FAHClient only works for NVIDIA GPUs after FAHClient Restart

[skim]

......................................................

[SteveWillis]

I see it's been a while since you posted, with no response. I've never had a similar problem with my Linux Mint rigs but if I was having that problem the first thing I'd try would be to reinstall the F@H software. Good luck!

[bruce]

I see it's been a while since you posted, with no response. I've never had a similar problem with my Linux Mint rigs but if I was having that problem the first thing I'd try would be to reinstall the F@H software. Good luck!

That never hurts, but given that the error messages are
11:05:35: CUDA: Not detected: cuInit() returned 100
11:05:35: OpenCL: Not detected: clGetPlatformIDs() returned -1001
I'd start by reinstalling the NVidia drivers. They may have been installed in a strange place or there may be a step that finalizes the installation that failed.

[skim]

.............................................

[bruce]

There have been other reports of error messages associated with CUDA/OpenCL during FAHClient's initialization but those reports also showed that the GPUs still could fold.

You're suggesting that maybe this is a race condition. (Provisionally I'm calling this a race condition although it may not be). Moreover, the I don't understand why it is connected with the initialization of X since FAH can be installed on server versions of Linux where X is never initialized.

Perhaps you can gather a bit more information. What happens if, instead of restarting the FAHClient, you simply restart the GPU slots with a PAUSE/UNPAUSE. Does FAHClient now see GPUs as operational?

Perhaps when the errors
> CUDA: Not detected: cuInit() returned 100
> OpenCL: Not detected: clGetPlatformIDs() returned -1001
are detected, all FAHClient needs to do is to retry GPU detection after a brief wait.

[skim]

...........................................

[skim]

................................