Change default wall boundary condition to bounce back.

[schmie]

Reported by miguel on 30 Sep 2011 10:22 UTC
Bounce back wall boundary condition seems to reproduce more accurately Poiseuille velocity profiles than our current default. It would, therefore, make sense to make it default.

The (slightly outdated) patch attached modifies the current implementation of SimpleBounceBack to match the new design of streaming and collision.

Todo:

• Merge patch (things have changed in lb.h since patch was created).
• Make bounce back default.
• Amend regression test and perhaps unit tests.

[schmie]

Comment by hywel on 7 Oct 2011 09:00 UTC
Yup, I agree with the plan.

[schmie]

Modified by hywel on 13 Oct 2011 11:38 UTC

[schmie]

Comment by miguel on 18 Oct 2011 14:52 UTC
I'm attaching an updated version of Hywel's patch (I had to deal with encapsulation of velocity distribution functions and the place where the non-equilibrium one is computed, please review).

Unit tests are fine, but regression test is now broken on both snapshot and image comparison. Shall I simply overwrite the CleanSnapshots/ folder with the new results and forget about images (cuz of streaklines issue)?

[schmie]

Comment by miguel on 18 Oct 2011 14:58 UTC
I forgot about creating the relevant unit test :( Working on that now.

[schmie]

Comment by hywel on 20 Oct 2011 13:17 UTC
Diff looks good but I don't like the formatting of the ternary operator. I know it's pedantic but
" ?
:
;"
is much nicer in my opinion.

Otherwise fine, no need for a re-review of this bit.

[schmie]

Modified by miguel on 27 Oct 2011 17:43 UTC

[schmie]

Comment by miguel on 27 Oct 2011 17:46 UTC
Please review previous patch, Hywel.

[schmie]

Comment by hywel on 27 Oct 2011 19:19 UTC
Replying to miguel:

Please review previous patch, Hywel.

=== !StreamerTests ===

• 288 - can't remember if we agreed on numWallSites or wallSiteCount
• I think we're using the allowedError enough that it should be a class-level constant
• 383 - I think you mean f post-collision not fOld.

Otherwise absolutely fine, no need to re-review. Cheers!

[schmie]

Modified by hywel on 28 Oct 2011 13:25 UTC

[schmie]

Comment by miguel on 28 Oct 2011 17:03 UTC
Changes pushed 272dc2528da0. This will break diffTest. Rupert and I are working on it at the moment.

[schmie]

Comment by rupert on 1 Nov 2011 13:59 UTC
The subrepo problems are resolved in changeset bae52c3880fa

James- please can you check this before resolving?

[schmie]

Comment by jamespjh on 3 Nov 2011 11:07 UTC
Subject to my hatred of subrepository.

[schmie]

Modified by jamespjh on 3 Nov 2011 11:28 UTC