offload/test/offloading/small_trip_count.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 // RUN: env LIBOMPTARGET_INFO=16 LIBOMPTARGET_MIN_THREADS_FOR_LOW_TRIP_COUNT=8 \
   6 // RUN:   %libomptarget-run-generic 2>&1 | %fcheck-generic --check-prefix=EIGHT
   7
   8 // REQUIRES: gpu
   9
  10 #define N 128
  11
  12 __attribute__((optnone)) void optnone() {}
  13
  14 int main() {
  15   // DEFAULT: Launching kernel {{.+_main_.+}} with 4 blocks and 32 threads in SPMD mode
  16   // EIGHT: Launching kernel {{.+_main_.+}} with 16 blocks and 8 threads in SPMD mode
  17 #pragma omp target teams distribute parallel for simd
  18   for (int i = 0; i < N; ++i) {
  19     optnone();
  20   }
  21   // DEFAULT: Launching kernel {{.+_main_.+}} with 4 blocks and 32 threads in SPMD mode
  22   // EIGHT: Launching kernel {{.+_main_.+}} with 16 blocks and 8 threads in SPMD mode
  23 #pragma omp target teams distribute parallel for simd
  24   for (int i = 0; i < N - 1; ++i) {
  25     optnone();
  26   }
  27   // DEFAULT: Launching kernel {{.+_main_.+}} with 5 blocks and 32 threads in SPMD mode
  28   // EIGHT: Launching kernel {{.+_main_.+}} with 17 blocks and 8 threads in SPMD mode
  29 #pragma omp target teams distribute parallel for simd
  30   for (int i = 0; i < N + 1; ++i) {
  31     optnone();
  32   }
  33   // DEFAULT: Launching kernel {{.+_main_.+}} with 32 blocks and 4 threads in SPMD mode
  34   // EIGHT: Launching kernel {{.+_main_.+}} with 32 blocks and 4 threads in SPMD mode
  35 #pragma omp target teams distribute parallel for simd thread_limit(4)
  36   for (int i = 0; i < N; ++i) {
  37     optnone();
  38   }
  39 }
  40