Opened 7 years ago
Last modified 3 years ago
#4990 assigned enhancement
Should we make IDA the default ODE mode solver?
| Reported by: | Francesco Casella | Owned by: | Andreas Heuermann |
|---|---|---|---|
| Priority: | high | Milestone: | |
| Component: | Run-time | Version: | |
| Keywords: | Cc: | Martin Sjölund, Adrian Pop |
Description
The IDA solver has been implemented as an alternative to DASSL for quite some time. IDA is in fact DASSL's successor: it implements the same basic algorithm, but it is written natively in C and, contrary to DASSL, it is actively maintained as a part of the SUNDIALS suite, so it is in general expected to give better performance than the old DASSL. In principle, once we have IDA implemented I see no reason to keep DASSL as the default option. In fact, I have some test cases where DASSL fails and IDA works just fine.
Unfortunately, 99% of the users will just use the default solver and, in case of simulation issues, won't probably think of solving them by simply switching to IDA.
So, the question is: should we make ida the default solver for ode mode simulations, in place of dassl?
I guess in order to answer this question we need to run the whole testsuite with ida as a default solver and look for regressions and improvements.
Currently, we have Hudson tasks running the whole testsuite with ida and daeMode (https://libraries.openmodelica.org/branches/daemode/ and https://libraries.openmodelica.org/branches/newInst-daemode/), but I don't think we should make daeMode the default solver option.
@sjoelund.se, could you set up a task to be run on demand such as https://libraries.openmodelica.org/branches/daemode/, but only with -s=ida as default simflag and no daeMode?
I could then analyze the results, so we can take an informed decision. In fact, it could be that ida has few recessions, or possibly none, and a lot of improvements compared to dassl already. On the other hand, if there are issues left we could try to solve them, and/or to give suggestions to the end users when should they perfer IDA over DASSL.
Change History (6)
comment:1 by , 6 years ago
| Milestone: | 1.13.0 → 1.14.0 |
|---|
comment:2 by , 5 years ago
| Milestone: | 1.14.0 → 1.16.0 |
|---|
Releasing 1.14.0 which is stable and has many improvements w.r.t. 1.13.2. This issue is rescheduled to 1.16.0
comment:4 by , 4 years ago
| Owner: | changed from to |
|---|---|
| Status: | new → assigned |
comment:5 by , 4 years ago
| Milestone: | 1.17.0 → 1.18.0 |
|---|
Retargeted to 1.18.0 because of 1.17.0 timed release.
Rescheduled to 1.14.0 after 1.13.0 releasee