Hi,
A.
Regarding SGLD, Ross's statement is correct with a couple of caveats
mostly due to me not finding the time to finish off some issues:
1) Romelia had signed off on me doing some code simplification
and cleanup involving sgld data fields and their reporting.
Romelia can you confirm that the situation is still as in our email on
Date: Thu, 22 Aug 2013 05:57:29 +0000
?
2) I owe Xiongwu an email on likely minor bugs in sander and pmemd SGLD.
I will send this soon.
3) We need to touch base on documentation.
B.
Regarding communication, we got into a situation that was relatively
easily resolved but could have become messy. Here are some highlights
and specific good communications in () that could have avoided the situation.
Romelia and Xiongwu agreed at the last meeting that Romelia would
add sgld to pmemd. (This list could have been notified of the plan.)
She started working on this in a private branch.
(This list could have been notified of the start of work.)
Dave asked me to add sgld to pmemd in May.
(This list could have been notified of the plan and start of work.)
I did a thorough grepping of the repository as my first step but
Romelia's work was in a private branch. I work in the master
branch, i commit frequently, and my detailed logs indicated the plan to
add sgld to pmemd. Romelia and Ross saw this and we got coordinated, but
we never did notify this list of the sgld work, so days ago Dave requested
work and Xiongwu volunteered to do work already done by Romelia.
The notification mechanism doesn't have to be this list, but it's
clear that some mechanism is needed. And the mechanism should include
public status info.
Happy New Year!
scott
On Mon, Jan 06, 2014 at 05:12:02PM -0800, Ross Walker wrote:
> I am copying in Romelia here since it seems the email she had registered
> for the developers list has 'expired'. Romelia can confirm for sure but I
> believe that the current situation with regards to SGLD is that it is done
> in the CPU code but there is no GPU version of it. This is likely how it
> will get released unless we magically find time to add it to the GPU code
> before the code freeze. It likely still needs to be documented though.
>
> All the best
> Ross
>
>
> On 1/3/14 12:21 PM, "Wu, Xiongwu (NIH/NHLBI) [E]" <wuxw.nhlbi.nih.gov>
> wrote:
>
> >Hi Jason,
> >Thanks for the information. I will work based on Scott's SGLD branch.
> >Romelia, it would be very helpful if you have an implementation to share.
> >Thanks!
> >Xiongwu
> >
> >> -----Original Message-----
> >> From: Jason Swails [mailto:jason.swails.gmail.com]
> >> Sent: Friday, January 03, 2014 1:14 PM
> >> To: amber-developers.ambermd.org
> >> Subject: Re: [AMBER-Developers] New Year fixes for Amber
> >>
> >> On Fri, 2014-01-03 at 12:23 -0500, David A Case wrote:
> >> > On Fri, Jan 03, 2014, Wu, Xiongwu (NIH/NHLBI) [E] wrote:
> >> >
> >> > > I will implement SGLD to pmemd and pmemd.cuda if no one else is
> >> doing it.
> >> > > Thanks and see you at the Amber meeting.
> >> >
> >> > Thanks!...note that there is a very beginning (look for "ifdef SGLD")
> >> in
> >> > $AMBERHOME/src/pmemd/src.
> >>
> >> To add on to this a little, I know Scott Brozell started to work on
> >> that
> >> while he was at Rutgers awhile ago. He started a branch (sgld) which
> >> he
> >> pushed to the main git repo, and it may be a better starting point
> >> (unless master cannot be easily merged, then it may be worth starting
> >> over).
> >>
> >> IIRC, though, he stopped after he was informed that Romelia from Ross's
> >> lab was working on the implementation. I'm not sure where we are with
> >> this now, but it may be worth coordinating efforts.
_______________________________________________
AMBER-Developers mailing list
AMBER-Developers.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber-developers
Received on Tue Jan 07 2014 - 03:00:03 PST
