| 1 | // Test target codegen - host bc file has to be created first. |
|---|
| 2 | // RUN: %clang_cc1 -verify -fopenmp -fopenmp-version=45 -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 -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 -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 -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 -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 | // expected-no-diagnostics |
|---|
| 8 | #ifndef HEADER |
|---|
| 9 | #define HEADER |
|---|
| 10 | |
|---|
| 11 | int a; |
|---|
| 12 | |
|---|
| 13 | int foo(int *a); |
|---|
| 14 | |
|---|
| 15 | int main(int argc, char **argv) { |
|---|
| 16 | int b[10], c[10], d[10]; |
|---|
| 17 | #pragma omp target teams map(tofrom:a) |
|---|
| 18 | #pragma omp distribute parallel for firstprivate(b) lastprivate(c) if(a) |
|---|
| 19 | for (int i= 0; i < argc; ++i) |
|---|
| 20 | a = foo(&i) + foo(&a) + foo(&b[i]) + foo(&c[i]) + foo(&d[i]); |
|---|
| 21 | return 0; |
|---|
| 22 | } |
|---|
| 23 | |
|---|
| 24 | // CHECK: [[MEM_TY:%.+]] = type { [128 x i8] } |
|---|
| 25 | // CHECK-DAG: [[SHARED_GLOBAL_RD:@.+]] = common addrspace(3) global [[MEM_TY]] zeroinitializer |
|---|
| 26 | // CHECK-DAG: [[KERNEL_PTR:@.+]] = internal addrspace(3) global i8* null |
|---|
| 27 | // CHECK-DAG: [[KERNEL_SIZE:@.+]] = internal unnamed_addr constant i{{64|32}} 84 |
|---|
| 28 | // CHECK-DAG: @__omp_offloading_{{.*}}_main_l17_exec_mode = weak constant i8 1 |
|---|
| 29 | |
|---|
| 30 | // CHECK-LABEL: define internal void @__omp_offloading_{{.*}}_main_l17_worker( |
|---|
| 31 | |
|---|
| 32 | // CHECK: define weak void @__omp_offloading_{{.*}}_main_l17([10 x i32]* dereferenceable(40) %{{.+}}, [10 x i32]* dereferenceable(40) %{{.+}}, i32* dereferenceable(4) %{{.+}}, i{{64|32}} %{{.+}}, [10 x i32]* dereferenceable(40) %{{.+}}) |
|---|
| 33 | // CHECK: call void @__kmpc_get_team_static_memory(i16 0, i8* addrspacecast (i8 addrspace(3)* getelementptr inbounds ([[MEM_TY]], [[MEM_TY]] addrspace(3)* [[SHARED_GLOBAL_RD]], i32 0, i32 0, i32 0) to i8*), i{{64|32}} 84, i16 1, i8** addrspacecast (i8* addrspace(3)* [[KERNEL_PTR]] to i8**)) |
|---|
| 34 | // CHECK: [[PTR:%.+]] = load i8*, i8* addrspace(3)* [[KERNEL_PTR]], |
|---|
| 35 | // CHECK: [[STACK:%.+]] = bitcast i8* [[PTR]] to %struct._globalized_locals_ty* |
|---|
| 36 | // CHECK: [[ARGC:%.+]] = load i32, i32* %{{.+}}, align |
|---|
| 37 | // CHECK: [[ARGC_ADDR:%.+]] = getelementptr inbounds %struct._globalized_locals_ty, %struct._globalized_locals_ty* [[STACK]], i{{32|64}} 0, i{{32|64}} 0 |
|---|
| 38 | // CHECK: store i32 [[ARGC]], i32* [[ARGC_ADDR]], |
|---|
| 39 | // CHECK: getelementptr inbounds %struct._globalized_locals_ty, %struct._globalized_locals_ty* [[STACK]], i{{32|64}} 0, i{{32|64}} 1 |
|---|
| 40 | // CHECK: getelementptr inbounds %struct._globalized_locals_ty, %struct._globalized_locals_ty* [[STACK]], i{{32|64}} 0, i{{32|64}} 2 |
|---|
| 41 | // CHECK: call void @__kmpc_for_static_init_4( |
|---|
| 42 | |
|---|
| 43 | // CHECK: call void @__kmpc_serialized_parallel( |
|---|
| 44 | // CHECK: call void [[PARALLEL:@.+]]( |
|---|
| 45 | // CHECK: call void @__kmpc_end_serialized_parallel( |
|---|
| 46 | |
|---|
| 47 | // CHECK: call void @__kmpc_for_static_fini(%struct.ident_t* @ |
|---|
| 48 | |
|---|
| 49 | // CHECK: call void @__kmpc_restore_team_static_memory(i16 0, i16 1) |
|---|
| 50 | |
|---|
| 51 | // CHECK: define internal void [[PARALLEL]]( |
|---|
| 52 | // CHECK-NOT: call i8* @__kmpc_data_sharing_push_stack( |
|---|
| 53 | |
|---|
| 54 | // CHECK-NOT: call void @__kmpc_data_sharing_pop_stack( |
|---|
| 55 | |
|---|
| 56 | #endif |
|---|
| 57 | |
|---|