Message: Re: Xeon Phi performance Not Logged In (login)
 Next-in-Thread Next-in-Thread
 Next-in-Forum Next-in-Forum

Feedback Re: Xeon Phi performance 

Forum: Multithreading
Re: None Xeon Phi performance
Re: None Re: Xeon Phi performance (Makoto Asai)
Re: None Re: Xeon Phi performance
Date: 25 Jun, 2015
From: Andrea Dotti <Andrea Dotti>

Hello Padraig, please note the following: what was suggested by Makoto, can improve the linearity of the speedup (e.g. doubling the number of threads you should get the double of performances as long as physical cores are available). See for example: https://twiki.cern.ch/twiki/bin/view/Geant4/MultiThreadingTaskForce#CPU_and_Memory_Performances Having good linearity is a prerequisite to obtain good performances on Xeon Phi (otherwise you are not using efficiency all hardware threads). If you have a good linearity and you manage to use the maximum number of threads allowed on the card, you may still get performances that are less than the theoretical peak performances. This is because we have not yet optimized Geant4 code for Intel Xeon Phi.

Independently of this, I would like to comment that the "extremely optimized" applications I've seen run on Xeon Phi a factor 2 faster than on the host. You should never expect very large performances (this is because every improvement on the Xeon Phi will also benefit the host). If I understand correctly your post you get more or less the same performances on the host and on the card.

I hope this clarifies a bit, please do not hesitate to contact us if you want more information on these aspects. Andrea

 Add Message Add Message
to: "Re: Xeon Phi performance"

 Subscribe Subscribe

This site runs SLAC HyperNews version 1.11-slac-98, derived from the original HyperNews