[Dune] accessing the subIndexes using ALUGRID

[Martin Nolte]

Hallo Nagaiah,

I did some first tests with your code (using GCC 4.4) on my current versions of 
ALUGrid, UG, and DUNE.

First of all, your test was way too fast (it took less than 1s on my machine), 
so I added a loop to slow down the code. I also added another loop for the 
different codimensions (out of curiosity).

My first results are as follows:

                    | codim 1 | codim 2 | codim 3
-------------------+---------+---------+---------
AlbertaGrid        |   24.6s |   22.9s |   21.5s
ALUSimplexGrid     |   25.4s |   33.7s |   28.2s
UGGrid (simplices) |   18.8s |   36.2s |   12.3s
-------------------+---------+---------+---------
ALUCubeGrid        |    6.2s |    8.7s |    8.6s
UGGrid (cubes)     |    5.2s |   10.5s |    4.1s
-------------------+---------+---------+---------
SGrid              |    4.3s |    9.1s |    4.7s
YaspGrid           |    0.9s |    1.9s |    0.9s

Please note that CPU times are still very small, so that measurements may vary 
noticably between subsequent runs.

These numbers indicate that element iteration and obtaining indices for all 
subentities is faster on UGGrid than on ALUGrid. Only in the case of vertices, 
however, this effect seems to be relevant. Moreover, you should note that we are 
only testing the macro grid, which might also have an impact on performance numbers.

Summarizing the results, I would say that, except for SGrid, all grids are 
reasonable fast in this test and I don't see urgent need to change anything. 
After some further improvements, however, I could imagine this test to be useful 
for maintaining (or improving) the current performance of grid iterators and 
index access.

Best,

Martin


On 03/02/2012 03:12 PM, Nagaiah Chamakuri wrote:
> Dear Martin Nolte,
>
> I hope you remembered me (we met during the first DUNE user meeting).
>
> I wrote a issue about accessing the subIndexes using the ALUGRID and I wrote
> to DUNE mailing list but till now I did not hear about the solution for this issue.
> (http://lists.dune-project.org/pipermail/dune/2011-December/009888.html )
>
> Dr. Andreas Dedner responded to this email, but due to missing installation of UG
> he did not test the example code which I attached to that email.
>
> The issue is only trigger for 3D elements.
> (basically accessing the vertex index while using the element iterator)
> If you dont mind could test that simple example code.
> Please let me know if you find the similar results what I mentioned in that mail.
>
> Thank you very much in advance.
>
> Best Regards
> Nagaiah
>
> --
> Dr. Chamakuri Nagaiah
> Institute of Mathematics and Scientific Computing
> Heinrichstrasse-36, University of Graz,
> A-8010, Graz, Austria. Ph(office) : (0043) 316 380 5063
>

-- 
Dr. Martin Nolte <nolte at mathematik.uni-freiburg.de>

Universität Freiburg                                   phone: +49-761-203-5630
Abteilung für angewandte Mathematik                    fax:   +49-761-203-5632
Hermann-Herder-Straße 10
79104 Freiburg, Germany
-------------- next part --------------
A non-text attachment was scrubbed...
Name: attachment2.cc
Type: text/x-c++src
Size: 2629 bytes
Desc: not available
URL: <https://lists.dune-project.org/pipermail/dune/attachments/20120305/4e952da8/attachment.cc>
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: unitcube3.dgf
URL: <https://lists.dune-project.org/pipermail/dune/attachments/20120305/4e952da8/attachment.ksh>