1 // expected-no-diagnostics 2 #ifndef HEADER 3 #define HEADER 4 // Test host codegen. 5 // RUN: %clang_cc1 -DCK1 -verify -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix CK1 --check-prefix CK1-64 --check-prefix HCK1 --check-prefix HCK1-64 6 // RUN: %clang_cc1 -DCK1 -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s 7 // RUN: %clang_cc1 -DCK1 -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix CK1 --check-prefix CK1-64 --check-prefix HCK1 --check-prefix HCK1-64 8 // RUN: %clang_cc1 -DCK1 -verify -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix CK1 --check-prefix CK1-32 --check-prefix HCK1 --check-prefix HCK1-64 9 // RUN: %clang_cc1 -DCK1 -fopenmp -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s 10 // RUN: %clang_cc1 -DCK1 -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix CK1 --check-prefix CK1-32 --check-prefix HCK1 --check-prefix HCK1-64 11 12 // RUN: %clang_cc1 -DCK1 -verify -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix SIMD-ONLY1 13 // RUN: %clang_cc1 -DCK1 -fopenmp-simd -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s 14 // RUN: %clang_cc1 -DCK1 -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix SIMD-ONLY1 15 // RUN: %clang_cc1 -DCK1 -verify -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix SIMD-ONLY1 16 // RUN: %clang_cc1 -DCK1 -fopenmp-simd -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s 17 // RUN: %clang_cc1 -DCK1 -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix SIMD-ONLY1 18 19 // Test target codegen - host bc file has to be created first. (no significant differences with host version of target region) 20 // RUN: %clang_cc1 -DCK1 -verify -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm-bc %s -o %t-ppc-host.bc 21 // RUN: %clang_cc1 -DCK1 -verify -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -fopenmp-is-device -fopenmp-host-ir-file-path %t-ppc-host.bc -o - | FileCheck %s --check-prefix CK1 --check-prefix CK1-64 --check-prefix TCK1 --check-prefix TCK1-64 22 // RUN: %clang_cc1 -DCK1 -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -fopenmp-is-device -fopenmp-host-ir-file-path %t-ppc-host.bc -o %t %s 23 // RUN: %clang_cc1 -DCK1 -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -fopenmp-is-device -fopenmp-host-ir-file-path %t-ppc-host.bc -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix CK1 --check-prefix CK1-64 --check-prefix TCK1 --check-prefix TCK1-64 24 // RUN: %clang_cc1 -DCK1 -verify -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm-bc %s -o %t-x86-host.bc 25 // RUN: %clang_cc1 -DCK1 -verify -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -fopenmp-is-device -fopenmp-host-ir-file-path %t-x86-host.bc -o - | FileCheck %s --check-prefix CK1 --check-prefix CK1-32 --check-prefix TCK1 --check-prefix TCK1-32 26 // RUN: %clang_cc1 -DCK1 -fopenmp -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -fopenmp-is-device -fopenmp-host-ir-file-path %t-x86-host.bc -o %t %s 27 // RUN: %clang_cc1 -DCK1 -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -fopenmp-is-device -fopenmp-host-ir-file-path %t-x86-host.bc -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix CK1 --check-prefix CK1-32 --check-prefix TCK1 --check-prefix TCK1-32 28 29 // RUN: %clang_cc1 -DCK1 -verify -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm-bc %s -o %t-ppc-host.bc 30 // RUN: %clang_cc1 -DCK1 -verify -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -fopenmp-is-device -fopenmp-host-ir-file-path %t-ppc-host.bc -o - | FileCheck %s --check-prefix SIMD-ONLY1 31 // RUN: %clang_cc1 -DCK1 -fopenmp-simd -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -fopenmp-is-device -fopenmp-host-ir-file-path %t-ppc-host.bc -o %t %s 32 // RUN: %clang_cc1 -DCK1 -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -fopenmp-is-device -fopenmp-host-ir-file-path %t-ppc-host.bc -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix SIMD-ONLY1 33 // RUN: %clang_cc1 -DCK1 -verify -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm-bc %s -o %t-x86-host.bc 34 // RUN: %clang_cc1 -DCK1 -verify -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -fopenmp-is-device -fopenmp-host-ir-file-path %t-x86-host.bc -o - | FileCheck %s --check-prefix SIMD-ONLY1 35 // RUN: %clang_cc1 -DCK1 -fopenmp-simd -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -fopenmp-is-device -fopenmp-host-ir-file-path %t-x86-host.bc -o %t %s 36 // RUN: %clang_cc1 -DCK1 -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -fopenmp-is-device -fopenmp-host-ir-file-path %t-x86-host.bc -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix SIMD-ONLY1 37 // SIMD-ONLY1-NOT: {{__kmpc|__tgt}} 38 39 #ifdef CK1 40 41 // HCK1: define{{.*}} i32 @{{.+}}target_teams_fun{{.*}}( 42 int target_teams_fun(int *g){ 43 int n = 1000; 44 int a[1000]; 45 int te = n / 128; 46 int th = 128; 47 // discard n_addr 48 // HCK1: alloca i32, 49 // HCK1: [[TE:%.+]] alloca i32, 50 // HCK1: [[TH:%.+]] = alloca i32, 51 // discard capture expressions for te and th 52 // HCK1: = alloca i32, 53 // HCK1: = alloca i32, 54 // HCK1: [[N_CAST:%.+]] = alloca i{{32|64}}, 55 // HCK1: [[TE_CAST:%.+]] = alloca i{{32|64}}, 56 // HCK1: [[TH_CAST:%.+]] = alloca i{{32|64}}, 57 // HCK1: [[N_PAR:%.+]] = load{{.+}}, {{.+}} [[N_CAST]], 58 // HCK1: [[TE_PAR:%.+]] = load{{.+}}, {{.+}} [[TE_CAST]], 59 // HCK1: [[TH_PAR:%.+]] = load{{.+}}, {{.+}} [[TH_CAST]], 60 // HCK1: call i32 @__tgt_target_teams(i64 -1, i8* @{{[^,]+}}, i32 4, i8** %{{[^,]+}}, i8** %{{[^,]+}}, 61 62 // HCK1: call void @[[OFFL1:.+]](i{{32|64}} [[N_PAR]], {{.+}}, i{{32|64}} [[TE_PAR]], i{{32|64}} [[TH_PAR]]) 63 #pragma omp target teams distribute parallel for num_teams(te), thread_limit(th) 64 for(int i = 0; i < n; i++) { 65 a[i] = 0; 66 #pragma omp cancel for 67 } 68 69 // HCK1: call i32 @__tgt_target_teams(i64 -1, i8* @{{[^,]+}}, i32 3, i8** %{{[^,]+}}, i8** %{{[^,]+}}, i{{64|32}}* {{.+}}@{{[^,]+}}, i32 0, i32 0), i64* {{.+}}@{{[^,]+}}, i32 0, i32 0), 70 // HCK1: call void @[[OFFL2:.+]](i{{64|32}} %{{.+}}) 71 {{{ 72 #pragma omp target teams distribute parallel for is_device_ptr(g) 73 for(int i = 0; i < n; i++) { 74 a[i] = g[0]; 75 } 76 }}} 77 78 // outlined target regions 79 // HCK1: define internal void @[[OFFL1]](i{{32|64}} [[N_ARG:%.+]], i{{32|64}} [[TE_ARG:%.+]], i{{32|64}} [[TH_ARG:%.+]]) 80 // TCK1: define weak void @{{.+}}target_teams_fun{{.*}}(i{{32|64}} [[N_ARG:%.+]], {{.+}}, i{{32|64}} [[TE_ARG:%.+]], i{{32|64}} [[TH_ARG:%.+]]) 81 // CK1: [[N_ADDR:%.+]] = alloca i{{32|64}}, 82 // CK1: [[TE_ADDR:%.+]] = alloca i{{32|64}}, 83 // CK1: [[TH_ADDR:%.+]] = alloca i{{32|64}}, 84 // TCK1: store {{.+}} [[N_ARG]], {{.+}} [[N_ADDR]], 85 // CK1: store{{.+}} [[TE_ARG]], {{.+}} [[TE_ADDR]], 86 // CK1: store{{.+}} [[TH_ARG]], {{.+}} [[TH_ADDR]], 87 // CK1-64: [[TE_CONV:%.+]] = bitcast{{.+}} [[TE_ADDR]] to 88 // CK1-64: [[TH_CONV:%.+]] = bitcast{{.+}} [[TH_ADDR]] to 89 // CK1-64: [[TE_VAL:%.+]] = load i32, i32* [[TE_CONV]], 90 // CK1-64: [[TH_VAL:%.+]] = load i32, i32* [[TH_CONV]], 91 // CK1-32: [[TE_VAL:%.+]] = load i32, i32* [[TE_ADDR]], 92 // CK1-32: [[TH_VAL:%.+]] = load i32, i32* [[TH_ADDR]], 93 // CK1: {{%.+}} = call i32 @__kmpc_push_num_teams({{.+}}, {{.+}}, i32 [[TE_VAL]], i32 [[TH_VAL]]) 94 // CK1: call void {{.+}} @__kmpc_fork_teams({{.+}}, i32 2, {{.+}} @[[OUTL1:.+]] to {{.+}}, {{.+}}, {{.+}}) 95 // CK1: ret void 96 97 // CK1: define internal void @[[OUTL1]]({{.+}}) 98 // CK1: call void @__kmpc_for_static_init_4( 99 // CK1: call void {{.+}} @__kmpc_fork_call( 100 // CK1: call void @__kmpc_for_static_fini( 101 // CK1: ret void 102 103 // HCK1: define internal void @[[OFFL2]]( 104 // TCK1: define weak void @{{.+}}target_teams_fun{{.+}}( 105 // CK1: call void {{.+}} @__kmpc_fork_teams({{.+}}, i32 3, {{.+}} @[[OUTL2:.+]] to {{.+}}, {{.+}}, {{.+}}) 106 // CK1: ret void 107 108 // CK1: define internal void @[[OUTL2]]({{.+}}) 109 // CK1: call void @__kmpc_for_static_init_4( 110 // CK1: call void {{.+}} @__kmpc_fork_call( 111 // CK1: call void @__kmpc_for_static_fini( 112 // CK1: ret void 113 114 return a[0]; 115 } 116 117 #endif // CK1 118 #endif // HEADER 119