FEMT: Improvement of Distribution inputs

Introduction

In FeMT 2022.3, to take into account the new management of the distribution via the Flux Supervisor, 3 Input Parameters are displayed:
  • Distribution (local) : Yes or No

    (Yes by default)

  • No. Concurrent Flux
    • read only, depending of the configuration of Options Supervisor
  • Multithreading No. Cores
    • if local distribution activated: read only, depending of the configuration of Options Supervisor
    • else the user can choose the number of cores.

The two last parameters are in "read only", and these displayed values depend to the configuration of the Options Supervisor.

Input parameter: Distribution (local)

  • If the Distribution (local) is set to Yes


    • If the distribution is activated in the Flux Supervisor that is to say that it has already been configured (Number of Concurrent Flux and Number of cores per concurrent Flux) and that you want to use distributed computing, then the solving will be distributed.
    • If the distribution is not activated in the Flux Supervisor then the solving will be sequential and not distributed even if the Distribution (local) is set to Yes.
    Note:
    • In order to activate the distribution in the Flux Supervisor:
      • either use the Distribution Manager button
      • or use the Options button and in System / Parallel computing, click on the Set local resources button

      and possibly click on the Allow button and then define the local ressources (Number of Concurrent Flux and Number of cores per concurrent Flux) and click on the Use button

    • In order to deactivate the distribution in the Flux Supervisor:
      • either use the Distribution Manager button
      • or use the Options button and in System / Parallel computing, click on the Set local resources button

        and possibly click on the Do Not Allow button or click on the Stop Using button

  • If the Distribution (local) is set to No then the solving will be sequential, not distributed.


For more information on distribution configuration, please see Parametric distribution in Flux.

Note: The computing distribution is not implemented in Skew.

Input parameter: No. Concurrent Flux



This field defines the number of Flux that can be running at the same time during a distributed computation.

This number of concurrent Flux is not accessible to the user here, it is grayed out.

  • If the Distribution (local) is set to Yes
    • If in the field No. Concurrent Flux there is a value n, then this indicates that the distribution has already been configured i.e. that this value has been previously set in the Flux Supervisor. So, for the solving, n Flux will be run in parallel.
      Warning: A great number may render difficult the use of the machine for other tasks until the solving completes.
    • If in the field No. Concurrent Flux there is no value, then this indicates that the distribution has not been activated in the Flux Supervisor. And so the solving will be sequential and not distributed even if the Distribution (local) is set to Yes.
    Note:
    • In order to configure the distribution in the Flux Supervisor and define the number of concurrent Flux:
      • either use the Distribution Manager button
      • or use the Options button and in System / Parallel computing, click on the Set local resources button

      and possibly click on the Allow button and then define the Number of Concurrent Flux and click on the Use button

    • In order to deactivate the distribution in the Flux Supervisor:
      • either use the Distribution Manager button
      • or use the Options button and in System / Parallel computing, click on the Set local resources button

      and possibly click on the Do Not Allow button or click on the Stop Using button

  • If the Distribution (local) is set to No then this field is not to be filled in, it will not be used as the solving will be sequential and not distributed.

Input parameter: Multithreading (No. Cores)

This field defines the number of cores used during a solving. More precisely, it is the number of cores used by the application for multithreaded algorithms. The application uses at least 1 core.

  • If the Distribution (local) is set to Yes
    • If in the field Multithreading (No. Cores per Flux) there is a value m, (note: this value is not accessible to the user here, it is grayed out), then this indicates that the distribution has already been configured i.e. that this value has been previously set in the Flux Supervisor. So, for the solving, m cores per concurrent Flux will be used.


    • If in the field Multithreading (No. Cores per Flux) there is no value, then this indicates that the distribution has not been activated in the Flux Supervisor. And so the solving will be sequential and not distributed even if the Distribution (local) is set to Yes.
    Note:
    • In order to configure the distribution in the Flux Supervisor and define the number of cores per concurrent Flux:
      • either use the Distribution Manager button
      • or use the Options button and in System / Parallel computing, click on the Set local resources button

      and possibly click on the Allow button and then define the Number of cores per concurrent Flux and click on the Use button

    • In order to deactivate the distribution in the Flux Supervisor:
      • either use the Distribution Manager button
      • or use the Options button and in System / Parallel computing, click on the Set local resources button

      and possibly click on the Do Not Allow button or click on the Stop Using button

  • If the Distribution (local) is set to No then the solving will be sequential and not distributed.

    However, even if the solving is sequential, the number of cores will be used for multithreaded algorithms.



    Default values are set for the field Multithreading (No. Cores):
    • For a 2D project, the number of cores is 1 by default
    • For a Skew project, the number of cores is 2 by default
    • For a 3D project, the number of cores is 4 by default

    This value corresponding to the number of cores can be modified by the user. The maximum value the user can enter is of course the total number of cores of the machine.

Warning: Using all cores might induce a heavy load on the machine (volume mesh, part of solving process, inductance computation).