Just for grins, I wrote a little module that converts n non-degenerate cubes to 2*n prisms, represented as degenerate cubes, and set it up in a net that generates 1000000 irregular cubes with irregular positions. It takes 30 sec. or so for the cubes and less than twice that for the prisms - which is what I'd expect. Pictures look identical. Here it is, if you are interested.
Greg
(See attached file: prisms.tar.gz)
Neil Carlson
<[EMAIL PROTECTED]> To:
[email protected]
Sent by: cc:
[EMAIL PROTECTED] Subject: Re: [opendx-dev]
adding a new element type
son.ibm.com
05/30/2002 12:28 PM
Please respond to
opendx-dev
On Thu, 2002-05-30 at 09:28, Gregory D Abram wrote:
>
> Hard, I'm afraid. Each module does its own thing on each element type -
> eg. MapToPlane has code to MTP each element type, Isosurface has code to
> isosurface each element type etc, etc, etc. Often, there are several
> methods, depending on whether the positions component is regular,
> partitially regular, or fully irregular. I'm curious - where do problems
> arise with degenerate cubes? I'll grant the 1/3 overhead for the
> connections component and neighbors. Anything else?
At this point I don't know what's going wrong, and need to do further
testing. It appears to work for tiny meshes (thousands of cells), but
for larger meshes (100 of thousands of cells) it zips through the import
and other modules, but then bogs down in the image module for ~15min
before the image window finally displays. This is on a 1.6GHz athlon
with 1GB memory. My *wild* speculation is that there is some n^2
(n^3,...) loop related to coping with degenerate elements. But I've
got more testing to do...
--
Neil N. Carlson Motorola, Los Alamos Research Park
Motorola Labs / PSRL 4200 W. Jemez Road, Suite 300
Computational Nanoscience Group Los Alamos, New Mexico 87544
[EMAIL PROTECTED] Fax: (505) 663-5150
Voice: (505) 663-5106 Pager: [EMAIL PROTECTED] or
888-946-2817
prisms.tar.gz
Description: Binary data
