1 // Test target codegen - host bc file has to be created first.
2 // RUN: %clang_cc1 -verify -fopenmp -fopenmp-version=45 -fopenmp-cuda-mode -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=nvptx64-nvidia-cuda -emit-llvm-bc %s -o %t-ppc-host.bc
3 // RUN: %clang_cc1 -verify -fopenmp -fopenmp-version=45 -fopenmp-cuda-mode -x c++ -triple nvptx64-unknown-unknown -fopenmp-targets=nvptx64-nvidia-cuda -emit-llvm %s -fopenmp-is-device -fopenmp-host-ir-file-path %t-ppc-host.bc -o - | FileCheck %s --check-prefix CHECK --check-prefix CHECK-64
4 // RUN: %clang_cc1 -verify -fopenmp -fopenmp-version=45 -fopenmp-cuda-mode -x c++ -triple i386-unknown-unknown -fopenmp-targets=nvptx-nvidia-cuda -emit-llvm-bc %s -o %t-x86-host.bc
5 // RUN: %clang_cc1 -verify -fopenmp -fopenmp-version=45 -fopenmp-cuda-mode -x c++ -triple nvptx-unknown-unknown -fopenmp-targets=nvptx-nvidia-cuda -emit-llvm %s -fopenmp-is-device -fopenmp-host-ir-file-path %t-x86-host.bc -o - | FileCheck %s --check-prefix CHECK --check-prefix CHECK-32
6 // RUN: %clang_cc1 -verify -fopenmp -fopenmp-version=45 -fopenmp-cuda-mode -fexceptions -fcxx-exceptions -x c++ -triple nvptx-unknown-unknown -fopenmp-targets=nvptx-nvidia-cuda -emit-llvm %s -fopenmp-is-device -fopenmp-host-ir-file-path %t-x86-host.bc -o - | FileCheck %s --check-prefix CHECK --check-prefix CHECK-32
7
8 // RUN: %clang_cc1 -verify -fopenmp -fopenmp-cuda-mode -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=nvptx64-nvidia-cuda -emit-llvm-bc %s -o %t-ppc-host.bc
9 // RUN: %clang_cc1 -verify -fopenmp -fopenmp-cuda-mode -x c++ -triple nvptx64-unknown-unknown -fopenmp-targets=nvptx64-nvidia-cuda -emit-llvm %s -fopenmp-is-device -fopenmp-host-ir-file-path %t-ppc-host.bc -o - | FileCheck %s --check-prefix CHECK --check-prefix CHECK-64
10 // RUN: %clang_cc1 -verify -fopenmp -fopenmp-cuda-mode -x c++ -triple i386-unknown-unknown -fopenmp-targets=nvptx-nvidia-cuda -emit-llvm-bc %s -o %t-x86-host.bc
11 // RUN: %clang_cc1 -verify -fopenmp -fopenmp-cuda-mode -x c++ -triple nvptx-unknown-unknown -fopenmp-targets=nvptx-nvidia-cuda -emit-llvm %s -fopenmp-is-device -fopenmp-host-ir-file-path %t-x86-host.bc -o - | FileCheck %s --check-prefix CHECK --check-prefix CHECK-32
12 // RUN: %clang_cc1 -verify -fopenmp -fopenmp-cuda-mode -fexceptions -fcxx-exceptions -x c++ -triple nvptx-unknown-unknown -fopenmp-targets=nvptx-nvidia-cuda -emit-llvm %s -fopenmp-is-device -fopenmp-host-ir-file-path %t-x86-host.bc -o - | FileCheck %s --check-prefix CHECK --check-prefix CHECK-32
13
14 // expected-no-diagnostics
15 #ifndef HEADER
16 #define HEADER
17
18 // Check that the execution mode of all 2 target regions on the gpu is set to NonSPMD Mode.
19 // CHECK-DAG: {{@__omp_offloading_.+l32}}_exec_mode = weak constant i8 0
20 // CHECK-DAG: {{@__omp_offloading_.+l37}}_exec_mode = weak constant i8 0
21 // CHECK-DAG: {{@__omp_offloading_.+l42}}_exec_mode = weak constant i8 0
22 // CHECK-DAG: {{@__omp_offloading_.+l47}}_exec_mode = weak constant i8 0
23
24 #define N 1000
25
26 template<typename tx>
ftemplate(int n)27 tx ftemplate(int n) {
28 tx a[N];
29 short aa[N];
30 tx b[10];
31
32 #pragma omp target simd
33 for(int i = 0; i < n; i++) {
34 a[i] = 1;
35 }
36
37 #pragma omp target simd
38 for (int i = 0; i < n; i++) {
39 aa[i] += 1;
40 }
41
42 #pragma omp target simd
43 for(int i = 0; i < 10; i++) {
44 b[i] += 1;
45 }
46
47 #pragma omp target simd reduction(+:n)
48 for(int i = 0; i < 10; i++) {
49 b[i] += 1;
50 }
51
52 return a[0];
53 }
54
bar(int n)55 int bar(int n){
56 int a = 0;
57
58 a += ftemplate<int>(n);
59
60 return a;
61 }
62
63 // CHECK-LABEL: define {{.*}}void {{@__omp_offloading_.+l32}}(
64 // CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 false)
65 // CHECK-NOT: call void @__kmpc_for_static_init
66 // CHECK-NOT: call void @__kmpc_for_static_fini
67 // CHECK: call void @__kmpc_target_deinit({{.*}}, i1 true, i1 false)
68 // CHECK: ret void
69
70 // CHECK-LABEL: define {{.*}}void {{@__omp_offloading_.+l37}}(
71 // CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 false)
72 // CHECK-NOT: call void @__kmpc_for_static_init
73 // CHECK-NOT: call void @__kmpc_for_static_fini
74 // CHECK: call void @__kmpc_target_deinit({{.*}}, i1 true, i1 false)
75 // CHECK: ret void
76
77 // CHECK-LABEL: define {{.*}}void {{@__omp_offloading_.+l42}}(
78 // CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 false)
79 // CHECK-NOT: call void @__kmpc_for_static_init
80 // CHECK-NOT: call void @__kmpc_for_static_fini
81 // CHECK: call void @__kmpc_target_deinit({{.*}}, i1 true, i1 false)
82 // CHECK: ret void
83
84 // CHECK-LABEL: define {{.*}}void {{@__omp_offloading_.+l47}}(
85 // CHECK: call i32 @__kmpc_target_init({{.*}}, i1 true, i1 false, i1 false)
86 // CHECK-NOT: call void @__kmpc_for_static_init
87 // CHECK-NOT: call void @__kmpc_for_static_fini
88 // CHECK-NOT: call void @__kmpc_nvptx_end_reduce_nowait(
89 // CHECK: call void @__kmpc_target_deinit({{.*}}, i1 true, i1 false)
90 // CHECK: ret void
91
92 #endif
93