Re: [Phonopy-users] Phonopy performance scaling

[Atsushi T. <atz...@gm...>]

My question was not answered although I answered your question.

Togo

On Tue, Jul 20, 2021 at 7:03 PM Rebecca Fair - STFC UKRI
<reb...@st...> wrote:
>
> Thanks for your reply, I just wanted to check I was making a fair comparison with Phonopy and was not missing anything, if Phonopy doesn't implement parallelisation over q-points that is an appropriate explanation.
>
> Becky
>
> -----Original Message-----
> From: Atsushi Togo <atz...@gm...>
> Sent: 20 July 2021 10:44
> To: phonopy-users <pho...@li...>
> Subject: Re: [Phonopy-users] Phonopy performance scaling
>
> In that case, distribution over q-points is probably the best and it is not implemented in phonopy. Do you really think your work is meaningful?
>
> Togo
>
> On Tue, Jul 20, 2021 at 6:08 PM Rebecca Fair - STFC UKRI <reb...@st...> wrote:
> >
> > I'm using Python's timeit to time the execution of the run_qpoints function for 25,000 q-points, and I'm submitting jobs to STFC's SCARF cluster with different numbers of allocated cores (all exclusive node access). For example, for a 40 atom unit cell of ScO it takes an average of 89s, but this doesn't improve with increasing numbers of cores. I've tried setting the OMP_NUM_THREADS variable to the same as the number of requested cores, and this changes the average time to approximately 68, 68, 60, 62, 68, 76, 82 seconds for 1, 2, 4, 8, 12, 16 and 24 cores respectively. So there does appear to be some sort of performance improvement with 4, 8 cores when setting the environment variable, but not by much. I was wondering if there is anything else I should be doing?
> >
> > Thanks,
> >
> > Becky
> >
> > -----Original Message-----
> > From: Atsushi Togo <atz...@gm...>
> > Sent: 20 July 2021 09:42
> > To: pho...@li...
> > Subject: Re: [Phonopy-users] Phonopy performance scaling
> >
> > Hi,
> >
> > Can you define your "performance"?
> >
> > Togo
> >
> > On Tue, Jul 20, 2021 at 5:21 PM Rebecca Fair - STFC UKRI <reb...@st...> wrote:
> > >
> > > Hi,
> > >
> > >
> > >
> > > I’m doing some performance benchmarking of the eigenvector/value calculation of a piece of software we’re developing, Euphonic, against some other codes such as Phonopy and CASTEP.
> > >
> > >
> > >
> > > Unfortunately I’m unable to get any speedup of the run_qpoints function for Phonopy at all, even using largeish 40 atom unit cells. I’m compiling Phonopy 2.9.3 with Intel 17.0.2, I have set cc=icc and am using the following flags:
> > >
> > >
> > >
> > > extra_compile_args_phonopy = [‘-qopenmp’, ‘-O3’]
> > >
> > > extra_link_args_phonopy = [‘-qopenmp’]
> > >
> > >
> > >
> > > Is there anything I’m missing? Or is this expected?
> > >
> > >
> > >
> > > Best wishes,
> > >
> > >
> > >
> > > Becky
> > >
> > >
> > >
> > > This email and any attachments are intended solely for the use of the named recipients. If you are not the intended recipient you must not use, disclose, copy or distribute this email or any of its attachments and should notify the sender immediately and delete this email from your system. UK Research and Innovation (UKRI) has taken every reasonable precaution to minimise risk of this email or any attachments containing viruses or malware but the recipient should carry out its own virus and malware checks before opening the attachments. UKRI does not accept any liability for any losses or damages which the recipient may sustain due to presence of any viruses.
> > >
> > > _______________________________________________
> > > Phonopy-users mailing list
> > > Pho...@li...
> > > https://lists.sourceforge.net/lists/listinfo/phonopy-users
> >
> >
> >
> > --
> > Atsushi Togo
> >
> >
> > _______________________________________________
> > Phonopy-users mailing list
> > Pho...@li...
> > https://lists.sourceforge.net/lists/listinfo/phonopy-users
> >
> > _______________________________________________
> > Phonopy-users mailing list
> > Pho...@li...
> > https://lists.sourceforge.net/lists/listinfo/phonopy-users
>
>
>
> --
> Atsushi Togo
>
>
> _______________________________________________
> Phonopy-users mailing list
> Pho...@li...
> https://lists.sourceforge.net/lists/listinfo/phonopy-users
>
> _______________________________________________
> Phonopy-users mailing list
> Pho...@li...
> https://lists.sourceforge.net/lists/listinfo/phonopy-users



-- 
Atsushi Togo