- Rename `find_elem` to `elem_offset` (matching `elem_exists`).
- Remove unused `SDNA_StructMember` return argument.
- Return an offset instead of a pointer which was being converted
back into an offset by one caller,
in this case there was no way to tell the difference between
and element that doesn't exist and a struct member
at the start of the array.
Resolves UBSan warning raised in T81340.
It looks like this code was left over from tabbed panels in the
properties editor. It wasn't used anywhere except for in one line of
the horizontally-aligned panel code that was recently removed.
Differential Revision: https://developer.blender.org/D8651
Running Blender on Ampere cards was already possible with ptx, this fix is
needed to support building CUDA binaries.
Note the CUDA version used for official Blender builds is still 10, this is
merely the change to make it possible for those using CUDA 11 and specifying
the sm_8x kernels to be compiled.
Found by Milan Jaros.
This adds the option to either smooth the entire boundary, or to keep
corners sharp, for the Subdivision Surface and Multiresolution modifiers.
This mainly helps with compatibility with other software. The default
behavior remains to smooth the entire boundary.
Differential Revision: https://developer.blender.org/D8485
This makes subdivision surfaces compatible with the old subdivision
surface modifier and other applications that do not use the limit surface.
This option is available on the Subdivision Surface modifier.
Differential Revision: https://developer.blender.org/D8413
On user level this fixes dead-lock of OpenCL render on Intel Iris GPUs.
Note that this patch does not include change in the logic which allows
or disallows OpenCL platforms to be used, that will happen after the
kernel fix is known to be fine for the currently officially supported
platforms.
The dead-lock was caused by wrong usage of memory barriers: as per the
OpenCL specification the barrier is to be executed by the entire work
group. This means, that the following code is invalid:
void foo() {
if (some_condition) {
return;
}
barrier(CLK_LOCAL_MEM_FENCE);
}
void bar() {
foo();
}
The Cycles code was mentioning this as an invalid code on CPU, while in
fact this is invalid as per specification. From the implementation side
this change removes the ifdefs around the CPU-only barrier logic, and
is implementing similar logic in the shader setup kernel.
Tested on NUC8i7HVK NUC.
The root cause of the dead-lock was identified by Max Dmitrichenko.
There is no measurable difference in performance of currently supported
OpenCL platforms.
Differential Revision: https://developer.blender.org/D9039
Implemented Brown-Conrady lens distortion model with 4 radial and
2 tangential coefficients to improve compatibility with other software,
such as Agisoft Photoscan/Metashapes, 3DF Zephir, RealityCapture,
Bentley ContextCapture, Alisevision Meshroom(opensource).
Also older programs: Bundler, CPMVS.
In general terms, most photogrammetric software.
The new model is available under the distortion model menu in Lens
settings.
For tests and demos check the original patch.
Reviewed By: sergey
Differential Revision: https://developer.blender.org/D9037
The "UI_panel_set_expand_from_list_data" doesn't need to be in the
public API since it's just called every time an instanced panel is added.
This commit just sets the expansion automatically and adjusts some
naming to account for the moved function.
Before this change, when users switch from edit mode to sculpt mode, the
entire mesh would be visible. Even if in the edit mesh mode part of it was
set to invisible.
With this change the visibility is preserved, by creating a separate face set
for the visible and invisible parts of the mesh and setting their initial visibility.
Implementation details: This adds a function to initialize a new Face Set
datalayer taking the current mesh visibility into account which is stored
in the ME_HIDE flag of the vertices.
Reviewed By: sergey
Differential Revision: https://developer.blender.org/D8901
This tool projects all vertices to the right of the plane defined by the
line gesture towards the plane. By doing this, this tool can create cuts
and plane surfaces in the mesh without modifying the geometry or using
boolean operations, so it is much faster than bisecting the mesh for
cases where the geometry was going to be remeshed afterwards.
Added as experimental as it does not have icon.
Reviewed By: sergey, Severin
Differential Revision: https://developer.blender.org/D9021
`SOCKET_OFFSETOF` was added in the initial commit {rBec51175f1fd6c91d5}
when `offsetof` [1] was not supported well enough. GCC and LLVM
support it since C++17.
Other two changes: type and size check can be done without creating
an invalid address too.
[1] https://cppreference.com/w/cpp/types/offsetof
Reviewed By: campbellbarton, brecht
Maniphest Tasks: T81100
Differential Revision: https://developer.blender.org/D9042
This took more than a day to fully investigate and understand, one of
the reasons being that the probability of the issue to show up was
extremely low, and subjected to very specific random factors.
Root of the issue is that, in some very rare cases, a newly read ID
might get the exact same memory address as the one it had when the blend
file was saved.
In that case, `BKE_workspace_active_set` would return immediately, since
the pointer to the active workspace would remain unchanged. But that
lead to having an unset NULL active layout pointer, which would crash
when attempting to get e.g. the active screen.
For the record, I ran into this when running a specific build (master
with one flag added to the `LIB_ID_CREATE` ones, with value `1 << 3`),
using a specific set of options (`--background --factory-startup -noaudio`),
and passing the .blend file from T80090 as argument.
Remove the attempt to update the active layout pointers of each window
from whithin `direct_link_workspace`.
This piece of code was a nonsense for at least to critical reasons:
* Do not, never, ever, access data from another datablock within the
direct_link_... functions. Just don't. Don't try to be smart.
* Since it was trying (and failing) to update the active layout of every
window for every workspace, it was effectively setting those
`act_layout` pointers to NULL (remapping can only ever happen once,
trying to remap and already remapped new pointer is bound to fail in
any case).
Luckily (and funnily), this piece of code was actually harmless, since
setting the active layout would be overridden/redone later, in
`lib_link_windowmanager`, when updating their `workspace_hook` in
`lib_link_workspace_instance_hook`.
Note that the similar horror with `WorkSpaceDataRelation->parent` (which
points at a window) is kept for now, because that one is needed. Hope to
refactor it soon though.
Although I haven't seen this cause any visible errors, there is some
incorrect handling for setting panel expansion during search:
- Properly check if child panel is active first
- Don't stop traversal at headerless panels that could theoretically
have children
Add an option to disable Alembic vertex interpolation.
Bump subversion from 5 to 6.
Alembic stores mesh samples at specific time keys; when a frame in
Blender maps to a timecode between two samples, Blender will interpolate
the mesh vertex positions. This interpolation only happens when the mesh
has a constant topology, but sometimes this was not detected properly
when the vertices change order, but the number of mesh elements remains
the same. This would result in a mesh with jumbled up vertices (T71981).
With this patch, users have the ability to disable vertex interpolation.
An alternative would be to have better detection of topology changes,
but that that'll cause a considerable slowdown.
Maniphest Tasks: T71981
Differential Revision: https://developer.blender.org/D9041
This modifier can only be added to Volume objects. It takes a mesh
as input and generates a "density" grid near the surface or in
the enclosed volume.
Ref T73201.
Reviewers: brecht
Differential Revision: https://developer.blender.org/D9032
Previously, one could only select a volume object in the outliner
or by clicking on the object origin. This patch allows you to click
on the actual volume.
Furthermore, the generated (invisible) mesh that is used for
selection is also used to draw an outline for the volume object now.
Reviewers: brecht
Differential Revision: https://developer.blender.org/D9022
When loading large files that are more than a couple weeks old
(such that DNA has changed in that time), a significant amount of
time is spent in `DNA_struct_reconstruct`. This function takes a struct
in the old layout and creates a struct in the new layout from it.
This was slow because it was computing the diff between the struct
layouts every time a struct is updated. Now the steps for the struct
reconstruction is computed only once per struct. This information is
then used to actually reconstruct all structs that changed.
I measured about 10-20% speedup when loading Spring files.
E.g. `10.6s -> 8.7s` for `06_055_A.anim.blend` in BKE_blendfile_read`.
This percentage varies a lot based on the number of blocks that have
to be reconstructed and how much DNA has changed since they have
been written. In none of my tests was the new code slower than the
old code.
Reviewers: campbellbarton
Differential Revision: https://developer.blender.org/D8959
The data layout remains exactly the same..
This change just gives all the elements in `SDNA->structs` names,
making it more comfortable to work with the data.
Reviewers: campbellbarton
Differential Revision: https://developer.blender.org/D8926
The behaviour is a bit different from other find modules:
`FFMPEG_FIND_COMPONENTS` should contain the
list of required libraries before calling `find_package(FFmpeg)`.
If not set, a default set of most common libraries is searched for:
`avcodec avdevice avfilter avformat avutil`
http://ffmpeg.org/download.html
Ref D8855
Reviewed By: brecht
Differential Revision: https://developer.blender.org/D8918
Add a dependency graph relation Force Object Animation → Scene Rigid
Body World Rebuild. This ensures that the rigid body world is rebuilt
when a force object is re-tagged for animation updates.
The extra relation doesn't add any new calculations when the animation
is running, as the Time Source node already had a relation to the
scene's `RIGIDBODY_REBUILD` node.
The relation is created directly to the `RIGIDBODY_REBUILD` Operation. I
would have liked to target the containing Component instead. However,
that has the `RIGIDBODY_SIM` operation as entry node, which isn't enough
to actually fix T80121.
Reviewers: Sergey
Differential Revision: https://developer.blender.org/T80121
Draw dashed lines for object child hierarchies in the outliner. This
makes the difference between collection and object hierarchies more
clear. Dashes are also consistent with viewport relationship lines.
Differential Revision: https://developer.blender.org/D8997
