Looking forward to implement sparse voxel octree code with iterative mesh refinement with Chapel. I forgot about it, because it was unusable for GPUs back when

Looking forward to implement sparse voxel octree code with iterative mesh refinement with Chapel. I forgot about it, because it was unusable for GPUs back when I first had contact. And now it seems good enough for non-GPU code and maybe even GPU code in comparison to C++. I expect a few 10s of TB ram to be required for a specific dataset, which practically rules out any shared-memory approach. Clusters should be better above a few hundred GB anyway, maybe less depending on network bandwidth vs. RAM needs.