Opened 11 years ago
Closed 11 years ago
#2330 closed discussion (invalid)
Avoid duplicate code for tearing
| Reported by: | Christian Schubert | Owned by: | Patrick Täuber |
|---|---|---|---|
| Priority: | high | Milestone: | 1.9.0 |
| Component: | Backend | Version: | trunk |
| Keywords: | Cc: | Lennart Ochel, Jens Frenkel, Volker Waurich |
Description (last modified by )
Hi,
I've seen that some tearing algorithms now appear multiple times (Changeset r16937).
That might not be the best solution. We should either fix the old ones, or if that is too complicated delete them, before reimplementation.
I also added Volkers diploma thesis which he presented at the OM Developers Meeting in February 2013. There he reviewed many classical tearing algorithms. The current implementation in omc is an improved version of the carpanzano algorithm.
As far as I know, the problem is not the tearing algorithm itself but its integration (no analytical jacobian, unknown bugs as found in ticket #2324, ...). I think concentrating on these issues would be more helpful to the OpenModelica Compiler.
If you have any questions on tearing don't hestitate to ask Volker.
Attachments (1)
Change History (3)
by , 11 years ago
| Attachment: | DA_Waurich.pdf added |
|---|
comment:1 by , 11 years ago
| Description: | modified (diff) |
|---|
comment:2 by , 11 years ago
| Resolution: | → invalid |
|---|---|
| Status: | new → closed |
There is no duplicated code in the tearing section. The method cellier2 is just a modified heuristic of the cellier approach without duplicating the algorithm itself.
Instead there is a lot of dead and unfinished code that we are going to finalize.
diploma thesis of Volker Waurich