Moved to http://lakeat.co.nf/
Enosh, a structured grid generator (V)
Posted March 5, 2014 at 17:14 by lakeat
I have made significant improvement on the mesh export. The manual building up of OpenFOAM mesh now turns up really fast. I got a feeling that it could possibly beat even OpenFOAM's native blockMesh generator. In my test of a 3,200,000 mesh (for flow around a square cylinder). The building process took 90.51s in total. Here are more details:
...... [1] Done assembling the points. ( 7.09000s)
...... [2] Done assembling all cells & intnl cntrs. ( 40.62000s)
...... [3] Done treatment of cntrs and init o&nbs. ( 27.74000s)
...... [4] Done assembling boundary cntrs and info. ( 9.77000s)
...... [5] Done assembling faces. ( 1.29000s)
...... [6] Done assembling owners on frontAndBack. ( 0.23000s)
...... [7] Done face orientation correction. ( 3.77000s)
This is a very exciting progress. As in the old days, I have to wait for 10 minutes for just a 24,000 grid sized mesh. So the speed now scales roughly around O(N), which is way better than O(NlogN) or O(n^2).
The renumber (e.g., using Cuthill–McKee algorithm) is needed, but it is not a big deal at this moment.
As for a smaller mesh, using the same blocks as before (8 blocks around the square), 240,000 mesh size, it took:
...... [1] Done assembling the points. ( 0.53000s)
...... [2] Done assembling all cells & intnl cntrs. ( 2.60000s)
...... [3] Done treatment of cntrs and init o&nbs. ( 1.81000s)
...... [4] Done assembling boundary cntrs and info. ( 0.91000s)
...... [5] Done assembling faces. ( 0.10000s)
...... [6] Done assembling owners on frontAndBack. ( 0.01000s)
...... [7] Done face orientation correction. ( 0.28000s)
...... [1] Done assembling the points. ( 7.09000s)
...... [2] Done assembling all cells & intnl cntrs. ( 40.62000s)
...... [3] Done treatment of cntrs and init o&nbs. ( 27.74000s)
...... [4] Done assembling boundary cntrs and info. ( 9.77000s)
...... [5] Done assembling faces. ( 1.29000s)
...... [6] Done assembling owners on frontAndBack. ( 0.23000s)
...... [7] Done face orientation correction. ( 3.77000s)
This is a very exciting progress. As in the old days, I have to wait for 10 minutes for just a 24,000 grid sized mesh. So the speed now scales roughly around O(N), which is way better than O(NlogN) or O(n^2).
The renumber (e.g., using Cuthill–McKee algorithm) is needed, but it is not a big deal at this moment.
As for a smaller mesh, using the same blocks as before (8 blocks around the square), 240,000 mesh size, it took:
...... [1] Done assembling the points. ( 0.53000s)
...... [2] Done assembling all cells & intnl cntrs. ( 2.60000s)
...... [3] Done treatment of cntrs and init o&nbs. ( 1.81000s)
...... [4] Done assembling boundary cntrs and info. ( 0.91000s)
...... [5] Done assembling faces. ( 0.10000s)
...... [6] Done assembling owners on frontAndBack. ( 0.01000s)
...... [7] Done face orientation correction. ( 0.28000s)
Total Comments 0
