Original (1.0) java was slow.  With 1.1 and above performance is
substantially better improved.  I have looked at benchmarks for both
native "C" and java in doing matrix manipulation (matrix multiplication,
rotations and transformations) any haven't seen any substantial
difference between the two (about half the time the Java code won out,
but always, the two were close).  I believe the term that is used is
"run time compiled byte code"?    What specific ray trace mathmatical
operations would be needed to give rough equivalency for a native .vs.
Java raytracer?  Matrix rotations?  Intersection testing?  Is there a
single small package in POY-RAY that could be ported to Java so that a
estimate of the performance could be made?

My desire in coding up POV ray was not so much a desire to see real
world processing time improvements, but to educate myself on
implementing a substantial practical algorithm on a distributed basis. 
The selection of Java, as was pointed out, for ease in implementing the
distributed architecture.  

So with the sage advice I have received from my esteemed collages.  I
dare to ask yet further questions about group think coding projects. 
What does the structure of the POV-RAY software development team look
like?  Is there one guy effectively doing all the work?  How are things
like archiving, bug reporting, patching, ... put together?

Secondly, is there a better build for one platform or another that would
be better as a starting place for a Java port?

Thirdly,
Jon, R-cast looks interesting, is the template document on the web serve
as the reference document for the development team?  Is that how a port
or project of these magnitudes gets started?

Thanks
Mike

Post a reply to this message