Hello everyone,
In order to speed up the SQS generation for 64 atoms, I would like to use more than 1 core to run gensqs. Is there any possibility to perform parallel calculations on multiple cores?
I have found the possibility to use these two parameters, but I don’t know how to use them properly:
-tp=[int] Total number of processes (for parallel operation)
-ip=[int] Index of current process (0,...,tp-1) (for parallel operation)
Many thanks for your help
Stefan
You may want to give mcsqs a try (it is more efficient).
Both commands have the -ip option, which you can use to give each running process a different id number. With gensqs you also need to give the total number of processes. Each running copy works independently, so you can run them on different processors (MPI is not needed). The ip just ensures that the file outputs don’t overwrite each other.
Since the search is via a stochastic algorithm, each core explores at random, independently.
To find the best solution so far, just run
mcsqs -best
If mcsqs doesn’t know the total number of cores running, by specifying -tp to every process, how does allocate the portion of search space to each process?
How should one gather the output then, since there are so many files. Will the best match appear in one of the files?
Hi
when i type mcsqs -best , it returns
Monte Carlo generator of Special Quasirandom Structures 3.04, by Axel van de Walle
-n=[int] nb of atom/unit cell
-l=[string] Input file defining the random structure (default: rndstr.in)
…
How do I solve this problem?
thank you
mcsqs -best works on the beta version of the code, try downloading and compiling that.
OK,I see,Thanks.