Roofline performance analysis for xeon-phi

Roofline performance analysis for xeon-phi

How can I perform the new roofline performance analysis on a remote xeon-phi MIC card through GUI or commandline?

4 posts / 0 new
Last post
For more complete information about compiler optimizations, see our Optimization Notice.

 

Hi Saumya,

Here is our getting started guide:

https://software.intel.com/en-us/articles/getting-started-with-intel-advisor-roofline-feature

Let us know if you have any further questions!

Regards,

Kevin

Hello Kevin, thank you for this documentation. But I had gone through it. I was able to find the roofline performance model for my program on

host system using this document.

But, how to find the same for a program running on a remote Xeon-phi MIC? Is there some option through which we could specify the target

system?

Soumya B.

I have the same desire - I want to run Advisor XE roofline on Intel Xeon Phi.    I am running on Intel Xeon Phi (KNL) intel mpi with 128 mpi ranks.  The systems I use only have batch submitting interface so I need to use command line interface.  Through previous experience I learned Advisor XE only works for MPI rank 0.   I found this intel article answers our questions ( I found your question first - but I kept looking until I found this other document and thought I would post the link for anyone coming across your question first like I did).

https://software.intel.com/en-us/get-started-with-advisor-command-line-a...

So I will have to submit two batch jobs - first collect advisor xe survey for rank 0, then collect advisor xe trip-counts/flops for rank 0.  Then we should be able to get our data for roofline analysis. I will be submitting jobs today to try this.

Leave a Comment

Please sign in to add a comment. Not a member? Join today