[gmx-developers] parallel make problems

[Mark Abraham]

On Mon, Jun 17, 2013 at 6:16 PM, Manuel Nuno Melo <m.n.melo at rug.nl> wrote:

> Hi,
>
> I have also had linking problems when making in parallel. In my case they
> could be traced back to the option to let GMX download/build its own fftw
> (-DGMX_BUILD_OWN_FFTW=ON).
>
> It seems that only one of make's threads starts building fftw, while the
> others go ahead building/linking GMX. Since fftw compilation is not ready
> by the time it is needed, GMX linking is botched.
>

Yes, Rossen first showed this to me. I don't know if the underlying issue
is that the dependency cannot be described properly, or that we're not
doing it properly. If it's a problem, people are welcome to contribute a
fix! :-)

Mark


>
> Cheers,
> Manel
>
> > Hi,
> >
> > I too suspect filesystem issues or clock skews. I think I tested make -j
> and make -j 12. The cluster is currently down for maintenance, so I can't
> inspect the details at the moment.
> >
> > On 5 Apr 2013, at 13:14, Alexey Shvetsov <alexxy at omrb.pnpi.spb.ru>
> wrote:
> >
> > > Hi Erik
> > >
> > > What are underlaying filesystem on this cluster? If it slow or
> overloaded
> > > somehow it may lead to parallel make issues. Also it may be related to
> make
> > > version (some old versions may expose such behavior). How many make
> threads do
> > > you issued? I tryed with make -j64 and it builds fine with recent cmake
> > > (2.8.10) and make (3.82) utility.
> > >
> > >
> > > В письме от 5 апреля 2013 11:55:27 пользователь Erik Marklund написал:
> > >> Hi,
> > >>
> > >> Building gromacs 4.6.1 failed whenever I issued parallel make, i.e.
> make -j.
> > >> I reported this to the cluster admins since I had never seen such
> behaviour
> > >> before from gromacs' side, and here's their reply. I can't tell
> whether
> > >> gromacs is at fault or the cluster.
> > >>
> > >> Erik
> > >>
> > >> Begin forwarded message:
> > >>> Hi,
> > >>>
> > >>>> I was compiling gromacs on tintin's login node the other day and it
> > >>>>
> > >>>>  seems that parallel make, i.e. make -j, doesn't work on tintin. I
> > >>>>  got linker errors that never showed up when make was run serially.
> > >>>>  I've never encountered such behaviour before.
> > >>>
> > >>> Without any more information (or being able to look for actual files
> right
> > >>> now), I'd guess this is a problem with the makefiles rather than the
> > >>> actual make. It seems somewhat unexpected that CMake would create
> > >>> makefiles that aren't safe for parallel building, but it does seem
> the
> > >>> most likely culprit (assuming it doesn't let developers add rules
> > >>> directly to the makefile to work around problems, I don't remember if
> > >>> that's possible).
> > >>>
> > >>> That you only see the problem on tintin can likely be explained by
> timing
> > >>> or other non deterministic factors.
> > > --
> > > Best Regards,
> > > Alexey 'Alexxy' Shvetsov
>
> --
> gmx-developers mailing list
> gmx-developers at gromacs.org
> http://lists.gromacs.org/mailman/listinfo/gmx-developers
> Please don't post (un)subscribe requests to the list. Use the
> www interface or send it 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/20130617/a6fc81cc/attachment.html>