[Opm] Typical Run-time for SPE10

[Alf Birger Rustad]

Hi Kang,

Actually you do not need to do any code changes for the solvers that are supported. They can be chosen at run-time, with command-line parameters or a parameter file. I cannot remember this being documented on the internet yet though, but others may chime in with pointers on that. For the two-phase simulator, you can see its introduction here:
https://github.com/OPM/opm-simulators/pull/1235#issuecomment-311350607
Reading down the thread you will notice that there is a significant speed-up setting linear_solver_use_amg=true which will enable usage of Dune's AMG routine. The default preconditioner is incomplete LU factorization. Please note that the two-phase simulator is found in a separate binary, named flow_ebos_2p. If you adapt SPE10 Model2 to run with the two-phase simulator, we will be happy to accept the changes to the deck in opm-data.

Cheers,
Alf

________________________________
Fra: Kangqiao Hu <kangqiaohu at gmail.com>
Sendt: 21. august 2017 06:39
Til: Alf Birger Rustad
Kopi: Opm at opm-project.org
Emne: Re: [Opm] Typical Run-time for SPE10

Thanks Alf for the explanation. If I want to replace the current solver with a custom one, where is a good place to start with? Is NewtonIterationBlackoilInterleaved.cpp where the solvers are actually defined?

Regards
Kang

On Mon, Aug 14, 2017 at 4:18 AM, Alf Birger Rustad <abir at statoil.com<mailto:abir at statoil.com>> wrote:

Hi Kang,


The short answer is that flow does not support running the model yet. The long answer is that SPE10 Model 2 is a two-phase upscaling study, and hence by nature not fond of the preconditioning routines used by default in flow (or some of the commercial simulators for that matter). The most obvious remedy to this is to exploit the fact that SPE10 Model 2 is a two-phase case, and to use algebraic multigrid based preconditioning. We just very recently landed a true two-phase simulator version of flow, which actually supports an AMG based linear solver. Adapting the deck to this reality may finally make SPE10 Model 2 run well with flow.


Cheers,

Alf

________________________________
Fra: Opm <opm-bounces at opm-project.org<mailto:opm-bounces at opm-project.org>> på vegne av Kangqiao Hu <kangqiaohu at gmail.com<mailto:kangqiaohu at gmail.com>>
Sendt: 14. august 2017 05:13
Til: Opm at opm-project.org<mailto:Opm at opm-project.org>
Emne: [Opm] Typical Run-time for SPE10

Hi everyone,

I am running spe10 model2 using flow with single thread on a Intel i7-7820hk mobile processor, and the simulation has been running for more than 4 days and it seems to be far from complete. I was wondering what is the typical time of running this test case? Thanks!

Regards
Kang


-------------------------------------------------------------------
The information contained in this message may be CONFIDENTIAL and is
intended for the addressee only. Any unauthorized use, dissemination of the
information or copying of this message is prohibited. If you are not the
addressee, please notify the sender immediately by return e-mail and delete
this message.
Thank you



-------------------------------------------------------------------
The information contained in this message may be CONFIDENTIAL and is
intended for the addressee only. Any unauthorized use, dissemination of the
information or copying of this message is prohibited. If you are not the
addressee, please notify the sender immediately by return e-mail and delete
this message.
Thank you
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <//opm-project.org/pipermail/opm/attachments/20170821/18a64fb4/attachment.html>