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 - -disable-llvm-optzns | 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 - -disable-llvm-optzns | 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 - -disable-llvm-optzns | 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 - -disable-llvm-optzns | 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 - -disable-llvm-optzns | 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 - -disable-llvm-optzns | 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 SPMD Mode.
19 // CHECK-DAG: {{@__omp_offloading_.+l33}}_exec_mode = weak constant i8 0
20 // CHECK-DAG: {{@__omp_offloading_.+l38}}_exec_mode = weak constant i8 0
21
22 template<typename tx>
ftemplate(int n)23 tx ftemplate(int n) {
24 tx a = 0;
25 short aa = 0;
26 tx b[10];
27
28 #pragma omp target parallel if(target: 0)
29 {
30 a += 1;
31 }
32
33 #pragma omp target parallel map(tofrom: aa)
34 {
35 aa += 1;
36 }
37
38 #pragma omp target parallel map(tofrom:a, aa, b) if(target: n>40)
39 {
40 a += 1;
41 aa += 1;
42 b[2] += 1;
43 }
44
45 return a;
46 }
47
bar(int n)48 int bar(int n){
49 int a = 0;
50
51 a += ftemplate<int>(n);
52
53 return a;
54 }
55
56 // CHECK-NOT: define {{.*}}void {{@__omp_offloading_.+template.+l17}}
57
58 // CHECK-LABEL: define {{.*}}void {{@__omp_offloading_.+template.+l33}}(
59 // CHECK: [[AA_ADDR:%.+]] = alloca i16*, align
60 // CHECK-NOT: call i8* @__kmpc_data_sharing_push_stack
61 // CHECK: store i16* {{%.+}}, i16** [[AA_ADDR]], align
62 // CHECK: [[AA:%.+]] = load i16*, i16** [[AA_ADDR]], align
63 // CHECK: [[THREAD_LIMIT:%.+]] = call i32 @llvm.nvvm.read.ptx.sreg.ntid.x()
64 // CHECK: call void @__kmpc_spmd_kernel_init(i32 [[THREAD_LIMIT]], i16 1)
65 // CHECK: call void @__kmpc_data_sharing_init_stack_spmd
66 // CHECK: br label {{%?}}[[EXEC:.+]]
67 //
68 // CHECK: [[EXEC]]
69 // CHECK: {{call|invoke}} void [[OP1:@.+]]({{.+}}, {{.+}}, i16* [[AA]])
70 // CHECK: br label {{%?}}[[DONE:.+]]
71 //
72 // CHECK: [[DONE]]
73 // CHECK: call void @__kmpc_spmd_kernel_deinit_v2(i16 1)
74 // CHECK: br label {{%?}}[[EXIT:.+]]
75 //
76 // CHECK: [[EXIT]]
77 // CHECK: ret void
78 // CHECK: }
79
80 // CHECK: define internal void [[OP1]](i32* noalias %.global_tid., i32* noalias %.bound_tid., i16* {{[^%]*}}[[ARG:%.+]])
81 // CHECK: = alloca i32*, align
82 // CHECK: = alloca i32*, align
83 // CHECK: [[AA_ADDR:%.+]] = alloca i16*, align
84 // CHECK: store i16* [[ARG]], i16** [[AA_ADDR]], align
85 // CHECK: [[AA:%.+]] = load i16*, i16** [[AA_ADDR]], align
86 // CHECK: [[VAL:%.+]] = load i16, i16* [[AA]], align
87 // CHECK: store i16 {{%.+}}, i16* [[AA]], align
88 // CHECK: ret void
89 // CHECK: }
90
91 // CHECK-LABEL: define {{.*}}void {{@__omp_offloading_.+template.+l38}}(
92 // CHECK: [[A_ADDR:%.+]] = alloca i32*, align
93 // CHECK: [[AA_ADDR:%.+]] = alloca i16*, align
94 // CHECK: [[B_ADDR:%.+]] = alloca [10 x i32]*, align
95 // CHECK: store i32* {{%.+}}, i32** [[A_ADDR]], align
96 // CHECK: store i16* {{%.+}}, i16** [[AA_ADDR]], align
97 // CHECK: store [10 x i32]* {{%.+}}, [10 x i32]** [[B_ADDR]], align
98 // CHECK: [[A:%.+]] = load i32*, i32** [[A_ADDR]], align
99 // CHECK: [[AA:%.+]] = load i16*, i16** [[AA_ADDR]], align
100 // CHECK: [[B:%.+]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align
101 // CHECK: [[THREAD_LIMIT:%.+]] = call i32 @llvm.nvvm.read.ptx.sreg.ntid.x()
102 // CHECK: call void @__kmpc_spmd_kernel_init(i32 [[THREAD_LIMIT]], i16 1)
103 // CHECK: call void @__kmpc_data_sharing_init_stack_spmd
104 // CHECK: br label {{%?}}[[EXEC:.+]]
105 //
106 // CHECK: [[EXEC]]
107 // CHECK: {{call|invoke}} void [[OP2:@.+]]({{.+}}, {{.+}}, i32* [[A]], i16* [[AA]], [10 x i32]* [[B]])
108 // CHECK: br label {{%?}}[[DONE:.+]]
109 //
110 // CHECK: [[DONE]]
111 // CHECK: call void @__kmpc_spmd_kernel_deinit_v2(i16 1)
112 // CHECK: br label {{%?}}[[EXIT:.+]]
113 //
114 // CHECK: [[EXIT]]
115 // CHECK: ret void
116 // CHECK: }
117
118 // CHECK: define internal void [[OP2]](i32* noalias %.global_tid., i32* noalias %.bound_tid., i32* {{[^%]*}}[[ARG1:%.+]], i16* {{[^%]*}}[[ARG2:%.+]], [10 x i32]* {{[^%]*}}[[ARG3:%.+]])
119 // CHECK: = alloca i32*, align
120 // CHECK: = alloca i32*, align
121 // CHECK: [[A_ADDR:%.+]] = alloca i32*, align
122 // CHECK: [[AA_ADDR:%.+]] = alloca i16*, align
123 // CHECK: [[B_ADDR:%.+]] = alloca [10 x i32]*, align
124 // CHECK: store i32* [[ARG1]], i32** [[A_ADDR]], align
125 // CHECK: store i16* [[ARG2]], i16** [[AA_ADDR]], align
126 // CHECK: store [10 x i32]* [[ARG3]], [10 x i32]** [[B_ADDR]], align
127 // CHECK: [[A:%.+]] = load i32*, i32** [[A_ADDR]], align
128 // CHECK: [[AA:%.+]] = load i16*, i16** [[AA_ADDR]], align
129 // CHECK: [[B:%.+]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align
130 // CHECK: store i32 {{%.+}}, i32* [[A]], align
131 // CHECK: store i16 {{%.+}}, i16* [[AA]], align
132 // CHECK: [[ELT:%.+]] = getelementptr inbounds [10 x i32], [10 x i32]* [[B]],
133 // CHECK: store i32 {{%.+}}, i32* [[ELT]], align
134 // CHECK: ret void
135 // CHECK: }
136 #endif
137