Added Documentation for Parallel execution with Numba and OpenMP (#1481)

* removed openmp requirements from the installation docs * added section for parallel execution * added parallelisation section to toctree * added ref to parallelisation section in using formal integral section * empty commit * made suggested changes
tardis-sn · Mar 19, 2021 · 47ed0ce · 47ed0ce
1 parent cf8900f
commit 47ed0ce
Show file tree

Hide file tree

Showing 4 changed files with 40 additions and 26 deletions.
diff --git a/docs/index.rst b/docs/index.rst
@@ -60,6 +60,7 @@ Mission Statement
     using/components/index
     using/visualization/index
     using/interaction/index
+    using/parallelisation/index
 
 
 .. toctree::

diff --git a/docs/installation.rst b/docs/installation.rst
@@ -78,31 +78,6 @@ then install TARDIS, as follows::
     python setup.py install
 
 
-.. _install_openmp:
-
-Enabling parallel execution with OpenMP
----------------------------------------
-
-
-Manually, cloning the repository enables other options such as running the code in parallel (enabling OpenMP).
-In general, we encourage downloading the compilers from `conda` as we then can ensure that they work with TARDIS.
-Within the TARDIS conda environment do::
-
-    conda install -c conda-forge compilers
-
-For macOS::
-
-    conda install -c conda-forge llvm-openmp
-
-For Linux::
-
-    conda install -c conda-forge openmp
-
-To compile TARDIS for parallel execution::
-
-    python setup.py install --with-openmp
-
-
 .. _troubleshooting_inst_label:
 
 Installation Troubles (FAQ)

diff --git a/docs/using/interaction/using_formal_integral.rst b/docs/using/interaction/using_formal_integral.rst
@@ -23,4 +23,6 @@ Note that the integrated spectrum (``simulation.runner.spectrum_integrated``)
 is a lazy property so it will be only generated (and then cached) once it is
 accessed. The spectrum integration routine is Open-MP parallelized, so this
 process may be significantly sped-up if TARDIS is built with the
-``--with-openmp`` option and more then one thread is used.
+``--with-openmp`` option and more then one thread is used. More instructions on
+how to enable parallelisation of the code is given in the :ref:`parallelisation`
+section.
diff --git a/docs/using/parallelisation/index.rst b/docs/using/parallelisation/index.rst
@@ -0,0 +1,36 @@
+.. _parallelisation:
+
+*****************************************
+Parallel execution with Numba and OpenMP
+*****************************************
+
+Enabling parallel execution with OpenMP
+========================================
+
+Manually, cloning the repository enables other options such as running the code in parallel (enabling OpenMP).
+In general, we encourage downloading the compilers from `conda` as we then can ensure that they work with TARDIS.
+Within the TARDIS conda environment do::
+
+    conda install -c conda-forge compilers
+
+For macOS::
+
+    conda install -c conda-forge llvm-openmp
+
+For Linux::
+
+    conda install -c conda-forge openmp
+
+To compile TARDIS for parallel execution::
+
+    python setup.py install --with-openmp
+
+
+
+Numba Usage Guide
+===========
+
+The ``montecarlo`` section of the Configuration file accepts the parameter ``nthreads`` which sets the number of
+threads to be used for parallelisation. Setting the value of the parameter between 1 and the environment variable
+``NUMBA_NUM_THREADS`` (which is, by default, the number of CPU cores on your system) will automatically invoke Numba
+to parallelise the code. (See :ref:`config-file` section).