openmp/libomptarget/test/offloading/default_thread_limit.c

   1 // clang-format off
   2 // RUN: %libomptarget-compile-generic
   3 // RUN: env LIBOMPTARGET_INFO=16 \
   4 // RUN:   %libomptarget-run-generic 2>&1 | %fcheck-generic --check-prefix=DEFAULT
   5
   6 // UNSUPPORTED: nvptx64-nvidia-cuda
   7 // UNSUPPORTED: nvptx64-nvidia-cuda-LTO
   8 // UNSUPPORTED: aarch64-unknown-linux-gnu
   9 // UNSUPPORTED: aarch64-unknown-linux-gnu-LTO
  10 // UNSUPPORTED: x86_64-pc-linux-gnu
  11 // UNSUPPORTED: x86_64-pc-linux-gnu-LTO
  12
  13 __attribute__((optnone)) int optnone() { return 1; }
  14
  15 int main() {
  16   int N = optnone() * 4098 * 32;
  17
  18 // DEFAULT: [[NT:(128|256)]] (MaxFlatWorkGroupSize: [[NT]]
  19 #pragma omp target teams distribute parallel for simd
  20   for (int i = 0; i < N; ++i) {
  21     optnone();
  22   }
  23 // DEFAULT: [[NT:(128|256)]] (MaxFlatWorkGroupSize: [[NT]]
  24 #pragma omp target teams distribute parallel for simd
  25   for (int i = 0; i < N; ++i) {
  26     optnone();
  27   }
  28 // DEFAULT: [[NT:(128|256)]] (MaxFlatWorkGroupSize: [[NT]]
  29 #pragma omp target teams distribute parallel for simd
  30   for (int i = 0; i < N; ++i) {
  31     optnone();
  32   }
  33 // DEFAULT: [[NT:(128|256)]] (MaxFlatWorkGroupSize: [[NT]]
  34 #pragma omp target
  35 #pragma omp teams distribute parallel for
  36   for (int i = 0; i < N; ++i) {
  37     optnone();
  38   }
  39 // DEFAULT: 42 (MaxFlatWorkGroupSize: 1024
  40 #pragma omp target thread_limit(optnone() * 42)
  41 #pragma omp teams distribute parallel for
  42   for (int i = 0; i < N; ++i) {
  43     optnone();
  44   }
  45 // DEFAULT: 42 (MaxFlatWorkGroupSize: 42
  46 #pragma omp target thread_limit(optnone() * 42) ompx_attribute(__attribute__((amdgpu_flat_work_group_size(42, 42))))
  47 #pragma omp teams distribute parallel for
  48   for (int i = 0; i < N; ++i) {
  49     optnone();
  50   }
  51 // DEFAULT: 42 (MaxFlatWorkGroupSize: 42
  52 #pragma omp target ompx_attribute(__attribute__((amdgpu_flat_work_group_size(42, 42))))
  53 #pragma omp teams distribute parallel for
  54   for (int i = 0; i < N; ++i) {
  55     optnone();
  56   }
  57 // DEFAULT: MaxFlatWorkGroupSize: 1024
  58 #pragma omp target
  59 #pragma omp teams distribute parallel for num_threads(optnone() * 42)
  60   for (int i = 0; i < N; ++i) {
  61     optnone();
  62   }
  63 // DEFAULT: MaxFlatWorkGroupSize: 1024
  64 #pragma omp target teams distribute parallel for thread_limit(optnone() * 42)
  65   for (int i = 0; i < N; ++i) {
  66     optnone();
  67   }
  68 // DEFAULT: MaxFlatWorkGroupSize: 1024
  69 #pragma omp target teams distribute parallel for num_threads(optnone() * 42)
  70   for (int i = 0; i < N; ++i) {
  71     optnone();
  72   }
  73 // DEFAULT: 9 (MaxFlatWorkGroupSize: 9
  74 #pragma omp target
  75 #pragma omp teams distribute parallel for num_threads(9)
  76   for (int i = 0; i < N; ++i) {
  77     optnone();
  78   }
  79 // DEFAULT: 4 (MaxFlatWorkGroupSize: 4
  80 #pragma omp target thread_limit(4)
  81 #pragma omp teams distribute parallel for
  82   for (int i = 0; i < N; ++i) {
  83     optnone();
  84   }
  85 // DEFAULT: 4 (MaxFlatWorkGroupSize: 4
  86 #pragma omp target
  87 #pragma omp teams distribute parallel for thread_limit(4)
  88   for (int i = 0; i < N; ++i) {
  89     optnone();
  90   }
  91 // DEFAULT: 9 (MaxFlatWorkGroupSize: 9
  92 #pragma omp target teams distribute parallel for num_threads(9)
  93   for (int i = 0; i < N; ++i) {
  94     optnone();
  95   }
  96 // DEFAULT: 4 (MaxFlatWorkGroupSize: 4
  97 #pragma omp target teams distribute parallel for simd thread_limit(4)
  98   for (int i = 0; i < N; ++i) {
  99     optnone();
 100   }
 101 }
 102