Re: [HTCondor-users] Bug report: condor_gpu_discovery.exe crashes on Windows 7 & 8.1 32 & 64 Bit in OpenCL detection

["John (TJ) Knoeller" <johnkn@xxxxxxxxxxx>]

Hi, Ernst-Georg.   I'm the one who touched the condor_gpu_discovery code 
last, so I'll
be the one to chase this crash.   The first step is to try and reproduce 
the problme.
Can you tell me what version of the OpenCL libraries you have installed?
Is it the OpenCL that wraps around CUDA devices? or the one for AMD devices?

Could you send me that actual output of condor_gpu_discovery from when 
it  crashes?

also could you try running on the command line with the -diagnostic flag 
and send me that output as well?

thanks.
-tj

On 4/25/2014 5:36 AM, Ernst-Georg Schmid wrote:
> Hello,
>
> I have tried condor_gpu_discovery.exe from 8.1.5 on various Windows Machines with Windows 7 and 8.1, 32 and 64 bit.
>
> CUDA detection works as expected with Nvidia cards.
>
> OpenCL detection crashes.
>
> When run with '-verbose', the OpenCL output looks ok, until the number of devices present is reported. There a random number is shown that has nothing to do with the actual number of installed GPUs and then Windows reports that condor_gpu_discovery.exe has stopped working.
>
> This happens with Nvidia, AMD and Intel devices. Since I can compile and run 32 and 64 bit OpenCL programs fine on all these machines, I assume that something in condor_gpu_discovery.exe is broken.
>
> The wrong number of devices might be random memory content which would indicate an off-by-one-error or something like this and would also explain the immediate crash afterwards.
>
> Best regards,
>
> Ernst-Georg Schmid
>
> _______________________________________________
> HTCondor-users mailing list
> To unsubscribe, send a message to htcondor-users-request@xxxxxxxxxxx with a
> subject: Unsubscribe
> You can also unsubscribe by visiting
> https://lists.cs.wisc.edu/mailman/listinfo/htcondor-users
>
> The archives can be found at:
> https://lists.cs.wisc.edu/archive/htcondor-users/