When loading an unknown node type from a newer Blender version, the node
storage data cannot be properly loaded. Re-saving the file will then
crash if saving with the same node type idname, since new Blender
versions cannot find the expected storage data.
Loading in older versions should replace unknown node types with a dummy
"Undefined" node that will get loaded as `NodeTypeUndefined` in newer
versions as well. Custom node types are exempted from this since they
store all data as generic ID properties and can always be fully
serialized.
Pull Request: https://projects.blender.org/blender/blender/pulls/114803
Some left-over code from old refactoring/improvements of old addresses
remapping in library reading code was causing some wrong remapping
between different IDs.
The code was actually not doing what its comment was describing, since
it would cause a random remapping to a different new address than the
expected one:
Initial state, before removed buggy line:
`old_addr_1 -> new_addr_1`
`old_addr_2 -> new_addr_2`
End state, if `old_addr_2` == `new_addr_1`:
`old_addr_1 -> new_addr_2`
`old_addr_2 -> new_addr_2`
From the description of the removed line, that behavior was actually
already covered by adding the remapping rule in the line above, since
adding a remapping overwrite an existing one if needed, so the line
above would do (in case a placeholder ID existed before the actual ID
got read):
Initial state, before adding new remapping rule:
`old_addr_1 -> new_addr_placeholder_1`
End state, after adding remapping to newly read ID:
`old_addr_1 -> new_addr_1`
NOTE: Noticed thanks to rare failures of the liboverride tests on
buildbots. The issue was less than 1%-reproducible with a debug build
and ASAN, but had about 2-4% repducibility with release builds.
Pull Request: https://projects.blender.org/blender/blender/pulls/114784
Caused by 6faa39edb7
In that commit it was assumed that the view offset does not need to be
updated if the operator is `V3D_OP_MODE_ROTATE` instead of simply
checking `this->use_dyn_ofs`.
Since `use_dyn_ofs` is always `True` when using Auto Depth or Orbit
Arround Select, the offset should always be updated in these cases.
Pull Request: https://projects.blender.org/blender/blender/pulls/114726
The initial pipeline state was not set and could lead to undetected
state changes.
This fixes several issues, including:
* Fix incorrect colors in the UI, that would be used everywhere, and
could be fixed by restarting Blender.
* Overlay extra objects where drawn below the grid.
Pull Request: https://projects.blender.org/blender/blender/pulls/114792
This fixes a bug where the rgb UsdUVTexture shader output
attributes were created with incorrect type color3f when
connected to UsdPreviewSurface diffuseColor and emissiveColor
inputs. The UsdPreviewSurface specification requires the
UsdUVTexture rgb output to be of type float3.
This was happening because the UsdUVTexture outputs were created
implicitly by the call to UsdShadeInput::ConnectToSource(). I.e.,
because the diffuseColor and emissiveColor inputs are of type
color3f, this was the type assigned by default to the rgb source
attributes as well.
Now explicitly creating the UsdUVTexture shader output attributes
with the correct types.
Pull Request: https://projects.blender.org/blender/blender/pulls/114728
The compositor executes superfluously on scene changes when adding or
removing objects.
This was due to an explicit handling of a certain notifier, which was
introduced in commit 1e948b251d. The exact reasoning behind the
addition is unknown, so this patch removes it until we have an informed
decision to keep it.
Devices that don't support GPU_DEPTH_COMPONENT24 natively will use
GPU_DEPTH_COMPONENT32F in stead. When done the current code path raised
an error as the conversion type wasn't set.
Issue detected on AMD RX5700 device.
Pull Request: https://projects.blender.org/blender/blender/pulls/114786
It was weird concept which was calculating scene FPS, but from
the depgraph side. It was misleading, especially for the initial
evaluation of FPS of nan.
If we want FPS counters this needs to be done on a higher level
in the code.
Pull Request: https://projects.blender.org/blender/blender/pulls/114782
Caused by 6faa39edb7
In that commit it was assumed that the view offset does not need to be
updated if the operator is `V3D_OP_MODE_ROTATE` instead of simply
checking `this->use_dyn_ofs`.
Since `use_dyn_ofs` is always `True` when using Auto Depth or Orbit
Arround Select, the offset should always be updated in these cases.
Use general util to build mesh topology map instead of array of vectors.
This util can be optimized in future by better multithreading and new
algorithms with better CPU catch heuristics. For now this will provide
better usage of memory without a lot of small allocations.
Pull Request: https://projects.blender.org/blender/blender/pulls/114683
3 component texture formats are often not supported by vulkan devices.
Sometimes the support is less than 5%. The 4 components variants have
more than 90% support.
This PR builds on top of the existing vulkan data conversion to add
the ability to convert between RGB16F<=>RGBA16F and RGB32F<=>RGBA32F
texture formats.
This allows using color management other then Standard/sRGB. Most places
the 3 component texture formats are phased out, but OCIO, external
render engines and real time compositor uses them.
Pull Request: https://projects.blender.org/blender/blender/pulls/114708
Currently, atomic additions in vertex normal calculation introduce
non-determinism that can influence the result of other operations,
sometimes causing non-reproduceable renders (in cases like #100250
and #101726). This is because the order used during threading when
accumulating face normals changes.
This commit fixes that non-determinism, using a vertex to face topology
map to calculate vertex normals without atomics. When the map is already
available, this can be faster too.
In the longer term future, this method of vertex normal calculation
means it will be easier to do partial recalculations when only part
of a mesh changes. That might be essential for cases like transforming
small selections in a non-BMesh edit mode.
As an experiment I tried a "fast" code path that avoids weighting face
normals by the corresponding corner angle when averaging them to
create vertex normals. This significantly reduces the necessary
computation and memory bandwidth for vertex normal calculation.
The results are shown below too, but it's not part of this PR.
I measured the FPS with two smooth shaded 16 million face grids:
| | Before | After | After (fast) |
| -------------------------------- | ------ | ----- | ------------ |
| Created from scratch every frame | 0.96 | 0.91 | 0.92 |
| Deformed by geometry nodes | 0.99 | 1.32 | 1.40 |
Though many other things besides normals are being tested here,
it's clear that the performance difference isn't that large either way,
though there is an observable regression with meshes created from
scratch, and there is a noticeable improvement when the topology
stays the same and is persistent.
Pull Request: https://projects.blender.org/blender/blender/pulls/111718
When VSE cache is invalidated (on most VSE edit operation e.g.), or
Blender quits, in some cases (complex and very long edits), the
prefetch job would block several seconds after being requested to stop.
This was because one codepath would keep looping over all frames without
checking for the `stop` flag.
The problem is in the way of identifying "fast" intersections through bounds.
In the existing code, before testing the intersections (to identify
holes) the polys are sorted according to the bounds
(in the order x1 < x2 || y1 < y2).
Then a for loop is used on the order returned by sort.
Each time the bound of a polygon intersects with another, it is joined
and the bound is added.
The problem with this solution is that some bounds may not intersect
with the first, but could intersect one that is joined to the first,
which, as it is cleared, makes the intersection undetected.
The solution is to remove this code with `qsort` and create a
"target_map" that identifies a source polygon and a dest polygon.
Pull Request: https://projects.blender.org/blender/blender/pulls/114600
This PR solves an issue when using texture pools with textures that are not
natively supported by the device. In the previous implementation the
internal `Texture::format_` was changed. Any check if the texture could
be reused would fail as its format would be different that the requested one.
This PR fixes this by separating the requested format `Texture::format_` and
how it is stored on the GPU `VKTexture::device_format_`.
This solves the next artifacts:
* Workbench flickering artifacts on AMD/Intel GPUs
* Workbench TAA on AMD/Intel GPUs
* Overlays were not always drawn fully solid
Pull Request: https://projects.blender.org/blender/blender/pulls/114697
OpenGL uses a depth range between -1 and 1, which is then normalized.
Metal & Vulkan uses a depth range between 0 and 1, which is already normalized.
The final plan would be to default to a depth range between 0 and 1, but
for now the depth ranges are retargetted so they won't be clipped away.
This solves the next issues for users:
- Navigate control will be rendered correctly
- Ortographic view clipping artifacts
- EEVEE light evaluation
Retargetting happens at the end of the vertex stage or when a geometry
stage is present at the end of the geometry stage. Derivatives using
depth would have a different value compared to OpenGL, but would match
Metal backend. OpenGL performs clipping and generates derivatives based
on the original depth value.
`gl_FragCoord` and clipping would have some precision differences as clipping
and normalizing are done in a different order but would match Metal.
Geometry shaders should use `gpu_EmitVertex` to ensure that the retargetting
is done per vertex.
Pull Request: https://projects.blender.org/blender/blender/pulls/114669
No functional changes.
Extract code from the function `insert_keyframe_value`
to make it easier to use in other places.
I created a new function `remap_driver_frame` that
removes 3 parameters from the function.
And I removed the requirement for a `ReportList`
by just returning false and letting the caller
create a message.
Pull Request: https://projects.blender.org/blender/blender/pulls/114700
