fopenmp=libiomp5 -x c++ -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -g -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s 4 // 5 // expected-no-diagnostics 6 #ifndef HEADER 7 #define HEADER 8 9 // CHECK: [[IDENT_T_TY:%.+]] = type { i32, i32, i32, i32, i8* } 10 // CHECK-LABEL: define {{.*void}} @{{.*}}without_schedule_clause{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}}) 11 void without_schedule_clause(float *a, float *b, float *c, float *d) { 12 // CHECK: [[GTID:%.+]] = call i32 @__kmpc_global_thread_num([[IDENT_T_TY]]* [[DEFAULT_LOC:[@%].+]]) 13 #pragma omp for 14 // CHECK: call void @__kmpc_for_static_init_4([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]], i32 34, i32* [[IS_LAST:%[^,]+]], i32* [[OMP_LB:%[^,]+]], i32* [[OMP_UB:%[^,]+]], i32* [[OMP_ST:%[^,]+]], i32 1, i32 1) 15 // UB = min(UB, GlobalUB) 16 // CHECK-NEXT: [[UB:%.+]] = load i32* [[OMP_UB]] 17 // CHECK-NEXT: [[UBCMP:%.+]] = icmp sgt i32 [[UB]], 4571423 18 // CHECK-NEXT: br i1 [[UBCMP]], label [[UB_TRUE:%[^,]+]], label [[UB_FALSE:%[^,]+]] 19 // CHECK: [[UBRESULT:%.+]] = phi i32 [ 4571423, [[UB_TRUE]] ], [ [[UBVAL:%[^,]+]], [[UB_FALSE]] ] 20 // CHECK-NEXT: store i32 [[UBRESULT]], i32* [[OMP_UB]] 21 // CHECK-NEXT: [[LB:%.+]] = load i32* [[OMP_LB]] 22 // CHECK-NEXT: store i32 [[LB]], i32* [[OMP_IV:[^,]+]] 23 // Loop header 24 // CHECK: [[IV:%.+]] = load i32* [[OMP_IV]] 25 // CHECK-NEXT: [[UB:%.+]] = load i32* [[OMP_UB]] 26 // CHECK-NEXT: [[CMP:%.+]] = icmp sle i32 [[IV]], [[UB]] 27 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]] 28 for (int i = 33; i < 32000000; i += 7) { 29 // CHECK: [[LOOP1_BODY]] 30 // Start of body: calculate i from IV: 31 // CHECK: [[IV1_1:%.+]] = load i32* [[OMP_IV]] 32 // CHECK-NEXT: [[CALC_I_1:%.+]] = mul nsw i32 [[IV1_1]], 7 33 // CHECK-NEXT: [[CALC_I_2:%.+]] = add nsw i32 33, [[CALC_I_1]] 34 // CHECK-NEXT: store i32 [[CALC_I_2]], i32* [[LC_I:.+]] 35 // ... loop body ... 36 // End of body: store into a[i]: 37 // CHECK: store float [[RESULT:%.+]], float* {{%.+}} 38 a[i] = b[i] * c[i] * d[i]; 39 // CHECK: [[IV1_2:%.+]] = load i32* [[OMP_IV]]{{.*}} 40 // CHECK-NEXT: [[ADD1_2:%.+]] = add nsw i32 [[IV1_2]], 1 41 // CHECK-NEXT: store i32 [[ADD1_2]], i32* [[OMP_IV]] 42 // CHECK-NEXT: br label %{{.+}} 43 } 44 // CHECK: [[LOOP1_END]] 45 // CHECK: call void @__kmpc_for_static_fini([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]]) 46 // CHECK: call {{.+}} @__kmpc_cancel_barrier([[IDENT_T_TY]]* [[DEFAULT_LOC_BARRIER:[@%].+]], i32 [[GTID]]) 47 // CHECK: ret void 48 } 49 50 // CHECK-LABEL: define {{.*void}} @{{.*}}static_not_chunked{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}}) 51 void static_not_chunked(float *a, float *b, float *c, float *d) { 52 // CHECK: [[GTID:%.+]] = call i32 @__kmpc_global_thread_num([[IDENT_T_TY]]* [[DEFAULT_LOC:[@%].+]]) 53 #pragma omp for schedule(static) 54 // CHECK: call void @__kmpc_for_static_init_4([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]], i32 34, i32* [[IS_LAST:%[^,]+]], i32* [[OMP_LB:%[^,]+]], i32* [[OMP_UB:%[^,]+]], i32* [[OMP_ST:%[^,]+]], i32 1, i32 1) 55 // UB = min(UB, GlobalUB) 56 // CHECK-NEXT: [[UB:%.+]] = load i32* [[OMP_UB]] 57 // CHECK-NEXT: [[UBCMP:%.+]] = icmp sgt i32 [[UB]], 4571423 58 // CHECK-NEXT: br i1 [[UBCMP]], label [[UB_TRUE:%[^,]+]], label [[UB_FALSE:%[^,]+]] 59 // CHECK: [[UBRESULT:%.+]] = phi i32 [ 4571423, [[UB_TRUE]] ], [ [[UBVAL:%[^,]+]], [[UB_FALSE]] ] 60 // CHECK-NEXT: store i32 [[UBRESULT]], i32* [[OMP_UB]] 61 // CHECK-NEXT: [[LB:%.+]] = load i32* [[OMP_LB]] 62 // CHECK-NEXT: store i32 [[LB]], i32* [[OMP_IV:[^,]+]] 63 // Loop header 64 // CHECK: [[IV:%.+]] = load i32* [[OMP_IV]] 65 // CHECK-NEXT: [[UB:%.+]] = load i32* [[OMP_UB]] 66 // CHECK-NEXT: [[CMP:%.+]] = icmp sle i32 [[IV]], [[UB]] 67 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]] 68 for (int i = 32000000; i > 33; i += -7) { 69 // CHECK: [[LOOP1_BODY]] 70 // Start of body: calculate i from IV: 71 // CHECK: [[IV1_1:%.+]] = load i32* [[OMP_IV]] 72 // CHECK-NEXT: [[CALC_I_1:%.+]] = mul nsw i32 [[IV1_1]], 7 73 // CHECK-NEXT: [[CALC_I_2:%.+]] = sub nsw i32 32000000, [[CALC_I_1]] 74 // CHECK-NEXT: store i32 [[CALC_I_2]], i32* [[LC_I:.+]] 75 // ... loop body ... 76 // End of body: store into a[i]: 77 // CHECK: store float [[RESULT:%.+]], float* {{%.+}} 78 a[i] = b[i] * c[i] * d[i]; 79 // CHECK: [[IV1_2:%.+]] = load i32* [[OMP_IV]]{{.*}} 80 // CHECK-NEXT: [[ADD1_2:%.+]] = add nsw i32 [[IV1_2]], 1 81 // CHECK-NEXT: store i32 [[ADD1_2]], i32* [[OMP_IV]] 82 // CHECK-NEXT: br label %{{.+}} 83 } 84 // CHECK: [[LOOP1_END]] 85 // CHECK: call void @__kmpc_for_static_fini([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]]) 86 // CHECK: call {{.+}} @__kmpc_cancel_barrier([[IDENT_T_TY]]* [[DEFAULT_LOC_BARRIER:[@%].+]], i32 [[GTID]]) 87 // CHECK: ret void 88 } 89 90 // CHECK-LABEL: define {{.*void}} @{{.*}}static_chunked{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}}) 91 void static_chunked(float *a, float *b, float *c, float *d) { 92 // CHECK: [[GTID:%.+]] = call i32 @__kmpc_global_thread_num([[IDENT_T_TY]]* [[DEFAULT_LOC:[@%].+]]) 93 #pragma omp for schedule(static, 5) 94 // CHECK: call void @__kmpc_for_static_init_4u([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]], i32 33, i32* [[IS_LAST:%[^,]+]], i32* [[OMP_LB:%[^,]+]], i32* [[OMP_UB:%[^,]+]], i32* [[OMP_ST:%[^,]+]], i32 1, i32 5) 95 // UB = min(UB, GlobalUB) 96 // CHECK: [[UB:%.+]] = load i32* [[OMP_UB]] 97 // CHECK-NEXT: [[UBCMP:%.+]] = icmp ugt i32 [[UB]], 16908288 98 // CHECK-NEXT: br i1 [[UBCMP]], label [[UB_TRUE:%[^,]+]], label [[UB_FALSE:%[^,]+]] 99 // CHECK: [[UBRESULT:%.+]] = phi i32 [ 16908288, [[UB_TRUE]] ], [ [[UBVAL:%[^,]+]], [[UB_FALSE]] ] 100 // CHECK-NEXT: store i32 [[UBRESULT]], i32* [[OMP_UB]] 101 // CHECK-NEXT: [[LB:%.+]] = load i32* [[OMP_LB]] 102 // CHECK-NEXT: store i32 [[LB]], i32* [[OMP_IV:[^,]+]] 103 104 // Outer loop header 105 // CHECK: [[O_IV:%.+]] = load i32* [[OMP_IV]] 106 // CHECK-NEXT: [[O_UB:%.+]] = load i32* [[OMP_UB]] 107 // CHECK-NEXT: [[O_CMP:%.+]] = icmp ule i32 [[O_IV]], [[O_UB]] 108 // CHECK-NEXT: br i1 [[O_CMP]], label %[[O_LOOP1_BODY:[^,]+]], label %[[O_LOOP1_END:[^,]+]] 109 110 // Loop header 111 // CHECK: [[O_LOOP1_BODY]] 112 // CHECK: [[IV:%.+]] = load i32* [[OMP_IV]] 113 // CHECK-NEXT: [[UB:%.+]] = load i32* [[OMP_UB]] 114 // CHECK-NEXT: [[CMP:%.+]] = icmp ule i32 [[IV]], [[UB]] 115 // CHECK-NEXT: br i1 [[CMP]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]] 116 for (unsigned i = 131071; i <= 2147483647; i += 127) { 117 // CHECK: [[LOOP1_BODY]] 118 // Start of body: calculate i from IV: 119 // CHECK: [[IV1_1:%.+]] = load i32* [[OMP_IV]] 120 // CHECK-NEXT: [[CALC_I_1:%.+]] = mul i32 [[IV1_1]], 127 121 // CHECK-NEXT: [[CALC_I_2:%.+]] = add i32 131071, [[CALC_I_1]] 122 // CHECK-NEXT: store i32 [[CALC_I_2]], i32* [[LC_I:.+]] 123 // ... loop body ... 124 // End of body: store into a[i]: 125 // CHECK: store float [[RESULT:%.+]], float* {{%.+}} 126 a[i] = b[i] * c[i] * d[i]; 127 // CHECK: [[IV1_2:%.+]] = load i32* [[OMP_IV]]{{.*}} 128 // CHECK-NEXT: [[ADD1_2:%.+]] = add i32 [[IV1_2]], 1 129 // CHECK-NEXT: store i32 [[ADD1_2]], i32* [[OMP_IV]] 130 // CHECK-NEXT: br label %{{.+}} 131 } 132 // CHECK: [[LOOP1_END]] 133 // Update the counters, adding stride 134 // CHECK: [[LB:%.+]] = load i32* [[OMP_LB]] 135 // CHECK-NEXT: [[ST:%.+]] = load i32* [[OMP_ST]] 136 // CHECK-NEXT: [[ADD_LB:%.+]] = add i32 [[LB]], [[ST]] 137 // CHECK-NEXT: store i32 [[ADD_LB]], i32* [[OMP_LB]] 138 // CHECK-NEXT: [[UB:%.+]] = load i32* [[OMP_UB]] 139 // CHECK-NEXT: [[ST:%.+]] = load i32* [[OMP_ST]] 140 // CHECK-NEXT: [[ADD_UB:%.+]] = add i32 [[UB]], [[ST]] 141 // CHECK-NEXT: store i32 [[ADD_UB]], i32* [[OMP_UB]] 142 143 // CHECK: [[O_LOOP1_END]] 144 // CHECK: call void @__kmpc_for_static_fini([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]]) 145 // CHECK: call {{.+}} @__kmpc_cancel_barrier([[IDENT_T_TY]]* [[DEFAULT_LOC_BARRIER:[@%].+]], i32 [[GTID]]) 146 // CHECK: ret void 147 } 148 149 void parallel_for(float *a) { 150 #pragma omp parallel 151 #pragma omp for schedule(static, 5) 152 // CHECK-NOT: __kmpc_global_thread_num 153 for (unsigned i = 131071; i <= 2147483647; i += 127) 154 a[i] += i; 155 } 156 157 #endif // HEADER 158 159