[CP2K:6648] Re: Compilations with Intel (XE 2013) for CP2K-trunk (2.7dev) & regtests errors

[Iain Bethune]

Great, that sounds reasonable.  The default versions included in the main source actually work quite well for most Intel CPUs.  10% gain is pretty good.

- Iain

--

Iain Bethune
Project Manager, EPCC

Email: ibet... at epcc.ed.ac.uk
Twitter: @IainBethune
Web: http://www2.epcc.ed.ac.uk/~ibethune
Tel/Fax: +44 (0)131 650 5201/6555
Mob: +44 (0)7598317015
Addr: 2404 JCMB, The King's Buildings, Peter Guthrie Tait Road, Edinburgh, EH9 3FD

> On 15 Jun 2015, at 13:42, Rolf David <rolf.d... at gmail.com> wrote:
> 
> Hi, 
> 
> So my results:
> With libgrid
> FC_comp="ifort -free -fpp"
> 
> FCFLAGS="-O3 -axAVX -xSSE4.2 -funroll-loops"
> 
> & make -j 1 alone in a node (16 cores but only one used 1)
> 
> 
> 
> and then H2O-64.inp, 16 cores.
> 
> 
> I go from 59 sec to 53 sec (10%) and from 56 to 55 sec (3%). So I guess it works.
> Precious time won.
> 
> Thanks again.
> 
> Rolf
> 
> 
> On Monday, June 15, 2015 at 1:56:27 PM UTC+2, IBethune wrote:
> Hi Rolf, 
> 
> libgrid should affect the performance in calculate_rho_elec and integrate_v_rspace.  For H2O-64 on low processor counts (e.g. 24) these should take up >50% of the runtime.  Larger H2O-XXX tend to mask the effect as the cubically scaling operations in OT tend to dominate. 
> 
> - Iain 
> 
> -- 
> 
> Iain Bethune 
> Project Manager, EPCC 
> 
> Email: ibe... at epcc.ed.ac.uk 
> Twitter: @IainBethune 
> Web: http://www2.epcc.ed.ac.uk/~ibethune 
> Tel/Fax: +44 (0)131 650 5201/6555 
> Mob: +44 (0)7598317015 
> Addr: 2404 JCMB, The King's Buildings, Peter Guthrie Tait Road, Edinburgh, EH9 3FD 
> 
> > On 15 Jun 2015, at 12:08, Rolf David <rolf... at gmail.com> wrote: 
> > 
> > Hi again, 
> > 
> > The 'workaround' proposed by Alfio 
> > 
> > > cat no.wlm 
> > batch_cmd() { 
> > $@ 
> > } 
> >   
> > Therefore you can use "-j 100 -w no". 
> > Note that I have never tried such a case, so I'm not sure it will work out-of-the-box. Let me how it goes. 
> > 
> > worked well in my case. I was able to compile. And run tiny1 all of it. And tiny2 in standard after. 
> > 
> > I'm moving on onto small1. 
> > 
> > Also I've compiled libgrid (on a compute node) and successfully integrated into CP2K. Is there a "benchmark" to see the effect, I mean a file in the tests-folder where people get difference without and with libgrid ? 
> > I run H2O-512.inp, but not noticeable difference. I run a test I had (QM/MM with hybrid functional) and I didn't see noticeable effect (on short tests), and in the readme it said about "Gaussian to Plane wave transformations", so I assume a speed up in some routine in GPW (or even GAPW no ?) 
> > Also, Iain said (https://groups.google.com/forum/#!searchin/cp2k/libgrid/cp2k/DU3KNkwM4as/8_bO8zjWZ0sJ) and here again, it's performance-critical. 
> > 
> > So if I have a "working" benchmark, I can see if I miscompiled it (no error in the out of the libgrid compilation), or maybe wrong compiler option and subroutines affected:  integrate_v_rspace for example  ? 
> > 
> > Regards, 
> > 
> > Rolf 
> > 
> > -- 
> > You received this message because you are subscribed to the Google Groups "cp2k" group. 
> > To unsubscribe from this group and stop receiving emails from it, send an email to cp2k+... at googlegroups.com. 
> > To post to this group, send email to cp... at googlegroups.com. 
> > Visit this group at http://groups.google.com/group/cp2k. 
> > For more options, visit https://groups.google.com/d/optout. 
> 
> 
> -- 
> The University of Edinburgh is a charitable body, registered in 
> Scotland, with registration number SC005336. 
> 
> 
> -- 
> You received this message because you are subscribed to the Google Groups "cp2k" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to cp2k+uns... at googlegroups.com.
> To post to this group, send email to cp... at googlegroups.com.
> Visit this group at http://groups.google.com/group/cp2k.
> For more options, visit https://groups.google.com/d/optout.


-- 
The University of Edinburgh is a charitable body, registered in
Scotland, with registration number SC005336.