Any link to tests for strong or weak type of scaling for MPI performance of FEniCS? Е.g. deal.ii claims to scale at least up to 16k of processes and provide measurments for both type of scaling for 4k in tutorial http://dealii.org/developer/doxygen/deal.II/step_40.html#Results
Table update with links http://htmlpreview.github.io/?https://github.com/kostyfisik/FEA-compare/blob/master/table.html пт, 16 окт. 2015 г. в 12:55, Anders Logg <[email protected]>: > The MPI entry for FEniCS should be changed from "Yes?" to "Yes". (Or maybe > even "Yes!" :-) > -- > Anders > > > fre 16 okt. 2015 kl 11:26 skrev Charles Cook <[email protected]>: > >> Hello Martin, >> >> >> >> Could you elaborate? >> >> >> >> > Any order in principle, but this is not quite completed yet (working >> on it!). >> >> >> >> >> >> Thank you, >> >> >> >> Charles Cook, Ph.D. >> >> >> >> *From:* [email protected] [mailto: >> [email protected]] *On Behalf Of *Martin Sandve Alnæs >> *Sent:* Thursday, October 15, 2015 11:40 AM >> *To:* Konstantin Ladutenko <[email protected]> >> *Cc:* Anders Logg <[email protected]>; [email protected] >> *Subject:* Re: [FEniCS] FEA Software Comparison >> >> >> >> >> >> On 15 October 2015 at 17:06, Konstantin Ladutenko < >> [email protected]> wrote: >> > Dear Anders, >> >> >> >> Thank you for you contibution! I have updated the table it is still >> available at GitHub. >> >> >> >> Thre is now also an export to html, it looks to be quite nice >> http://htmlpreview.github.io/?https://github.com/kostyfisik/FEA-compare/blob/master/table.html >> >> >> >> (sorry, I messed a header a little bit and was not able to find a >> solution fast) >> >> >> >> Best regards, >> >> Kostya >> >> >> >> ср, 14 окт. 2015 г. в 17:02, Anders Logg <[email protected]>: >> >> Looks like a useful comparison. Here are some suggestions: >> >> ... >> > Elements: Add quads and hexes since this has been added recently (Chris >> and Martin can comment) >> >> >> > Quads and hexes are not fully functional yet. >> >> >> >> >> >> Mapping: Any order or only isoparametric? (Martin can comment) >> >> >> >> Any order in principle, but this is not quite completed yet (working on >> it!). >> >> >> >> Martin >> >
_______________________________________________ fenics mailing list [email protected] http://fenicsproject.org/mailman/listinfo/fenics
