Nvc++ seems to ignore std::execution::par if --stdpar is not specified

Accelerated Computing HPC Compilers nvc, nvc++ and nvfortran
1

Hi,

I compile my code with the following commands …

includes="$NVCOMPILERS/$NVARCH/20.5/compilers/include-stdpar"
compile="nvc++ -Wall -fast -I $includes"

$compile -DPOLICY=std::execution::par -o main_par main.cpp

This enables me to define different policies each time so I can compare performance.

The loop of my code (ignoring various dependencies) is …

std::transform
(
  POLICY,
  thrust::counting_iterator<unsigned>(0),
  thrust::counting_iterator<unsigned>(num_samples),
  result->data(),
  [=](unsigned index) -> result_scalar_type
  {
    result_scalar_type k = (index * max_k) / num_samples;
    result_scalar_type n = initial_value; 

    for (unsigned i = 0; i < iterations ; i++)
      n = k * n * (1.0 - n);

    return n;
  }
);

Whereas if I compile with --stdpar the code is parallelized on the GPU, without the --stdpar there is no difference between the par and seq execution policies. I would have expected this to be spread across the cpu cores & threads?

Am I doing something wrong?

Thanks,

Leigh.