This adds a new `Map.lookup_try` method that returns a `std::optional<Value>`
for a given key. If the key is not in the map, `std::nullopt` is returned. If it
is in the map, then a copy of the value is returned. Note that the copy is
necessary because `std::optional` can't contain reference types.
This method helps a lot in #132219 to reduce boilerplate.
Pull Request: https://projects.blender.org/blender/blender/pulls/132278
This optimizes the move-constructor for `blender::Vector` when all of the
following are true:
* The source and destination vector have exactly the same type.
* The stored type is trivial.
* The inline buffer is `<= 32` bytes large (this value is a heuristic that could
be changed).
The basic idea of the optimization is that under these circumstances one can
just copy the entire inline-buffer over instead of only copying it partially
based on the vector size. While that can mean that more bytes have to be copied,
the machine code that does the copying can be more efficient due to less
branching and the hardcoded size.
The performance impact is quite measurable. Note that the speedup depends on how
many elements are in vector and thus how many elements of the inline buffer are
used. The following table shows the move construction performance of a
`Vector<void *, 4>`. Starting at 5 elements, the performance doesn't change much
anymore, because the inline buffer is just ignored.
| Elements | Old | New |
|----------|------|------|
| 0 | 20.3 | 14.6 |
| 1 | 22.7 | 21.5 |
| 2 | 36.4 | 21.6 |
| 3 | 36.4 | 21.5 |
| 4 | 36.5 | 21.6 |
| 5 | 21.4 | 21.1 |
| 6 | 21.3 | 21.1 |
| 7 | 21.4 | 21.1 |
| 8 | 21.5 | 21.0 |
| 9 | 21.4 | 20.9 |
| 10 | 21.3 | 20.9 |
The binary size stays effectively unchanged (< 2kb change).
Pull Request: https://projects.blender.org/blender/blender/pulls/131841
Caused by 024d7d12e2.
I forgot that the BVH cache value isn't reset to default if it isn't
shared, which can happen after remeshing in some cases. To fix this,
switch to unique_ptr which handles ownership properly on assignment,
removing this class of bug from this code.
Pull Request: https://projects.blender.org/blender/blender/pulls/132040
Previously, calling `clear()` on `Map`, `Set` or `VectorSet` would remove all
elements but did not free the already allocated capacity. This is fine in most
cases, but has very bad and non-obvious worst-case behavior as can be seen in
#131793. The issue is that having a huge hash table with only very few elements
is inefficient when having to iterate over it (e.g. when clearing).
There used to be a `clear_and_shrink()` method to avoid this worst-case
behavior. However, it's not obvious that this should be used to improve
performance.
This patch changes the behavior of `clear` to what `clear_and_shrink` did before
to avoid accidentally running in worst-case behavior. The old behavior is still
available with the name `clear_and_keep_capacity`. This is more efficient if
it's known that the hash-table is filled with approximately the same number of
elements or more again.
The main annoying aspect from an API perspective is that for `Vector`, the
default behavior of `clear` is and should stay to not free the memory. `Vector`
does not have the same worst-case behavior when there is a lot of unused
capacity (besides taking up memory), because the extra memory is never looked
at. `std::vector::clear` also does not free the memory, so that's the expected
behavior. While this patch introduces an inconsistency between `Vector` and
`Map/Set/VectorSet` with regards to freeing memory, it makes them more
consistent in that `clear` is the better default when reusing the data-structure
repeatedly.
I went over existing uses of `clear` to see if any of them should be changed to
`clear_and_keep_capacity`. None of them seemed to really benefit from that or
showed that it was impossible to get into the worst-case scenario. Therefore,
this patch slightly changes the behavior of these calls (only performance wise,
semantics are exactly the same).
Pull Request: https://projects.blender.org/blender/blender/pulls/131852
`OtherInlineBufferCapacity <= InlineBufferCapacity` implies
`size <= InlineBufferCapacity` and can be done at compile time.
So if that is true already (as it commonly is), the `else` branch
is not necessary. This can improve performance in theory, although
I couldn't measure that in practice.
Additionally, this change also reduces the binary size (because of the
omitted else branch). Since this code is inlined fairly often, the impact
is measurable: the binary size is reduced by 50kb.
This enables moving elements form one vector to another.
Usually this is being doing by extending a vector with the content
from the secondary vector and then clearing the secondary vector.
However sometimes this being performed to transfer ownership of managed elements,
if elements are copied from the secondary vector, but not cleared, this could
lead to 2 vectors to share ownership of objects.
Pull Request: https://projects.blender.org/blender/blender/pulls/131560
In Blender 4.2 in Grease Pencil draw mode it was possible to
erase strokes using box and lasso gestures.
Under the hood, these were just using the selection operators
that had special deletion handling if the object was in draw mode.
Rather than hacking this into the selection operations, this adds
two new operators:
* `grease_pencil.erase_lasso`
* `grease_pencil.erase_box`
When using one of the erase operations with auto-key, the previous
keyframe will be duplicated to the current frame (for the drawings
that are affected by the eraser).
They are mapped to the same shortcuts than the selection
operators in Blender 4.2.
* Lasso erase: `Ctrl`+`Alt`+`RMB`.
* Box erase: `B`.
This is part of #130518.
Pull Request: https://projects.blender.org/blender/blender/pulls/131504
This patch adds an option to use the new perlin noise as a noise function for the FCurve modifier.
That fixes the issue of the old noise function which could produce values outside the -0.5/0.5 range (For a strength of 1).
The old way is still preserved and will be automatically used for old files.
Because of the different noise function, I've added two more parameters to the noise modifier.
The default values for those parameters are chosen to be as close to the original noise modifier as possible.
However the match is not 1:1
## Depth
With the new noise, the parameters for "Lacunarity" and "Roughness" are available to the user.
By default they are set in such a way that changing the "Depth" results in a similar noise pattern.
## Offset and Scale
The offset and scale parameters now behave differently. Before they would also shift the phase in some way, now all they do is shift or scale the noise.
## Amplitude
The biggest difference is in the amplitude of the noise.
Overall the amplitude is a lot smaller in general than with the legacy noise, but it does fix the issue of the amplitude reaching outside the -0.5/0.5 range.
By shifting the phase I was able to find points where the amplitude reaches 0.5 but never above that.
Pull Request: https://projects.blender.org/blender/blender/pulls/129535
These functions showed up in a profile of a geometry nodes setup
using BVH sampling. They are so simple that function call overhead
is likely to be significant, and giving the compiler a view of what they
do is also likely to be helpful.
The BLI bicubic interpolation function with border boundary is wrong
around edges. It returns zero, where interpolated values should exist.
This is due to a wrong early exit condition, where only a 2x2 window is
assumed instead of the 4x4 window considered by bicubic interpolation.
To fix this, assume a 2 pixel region in the early exit condition for
both directions.
Pull Request: https://projects.blender.org/blender/blender/pulls/131320
The BLI wrapmode variant of nearest interpolation for border boundary is
wrong for coordinates between [-1, 0]. That's because the negative
comparison in wrap_coord is done on the integer rounded coordinates,
which will be zero in this case, so the condition will not fail. To fix
this, do the comparison on the original coordinates instead.
Pull Request: https://projects.blender.org/blender/blender/pulls/131306
This applies the fix from 98334b8f7d to the c++ implementation.
Two notes:
* We need to work towards unifying these implementations which would have avoided this bug.
* In the C++ implementation, one can't use `math::dot` and `math::normalize` for quaternions
where this function is implemented, because of include dependency order.
Both these things should be resolved, but right now correctness has priority for me.
Pull Request: https://projects.blender.org/blender/blender/pulls/131296
`StringRefNull ` expects a null terminated string and `StringRefNull (const char *)` initializer
already realizes a runtime assert to avoid `nullptr` c strings, however this stills allows
initializing `StringRefNull ` with a explicit `nullptr`.
Deleting `StringRefNull(std::nullptr_t)` gives a error if `StringRefNull` is initialized with a implicit `nullptr`.
This follows the same deleted `std::string(std::nullptr_t)` initializer in c++23.
Pull Request: https://projects.blender.org/blender/blender/pulls/130935
Add `bool MutableSpan<T>::contains(const T &value)` function, which is an
exact copy of the function from the `Span<T>` class. This makes it possible
to call `.contains()` directly on the `MutableSpan`, instead of having to
call `.as_span()` first.
Note that the `contains_ptr()` function was already copied from `Span` to
`MutableSpan` before.
No functional changes to the affected code, just the removal of the now
no longer necessary `.to_span()` call.
Pull Request: https://projects.blender.org/blender/blender/pulls/131149
Remove assert statement from make_update.py, making it ready for any
architecture.
Add riscv 32, 64 and 128 bit cpu architecture with little/big endian
to Blender's BLI_build_config.h and Libmv's build_config.h
Tested (to compile) on riscv64 little endian machine.
Pull Request: https://projects.blender.org/blender/blender/pulls/130920
Implements #130836:
interpolate_*_wrapmode_fl take InterpWrapMode wrap_u and wrap_v arguments.
U and V coordinate axes can have different wrap modes: clamp/extend,
border/zero, wrap/repeat.
Note that this removes inconsistency where cubic interpolation was
returning zero for samples completely outside the image, but all other
functions were not, and the behavior was not matching the function
documentation either.
Use the new functions in the new compositor CPU backend.
Possible performance impact for other places (e.g. VSE): measured on
4K resolution, transformed (scaled and rotated) 4K EXR image:
- Nearest filter: no change,
- Bilinear filter: no change,
- Cubic BSpline filter: slight performance decrease, IMB_transform
19.5 -> 20.7 ms (Ryzen 5950X, VS2022). Feels acceptable.
Pull Request: https://projects.blender.org/blender/blender/pulls/130893
Add a `short2` attribute type, intended to store the existing
`CD_CUSTOMLOOPNORMAL` custom data type in order to move the custom
normal storage to generic attributes. This type probably won't get much
use besides that, but generally we don't have reasons not to add these
generic types, besides binary size. In the future we should consolidate
usage of `convert_to_static_type` to avoid large binary size increases
(about 700KB here) for this sort of addition.
This is the first step of #130484.
Pull Request: https://projects.blender.org/blender/blender/pulls/130530
This adds support for constructing a `Map` directly from key-value-pairs.
If the same key exists multiple times, only the first one is taken into account.
Note: the keys and values are copied into the map (not moved). This is because
an `std::initializer_list` is read-only. If move-behavior is needed, it is better to
use the existing `map.add*` functions.
```cpp
Map<int, std::string> map = {{1, "where"}, {3, "when"}, {5, "why"}};
````
Pull Request: https://projects.blender.org/blender/blender/pulls/130470
Cleanup to simplify code by using common terms like _first_ and _last_ in context
of predicate applying in a range just like we do for spans and range containers.
Pull Request: https://projects.blender.org/blender/blender/pulls/130380
Doing something like `for (auto key : get_map().keys())` does not work, because the lifetime
of the `Map` returned by `get_map()` is not automatically extended for the duration of
the loop. This could cause crashes at run-time.
Now this pattern will not compile anymore. Note that there are possible valid use-cases
of this iterating over a `Map` that won't compile anymore either. This is fairly uncommon
though, and just storing the map in a separate temporary variable is a simple fix.
This would have been a valid usage for example: `do_something(get_map().keys())`.
This would have prevented e.g. 9840dd233c.
Pull Request: https://projects.blender.org/blender/blender/pulls/129880
This implements the proposal from #124512. For that it contains the following
changes:
* Remove the global override of `new`/`delete` when `WITH_CXX_GUARDEDALLOC` was
enabled.
* Always use `MEM_CXX_CLASS_ALLOC_FUNCS` where it is currently used. This used
to be guarded by `WITH_CXX_GUARDEDALLOC` in some but not all cases. This means
that a few classes which didn't use our guarded allocator by default before,
are now using it.
Pull Request: https://projects.blender.org/blender/blender/pulls/130181
The `_EXCEPTION_POINTERS` structure is valid only within the context of
the `UnhandledExceptionFilter` function.
After the function exits, the memory referenced by the pointer might no
longer be valid or the exception information.
The solution for this was to create `BLI_system_backtrace_with_os_info`
and passing a system-specific data as the second argument.
`BLI_system_backtrace` calls `BLI_system_backtrace_with_os_info` with a
null second argument internally.
Pull Request: https://projects.blender.org/blender/blender/pulls/129999
Catch any exception from the JSON parser and handle it by returning a
null value from the deserialization function. Let the asset indexer
regenerate the index file to handle the error.
At least for the asset index case we don't care about the exact parsing
error, so this simple error handling strategy is fine. Should more
precise error reporting be necessary for other use-cases this can be
added still, but I think these errors are usually a bit too low level to
expose to users.
Pull Request: https://projects.blender.org/blender/blender/pulls/129879
