https://hgpu.org/?p=4255
Efficient Embarrassingly Parallel on Graphics Processor Unit