Based on @fclem's suggestion in D6421, this commit implements support for
storing all tiles of a UDIM texture in a single 2D array texture on the GPU.
Previously, Eevee was binding one OpenGL texture per tile, quickly running
into hardware limits with nontrivial UDIM texture sets.
Workbench meanwhile had no UDIM support at all, as reusing the per-tile
approach would require splitting the mesh by tile as well as texture.
With this commit, both Workbench as well as Eevee now support huge numbers
of tiles, with the eventual limits being GPU memory and ultimately
GL_MAX_ARRAY_TEXTURE_LAYERS, which tends to be in the 1000s on modern GPUs.
Initially my plan was to have one array texture per unique size, but managing
the different textures and keeping everything consistent ended up being way
too complex.
Therefore, we now use a simpler version that allocates a texture that
is large enough to fit the largest tile and then packs all tiles into as many
layers as necessary.
As a result, each UDIM texture only binds two textures (one for the actual
images, one for metadata) regardless of how many tiles are used.
Note that this rolls back per-tile GPUTextures, meaning that we again have
per-Image GPUTextures like we did before the original UDIM commit,
but now with four instead of two types.
Reviewed By: fclem
Differential Revision: https://developer.blender.org/D6456
Seems that previous fix didn't work in all cases: Debian's build
environment didn't fully detect endianess, possibly due to typo,
possibly due to difference in various environments.
Using define magic from a more battle-tested project seems a safe
way to go.
There are more changes than just PPC since the upstream commit contains
full re-synchronization of all defines.
This commit updates numaapi to a latest library version from upstream.
rB4cde92303f3d made a MESH-only check (but that is no longer mandatory
since the Mesh is fetched by MOD_deform_mesh_eval_get instead of ctx-
>object->data now...). This would fail for projecting beveled curves
(resulting mesh missing), now also get the mesh for curves if we use
MOD_SHRINKWRAP_PROJECT (and only then).
Maniphest Tasks: T68489
Differential Revision: https://developer.blender.org/D6548
'Shift + S' is taken by toggeling Smooth Stroke in the keymap, should be
'Shift + Alt + S' (this also goes more in line with other editors, where
Alt is added to the shortcut used by 'Save' to do 'Save As')
Maniphest Tasks: T72803
Differential Revision: https://developer.blender.org/D6572
when using UDIMS
A more generic approach might be considered in the future (I assume
there are other operators around that need an update in shifting their
uvs) though.
Maniphest Tasks: T72648
Differential Revision: https://developer.blender.org/D6570
The Falloff property in the pose brush controls how much deformation
propagates through the ik chain, not by using the distance to the
vertices, so this option does not make sense (it does nothing in the
pose brush code and it disables the 3d paint cursor).
Reviewed By: jbakker
Maniphest Tasks: T73070
Differential Revision: https://developer.blender.org/D6574
When a node was partially/fully hidden, this was causing the mask flags
to update incorrectly because it was not checking all vertices, so they
were assigned the fully_masked state and not updating in the transform
tool and mesh filter.
Reviewed By: jbakker
Maniphest Tasks: T73094
Differential Revision: https://developer.blender.org/D6573
The PBVH usually is rebuild after a topology change, so it does not make
sense to keep the previous dynamic mesh preview vertex list. This may
cause a crash is the number of vertices of the new mesh (and preview) is
larger previous one. Now the list is deleted with the PBVH and a new one
will be generated using the new mesh when the cursor is updated.
Reviewed By: jbakker
Maniphest Tasks: T71712
Differential Revision: https://developer.blender.org/D6476
Commit baeb11826b switched memory
allocation for the motion transform to use CUDA directly, instead of going
through abstractions. But no CUDA context was set active before those
were called, so the calls failed. This fixes that by binding a context beforehand.
The `do_versions_after_linking_280()` function was missing a placeholder
for the newly added versioning code. I copied the comments from the
`blo_do_versions_280()` function.
No functional changes.
During development of Blender 2.80 the `Object.hide` property was removed,
and later reintroduced in rB5e968a996a53 as `Object.hide_viewport`. Of
course there are some technical details missing in this summary, but this
is the view that's given in the 2.80 release notes.
FCurves on `Object.hide` weren't updated in versioning code, resulting in
the property no longer being animated. This commit corrects the RNA path
of such FCurves.
When splitting a Quad View by dragging a corner an quad area can become
negative size because of a one pixel offset to calculate the start of
the area.
This patch solves this to make sure that there are no negative areas. By
adapting the `ARegion.winrct`.
Reviewed By: Severin
Differential Revision: https://developer.blender.org/D6579
The original code had 0 as a magic number in the test whether the weight
belongs to a locked group, instead of comparing it to the actual group
number.
Thanks @mano-wii for providing the diff.
The Viewport Render Keyframes operator didn't include keyframes from
grease pencil animation or from mesh shape key animation (see T72861).
To find all related datablocks that could be considered as part of the
selected object, `BKE_library_foreach_ID_link(..., IDWALK_RECURSE)` is
used. To prevent relations like constraint targets from being visited,
the recursion is stopped when it finds a different object.
Reviewed By: mont29
Differential Revision: https://developer.blender.org/D6558
Never treat one of those horrorible 'IDs that are not real IDs' as
regular ID, and expect ID management code to do so. Unless there is a
very good reason, one should never explicitely pass those fake IDs to ID
management code directly.
In that specific case, user count is sort of 'disabled' in libquery
code, because master collections are not in bmain (`LIB_TAG_NO_MAIN`).
When sculpting and doing a image render the workbench used the sculpt
batch what is created by a different GPU context and would not show up.
In debug builds an assert for this case is checked. I rechecked all
`use_sculpt_pbvh` that it also checked if it was rendering. Only the
workbench deferred didn't do this for the default render mode.
With this change the user can render a workbench render directly from
sculpt mode.
