Use multiple graphics card/multiple instances

hydradarpa / kalman-hydra

Python and C++ code for Hydra optical flow, behavior analysis with extended Kalman filter tracker

2 stars 0 forks source link

Use multiple graphics card/multiple instances #11

Closed benlansdell closed 8 years ago

benlansdell commented 8 years ago

See:

http://www.virtualgl.org/vgldoc/2_2_1/#hd005002

With each block we run separate mesh initializations, and then glue those meshes together (create a mapping between the two).

benlansdell commented 8 years ago

Need to check performance hit from using same GPU for multiple instances...

benlansdell commented 8 years ago

OpenGL doesn't play well with threads (https://github.com/vispy/vispy/issues/632), so the best option might be to very crudely parallelize things. This would be done by splitting the video into chunks and running the chunks separately, with independent meshes. Once the chunks are done, a mapping between the meshes at the frames that overlap would be determined. Open a new issue for this...