Using sum_hills and multiple walkers

[carolynl...@gmail.com]

Hi,

I am guess that if you are using multiple walkers to generate multiple HILL files, then using stride to look for convergence will not work, because sum_hills is likely just concatenating the five files and therefore history of when hills are being added is not being correctly interpreted.  True?

e.g.

plumed sum_hills --hills HILLS.0,HILLS.1,HILLS.2,HILLS.3,HILLS.4 --stride=500 --mintozero

Will not produce an accurate depiction of how the system is converging.

[carolynl...@gmail.com]

I believe I have found a solution to my problem.  

I notice that HILLS.* files generated by using walkers include the system clock time as the final field.  (I am sure this is necessary so that when a walker reads a file generated by another walker, it knows which hills it has already and which it needs to update its grid.)

Thus, with a small amount of scripting, I can create a single concatenated HILLS file that is sorted by that last field.  Assuming that all the walkers are reading each other's files at a reasonable rate, this represents well how the collective FES is being constructed and this can be used with "stride" to look for convergence.

-Carolyn