Dear all,
First post here :)
I've been crunching for E@H for about a month now. I added a Raspberry Pi 3 to my cluster recently, but as you can imagine the four CPU cores don't allow me to contribute much to this project... Since the Raspberry Pi 3 has a decent VideoCore GPU, and there's one existing implementation of OpenCL for it, I thought I'd give it a try:
https://github.com/doe300/VC4CL
So my question is: do you know whether GPU jobs would run on my Raspberry Pi 3, knowing that VC4CL is compatible with up to OpenCL 1.2 but only supports the embedded profile ?
Best regards,
Samuel (a former member of the Fermi-LAT collaboration ;) )
Copyright © 2024 Einstein@Home. All rights reserved.
You might try posting your
)
You might try posting your question in this thread instead as it's more about ARM and Pi devices and has several active crunchers using them:
https://einsteinathome.org/content/parallella-raspberry-pi-fpga-all-stuff
Dear Mikey, Thank you for
)
Dear Mikey,
Thank you for your quick answer, really appreciated.
As you can see my credit is 0 (my computers are part of a pool, I'm not earning any credit), so I doubt I could post in that thread, I read it prior to posting here though.
Could you please relay my message over there, or draw the attention of those ARM crunchers to this thread ?
Best regards,
Samuel
The only gpu application for
)
The only gpu application for SoC devices here at Einstein is the one for the embedded Intel gpu on the BRP project.
Until the project devs make something else for the embedded gpus on other non-Intel, non-ATI SoC platforms you are out of luck here.
I got lucky with Guarav Khanna's Nvidia application for my Nvidia Jetson Nano but that SoC has a real Maxwell class gpu in it. It uses the standard Nvidia drivers. But Guarav's BRP application is actually the code for the BRP cpu application compiled for the gpu.
Dear Keith, Thank you for
)
Dear Keith,
Thank you for this detailed explanation.
This first foray into GPGPU is a difficult one indeed ! :)
Best regards,
Samuel