[gmx-developers] ICC 14 support

Mark Abraham mark.j.abraham at gmail.com
Tue Sep 30 16:59:48 CEST 2014 

On Tue, Sep 30, 2014 at 4:48 PM, Kevin Chen <fch6699 at gmail.com> wrote:

> Have you guys compared GROMACS performance between icc and gcc? Per
> Szilárd's note, seems gcc is better, am I right?
>

Yeah, I'd take that one to the bank. Of course, we're always interested to
hear of observations (either way).

Mark


> Cheers,
>
> Kevin
>
>
> -----Original Message-----
> From: gromacs.org_gmx-developers-bounces at maillist.sys.kth.se [mailto:
> gromacs.org_gmx-developers-bounces at maillist.sys.kth.se] On Behalf Of
> Szilárd Páll
> Sent: Monday, September 29, 2014 3:50 PM
> To: Discussion list for GROMACS development
> Subject: Re: [gmx-developers] ICC 14 support
>
> BTW: you'll have less trouble when combining with CUDA as well as better
> performance with gcc!
>
> Cheers,
> --
> Szilárd
>
>
> On Mon, Sep 29, 2014 at 10:14 PM, Kevin Chen <fch6699 at gmail.com> wrote:
> > Hi Roland,
> >
> >
> >
> > Thanks for the reply! Looks like the error messages were generated by
> > Cuda6.0 (see errors below) instead of GROMACS. Switching back to
> > ICC13.1 and turning off static-libraries worked for us.
> >
> >
> >
> > Best,
> >
> >
> >
> > Kevin
> >
> >
> >
> >
> >
> >
> >
> > ======================================================================
> > ===============================================================
> >
> >
> >
> > “In file included from /opt/apps/cuda/6.0/include/cuda_runtime.h(59),
> >
> > /opt/apps/cuda/6.0/include/host_config.h(72): catastrophic error:
> > #error
> > directive: -- unsupported ICC configuration! Only ICC 13.1 on Linux
> > x86_64 is supported!
> >
> >   #error -- unsupported ICC configuration! Only ICC 13.1 on Linux
> > x86_64 is supported!
> >
> >    ^
> >
> >
> >
> > /opt/apps/cuda/6.0/include/host_config.h(72): catastrophic error:
> > #error
> > directive: -- unsupported ICC configuration! Only ICC 13.1 on Linux
> > x86_64 is supported!
> >
> >   #error -- unsupported ICC configuration! Only ICC 13.1 on Linux
> > x86_64 is supported!
> >
> >    ^
> >
> >
> >
> >                  from
> >
> /admin/build/admin/rpms/stampede/BUILD/gromacs-5.0.1/src/gromacs/gmxlib/gpu_utils/
> gpu_utils.cu(0):
> >
> > /opt/apps/cuda/6.0/include/host_config.h(72): catastrophic error:
> > #error
> > directive: -- unsupported ICC configuration! Only ICC 13.1 on Linux
> > x86_64 is supported!
> >
> >   #error -- unsupported ICC configuration! Only ICC 13.1 on Linux
> > x86_64 is supported!
> >
> >    ^
> >
> >
> >
> > In file included from /opt/apps/cuda/6.0/include/cuda_runtime.h(59),
> >
> >                  from
> >
> /admin/build/admin/rpms/stampede/BUILD/gromacs-5.0.1/src/gromacs/gmxlib/cuda_tools/
> pmalloc_cuda.cu(0):
> >
> > /opt/apps/cuda/6.0/include/host_config.h(72): catastrophic error:
> > #error
> > directive: -- unsupported ICC configuration! Only ICC 13.1 on Linux
> > x86_64 is supported!
> >
> >
> >
> > “
> >
> >
> >
> >
> >
> >
> >
> >
> >
> > From: gromacs.org_gmx-developers-bounces at maillist.sys.kth.se
> > [mailto:gromacs.org_gmx-developers-bounces at maillist.sys.kth.se] On
> > Behalf Of Roland Schulz
> > Sent: Monday, September 29, 2014 11:16 AM
> > To: gmx-developers at gromacs.org
> > Subject: Re: [gmx-developers] ICC 14 support
> >
> >
> >
> > Hi,
> >
> >
> >
> > what problem do you have with ICC14? Both ICC14 and ICC15 should work
> fine.
> > There was an issue with ICC14+static-libraries (9e8061e13f48, 4.6.7
> > and
> > 5.0.1) and ICC14+unit-tests(b0e60e91add6, 5.0.1). Both are fixed in
> > release-5-0 and will be included 5.0.2. You can either use the
> > release-5-0 branch, just apply the patch, wait till 5.0.2 (should be
> > soon), or don't use static-libraries (default) and don't try to run
> > the unit-tests (the compiler issue isn't present in the main code thus
> > even though the unit tests fail the actual program is OK).
> >
> >
> >
> > Roland
> >
> >
> >
> > On Mon, Sep 29, 2014 at 10:35 AM, Kevin Chen <fch6699 at gmail.com> wrote:
> >
> > Hi Guys,
> >
> > Was wondering if GROMACS will support ICC 14 in the near future?
> >
> > Kevin Chen, Ph.D.
> > HPC Applications, TACC
> >
> >
> > -----Original Message-----
> > From: gromacs.org_gmx-developers-bounces at maillist.sys.kth.se
> > [mailto:gromacs.org_gmx-developers-bounces at maillist.sys.kth.se] On
> > Behalf Of Alexey Shvetsov
> > Sent: Sunday, September 28, 2014 3:09 PM
> > To: gmx-developers at gromacs.org
> > Subject: Re: [gmx-developers] Possible bug in gmx
> >
> > Mark Abraham писал 28-09-2014 18:17:
> >> How about a redmine issue - this thread's not about GROMACS
> >> development, per se ;-)
> >
> > Sorry about that =D
> >
> > Redmine issue http://redmine.gromacs.org/issues/1607
> > With relevant tpr file attached
> >>
> >> Mark
> >>
> >> On Sun, Sep 28, 2014 at 3:18 PM, Alexey Shvetsov
> >> <alexxy at omrb.pnpi.spb.ru> wrote:
> >>
> >>> Hi Berk!
> >>>
> >>> Its not a cut and paste error, also there are no pdb dumps.
> >>> Also I see this error before with other systems.
> >>>
> >>> I can provide tpr file for that system
> >>>
> >>> https://biod.pnpi.spb.ru/~alexxy/gmx/psa_pep_ctrl.md_npt.tpr [1]
> >>>
> >>> Berk Hess писал 28-09-2014 16:37:
> >>>
> >>> Hi,
> >>>
> >>> I assume that your old and be coordinates being identical is correct
> >>> and not a cut-and-paste error.
> >>> This seems a bit strange or do you freeze part of the system?
> >>> The only things moving here are then the domain boundaries and I
> >>> don't see an issue there, since they only moved a little.
> >>>
> >>> Do you have any more output besides the error message? PDB dump
> >>> files maybe?
> >>>
> >>> Cheers,
> >>>
> >>> Berk
> >>>
> >>> On 09/28/2014 02:22 PM, Alexey Shvetsov wrote:
> >>> Hi,
> >>>
> >>> just wanna add that this error seems to be reproducable even on
> >>> single node. Also i get same error for gpu runs.
> >>> However i dont see it in large systems (800k+ atoms) running on
> >>> large number of cpus (512+)
> >>>
> >>> Alexey Shvetsov писал 28-09-2014 13:44:
> >>> Hi,
> >>>
> >>> DD grid is
> >>>
> >>> Domain decomposition grid 4 x 1 x 1, separate PME ranks 0 PME domain
> >>> decomposition: 4 x 1 x 1
> >>>
> >>> for 4 node setup
> >>>
> >>> and
> >>>
> >>> Domain decomposition grid 4 x 2 x 1, separate PME ranks 0 PME domain
> >>> decomposition: 4 x 2 x 1
> >>>
> >>> for 8 node setup
> >>>
> >>> It's reproducable with 5.0 release and latest git master. I try to
> >>> check if its reproducable with 1 node. Also i can provide tpr file
> >>> for this system
> >>>
> >>> Mark Abraham писал 28-09-2014 13:28:
> >>> Hi,
> >>>
> >>> It's hard to say on that information. There were some issues fixed
> >>> in the lead-up to GROMACS 5 with DD not always working with 2
> >>> domains in a direction, but that's a pure guess. I'd assume you can
> >>> reproduce this with release-5-0 branch. Do you observe it with a
> single domain?
> >>> If not, then it's surely a bug (and should be submitted to redmine).
> >>>
> >>> Mark
> >>>
> >>> On Sun, Sep 28, 2014 at 11:18 AM, Alexey Shvetsov
> >>> <alexxy at omrb.pnpi.spb.ru> wrote:
> >>>
> >>> Hi all!
> >>>
> >>> I'm doing some tests with small peptide and constantly getting this
> >>> error.
> >>> I get it with few systems.
> >>>
> >>> Systems sizes are around 10k or 20k
> >>> I run it on 4 or 8 old nodes each with two xeon 54xx series
> >>>
> >>> starting mdrun '2ZCH_3 in water'
> >>> 50000000 steps, 100000.0 ps (continuing from step 1881000, 3762.0
> >>> ps).
> >>>
> >>> Step 13514000:
> >>> The charge group starting at atom 6608 moved more than the distance
> >>> allowed by the domain decomposition (1.112924) in direction X
> >>> distance out of cell -1.193103 Old coordinates: 5.467 0.298 3.636
> >>> New
> >>> coordinates: 5.467 0.298 3.636 Old cell boundaries in direction X:
> >>> 4.037 5.382 New cell boundaries in direction X: 4.089 5.452
> >>>
> >>>
> >> ---------------------------------------------------------------------
> >> -
> >> ----
> >>> MPI_ABORT was invoked on rank 3 in communicator MPI_COMM_WORLD with
> >>> errorcode 1.
> >>>
> >>> NOTE: invoking MPI_ABORT causes Open MPI to kill all MPI processes.
> >>> You may or may not see output from other processes, depending on
> >>> exactly when Open MPI kills them.
> >>>
> >>>
> >> ---------------------------------------------------------------------
> >> -
> >> ----
> >>>
> >>> -------------------------------------------------------
> >>> Program mdrun_mpi, VERSION 5.1-dev-20140922-20c00a9-dirty-unknown
> >>> Source code file:
> >>>
> >>>
> >> /var/tmp/alexxy/portage/sci-chemistry/gromacs-9999/work/gromacs-9999/
> >> s
> >> rc/gromacs/mdlib/domdec.cpp,
> >>> line: 4388
> >>>
> >>> Fatal error:
> >>> A charge group moved too far between two domain decomposition steps
> >>> This usually means that your system is not well equilibrated For
> >>> more information and tips for troubleshooting, please check the
> >>> GROMACS website at http://www.gromacs.org/Documentation/Errors [2]
> >>> [1]
> >>> -------------------------------------------------------
> >>>
> >>> -- Best Regards,
> >>> Alexey 'Alexxy' Shvetsov, PhD
> >>> Department of Molecular and Radiation Biophysics FSBI Petersburg
> >>> Nuclear Physics Institute, NRC Kurchatov Institute, Leningrad
> >>> region, Gatchina, Russia mailto:alexxyum at gmail.com
> >>> mailto:alexxy at omrb.pnpi.spb.ru
> >>> -- Gromacs Developers mailing list
> >>>
> >>> * Please search the archive at
> >>> http://www.gromacs.org/Support/Mailing_Lists/GMX-developers_List
> >>> [3] [2]
> >>> before posting!
> >>>
> >>> * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists [4]
> >>> [3]
> >>>
> >>> * For (un)subscribe requests visit
> >>>
> >>>
> >> https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-develope
> >> r
> >> s
> >>> [5]
> >>> [4] or send a mail to gmx-developers-request at gromacs.org.
> >>>
> >>> Links:
> >>> ------
> >>> [1] http://www.gromacs.org/Documentation/Errors [2] [2]
> >>> http://www.gromacs.org/Support/Mailing_Lists/GMX-developers_List [3]
> >>> [3] http://www.gromacs.org/Support/Mailing_Lists [4] [4]
> >>>
> >> https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-develope
> >> r
> >> s
> >>> [5]
> >>
> >>  -- Best Regards,
> >>  Alexey 'Alexxy' Shvetsov, PhD
> >>  Department of Molecular and Radiation Biophysics  FSBI Petersburg
> >> Nuclear Physics Institute, NRC Kurchatov Institute,  Leningrad
> >> region, Gatchina, Russia  mailto:alexxyum at gmail.com
> >> mailto:alexxy at omrb.pnpi.spb.ru
> >>
> >>  --
> >>  Best Regards,
> >>  Alexey 'Alexxy' Shvetsov, PhD
> >>  Department of Molecular and Radiation Biophysics  FSBI Petersburg
> >> Nuclear Physics Institute, NRC Kurchatov Institute,  Leningrad
> >> region, Gatchina, Russia  mailto:alexxyum at gmail.com
> >> mailto:alexxy at omrb.pnpi.spb.ru
> >>  --
> >>  Gromacs Developers mailing list
> >>
> >>  * Please search the archive at
> >> http://www.gromacs.org/Support/Mailing_Lists/GMX-developers_List [3]
> >> before posting!
> >>
> >>  * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists [4]
> >>
> >>  * For (un)subscribe requests visit
> >>
> >> https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-develope
> >> r s [5] or send a mail to gmx-developers-request at gromacs.org.
> >>
> >>
> >> Links:
> >> ------
> >> [1] https://biod.pnpi.spb.ru/~alexxy/gmx/psa_pep_ctrl.md_npt.tpr
> >> [2] http://www.gromacs.org/Documentation/Errors
> >> [3] http://www.gromacs.org/Support/Mailing_Lists/GMX-developers_List
> >> [4] http://www.gromacs.org/Support/Mailing_Lists
> >> [5]
> >> https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-develope
> >> r
> >> s
> >
> > --
> > Best Regards,
> > Alexey 'Alexxy' Shvetsov, PhD
> > Department of Molecular and Radiation Biophysics FSBI Petersburg
> > Nuclear Physics Institute, NRC Kurchatov Institute, Leningrad region,
> > Gatchina, Russia mailto:alexxyum at gmail.com
> > mailto:alexxy at omrb.pnpi.spb.ru
> > --
> > Gromacs Developers mailing list
> >
> > * Please search the archive at
> > http://www.gromacs.org/Support/Mailing_Lists/GMX-developers_List
> > before posting!
> >
> > * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
> >
> > * For (un)subscribe requests visit
> > https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-developer
> > s or send a mail to gmx-developers-request at gromacs.org.
> >
> > --
> > Gromacs Developers mailing list
> >
> > * Please search the archive at
> > http://www.gromacs.org/Support/Mailing_Lists/GMX-developers_List
> > before posting!
> >
> > * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
> >
> > * For (un)subscribe requests visit
> > https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-developer
> > s or send a mail to gmx-developers-request at gromacs.org.
> >
> >
> >
> >
> >
> > --
> > ORNL/UT Center for Molecular Biophysics cmb.ornl.gov 865-241-1537,
> > ORNL PO BOX 2008 MS6309
> >
> >
> > --
> > Gromacs Developers mailing list
> >
> > * Please search the archive at
> > http://www.gromacs.org/Support/Mailing_Lists/GMX-developers_List
> > before posting!
> >
> > * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
> >
> > * For (un)subscribe requests visit
> > https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-developer
> > s or send a mail to gmx-developers-request at gromacs.org.
> --
> Gromacs Developers mailing list
>
> * Please search the archive at
> http://www.gromacs.org/Support/Mailing_Lists/GMX-developers_List before
> posting!
>
> * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
>
> * For (un)subscribe requests visit
> https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-developers
> or send a mail to gmx-developers-request at gromacs.org.
>
> --
> Gromacs Developers mailing list
>
> * Please search the archive at
> http://www.gromacs.org/Support/Mailing_Lists/GMX-developers_List before
> posting!
>
> * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
>
> * For (un)subscribe requests visit
> https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-developers
> or send a mail to gmx-developers-request at gromacs.org.
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://maillist.sys.kth.se/pipermail/gromacs.org_gmx-developers/attachments/20140930/9ce841f4/attachment-0001.html>

More information about the gromacs.org_gmx-developers mailing list