Benchmark problems

The graphical user interface of Elmer
Post Reply
asandip
Posts: 45
Joined: 09 May 2019, 22:06
Antispam: Yes

Benchmark problems

Post by asandip » 29 May 2019, 01:05

Hello,

Are the case studies listed in Elmer_GUI_Tutorials.pdf benchmark problems?

- Anjali

mzenker
Posts: 1955
Joined: 07 Dec 2009, 11:49
Location: Germany

Re: Benchmark problems

Post by mzenker » 29 May 2019, 09:50

Hi,

what do you mean by "benchmark problem"?

Matthias

kevinarden
Posts: 253
Joined: 25 Jan 2019, 01:28
Antispam: Yes

Re: Benchmark problems

Post by kevinarden » 29 May 2019, 12:15

A "benchmark problem" can be a comparison of a result to a known or accepted answer. A test to see if the correct solution is obtained. It can also be a comparison of performance to another code, such as solution time.

mzenker
Posts: 1955
Joined: 07 Dec 2009, 11:49
Location: Germany

Re: Benchmark problems

Post by mzenker » 29 May 2019, 13:57

Thanks for clarifying.
The definitive answer would have to come from the Elmer team, but I would assume that the tutorials cases have been checked to give correct results, and I don't think they are especially set up to demonstrate the speed of ElmerSolver.

Matthias

asandip
Posts: 45
Joined: 09 May 2019, 22:06
Antispam: Yes

Re: Benchmark problems

Post by asandip » 29 May 2019, 17:54

Thank you for your response.

Hoping the Elmer team responds to this question.

kevinarden
Posts: 253
Joined: 25 Jan 2019, 01:28
Antispam: Yes

Re: Benchmark problems

Post by kevinarden » 29 May 2019, 21:13

Can't speak for the team, but I believe the "benchmarks", tests, are located in the depository at

https://github.com/ElmerCSC/elmerfem/tr ... /fem/tests

The tutorials are just intended to learn, although they may also be numerically correct.

raback
Site Admin
Posts: 3463
Joined: 22 Aug 2009, 11:57
Antispam: Yes
Location: Espoo, Finland
Contact:

Re: Benchmark problems

Post by raback » 30 May 2019, 18:53

Hi

The ElmerGUI tutorials are intended to be educational simple cases that can be defined in rather short time by first time users of Elmer. They have been part of the material that we use on the courses.

Usually tutorials are not benchmark problems that would verify performance or correctness of code. Correctness has been usually checked for in earlier phases and making a graphical tutorial is often not related to that. There are some cases that have reference results from literature, such as bending of the beam or capacitance of two balls. Some are so coarse (e.g. the Rayleigh-Benard instability case) that having the same solution on some other software is quite improbable.

The problem with tutorials is that there is no automated procedure to run them when new versions are created. So rarely it happens that a tutorial case gets broken.

The consistency test, on the other hand, are run every time there is a new commit to code. Deverlopers do it themselves but also Travis does this automatically as can be seen in:
https://github.com/ElmerCSC/elmerfem/branches

Among the consistency tests there are also benchmarks for performance and verification but that's a minority of the ~600 tests.

As there are much more tests than tutorials it is hoped that after the user gets a feeling how the code operates he/she can pick needed functionality directly from the tests.

-Peter

asandip
Posts: 45
Joined: 09 May 2019, 22:06
Antispam: Yes

Re: Benchmark problems

Post by asandip » 31 May 2019, 00:46

Thank you, Peter! That answers my question.

Post Reply