v4lloopback, syphon, spout output

[drlight-code]

We need v4lloopback output in order to route image data to veejay / MapMap. We have to think about some issues here. Foremost, will there be one etudes-host instance per running etude, or just one host application running them all?

In general I would not mind to have one etude per instance, it would keep things very simple. In that case however our naming scheme would be different, since one etude actually would be identified by a port with a separate listening osc server instance rather than by a name, but that could be quickly adapted and would be fine actually.

We could also have one host application which holds several etudes running in parallel which would make the application more complex, however would have some benefits:

• resources / result images could be shared between etudes, question do we want that
• tearless synchronization of the different outputs to v4l might be simpler
• for a lot of opengl applications the context switching might become expensive

[drlight-code]

An interesting mail on mapmap-list describes new developments in DRI which would allow sharing of image data using the GPU, similar to syphon on MacOS X:

• https://github.com/pcwalton/linux-drm-sharing-test
• http://keithp.com/blogs/DRI-Next/

[drlight-code]

Rudimentary v4lloopback output working, still lots of things to be done, such as

• [ ] use the dma-buf interface to share memory regions between GPU and loopback driver
• [ ] ping-pong FBO rendering and interleaved readout for all etudes which output
• [ ] tight frame synchronization, no tearing
• [ ] supporte/negotiate different pixel formats and color spaces, on-the-fly conversion
• [ ] get input on the veejay side working

[drlight-code]

syphon and spout working :)