[bastil]

HI 1.6-ext-team,

congratulations for the new release. This will be a large step for OpenFOAM. I have some questions regarding the release notes:

Improvements in accuracy and stability on tetrahedral and tet-dominant meshes

Do I need to use special numerical settings to get this? (eg. reconCentral Scheme?) or will it be available for all numerics? I was having some stability problems with 1.5-dev and I am looking forward testing this.

Demonstration solver for fully implicit block-matrix coupling

Is it possible to do implicit coupling of pressure and velocity? THat is what I am finally behind but the example solver does not do do this and I guess that 1.6-ext is not yet ready for this?

Turbomachinery features: GGI interface, cyclic GGI, partial overlap GGI

How much improvement has been done there from the latest 1.5-dev? Has the parallel speedup improved?

New parallel decomposition and reconstruction formulation

Cool. Does this mean you can decompose cases in parallel?

Regards and Thanks for all.

Bastian

[hjasak]

Tet meshes: it's in the settings. You need:

div(phi,U) Gauss reconCentral cellLimited leastSquares 1.0;

interpolate(U) reconCentral phi leastSquares;

Block implicit p-U coupling

Not there yet: as Martin says, first you make it run and then you make it fast! What has been released is the block matrix machinery and you can try to make the top-level solver using that. At Wikki we are looking for additional funding for the block-coupled p-U work (I am very confident I know what is required), but this will take some time. I'm sure I'll drop a line on the forum once this gets working, it is one of the few missing pieces of numerics that I rally fancy doing.

Turbomachinery

The big new feature is partial overlap GGI (Oliver Borm, thank you for doing this) and it looks really cool. It also allows good things to happen with non-matching rotor-stator pitch and (ulike some other techniques, eg. "frozen rotor") it looks physical to me. There is still outstanding work on parallel acceleration, which is sitting on Martin's and mine To-Do list, and this will come out in an update.

Parallelisation

We need to get a good set of testing results - that will tell us how much improvement has been made. "Decomposition-in-parallel": not sure what you mean exactly. There has been some discussion with Eric, but we've never really pushed it all the way.

So much for the moment...

Hrv

[elisabet]

Hi Hrv,

I'm interested in solving, with a block matrix, two scalar equations with a source term that depends on both scalars. I've seen your work at M&C 2009 (by Ivor Clifford) and the one at ECCOMAS 2010 (by Kathrin Kissling and Julia Springer). I've also looked in this forum but I always get a 'this will come soon'. Finally I read this:

What has been released is the block matrix machinery and you can try to make the top-level solver using that.

what sounds quite well. I'm still working with OF-1.5-dev and I've not seen anything corresponding to such 'block matrix machinery'. Could you please provide me a brief summary on how to deal with it?

thank you very much!

elisabet

[elisabet]

Sorry, I've got into OF-1.6-ext and saw it. Thanks!

elisabet

[hjasak]

Good hunting, Elisabet,

Hrv