[gmx-developers] restarting from checkpoint--option to append to output files?

[Mark Abraham]

Anton Feenstra wrote:
> Daniel Larsson wrote:
>> Another thing that also was mentioned during the workshop and that I 
>> would love to have is the possibility to feed the analysis tools with 
>> a set of trajectory and/or energy files from a continued simulation 
>> without having to concatenate them first.
>>
>> The two main reasons would be that it is always good to keep the 
>> original files to minimize the risk of human errors and the other 
>> would be that long trajectories might result in very large files that 
>> are awkward to handle.
> 
> There already is a 'multiple file' type for the file options, where you 
> could do something like '-f 1.xtc 2.xtc 3.xtc' or '-f 1.xtc -o out.xvg 
> -f 2.xtc'. It is currently probably only used by trjcat and eneconv, but 
> it should be straightforward to use it in other tools. We'd need library 
> functions to handle multiple inputs as well, currently, IIRC, trjcat and 
> eneconv handle this internally.

I have made a start on C routines to implement multi-I/O ability for 
trajectory files. I have a working multiple-input-file 3.3.1 g_rms based 
on said routines, but it's a long way from being worth my time to port 
all such utilities - especially with version 4 due soon.

Mark