next up previous
Next: CLUSTEREASY Variables Up: CLUSTEREASY: The Parallel Computing Previous: Modifying Model Files for


Customizing CLUSTEREASY

LATTICEEASY is written to be easily customizable. The most common ways to insert your own code into a particular model are through the model_initialize() and model_output() functions. This section lists the most important new features of CLUSTEREASY that you will need to know to write such functions.

First, however, we can give one general warning about output in MPI. It is generally not reliable to have more than one processor writing data to the same file. Even if you are careful to ensure that the processors write the data in the correct order, file caching in the operating system can still cause unexpected results. We found that the only reliable method of writing to a file was to copy the output data to the root processor and do all the writing from there. (You could also have each processor write to a separate file, but we wanted to produce output whose format didn't depend on the number of processors.)



Subsections


Go to The LATTICEEASY Home Page
Go to Gary Felder's Home Page
Send email to Gary Felder at gfelder@email.smith.edu
Send email to Igor Tkachev at Igor.Tkachev@cern.ch

This documentation was generated on 2008-01-21