[gmx-developers] last days for 5.1

[Szilárd Páll]

Hi,


--
Szilárd

On Wed, Aug 5, 2015 at 9:15 AM, Mark Abraham <mark.j.abraham at gmail.com>
wrote:

> Hi all,
>
> I know we're all keen to get 5.1 shipped and move on to new work, so I
> thought I'd list the things I know would be good to get done so we can feel
> happy about shipping the real release, e.g. early next week. Then you can
> get some brain-share from me for new and exciting things ;-)
>
> Performance regression testing - I've done some manual testing here,
> things look fine (several issues flagged by Szilard over the lifetime of
> 5.1 have all been resolved, one way or another) but there are a few things
> left I want to check.
>

AFAICT there are still a few relatively small issues in redmine, most of
them by me. I'd say at least 1782 and 1784 would be best fixed before
release. I've a draft for the latter in gerrit, but needs a bit more
attention. Will squeeze out some time today to get it ready. May need help
with testing (and this may very well fix the DD+OpenCL issue).


>
> I'll re-run some ensemble-checking tests on a range of hardware. This is
> important to do with all the bells and whistles on/default, because we have
> to disable some of those for the regression testing to work reliably.
>
> I will also try re-run a battery of free-energy tests we ran while fixing
> issues just before 5.0. Nothing should have changed, but we may as well
> find out.
>
> I've checked the build on BlueGene/Q (found a few issues now on gerrit or
> redmine, some solved already), and will check it on a few GPU clusters.
> Erik can you check in on K, please?
>
> We need to add some mdrun-time check that does not try to run (AMD?)
> OpenCL kernels with MacOS 10.3 (but only the necessary combo should be
> prohibited, and preferably not all of mdrun). A cmake-time warning for that
> combo would also be nice.
>
> We changed some default behaviour for how mdrun picks things like numbers
> of cores per rank, etc. Carsten, can you please try a low-key gmx tune_pme
> run to see we haven't broken things there?
>
> There's a few TODOs in docs/user-guide/mdrun-performance.rst that I'd like
> to deal with (and a couple of pages to add), so that version-specific
> information isn't living on a wiki page with switch(version). Szilard, are
> there things that you'd like to add for running with GPUs?
>

I'll try to find some time to read through it. One thing we definitely need
to emphasize is the new PP rank to GPU mapping automation.

--
Szilárd


>
> http://www.gromacs.org/Documentation/How-tos/Tool_Changes_for_5.0 is
> starting to leak a bit, because we have changes for 5.0 and probably some
> changes for 5.1 that would need similar documentation, and some of those
> areas overlap (IIRC gmx distance and gmx pairlist?). So I'll have a think
> about what we can do better here. Thoughts, anyone?
>
> Anything I've forgotten?
>
> Otherwise, please keep an eye on
> https://gerrit.gromacs.org/#/q/status:open+project:gromacs+branch:release-5-0
> and
> https://gerrit.gromacs.org/#/q/status:open+project:gromacs+branch:release-5-1
> and review as you feel able!
>
> Cheers,
>
> Mark
>
> --
> 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/20150805/2afcb31a/attachment.html>