CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   OpenFOAM Running, Solving & CFD (http://www.cfd-online.com/Forums/openfoam-solving/)
-   -   Block matrices vs segregated solvers (http://www.cfd-online.com/Forums/openfoam-solving/58215-block-matrices-vs-segregated-solvers.html)

fisher December 13, 2008 08:13

Hello, in these two threads a
 
Hello,
in these two threads a block matrix implementation was announced that whould allow implicit curl operators and alike:

http://www.cfd-online.com/OpenFOAM_D...es/1/1553.html
"The current matrix support cannot handle that but the new version (block matrix) will come in due course)." (Dec. 2005)

http://www.cfd-online.com/OpenFOAM_D...es/1/8121.html
"The guy is talking about a block solution of the compressible flow system ... This is coming with my new block matrix implementation." (June 2008)

Have block matrices already been implemented in the meantime? Is any work available for testing and to contribute?

Regards,
Olaf

hjasak December 13, 2008 12:31

This will (in part) be present
 
This will (in part) be presented at the International Conference on Mathematics, Computational Methods & Reactor Physics (M&C 2009) Saratoga Springs, New York, May 3-7, 2009.

Further publications are likely to follow. Do you have any special need/interest/way to contribute?

Hrv

fisher December 13, 2008 20:12

Hello Hrvoje, I'm mostly aski
 
Hello Hrvoje,
I'm mostly asking for curiosity, because I'd like to compare the block matrix performance with the current implementation for standard cfd tasks. Furthermore, I'm also getting interested in electrodynamics simulations which are currently impossible as far as I can see.

If any code is available for early tests, I'd like to give it a try. But I am also sufficiently busy to wait until May.

Thank you
Olaf

jploz June 11, 2009 11:53

Actual situation
 
Hello,

regarding the previous posts: how is the current situation on block matrices? Are any implementations already available in the development version?

In another post was stated: "but it will make you happy to know that some people are already running block-coupled solutions". Is something of that accessable for the public?

Thanks in advance for any hints.

Best regards.


luca June 12, 2009 05:16

yes we are working on coupled solver based on OpenFOAM. In details, we are developing coupled pressure-based solver. We have approached to the problem in a hybrid way: in fact at the moment we use external linear solver to solve the block matrix assembled using OpenFOAM explicit operators and topology. The new solver is under testing, there are still open issue that are need to be solved so we'll keep you updated on any further developments. For other informations you can contact us.

Luca

mediode June 11, 2010 05:45

publication
 
For everybody interested, the link to the pub above mentionned:

http://powerlab.fsb.hr/ped/kturbo/op...lockSolver.pdf

m*

alberto June 12, 2010 13:53

The block solver is a very interesting feature, but there has been no estimated time of availability for a long time. It has been years the block-solver has been under discussion as something supposed to come soon.

Trying to obtain the code for joint research purposes has not been possible either, at least for what concerns my experience. I would think it should be easier in open source projects, but maybe others were luckier and had more interesting projects (or funding) to convince the developers at Wikki :D

Really no pun intended, but it would be really useful to have at least a vague road-map of what we can expect coming "in the near future" (meaning 1 year, for example). It would help users to decide, and avoid confusion.

As usual, just my two cents.

Best,

arjun June 12, 2010 19:48

Quote:

Originally Posted by alberto (Post 262760)

Trying to obtain the code for joint research purposes has not been possible either, at least for what concerns my experience. I would think it should be easier in open source projects, but maybe others were luckier and had more interesting projects (or funding) to convince the developers at Wikki :D

edited to remove post.

alberto June 12, 2010 20:55

Quote:

Originally Posted by arjun (Post 262771)
if you only need amg solver then pyAMG is good choice. I think they support block solvers too.

I was talking about the block solver (aka coupled solver for equations, not zones) they're working on :)

arjun June 13, 2010 04:15

Quote:

Originally Posted by alberto (Post 262772)
I was talking about the block solver (aka coupled solver for equations, not zones) they're working on :)

sorry i mis-understood it as block solver where as you were talking of coupled algo.


All times are GMT -4. The time now is 23:58.