[gmx-developers] time to make 5.1 happen
hess at kth.se
Tue Feb 10 16:06:23 CET 2015
My pull code change (together with the g_wham change) should go in. It
has been hanging in Gerrit for half a year.
And my LINCS changes might be nice for performance.
On 02/10/2015 04:00 PM, Mark Abraham wrote:
> Sorry for quietness on the management front, I've had a bunch of
> non-code tasks lately. We're a few weeks behind on my original
> timeline for 5.1-beta1. Various people had expressed hopes of putting
> code on gerrit, but few things have actually appeared, so I think we
> should pretty much regard the feature stuff on gerrit now as the
> maximum extent of 5.1. There's still a bunch of dead / experimental
> patches on gerrit that won't make the 5.1 cut, of course. Please speak
> up super-fast if that's a problem - you should have had code in gerrit
> a month and more ago!
> So I plan to roll 5.1-beta1 from master HEAD on Friday, and branch
> release-5-1 and Jenkins infrastructure at that time. Per my original
> suggestion, release-5-1 will be open for "feature"-type content for a
> few weeks while we deal with remaining code review of things currently
> on master. Once we release 5.1-rc1 (perhaps after another beta if
> appropriate) then features are frozen. The gerrit cherry-pick feature
> should be good for people to make clear that they are proposing a
> "feature" patch for release-5-1.
> I'd like to see a couple more of the new Sphinx-based docs commits at
> get merged, so that we can invite users to give feedback or contribute
> It would be great for our colleagues at StreamComputing if we get
> their OpenCL work out to the world in the first beta; I know Berk has
> given some feedback but is basically OK with the commit. Szilard has a
> few outstanding questions we can probably resolve quickly. I have a
> reservation about whether using more than one Nvidia GPU works, but
> since that is in any case not a useful configuration for performance,
> we can hack a temporary fatal error in front of that and ship
> something usable.
> Szilard's CUDA work needs to get in, but the remaining issues there
> are minor, I think.
> Berk's halo-decomposition is not yet clearly useful (right, Berk?) so
> perhaps we will leave that on the shelf for now.
> Erik's SIMD support patches need one more reviewer. It would be nice
> to be able to say to hardware devs that we already have preliminary
> support in released code when early-phase testing is on the cards.
> Happy hacking!
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the gromacs.org_gmx-developers