Compute Mandelbrot Set using adaptive mesh refinement technique and C++/kokkos library for multi-architecture execution. This is not intended to be efficient computation, but just a simple illustrative example of use of class Kokkos::UnorderedMap container (multi-architecture hash map data container). Mandelbrot set is computed iteratively, refined cells are inserted into the hashmap based data structure.
The following source code is self-contained; C++/kokkos library sources are included as a git submodule.
Example of Mandelbrot set with level_min=6 and level_max=11
./compute_mandelbrot_amr 6 11
git clone git@github.com:pkestene/AMR_mandelbrot.gitTo build kokkos/OpenMP backend
mkdir -p _build/openmp; cd _build/openmp
cmake -DAMR_MANDELBROT_KOKKOS_BUILD=ON -DAMR_MANDELBROT_KOKKOS_BACKEND=OpenMP ../..
make
# run application
./compute_mandelbrot_amr 6 11Optionally, you can (recommended) activate HWLOC support by using the flag -DKOKKOS_ENABLE_HWLOC=ON.
Obviously, you need to have Nvidia/CUDA driver and toolkit installed on your platform. Then you need to
# example build for cuda
mkdir -p _build/cuda; cd _build/cuda
cmake -DAMR_MANDELBROT_KOKKOS_BUILD=ON -DAMR_MANDELBROT_KOKKOS_BACKEND=Cuda -DKokkos_ENABLE_CUDA_LAMBDA=ON -DKokkos_ENABLE_HWLOC=ON ../..
make
# run application
./compute_mandelbrot_amr 6 11