https://hgpu.org/?p=10091
Methods for Optimizing OpenCL Applications on Heterogeneous Multicore Architectures