clang/test/OpenMP/allocate_codegen.cpp

   1 // RUN: %clang_cc1 -verify -fopenmp -triple x86_64-apple-darwin10.6.0 -emit-llvm -o - %s | FileCheck %s
   2 // RUN: %clang_cc1 -fopenmp -triple x86_64-apple-darwin10.6.0 -x c++ -std=c++11 -emit-pch -o %t %s
   3 // RUN: %clang_cc1 -fopenmp -triple x86_64-apple-darwin10.6.0 -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s
   4 // RUN: %clang_cc1 -verify -fopenmp -triple x86_64-unknown-linux-gnu -emit-llvm -o - %s | FileCheck %s
   5 // RUN: %clang_cc1 -fopenmp -fnoopenmp-use-tls -triple x86_64-unknown-linux-gnu -x c++ -std=c++11 -emit-pch -o %t %s
   6 // RUN: %clang_cc1 -fopenmp -fnoopenmp-use-tls -triple x86_64-unknown-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s
   7
   8 // RUN: %clang_cc1 -verify -fopenmp-simd -triple x86_64-apple-darwin10.6.0 -emit-llvm -o - %s | FileCheck --check-prefix SIMD-ONLY0 %s
   9 // RUN: %clang_cc1 -fopenmp-simd -triple x86_64-apple-darwin10.6.0 -x c++ -std=c++11 -emit-pch -o %t %s
  10 // RUN: %clang_cc1 -fopenmp-simd -triple x86_64-apple-darwin10.6.0 -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck --check-prefix SIMD-ONLY0 %s
  11 // RUN: %clang_cc1 -verify -fopenmp-simd -triple x86_64-unknown-linux-gnu -emit-llvm -o - %s | FileCheck --check-prefix SIMD-ONLY0 %s
  12 // RUN: %clang_cc1 -fopenmp-simd -fnoopenmp-use-tls -triple x86_64-unknown-linux-gnu -x c++ -std=c++11 -emit-pch -o %t %s
  13 // RUN: %clang_cc1 -fopenmp-simd -fnoopenmp-use-tls -triple x86_64-unknown-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck --check-prefix SIMD-ONLY0 %s
  14 // SIMD-ONLY0-NOT: {{__kmpc|__tgt}}
  15 // expected-no-diagnostics
  16
  17 #ifndef HEADER
  18 #define HEADER
  19
  20 enum omp_allocator_handle_t {
  21   omp_null_allocator = 0,
  22   omp_default_mem_alloc = 1,
  23   omp_large_cap_mem_alloc = 2,
  24   omp_const_mem_alloc = 3,
  25   omp_high_bw_mem_alloc = 4,
  26   omp_low_lat_mem_alloc = 5,
  27   omp_cgroup_mem_alloc = 6,
  28   omp_pteam_mem_alloc = 7,
  29   omp_thread_mem_alloc = 8,
  30   KMP_ALLOCATOR_MAX_HANDLE = __UINTPTR_MAX__
  31 };
  32
  33 struct St {
  34   int a;
  35 };
  36
  37 struct St1 {
  38   int a;
  39   static int b;
  40 #pragma omp allocate(b) allocator(omp_default_mem_alloc)
  41 } d;
  42
  43 int a, b, c;
  44 #pragma omp allocate(a) allocator(omp_large_cap_mem_alloc)
  45 #pragma omp allocate(b) allocator(omp_const_mem_alloc)
  46 #pragma omp allocate(d, c) allocator(omp_high_bw_mem_alloc)
  47
  48 template <class T>
  49 struct ST {
  50   static T m;
  51 #pragma omp allocate(m) allocator(omp_low_lat_mem_alloc)
  52 };
  53
  54 template <class T> T foo() {
  55   T v;
  56 #pragma omp allocate(v) allocator(omp_cgroup_mem_alloc)
  57   v = ST<T>::m;
  58   return v;
  59 }
  60
  61 namespace ns {
  62 int a;
  63 }
  64 #pragma omp allocate(ns::a) allocator(omp_pteam_mem_alloc)
  65
  66 // CHECK-NOT:  call {{.+}} {{__kmpc_alloc|__kmpc_free}}
  67
  68 int main() {
  69   static int a;
  70 #pragma omp allocate(a) allocator(omp_thread_mem_alloc)
  71   a = 2;
  72   double b = 3;
  73 #pragma omp allocate(b)
  74   return (foo<int>());
  75 }
  76
  77 // CHECK-LABEL: define {{[^@]+}}@main
  78 // CHECK-SAME: () #[[ATTR0:[0-9]+]] {
  79 // CHECK-NEXT:  entry:
  80 // CHECK-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
  81 // CHECK-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1:[0-9]+]])
  82 // CHECK-NEXT:    store i32 0, ptr [[RETVAL]], align 4
  83 // CHECK-NEXT:    store i32 2, ptr @_ZZ4mainE1a, align 4
  84 // CHECK-NEXT:    [[DOTB__VOID_ADDR:%.*]] = call ptr @__kmpc_alloc(i32 [[TMP0]], i64 8, ptr null)
  85 // CHECK-NEXT:    store double 3.000000e+00, ptr [[DOTB__VOID_ADDR]], align 8
  86 // CHECK-NEXT:    [[CALL:%.*]] = call noundef i32 @_Z3fooIiET_v()
  87 // CHECK-NEXT:    store i32 [[CALL]], ptr [[RETVAL]], align 4
  88 // CHECK-NEXT:    call void @__kmpc_free(i32 [[TMP0]], ptr [[DOTB__VOID_ADDR]], ptr null)
  89 // CHECK-NEXT:    [[TMP2:%.*]] = load i32, ptr [[RETVAL]], align 4
  90 // CHECK-NEXT:    ret i32 [[TMP2]]
  91
  92 // CHECK: define {{.*}}i32 @{{.+}}foo{{.+}}()
  93 // CHECK:      [[GTID:%.+]] = call i32 @__kmpc_global_thread_num(ptr @{{.+}})
  94 // CHECK-NEXT: [[V_VOID_ADDR:%.+]] = call ptr @__kmpc_alloc(i32 [[GTID]], i64 4, ptr inttoptr (i64 6 to ptr))
  95 // CHECK-NOT:  {{__kmpc_alloc|__kmpc_free}}
  96 // CHECK:      store i32 %{{.+}}, ptr [[V_VOID_ADDR]],
  97 // CHECK-NEXT: [[V_VAL:%.+]] = load i32, ptr [[V_VOID_ADDR]],
  98 // CHECK-NEXT: call void @__kmpc_free(i32 [[GTID]], ptr [[V_VOID_ADDR]], ptr inttoptr (i64 6 to ptr))
  99 // CHECK-NOT:  {{__kmpc_alloc|__kmpc_free}}
 100 // CHECK:      ret i32 [[V_VAL]]
 101
 102 // CHECK-NOT:  call {{.+}} {{__kmpc_alloc|__kmpc_free}}
 103 extern template int ST<int>::m;
 104
 105 // CHECK: define{{.*}} void @{{.+}}bar{{.+}}(i32 noundef %{{.+}}, ptr noundef {{.+}})
 106 void bar(int a, float &z) {
 107 // CHECK: [[A_VOID_PTR:%.+]] = call ptr @__kmpc_alloc(i32 [[GTID:%.+]], i64 4, ptr inttoptr (i64 1 to ptr))
 108 // CHECK: store i32 %{{.+}}, ptr [[A_VOID_PTR]],
 109 // CHECK: [[Z_VOID_PTR:%.+]] = call ptr @__kmpc_alloc(i32 [[GTID]], i64 8, ptr inttoptr (i64 1 to ptr))
 110 // CHECK: store ptr %{{.+}}, ptr [[Z_VOID_PTR]],
 111 #pragma omp allocate(a, z) allocator(omp_default_mem_alloc)
 112   // CHECK: call void @__kmpc_free(i32 [[GTID]], ptr [[Z_VOID_PTR]], ptr inttoptr (i64 1 to ptr))
 113   // CHECK: call void @__kmpc_free(i32 [[GTID]], ptr [[A_VOID_PTR]], ptr inttoptr (i64 1 to ptr))
 114   // CHECK: ret void
 115 }
 116 #endif