This is to avoid use after free when the `GPUPass` gets compiled after the
original `GPUMaterial` used to create it was freed.
The issue was introduced by rBfa3bd17ae873
The subdivision is always recomputed on the CPU when displaying stats
if the mesh is animated which leads to bad performance.
This caches the subdivision topology counters from the draw code in the
mesh runtime and uses them for the viewport statistics.
Differential Revision: https://developer.blender.org/D14774
The crash is caused as the subdivision wrapper does not have loop
normals, which are generally computed at the end of the modifier stack
evaluation via `mesh_calc_modifier_final_normals`. (Note that they are
initially computed, but deleted by the subdivision wrapper creation.)
This records in the mesh runtime whether loop normals should have been
computed and computes them alongside the subdivision wrapper.
Differential Revision: https://developer.blender.org/D14489
There are a number of shaders, most notably grid_frag.glsl, which rely on default assignments to uniform values within shaders. This is not currently supported by the shader uniform push model implemented for the Metal backend, wherein uniform updates are pushed as a singular block of data. Any default assignment would become over-written.
As such, adding assignments of these default values in the high-level, to ensure the correct value is written for all APIs. This likely impacts Vulkan push-constants as well.
Authored by Apple: Michael Parkin-White
Ref T96261
Reviewed By: fclem
Maniphest Tasks: T96261
Differential Revision: https://developer.blender.org/D14555
This covers implementation of the GPUTexture abstraction for the Metal backend, with additional utility functionality as required.
Some components have been temporarily disabled pending dependencies on upcoming Metal backend components, and these will be addressed as the backend is fleshed out.
One core challenge addressed in the Metal backend is the requirement for read/update routines for textures. MTLBlitCommandEncoders offer a limited range of the full functionality provided by OpenGLs texture update and read functions such that a series of compute kernels have been implemented to provide advanced functionality such as data format conversion and partial/swizzled component updates.
This diff is provided in full, but if further division is required for purposes of code review, this can be done.
Authored by Apple: Michael Parkin-White
Ref T96261
Reviewed By: fclem
Maniphest Tasks: T96261
Differential Revision: https://developer.blender.org/D14543
When more than one, consecutive, subdivision modifier is used on a Mesh,
the last subsurf modifier is used for GPU subdivision even though it
might be disabled. This is because retrieving the last subsurf modifier
in the draw code did not check whether the modifier was disabled or not.
To fix this, the session UUID of the modifier which delegated evaluation
to the GPU code is cached and used in the draw to select the right subsurf
modifier.
Differential Revision: https://developer.blender.org/D14488
Adds a column to the right in the Library Overrides Hierarchies view
mode to toggle editability of library overrides.
Note that making a library override non-editable currently involves
clearing all overridden properties. This is an arguable design choice,
we should probably at least warn the user before doing this.
Part of T95802.
Reviewed by: Bastien Montagne
Differential Revision: https://developer.blender.org/D14653
The code to compare buttons from the previous to the current frame, to
see if they match (an thus should keep the same state) was quite
generic, and didn't allow much flexibility/customization. For some
cases this isn't enough, and a more specific comparison is needed. Say
if some buttons don't actually store comparable data themselves, only
via the button context. This was the case in D14653.
It was possible to set context pointers for buttons via the layout, but
not for buttons in places where the layout system wasn't used (where
buttons are placed manually). This is needed for buttons in the
Outliner, see D14653.
Goals:
* Better high level control over where devirtualization occurs. There is always
a trade-off between performance and compile-time/binary-size.
* Simplify using array devirtualization.
* Better performance for cases where devirtualization wasn't used before.
Many geometry nodes accept fields as inputs. Internally, that means that the
execution functions have to accept so called "virtual arrays" as inputs. Those
can be e.g. actual arrays, just single values, or lazily computed arrays.
Due to these different possible virtual arrays implementations, access to
individual elements is slower than it would be if everything was just a normal
array (access does through a virtual function call). For more complex execution
functions, this overhead does not matter, but for small functions (like a simple
addition) it very much does. The virtual function call also prevents the compiler
from doing some optimizations (e.g. loop unrolling and inserting simd instructions).
The solution is to "devirtualize" the virtual arrays for small functions where the
overhead is measurable. Essentially, the function is generated many times with
different array types as input. Then there is a run-time dispatch that calls the
best implementation. We have been doing devirtualization in e.g. math nodes
for a long time already. This patch just generalizes the concept and makes it
easier to control. It also makes it easier to investigate the different trade-offs
when it comes to devirtualization.
Nodes that we've optimized using devirtualization before didn't get a speedup.
However, a couple of nodes are using devirtualization now, that didn't before.
Those got a 2-4x speedup in common cases.
* Map Range
* Random Value
* Switch
* Combine XYZ
Differential Revision: https://developer.blender.org/D14628
The new name is `Internal Dependencies`.
* Is more inline with the general goal of the panel.
* Can also show which external objects are used in the node tree in the future.
This name was choosen in the geometry nodes submodule meeting (2022-04-25).
Differential Revision: https://developer.blender.org/D14752
This macro allows defining a default argument for when the translation
unit is compiled in C++. Otherwise (in C), the argument has to be passed
explicitly.
A couple of benefits:
* Default arguments are a nice quality-of-life feature in C++. It's
annoying if these can't be used in C++ files, just because the header
with the function declaration still needs to be C compatible.
* Adds useful information to the API declaration. E.g. that an argument
can be nullptr.
* Should help us to move to using default arguments more, helping
readability (arguably)
Used in D14653.
Reviewed By: LazyDodo
Differential Revision: https://developer.blender.org/D14654
This commit adds support for the curves object to the apply modifier
operator. A warning is added when the evaluated result of the modifier
doesn't contain any curves data.
Differential Revision: https://developer.blender.org/D14730
Motion paths can now be initialised to more sensible frame ranges,
rather than simply 1-250:
- Scene Frame Range
- Selected Keyframes
- All Keyframes
Reviewed By: sybren, looch, dfelinto, pablico
Maniphest Tasks: T93047
Differential Revision: https://developer.blender.org/D13687
Add missing check as the context override dict may have been copied
since it was assigned, also initialize the context manager with
PyType_Ready, while it didn't cause any errors - it's expected
that all types are initialized.
Failure to clip automatic-names meant named could end with a "." for e.g.
Error in [0] meant the clipped text was copied then
immediately overwritten.
[0]: 354e6b9c18
- De-duplicate txt_new_linen & txt_new_line.
- Don't accept NULL as input for txt_new_linen & txt_new_line
(callers can pass in an empty string instead).
- Avoid duplicate strlen calls in txt_new_linen.
- Use memcpy when the length of the string is known in txt_delete_sel &
BKE_text_load_ex.
`attribute_try_create` didn't understand that the vertex group
attribute already existed because it only looks for names in custom
data layers when the domain matches. Using `attribute_exists`
unfortunately requires another loop through all attribute names,
but that should be optimized separately in the future anyway.
Differential Revision: https://developer.blender.org/D14756
Added a fallback path to compute the
cursor radius for when the stroke
starts over a blank area of space
(in which case SCULPT_cursor_geometry_update
fails).
This commit makes the `MeshVertex.normal` property read-only.
In practice it already was, since the value could be overwritten at any
time if code requests updated normals, so this just clarifies that
position and avoids misleading users.
Differential Revision: https://developer.blender.org/D14696
auto-iteration property in mask filter
Note: Auto-iteration is still set manually
for increase/decrease contrast. These should
probably become their own operators.
by showing redo panel.
The A hotkey has "auto iteration" enabled by default,
which calculates the number of times to run the filter
using a heuristic based on vertex count.
To make clear to the user what is going on the redo
panel is now shown for the mask filter operator.
NOTE: I discovered the source of the bug where sculpt
operators' redo panels were greyed out. The name
fed to SCULPT_undo_push_begin must match the operator
name. I've added a comment in sculpt_intern explaining
this.
Remembering the number of curves of every type makes it fast to know
whether processing specific to a single curve type has to be done.
This information was accessed in quite a few places, so this should be
an overall reduction in overhead for the new curves type.
The cache is computed eagerly, in other words every time after changing
the curve types. In order to reduce verbosity I added helper functions
for some common ways to set the types.
Differential Revision: https://developer.blender.org/D14732
