Why does NeoPz Lib use rather Skyline storage than YSMP one?

[MrNGS]

Hi, this is a rich library! ¡May GOD bless those who take care of it! While going over many custom examples, i noticed the use of Skyline strategy storage everywhere --almost by default, buy why do you guys use it over the YSMP one? Is not the latter less expensive as far as the memory cost concerned? EDIT: like to know if there are reasons for this use of Skyline over Ysmp.. Another question where to find the shape funcions for FM* and FM*? Is there any similarity between FE Mini and FM? Thank you, Meddean

[MrNGS]

Hello. Quick question: Why this line in pzgeoelrefless.h file 147 for (i = 0;i < NNodes();i++) out << NodePtr(i)->Id() << " "; is giving me a warning? I am not sure but when I get rid off the / ->Id() / access, the warning went away... What is the right way to handle it?

In file TPZSBFemElementGroup.cpp, I think the Lapack/blass called functions should be put in guard, because they trigger error compilation when the Blas/Lapack lib are not activated. The same goes for the eigSolv line 268 globmatkeep.SolveEigenProblem(eigenvalues, eigenVectors); Also, the class TPZBuildSBFem.cpp raises concern about this method: 372 cmesh.ApproxSpace().SetAllCreateFunctionsSBFem(dim); 473 cmesh.ApproxSpace().SetAllCreateFunctionsSBFem(dim); 608 cmesh.ApproxSpace().SetAllCreateFunctionsSBFem(dim); I was just trying to build the library in serial, and noticed these warnings. I do really to learn the detail of that project called iRMS but too much issues are raised there! If someone could take a look at it and update it, it would be more than helpful. Thank you

[philippedevloo]

Thanks for the comments. In fact the SBFem related files should not be included in the library when Lapack or Intel MKL is not linked. They depend on eigenvalue computations in a hard way.

I dont understand the warning. Do you mean a compile time warning? Which compiler/OS?

IRMS is a reservoir modeling project developed by Omar Duran Triana. It is rather complex. I suggest you start by studying his PhD thesis and see the IRMS as an implementation of the algorithms described there

[philippedevloo]

We use skyline quite a bit because it is "the best" in house solver. The ysmp solver requires linking to MKL to use Pardiso to invert the systems. But Pardiso is at least 10x faster than our skyline solver...

[MrNGS]

Thank you so much @philippedevloo for taking time to those kind response. YES, I noticed Omar's work on the subject, and delved into his branch and got his documentation on the matter. I see even that his dissertation is available and I'm going to have it. Your library provides a lot of so important and interesting stuff about fem tools and techniques. For the warnings or typos, I am not sure maybe it is just because a novice on the subject. I have actually started with the project CursoPz but as I am rather interested in saturated/unsaturated flows, i checked irms and the creepstokes, the latter so far compiles without issues but irms did not. And as i am learning and studying, i'll be more clear about the issues i am encountering, and then I will ask the right questions and be more precise. Today at my univ i will be having C. Douglas et al. paper, that work also seems to me so pioneering. I've discovered the repository last weak and it seems my summer will be FULL. Again, many thanks indeed!
Meddean

[MrNGS]

For the OS, I am working with Ubuntu 18.04.2 LTS 3,7 Gio Intel® Core™ i5-2520M CPU @ 2.50GHz × 4 Intel® Sandybridge Mobile 64-bit 314,0 Go

[OmarDuran]

Dear Meddean,

Currently, I am going to solve all the issues of IRMS and recreate it an external project here (https://github.com/labmec). Thus, you can compile and use it. It is going to take me a week. I will notify you when it is done.

Kind regards,

On Mon, 10 Jun 2019 at 04:33, Meddean notifications@github.com wrote:

Thank you so much @philippedevloo https://github.com/philippedevloo for taking time to those kind response. YES, I noticed Omar's work on the subject, and delved into his branch and got his documentation on the matter. I see even that his dissertation is available and I'm going to have it. Your library provides a lot of so important and interesting stuff about fem tools and techniques. For the warnings or typos, I am not sure maybe it is just because a novice on the subject. I have actually started with the project CursoPz but as I am rather interested in saturated/unsaturated flows, i checked irms and the creepstokes, the latter so far compiles without issues but irms did not. And as i am learning and studying, i'll be more clear about the issues i am encountering, and then I will ask the right questions and be more precise. Today at my univ i will be having C. Douglas et al. paper, that work also seems to me so pioneering. I've discovered the repository last weak and it seems my summer will be FULL. Again, many thanks indeed! Meddean

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/labmec/neopz/issues/23?email_source=notifications&email_token=AC2GHI5KJDNIJALEOVYQVWLPZX7URA5CNFSM4HWICEBKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGODXJEYEQ#issuecomment-500321298, or mute the thread https://github.com/notifications/unsubscribe-auth/AC2GHI6YT43EFHPFKVIYK33PZX7URANCNFSM4HWICEBA .

--

Omar Y. Durán Triana https://www.researchgate.net/profile/Omar_Duran_Triana Visitor Researcher and Petroleum Engineer Phone +55 19 35211149 Mobile +55 19 994137841

Computational Mechanics Laboratory http://www.labmec.org.br/wiki/ The State University of Campinas Rua Josiah Willard Gibbs, 85, SP - Brazil

[MrNGS]

How kind and generous of you, dear Omar! I am grateful beyond words! Blessed be! All the best, Meddean