https://hgpu.org/?p=18863
Transparent Acceleration for Heterogeneous Platforms With Compilation to OpenCL