Personal tools
You are here: Home Projects LLNL Project Development Code Development Meeting Minutes -- 4/11/06
Document Actions

Code Development Meeting Minutes -- 4/11/06

by Dan Reynolds last modified 2006-05-19 08:51

Notes from meeting with (majority of) Code Development team. Discussion included regression testing, Enzo documentation/method paper, and some talk regarding compartmentalization.

1. Regression Testing
- James has begun a document discussing regression testing for Enzo.
- Regression testing will be based on testing of different categories, from simplest (compile-ability on various platforms) to most difficult (accuracy/stability/performance of large scale runs).
- The current instantiation of testing will not be fully automatic, but is based on Makefiles and scripts that need only be run.
- James will place document up on Plone when it is ready for dissemination within the group.
- Alexei suggested that a common testing infrastructure would be helpful, so that additional tests may be incorporated easily to the test suite.

Needs: Suggestions of things to test as well as input files and code test files should be sent to James for inclusion within the regression tests.


2. Enzo Documentation and/or Method Paper
- We discussed the outline presented by Dave Collins discussing the overall structure and needs of an Enzo method paper.
- Dave is currently working on section 2 of the Developer's guide (regarding setting up new initial/boundary conditions that are not currently implemented)
- Dave will post the updated outline to the Plone site when he is finished with edits, at which time we will distribute sections to the relevant experts to fill in equation/algorithm details.


3. Compartmentalization
- We did not discuss much of this, since Robert was unable to attend the meeting.
- In discussions on this topic, Dan brought up the desire to separate grid information from grid data, i.e. do not have Grid.h contain any data whatsoever, just grid descriptions. Then let individual data arrays be instantiations of an 'EnzoVector' class (for example) that derives from the Grid, so it would know all of it's grid structure, but so that radiation arrays are separated from magnetic field arrays, etc.
- Such 'compartmentalizations' would allow for much simpler development of new modules within Enzo; however, it will take significant planning and Enzo expertise to manage such a separation of functionality within Enzo.


Powered by Plone CMS, the Open Source Content Management System

This site conforms to the following standards: