14

I know this question has been asked and answered before here: Can I use the GPU for calculations?

but there have been a number of developments recently, including a hint that there will be more forthcoming. Does anyone know if this is likely in the next few months - that is, OpenCL on the RPI?

Community
  • 1
cjm2671
  • 271
  • 2
  • 3
  • 7
  • I highly doubt there will be a solid OpenCL implementation in the next 6 months. From what I have seen there is no indication that either the Pi Foundation or Broadcom are working on this. –  Apr 21 '14 at 16:10
  • There is a experimental and abandoned OpenCL compiler here : https://www.raspberrypi.org/forums/viewtopic.php?f=29&t=78919 – flakeshake Dec 10 '15 at 13:58

3 Answers3

3

officially not, if you follow the Raspberry Pi blog you can notice that exist some ASM Compiler for QPUs (the hearts of GPU of Raspberry Pi), you need compile your ASM code (at register level), if you want to try this method I can recommend this link:

http://rpiplayground.wordpress.com/2014/05/03/hacking-the-gpu-for-fun-and-profit-pt-1/

however, if you want to use OpenCL I fear you need to wait more.

tejonbiker
  • 31
  • 2
3

The Raspberry Pi foundation has been endorsing GPGPU on the Pi since 2014 , shortly after Broadcom released documentation for the QPU units inside the GPU.

An experimental OpenCL compiler was created by Simon J. Hall (the winner of the tightly related 2014 10,000 $ competition to make Quake run acceptably without using the GPU BLOB) : see here.

flakeshake
  • 6,235
  • 1
  • 14
  • 33
0

I see two possible ways to use the GPU, sadly none of them involve OpenCL:

  1. Abuse vertex or fragment sharers for your calculations.
  2. Try installing Android and using RenderScript (Android's own GPGPU language). Though I don't actually know whether the Pi's GPU is supported.
Daniel Eberts
  • 151
  • 3
  • OpenCL is yet another RenderScript equivalent that is widely accepted and implemented by almost all big player GPU and CPU manufacturers including AMD (for both Radeon GPUs and CPU cores,) nVidia (for both GPUs and Tegra ARM CPU cores) as well as Intel (for both Intel HD/Iris integrated GPU cores and CPU cores). I have once tried a OpenCL AES demo program and it brought both my quad-core Intel Core 2 and my GeForce GTX 650 Ti to maximum when encrypting a 4GB disk image. – Maxthon Chan Apr 11 '15 at 04:54