[gmx-developers] [gmxana] gmx cluster is very buggy

[Erik Lindahl]

Hi,

> On 13 May 2016, at 07:01, David van der Spoel <spoel at xray.bmc.uu.se> wrote:
> 
> On 13/05/16 05:38, Yunlong Liu wrote:
>> 
>> Hi Gromacs developers,
>> 
>> I recently used gmx cluster for cluster analysis but I got some very
>> buggy result. The RMSD distribution I got is from 0.01 - 0.5 and I set a
>> cutoff of 0.1. Why did the program give me only one cluster? When I
>> remove the cutoff tag, (I thought the default value is 0.1 as well),
>> then it gave me 30+ clusters.
> If you suspect a bug please file a bug report on redmine.gromacs.org <http://redmine.gromacs.org/>

No, please do *NOT* use redmine unless you are virtually certain it is a bug. It is not a substitute for reading documentation and understanding the algorithms used in the program, or trying to get help on this list.

In particular when we’re dashing for a release we’d like to avoid lots of extra reports that turn out not to be bugs at all - such bug reports will simply be rejected without any further comments.

Cheers,

Erik


>> 
>> Also, I found "-om" tag and "-dm" tag is almost useless.
>> 
>> *** A developer call ***
>> As the trajectories getting larger and larger, a single core analysis
>> program is more and more slow. For some analysis program, like "cluster
>> analysis", they have such intrinsic properties that naturally support
>> parallelism. Why not parallelize them to increase speed?

Many of them are already parallelized, and more are added for every release.

Cheers,

Erik
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://maillist.sys.kth.se/pipermail/gromacs.org_gmx-developers/attachments/20160513/25af9fba/attachment.html>