This patch improves working with grease pencil layers in geometry nodes.
* Allow layers to have duplicate names in geometry nodes. In original data, unique names are enforced.
* This allows e.g. duplicating layers and then merging them by name in the end.
* It also resolves a big serial bottleneck when working with many grease pencil layers in geometry nodes. Enforcing unique names is inefficient.
* New `Merge Layers` node that can merge multiple layers by name or by a custom group id.
* Applying a grease pencil modifier now first merges all layers with the same name to ensure all names are unique.
Co-authored-by: Jacques Lucke <jacques@blender.org>
Pull Request: https://projects.blender.org/blender/blender/pulls/127873
A driver (package) installed by the user can have many different drivers
and they can all report a different version. For AMD the version we
reported was from their Vulkan driver. This version isn't useful during
bug triaging.
This PR will use the driver info and driver name from the driver
properties to construct a driver version string that will be used for
reporting.
Pull Request: https://projects.blender.org/blender/blender/pulls/128232
347ec1acd7 switched to iterating over nodes with index
masks but introduced nested lambdas with the inner one slicing
the index mask with a range provided by `parallel_for`. That was
done to avoid calling `.local()` more than necessary but in practice
that doesn't have an effect, especially with a grain size of 1.
That nested iteration is cumbersome and error-prone. For the
simple parallel loops with thread local data, switch to just using the
`GrainSize` argument to `mask.foreach_index` instead.
Pull Request: https://projects.blender.org/blender/blender/pulls/128221
In this case there are two dependency graphs, each with a separate
object containing the original mesh. Leaving edit mode modifies the
original mesh by deleting the edit mesh data. The active dependency
graph's object was tagged for that change, but the other dependency
graph didn't know about the change because its object wasn't tagged.
The solution is to tag the original mesh which will cause any
dependency graph using it to properly reevaluate.
Pull Request: https://projects.blender.org/blender/blender/pulls/128192
This commit forcibly rebuilds the PBVH whenever the number of verts is
changed by an operation, additionally, the related deform variables are
freed when undoing geometry steps now to ensure data remains consistent.
Pull Request: https://projects.blender.org/blender/blender/pulls/128145
Customizable highlighting of editor edges that can be used to help
differentiate between areas, indicate active area, and/or help users
with visual impairment. Can be completely removed as well. This
replaces the active area header highlighting, which doesn't work for
transparent headers.
Pull Request: https://projects.blender.org/blender/blender/pulls/116684
A combination of a few changes:
- Don't apply hardness again, the "cube tip" calculation already factors it in.
- Avoid multiplication by brush radius, just use the 0-1 distances from the
cube tip calculation directly.
- Add filtering by brush radius to fix the second part of #128004.
- Change brush radius filtering to be consistent with previous code using
less than or equal instead of less than.
Note that the current radius filtering in BKE_brush_calc_curve_factors is
unnecessary since we already filter by the radius in a separate step.
There were multiple issues that popped up when following the reproduction steps
in the report:
* Somehow, `socket.socket_data` was set for a custom socket type. I don't know
how that happened yet. I added some code to clean it up on file-load.
* `declaration_for_interface_socket` did not work yet when the socket type is
not registered. Now it just creates a custom socket with the correct idname,
even if that's not registered yet.
* There was a missing node tree update when changing `bl_socket_idname` of
interface sockets.
Pull Request: https://projects.blender.org/blender/blender/pulls/128191
Previous change in there (d0cf4a4a8b, in 4.3 only) reverted
retiming keys to have one draw call per strip again. Which gets fairly
expensive if many strips have retiming keys.
However! All the strips are already drawn in two parts: all the regular
strips first, and then all the strips that are being "dragged over"
the other strips. So we can totally draw retiming keys for the whole
batch of strips at once, and it will just work out fine.
Viewing whole Sprite Fright Edit v135 timeline, on PC (Ryzen 5950X,
Win10/VS2022): draw_timeline_seq 7.4ms -> 5.8ms
(sequencer_retiming_keys_draw part 1.7ms -> 0.3ms)
Pull Request: https://projects.blender.org/blender/blender/pulls/128170
After 4bf08198a7 it it no longer necessary to flip gizmo bounds if they are not valid.
Note: it is still possible to create an invalid crop and "losing" the gizmo if it is dragged and dropped outside the compositor canvas. This will be addressed in a different patch.
Pull Request: https://projects.blender.org/blender/blender/pulls/127656
While we continue to improve our tooltips, this PR appends a period to
descriptions if it does not already end in punctuation (.!?), or does
not contain a bullet character. This only does so under the same
circumstances as in earlier code, removed in #125460.
Pull Request: https://projects.blender.org/blender/blender/pulls/128097
The material handling on curves was not super strong yet because there was not a
lot of need for it. However, now with Grease Pencil it's much more likely that
material selections are used on curves.
The patch is larger than one might expect at first, because we have to pass more
information into the field context in many places, because the materials are
stored on `Curves` and not `CurvesGeometry`.
Related to #128109.
Pull Request: https://projects.blender.org/blender/blender/pulls/128182
Immediate mode uses the old 'resource tracker' which has been replaced
by swap chain resource pools. This PR optimizes immediate mode buffers
by utilizing resource pools.
Pull Request: https://projects.blender.org/blender/blender/pulls/128188
This implements versioning code to go from legacy to layered action.
The versioning is only triggered when the experimental flag for
Multi-Slot actions is enabled.
All the actions are converted in place, which should be fine because
of backwards and forwards compatibility with layered actions.
Pull Request: https://projects.blender.org/blender/blender/pulls/127842
The issue has been introduced in aca27e5462.
Before that change the memory used for baking was cleared-alloced,
but now it relies on the default object initialization. This could
leave fields uninitialized.
Pull Request: https://projects.blender.org/blender/blender/pulls/128181
Anim: Show Action Slot type in panel + change type icon RNA wrapper name
This PR combines two changes, both necessary to properly show the
Action Slot ID type in the GUI:
1. Expose the slot's ID type as an RNA enum property `id_root`, in the
same way that the DNA `Action::idtype` is exposed as
`Action.id_root` in RNA. The only differences are that this
property is read-only on slots (it is supposed to be correct, and
if it's not, I'd prefer a bug report over people working around
this), and that the default value is named differently (`ANY` on
Actions, `UNSPECIFIED` on Slots). The latter is done to clarify
that this is not a situation that will last for long; as soon as
the slot is assigned to any ID, it will be locked to that ID type.
2. Rename the RNA property `ActionSlot.idtype_icon` to
`ActionSlot.id_root_icon` to be consistent with the above.
And of course the type is drawn in the Slot properties panel in the
Action editor. This now replaces the icon in the slot name (in the
same panel), so that it's clearer that that is the slot name, and not
an object/material/whatevertype name.
Previously we were using a bespoke hodgepodge of
`Action::is_action_legacy()` and `Action::is_action_layered()`,
sometimes in combination with checking for the Baklava feature flag,
when what we really meant is "Should this action be treated as legacy
or not?"
This commit changes the places where that's semantically what we meant
to use `action_treat_as_legacy()`. Some of those places were already
correct, using a compound conditional, but some of them weren't, and
thus were not always branching correctly. For those latter cases,
this commit is a bug fix.
Importantly, not all uses of bare `Action::is_action_legacy()` or
`Action::is_action_layered()` are semantically incorrect: there are many
places where that is the right thing to do. This commit takes care not
to touch those places.
Pull Request: https://projects.blender.org/blender/blender/pulls/128174
The additional division modes were implemented with float
division then rounding to match float > int conversion in
existing float to int node.
As pointed out in the bug report the precision here is limited.
This patch replaces the float division with integer math which
increases the precision to much higher numbers.
Divide Round by due to the way it is calculated has less precision
than Divide Floor and Divide Ceil.
Pull Request: https://projects.blender.org/blender/blender/pulls/128123
This improve the API in multiple aspects:
* No need for an additional `lookup` call to get the current attribute. This
would internally iterate over all attributes again. This leads to O(n^2)
behavior. Note that there are still other reasons for O(n^2) behavior when
processing attributes (where n is the number of attributes).
* Remove the need to return a value from the iteration code to indicate that the
iteration should continue. This is now the default behavior. The iteration can
still be stopped by calling `iter.stop()`.
* Easier access to `is_builtin` property.
* Iterator callback only has a single parameter instead of two (of which one is
sometimes unused).
Pull Request: https://projects.blender.org/blender/blender/pulls/128128
This bug was introduced in #126561. The issue was that the UI action
selectors no longer ensured that animdata existed before attempting
to assign an action, which would then cause the assignment to fail.
However, this was only the case in Blender builds without experimental
features, since the new Baklava code paths would otherwise be taken,
which *do* handle this correctly.
This fixes the issue by ensuring animdata in the non-experimental code
paths again.
Pull Request: https://projects.blender.org/blender/blender/pulls/128083
Currently the log only contained the first compatible device. It is
more important to the user to know which device is used.
This PR increases the level of the first compatible device so it is only
visible when increasing the log level. It reports the device, driver and
vendor when starting blender with `--debug-gpu`.
Pull Request: https://projects.blender.org/blender/blender/pulls/128168
Resource binding was over-complicated as I didn't understood the state
manager and vulkan to make the correct decisions at that time. This
refactor will remove a lot of the complexity and improves the performance.
**Performance**
The performance improvement is noticeable in complex grease pencil
scenes.
Grease pencil benchmark file picknick:
- `NVIDIA Quadro RTX 6000` 17 fps -> 24 fps
- `Intel(R) Arc(tm) A750 Graphics (DG2)` 6 -> 21 fps
**Bottle-neck**
The performance improvements originates from moving the update entry
point from state manager to shader interface. The previous implementation
(state manager) had to loop over all the bound resources and find in the
shader interface where it was located in the descriptor set. Ignoring
resources that were not used by the shader. But also making it hard
to determine if descriptor sets actually changed. Previous implementation
assumed descriptor sets always changed.
When descriptor set changed a new descriptor set needed to be allocated.
Most drivers this is a fast operation, but on Intel/Mesa this was measurable
slow. Using an allocation pool doesn't fit the Vulkan API as you are only
able to reuse when the layout matches exactly. Of course doable, but requires
another structure to keep track of the actual layouts.
**Solution**
By using the shader interface as entry point we can:
1. Keep track if there are any changes in the state manager. If not and the
layout is the same, the previous shader can be reused.
2. In stead of looping over each bound resource, we loop over bind points.
**Future extensions**
Bundle all descriptor set uploads just before use. This would be more
in line with how 'modern' Vulkan should be implemented. This PR already
separates the uploading from the updating and technically allows to upload
more than one descriptor set.
Instead of looking 1 set back we should measure if we can handle multiple
or keep track of the different layouts resources to improve the performance
even further.
Optional use `VK_KHR_descriptor_buffer` when available.
Pull Request: https://projects.blender.org/blender/blender/pulls/128068
Use the `foreach_action_slot_use(id, callback)` iterator function to find
all action/slot uses of an ID. That already handles the action constraint,
and thus solves the issue.
Pull Request: https://projects.blender.org/blender/blender/pulls/128090
Replaced "Convert stroke to **perimeter**" with "Convert stroke to **outline**"
in the outline tooltip description.
It was suggested that the description is not accurate and could be misleading
between the two terms. Thus, I simply just replaced the word across all tooltip
descriptions to keep it consistent and easily understood.
Resolves#126814.
Pull Request: https://projects.blender.org/blender/blender/pulls/127990
This adds an eyedropper button to properties where you can choose a bone.
Clicking it activates the modal operation and you can pick:
* Bones from the Outliner
* Bones from the 3D Viewport if the Armature is in Pose Mode or Edit Mode
## Limitation
Picking from the 3D viewport while in Object Mode doesn't work due to technical reasons.
The selection buffer is only filled with pose bones if the armature is in pose mode.
Using the picker while in object mode gives you the Armature object, not its bones.
So you cannot use the bone picker to constrain an object to a bone.
UNLESS you pin the panel with the object, go to pose mode and then pick.
There is a warning printed for those cases to tell the user why this hasn't worked.
Pull Request: https://projects.blender.org/blender/blender/pulls/121523
+0.5 should be added to integer pixel coordinate (sampling at pixel
center). This is only really visible in very specific scenarios, but
is more correct.
Previous fix (b17734598d) tried to fix this, but it seems that
depending on CPU and video width, ffmpeg might need at least 64
byte alignment, even if CPU we're running on is only AVX2 (32 byte
alignment). That is because some other parts of ffmpeg code
statically pick 64 or 32 byte alignment internally, depending on whether
AVX512 support is even compiled in (even if CPU might not have it).
I have checked whether this does not negatively affect a platform where
SIMD alignment is always 16 (Mac), and it does not, everything works as
expected.
Pull Request: https://projects.blender.org/blender/blender/pulls/128107
