A new release of my MegaPOV port to PVM is available here:

http://www.wozzeck.net/images/pmp/

It is still based on MegaPOV 0.5a
Slaves now report errors instead of letting the master wait for them
indefinitely.
Radiosity has been improved a bit, and some bugs have been found, but
there are still some problems.

Although the octrees are the same after the "preview" passes, and even
if the samples calculated during the final passes are not stored, there
are still differences between adjacent blocks. Any idea?
Another idea is that all the "preview" passes are currently made at once
for a given block. This means that when a slave starts the preview
renderings of a block, it does the first to the last preview passes,
then shares the octree. Would it be MUCH better if the preview passes
were actually ran separately? This would make the code much more complex
and increase the overhead, but if it solved the problem, it would become
worth doing.

Thanks in advance for your feedback.
-- 

      __  __ __  __  _
|  | /  \  /  / |_  /  |/
\/\/ \__/ /_ /_ |__ \_ |\

Post a reply to this message