This enables most of the GPU compiler's optimizations while -ffast-math
isn't set at DPC++ level.
It brings an overall 1% speedup and currently doesn't change the unit
tests pass rate.
When attempting to load a blend file that is not found or otherwise
invalid we are currently showing an (ugly) popup menu report and then
clearing the item from the Recent Files list. Clearing the list could
be unwanted in the case of files that are temporarily offline. And we
now have the ability to only clear "Items not found" using the "Clear
Recent Files List" operator. This PR removes that clearing.
Pull Request: https://projects.blender.org/blender/blender/pulls/127882
The Scale node in the Scale To Render size mode is not precise in case
of the use of render size percentage. This is due to floating point
imprecisions and we mitigate it by rounding instead of flooring after
multiplying by the percentage.
The Kuwahara node produces NaNs if the pixels have a very high local
standard deviation and sharpness is also high. This is because the
weighted sum of the Kuwahara sectors can have a zero total weight, which
causes zero division. To fix this, we return the original color if the
total weight is zero.
Pull Request: https://projects.blender.org/blender/blender/pulls/128378
This changes the name from `For Each Geometry Element` to just `For Each
Element`. While we might have more foreach zones in the future, it's not clear
yet whether the longer name will be necessary to differentiate the zones. So
better just use the better looking name for now.
Pull Request: https://projects.blender.org/blender/blender/pulls/128384
Adds `BKE_grease_pencil_copy_layer_parameters` and
`BKE_grease_pencil_copy_layer_group_parameters` to copy
the parameters of layers and layer groups without copying
the data.
We'd first set the brush, then change tools, which would again set the
brush to what it was before (switching tools tries to remember its last
used brush). Instead, switch the brush after switching the tool.
Also added a brush API check to test if the brush is valid for the
current mode, so that we don't end up switching tools in that case. This
matches previous behavior where this would be checked as part of setting
the brush already, causing the function to early exit before the tool is
changed.
* Match name with the extension ID.
* More strict double comment for optional parameters.
* Comment out tag since it is optional.
* More complete explanation for the paths_exclude_pattern.
This enables three additional math optimizations:
-ffp-contract=fast (enables FMA generation)
-freciprocal-math (enables x/y -> x*(1/y))
-fassociative-math (enables e.g. a*b + c*b -> (a+c)*b)
These are used on Windows and HIP anyways, so our code can't expect exact IEEE
semantics in any case.
The only difference between the new set and -ffast-math is that we don't use
-ffinite-math-only since this causes issues with the BVH (see ce1f2e271d) and
breaks e.g. isnan.
This causes a ~1.5% speedup in my very quick test, but might be higher for some
more math-intensive cases.
Pull Request: https://projects.blender.org/blender/blender/pulls/128342
The only difference between Windows+Clang and the others is a prefix, so use
some CMake logic to just prepend that to all flags instead of duplicating them.
Pull Request: https://projects.blender.org/blender/blender/pulls/128342
The Alpha Over node has an unexpected output size in GPU mode. If the
Factor and first Image inputs were connected to the same output, the
size of the output will assume the size of the second Image input, even
though the first Image input has a higher domain priority.
That's because the Alpha Over node is compiled into a Shader Operation
which has a single operation input for all inputs connected to the same
output, but only the domain priority of the first input was considered,
so in the case of the Alpha Over node, the domain priority of the Factor
input was assumed, ignoring the priority of the first Image output.
To fix this, we consider the priority of all inputs connected to the
same output, and assume the highest priority of all.
Pull Request: https://projects.blender.org/blender/blender/pulls/128366
The problem with the Integer Math node is that it is less flexible in common
cases because the other input has to be an integer too. There is no simple way
to switch the node to be a Float Math node (yet), so better insert that
directly. Using the Integer Math node is a more explicit choice now.
Searching for e.g. `iadd` prioritizes the Integer Math node thanks to prefix
search.
Pull Request: https://projects.blender.org/blender/blender/pulls/128370
This solves an issue when both linear gizmos happen to be at the same position.
In this case the gizmos became fairly unusable because they overlapped a lot.
They can still overlap of course, but it's more likely that they start at the
same position than that they end at the same position.
Note, when using a transform gizmo, the collision between the scale and
translate gizmo is resolved already. The issue only happens when two independent
Linear Gizmo nodes are used.
Pull Request: https://projects.blender.org/blender/blender/pulls/128371
Calling `action.fcurves.clear()` would clear the F-Curves, but didn't
update the F-Curve groups. This meant that the groups data still had their
original length & offsets into the F-Curves array, causing subsequent
F-Curve creation to crash Blender.
The unit test for this also covers the previous commit.
Pull Request: https://projects.blender.org/blender/blender/pulls/128375
Fix an issue where Python's `rna_struct.keyframe_insert(path, index=-1)`
would not create any keys. This was caused by the new layered animation
code not taking -1 as an 'all array elements' wild-card.
The accompanying unit test will follow in another commit, as its success
depends on another bugfix as well.
This was broken by 8027720e51 because
we forget to update the label in the declaration.
This is now reverted for the Mesh Line node, because there is no way to
change the label in the declaration yet. A new patch can be created that properly
updates the Mesh Line node to use the declaration.
It is possible to have IK solver without actual bones. There was a relation
created for this case, but it was originating from a wrong operation: it
was possible to run into situation when init is not yet done, but cleanup
is already run.
Fix the relation which ensures the order to go from IK initialization and
not the pose initialization. This should ensure proper cleanup order.
There is a relation from pose initialization to IK initialization so this
change should not cause any missing relations.
This should fix the intermittent failure of blendfile_versioning_1_over_8.
Pull Request: https://projects.blender.org/blender/blender/pulls/128267
* Handle the case where `drawing_array_num` is 0.
* Reset the pointer of the active node (because it is no longer valid)
* Cleanup copy-paste error in docstring
Blender crashes when viewing invalid images in the compositor. That's
because cached images will not be null, but will be unallocated, so to
fix this, we check allocation status as well as nullptr in the image
node.
