https://hgpu.org/?p=4350
The accelerating implementation of BLAST with stream processor