From: John Stone (
Date: Wed Feb 21 2007 - 11:45:22 CST

  Actually neither one of these suggestions is really workable,
as the safety checks prevent IEEE NANs from being used in the vertex
coordinate interpolation. I don't want to add any additional flags
controlling the safety checks as they are in a performance-critical
region of the code. Setting the gridpoint value to zero doesn't affect
the need for the test, as the test is required in conditions when the
_difference_ between gridpoint values or positions is nearly zero, which
can occur for many combinations of voxel values, for example. I'm going to
test an alternative equation for stability under conditions of IEEE NAN
input values and see if that works. If it does, then I'll remove the
safety checks and the quality of the surface should degrade smoothly up
until NAN occurs. Anyway, give me some time to work on it
and I'll let you know how it goes.


On Wed, Feb 21, 2007 at 10:44:21AM -0500, Axel Kohlmeyer wrote:
> On Wed, 21 Feb 2007, Avell Diroll wrote:
> JS> > check just slightly differently to allow you to get your surface even though
> JS> > this specific case is definitely marginal at best. Try rendering the
> JS> > isosurface for isovalues 0, 0.0001, and 0.000001 and you'll see how
> JS> > small the volume gradient is at the outer edges of your box.
> how about adding a 'cutoff' threshold that will set all gridpoints to
> zero for the purpose of isosurface rendering if the absolute value of
> the grid point is below that threashold?
> we're doing something similar in the gradient corrected functionals in
> the density functional theory codes. below a certain density the data
> is too 'noisy' and you can get huge gradients due to that noise. since
> the density is _very_ low, it is error from ignoring that contribution
> (can be checked with much larger basis sets) is very small than the
> one introduced from the spuriously high gradients.
> axel.
> --
> =======================================================================
> Axel Kohlmeyer
> Center for Molecular Modeling -- University of Pennsylvania
> Department of Chemistry, 231 S.34th Street, Philadelphia, PA 19104-6323
> tel: 1-215-898-1582, fax: 1-215-573-6233, office-tel: 1-215-898-5425
> =======================================================================
> If you make something idiot-proof, the universe creates a better idiot.

NIH Resource for Macromolecular Modeling and Bioinformatics
Beckman Institute for Advanced Science and Technology
University of Illinois, 405 N. Mathews Ave, Urbana, IL 61801
Email:                 Phone: 217-244-3349
  WWW:      Fax: 217-244-6078