librelist archives

« back to archive

[tisaac@ices.utexas.edu: Re: Balance algorithm question]

[tisaac@ices.utexas.edu: Re: Balance algorithm question]

From:
Tobin Isaac
Date:
2014-09-22 @ 17:52
For those interested in the algorithms in p4est, I should point out an
error that was recently found in the description of the 2:1 balance
algorithm in our IPDPS 2012 paper.  Thanks for John Mousel for
noticing this.

Best regards,
  Toby

----- Forwarded message from Tobin Isaac <tisaac@ices.utexas.edu> -----

Date: Mon, 22 Sep 2014 12:12:02 -0500
From: Tobin Isaac <tisaac@ices.utexas.edu>
To: John Mousel <john.mousel@gmail.com>
Subject: Re: Balance algorithm question


Hi John,

Yes, line 3 of algorithm 7 should include newly added octants.  Thanks
for pointing out this mistake, I'll include a note of it on my website
and on the p4est website.

Best regards,
  Toby

On Mon, Sep 22, 2014 at 12:02:12PM -0500, John Mousel wrote:
> Hi Tobin,
> 
> I'm trying to implement the 2:1 balancing algorithm (Figure 7) in your
> paper: "Low-Cost Parallel Algorithms for 2:1 Octree Balance" and am running
> into a slight issue. It seems my balancing does not propagate far enough.
> Looking at the algorithm in Figure 6, newly added octants also add their
> coarse neighborhood octants. However, algorithm 7 does not seem to state
> this as necessary. Should line 3 (for all o in R do) include newly added
> octants?
> 
> I'd appreciate any clarification you could provide.
> 
> Thanks,
> John Mousel



----- End forwarded message -----