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

None Re: Xeon Phi performance 

Forum: Multithreading
Re: None Xeon Phi performance
Re: None Re: Xeon Phi performance (Makoto Asai)
Date: 25 Jun, 2015
From: <padraig.looney@nhs.net>

The geometry is a pencil beam impacting a sheet of material and a sheet of material as a detector. The energy is around 30keV.

I had noticed, from using intel's Vtune that there was a considerable amount of time spent in the SetSeed method. I attach a summary of the results from Vtune before and after your suggestion.

After applying your suggestion the performance for 1 billion events was

CPU: 24 minutes

MIC: 19 minutes

Is there any place in documentation that is relevant to this problem that you would recommend?

Is the method

runManager->SeedOncePerCommunication();

related to this? After noticing the time spend in the SetSeed I tried to use this but it did not improve the runtime.

Many thanks for your help.

   Attachment:
      http://hypernews.slac.stanford.edu/HyperNews/geant4/get/AUX/2015/06/25/03.39-58643-vtuneSeedPerEvent.png
      http://hypernews.slac.stanford.edu/HyperNews/geant4/get/AUX/2015/06/25/03.39-32377-seedPerRun.png

Inline Depth:
 1 1
 All All
Outline Depth:
 1 1
 2 2
 All All
Add message: (add)

1 Feedback: Re: Xeon Phi performance   (Andrea Dotti - 25 Jun, 2015)
2 None: Re: Xeon Phi performance   (Makoto Asai - 25 Jun, 2015)
 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