[gmx-developers] freeenergy.relative unstable?

[Szilárd Páll]

Hi,

I can still reliably reproduce the issue with the latest code on an
AMD Piledriver processor (FX-8350). I think it would be quite
important to fix this because in this state the regression test can
not serve as a check for the correct functioning of a GROMACS
installation.

This is how checkpot.out looks like:
Bond             step  16:       49.0311,  step  16:      48.9803
Bond             step  17:       45.1255,  step  17:      45.0732
LJ (SR)          step  24:       5586.75,  step  24:      5592.45
LJ (SR)          step  25:       5588.03,  step  25:      5594.11
LJ (SR)          step  26:       5590.09,  step  26:      5596.53
LJ (SR)          step  27:       5593.28,  step  27:       5600.1
LJ (SR)          step  28:       5597.99,  step  28:      5605.17
LJ (SR)          step  29:       5604.25,  step  29:      5611.79
LJ (SR)          step  30:       5611.68,  step  30:      5619.58
LJ (SR)          step  31:        5620.3,  step  31:      5628.52
LJ (SR)          step  32:       5629.74,  step  32:      5638.21
LJ (SR)          step  33:       5639.08,  step  33:      5647.81
LJ (SR)          step  34:       5648.18,  step  34:      5657.17
LJ (SR)          step  35:       5656.94,  step  35:      5666.15
LJ (SR)          step  36:        5665.2,  step  36:      5674.66
LJ (SR)          step  37:       5673.13,  step  37:      5682.82
LJ (SR)          step  38:          5681,  step  38:      5690.93
LJ (SR)          step  39:       5689.07,  step  39:      5699.21
LJ (SR)          step  40:       5697.36,  step  40:      5707.73

Cheers,
--
Szilárd


On Fri, Oct 18, 2013 at 9:30 PM, Shirts, Michael (mrs5pt)
<mrs5pt at eservices.virginia.edu> wrote:
>
> Given the fact the potential generally doesn't diverge immediately, and
> these code changes involve changes in the order of operation of the
> calculations, I would expect it's probably a numerical instability issue. I
> can have a look at this case more carefully as we get ready for 5.0.
>
> Best,
> ~~~~~~~~~~~~
> Michael Shirts
> Assistant Professor
> Department of Chemical Engineering
> University of Virginia
> michael.shirts at virginia.edu
> (434)-243-1821
>
>
>> From: Szilárd Páll <pall.szilard at gmail.com>
>> Reply-To: Discussion list for GROMACS development <gmx-developers at gromacs.org>
>> Date: Fri, 18 Oct 2013 20:12:35 +0200
>> To: Discussion list for GROMACS development <gmx-developers at gromacs.org>
>> Subject: [gmx-developers] freeenergy.relative unstable?
>>
>> Hi,
>>
>> The freeenergy.relative test has been failing on a regular basis, some
>> of the recent cases are:
>> http://jenkins.gromacs.org/job/Gromacs_Gerrit_4_6/4008/
>> http://jenkins.gromacs.org/job/Gromacs_Gerrit_4_6/4014/
>> http://jenkins.gromacs.org/job/Gromacs_Gerrit_4_6/4019/
>> http://jenkins.gromacs.org/job/Gromacs_Gerrit_4_6/4064/
>> http://jenkins.gromacs.org/job/Gromacs_Gerrit_4_6/4084/
>> http://jenkins.gromacs.org/job/Gromacs_Gerrit_4_6/4095/
>>
>> Based on these the failure seems to occur more often on the Sandy
>> Bridge test machine (and 4/5 with gcc 4.8).
>>
>> Is there any concern that there may be a slight bug present or can we
>> safely blame this on the instability of this particular test?
>>
>> Cheers,
>> --
>> Szilárd
>> --
>> 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.