Any Comsol users switched to Deal.II to share their experience? Or using 
both Deal.II and Comsol on everyday tasks?

Some questions behind this: 
1) How I can motivate my colleagues to try Deal.II? 
2) What are possible arguments to invest time into Deal.II instead of 
buying one more Comsol license?

My thought on this (still I am not actually using any of them, so any 
comments are welcome):

1) License cost?

Comsol license is quite expensive. However, if you put the scientist salary 
on the board, it is not that straightforward.  With Comsol it is quite easy 
to become productive from day zero. To achieve the same technical level 
with Deal.II you will need few weeks or even months of everyday use... 
There are so many technical details under the hood hidden from Comsol user 
which you are forced to be aware of with Deal.II. And even after mastering 
them both it looks like using Comsol to create a new model (with new 
geometry and physics) is cheaper compared to deal.ii as soon as you already 
have a license. And this works even better when you scale up - for sure you 
need much less number of licenses than a stuff count in the lab..

2) Control. 

This is a real benefit from my point of view. You can put the C++ into 
version control system and get all the improvements with simple diff. 
Probably Comsol has some kind of "model review mode" to show all check 
boxes and list choices, however, I was not able to locate this feature fast.

3) Adaptive mesh with hanging nodes.

It seems to be a unique Deal.II feature to be able to do a local refinement 
or coarsening of the model. For sure it is very good, especially for 
large-scale simulations.

The side effect is that it is very natural to do hp-refinement. For time 
domain problems it is easy for Deal.II to follow with dense mesh alongside 
with places, where it is actually needed.

4) Large models for HPC

This is mostly related to engenering tasks. However, if you have a large 
problem it can be easier to solve it with Deal.II using HPC cluster.  Using 
100-1000 cores or more to get the result can speed up the progress a lot, 
and such kind of clusters are quite affordable our days, it is quite easy 
to get a HPC access for academic research. It looks like it is not feasible 
with Comsol (at list I was not able to get data on more than a 20x parallel 
speedup with 0.8 efficiency for Comsol 4.2  in the internet). 

5) Comsol is an integrated environment.

So it has the drawing, meshing, setting the model, evaluation, 
visualisation in the same window, which is rather convenient. Actually this 
an argument aginst Deal.II, where you should mostly rely on external tools 
for most of the tasks. It is not a crucial problem, however, it complicates 
it a bit, e.g. to set up a working environment or to become really 
productive...

Any comments on this? More killer features of Deal.ii are welcome!

Best regards,
Konstantin Ladutenko

-- 
The deal.II project is located at http://www.dealii.org/
For mailing list/forum options, see 
https://groups.google.com/d/forum/dealii?hl=en
--- 
You received this message because you are subscribed to the Google Groups 
"deal.II User Group" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to dealii+unsubscr...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

Reply via email to