OpenGL/OpenCL interoperability on headless server?

Question

Is there a way to run app that uses OpenGL/OpenCL interoperability on a headless server? Maybe without hardware acceleration. Maybe using some library for offscreen rendering.

Now there are plenty of popular cloud continuous integration services like drone.io or travis-ci. But testing projects that use OpenGL/OpenCL seems to be a problem still.

Answer 1

OpenGL/OpenCL interop usually is tied to a particular combination of OpenGL + OpenCL implementation. Normally the only combination of implementations working together are those, found in the same GPU driver package.

Mesa actually has some OpenCL support but I honestly don't know how well it works, what it constraints are and so on. Of course OpenCL is mostly useful only if you have a massively parallelizing processor available, like a GPU or a CPU cluster. There is a OpenCL implementation made by Intel targeting their multicore CPUs, but I don't know how well (if at all) it works with arbitrary OpenGL implementations.

If you have a OpenCL capable GPU available, then the easiest things to do, would be starting an X11 server with no or just a dummy screen configured (yes, this is possible).