[aspect-devel] [CIG-MC] Progress in writing the mantle convection code Aspect

Timo Heister heister at clemson.edu
Tue Oct 1 12:03:49 PDT 2013


Hey,

I think it is a very good idea to communicate better what we are all working on!

> [...]
> These are the two big projects currently on my plate.

I would like to add what I am working on related to Aspect right now:
1/ Performance improvements especially in 3d (temperature assembly is
way too slow and I have several things to test here)
2/ 3d examples. We currently have no good examples for 3d setups that work well.

A bit further down the line:
3/ Improvements to compositional fields if they are to represent a
level set (as in 0=no, 1=yes and not a continuous
range/concentration).
4/ optionally use PETSc instead of Trilinos for linear algebra (a lot
of work happened inside deal.II already, but there is stuff left to
do). This allows us to analyze performance bottlenecks by comparing
PETSc and Trilinos.
5/ 3d scalability tests and optimizations.

-- 
Timo Heister
http://www.math.clemson.edu/~heister/


More information about the Aspect-devel mailing list