1 // RUN: %clang_cc1 -verify -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=nvptx64-nvidia-cuda -emit-llvm-bc %s -o %t-ppc-host.bc
2 // RUN: %clang_cc1 -verify -fopenmp -x c++ -triple nvptx64-nvidia-cuda -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
3 // expected-no-diagnostics
4
foo(int & a)5 int foo(int &a) { return a; }
6
bar()7 int bar() {
8 int a;
9 return foo(a);
10 }
11
12 // CHECK: define weak void @__omp_offloading_{{.*}}maini1{{.*}}_l[[@LINE+5]](i32* nonnull align {{[0-9]+}} dereferenceable({{[0-9]+}}) %{{.*}})
13 // CHECK-NOT: @__kmpc_data_sharing_coalesced_push_stack
14
maini1()15 int maini1() {
16 int a;
17 #pragma omp target parallel map(from:a)
18 {
19 int b;
20 a = foo(b) + bar();
21 }
22 return a;
23 }
24
25 // parallel region
26 // CHECK: define {{.*}}void @{{.*}}(i32* noalias {{.*}}, i32* noalias {{.*}}, i32* nonnull align {{[0-9]+}} dereferenceable({{[0-9]+}}) %{{.*}})
27 // CHECK-NOT: call i8* @__kmpc_data_sharing_coalesced_push_stack(
28 // CHECK: [[B_ADDR:%.+]] = alloca i32,
29 // CHECK: call {{.*}}[[FOO:@.*foo.*]](i32* nonnull align {{[0-9]+}} dereferenceable({{[0-9]+}}) [[B_ADDR]])
30 // CHECK: call {{.*}}[[BAR:@.*bar.*]]()
31 // CHECK-NOT: call void @__kmpc_data_sharing_pop_stack(
32 // CHECK: ret void
33
34 // CHECK: define {{.*}}[[FOO]](i32* nonnull align {{[0-9]+}} dereferenceable{{.*}})
35 // CHECK-NOT: @__kmpc_data_sharing_coalesced_push_stack
36
37 // CHECK: define {{.*}}[[BAR]]()
38 // CHECK: alloca i32,
39 // CHECK: [[A_LOCAL_ADDR:%.+]] = alloca i32,
40 // CHECK: [[RES:%.+]] = call i8 @__kmpc_is_spmd_exec_mode()
41 // CHECK: [[IS_SPMD:%.+]] = icmp ne i8 [[RES]], 0
42 // CHECK: br i1 [[IS_SPMD]], label
43 // CHECK: br label
44 // CHECK: [[RES:%.+]] = call i8* @__kmpc_data_sharing_coalesced_push_stack(i64 128, i16 0)
45 // CHECK: [[GLOBALS:%.+]] = bitcast i8* [[RES]] to [[GLOBAL_ST:%.+]]*
46 // CHECK: br label
47 // CHECK: [[ITEMS:%.+]] = phi [[GLOBAL_ST]]* [ null, {{.+}} ], [ [[GLOBALS]], {{.+}} ]
48 // CHECK: [[A_ADDR:%.+]] = getelementptr inbounds [[GLOBAL_ST]], [[GLOBAL_ST]]* [[ITEMS]], i{{[0-9]+}} 0, i{{[0-9]+}} 0
49 // CHECK: [[TID:%.+]] = call i32 @llvm.nvvm.read.ptx.sreg.tid.x()
50 // CHECK: [[LID:%.+]] = and i32 [[TID]], 31
51 // CHECK: [[A_GLOBAL_ADDR:%.+]] = getelementptr inbounds [32 x i32], [32 x i32]* [[A_ADDR]], i32 0, i32 [[LID]]
52 // CHECK: [[A_ADDR:%.+]] = select i1 [[IS_SPMD]], i32* [[A_LOCAL_ADDR]], i32* [[A_GLOBAL_ADDR]]
53 // CHECK: call {{.*}}[[FOO]](i32* nonnull align {{[0-9]+}} dereferenceable{{.*}} [[A_ADDR]])
54 // CHECK: br i1 [[IS_SPMD]], label
55 // CHECK: [[BC:%.+]] = bitcast [[GLOBAL_ST]]* [[ITEMS]] to i8*
56 // CHECK: call void @__kmpc_data_sharing_pop_stack(i8* [[BC]])
57 // CHECK: br label
58 // CHECK: ret i32
59