- Upgrade Recast library to latest portable version
- Implement recast_qsort based on FreeBSD qsort.c to have
portable thread safe quick sort for use in conversion routine.
- Better default value for the Build Navigation Mesh operator
Speaker objects fully functional!
Minor changes:
* Fixed three memory bugs found via valgrind.
* Fixed bug with jack transport crashing after file loading.
* Sound NLA Strips now start at CFRA instead of 0.
Implemented basic audio animation.
* AnimatableProperty: Propper cache writing and spline interpolation for reading (the solution for stair steps in audio animation)
* Animatable properties so far are: volume, pitch, panning
* Users note: Changing the pitch of a sound results in wrong seeking, due to the resulting playback length difference.
* Users note: Panning only works for mono sources, values are in the range [-2..2], this basically controls the angle of the sound, 0 is front, -1 left, 1 right and 2 and -2 are back. Typical stereo panning only supports [-1..1].
* Disabled animation of audio related ffmpeg output parameters.
* Scene Audio Panel: 3D Listener settings also for Renderer, new Volume property (animatable!), Update/Bake buttons for animation problems, moved sampling rate and channel count here
This fixes bug #26764 and several others like it, where modifier
properties (and others, but most visibly modifiers) would not do
anything when animated or driven, as modifier properties require the
RNA update calls to tag the modifiers to get recalculated.
While just adding a call to RNA_property_update() could have gotten
this working (as per the Campbell's patch attached in the report, and
also my own attempt #25881). However, on production rigs, the
performance cost of this is untenatable (on my own tests, without
these updates, I was getting ~5fps on such a rig, but only 0.9fps or
possibly even worse with the updates added).
Hence, this commit adds a property-update caching system to the RNA
level, which aims to reduce to the number of times that the update
functions end up needing to get called.
While this is much faster than without the caching, I also added an
optimisation for pose bones (which are numerous in production rigs) so
that their property updates are skipped, since they are useless to the
animsys (they only tag the depsgraph for updating). This gets things
moving at a more acceptable framerate.
- Sequencer dynamics: Now it's possible to change the output channels and the resampling quality also increased (previously maximum quality was 44,1 kHz)
- Changed two buffers to use ffmpeg allocation, not sure if that helps somehow.
- Sync Markers option works for local markers (or any other list of
markers in future) too now.
- Apply Pose to Restpose operator now displays a warning if an action
was found (warning about the action now being invalid)
Drivers for Scene, World, and Compositing Nodes now "work" (well,
sort-of)! Previously they were strictly restricted to object-
accessible data only; now they can function across the board (give or
take some weak spots).
Although there is still no depsgraph support so that these properties
update properly when their source controls are changed (this will
probably require a lot more work), they can still update under other
circumstances (i.e. frame change and/or manual refresh flushing via
mouse movement, etc.)
As the depsgraph tagging support is lacking, these just get always
executed for now, which might potentially be quite sluggish, though it
is hoped that there are so few of these top-level datablocks with
drivers hooked up that this is barely an issue in practice. At least I
haven't noticed any substantial slowdowns for animation playback, so
it should probably be fine.
These should not have any effect on render results, except in some cases with
you have overlapping faces, where the noise seems to be slightly reduced.
There are some performance improvements, for simple scenes I wouldn't expect
more than 5-10% to be cut off the render time, for sintel scenes we got about
50% on average, that's with millions of polygons on intel quad cores. This
because memory access / cache misses were the main bottleneck for those scenes,
and the optimizations improve that.
Interal changes:
* Remove RE_raytrace.h, raytracer is now only used by render engine again.
* Split non-public parts rayobject.h into rayobject_internal.h, hopefully
makes it clearer how the API is used.
* Added rayintersection.h to contain some of the stuff from RE_raytrace.h
* Change Isect.vec/labda to Isect.dir/dist, previously vec was sometimes
normalized and sometimes not, confusing... now dir is always normalized
and dist contains the distance.
* Change VECCOPY and similar to BLI_math functions.
* Force inlining of auxiliary functions for ray-triangle/quad intersection,
helps a few percentages.
* Reorganize svbvh code so all the traversal functions are in one file
* Don't do test for root so that push_childs can be inlined
* Make shadow a template parameter so it doesn't need to be runtime checked
* Optimization in raytree building, was computing bounding boxes more often
than necessary.
* Leave out logf() factor in SAH, makes tree build quicker with no
noticeable influence on raytracing on performance?
* Set max childs to 4, simplifies traversal code a bit, but also seems
to help slightly in general.
* Store child pointers and child bb just as fixed arrays of size 4 in nodes,
nearly all nodes have this many children, so overall it actually reduces
memory usage a bit and avoids a pointer indirection.
Since 2.5x blender has been using CD_MASK_BAREMESH for updating objects since object_handle_update() no longer has access to G.curscreen to calculate the mask from viewports.
The problem with this is after an initial calculation, CD_MASK_MTFACE may be required on draw, so it would recalculate the modifier stack multiple times per frame.
One case which caused this is armature animated mesh with texface in a dupligroup.
Fix this by having customdata_mask member in the scene, this isn't great design but at least fixes the bug and only changes a few files.
Surprising this wasnt noticed in a much more obvious case:
- Key Location, Move, Rotate, Undo-Rotate >> Resets to keyed location as well.
This was happening because DAG_on_load_update() was called on read_undosave(), flagging 'ob->adt->recalc |= ADT_RECALC_ANIM;'
Fix by adding an option to DAG_on_load_update(), not to recalculate time flags.
only tags the ID and does the actual flush/update delayed, before the next
redraw. For objects the update was already delayed, just flushing wasn't
yet.
This should help performance in python and animation editors, by making
calls to RNA property update quicker. Still need to add calls in a few
places where this was previously avoided due to bad performance.
Also use const char in many other parts of blenders code.
Currently this gives warnings for setting operator id, label and description since these are an exception and allocated beforehand.
Both stored the filename of the blend file, but G.sce stored the last opened file.
This will make blender act differently in some cases since a relative path to the last opened file will no longer resolve (which is correct IMHO since that file isnt open and the path might not even be valid anymore).
Tested linking with durian files and rendering to relative paths when no files is loaded however we may need to have some operators give an error if they are used on the default startup.blend.
- removed deprecated bitmap arg from IMB_allocImBuf (plugins will need updating).
- mostly tagged UNUSED() since some of these functions look like they may need to have the arguments used later.
