Version 4 (modified by ole, 15 years ago) (diff) |
---|
ANUGA FAQ
Miscellaneous Questions
This part of the FAQ is for questions that have no obvious place elsewhere. This is also a good place to start a question that you would like answered.
How do I get a question answered?
The best way is to start a new question on this page. Just cut-and-paste this question/answer to somewhere below here and leave the answer part blank. Or perhaps you can discuss the question in more detail.
Why does ANUGA get so slow when meshes are refined
Two reasons:
- With smaller triangles timesteps tend to become smaller due to the CFL condition
- The computation time per timestep grows more than linearly with the number of triangles. Here is why
Let h be the grid size. The number of cells (triangles) N will normally be proportional to h^{-2} The size of the timesteps will be proportional to the grid size, and so the number of timesteps is proportional to h^{-1} = N^{1/2} So the computational cost (solution time) will be proportional to (number of cells) x (no of timesteps) = N x N^{1/2} = N^{3/2} log (Time) = 3/2 log( N). So increasing the grid size from 100 cells to 1000 cells grows the computational time by 10^{3/2} (about 30). == Can ANUGA run in parallel on multiple cpus == Yes - there is a module using MPI (via pypar) to allow ANUGA to run in Parallel. This is not part of the standard distribution on sourceforge but can be obtained directly from the ANUGA repository (https://datamining.anu.edu.au/svn/ga/anuga_core/source/anuga_parallel) Nils Goseberg has kindly put together some instructions which are available at
ANUGA FAQ
Attachments (1)
- Installation of ANUGA parallel.doc (41.0 KB) - added by ole 15 years ago.
Download all attachments as: .zip