lp:ffdm
- Get this branch:
- bzr branch lp:ffdm
Branch merges
Recent revisions
- 11. By Miro
-
I added function space for scalar functions and also saving routine -
this needs further optimazation; I will try design these things later
with inheritance in mind. Later means when I see a correct solution to
laplace u = f with DirichletBC. - 10. By Miro
-
in derivatives i started using thisArea criterion again, it works nicely
on hexagon, I have to test square with 'center' node sitting on the edge - 9. By Miro
-
in one_sided_
derivative I fixed problem of derivative in "impossible"
directions. Still plenty of work on central derivatives and also
different versions for these functions based on the input data. I am
going to take a brake from this problem and focus on saving data or
marking the nodes as boundaries for BC. - 8. By Miro
-
central diff now working (so so). There is one interesting issue; when I
make radius smaller in main, function in main gets the y-derivative of
(x-1)**2 @ [0,0] correct no matter how small the radius is, whereas the
templated one_sided_derivative functions starts to diverge at
radius=1E-12. All I was able to figure out is that values
function<double>() and what() differ; 0 vs. 1E-17. But I can't figure
why that is. So for now, element size smaller than 1e-11 is at your own
risk. - 5. By Miro
-
I started to play with derivatives (far from done), also I added Vector
class just for fun. - 4. By Miro
-
improved performance of adding neighbors( I can get through 1e6
triangles in less than 10s). next up are the classes for evaluation of
functions and fields on nodes, and derivatives
Branch metadata
- Branch format:
- Branch format 7
- Repository format:
- Bazaar repository format 2a (needs bzr 1.16 or later)