[Soc-2008-dev] Week 3 Report

Davide Vercelli davide.vercelli at gmail.com
Sat Jun 14 09:58:27 CEST 2008


Week 3 Report
=============

Please note: I am note 100% sure I can attend the Sunday Meeting this
week. Sorry about that.

This week
---------
The goals set last week were all met:
- found and fixed the pretty serious bug I was looking for: it turned
out that I could get negative light contributions, which pretty much
explains all the major rendering errors I have had so far
- added code to compute the material error bound, that is the maximum
contribution that a subtree of light can give; this involves computing
the maximum cosine between the normal and any point inside a bounding
box (the "cosine bound" I mentioned last week and was told not being
clear)
- as for the debug tools I needed, as I already pointed out on this
list, Matt Ebb was kind enought to show me they were already in place,
which is very good news indeed!

Also:
- reintroduced "constant sampling" for lights, in order to perform
fair comparisons with lightcuts (will be removed in the end)
- added end-of-render statistics

Bonus: I finally updated the wiki, including adding a short project
faq and an updated schedul (albeit still pretty generic):
http://wiki.blender.org/index.php/User:UncleZeiv/SummerOfCode2008

Next week
---------
Fixing the bug mentioned before removed some blatant rendering errors;
adding the material error bound, though, some differences with respect
to the "vanilla" rendered images became apparent. This may have many
causes that are too long to discuss here, but I have many ideas of
what could be going on so I am not worried yet.

It's time to add support for other types of lights and/or shaders. My
intention is to add support for directional lights first: should be
fairly easy but requires a bit of a refactoring since I have to handle
more than one tree. Directional lights would finally enable more
interesting rendering and comparisons, e.g. using environment maps.

Issues
------
None so far.

Schedule
--------
Last week I was a bit stuck both on writing new code (cosine bounding
stuff) and on bug hunting. Now things are definitely back on track.

Davide


More information about the Soc-2008-dev mailing list