https://hgpu.org/?p=9438
On-Demand Generating and Scheduling Optimised Parallel Applications on Heterogeneous Platforms