QuantumATK Forum

QuantumATK => General Questions and Answers => Topic started by: Choongman Moon on February 16, 2017, 10:39

Title: Change in "number of processors per kpoint" doesn't seem to be applied.
Post by: Choongman Moon on February 16, 2017, 10:39

Dear all,

I'm a quite new to VNL-ATK and I'm trying to optimize the structure of graphene.
I changed "number of processor per kpoint" from default value to 4 (using "scriptor" or "editor") and checked log file after calculation.
But in the log file, I got the warning message like below,

+------------------------------------------------------------------------------+
| DiagonalizationSolver parallelization report.                                |
+------------------------------------------------------------------------------+
| Total number of processes: 4                                                 |
| Total number of k-points: 85                                                 |
| Processes per k-point: 1                                                     |
| Number of process groups: 4                                                  |
+------------------------------------------------------------------------------+
+------------------------------------------------------------------------------+
| WARNING: Sub-optimal distribution of processes over k-points.                |
| For optimal performance make sure the number of processes is                 |
| a multiple of the number of k-points times processes_per_kpoint.             |
| 3 process(es) have fewer k-points assigned than others                       |
| and will be partially idle.                                                  |
+------------------------------------------------------------------------------+
| WARNING: Sub-optimal process distribution for the DiagonalizationSolver.     |
| Number of processes per k-point has been adapted for better performance.     |
| For an optimal process distribution, use 4 MPI process(es),                  |
| and set processes_per_kpoint=1                                               |
+------------------------------------------------------------------------------+

Calculation is finished anyway, but the changed value of "number of processors per kpoint" is not reflected.
Is this a bug or am I making mistake?
Any help or comments would be greatly appreciated.

Title: Re: Change in "number of processors per kpoint" doesn't seem to be applied.
Post by: Ulrik G. Vej-Hansen on February 16, 2017, 10:52

I would say you are making a mistake. When you have 85 k-points and less than 86 MPI processes, it is much more efficient to parallelize over k-points. That is why the program actually overrides your choice in this case. You need a much bigger system for more than one process per k-point to make sense.

Title: Re: Change in "number of processors per kpoint" doesn't seem to be applied.
Post by: Choongman Moon on February 17, 2017, 04:42

Thank you for your reply
It really helps.