> On Jan 6, 2014, at 3:59 PM, Ross Walker <ross.rosswalker.co.uk> wrote:
>
> The Monte Carlo barostat - both CPU and GPU.
Oh. I haven't seen any commits about that. It should be working in master now...
>
>
>> On 1/6/14 9:41 AM, "Ross Walker" <ross.rosswalker.co.uk> wrote:
>>
>> Ben Madej has been working extensively on this over the last few months so
>> I will let him provide an update here but my understanding is that this is
>> fixed although what is in the master branch is NOT up to date.
>>
>>
>> NOTE to all - the GPU code (and quite a bit of the PMEMD code) in the
>> master is VERY out of date and you should not be using this as your
>> reference for how things stand. It was left like this because a lot of
>> people use the git tree for production runs and I didn't want to cause
>> issues with development code. The current state of the art for pmemd and
>> pmemd.cuda is thus the GPU_Tachyon branch and this is where development
>> has been going on. This week I plan to replace the master for
>> $AMBERHOME/src/pmemd/ with the contents of GPU_Tachyon.
>>
>> All the best
>> Ross
>>
>>> On 1/4/14 4:50 PM, "Jason Swails" <jason.swails.gmail.com> wrote:
>>>
>>>> On Thu, Jan 2, 2014 at 10:20 AM, case <case.biomaps.rutgers.edu> wrote:
>>>>
>>>> Hi everyone, and happy new year!
>>>>
>>>> Below are some Amber updates that would be great to have some attention
>>>> paid to, somewhat in order of decreasing priority. For the top few, it
>>>> would
>>>> be really great to have fixes available for Amber14.
>>>>
>>>> 1. Monte-Carlo barostat for pmemd and pmemd.cuda (and sander, with
>>>> lower
>>>> priority). There are reference implementations in mdgx and openMM;
>>>> Jason
>>>> started a port to pmemd, but the code doesn't work. This would be a
>>>> big
>>>> win
>>>> both for efficiency and for physical realism.
>>>
>>> Pawel seems to have found the problem, so this should be fixed before the
>>> Amber meeting. We'll have to work with Scott in order to get the CUDA
>>> code
>>> taking maximum advantage of the performance benefits. The CUDA code will
>>> work with the fixed barostat, but it still computes the virial which
>>> negates one of the two advantages of the MC barostat in the first place
>>> (the second, of course, being that the MC barostat is rigorously correct
>>> from a stat mech viewpoint).
>>>
>>> Cheers, and happy new year!
>>>
>>> Jason
>>>
>>> --
>>> Jason M. Swails
>>> BioMaPS,
>>> Rutgers University
>>> Postdoctoral Researcher
>>> _______________________________________________
>>> AMBER-Developers mailing list
>>> AMBER-Developers.ambermd.org
>>> http://lists.ambermd.org/mailman/listinfo/amber-developers
>>
>>
>>
>> _______________________________________________
>> AMBER-Developers mailing list
>> AMBER-Developers.ambermd.org
>> http://lists.ambermd.org/mailman/listinfo/amber-developers
>
>
>
> _______________________________________________
> AMBER-Developers mailing list
> AMBER-Developers.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber-developers
_______________________________________________
AMBER-Developers mailing list
AMBER-Developers.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber-developers
Received on Mon Jan 06 2014 - 13:30:03 PST
