Altering the Architecture of a Submitted Job

[CDon]

Hey, when working on DevCloud, I wanted to submit my code to the job's server to leverage GPU resources using qsub. I am able to send the job using a GPU via:

import cfxmagic
%%qsub -l nodes=1:gpu:ppn=2 -l walltime=24:00:00 -d .

However, there were a few things I wanted to alter and check when running these submissions:

1. How can I tell that the GPU is actually being leveraged in the job's server and/or how much of the GPU is being used by the code?

2. For the purpose of being able to apply my code across various kinds of hardware, I wanted to alter the hardware I request with the qsub command. Using 

http://docs.adaptivecomputing.com/torque/4-1-3/Content/topics/2-jobs/requestingRes.htm#qsub, I was able to see that I can alter some information like setting a memory cap. However, I cannot see anything that can alter the number of cores I provision. I can alter the number of nodes with the ppn being fixed; however, I believe that we have 24 cores maximum, which doesn't seem to add up with the 2 nodes maximum each with 2 ppn. Is there some way I can request a specific number of cores under the 24 maximum?

Thanks for all your help!

[JaideepK_Intel]

Hi,

Thank you for posting Intel forums.

We're working on this internally. We'll get back to you soon with an update.

Thanks,

Jaideep

[JaideepK_Intel]

Hi,

Sorry for the delay,

1) To know whether GPU is actually being leveraged in the job's server and/or how much of the GPU is being used by the code,

you can use Intel's powerful analyzer VTune Profiler. It will give you an idea about the same in the GUI.

Please follow the below link for more details.

https://software.intel.com/content/www/us/en/develop/documentation/vtune-help/top/analyze-performance/accelerators-group/gpu-offload-analysis.html

2)We can not get control over the cores in a CPU but you can choose a CPU with desired number of cores. 

Since Devcloud is a worldwide academic resource shared among several thousand users we have standardized what qsub can request in terms of cores, lob limts etc...

These CPUs have 24cores:

qsub -l nodes=1:i9-10920x:ppn=2 -d . -I

qsub -l nodes=1:gold6128:ppn=2 -d . -I

Thanks,

jaideep

[JaideepK_Intel]

Hi,

If this resolves your issue, make sure to accept this as a solution. This would help others with similar issue. 

Thank you,

Jaideep

[JaideepK_Intel]

Hi,

I assume that your issue is resolved. If you need any additional information, please post a new question as this thread will no longer be monitored by Intel.

Thanks,

Jaideep