Stop and restart
================

An important feature when running a simulation is the ability to stop and restart it. This section explain how you can do it with **TRUST**.

Stop a Running Calculation
--------------------------

When using **TRUST**, your calculation will automatically stop if it has reached:

-  the end of the calculation time.

-  the maximal allowed cpu time.

-  the maximal number of iterations.

-  the threshold of convergence.

You may use the ``my_data_file.stop`` file, if you want to stop properly your running calculation (i.e. with all saves).

When the simulation is running, you can see the **0** value in that file. To stop it, put a **1** instead of the **0** and at the next iteration the calculation will stop properly. 

When you don’t change anything to that file, at the end of the calculation, you can see that it is writen **Finished correctly**.

Save
----

**TRUST** makes automatic backups during the calculation. The unknowns (velocity, temperature,...) are saved in:

-  one **.xyz** file, written:

  -  at the end of the calculation.

  -  but, user may disable it with the specific keyword **EcritureLectureSpecial 0** added just before the **Solve** keyword.

-  one (or several in case of parallel calculation) **.sauv** files, written:

  -  at the start of the calculation.

  -  at the end of the calculation.

  -  each 23 hours of CPU, to change it, uses **periode_sauvegarde_securite_en_heure** keyword (default value 23 hours).

  -  user may also specify a time physical period with **dt_sauv** keyword.

  -  periodically using **tcpumax** keyword for which calculation stops after the specified time (default value :math:`10^{30}`). Use it for calculation on CCRT/TGCC and CINES clusters for example.


.. note:: 

  By default, the name for the **.sauv** files is **filename_problemname.sauv** for sequential calculation, **filename_problemname_000n.sauv** for parallel calculation (one per process). The format of theses files is binary and the files are completed during successive saves.

You can change the behaviour using the following keywords just before the **solve** instruction:

.. code-block:: bash

  sauvegarde binaire|xyz filename .sauv|filename .xyz

with **xyz**: the **.xyz** file is written instead of the **.sauv** files.

.. note::

  You can use **sauvegarde_simple** instead of **sauvegarde** where the .sauv or .xyz file is deleted before saves, in order to keep disk space:

.. code-block:: bash

  sauvegarde_simple binaire|xyz filename .sauv|filename .xyz

For more information, go see the :doc:`../reference/index`.

Resume
------

To resume your calculation, you may:

-  change your initial time, the new initial time will be the real final calculation time of the previous calculation (see the .err file).

-  change your final calculation time to the new wanted value.

-  add the following block just before the **Solve** keyword:

  .. code-block:: bash

    reprise binaire|xyz filename .sauv|filename .xyz


  .. note ::

    Instead of **reprise** keyword, you can use **resume_last_time** where **tinit** is automatically set to the last time of saved files (but you may change **tmax**):

    .. code-block:: bash

      resume_last_time binaire|xyz filename .sauv|filename .xyz

You can resume your calculation:

-  from .sauv file(s) (one file per process): you can only resume the calculation with the **same number of equations** on **the same number of processes**.

-  or from a .xyz file: here you can resume your calculation by **changing the number of equations solved** and/or with a **different number of processes**.

.. note::

  You can run a calculation with initial condition read into a save file (.xyz or .sauv) from a previous calculation using **Champ_Fonc_reprise** or read a into a MED file with **Champ_Fonc_MED**.