offload/test/offloading/small_trip_count_thread_limit.cpp

   1 // clang-format off
   2 // RUN: %libomptarget-compilexx-generic
   3 // RUN: env LIBOMPTARGET_INFO=16 \
   4 // RUN:   %libomptarget-run-generic 2>&1 | %fcheck-generic
   5
   6 // REQUIRES: gpu
   7
   8 int main(int argc, char *argv[]) {
   9   constexpr const int block_size = 256;
  10   constexpr const int grid_size = 4;
  11   constexpr const int count = block_size * grid_size;
  12
  13   int *data = new int[count];
  14
  15 #pragma omp target teams distribute parallel for thread_limit(block_size) map(from: data[0:count])
  16   for (int i = 0; i < count; ++i)
  17     data[i] = i;
  18
  19   for (int i = 0; i < count; ++i)
  20     if (data[i] != i)
  21       return 1;
  22
  23   delete[] data;
  24
  25   return 0;
  26 }
  27
  28 // CHECK: Launching kernel {{.*}} with [4,1,1] blocks and [256,1,1] threads in SPMD mode