[Soc-2008-dev] Week 11 report

[Nicholas Bishop]

a) Previous week:
* Fixed sculpt mode's partial redraw mode when used with multires
* Some speed improvements (less updating when it's known that
displacements haven't changed)
* Some bug fixes

I also spent some time looking at speed and data structure issues. As
I explained last week, there are speed issues with the way level
updates and subdivision are currently handled. I've been experimenting
with some pretty radical changes, such as directly subdividing
displacements (rather than vertex coordinates) and also faster regular
subdivision of the mesh. Part of the problem is that CCGSubsurf was
designed for a different use than what multires is using it for. Its
partial-update system (used to good effect by the subsurf modifier in
editmode) is actually not of much to the multires modifier.
Speed-wise, CCGSubsurf is slower than an iterative subdivision
approach, and requires a lot of memory if applied to a dense base mesh
(this claim is based on comparing the old multires in trunk to the
subsurf modifier.) Based on this, I'm tentatively planning to write
new subsurf code. This will have to wait until after GSoC, of course.
The old multires subdivision code was a fairly basic Catmull-Clark
implementation, and as it noted above, it is fast. However, it had
some serious problems, notably in it's handling of customdata (and
especially UVs), and it's vertex ordering would be seriously
inconvenient to map displacements to. So something entirely new is
needed; I'll be looking into this more after the 18th.

b) Next week:
Next week is the last week of GSoC. My top priority is fixing file
loading for old files. I've been trying to make progress with the
conversion, but the data storage is very different between the two
formats, and I'm having difficulty wrapping my brain around the best
approach (or really, any approach.) There's also a data-loss bug in
the file save code that needs to be fixed.

On Fri, Aug 8, 2008 at 6:13 PM, Ian Thompson <quornian at googlemail.com> wrote:
> Week 11 Report
> ==============
>
> This Week
> ---------
> The marker system is now in place. The behaviour of markers is dictated by their
> flags and groupings. Placing markers with the TMARK_EDITALL flag set allows a
> number of spans to be edited from one (as in the Find All case). The TMARK_TEMP
> flag denotes markers which should be removed once focus is taken elsewhere.
> Together with grouping information (currently the upper two bytes of a marker's
> flags) these provide a platform for code templates.
>
> The textplugin_templates.py contains an implementation of such a system, which
> may be extended by the user with a simple syntax (to be documented).
>
> Find all is implemented as an option ("Mark All") on the find and replace panel,
> which has also seen some improvements this week. It makes use of markers to edit
> all occurrences at once.
>
> The last few word-wrap functions (Home, End, etc.) have also now been completed.
>
> Next Week
> ---------
> Obtain as much feedback as possible. Solve any remaining issues and adjust any
> behaviour that may confuse or hinder users. Finish writing integration docs and
> write full user documentation ready for the wiki.
>
> Issues
> ------
> None
>
> Schedule
> --------
> The main code is complete. Time now to polish things off and get ready for
> integration. This is the final week and I'm on target to finish by the deadline.
> _______________________________________________
> Soc-2008-dev mailing list
> Soc-2008-dev at blender.org
> http://lists.blender.org/mailman/listinfo/soc-2008-dev
>