[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
doubles vs. floats [was Re: ElevationGrid II]
Martin Reddy wrote:
> Gavin Bell wrote:
> > -- Double-precision coordinates (DoubleCoordinate maybe?) to support CAD
> > applications that require very-high-precision models. I'm suprised a
> > VRML-For-CAD working group hasn't started to talk about this, and about
> > adding trimmed NURBS and other features common to CAD applications.
>
> This is one of the issues that the GeoVRML Working Group is interested in
> looking at. Single-precision floats don't provide enough accuracy to model
> the Earth down to anywhere near meter accuracy (diameter of Earth ~= 12,000
> km = 12,000,000 m = 8 digits of precision. Single precision values only
> store around 6 digits of precision).
>
> At SRI we have gotten round this limitation for our VRML terrain
> visualisation application by implementing a hierarchy of local coordinate
> systems, each of which are then transformed into the global coordinate
> system. But support for SFDouble/MFDouble types in VRML would make life way
> easier!
a. is anyone looking at a pairwise MFFloat convention to act as doubles?
or better yet, pairwise representations for general quadrant/precise relative
location? in other words, think about how we might PROTO a base type.
b. until this group starts nailing down coordinate system conventions, and
maybe picks the most likely/useful one(s) to build as initial shared
implementation(s), unfortunately i think this conversation will remain
wishful thinking on a "he said/she said this/that about doubles/floats"
level. it would be much more interesting to lay out the existing candidate
representations side by side, compare/contrast strengths/weaknesses, and
look at actual results.
c. there is a lot of work going on already. why don't we build a group
plan about how to best explore the double vs. floats question?
all the best, Don
--
Don Brutzman Naval Postgraduate School, Code UW/Br Root 200 work 408.656.2149
Monterey California 93943-5000 USA fax 408.656.3679
Virtual worlds/underwater robots/Internet http://www.stl.nps.navy.mil/~brutzman