[gmx-developers] Rewind and reread trj from C++ api

[Berk Hess]

On 11/13/2014 05:23 PM, Teemu Murtola wrote:
>
> Hi,
>
> On Nov 13, 2014 9:51 AM, "Alexey Shvetsov" <alexxy at omrb.pnpi.spb.ru 
> <mailto:alexxy at omrb.pnpi.spb.ru>> wrote:
> > Personaly i thougth about option 2 (store intermediate data). 
> However in some cases it will lead to very large memory footprint.
>
> Sure, but since you haven't told us anything about what you want to 
> do, it's impossible to say whether that applies to your case. And 
> there certainly are cases where this is possible.
>
> > Example when you need to rewind and reread trajectory already 
> present in gromacs tools (e.g. g_covar it computes average structure 
> and after that computes displacements over average possitions).
>
> That is a particularly bad example, because
> 1) it is possible to compute the average and covariance in a single 
> pass, so my approach 1 results in much more effective analysis, and
>
IIRC g_covar does this because you (might) want to fit to the average 
structure and fitting is a non-linear operation. I would think that 
otherwise I would have implemented a single pass, as you write.

Cheers,

Berk
>
> 2) even if it weren't, there would be essentially no overhead to write 
> two different tools, one to compute the average, and another one that 
> takes the average structure as input and computes the covariance.
>
> > Btw is it possible to call AnalyzeFrame recursively from C++?
>
> That depends fully on what you put into that function... The framework 
> itself surely doesn't call it recursively, but if you don't use any 
> constructs that expects the guarantees that the framework provides and 
> your recursion breaks, why not?
>
> > Using python wrapper is a good idea, however it currenlty in early 
> stage (e.g no selections).
>
> If you followed my recommendation, you wouldn't need selection support 
> to implement my suggestion 4 (as long as you wrote your core tool in 
> C++, and only the glue with Python). But if you don't even want to 
> explore this approach for the bindings, there's not much I can do...
>
> Best regards,
> Teemu
>
>
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://maillist.sys.kth.se/pipermail/gromacs.org_gmx-developers/attachments/20141113/42fd6232/attachment.html>