Last June the answer was a qualified no: Can I use the GPU for calculations?

However, the Videocore driver code was open sourced towards the end of 2012: http://www.raspberrypi.org/archives/2221 - does that mean we're now likely to see it happen?

If so, given existing libraries (http://mathema.tician.de/software/pyopencl, http://www.jocl.org/) might already work, but with a CPU implementation, it would seem reasonable to start using them in the hope they'll soon be able to use the GPU.

  • 1
    If someone would like to create "opencl" and/or "gpgpu" tags feel free to re-tag this question. Commented Mar 29, 2013 at 11:16
  • Note that it's not accurate to say that "Videocore driver code was open sourced". Only part of it - running on ARM side was (and really important code is running on VideoCore itself and community have no way to do any changes to it). So this may be not enough for such implementation. Commented Mar 29, 2013 at 11:39

2 Answers 2


The situtation dramatically changed in 2014 with the official release of documentation on the QPU units in the Pi GPU. The blog featured several GPGPU projects after that. Since then , the number of Pi GPGPU projects has risen.

See for example "GPGPU python library for the Raspberry Pi" : https://github.com/nineties/py-videocore


OpenMAX is supported. It does not let you do arbitrary computations on the GPU, but it seems to contain compression and decompression of audio, video and images. This thread is about JPEG decompression with OpenMAX on the Pi.

Here is a list of API:s supported by the VideoCore.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.