Even if related, they don't have the same performance
impact.
To avoid any performance hit, we replace the Diffuse
by a Subsurface Closure for legacy EEVEE and
use the subsurface closure only where needed for
EEVEE-Next leveraging the random sampling.
This increases the compatibility with cycles that
doesn't modulate the radius of the subsurface anymore.
This change is only present in EEVEE-Next.
This commit changes the principled BSDF code so that
it is easier to follow the flow of data.
For legacy EEVEE, the SSS switch is moved to a
`radius == -1` check.
Update the command line help message to reflect the actual image output
formats available. Remove mention of IRIZ and DDS, rename MPEG to
FFMPEG. HDR and TIFF are always valid now.
Pull Request: https://projects.blender.org/blender/blender/pulls/115987
Before 111e586424, the normals were written back to the mesh
from the values stored in the undo step. Now though, due to caching of
normals and better const correct-ness, this is not so simple or safe.
That still may be a nice optimization to apply in the future, but for
now the simplicity of just recalculating them is much more feasible.
Since the recalculation is localized to the brush action, performance
should be okay. And maybe normals won't have to be stored in undo steps
in the future as well, since they're derived data.
`set_default_remaining_node_outputs` didn't work because the mapping
between the original node group sockets and the lazy function outputs
wasn't set up during the construction of the node type, as done by the
bake node and others.
Instead of storing a boolean "update tag" for every vertex, just
recalculate the normals of all the faces and vertices in affected PBVH
nodes. This avoids the overhead of tracking position updates at the
vertex level, and simplifies the normal calculation, since we can
just calculate values for all the normals in a node.
The main way we gain performance is avoiding building a `VectorSet`
for all vertices and faces that need updates in the entire mesh. That
process had to be single threaded, and was a large bottleneck when many
vertices were affected at a time.
That `VectorSet` was necessary for thread safety deduplication of work
though, because neighboring nodes can't calculate the normals of the
same faces or vertices at the same time. (Normals need to be calculated
for all faces connected to moved vertices and all vertices connected to
those faces). In this PR, we only build a set of the *boundary* faces
and vertices. We calculate those in a separate step, which duplicates
work from the non-boundary calculations, but at least it's threadsafe.
I measured normal recalculation timing when sculpting on a 16 million
vertex mesh. The removal of the `vert_bitmap` array also saves 16 MB
of memory.
| Nodes | Affected Vertices | Before (ms) | After (ms) |
| ----- | ------------ | ----------- | ---------- |
| 4 | 15625 | 0.208 | 0.304 |
| 35 | 136281 | 2.98 | 0.988 |
| 117 | 457156 | 15.0 | 3.21 |
| 2455 | 9583782 | 566 | 84.0 |
Pull Request: https://projects.blender.org/blender/blender/pulls/116209
Code inside `IMB_transform` (which is pretty much only used inside VSE
to do translation/rotation/scale of image or movie strips) was not
correctly doing mapping between pixel and texel spaces. This is similar
to e.g. GPU rasterization rules and has to do with whether some
coordinate refers to pixel/texel "corner" or "center" etc. It's a long
topic, but short summary would be:
- Coordinates refer to pixel/texel corner,
- Do sampling at pixel centers,
- Bilinear filter should use floor(x-0.5) and floor(x-0.5)+1 texels.
Also, there was a sign error introduced in Subsampling 3x3 filter, in
commit b3fd169259.
After making the PR, I found out that this seems to fix#90785, #112923
and possibly some others.
Long explanation with lots of images is in the PR.
Pull Request: https://projects.blender.org/blender/blender/pulls/116628
This exposes the internal `ID.session_uuid` in the Python API as `ID.session_uid`.
The exposed name is not `session_uuid`, because it's not actually universally unique,
and we want to change the internal name too.
The reason for exposing this is to allow Python scripts to call operators that take the
session id as input. A fair number of operators do this as you can see when searching
for `WM_operator_properties_id_lookup`.
Pull Request: https://projects.blender.org/blender/blender/pulls/116888
As discussed in #105407, it can be useful to support returning
a fallback value specified by the user instead of failing the driver
if a driver variable cannot resolve its RNA path. This especially
applies to context variables referencing custom properties, since
when the object with the driver is linked into another scene, the
custom property can easily not exist there.
This patch adds an optional fallback value setting to properties
based on RNA path (including ordinary Single Property variables
due to shared code and similarity). When enabled, RNA path lookup
failures (including invalid array index) cause the fallback value
to be used instead of marking the driver invalid.
A flag is added to track when this happens for UI use. It is
also exposed to python for lint type scripts.
When the fallback value is used, the input field containing
the property RNA path that failed to resolve is highlighted in red
(identically to the case without a fallback), and the driver
can be included in the With Errors filter of the Drivers editor.
However, the channel name is not underlined in red, because
the driver as a whole evaluates successfully.
Pull Request: https://projects.blender.org/blender/blender/pulls/110135
In order to prepare for introduction of fallbacks, refactor the
function to return its status as an enumeration value. Also, move
the index range check inside the function from python-specific code.
F-Curves with a broken rna path are highlighted with a red underline
in the channel list of the animation editors. I think it makes sense
to also apply this to drivers that failed to evaluate and were disabled.
Otherwise, it is not apparent which drivers are broken without checking
every one manually, or applying the errors filter.
`armature.collections.move(x, y)` now retains the active collection. Due to
the move, the active collection index can change. The index is now updated
to accomodate for this.
Ensure that the active bone collection doesn't change, when adding a new
bone collection via low-level functions (for example
`armature.collections.new()` in Python).
The change to the active bone collection happened because adding a new
child may change the index of the active bone collection.
The graph editor supports an Extend transformation mode, which is
essentially Move that only affects keyframes either before or
after the current frame, based on the mouse cursor position.
This fixes a bug in this mode where it doesn't respect the snap setting.
Converting a near-identity matrix to a euler could be off by over
4.5 degrees when the result of the hypotenuse calculated from the
matrix was small (around 2e-6).
Resolve by increasing the epsilon ~20x to 3.75e-05 which results
in ~1/20th the error.
This was tested against many generated near-identity matrices
(which tend to cause arithmetic error) to ensure this change doesn't
cause worse results in other cases (see report for details).
The GPU Directional Blur node does not match CPU. This is because GPU
accumulates the scale, while the CPU increments it. This patch
incremenets the scale for the GPU to make them match.
This commit fixes missing updates during playback when `frame_current`
property of the active scene is used.
Applied the same check for the depends-on-time as for explicitly
specified scene.
Pull Request: https://projects.blender.org/blender/blender/pulls/116886
Cancel drawing operation when layer is locked/hidden (similar to GPv2).
Otherwise `get_editable_drawing_at` returns nullptr which results in crash
and assert hit.
Pull Request: https://projects.blender.org/blender/blender/pulls/116779
This avoid the cost of creating the tiles themselves which uses a lot
texture write. This was a bottleneck on Apple GPUs.
Also the per pixel classification allows us to remove certain checks in
the deferred lighting shader making it faster.
### TODO
- [x] Add gl_FragStencilRefARB support on other backend
- [x] Add workaround for when gl_FragStencilRefARB isnt supported
Pull Request: https://projects.blender.org/blender/blender/pulls/116704
Reloading scripts lead to wmDropBox::ot pointing to freed operators.
Resolve by following the logic of key-map operator initialization
to ensure changes to registered operator triggers a refresh of the
operator pointer & properties.
Since wmDropBox are setup with the editors and aren't re-initialized
when the file is reloaded for e.g. it doesn't make sense to use the
current context when setting up their properties.
Note that this doesn't fix any user visible bugs and is more a issue of
correctness.
