Op woensdag 26 februari 2003 20:07, schreef Miguel:
> While I was traveling last week I had the opportunity to spend a lot of
> time thinking about 3d rendering in Jmol.
>
> My thoughts in random order:
> - It is ridiculous to require Java3D.
> * It severely limits the user-base.
> * I'm not a member of the "graphics community", but it doesn't
> seem to me that it has very much momentum.
My feeling is just that...
> - We need a solution that supports 3D as an
> applet with an old browser JVM.
Agreed. The need for 3D display applet is still very high...
On thursday I had two people over about crystal structure display.
There were basically two things missing:
1. building crystals using symmetry operations (space groups)
I've already discussed this with Fabian in the past and I plan
to at least implement this in v7 partially.
2. Display over isosurfaces, which they need for energy band plotting
I think.
The latter would be great to have implemented.
> - I worked out most of the design for a z-buffer scheme that
> (I believe) would work quite well for molecular viewing.
Can you email how it works?
> - I also worked out a more complicated scheme which would offer
> significantly better performance
> * rendering times would be cut by 25%-50%
> * at a cost of significantly more complex code
As long as the algorithm is very well described, and the code
is written is good as possible (variable names, method names etc)
I see no problem there...
> - performance is still going to be significantly slower than the
> (rather tight) C code of RasMol ... my wild guess is 5 times slower.
Indeed, but Rasmol won't work inside a webbrowser... it's a trade off
I guess...
> - I took a look at the OpenRasmol 2.7.1 code and uncovered one
> rendering performance *issue* with cylindrical bonds.
> That will be an area where Jmol could "catch up" a bit.
;)
> - Efficient 3D implementation of the following important
> shapes is rather straightforward
> * shaded spheres (aka Atoms)
> * shaded cylinders (aka Bonds)
> * lines
> * text
> - intersections these shapes would be *correct*
> * atoms displayed at 100% vdwRadius would do the "right thing"
A so called space filling rendering is indeed 'missing' at this time...
I would put higher priority on isosurfaces though...
> - I've not put much thought into the other shapes that are
> in RasMol (ribbons, strands, etc), but I don't think they
> would be too difficult.
This is an option that is often used by biochemist, molecular biologist
and all the other names for those people...
> - I think I could implement it in a month or two of focused work.
>
> So, my questions for you are as follows:
>
> Do *you* think that true 3D rendering functionality this would be
> valuable for the Jmol viewer? (By true 3D I mean that objects
> intersect correctly, lines/vectors pass through atoms and
> come out the other side, etc.)
The current Jmol has worked for me as it is... I've never needed
anything more fancy up to now. As said isosurfaces seems to me
one one the interesting things missing...
> For the work that *you* are doing, is true 3D rendering important?
No, rendering of things like hydrogen bonding is actually the only
thing I now miss in my molecular crystal structure work myself...
> Or is the pseudo-3D rendering of Jmol good enough?
For me it is.
> For example, is it important to render "charge fields" correctly for the
> things you are doing?
If you mean charge fields rendered as isosurface....
> If *you* had some "skilled talent" to allocate for a month or two, is
> this the area where you would choose spend it?
> If not, what would *you* have me work on?
No, *if* I had money, I would have you spend your time on integrating
Jmol work CDK, i.e. having it use a common rendering API, the CDK core
classes, port the IO to CDK, etc...
And yes, a port of the scripting engine... I would just love to have a
scripting mode in JChemPaint, so that in the future (combined with a
speech interpreter) I could just say: add a methyl group at atom 5,
make atom 3 an oxygen, generate a reaction where electrons move
from bond 7 to atom 2....
> It seems to me that this is the last major piece that we need to start to
> make Jmol an acceptable RasMol/Chime substitute. Clearly it would take a
> lot more polishing, but all the pieces would be there. What other pieces
> are missing?
*If* the work *had* to be in the area of having Jmol substitute Rasmol/Chime ,
I would have you work on display of cartoons of enzymes, backbond, helices,
sheet etc...
> Do *you* have any interest in a RasMol/Chime substitute?
Yes. But as an substitute for displaying small molecules it already does
everything I need...
Egon
-------------------------------------------------------
This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
http://thinkgeek.com/sf
_______________________________________________
Jmol-developers mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/jmol-developers
