Re: 256^3

• To: Robert S Laramee <rlaramee@xxxxxxxxxxxxxxx>
• Subject: Re: 256^3
• From: "From: Bill Hibbard <hibbard@xxxxxxxxxxxxxxxxx>" <billh@xxxxxxxxxxxxx>
• Date: Sat, 14 Apr 2001 11:57:57 -0500 (CDT)

Hi Bob,

> Has anyone tried rendering (anything) with a 256^3 volume data set using
> VisAD?  In particular, has anyone tried the 256 x 256 x 113 cadaver head
> that appears in many-a-graphics research paper? If so, what hardware were
> you using?
> 
> So far we haven't gone past 128^3 resolution.

How much memory did that need (what was your -mxNNNm)?

256^3 is 16m grid points. A VisAD FlatField will let you
pack one value per grid point. Assuming you use a Linear3DSet
that will let you store the data in 16 MB. A change we made
last year in response to NCAR's concerns eliminated the need
for the display logic to create a temporary Gridded3DSet with
16m x 3 floats (192MB), but the iso-surface geometry and its
helper arrays will be pretty large (how large depends on the
geometry). Also, since Java3D bugs prevented us from using
Java3D version 1.2 we were unable to take advantage of its
memory efficiency. When I get a breather in my work I will
take advantage of these memory improvements for Java3D version
1.2.1.

Cheers,
Bill
----------------------------------------------------------
Bill Hibbard, SSEC, 1225 W. Dayton St., Madison, WI  53706
hibbard@xxxxxxxxxxxxxxxxx  608-263-4427  fax: 608-263-6738
http://www.ssec.wisc.edu/~billh/vis.html

• References:
  • 256^3
    • From: Robert S Laramee