https://hgpu.org/?p=27718
Efficient OpenCL system integration of non-blocking FPGA accelerators