https://hgpu.org/?p=22164
Automated Partitioning of Data-Parallel Kernels using Polyhedral Compilation