Blender Git Commits

seems to run faster now, but still have a big memory leak

minor correctional fixes, enable automerge workaround again

cache_offset_map, store offset instead of calculating, but only 1fps speed gain, hrm

oddly, in rigidbody.c not everywhere MEM_ functions were used, fixed that due to odd crash

automerge fix applied again (was not working any more)

fixes to allow other constraint types and (mostly) correct outer constraints again

further work at speedup, automerge needs to be tackled still

first attempt to increase performance: disable storage of fracture results, this costs more performance instead improving it, and added a boundbox test for constraints calculation (but often those intersect each other, so not much performance gain, also need to fix for not touching islands which possibly should act as fixed too (using higher contact distance)

partial success with speedup attempt, but automerge still slow and wrong

some stabilization (refreshing modifiers crashed before) but still error with baked data and memory leak

attempt for fixing bake error, unfortunately no luck yet

some attempts to increase performance, only a bit faster now

merge related fixes

correction: calculate only (many) fixed constraints faster, the other outer constraint types connect to 1 shard basically and work incorrectly if calculation stops at 0 shared vertices.

merge related compile fix

small performance plus when building constraints, stop when 0 shared vertices occur (because meshislands are sorted by distance, kdtree)

disable auto-refresh in rigidbody modifier (can be very slow, but ... always remember to hit refresh before starting the sim now !)

re-introduce Group Contact Distance (has effect for fixed type only) fix for changing passive to active unexpectedly

adapt change from BMesh Operator "recalc face normals", removed some printouts

a bit more stabilization for outer constraints, but cache shows still unexpected behavior