Dev:SundayMeetingAgenda/2007-09-23rd

FTP

Ton introduced a new temporary FTP space and sent use instructions by e-mail on [BF-committers]. Extract: A much wanted request:

ftp to: download.blender.org user: anonymous pass: email address

When logged in, go to the directory 'incoming'. What you put there is visible here: http://download.blender.org/ftp/incoming/

This upload space deletes all old files after a week, so don't use it for permanent storage. You can ask me, Nathan Letwory, Chris Want or Kent Mein to move it to a permanent location on the server.

Physics Sprints

The two days before next BConf will be held an experiment, a 'sprint' to try to speedup development of Blender Physics; so far, will be involved: Bjornmose, Janne, Genscher (TBC, no answer from University), Ton, Brecht, Nils, Campbell.

Many topics have been proposed for this sprint:

• Unified baking and rigid bodies integration (mfoxdogg)
• Design, evaluation of bottlenecks, and action plans (Ton), units of measurements common to all dynamic stuff (Matt backed by Janne, while Ton would rather prefer a unit-free implementation).
• Ton expressed the need for a single Physics API, for collision and spatial lookups for example, or a help library with basic calls.
• Matt stressed that bake modifier would be interesting, as baking armature deformation would speed-up many characters, through baking to a point cache file to stream from disk.
• External systems like bullet (bjornmose) has also been listed.
• Ton also proposed to design physics UI stuff, in order to have less buttons, more tools and editing things in 3d window; for example, a rigidbody toolkit should allow clickety-click 'n go.
• mfoxdogg then offered Physics Threading as a topic, backed by Jahka while Genscher pointed that OpenMP is already in use with Cloth simulator and could also be with Fluids according to Nils. Janne added particles to the list of possible benefactors of OpenMP.
• Ton also offered to show the work of the physics-character dude who has code to solve motion interpolation physics based and possibly introduce controllers (something like constraints, but different). The idea would be to think over uniform constraint/controller stuff.

Mfoxdogg asked for the sprint to be streamed.

Current Projects

Matt stated that his HDR work is actually not finished and clunky.

Janne has been buzy with studies this week too, has made some progress on brecht's review and some bugs, and will try to commit changes, and make a merge from trunk this evening or tomorrow.

Brecht reported having eeshlo's micropolygon stuff and wanted to start a svn branch and do integration work. No doc is available but there are comments in the code though, and it closely follows the design from the "production rendering" book. As it has the standard renderman api, it can be used as a library. Brecht worked on it since two days and got some basic stuff working. According to brecht, the code could be ripped apart in order to make closer integration easier, but then the nice renderman architecture would be lost. Brecht plan would be to extend the renderman API with Blender specific calls.

Mathias is still hacking in blender's old yafray export code and pointed out that new code is not compatible at all with old one. Basically he rewrote the whole GI and material system of yafray, which are more physically based, would allow more intuitive GI setup, new GI algorithms, etc. Recent example of improvements: [1]. Ton mentioned that the project should get boosted, but unfortunately, only a couple of guys showed interest but no idea if one of them will really dig into the code... Only one guy is working on the python script so Yaf(a)ray would be usable from Blender. Martin pointed to Veach's extensive thesis on Monte Carlo rendering (http://graphics.stanford.edu/papers/veach_thesis/) but Mathias already studied it.

SoC results

Apparently mosani was still working on his SoC Project, and no news from maike either.

Red_fox's bevel code looks really close to be finished.

Hcube has just some linking issues to be solved (Ubuntu Linux, Feisty Dawn), and was asked to request help on the lists about the troubles that has been reported and that he can't witness himself. The discussion lasted after the meeting has been officially closed.