Merge pull request #799 from baagaard-usgs/docs-run-pylith-offline

Add instructions for running the PyLith binary offline

Author: baagaard-usgs

CHANGES.md

@@ -16,6 +16,7 @@ The version numbers are in the form `MAJOR.MINOR.PATCH`, where major releases in
   the other output files.
   * Consistency check for material properties and scales used in nondimensionalization (currently just the shear modulus).
   * Add section to User Guide on troubleshooting solver issues.
+  * Add section to User Guide on how to run PyLith binary offline.
   * Allow output on a finer resolution mesh than used in the simulation to facilitate accurate representation of fields with a basis order of 2 or greater.
 * **Fixed**
   * Fix inconsistency in normal direction on fault surfaces. Orientation was correct but direction was flipped at some locations. This affected local slip direction and the resulting deformation close to the fault. This bug fix was not in version 4.1.3.

docs/user/run-pylith/overview.md

@@ -19,6 +19,13 @@ The amount of speedup you will see by running in parallel depends on the process
 On most desktops and laptops, you can obtain reduced runtimes using about one-quarter to one-half of the number of physical cores.
 Once you saturate the memory bus, using additional processes (cores) will result in little, if any, speedup.
 
+### Running on a Offline Desktop or Laptop
+
+The MPI implementation (MPICH) used in the PyLith binary by default uses the network interface.
+If your desktop or laptop is offline, then you need to tell MPI to use sockets rather than the network interface.
+You can do this by setting the `IF_PROVIDER` environment variable.
+For example, in a terminal running zsh or tcsh, you can use `export IF_PROVIDER=sockets`.
+
 (sec-run-pylith-cluster)=
 ### Running in Parallel on a Cluster