Under some circumstances using task isolation can cause deadlocks.
Previously, our task pool implementation would run all tasks in an
isolated region. Now using task isolation is optional and can be
turned on/off for individual task pools.
Task pools that spawn new tasks recursively should never enable
task isolation. There is a new check that finds these cases at runtime.
Right now this check is disabled, so that this commit is a pure refactor.
It will be enabled in an upcoming commit.
This fixes T88598.
Differential Revision: https://developer.blender.org/D11415
Share the pointer with the original mesh instead, this matches behavior
of all other objects edit-mode data.
Duplicating the edit-mesh pointer makes updates to edit-mesh require
a COPY_ON_WRITE update, which is currently an expensive operation
(copying the entire mesh).
Notes:
- This change is from 802027f3f8
so the edit-meshes object pointer `BMEditMesh.ob` referenced the COW
version of the object. This pointer has since been removed, so the
copy is no longer needed.
- Having a separate edit-mesh pointer could be used so linked duplicates
could have their own generated meshes. For this to be supported,
many other changes would be needed: see D10920.
Resolve ownership ambiguity with shared physics pointers.
Previously, LIB_ID_CREATE_NO_MAIN allowed pointer sharing with
the source ID so physics caches can be shared between original and
evaluated data: (Object.soft.shared & Object.rigidbody_object.shared).
This only worked properly for LIB_TAG_COPIED_ON_WRITE ID's,
as LIB_TAG_NO_MAIN can be used in situations where the original ID's
lifetime limited by it's original data.
This commit adds `LIB_ID_COPY_SET_COPIED_ON_WRITE` so ID's only share
memory with original data for ID's evaluated in the depsgraph.
For all other uses, a full copy of physics data is made.
Ref D11228#287094
Issue was caused by anim handle being reset by
`DEG_evaluate_on_framechange()`
Preserve anim handle by backing it up in
`blender::deg::SequenceBackup`
Reviewed By: sergey
Differential Revision: https://developer.blender.org/D11059
Fix `nullptr` redeference when setting 'orig_data' pointers on CoW copies,
by stopping the loop also when `element_cow == nullptr`. This avoids a
crash of Blender when the original list of pointers is longer than the
CoW list of pointers.
I've also added a `BLI_assert()` that checks for equal lengths of the
two `ListBase`s, so that problems like these aren't hidden away completely.
The root cause of the crash was actually a modifier that was assigned to
an object of the wrong type (an Armature object doesn't support modifiers).
This caused the list of modifiers on the CoW copy to be shorter than the
list of modifiers on the original Object. It's still a mystery how that
object got that modifier in the first place.
This removes a lot of unnecessary code that is generated by
the compiler automatically.
In very few cases, a defaulted destructor in a .cc file is
still necessary, because of forward declarations in the header.
I removed some defaulted virtual destructors, because they are not
necessary, when the parent class has a virtual destructor already.
Defaulted constructors are only necessary when there is another
constructor, but the class should still be default constructible.
Differential Revision: https://developer.blender.org/D10911
Handle Lattice object the same way as Mesh objects. This is mostly to
execute the `object->data = data_eval;` line, which ensures that the
evaluated mesh is assigned to the evaluated object, and thus prevents
the lattice from un-deforming.
Rename:
- `BKE_animsys_store_rna_setting` → `BKE_animsys_rna_path_resolve`
- `BKE_animsys_read_rna_setting` → `BKE_animsys_read_from_rna_path`
- `BKE_animsys_write_rna_setting` → `BKE_animsys_write_to_rna_path`
The concept of "RNA setting" is unclear; the new names reflect better
what the functions actually do.
No functional changes.
Corrects incorrect usage of contraction for 'it is', when possessive 'its' was required.
Differential Revision: https://developer.blender.org/D9250
Reviewed by Campbell Barton
Note that possibility to pass the new ID pointer as parameter was kept,
as this is needed for some rather specific cases (like in depsgraph/COW,
when copying into already allocated memory).
Part of T71219.
This reverts {rB1693a5efe91999b60b3dc0bdff727473b3bd00bb}
and implements an alternative solution.
The old patch had the problem that the depsgraph would always
evaluate at the current frame of the original scene (even when
`DEG_evaluate_on_framechange` was used). Now it is possible
to evaluate the depsgraph at a specific frame without having to
change the original scene.
Reviewers: sergey, sybren
Differential Revision: https://developer.blender.org/D8616
This mainly removes the bmain argument, which can be
retrieved from the graph itself.
Also, I removed some outdated/unnecessary comments.
Reviewers: sergey, sybren
Differential Revision: https://developer.blender.org/D8614
Solves possible pointer-based comparison fiasco.
Another nice outcome of this is that topology cache will now be
preserved throughout the undo system. For example, undo of object
transform will not require topology cache to be re-created.
Differential Revision: https://developer.blender.org/D8493
The root of the issue comes to the fact that sub-data pointers were
used to match strips before/after copy-on-write. The undo system might
re-use sub-data pointers after re-allocating them, making it so that,
for example, pointer used by sound strip is later re-used by video
strip.
This fix takes an advantage of recently introduced per-sequence UUID
and uses it to match sequences before/after copy-on-write.
Will trigger code paths which makes sure UUIDs are generated and
are unique.
Enabled with --debug-depsgraph-uuid (which is also implied by
--debug-depsgraph).
A simulation data block has an embedded node tree, which requires
special handling in a couple of places. Some of those places were
missing beforehand.
This also adds a relation to make sure that the simulation is evaluated
after animations on the embedded node tree are evaluated.
This updates the usage of integer types in code I wrote according to our new style guides.
Major changes:
* Use signed instead of unsigned integers in many places.
* C++ containers in blenlib use `int64_t` for size and indices now (instead of `uint`).
* Hash values for C++ containers are 64 bit wide now (instead of 32 bit).
I do hope that I broke no builds, but it is quite likely that some compiler reports
slightly different errors. Please let me know when there are any errors. If the fix
is small, feel free to commit it yourself.
I compiled successfully on linux with gcc and on windows.
Every Particle Simulation node has a name (or a path when it is in a node group).
This name has to be used in the Simulation modifier on a point cloud to see
the particles.
Caching has been disabled for now, because it was holding back development
a bit. To reset the simulation, go back to frame 1.
Currently, there is no way to influence the simulation. There are just some
randomly moving points. Changing that is the next step.
Note that this is partially WIP code, we only take care of shapekeys
here for now.
Also, move this tagging for liboverride refresh into same chack as the
one for tagging editors, sounds more logical that way.
This makes any operation (including mere bone selection) several times
faster on some complex production character, since we typically now only
need to diff a single ID, instead of tens of them.
We decided to use our own map data structure in general for better
readability and performance.
Reviewers: sergey
Differential Revision: https://developer.blender.org/D7987
Now all overrides are handled that way. Performances of the process look
decent enough, even with production characters...
If performance issues still arise, we'll investigate other solutions.
This should also make T73154 obsolete now.
