.. _running:

###########
Running CMC
###########

Here we describe how to actually run CMC, assuming you've generated the initial conditions using COSMIC


===========
Running CMC
===========

With our initial conditions in place, the hard part is actually over!  Assuming 
the ``cmc`` executable is the directory you want to run from, we need only to 
call it with ``mpirun`` (or ``mpiexec``, depending on your mpi version):

.. code-block:: bash 

        mpirun -np <n_cores> ./cmc <params.ini> <output>

Where

 * <n_cores> is the number of cores you want to run on
 * <params.ini> is the path to your .ini file (examples `here <https://github.com/ClusterMonteCarlo/CMC-COSMIC/tree/master/examples>`_ or in the :ref:`inifile` page) 
 * <output> is the prefix for all the output files that will be produced by CMC

==============
Restarting CMC
==============

CMC allows for bit-for-bit restarting, by saving the state of each parallel 
process every few hours (2 by default).  These restart files will be saved in 
your run folder in a subfolder titled ``<output>-RESTART``.  These restart 
files will be numbered sequentially starting from 1, so you'll have to specifiy 
which restart 
file you want to use.  A typical restart file will have the form 
``<output>.restart.<n>-<proc>.bin``, where ``<n>`` is the number of the 
restart, and ``<proc>`` is the number of the mpi processor that file was 
produced by.

To restart CMC from one of these, you'll need the original .ini file and the 
path to the restart folder. 

.. code-block:: bash

        mpirun -np <n_cores> ./cmc -R <n> <params.ini> <new_output> <output>

where <new_output> is the new prefix for all the files that are restarted.  If 
for some reason you need to start from a restart file but with a `different` 
random seed, you can specifiy that on the command line with ``-n <new_seed>``

.. note::

        CMC cannot restart on different numbers of cores than the original run was performed on 

============================================
Example: Run Plummer Sphere to Core Collapse
============================================

Now we have all the tools we need to run CMC!  Let's try running the above Plummer sphere with :math:`N=10^4` particles to core collapse.

We've already generated the initial conditions above and saved them as 
``plummer.hdf5``, so all we need is to move them into a folder with the ``cmc`` 
executable and an appropriate ini file.  We can use the ``PlummerSphere.ini`` 
file located in the `CMC-COSMIC/examples 
<https://github.com/ClusterMonteCarlo/CMC-COSMIC/tree/master/examples>`_ 
folder.  

With that in place, we can run CMC on four cores with

.. code-block:: bash

        mpirun -np 4 ./cmc PlummerSphere.ini plummer

On the Vera cluster at CMU, this takes about 2 minutes to run to core collapse.

We can check that it's worked by looking at the Lagrange Radii (the <x> 
Lagrange radius is the radius enclosing <x> percent of the cluter mass).  If we 
plot the 0.1% 1%, 10% 50%, and 90% Lagrange radii, the classic collapse of the 
core (and expansion of the halo) appears:

.. ipython:: python
        
        import pandas as pd
        lag_rad = pd.read_table('source/example_output/plummer.lagrad.dat',header=1,delimiter=' ',index_col=False)
        plt.plot(lag_rad['#1:t'],lag_rad[['7:r(0.001)','12:r(0.01)','17:r(0.1)','21:r(0.5)','25:r(0.9)']],color='C1');
        plt.yscale('log');
        plt.xlabel("Time (relaxation time)",fontsize=15);
        @savefig plot_lagrad.png width=6in
        plt.ylabel("Virial Radii",fontsize=15);