[gmx-developers] Any known gcc 4.1 issues?

Szilárd Páll szilard.pall at cbr.su.se
Wed Aug 4 11:54:42 CEST 2010


Hi,

We are trying to figure out whether we should keep the explicit ban
through the build scripts on gcc 4.1.x or not. Does anyone know of any
bugs in gromacs that can (possibly) be attributed to this gcc version?

Roland/Xiaohu: do you still see the fluctuations mentioned in you earlier mail?

--
Szilárd



On Tue, Jul 6, 2010 at 12:59 AM, Roland Schulz <roland at utk.edu> wrote:
>
>
> On Mon, Jul 5, 2010 at 5:59 PM, <hess at sbc.su.se> wrote:
>>
>> Hi,
>>
>> Which Gromacs version is this?
>
>  Head from two weeks ago (commit: 0541796)
>>
>> And this is not with GBSA?
>
> no GBSA. PME.
>>
>> I assume this is in parallel?
>
> yes
>>
>> In all 4.0 versions there is a bug in the graph code,
>> but I have never seen that being triggered.
>
> you mean GCC 4 or GROMACS 4?
>>
>> You probably wouldn't know if 4.1.3 does better?
>
> no. Only tested those two GCC versions.
> BTW: Are MDP files supposed to be upward compatible? Because the default for
> nstcalcenergy is -1 in grompp. Thus if one has an old mdp file without the
> nstcalcenergy option one gets a different tpr file with grompp from HEAD
> than grompp 4.0.7. Should we set nstcalcenergy to 1 if it is missing or
> print a NOTICE if is it is missing and we set it to -1?
> (We noticed this while looking at the origin of the large
> energy fractionation. The  (short-term)  fluctuation are increased with
> nstcalcenergy -1 (it is using 10) but much smaller than the increase caused
> by GCC 4.1 - which are longer term fluctuations)
> Roland
>>
>> Berk
>>
>> > Hi,
>> >
>> > we have a simulation in our group (Xiaohu's system) which shows much
>> > larger
>> > energy fluctuations (on a ~10ns time-scale) if compiled with 4.1.2.
>> > These fluctuation are large enough for the system to explode
>> > infrequently.
>> > The system behaves as expected (without the large
>> > fluctuations/explosions)
>> > when compiled with 4.4.4.
>> >
>> > Roland
>> >
>> > On Thu, Jun 24, 2010 at 11:05 AM, Berk Hess <hess at cbr.su.se> wrote:
>> >
>> >> Hi,
>> >>
>> >> We just found out a that a bug that frequently showed up and we thought
>> >> might be due to a bug in gcc 4.1
>> >> was actually a bug in the GB sse2 (not sse3) code that only got
>> >> triggered with gcc 4.1 with -O2 or -O3.
>> >>
>> >> So my question is now:
>> >> Do you know of any issues with gcc 4.1.x on non-GB simulations?
>> >>
>> >> Thanks,
>> >>
>> >> Berk
>> >>
>> >> --
>> >> 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.
>> >>
>> >
>> >
>> >
>> > --
>> > ORNL/UT Center for Molecular Biophysics cmb.ornl.gov
>> > 865-241-1537, ORNL PO BOX 2008 MS6309
>> > --
>> > 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.
>>
>> --
>> 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.
>
>
>
> --
> ORNL/UT Center for Molecular Biophysics cmb.ornl.gov
> 865-241-1537, ORNL PO BOX 2008 MS6309
>
> --
> 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.
>



More information about the gromacs.org_gmx-developers mailing list