User level notes are in Wiki here;
http://wiki.blender.org/bin/view.pl/Blenderdev/Softbodies
And will be added in blender3d.org CMS later.
Tech level notes are still pending, but here's the most relevant ones;
- made ob->soft struct SoftBody to hold all settings, and read/save in
files
- added (temporal!) conversion for the old settings. So: read old files
with softbody experiments now, and save over!
- cleaned API calls for softbody, which are only 5 of them now:
sbNew()
sbFree()
sbObjectStep() (animation steps)
sbObjectToSoftbody() (full re-initialize data)
sbObjectReset() (only reset motion)
- API calls accepts time in frames now, within softbody.c it converts
Further, internally code was cleaned some (missing tabs etc). Also tried
to keep a well defined structure with hints how to add support for more
objects. Can write notes about that...
- added G.editMesh->derived pointer... idea is to use this
for mesh derived from editmesh instead of Mesh->derived (as the
derived mesh tends to vary depending on what it came from).
This part could be cleaner, also there may problems with it not
being invalidated correctly.
- And most importantly: In case people were beginning to worry all
these edits were just crazy zr stuff, the big point comes about:
Incremental subsurf calculation is now enabled. This gives massive
speed improvements when editing a large mesh.
For the eye-candy happy: try setting G.rt==52 before entering editmode
and the edges and vertices (in optimal mode) will switch to displaying
visually the age since a region has last been calculated. Lots of
fun!
overrated. Turns out there was a bug in recalc code where
in editmode mesh was getting reevaluated all the time (because
ob->disp doesn't get built in editmode). This should fix.
- Fix some normal calculation/backwards face drawing
DerivedMesh implementation of CCGSubSurf. Only used in editmode at
the moment, and does not draw smooth normals correctly.
- Just in its own this brings a rather large speedup for editing
subsurfs.
to determine if element should be drawn and to set draw options),
which meant some could be combined - eases implementation of new
DerivedMesh types and keeps more app specific logic out of DerivedMesh
implementations.
- removed DL_MESH displist type!!!! Now store a DerivedMesh directly.
- May still be some issues left having to do with releasing this
at the right time (old code just splashed free_displist all
over the place).
- switched drawmesh.c to use new functions
- WHEE! This is last such function that has to be written for the time
being, and also effectively marks the end of DL_MESH! DispListMesh still
hangs around as a kind of interchange format but should also die once
DerivedMesh provides more intimate mesh access.
DL_MESH (the displist type for DispListMesh) is still used to hold
the DispListMesh, but that will die as soon as new subsurf switches to
using a bonafide DerivedMesh implementation. There is also one silly
usage for the decimator that is easy to kill.
- Oh, and I also made potatoe mode draw mesh vertex colors - seemed to
be more uniform behavior.
- cancelled previous commit to add RE_findTFAce, instead just added
a MemArena to render struct... free'd at end of render, can be used
to store other data as well
- switch rendering to using DerivedMesh API... this is slightly more
inefficient now because it is doing some unnecessary copying. Can
be fixed by defining a DerivedMesh function to convert the object
into a render object (on todo list)
implementations (slightly redundant, could implement draw in terms
of first... can be cleaned later)
- switch to use of new APIs... all uses of EditVert->ssco are now
clean (read: ready to die as soon as a proper DerivedMesh implementation
is swapped in with ccgsubsurf)
implementations
- updated drawobject selection code to use DerivedMesh (last DispListMesh
usage in drawobject.c that needs to be replaced! Woot!)
- interface could still be a lot cleaner, mostly calculating indices for
the various things to pass to selection color routine is hacky.
this is to ease migration.
- updated particles and STL converter to use DerivedMesh interface. This
is a bit wasteful at the moment, but these are not key performance areas.
Can update once DerivedMesh has accessors for faces and face data.
into drawobject.c (just used for calculating weight
map)
- removed two_sided (replace with glLightModeli calls)
- huge rewrite of drawobject.c for meshes, extracting simple
drawing functions and then reworking to bring some order
and clarity back to the code.
A lot was changed here so it is likely I missed a few
things in testing although I tried to be very careful.
Please let me know if you find any changes in drawing.
