The `OUTLINER_OT_item_activate` operator, although it detects when
something changes, always returns `OPERATOR_FINISHED` and thus induces
the creation of undo steps.
So return `OPERATOR_CANCELLED` when nothing changes.
Ref T94080
Reviewed By: Severin
Maniphest Tasks: T94080
Differential Revision: https://developer.blender.org/D13638
- Fix assert on size.
- Fix void * casting.
- Pass extent by values.
- Add swap function to avoid letting the types copyable.
- Add back the GPUTexture * operator on TextureFromPool.
Does two main changes:
* Handle regions in the order as visible on screen. Practically this
just means handling overlapping regions before non-overlapping ones.
* Don't handle any other regions after having found one containing the
mouse pointer.
Fixes: T94016, T91538, T91579, T71899 (and a whole bunch of duplicates)
Addresses: T92364
Differential Revision: https://developer.blender.org/D13539
Reviewed by: Campbell Barton
Asserts that such events actually always lead to a handler return value
that actually keeps the event passing.
Reviewed by Campbell Barton as part of
https://developer.blender.org/D13539.
Though the edge vertices aren't really meant to have an order,
it can make a difference in operations when there isn't any other
information to make decisions from, like etruding a circle of
loose edges (the situation in the report). This commit changes
the order of the vertices in the final cyclic edge to go in the
same direction as all of the other edges.
The vertex and face normals from the input mesh
were used to calculate the normals on the result,
which could cause a crash because the result should
be about twice as large.
Also remove an unnecessary dirty tag, since it is handled
automatically when creating a new mesh or in the case
of the mirror modifier, when calculating the new custom
face corner normals.
Swap "active" and "selected" in the tooltip if the `use_reverse_transfer`
option is activated.
Reviewed By: mont29
Maniphest Tasks: T85233
Differential Revision: https://developer.blender.org/D13499
Also fixes similar issues regarding some liboverride menu entries.
Reviewed By: mont29
Maniphest Tasks: T93766
Differential Revision: https://developer.blender.org/D13513
Add `USD Preview Surface From Nodes` export option, to convert a
Principled BSDF material node network to an approximate USD Preview
Surface shader representation. If this option is disabled, the original
material export behavior is maintained, where viewport setting are saved
to the Preview Surface shader.
Also added the following options for texture export.
- `Export Textures`: If converting Preview Surface, export textures
referenced by shader nodes to a 'textures' directory which is a
sibling of the USD file.
- `Overwrite Textures`: Allow overwriting existing texture files when
exporting textures (this option is off by default).
- `Relative Texture Paths`: Make texture asset paths relative to the
USD.
The entry point for the new functionality is
`create_usd_preview_surface_material()`, called from
`USDAbstractWriter::ensure_usd_material()`. The material conversion
currently handles a small subset of Blender shading nodes,
`BSDF_DIFFUSE`, `BSDF_PRINCIPLED`, `TEX_IMAGE` and `UVMAP`.
Texture export is handled by copying texture files from their original
location to a `textures` folder in the same directory as the USD.
In-memory and packed textures are saved directly to the textures folder.
This patch is based, in part, on code in Tangent Animation's USD
exporter branch.
Reviewed By: sybren, HooglyBoogly
Differential Revision: https://developer.blender.org/D13647
Downgrade the Python zstandard from 0.17.0 to 0.16.0. The Python package
should be linked against the exact same version of libzstd as Blender is,
otherwise it will refuse to load from within the Blender executable.
Python zstandard 0.17.0 links to 1.5.1, whereas we need 1.5.0.
`chardet` was replaced by `charset_normalizer` for modern `requests`.
With this change, `{make,ninja} install` will also copy the latter into
Blender's install directory.
This opt-in functionnality enabled developper keep track of unused
resources present in the `GPUShaderCreateInfo` descriptors of their
shaders.
The output is pretty noisy at the moment so we do not enforce its usage.
While install_deps tries to stay as close as possible from official
Blender versions of the libraries, it also strives to use as many distro
packages as possible.
OSL 1.11.16.0 is the minimal version that builds with llvm13, which is
the default llvm/clang version in e.g. Debian testing.
Adds two new attribute outputs:
"Line" outputs the line number of the character.
"Pivot Point" outputs the selected pivot point position per char.
Some refactoring of the text layout code.
Differential Revision: https://developer.blender.org/D13694
`node_exec` had some code that was specific to texture/shader nodes.
These functions arent used outside there module so limit there declarations.
Also make a function static that is only used in `node_exec.c`
Reviewed By: JacquesLucke
Differential Revision: https://developer.blender.org/D13899
The `TreeElement.rnaptr` was only needed for RNA tree-elements. Now it
can be gotten through the new type specific classes, e.g.
`TreeElementRNAProperty.getPointerRNA()`.
Plan is to remove things like `TreeElement.directdata` and to instead
expose specific queries in the new type specific tree-element classes.
e.g. like here: `TreeElementSequence.getSequence()`
For now uses `tree_element_cast<>()` to get the new type specific
tree-element, later these should replace `TreeElement` all together.
Add function to safely request the type-specific C++ element from a
C-style `TreeElement`. Looks like this:
```
TreeElementFoo *te_foo = tree_element_cast<TreeElementFoo>(te);
```
The "cast" will return null if the tree-element doesn't match the
requested type.
This is useful for the transition from the C-style type to the new ones.
This uses a light parser / string modification pass to convert
C++ enum declaration syntax to GLSL compatible one.
GLSL having no support for enums, we are forced to convert the
enum values to a series of constant uints.
The parser (not really one by the way), being stupidly simple,
will not change anything to the values and thus make some C++
syntax (like omitting the values) not work.
The string replacement happens on all GLSL files on startup.
I did not measure significant changes in blender startup speed.
There is plans to do all of this at compile time.
We limit the scope of the search to `.h` and `.hh` files to prevent
confusing syntax in `.glsl` files.
There is basic error reporting with file, line and char logging
for easy debuggabiliy.
The requirements to use this enum sharing system are already listed in
`gpu_shader_shared_utils.h` and repeated on top of the preprocessor
function.
Remove small ray offsets that were used to avoid self intersection, and leave
that to the newly added primitive object/prim comparison. These changes together
significantly reduce artifacts on small, large or far away objects.
The balance here is that overlapping primitives are not handled well and should
be avoided (though this was already an issue). The upside is that this is
something a user has control over, whereas the other artifacts had no good
manual solution in many cases.
There is a known issue where the Blender particle system generates overlapping
objects and in turn leads to render differences between CPU and GPU. This will
be addressed separately.
Differential Revision: https://developer.blender.org/D12954
Caused by 0f89bcdbeb where it was needed for cage and evaluated mesh
to have same behavior in respect of having edit_mesh pointer assigned.
This change makes it so that edit_data is not implied to exist when the
edit_mesh pointer is not null. This was already the case in some other
code.
Those cases are almost always synptoms of either bug in code, or broken
files. Re-doin resync on them only costs time and causes extra trash
data as a result, without really helping in any way.
Both new normals (from rBb7fe27314b25) and vpaint (from rBf7bbc7cdbb6c)
RNA arrays were missing the `PROPOVERRIDE_IGNORE`. Those huge blobs of
geometry data should never be processed by liboverride code.
This enables support for node group assets. Previously, node group
assets only worked when the "extended asset browser" experimental
features is enabled.
Differential Revision: https://developer.blender.org/D13748
For node groups there is no good default preview generation.
Nevertheless, t would be useful to generate a preview image for a
node group by rendering an object in some cases.
This commit adds a new operator that allows updating the preview
image for the active asset by rendering the active object.
Note, the operator can also be used for other asset types, not just
node groups.
The operator can be found in a menu right below the refresh-preview
button. Currently it is the only operator in that menu. In the future,
more operators to create previews may be added.
Differential Revision: https://developer.blender.org/D13747
The VSE and node editor only uses an overlay buffer to draw to the screen. The
GPUViewport assumes that platforms clears all textures during creation, but
they do not on selected platforms. What would lead to drawing from
uncleared memory.
This patch fixes this by clearing all viewport textures during creation.
Also adds a few things to GPUShader for easily create shaders.
Heavy usage of macros to compose the createInfo and avoid
duplications and copy paste bugs.
This makes the link between the shader request functions
(in workbench_shader.cc) and the actual createInfo a bit
obscure since the names are composed and not searchable.
Reviewed By: jbakker
Differential Revision: https://developer.blender.org/D13910
Continuation of work started in 2e221de4ce and 249e4df110.
Adds new tree-element classes for RNA structs, properties and array
elements. This isn't exactly a copy and paste, even though logic should
effectively be the same. Further cleanups are included to share code in
a nice way, improve code with simple C++ features, etc.
This fixes failing test cases when using `make test`.
See {D13615} for more information.
The fix will perform the id remapping one item at a time. Although not
really nice, this isn't a bottleneck.
The failing test cases is because space_node stores pointers multiple
times and didn't update all pointers. It was not clear why it didn't do
it, but changing the behavior more to the previous behavior fixes the
issue at hand.
I prefer to remove the double storage of the node tree pointers (in
snode and path) to reduce pointer management complexity.
During sprite fright loading of complex scenes would spend a long time in remapping ID's
The remapping process is done on a per ID instance that resulted in a very time consuming
process that goes over every possible ID reference to find out if it needs to be updated.
If there are N of references to ID blocks and there are M ID blocks that needed to be remapped
it would take N*M checks. These checks are scattered around the place and memory.
Each reference would only be updated at most once, but most of the time no update is needed at all.
Idea: By grouping the changes together will reduce the number of checks resulting in improved performance.
This would only require N checks. Additional benefits is improved data locality as data is only loaded once
in the L2 cache.
It has be implemented for the resyncing process and UI editors.
On an Intel(R) Core(TM) i7-6700 CPU @ 3.40GHz 16Gig the resyncing process went
from 170 seconds to 145 seconds (during hotspot recording).
After this patch has been applied we could add similar approach
to references (references between data blocks) and functionality (tagged deletion).
In my understanding this could reduce the resyncing process to less than a second.
Opening the village production file between 10 and 20 seconds.
Flame graphs showing that UI remapping isn't visible anymore (`WM_main_remap_editor_id_reference`)
* Master {F12769210 size=full}
* This patch {F12769211 size=full}
Reviewed By: mont29
Maniphest Tasks: T94185
Differential Revision: https://developer.blender.org/D13615
This reverts commit 086f191169.
There was apparently a problem using APPEND which wasn't referenced
in the commit log.
Added comment noting the reason for the discrepancy.
Use more efficient logic for detecting when gizmos are under the cursor.
Even though this isn't a bottleneck, it runs on cursor motion in the
3D viewport, so avoiding any lag here is beneficial.
The common case for cursor motion without any gizmos was always
drawing two passes (one small, then again if nothing was found).
Now a single draw call at the larger size is used.
In isolation this gives around 1.2x-1.4x speedup.
When there are multiple gizmos a depth-buffer picking is used
(similar to object / bone selection) which is more involved but
still only performs 2x draw calls since the result is cached for reuse.
See note in gizmo_find_intersected_3d for a more detailed explanation.
Also restore the depth values in the selection result as they're
needed for gizmos to use selection bias.
Broken since support for GL_SELECT was removed.
Early on in 2.8x development gizmo-depth used GL_SELECT,
which has been removed. Bind the depth buffer so occlusion queries
use the front-most gizmo.
While this report only mentions face-maps, gizmo depth was ignored in
all cases. This wasn't noticeable in most cases though since the
transform gizmo for example was placed so gizmos didn't overlap.
When calling GPU_select_cache_begin, checking the selection mode used
the last used selection mode, not the one about to be used.
Using border select, then picking would not use the selection cache.
This wasn't noticeable by users as failing to use cache just completes
the selection without it (drawing the depth buffer unnecessarily).
The OSL image compilation step needed to be taught about the new UVTILE
format for UDIM textures.
A small missing feature from OIIO[1] means this is a bit uglier than it
needs to be. Once we update to a version of OIIO with the fix we can
remove the string replace part.
[1] 35cb6a83e2
Differential Revision: https://developer.blender.org/D13912
This was already done for APPLE & WIN32, which would
reference these libraries twice.
Now append BROTLI_LIBRARIES to FREETYPE_LIBRARIES when they're
required for linking.
No functional changes as all references to FREETYPE_LIBRARIES also
used BROTLI_LIBRARIES.
When LIBDIR existed, searching for system libraries would always
first search 'LIBDIR'.
This meant "WITH_SYSTEM_*" would still prefer LIBDIR versions of
libraries if they exist.
The presence of LIBDIR also ignored the setting for WITH_STATIC_LIBS
which is now restored to the cached value once pre-compiled libraries
have been handled.
Fix formula in function `SEQ_sound_update_length`.
Formula for sound strip length was changed in commit ded68fb102, when
strip is added to timeline, but it was not changed in function
mentioned above.
This change applies only for automatic proxy building, when strip
is added to timeline. Manual building process is not affected.
Don't build proxy file if movie is already fast enough to seek.
To determine seek performance, check if whole GOP can be decoded
in 100 milliseconds.
To consider some variation in GOP size, large number of packets are
read, assuming that each packet will produce 1 frame. While this is not
technically correct, it does give quite accurate estimate of maximum GOP
size.
This test will ensure consistent performance on wide array of machines.
Check should be done in order of few milliseconds.
Reviewed By: sergey
Differential Revision: https://developer.blender.org/D11671
This will allow building most deps with VS2019
SDL has some linker issues that are resolved in
a newer version, but that would be better handled
in a separate change.
VS2013 and VS2015 support which was broken has
been removed.
In order to use a workaround builtin uniform, we need to count it
just like other uniforms and give it some space in the name buffer.
This also fixes extensions being added after the uniform declaration.
All `#extension` directives are now part of the gl backend.
This commit moves the weld modifier code to the geometry module
so that it can be used in the "Merge by Distance" geometry node
from ec1b0c2014. The "All" mode is exposed in the node
for now, though we could expose the "Connected" mode in the future.
The modifier itself is responsible for creating the selections from
the vertex group. The "All" mode takes an `IndexMask` for the
selection, and the "Connected" mode takes a boolean array,
since it actually iterates over all edges.
Some disabled code for a BVH mode has not been copied over,
it's still accessible through the patches and git history anyway,
and it made the port slightly simpler.
Differential Revision: https://developer.blender.org/D13907
Previously it was only part of experimental features in beta, however now
renderers can render point clouds generated by geometry nodes. Adding or
converting a point cloud object directly is still hidden by default, since
there is no good way to edit it.
This implements a merge by distance operation for point clouds.
Besides the geometry input, there are two others-- a selection
input to limit the operation to certain points, and the merge
distance. While it would be a reasonable feature, the distance
does not support a field currently, since that would make
the algorithm significantly more complex.
All attributes are merged to the merged points, with the values
mixed together. This same generic method is used for all attributes,
including `position`. The `id` attribute uses the value from the
first merged index for each point.
For the implementation, most of the effort goes into creating a
merge map to speed up attribute mixing. Some parts are inherently
single-threaded, like finding the final indices accounting for the
merged points. By far most of the time is spend balancing the
KD tree.
Mesh support will be added in the next commit.
Differential Revision: https://developer.blender.org/D13649
Only show options that are valid for the used device (CPU, GPU, Multi).
Note: The panel isn't shown for OPTIX anymore, unless Multi device is used.
Reference: https://developer.blender.org/D13592
Make the Embree RTC_SCENE_FLAG_COMPACT flag optional and enabled per default.
Disabling it makes CPU rendering a bit faster in some scenes at the cost of a higher memory usage.
Barbershop renders about 3% faster, victor about 4% on CPU with compact BVH disabled.
Differential Revision: https://developer.blender.org/D13592
Several sub commands tried on their own
to locate python, given I wanted to look
in several locations for a broader libdir
compatibility this is best done in a
central location.
Python 3.9 is still preferred, but if
3.10-3.12 are available that be accepted
as well.
note: this is about the python version
make.bat uses to run various python helper
scripts, this change has no influence on
the python version blender itself uses.
With (center) position, radius and random value outputs.
Eevee does not yet support rendering point clouds, but an untested
implementation of this node was added for when it does.
Ref T92573
Movies with variable frame rate can cause mismatch of displayed frame
when proxies are used. Since proxies are not used for rendering, this
means, that output may be different than expected. This problem can be
avoided when timecodes are used.
Set used timecode to Record Run. Timecodes are built with proxies at
the same time, therefore if proxies are built and used this will
resolve possible mismatch of output.
Record run is chosen, because it will show frames based on time they
were encoded by encoder and should match behavior as if movie was
played back at normal speed. This change is done only for new strips
in order to not overwrite user defined settings.
Other minor changes:
- When proxies are enabled, size 25% is no longer set by default. It was mostly annoying anyway.
- Silence warning when timecode file is not present. This was introduced in 4adbe31e2f.
Previously use of timecodes was hard-coded in sequencer and this error would spam console if timecodes would be
enabled by default and proxies would be never built.
ref: T95093
Reviewed By: sergey
Differential Revision: https://developer.blender.org/D13905
Using a negative linesize to flip an image vertically is supported in
ffmpeg but not for every function.
This method treats frames that need and those that do not need alignment
the same. An RGBA frame buffer with alignment that ffmpeg decides is
optimal for the CPU and build options is allocated by ffmpeg.
The `sws_scale` does the colorspace transformation into this RGBA frame
buffer without flipping. Now the image is upside down and aligned.
The combined unaligning and vertical flipping is then done by
`av_image_copy_to_buffer` which seems to handle negative linesize
correctly.
Reviewed By: ISS
Differential Revision: https://developer.blender.org/D13908
Layer resync code would not yet fully properly deal with all possible
invalid status of ViewLayer comming from those older files.
Now put 2.80-doversion specific fixes into their own dedicated
function, so that they do not affect actual regular layer resync code
anymore. Also added some sanity-checks in main
`BKE_layer_collection_sync` code.
The Brotli library only needs to be explicitly linked when using the
statically linked libraries. When using system libs they're shared, and
the .so loading mechanism takes care of dependencies.
Currently the Boolean Math node only has 3 basic logic gates:
AND, OR, and NOT. This commit adds 6 additional logic gates
for convenience and ease of use.
- **Not And (NAND)** returns true when at least one input is false.
- **Nor (NOR)** returns true when both inputs are false.
- **Equal (XNOR)** returns true when both inputs are equal.
- **Not Equal (XOR)** returns true when both inputs are different.
- **Imply (IMPLY)** returns true unless the first input is true and
the second is false.
- **Subtract (NIMPLY)** returns true when the first input is true and
the second is false.
Differential Revision: https://developer.blender.org/D13774
ssize_t is a posix type pyconfig.h previously
supplied for MSVC, it appears to have stopped
doing this in the python 3.10 headers.
Py_ssize_t is the type of the field this macro
actually returns, so best to to use that in our
code as well.
This reverts commit 948211679f.
This commit introduced some regressions in the test suite.
As this change is a core part of blender Bastien and I decided to revert
it as the solution isn't clear and needs more investigation.
The following tests FAILED:
62 - blendfile_liblink (SEGFAULT)
63 - blendfile_library_overrides (SEGFAULT)
It fails in (id_us_ensure_real)
This makes optionnal the use of a different interface for the geometry
shader stage output. When the vertex and geometry interface instance name
matches, a `_in` and `_out` suffix is added to the end of the instance name.
This makes it easier to have optional geometry shader stages.
# Conflicts:
# source/blender/gpu/intern/gpu_shader_create_info.hh
During sprite fright loading of complex scenes would spend a long time in remapping ID's
The remapping process is done on a per ID instance that resulted in a very time consuming
process that goes over every possible ID reference to find out if it needs to be updated.
If there are N of references to ID blocks and there are M ID blocks that needed to be remapped
it would take N*M checks. These checks are scattered around the place and memory.
Each reference would only be updated at most once, but most of the time no update is needed at all.
Idea: By grouping the changes together will reduce the number of checks resulting in improved performance.
This would only require N checks. Additional benefits is improved data locality as data is only loaded once
in the L2 cache.
It has be implemented for the resyncing process and UI editors.
On an Intel(R) Core(TM) i7-6700 CPU @ 3.40GHz 16Gig the resyncing process went
from 170 seconds to 145 seconds (during hotspot recording).
After this patch has been applied we could add similar approach
to references (references between data blocks) and functionality (tagged deletion).
In my understanding this could reduce the resyncing process to less than a second.
Opening the village production file between 10 and 20 seconds.
Flame graphs showing that UI remapping isn't visible anymore (`WM_main_remap_editor_id_reference`)
* Master {F12769210 size=full}
* This patch {F12769211 size=full}
Reviewed By: mont29
Maniphest Tasks: T94185
Differential Revision: https://developer.blender.org/D13615
This patch migrates the draw manager hair refine compute shader to use
GPUShaderCreateInfo.
Reviewed By: fclem
Differential Revision: https://developer.blender.org/D13915
Allows to perform correction of coordinate delta/displacement in a
similar way of how sculpt mode handles sculpting on a deformed mesh.
An example of usecase of this is allowing riggers and sciprters to
improve corrective shapekey workflow.
The usage consists of pre-processing and access. For example:
object.crazyspace_eval(depsgraph, scene)
# When we have a difference between two vertices and want to convert
# it to a space to be stored, say, in shapekey:
delta_in_orig_space = rigged_ob.crazyspace_displacement_to_original(
vertex_index=i, displacement=delta)
# The reverse of above.
delta_in_deformed_space = rigged_ob.crazyspace_displacement_to_deformed(
vertex_index=i, displacement=delta)
object.crazyspace_eval_clear()
Fuller explanation with actual usecases and studio examples are written in
the comment:
https://developer.blender.org/D13892#368898
Differential Revision: https://developer.blender.org/D13892
Brotli seems to add a custom postfix to its static libraries by default,
but in Debian at least libraries are just named the same for both shared
and static versions, as usual.
So add standard name after static-specific ones.
Follow-up to rB4c617c06e9cb and rBa000de7c2a4d.
The evaluated mesh is a result of evaluated modifiers, and referencing
other evaluated IDs such as materials.
It can not be stored in the EditMesh structure which is intended to be
re-used by many areas. Such sharing was causing ownership errors causing
bugs like
T93855: Cycles crash with edit mode and simultaneous viewport and final render
The proposed solution is to store the evaluated edit mesh and its cage in
the object's runtime field. The motivation goes as following:
- It allows to avoid ownership problems like the ones in the linked report.
- Object level is chosen over mesh level is because the evaluated mesh
is affected by modifiers, which are on the object level.
This patch allows to have modifier stack of an object which shares mesh with
an object which is in edit mode to be properly taken into account (before
the change the modifier stack from the active object will be used for all
objects which share the mesh).
There is a change in the way how copy-on-write is handled in the edit mode to
allow proper state update when changing active scene (or having two windows
with different scenes). Previously, the copt-on-write would have been ignored
by skipping tagging CoW component. Now it is ignored from within the CoW
operation callback. This allows to update edit pointers for objects which are
not from the current depsgraph and where the edit_mesh was never assigned in
the case when the depsgraph was evaluated prior the active depsgraph.
There is no user level changes changes expected with the CoW handling changes:
should not affect on neither performance, nor memory consumption.
Tested scenarios:
- Various modifiers configurations of objects sharing mesh and be part of the
same scene.
- Steps from the reports: T93855, T82952, T77359
This also fixes T76609, T72733 and perhaps other reports.
Differential Revision: https://developer.blender.org/D13824
The Equalize Handles operator allows users to make selected handle
lengths uniform: either respecting their original angle from the key
control point or by flattening their angle (removing the overshoot
sometimes produced by certain handle types).
Design: T94172
Reviewed by: sybren
Differential Revision: https://developer.blender.org/D13702
Some IDs (like text ones) can be linked and only kept around thanks to
editors, allow making such IDs local in `BKE_lib_id_make_local_generic`.
Also refactor logic checking whether ID should be made directly local or
copied into its own util function, so that we can remain sure all
special-cases 'make local' code still uses the same logic here.
Currently there are many function declarations in `BKE_node.h` that
don't actually have implementations in blenkernel. This commit moves
the declarations to `NOD_composite.h`, `NOD_texture.h`, and
`NOD_shader.h` instead. This helps to clarify the purpose of the
different modules.
Differential Revision: https://developer.blender.org/D13869
Fixes T93680
For current drivers of Intel HD Graphics 4400 and 4600, various Program Introspection functions appear broken and return incorrect values, causing crashes in the current handling of SSBOs. Disable use of this feature on those devices. Add checks to features that use SSBOs (Hair and Subdivision Modifier).
Reviewed By: fclem, jbakker
Maniphest Tasks: T93680
Differential Revision: https://developer.blender.org/D13806
The point IBO should only have data for coarse vertices (or in general,
the vertices in the original mesh). As it used for displaying the
vertices for selection in edit mode, and as it indexes into the VBOs for
the positions and edit data, it is itself only indexed by coarse/
original vertex index.
For the subdivision case, this would allocate space for the final
subdivision vertex and reallocate to make room for loose geometry,
although only the first coarse vertex count amount of data would be.
Now just allocate for the required memory. Also reuse index buffer APIs
instead of doing manual work.
unity launches blender in background mode to do some
file conversions, ever since the launcher got introduced
this process broke.
The root cause here is: Unity looks up the default program
to launch .blend files with, which is now the launcher, then
launches it in background mode with a script to export the data.
The launcher however was designed to exit as quickly as
possible so there would not be an extra background process
lingering. It does not wait for blender to exit and does not
pass back any error codes.
This broke unity's workflow since it assumed if the process
exits and succeeds the data *must* be ready for reading which
no longer holds true.
This change keeps the launcher design as was previously,
*except* when launching in background mode, then it
waits and passes back any error codes, thus restoring
unity's workflow.
Differential Revision: https://developer.blender.org/D13894
Reviewed by: LazyDodo, Brecht
Previously weight paint wasn't hooked up to the "Smooth" and "Invert" modes.
With this patch it is not possible to use the "Smooth" and "Invert"
modifiers for the draw keybindings.
Reviewed By: Campbell Barton
Differential Revision: http://developer.blender.org/D13857
Add `libbrotlidec-static.a` and `libbrotlicommon-static.a` to the CMake
`$FREETYPE_LIBRARIES` variable; they'll be required when the Linux libs
for the FreeType upgrade lands (D13448).
The order of libraries is different compared to the similar lines in the
Windows and Apple CMake files, to prevent linker errors on Linux.
Some drivers/glsl compilers will not warn about multiple resources using
the same binding, creating silent errors.
This patch checks for this case and outputs a descriptive error message if
a particular createInfo merge error is founds.
Other validation can be added later.
This patch introduces an extrude node with three modes. The vertex mode
is quite simple, and just attaches new edges to the selected vertices.
The edge mode attaches new faces to the selected edges. The faces mode
extrudes patches of selected faces, or each selected face individually,
depending on the "Individual" boolean input.
The default value of the "Offset" input is the mesh's normals, which
can be scaled with the "Offset Scale" input.
**Attribute Propagation**
Attributes are transferred to the new elements with specific rules.
Attributes will never change domains for interpolations. Generally
boolean attributes are propagated with "or", meaning any connected
"true" value that is mixed in for other types will cause the new value
to be "true" as well. The `"id"` attribute does not have any special
handling currently.
Vertex Mode
- Vertex: Copied values of selected vertices.
- Edge: Averaged values of selected edges. For booleans, edges are
selected if any connected edges are selected.
Edge Mode
- Vertex: Copied values of extruded vertices.
- Connecting edges (vertical): Average values of connected extruded
edges. For booleans, the edges are selected if any connected
extruded edges are selected.
- Duplicate edges: Copied values of selected edges.
- Face: Averaged values of all faces connected to the selected edge.
For booleans, faces are selected if any connected original faces
are selected.
- Corner: Averaged values of corresponding corners in all faces
connected to selected edges. For booleans, corners are selected
if one of those corners are selected.
Face Mode
- Vertex: Copied values of extruded vertices.
- Connecting edges (vertical): Average values of connected selected
edges, not including the edges "on top" of extruded regions.
For booleans, edges are selected when any connected extruded edges
were selected.
- Duplicate edges: Copied values of extruded edges.
- Face: Copied values of the corresponding selected faces.
- Corner: Copied values of corresponding corners in selected faces.
Individual Face Mode
- Vertex: Copied values of extruded vertices.
- Connecting edges (vertical): Average values of the two neighboring
edges on each extruded face. For booleans, edges are selected
when at least one neighbor on the extruded face was selected.
- Duplicate edges: Copied values of extruded edges.
- Face: Copied values of the corresponding selected faces.
- Corner: Copied values of corresponding corners in selected faces.
**Differences from edit mode**
In face mode (non-individual), the behavior can be different than the
extrude tools in edit mode-- this node doesn't handle keeping the back-
faces around in the cases that the edit mode tools do. The planned
"Solidify" node will handle that use case instead. Keeping this node
simpler and faster is preferable at this point, especially because that
sort of "smart" behavior is not that predictable and makes less sense
in a procedural context.
In the future, an "Even Offset" option could be added to this node
hopefully fairly simply. For now it is left out in order to keep
the patch simpler.
**Implementation**
For the implementation, the `Mesh` data structure is used directly
rather than converting to `BMesh` and back like D12224. This optimizes
for large extrusion operations rather than many sequential extrusions.
While this is potentially more verbose, it has some important benefits:
First, there is no conversion to and from `BMesh`. The code only has
to fill arrays and it can do that all at once, making each component of
the algorithm much easier to optimize. It also makes the attribute
interpolation more explicit, and likely faster. Only limited topology
maps must be created in most cases.
While there are some necessary loops and allocations with the size of
the entire mesh, I tried to keep everything I could on the order of the
size of the selection rather than the size of the mesh. In that respect,
the individual faces mode is the best, since there is no topology
information necessary, and the amount of work just depends on the size
of the selection.
Modifying an existing mesh instead of generating a new one was a bit
of a toss-up, but has a few potential benefits:
- Avoids manually copying over attribute data for original elements.
- Avoids some overhead of creating a new mesh.
- Can potentially take advantage of future ammortized mesh growth.
This could be changed easily if it turns out to be the wrong choice.
Differential Revision: https://developer.blender.org/D13709
When moving to C++ field for initialization was removed.
Favor assignments to field names as it reads better and avoids bugs if
files are ever re-arranged as well as mistakes (see T94784).
Note that the generated optimized output is identical with GCC11.
This adds a selection field input to the node, faces that are selected and
meet the minimum vertex count threshold will be triangulated.
Differential Revision: https://developer.blender.org/D13804
Add a boolean option to have the Curve Handle Position input node return the
position of the handle relative to each point position.
Differential Revision: https://developer.blender.org/D12947
A large polygon in the file from the report caused `alloca`
to exceed the maximum stack size, causing a crash. Instead
of using `alloca`, use `blender::Array` with an inline buffer.
Based on a patch by Germano Cavalcante (@mano-wii).
Differential Revision: https://developer.blender.org/D13898
When enabled, it will keep contour around the object instead of hide them by rule of face mark,
so the object can always have full contour while filtering out some of the feature lines inside certain regions.
Reviewed By: Antonio Vazquez (antoniov), Aleš Jelovčan (frogstomp)
Differential Revision: https://developer.blender.org/D13847
Option to discard back faced triangles, this speeds up calculation especially for when you only want to show visible feature lines.
Reviewed By: Antonio Vazquez (antoniov), Aleš Jelovčan (frogstomp)
Differential Revision: https://developer.blender.org/D13848
Instead of splitting it at each occlusion change, it tolerates short segments of "zig-zag" occlusion incoherence and doesn't split the chain at these points, thus creating a much smoother result.
Reviewed By: Antonio Vazquez (antoniov), Aleš Jelovčan (frogstomp)
Differential Revision: https://developer.blender.org/D13851
This gives a modest speedup as calculating tessellation and face
normals at the same time can be more efficiently multi-threaded.
Also avoids calculating face normals twice,
oversight in d590e223da.
This commit improves NURBS knot generation by adding proper support
for the combination of the Bezier and cyclic options. In other cases
the resulting knot doesn't change. This cyclic Bezier knot is used to
create accurate accurate "Nurbs Circle", "Nurbs Cylinder" primitives.
"Nurbs Sphere" and "Nurbs Torus" primitives are also improved by
tweaking the spin operator.
The knot vector in 3rd order NURBS curve with Bezier option turned on
(without cyclic) is changed in comparison to previous calculations,
although it doesn't change the curve shape itself.
The accuracy of the of NURBS circle is fixed, which can be checked by
comparing with mesh circle. Tessellation spacing differences in
circular NURBS is also fixed, which is observable with the NURBS
cylinder and sphere primitives. These were causing seam-like effects.
This commit contains comments from Piotr Makal (@pmakal).
Differential Revision: https://developer.blender.org/D11664
Normal layers currently aren't stored in the undo step
mesh storage, since they are not stored in files at all.
However, the edit mesh expects normals to be fully
calculated, and does not keep track of a dirty state.
This patch updates the normals in the BMesh created
by loading an undo step.
Another option would be calculating the normals on
the undo mesh first, which might be better if Mesh
normal calculation is faster than BMesh calculation,
but the preferred method to access vertex normals fails
in this case, because the mesh runtime mutexes are not
initialized for undo-state meshes.
Differential Revision: https://developer.blender.org/D13859
From an error in rBcfa53e0fbeed, the vertex normals in `SculptSession`
seem to be used, but in the case when no "pbvh" is used, the value of
the pointer is never assigned.
Normals were not generally dirty before this "ensure" function with
regular sculpting operations, so this addition shouldn't have any cost.
Differential Revision: https://developer.blender.org/D13854
All other nodes with data type and domain choices have the domain
below the data type. Generally that order makes sense, because it's
consistent with nodes that have no domain drop-down.
Because this operator is used on original objects, it's best to tag
the normals dirty instead of explicitly calculating them, to avoid
unnecessarily storing normal layers on an original object (since
they might have to be recalculated during evaluation anyway).
There may be other places this change is helpful, but being
conservative is likely better for now.
Related to T95125
The new OBJ exporter did not handle object instances.
The fix is to use a dependency graph iterator, asking for instances.
Unfortunately that iterator makes a temporary copy of instance objects
that does not persist past the iteration, but we need to save all the
objects and meshes to write later, so the Object has to be copied now.
This changed some unit tests. Even though the tests don't have instancing,
the iterator also picks up some Text objects as Mesh ones (which is a good
thing), resulting in two more objects in the all_objects.obj file output.
With object collection properties open there was a huge delay when
switching active objects in a large scene, (~10k objects, ~5m vertices).
This is due to a non-optimal function to query all the collections the object is in.
To solve this the code can be simplified by using `bpy.types.Object.users_collection`
This returns all the collections the object is in removing the need to compute this in python.
The UI team requested adding woff2 support to freetype.
this required a new dependency brotli.
This changes adds brotili to the builder and bumps
freetype to version 2.11.0
As freetype now depends on other libraries, for consistency
all use of ${FREETYPE_LIBRARY} in cmake has been updated to
use ${FREETYPE_LIBRARIES} adjustments have been made in the
windows platform file, all other platforms use cmake's
FindFreeType.cmake which already sets this variable.
reviewed by: brecht
Differential Revision: https://developer.blender.org/D13448
This node can scale individual edges and faces. When multiple selected
faces/edges share the same vertices, they are scaled together.
The center and scaling factor is averaged in this case.
For some examples see D13757.
Differential Revision: https://developer.blender.org/D13757
Currently there is no way to flip normals in geometry nodes. This node
makes that possible by flipping the winding order of selected faces.
The node is purposely not called "Flip Normals", because normals are
derived data, changing them is only a side effect. The real change is
that the vertex and edge indices in the face corners of every selected
polygon are reversed, and face corner attribute data is reversed.
While there are existing utilities to flip a polygon and its custom
data, this node aims to process an attribute's data together instead
of processing all attributes separately for each index.
Differential Revision: https://developer.blender.org/D13809
Adds a second output to the Mesh Islands node that shows the total
number of islands as a field.
Differential Revision: https://developer.blender.org/D13700
This is not currently working, with an internal compiler error. However
we are currently using BVH2 instead of Metal RT. So this has no effect for
users, it's being committed to avoid the code getting outdated.
Ref T92573, T92212
Differential Revision: https://developer.blender.org/D13632
This shader needs to use the same interface as the OCIO shader. On Linux
and Windows this seems to be the case. On MacOS however there is a
mismatch that makes the overlay texture to be completely black when
switching to workbench in the Shader workspace.
This is just a temporarily work-around as this should be solved. Due to
the poor GPU debugging facilities on Mac we haven't been able to
pin-point the root cause.
Editing of generic attributes on the original objects in edit modes is
still very limited. However, when applying a geometry nodes modifier
that generates new attributes. These attributes will show up in the
Attributes panel.
Currently, our exporters are not capable of exporting generic attributes.
Therefore, for the time being, a work around is to apply geometry nodes
and then convert a generic attribute to a task specific attribute like a
uv map, vertex group or vertex color layer. Once more parts of Blender
support generic attributes, this will become less important.
Currently, only meshes are supported by the operator. However, it would
be relatively easy to extend it to other geometry types.
Differential Revision: https://developer.blender.org/D13838
Allow any non-modifier keyboard events to be used.
Note that the existing check for >= EVT_AKEY allowed NDOF and other
non-keyboard events (including modifiers, which didn't work).
- Use defines instead of magic numbers for F-Key & NDOF range checks.
- Use explicit values for NDOF event types.
- Minor clarification to doc-strings.
- Use doxy-sections.
There is probably a better solution that's possible, but the few other
things I tried didn't work, and the build error should be resolved for
the buildbots. This is similar to the "breaks" in the namespace for
functions declared in `ED_node.h`.
The code here was using velocity based interpolation copied from particles.
However there is no velocity cached in the rigid body point cache so the
results were non-sensical.
This adds a new curve primitive to generate arcs.
Radius mode (default): Generates a fixed radius arc on XY plane
with controls for Angle, Sweep and Invert.
Points mode: Generates a three point curve arc from Start to End
via Middle with an Angle Offset and option to invert the arc.
There are also outputs for arc center, radius and normal direction
relative to the Z-axis.
This patch is based on previous patches
D11713 and D13100 from @guitargeek. Thank you.
Reviewed By: HooglyBoogly
Differential Revision: https://developer.blender.org/D13640
This fixes a similar issue as the previous commit, but this time the
continuous notifiers would be sent after redoing. E.g. after moving an
object, and then modifying the transform in the "Adjust Last Operation"
panel.
The thumbnail caching continuously sends `ND_SPACE_FILE_PREVIEW`
notifiers via a timer. But this timer was never ended properly after
thumbnails are fully loaded into the cache.
Wouldn't actually cause a refresh or redraw, send and process the
notifiers.
I already tried to avoid this for the asset view template, but
apparently that wasn't working correctly. For the File/Asset Browser I
never applied that fix to avoid possible regressions before the release.
This commit moves code in all node editor files to the
`blender::ed::space_node` namespace, except for C API
functions defined in `ED_node.h`, which can only be moved
once all areas calling them are moved to C++.
The change is fairly straightforward, I just moved a couple
of "ED_" code blocks around to make the namespace more
contiguous, and there's the method for adding a pointer to
a struct in a C++ namespace in DNA.
Differential Revision: https://developer.blender.org/D13871
This patch fixes a correctness issue discovered in the `int4 select(...)` function on Apple Silicon machines, which causes bad bvh2 builds. Although the generated bvh2s give correct renders, the resulting runtime performance is terrible. This fix allows us to switch over to bvh2 on Apple Silicon giving a significant performance uplift for many of the standard benchmarking assets. It also fixes some unit test failures stemming from the use of MetalRT, and trivially enables the new pointcloud primitive.
Ref T92212
Reviewed By: brecht
Maniphest Tasks: T92212
Differential Revision: https://developer.blender.org/D13877
Custom bones are drawn by instancing the GPUBatch of the base object. To
access the mesh and its GPUBatch, `BKE_object_get_evaluated_mesh` was
used. However, since GPU subdivision support, this will return a
subdivision wrapper which will never be drawn, and thus will have an
invalid batch, which caused the crash.
`BKE_object_get_evaluated_mesh_no_subsurf` should be used instead, to
return the mesh that will be drawn, and have the subdivision evaluated
on the GPU. Note that the rest of the draw code is already using this
function.
This adds vertex creasing support for OpenSubDiv for modeling, rendering,
Alembic and USD I/O.
For modeling, vertex creasing follows the edge creasing implementation with an
operator accessible through the Vertex menu in Edit Mode, and some parameter in
the properties panel. The option in the Subsurf and Multires to use edge
creasing also affects vertex creasing.
The vertex crease data is stored as a CustomData layer, unlike edge creases
which for now are stored in `MEdge`, but will in the future also be moved to
a `CustomData` layer. See comments for details on the difference in behavior
for the `CD_CREASE` layer between egdes and vertices.
For Cycles this adds sockets on the Mesh node to hold data about which vertices
are creased (one socket for the indices, one for the weigths).
Viewport rendering of vertex creasing reuses the same color scheme as for edges
and creased vertices are drawn bigger than uncreased vertices.
For Alembic and USD, vertex crease support follows the edge crease
implementation, they are always read, but only exported if a `Subsurf` modifier
is present on the Mesh.
Reviewed By: brecht, fclem, sergey, sybren, campbellbarton
Differential Revision: https://developer.blender.org/D10145
This brush fixes the random spikes that
occasionally happen in multires models.
These spikes can be nearly impossible to
fix manually and can make working with
multires a nightmare.
Since vertex and face normals are now stored on the mesh when necessary,
we can expose them as contiguous arrays of vectors in the Python API.
This can give render engines and other addons easy access to they data
for fast access through a regular collection property.
While "Mesh Vertex" still has a "normal" property in RNA, that is only
maintained in order to avoid breaking the existing API, and accessing
it is less efficient than accessing the normals directly.
I made the normal arrays read-only, because modifying them could
put them in an invalid state. This is inline with how we treat the data
internally, and helps keep relationships between data clear.
Differential Revision: https://developer.blender.org/D13839
When toggling to a File Browser from an Asset Browesr, the asset indexer
would be used to load files. I couldn't spot issues with that on a
quick look, but this should still be corrected.
Adds an "Asset Indexing" option (enabled by default) to Preferences >
Experimental > Debugging. This is useful when working on the asset
library loading.
An external CMake script can be used to debug CMake code, modify/read
target properties, inter-target dependencies, non-cache variables etc.
Reviewed by: LazyDodo, brecht
Differential Revision: https://developer.blender.org/D13830
This patch fixes crash T94736 on Metal in which the launch_params were not being updated to reflect destruction of MetalMem objects.
Reviewed By: brecht
Differential Revision: https://developer.blender.org/D13875
When copying strips between 2 scenes, it wasn't possible to copy
animation curves along with strips.
In this patch curves are copied into clipboard `ListBase`. When pasted,
original curves are moved into temporary `ListBase` and curves in
clipboard are moved into scene action. This is because when strips from
clipboard have to be renamed, function `SEQ_ensure_unique_name()` does
fix RNA paths of curves, but this is done globally for all curves within
action. After strips are renamed, restore original curves from backup.
Note: This patch handles only fcurves. Drivers and actions are currently
not handled anywhere in VSE.
Fixes T77530
Reviewed By: sergey
Differential Revision: https://developer.blender.org/D13845
- Move functions that handle animation to own file - `animation.c`
- Refactor `SEQ_offset_animdata` and `SEQ_free_animdata` functions
- Add function `SEQ_fcurves_by_strip_get` to provide more granular
and explicit way for operators to handle animation
- Remove function `SEQ_dupe_animdata`, do curve duplication explicitly
in operator code, which makes more sense to do. Further this function
was also used for renaming strips which makes no sense.
- Refactor usage of function `SEQ_free_animdata` and remove XXX comment.
Now this functiuon is no longer called when `Sequence` data is freed
implicitly, it is done explicitly in high level function
`SEQ_edit_remove_flagged_sequences`
Reviewed By: sergey
Differential Revision: https://developer.blender.org/D13852
The ustring is not a trivially copyable object from the C++ standard
point of view, so using memcpy on it is strictly wrong. In practice,
however, this is OK since it is just a thin wrapper around char*.
For now use explicit cast to void* same as it was done in other places
of ccl::array implementation. But also localize the place where memory
copy happens to make it easier to support proper non-trivial C++
objects in the future.
This merge the description into one struct only that can be more easily
copied during `finalize()`.
The in and out layout parameters are better named and extended with the
invocation count (with fallback support)
Route cause was data alignment mismatch between GPU and CPU. This
mismatch would not allow us to bind the UBO where data wasn't available
on the GPU.
Fixed by using float4 in stead of float2. This could eventually be
packed, but that would lead to less readable code.
Cause was incorrect logic when generating the resource layout. It the
explicit_location_support setting was ignored and the binding were
generated for image, uniform buffers and storage buffers.
- Add BM_mesh_debug_print & BM_mesh_debug_info.
- Report flags in Mesh.cd_flag in BKE_mesh_debug_print
- Move custom data printing into customdata.cc (noted as a TODO).
Note that the term "runtime" has been removed from
`BKE_mesh_runtime_debug_print` since these are useful for debugging any
kind of mesh data.
Code that handled merging & initializing custom-data from other
meshes sometimes missed checks for this flag, causing bevel weights to
lost when the mesh was converted to a BMesh.
The following changes are a more general fix for T94197.
- Add BM_mesh_copy_init_customdata_from_mesh_array which initializes
custom-data from multiple meshes at once.
As well as initializing custom-data layers from Mesh.cd_flag.
This isn't essential for boolean, however it avoids the overhead of
resizing custom-data layers.
- Loading mesh data into a BMesh now respects Mesh.cd_flag
instead of only checking if the BMesh custom-data-layer exists.
Without this, the order of meshes passed to BM_mesh_bm_from_me could
give different (incorrect) results.
- Copying mesh data now copies `cd_flag` too. This is a precaution
as in my tests evaluating modifiers these values always matched.
Nevertheless it's correct to copy this value as custom-data it's
self is being copied.
The buffer passed as an argument to `GPUFrameBuffer.read_color` is used
in the return of the function and therefore, if not used, its refcount is
decremented.
So be sure to increment the refcount of the already existing objects that
will be used in the return of a function.
This was the case for multi input sockets that have a link already.
Since we have multi input sockets, the way we use `socket_is_available`
is not really giving the expected result on these.
When used for input sockets the intention is to find a free socket
(either for noodle **replacement**, then it is always available, or just
the next free available socket).
Now I would think without the intention to replace an existing link, a
multi input socket should still be available.
From the inside of the function, the `replace` argument turns [namewise]
to `allow_used`, which sounds a little different (so one might argue
that if `allow_used` is `False` this should also trigger for already
connected multi input sockets).
In the end, this is an issue with the variable naming though, cant think
of a usecase where the patch change would really go against intentions.
Maniphest Tasks: T93413
Differential Revision: https://developer.blender.org/D13866
Bug possibly introduced in {rBc57e4418bb85aec8bd3615fd775b990badb43d30}.
Interestingly, the orientation set before (NORMAL), even different from
the orientation that was actually used, was allowing the use of
"orient_matrix" ("orient_matrix_type" should have been NORMAL in that
case too).
In any case, make sure the "orient_matrix_type" and "orient_type" are the
same so that the "orient_matrix" is used.
These operations (sorting and selecting all nodes) should generally
be handled by the node editor and not outside code. They were not
called outside of the node editor, so they can be moved to the editor's
`intern` header.
This file was added nine years ago, and was unused then.
Now with active tools we use a different approach to create
toolbars, so the file is not relevant.
This node allows accessing data of other elements in the context geometry.
It is similar to the Transfer Attribute node in Index mode. The main difference
is that this node does not require a geometry input, because the context
is used.
The node can e.g. be used to generalize what the Edge Vertices node is doing.
Instead of only being able to get the position of the vertices of an edge,
any field/attribute can be accessed on the vertices.
Differential Revision: https://developer.blender.org/D13825
Adds a second output to the edge angle node that shows the signed angle
between the two faces, where Convex angles are positive and Concave angles
are negative. This calculation is slower than the unsigned angle, so it
was best to leave both for times where the unsigned angle will suffice.
Differential Revision: https://developer.blender.org/D13796
It's now easier than before to do the interpolation of attributes
only for the elements that are actually used in some cases.
This can result in a speedup because unnecessary computations
can be avoided. See the patch for a simple performance test.
Differential Revision: https://developer.blender.org/D13828
Cause of the issue isn't that clear, but the NVIDIA GLSL compiler
complained that it couldn't find an overloaded function when the second
parameter is an interger. This change fixes it by using a float.
Display exact integer values of a floating point fields without
a fraction if the step is also an exact integer. This is intended
for cases when the value can technically be fractional, but most
commonly is supposed to be integer.
This handling is not applied if the field has any unit except frames,
because integer values aren't special for quantities like length.
The fraction is discarded in the normal display mode and when copying
the value to clipboard, but not when editing to remind the user that
the field allows fractions.
Differential Revision: https://developer.blender.org/D13753
Fix a precision issue when stepping down from 1 to 0 via the left
decrement button and step 100 results in a small nonzero value.
The reason is that 0.01 can't be precisely represented in binary
and converting to double before multiplication reveals this.
Ref D13753
This patch converts GPU_SHADER_2D_IMAGE_MULTI_RECT_COLOR shader to use
the GPUShaderCreateInfo pattern. It can be used as a reference when
converting other shaders.
In this special case the flat uniform vector cannot be used anymore as it
doesn't fit as push constants. To solve this a uniform buffer is used.
The previous optimization did not work in general yet, unfortunately.
This change makes the code more correct, but also brings back
some unnecessary updates (e.g. when creating a node group).
For boolean operations only one of the meshes was checked to determine
if bevel weights should be created.
Now initialize custom data from both meshes flag.
Note that this is a localized fix to be back-ported, further changes
will be made so edit-mode conversion accounts for this
without the caller needing explicit checks for custom-data flags.
Asset indexing was disabled as ID property indexing wasn't supported.
Now that ID property support is added we can enable asset indexing.
Check {T91406} for more information about asset indexing.
Object/collection asset workflow would need the bounding box for snapping.
The bounding box is stored using ID properties in the scene. Currently ID properties
aren't stored in the asset index, what would break object snapping. For this reason
Asset Indexing is turned off in mater. This patch will introduce the indexing of ID
properties what will allow the indexing to be turned on again.
## Data Mapping ##
For data mapping we store the internal structure of IDProperty to the indexer (including meta-data) to be able to deserialize it back.
```
[
{
"name": ..,
"value": ..,
"type": ..,
/* `subtype` and `length` are only available for IDP_ARRAYs. */
"subtype": ..,
},
]
```
| **DNA** | **Serialize type** | **Note** |
| IDProperty.name | StringValue| |
| IDProperty.type | StringValue| "IDP_STRING", "IDP_INT", "IDP_FLOAT", "IDP_ARRAY", "IDP_GROUP", "IDP_DOUBLE"|
| IDProperty.subtype | StringValue| "IDP_INT", "IDP_FLOAT", "IDP_GROUP", "IDP_DOUBLE" |
| IDProperty.value | StringValue | When type is IDP_STRING |
| IDProperty.value | IntValue | When type is IDP_INT |
| IDProperty.value | DoubleValue | When type is IDP_FLOAT/IDP_DOUBLE |
| IDProperty.value | ArrayValue | When type is IDP_GROUP. Recursively uses the same structure as described in this section. |
| IDProperty.value | ArrayValue | When type is IDP_ARRAY. Each element holds a single element as described in this section. |
NOTE: IDP_ID and IDP_IDARRAY aren't supported. The entry will not be added.
Example
```
[
{
"name": "MyIntValue,
"type": "IDP_INT",
"value": 6,
},
{
"name": "myComplexArray",
"type": "IDP_ARRAY",
"subtype": "IDP_GROUP",
"value": [
[
{
"name": ..
....
}
]
]
}
]
```
## Considered alternatives ##
- Add conversion functions inside `asset_indexer`; makes generic code part of a specific solution.
- Add conversion functions inside `BLI_serialize`; would add data transformation responsibilities inside a unit that is currently only responsible for formatting.
- Use direct mapping between IDP properties and Values; leads to missing information and edge cases (empty primitive arrays) that could not be de-serialized.
Reviewed By: Severin, mont29, HooglyBoogly
Maniphest Tasks: T92306
Differential Revision: https://developer.blender.org/D12990
Motion paths themselves aren't getting saved (not sure if they are
without overrides), but being able to override options makes them
usable even if it's necessary to regenerate every edit session.
Differential Revision: https://developer.blender.org/D13842
Some new obj exporter tests were disabled because the normals were different
in the last decimal place on different platforms.
The old python exporter deduped normals with their coordinates rounded to
four decimal places. This change does the same in the new exporter.
On one test, this produced a file 25% smaller and even ran 10% faster.
This can simplify iterating through all of the indices in the vector,
which is fairly common, since one of the benefits of the data structure
is that all values are contiguous.
This node's UI uses a multi-select enum to allow adjusting the
type of both handle sides with the same node. Since usually the
user wants to affect both handles, and it's the multi-select behavior
isn't obvious, selecting both by default is an improvement.
This significantly reduces discontinuities on UV seams, by giving a better
match of the texture filtered colors on both sides of the seam. It works by
using pixels from adjacent faces across the UV seam.
This new option is called "Adjacent Faces" and is the default. The old option
is called "Extend", and extends border pixels outwards.
Differential Revision: https://developer.blender.org/D13303
Use GPU-side scaling to speed up the scaling itself, and to avoid having
to copy the image buffer using the CPU. Mipmapping is used to get decent
filtering when downscaling without ugly artifacts.
In my comparisons, there was barely any difference between the methods
for DPIs >= 1. Below that, the result looks a bit different due to the
different filtering method.
See D13144 for screen-recordings showing the difference.
Part of T92922.
Differential Revision: https://developer.blender.org/D13144
Reviewed by: Jeroen Bakker
Typo in rB605cdc4346e5f82, both `eBlendfileLinkAppendForeachItemFlag`
flags had the same value, effectively preventing to filter out direct
vs. indirect appended items.
Override layers are a standard feature of Alembic, where archives can override
data from other archives, provided that the hierarchies match.
This is useful for modifying a UV map, updating an animation, or even creating
some sort of LOD system where low resolution meshes are swapped by high resolution
versions.
It is possible to add UV maps and vertex colors using this system, however, they
will only appear in the spreadsheet editor when viewing evaluated data, as the UV
map and Vertex color UI only show data present on the original mesh.
Implementation wise, this adds a `CacheFileLayer` data structure to the `CacheFile`
DNA, as well as some operators and UI to present and manage the layers. For both
the Alembic importer and the Cycles procedural, the main change is creating an
archive from a list of filepaths, instead of a single one.
After importing the base file through the regular import operator, layers can be added
to or removed from the `CacheFile` via the UI list under the `Override Layers` panel
located in the Mesh Sequence Cache modifier. Layers can also be moved around or
hidden.
See differential page for tests files and demos.
Reviewed by: brecht, sybren
Differential Revision: https://developer.blender.org/D13603
This is a first part of the Shader Create Info system could be.
A shader create info provides a way to define shader structure, resources
and interfaces. This makes for a quick way to provide backend agnostic
binding informations while also making shader variations easy to declare.
- Clear source input (only one file). Cleans up the GPU api since we can create a
shader from one descriptor
- Resources and interfaces are generated by the backend (much simpler than parsing).
- Bindings are explicit from position in the array.
- GPUShaderInterface becomes a trivial translation of enums and string copy.
- No external dependency to third party lib.
- Cleaner code, less fragmentation of resources in several libs.
- Easy to modify / extend at runtime.
- no parser involve, very easy to code.
- Does not hold any data, can be static and kept on disc.
- Could hold precompiled bytecode for static shaders.
This also includes a new global dependency system.
GLSL shaders can include other sources by using #pragma BLENDER_REQUIRE(...).
This patch already migrated several builtin shaders. Other shaders should be migrated
one at a time, and could be done inside master.
There is a new compile directive `WITH_GPU_SHADER_BUILDER` this is an optional
directive for linting shaders to increase turn around time.
What is remaining:
- pyGPU API {T94975}
- Migration of other shaders. This could be a community effort.
Reviewed By: jbakker
Maniphest Tasks: T94975
Differential Revision: https://developer.blender.org/D13360
For an upcoming refactoring of library remapping we want to be able to test if the logic won't change.
It also increased my experience inside the remapping codebase and find out what exactly needed to
be refactored.
This patch adds test cases for the core functionality of `foreach_libblock_remap_callback`. The test cases
don't cover of all the branches. Also pre-, post-processing, referencing and proxies are not tested.
Reviewed By: mont29
Differential Revision: https://developer.blender.org/D13815
Allows conveniently selecting an inverse of a collection.
Reviewed By: Antonio Vazquez (antoniov)
Differential Revision: https://developer.blender.org/D13846
This patch improves conversion method from NURBS to Bezier curves,
resulting in exact shape between those two types when provided with
a 3rd degree NURBS curve. Part of T86086.
See the differential revision for more comparisons.
The node still cannot account properly for a NURBS "order" other
than 4 and it does not take into account control point weights.
Differential Revision: https://developer.blender.org/D13546
Fix the description for WORKSPACE_OT_reorder_to_back to say "last" in
list rather than "first"
See D13696 for more details.
Differential Revision: https://developer.blender.org/D13696
Reviewed by Aaron Carlisle
The crash is due to the fact that GPU subdivision extraction routines
for edit data (including UVs) only worked for BMesh. However, a Mesh
based version is still needed for texture painting. This adds the
missing components. This also ensures all data are properly initialized
(at least the ones revealed by the bug).
This puts the loop over the final subdivision quads outside of the mesh
iteration callback. This can also allow for easier parallel execution in
the future if need be.
Since the option to enable linkers are booleans,
it's possible to enable them all at once.
Now only the first enabled + available linker is used
(with priority given to link is with better performance).
Set the linker using CMAKE_*_LINKER_FLAGS instead of {C/CXX}FLAGS.
There is no advantage in using the CFLAGS to set the linker, it has the
downside of triggering a full rebuild when changing the linker.
Tested building Blender and the bpy.so Python module.
Ref D13833
Reviewed by: sergey, brecht
Conceptually, this is the geometry that data is taken from,
not the target of an operation, so rename it from "Target"
to "Source". This was common user feedback and agreed
on in a recent sub-module meeting.
Before rB644e6c7a3e99ae1d43ed, `fill` was used in the error
cases, but now `fill_indices` is used, which doesn't work when
the span is empty (when only one output is used). The fix is just
to check for that case.
The search list only displayed the "Result" output socket in this
case, which is unexpected since dragging from an input gives the
operations in the list as well. Also use integer mode when
connecting to boolean sockets.
I noticed these when doing final cleanup on rBcfa53e0fbeed.
One use was removed in that commit, the others were unused
going further back a few years.
Differential Revision: https://developer.blender.org/D13834
`TreeElement` isn't a trivial type anymore, so `MEM_delete()` should be
called, which calls the destructor.
AFAICS this would cause a memory leak, since the contained `unique_ptr`
is allocated but not destructed correctly - but it's not using the
guarded allocator so woudn't be reported.
Smart pointers should be the default choice for C++ owning pointers,
since they let you manage memory using RAII.
Also moved type factory methods into static class functions.
Basically this removes any C <-> C++ glue code. C++ types are accessed
directly via the public C++ APIs.
Contains some related changes like, moving functions that were
previously declared in a now removed header to a different file, whose
header is the more appropriate place (and the source file as well).
But generally I tried to avoid other changes.
The switch to how normals are kept has led to tiny differences in
the normal output values on different platforms. Disabling the failing
tests while working on a solution to this problem.
Part of a5cb7c1e62 is reverted since it
created unknown pragma warning on windows.
Use a trick to do self-assigning.
Reviewed by Jacques Lucke in chat.
`BKE_layer_collection_sync` was missing a specific handling for one of
those pre-master collection cases,
NOTE: It is a bit unfortunate to have to do 'do-version' code in BKE...
At some point might look into moving this into actual `do_version` file,
but this is not fully trivial not critical improvement for now.
Caused by rBa5c59fb90ef9.
Since Group Input and Output sockets happen to be of type `SOCK_CUSTOM`
[and since rBa5c59fb90ef9 custom py defined sockets are too :)] a check
introduced in rB513066e8ad6f that prevents connections for `SOCK_CUSTOM`
triggered.
Now refine the check, so it specifically looks for NODE_GROUP_INPUT /
NODE_GROUP_OUTPUT, too (this keeps the intention intact to not connect
group inputs to group outputs and vice versa, but allows custom py
defined sockets to connect again) and put it in new utility function.
Maniphest Tasks: T94827
Differential Revision: https://developer.blender.org/D13817
Blender.xcodeproj User-supplied CFBundleIdentifier value
'org.blenderfoundation.blender' in the Info.plist must be the same as
the PRODUCT_BUNDLE_IDENTIFIER build setting value ''.
Reviewed By: #platform_macos, brecht
Differential Revision: https://developer.blender.org/D13826
Didn't remove the key-value pair since old Xcode behavior is not
known.
warning: LSMinimumSystemVersion of '10.9.0' is less than the value of
MACOSX_DEPLOYMENT_TARGET '10.13' - setting to '10.13'. (in target
'blender' from project 'Blender')
Reviewed By: #platform_macos, brecht
Differential Revision: https://developer.blender.org/D13831
Fix assignment warning
source/blender/blenlib/tests/BLI_any_test.cc:56:5: warning: explicitly
assigning value of variable of type 'blender::Any<void, 8, 8>'
to itself [-Wself-assign-overloaded]
c = c;
Reviewed By: JacquesLucke
Differential Revision: https://developer.blender.org/D13835
2-point-curves are treated separately from 3plus-point-curves (assume a
lot of the twisting reduction can be skipped, so there is a dedicated
function for single segment curves).
And while using the 3plus-point-curves function [`make_bevel_list_3D`]
would actually work in this case, the dedicated function
`make_bevel_list_segment_3D` would only consider the tilt of the second
point and would just copy over the quat to the first point as well. Dont
see a reason for this, now consider the first point's tilt as well.
Maniphest Tasks: T94837
Differential Revision: https://developer.blender.org/D13813
Can give considerably faster linking, especially on system with many
cores.
The mold linker recently reached 1.0, see:
https://github.com/rui314/mold
The current stable release of GCC can't use this linker via
-fuse-ld=mold, so this patch uses the "-B" argument to add a binary
directory containing an alternate "ld" command that points to
"mold" (which is part of the default mold installation).
Some timing tests for linking full builds for AMD TR 3970X:
- BFD: 20.78 seconds.
- LLD: 12.16 seconds.
- GOLD: 7.21 seconds.
- MOLD: 2.53 seconds.
Ref D13807
Reviewed by: sergey, brecht
The mask is only used if it's not zero. Adding the normal mask made
it not zero, but it didn't include anything else, so all custom data
layers except normals were removed. The fix is to only add normals
to the mask when it should be used.
As described in T91186, this commit moves mesh vertex normals into a
contiguous array of float vectors in a custom data layer, how face
normals are currently stored.
The main interface is documented in `BKE_mesh.h`. Vertex and face
normals are now calculated on-demand and cached, retrieved with an
"ensure" function. Since the logical state of a mesh is now "has
normals when necessary", they can be retrieved from a `const` mesh.
The goal is to use on-demand calculation for all derived data, but
leave room for eager calculation for performance purposes (modifier
evaluation is threaded, but viewport data generation is not).
**Benefits**
This moves us closer to a SoA approach rather than the current AoS
paradigm. Accessing a contiguous `float3` is much more efficient than
retrieving data from a larger struct. The memory requirements for
accessing only normals or vertex locations are smaller, and at the
cost of more memory usage for just normals, they now don't have to
be converted between float and short, which also simplifies code
In the future, the remaining items can be removed from `MVert`,
leaving only `float3`, which has similar benefits (see T93602).
Removing the combination of derived and original data makes it
conceptually simpler to only calculate normals when necessary.
This is especially important now that we have more opportunities
for temporary meshes in geometry nodes.
**Performance**
In addition to the theoretical future performance improvements by
making `MVert == float3`, I've done some basic performance testing
on this patch directly. The data is fairly rough, but it gives an idea
about where things stand generally.
- Mesh line primitive 4m Verts: 1.16x faster (36 -> 31 ms),
showing that accessing just `MVert` is now more efficient.
- Spring Splash Screen: 1.03-1.06 -> 1.06-1.11 FPS, a very slight
change that at least shows there is no regression.
- Sprite Fright Snail Smoosh: 3.30-3.40 -> 3.42-3.50 FPS, a small
but observable speedup.
- Set Position Node with Scaled Normal: 1.36x faster (53 -> 39 ms),
shows that using normals in geometry nodes is faster.
- Normal Calculation 1.6m Vert Cube: 1.19x faster (25 -> 21 ms),
shows that calculating normals is slightly faster now.
- File Size of 1.6m Vert Cube: 1.03x smaller (214.7 -> 208.4 MB),
Normals are not saved in files, which can help with large meshes.
As for memory usage, it may be slightly more in some cases, but
I didn't observe any difference in the production files I tested.
**Tests**
Some modifiers and cycles test results need to be updated with this
commit, for two reasons:
- The subdivision surface modifier is not responsible for calculating
normals anymore. In master, the modifier creates different normals
than the result of the `Mesh` normal calculation, so this is a bug
fix.
- There are small differences in the results of some modifiers that
use normals because they are not converted to and from `short`
anymore.
**Future improvements**
- Remove `ModifierTypeInfo::dependsOnNormals`. Code in each modifier
already retrieves normals if they are needed anyway.
- Copy normals as part of a better CoW system for attributes.
- Make more areas use lazy instead of eager normal calculation.
- Remove `BKE_mesh_normals_tag_dirty` in more places since that is
now the default state of a new mesh.
- Possibly apply a similar change to derived face corner normals.
Differential Revision: https://developer.blender.org/D12770
Some of the message-bus macros are not safe to use in C++. This has come
up before, but no good solution was found. Now @LazyDodo, @HooglyBoogly
and I concluded this is the best duct tape "solution" for the moment.
The message-bus API should address this.
Today many users seem to think the output from
this node is a single curve with multiple splines.
This patch renames the geometry output socket
from "Curves" to "Curve Instances" to avoid confusion.
Differential Revision: https://developer.blender.org/D13693
We want to refactor quite some of the Outliner code using C++, this is a
logical step to help the transition to a new architecture.
Includes plenty of fixes to make this compile without warnings, trying
not to change logic. The usual stuff (casts from `void *`, designated
initializers, compound literals, etc.).
There were a couple of function name collisions which were caused
by sharing code with the mask modifier. I just removed the dependence
on the mask modifier now. The code that I duplicated for that purpose
is only in a legacy node, so it can be expected to be removed soonish.
And change install_deps.sh to build shared (instead of static) FFMPEG
libraries, for consistency with other library dependencies and to simplify
the logic. This may require users of install_deps.sh to rebuild FFMPEG.
This is the last step that lets us get rid of LIBPATH variables and
link_directories() entirely, as recommended by the CMake docs.
Some fixes were needed in the find FFMPEG module to make it actually work,
this code was unused up to now.
Followup to D8855.
Differential Revision: https://developer.blender.org/D9177
The fundamental limitation is that we can only have one instance
("dupli") generator at a time. Because the mesh output of a curve
object is output as an instances, the geometry set instances existed,
replacing the object as font instances. The "fix" is to reverse the
order. The behavior won't be perfect still, but at least the old
behavior will be preserved, which is really what matters for a
feature like this.
One way to take this change further would be completely disabling
regular geometry evaluation while this option is active. However,
it doesn't seem like that would actually improve the state of the code.
Differential Revision: https://developer.blender.org/D13768
When drawing windows on monitors that differ in DPI, we can sometimes
have UI elements draw at an incorrect scale. This patch just ensures
that `wm_window_make_drawable` always updates DPI.
See D10483 for more details.
Differential Revision: https://developer.blender.org/D10483
Reviewed by Brecht Van Lommel
Allow area Split to be initiated in any area and give better feedback
when not allowed.
See D13599 for more details and usage examples.
Differential Revision: https://developer.blender.org/D13599
Reviewed by Campbell Barton
And change install_deps.sh to build shared (instead of static) FFMPEG
libraries, for consistency with other library dependencies and to simplify
the logic. This may require users of install_deps.sh to rebuild FFMPEG.
This is the last step that lets us get rid of LIBPATH variables and
link_directories() entirely, as recommended by the CMake docs.
Some fixes were needed in the find FFMPEG module to make it actually work,
this code was unused up to now.
Followup to D8855.
Differential Revision: https://developer.blender.org/D9177
This is my attempt of adding defaults for the space clip editor struct
(in line with https://developer.blender.org/T80164).
It adds the default allocation for `SpaceClip` and
`node_composite_movieclip.cc`. This also solves the error below (for
C++ files using the DNA_default_alloc), which was put forward by
Sergey Sharybin.
Differential Revision: https://developer.blender.org/D13367
Reviewed by: Julian Eisel
While theorically fairly generic, current code is only enabled for
bledfile and liboverride views, and only used to display messages from
library IDs.
Reviewed By: Severin
Differential Revision: https://developer.blender.org/D13766
SVG files contained specific detailed pathnames on developers'
computers. These included full local user profile and path and should
not be in the release.
This patches corrects those lines. It also removes unused gradients from
the private icons SVG.
Differential Revision: https://developer.blender.org/D13344
Reviewed by: Yevgeny Makarov, Julian Eisel
This adds wrapper classes that make it easier to use GPU objects in C++.
####Motivations:####
- Easier handling of GPU objects.
- EEVEE rewrite already makes use of similar wrappers.
- There is the ongoing effort to use more C++ in the codebase
and lans to port more engines to it.
- The shader code refactor will make use of many UBOs with shared
struct declaration. This helps managing them.
- Safer handling of `TextureFromPool` which can't be bound as normal
texture (only texture ref) and can be better tracked in the future.
####Considerations:####
- I chose the `blender::draw` namespace because `blender::gpu` already has private classes (i.e: `gpu::Texture`).
- Theses are wrappers that manage a GPU object internally. They might be confused with actual `Texture`. However, the name `TextureWrapper` is a bit too much verbose in my opinion. I'm open to suggestion about better name.
Reviewed By: jbakker
Differential Revision: http://developer.blender.org/D13805
This patch implements the vector types (i.e:`float2`) by making heavy
usage of templating. All vector functions are now outside of the vector
classes (inside the `blender::math` namespace) and are not vector size
dependent for the most part.
In the ongoing effort to make shaders less GL centric, we are aiming
to share more code between GLSL and C++ to avoid code duplication.
####Motivations:
- We are aiming to share UBO and SSBO structures between GLSL and C++.
This means we will use many of the existing vector types and others
we currently don't have (uintX, intX). All these variations were
asking for many more code duplication.
- Deduplicate existing code which is duplicated for each vector size.
- We also want to share small functions. Which means that vector
functions should be static and not in the class namespace.
- Reduce friction to use these types in new projects due to their
incompleteness.
- The current state of the `BLI_(float|double|mpq)(2|3|4).hh` is a
bit of a let down. Most clases are incomplete, out of sync with each
others with different codestyles, and some functions that should be
static are not (i.e: `float3::reflect()`).
####Upsides:
- Still support `.x, .y, .z, .w` for readability.
- Compact, readable and easilly extendable.
- All of the vector functions are available for all the vectors types
and can be restricted to certain types. Also template specialization
let us define exception for special class (like mpq).
- With optimization ON, the compiler unroll the loops and performance
is the same.
####Downsides:
- Might impact debugability. Though I would arge that the bugs are
rarelly caused by the vector class itself (since the operations are
quite trivial) but by the type conversions.
- Might impact compile time. I did not saw a significant impact since
the usage is not really widespread.
- Functions needs to be rewritten to support arbitrary vector length.
For instance, one can't call `len_squared_v3v3` in
`math::length_squared()` and call it a day.
- Type cast does not work with the template version of the `math::`
vector functions. Meaning you need to manually cast `float *` and
`(float *)[3]` to `float3` for the function calls.
i.e: `math::distance_squared(float3(nearest.co), positions[i]);`
- Some parts might loose in readability:
`float3::dot(v1.normalized(), v2.normalized())`
becoming
`math::dot(math::normalize(v1), math::normalize(v2))`
But I propose, when appropriate, to use
`using namespace blender::math;` on function local or file scope to
increase readability.
`dot(normalize(v1), normalize(v2))`
####Consideration:
- Include back `.length()` method. It is quite handy and is more C++
oriented.
- I considered the GLM library as a candidate for replacement. It felt
like too much for what we need and would be difficult to extend / modify
to our needs.
- I used Macros to reduce code in operators declaration and potential
copy paste bugs. This could reduce debugability and could be reverted.
- This touches `delaunay_2d.cc` and the intersection code. I would like
to know @howardt opinion on the matter.
- The `noexcept` on the copy constructor of `mpq(2|3)` is being removed.
But according to @JacquesLucke it is not a real problem for now.
I would like to give a huge thanks to @JacquesLucke who helped during this
and pushed me to reduce the duplication further.
Reviewed By: brecht, sergey, JacquesLucke
Differential Revision: https://developer.blender.org/D13791
This patch implements the vector types (i.e:`float2`) by making heavy
usage of templating. All vector functions are now outside of the vector
classes (inside the `blender::math` namespace) and are not vector size
dependent for the most part.
In the ongoing effort to make shaders less GL centric, we are aiming
to share more code between GLSL and C++ to avoid code duplication.
####Motivations:
- We are aiming to share UBO and SSBO structures between GLSL and C++.
This means we will use many of the existing vector types and others
we currently don't have (uintX, intX). All these variations were
asking for many more code duplication.
- Deduplicate existing code which is duplicated for each vector size.
- We also want to share small functions. Which means that vector
functions should be static and not in the class namespace.
- Reduce friction to use these types in new projects due to their
incompleteness.
- The current state of the `BLI_(float|double|mpq)(2|3|4).hh` is a
bit of a let down. Most clases are incomplete, out of sync with each
others with different codestyles, and some functions that should be
static are not (i.e: `float3::reflect()`).
####Upsides:
- Still support `.x, .y, .z, .w` for readability.
- Compact, readable and easilly extendable.
- All of the vector functions are available for all the vectors types
and can be restricted to certain types. Also template specialization
let us define exception for special class (like mpq).
- With optimization ON, the compiler unroll the loops and performance
is the same.
####Downsides:
- Might impact debugability. Though I would arge that the bugs are
rarelly caused by the vector class itself (since the operations are
quite trivial) but by the type conversions.
- Might impact compile time. I did not saw a significant impact since
the usage is not really widespread.
- Functions needs to be rewritten to support arbitrary vector length.
For instance, one can't call `len_squared_v3v3` in
`math::length_squared()` and call it a day.
- Type cast does not work with the template version of the `math::`
vector functions. Meaning you need to manually cast `float *` and
`(float *)[3]` to `float3` for the function calls.
i.e: `math::distance_squared(float3(nearest.co), positions[i]);`
- Some parts might loose in readability:
`float3::dot(v1.normalized(), v2.normalized())`
becoming
`math::dot(math::normalize(v1), math::normalize(v2))`
But I propose, when appropriate, to use
`using namespace blender::math;` on function local or file scope to
increase readability.
`dot(normalize(v1), normalize(v2))`
####Consideration:
- Include back `.length()` method. It is quite handy and is more C++
oriented.
- I considered the GLM library as a candidate for replacement. It felt
like too much for what we need and would be difficult to extend / modify
to our needs.
- I used Macros to reduce code in operators declaration and potential
copy paste bugs. This could reduce debugability and could be reverted.
- This touches `delaunay_2d.cc` and the intersection code. I would like
to know @howardt opinion on the matter.
- The `noexcept` on the copy constructor of `mpq(2|3)` is being removed.
But according to @JacquesLucke it is not a real problem for now.
I would like to give a huge thanks to @JacquesLucke who helped during this
and pushed me to reduce the duplication further.
Reviewed By: brecht, sergey, JacquesLucke
Differential Revision: https://developer.blender.org/D13791
Fixes issue T94603
It adds a new compositor node called Scene Time which is already present as a geo node, having the same basic nodes available in all node trees is a nice thing to have.
Renames "Time" node to "Time Curve", this is done to avoid confusion between the Time node and the Scene Time node.
Reviewed By: jbakker
Maniphest Tasks: T94603
Differential Revision: https://developer.blender.org/D13762
This patch implements the vector types (i.e:float2) by making heavy
usage of templating. All vector functions are now outside of the vector
classes (inside the blender::math namespace) and are not vector size
dependent for the most part.
In the ongoing effort to make shaders less GL centric, we are aiming
to share more code between GLSL and C++ to avoid code duplication.
Motivations:
- We are aiming to share UBO and SSBO structures between GLSL and C++.
This means we will use many of the existing vector types and others we
currently don't have (uintX, intX). All these variations were asking
for many more code duplication.
- Deduplicate existing code which is duplicated for each vector size.
- We also want to share small functions. Which means that vector functions
should be static and not in the class namespace.
- Reduce friction to use these types in new projects due to their
incompleteness.
- The current state of the BLI_(float|double|mpq)(2|3|4).hh is a bit of a
let down. Most clases are incomplete, out of sync with each others with
different codestyles, and some functions that should be static are not
(i.e: float3::reflect()).
Upsides:
- Still support .x, .y, .z, .w for readability.
- Compact, readable and easilly extendable.
- All of the vector functions are available for all the vectors types and
can be restricted to certain types. Also template specialization let us
define exception for special class (like mpq).
- With optimization ON, the compiler unroll the loops and performance is
the same.
Downsides:
- Might impact debugability. Though I would arge that the bugs are rarelly
caused by the vector class itself (since the operations are quite trivial)
but by the type conversions.
- Might impact compile time. I did not saw a significant impact since the
usage is not really widespread.
- Functions needs to be rewritten to support arbitrary vector length. For
instance, one can't call len_squared_v3v3 in math::length_squared() and
call it a day.
- Type cast does not work with the template version of the math:: vector
functions. Meaning you need to manually cast float * and (float *)[3] to
float3 for the function calls.
i.e: math::distance_squared(float3(nearest.co), positions[i]);
- Some parts might loose in readability:
float3::dot(v1.normalized(), v2.normalized())
becoming
math::dot(math::normalize(v1), math::normalize(v2))
But I propose, when appropriate, to use
using namespace blender::math; on function local or file scope to
increase readability. dot(normalize(v1), normalize(v2))
Consideration:
- Include back .length() method. It is quite handy and is more C++
oriented.
- I considered the GLM library as a candidate for replacement.
It felt like too much for what we need and would be difficult to
extend / modify to our needs.
- I used Macros to reduce code in operators declaration and potential
copy paste bugs. This could reduce debugability and could be reverted.
- This touches delaunay_2d.cc and the intersection code. I would like to
know @Howard Trickey (howardt) opinion on the matter.
- The noexcept on the copy constructor of mpq(2|3) is being removed.
But according to @Jacques Lucke (JacquesLucke) it is not a real problem
for now.
I would like to give a huge thanks to @Jacques Lucke (JacquesLucke) who
helped during this and pushed me to reduce the duplication further.
Reviewed By: brecht, sergey, JacquesLucke
Differential Revision: http://developer.blender.org/D13791
The issue was caused by rBd09b1d2759861aa012ab2e7e4ce2ffa2.
Since this commit, the image users in gpu materials were updated
during depsgraph evaluation as well. However, there was a race
condition when one thread is deleting gpu materials in `BKE_material_eval`
while another thread is updating the image users at the same time.
The solution is to make sure that deleting gpu materials is done before
iterating over all gpu materials, by adding a new depsgraph relation.
The main issue was the use of `G_MAIN` during file load.
This patch refactors the code so that iterating over `G_MAIN`
is not necessary anymore. See D13800 for more details.
Differential Revision: https://developer.blender.org/D13800
The issue was caused by Cycles display driver not being able to restore
window's OpenGL context after disposing Cycles-side OpenGL context.
This is due to the window OpenGL re-activation needing to access window
manager which gets cleared out form global main during file reading.
Defer clearing window manager from the global main to until after all
screens are "exited". This allows Cycles to properly stop rendering,
dispose its OpenGL context, and restore window's drawable context.
It is unclear why it was required to clear window manager list early
on. Guess is that it comes from an original code in a1c8543f2a where
there was an early return which then got replaced with an actual logic
without changing the order of de-initialization and window manager list
clear.
Differential Revision: https://developer.blender.org/D13799
The direct cause of the bug in question was passing in the raw memory
buffer to sscanf. It should be called with a null-terminated buffer;
which isn't guaranteed when blindly trusting the file data.
When attempting to fuzz this code path, a variety of other crashes were
discovered and fixed.
Differential Revision: https://developer.blender.org/D11952
Remove code that very slightly darkened line on bottom of timeline, when
backdrop is enabled. Purpose of the code wasn't dodumented, and 2.79
doesn't seem to produce this darkened line.
Rename drawing functions to appropriate names.
Add maximum string length argument to UI_fontstyle_draw to reduce usage
of BLF_DRAW_STR_DUMMY_MAX. Reorders arguments to UI_fontstyle_draw_ex
See D13794 for more details.
Differential Revision: https://developer.blender.org/D13794
Reviewed by Campbell Barton
Reduction of the number of uses of the define BLF_DRAW_STR_DUMMY_MAX
by using actual sizes of static character arrays.
See D13793 for more details.
Differential Revision: https://developer.blender.org/D13793
Reviewed by Campbell Barton
Blender's compositor code already makes extensive use of
namespace which makes it very simple to enable unity build.
There was one duplicated function that has since to be moved
to a common header.
I saw roughly a 3x speedup of bf_compositor using ninja on
linux using i5 8250u (1:34 down to 0:34).
Reviewed By: LazyDodo
Differential Revision: https://developer.blender.org/D13792
With this change, compilation saw a 2.4x improvement.
This can be combined with unity build to give an overall 4x improvement
Depends on D13797
Reviewed By: LazyDodo
Differential Revision: https://developer.blender.org/D13798
Removal of unused tmp member of GHOST_TEventImeData. Not used now,
nor was it used by the commit that added it to begin with.
Differential Revision: https://developer.blender.org/D11799
Reviewed by Ray Molenkamp
our UNUSED macro is essentially a no-op for MSVC, which lead to
the situation where this well meant macro was emitting the
following warning:
C4189: 'UNUSED_i': local variable is initialized but not referenced
However since we have been on c++17 for a while now the UNUSED
macro can be replaced with the standard [[maybe_unused]] attribute
in cpp files.
This changes cleans up the use of the UNUSED macro in the
bf_nodes_geometry project.
Differential Revision: https://developer.blender.org/D12915
Reviewed by: JacquesLucke, Severin, Sergey, HooglyBoogly
Since CMake 3.16, CMake has native precompiled header (PCH) support.
This change swaps Blender's own PCH implementation with the native implementation.
Previously, PCH was only enabled on Windows however,
this new implementation works on all platforms.
For more information see https://cmake.org/cmake/help/latest/command/target_precompile_headers.html
On my system, Linux with ninja running on an i5 8250U
I saw a 60% reduction in compile times for `bf_freestyle` + linking time.
Reviewed By: LazyDodo, brecht
Differential Revision: https://developer.blender.org/D13797
The problem was the points were selected in edit mode and then sampled. Now, in draw mode, the points are always unselected to avoid this effect in the auto merge process.
The new triangulation mode for quads is the opposite of the current default
shortest diagonal mode. It is optimal for cloth simulations using quad meshes.
Differential Revision: http://developer.blender.org/D13777
Issue introduced in rB1d49293b80446b89b5b12fa0eeefaf14e5051e48
`drw_manager_init` must be called after `drw_context_state_init` as
`DST.draw_ctx.sh_cfg` (indicating when the view is clipped) must be set
first.
Differential Revision: https://developer.blender.org/D13795
After disconnecting hair on an object, if you then hide the particle system, and try connecting the hair again, the operator is cancelled due to `remap_hair_emitter` returning `false` because `target_psmd->mesh_final` is NULL, but `connect_hair` will still strip the `PSYS_GLOBAL_HAIR` flag, which will cause the hair in the hidden particle system to be positioned incorrectly. The correct behavior is to strip the flag only if `remap_hair_emitter` succeeds.
Differential Revision: https://developer.blender.org/D13703
When the 'threshold' is not used in the type we are comparing, just hide
it. This was obvious for some types (e.g. Materials), but maybe not so
on others (e.g. Polygon Sides) and potentionally confusing.
Reported by @hitrpr in chat.
Differential Revision: https://developer.blender.org/D13760
This is an update to the correct OCIO role.
It changes `SceneReference` to `scene_linear`
See https://opencolorio.readthedocs.io/en/latest/guides/authoring/overview.html#config-roles
> - reference - the color space against which the other color spaces are defined
>NOTE: The reference role has sometimes been misinterpreted as being the space in which “reference art” is stored in.
>
> - scene_linear - the scene-referred linear-to-light color space, often the same as the reference space
The current OCIO UX working group doc says:
>reference: This role has had multiple interpreted meanings over the years and is a common point of confusion. It is kept in OCIO for backwards compatibility, but the recommendation is that it is not used by apps.
Reviewed By: jbakker
Differential Revision: https://developer.blender.org/D11398
This puts all static functions in composite node files into a new
namespace. This allows using unity build which can improve
compile times significantly.
This is a follow up on rB1df8abff257030ba79bc23dc321f35494f4d91c5
but for compositor nodes.
The namespace name is derived from the file name.
That makes it possible to write some tooling that checks the names later on.
The filename extension (`cc`) is added to the namespace name as well.
his also possibly simplifies tooling but also makes it more obvious that this namespace is specific to a file.
Reviewed By: JacquesLucke, HooglyBoogly, jbakker
Differential Revision: https://developer.blender.org/D13466
If timeline contains scene strip outside of edited meta strip, this will
cause crash. This is because prefetchin ignored meta strips being edited
when rendering, but did check for scene strips only inside edited meta
strip.
Change active seqbase pointer when entering meta strip. This makes it
possible to prefetch only content that is being presented to user.
rBeed45d2a239a introduced a GPU backend for OpenSubDiv which lets us do
the subdivision at render time. However, some tools might still need to
have the subdivision data available on the CPU side. For this a
subdivision mesh wrapper was also introduced, and is computed whenever a
CPU side mesh is needed. The subdivision settings for this wrapper are
stored during modifier evaluation if GPU subdivision can be done.
The performance regression is due to the fact that although the
subdivision mesh was already computed on the CPU, and no subdivision
wrapper is generated, some checks for creating subdivision data in
`BKE_mesh_wrapper_ensure_subdivision` where still run, one of which is
very expensive.
To fix this we first check the runtime settings of the mesh to see if
subdivision is needed at all.
This commit adds topology information from mesh data structs to the
spreadsheet when the debug value `4001` is set. Eventually we could
expose these. For now it can be a useful tool for developers when
working on mesh algorithms.
Differential Revision: https://developer.blender.org/D13735
The value of this flag was only retrieved in `nodeGetActiveID`, which
wasn't used anywhere. Other than that, the `NODE_ACTIVE_ID` and
related functions seem to come from the Blender internal renderer.
Differential Revision: https://developer.blender.org/D13770
This commit moves the normal field input to `BKE_geometry_set.hh`
from the node file so that normals can be used as an implicit input to
other nodes.
Differential Revision: https://developer.blender.org/D13779
Slight change to our processing of Ctrl-C, Ctrl-V, and Ctrl-X so that
they will not be triggered if Alt is also pressed. This allows entry
of AltGr-C, -V, -X when using International keyboard layouts.
See D13781 for more details
Differential Revision: https://developer.blender.org/D13781
Reviewed by Brecht Van Lommel
Treat "/" as a key that should be evaluated by the Win IME system when
the input language is Chinese. This fixes a duplication of the input
character and results in the expected output of a Chinese wide comma.
See D13771 for more details.
Differential Revision: https://developer.blender.org/D13771
Reviewed by Brecht Van Lommel
Consider temporary directory to be variant part of session configuration
which gets communicated to the tile manager on render reset.
This allows to be able to render with one temp directory, change the
directory, render again and have proper render result even with enabled
persistent data.
For the ease of access to the temp directory expose it via the render
engine API (engine.temp_directory).
Differential Revision: https://developer.blender.org/D13790
The issue was caused by the recent changes in the way how the
render result is drawn: the display driver now could hold an
OpenGL resources. Those resources are not shared across contexts
so whenever OpenGL context is destroyed those resources are to
be destroyed as well (and not attempted to be re-used for a next
render).
Do such destruction and entire driver re-creation since it does
simplifies things from API usage point of view without causing
measurable slowdown.
Steps to reproduce the issue:
- Set the render resolution to 2x of Full HD
- Enable persistent data
- Render (F12)
- Render again
Observe OpenGL state being corrupted. Easy to see in debug mode
where IMM abstraction level reports issues about the buffer size
not being the proper size. This was caused by the display driver
trying to use VAO from the previous OpenGL context.
Differential Revision: https://developer.blender.org/D13789
The core issue is that flushing dependencies are created from an object
to a node tree when it contains e.g. a Texture Coordinate node.
That is an issue because the evaluation of the node tree itself does not
depend on the object (node tree evaluation is essentially a no-op).
Only other systems that parse and evaluate the node tree in a specific
context actually depend on e.g. the position of the referenced object.
It can even be the case that the node tree depends on objects that
the actual evaluator (geometry nodes modifier/material) does not depend
on, because a node is not connected to the output.
Geometry nodes makes the distinction between dependencies to the
node tree and to the evaluator already. Shader nodes do not.
Therefore, shader nodes need a flushing relation from node groups
to their parent node groups.
This brings back some unnecessary updates from rB7e712b2d6a0d
(e.g. when creating a node group from nodes that are not connected
to the output). This is a bit unfortunate, but refactoring how
dependencies work with shader nodes is a out of scope for this fix.
Since rBf9ccd26b037d, calling `data.path_resolve()` on custom properties
with `None` value do not cause a `ValueError` exception any more. This
is now taken into account in the keying sets targeting custom
properties.
Reviewed By: sybren
Differential Revision: https://developer.blender.org/D13787
rBd6891d9bee2b introduced a way to apply a single constraint from the
constraint stack. For this we want to work in the evaluated domain, in
particular the constraint target should be evaluated (the shrinkwrap
constraint needs to have access to the target's evaluated mesh).
Thx a lot to @sergey for handholding here!
Maniphest Tasks: T94600
Differential Revision: https://developer.blender.org/D13765
This moves the clear paths button ("X") to the same line of "Update All Paths",
and make it visible at all times.
1. The clear button affects all objects (by default). However the
Calculate/Update Paths only works on the selected objects/objects.
Better to not have them both on the same line.
2. The operator to clear object and pose paths can run even if the active
object/bone has no motion path. However the UI was not showing the button in
those cases.
Before:
{F12757500, size=full}
After:
{F12757502, size=full}
Differential Revision: https://developer.blender.org/D13609
Compositor node to convert between color spaces.
Conversion is skipped when converting between the same color spaces or to or from data spaces.
Implementation done for tiled and full frame compositor.
Reviewed By: Blendify, jbakker
Differential Revision: https://developer.blender.org/D12481
Switched populating GHOST_WintabInfoWin32 vector from resizing and
assigning to reserving and pushing.
Removed unnecessary state tracking for multiple button presses in a
single packet.
Paired initialization with definition, and added default initialization
for GHOST_WintabInfoWin32.
Currently the node link ui template only works with a few socket types.
This commit addes support for the rest of the socket type declarations.
As pointed out in D13776 currently after recent refactors
Shader nodes no longer display in the menu.
In the future more socket types will be used in the shader nodes
and makes the UI template work better for other node trees.
Differential Revision: https://developer.blender.org/D13778
Weightpaint gradient tool panel showed in other modes (and as a separate
panel).
Fix for fix, see
- rBf8a0e102cf5e
- rBe549d6c1bd2d
So now, check mode again and restrict to topbar (prevents an additional
panel since this is already included in the brush settings).
ref rB0837926740b3 in sculpt-dev branch, so thx @joeedh as well!
Maniphest Tasks: T94243
Differential Revision: https://developer.blender.org/D13630
Currently, most node buttons are defined in `drawnode.cc` however,
this is inconvenient because it requires editing many files when adding new nodes.
The goal is to minimize the number of files needed to add or update a node.
This commit moves most of the node layout functions for shader nodes into their respected
source/blender/nodes/shader/nodes file.
In the future, these functions will be simplified to node_layout.
Some nodes were left in `drawnode.cc` as this would require duplicating code
while this is likely fine it is best to leave that to a seperate commit.
Some software or processing tools (videogrammetry in this case) may
export malformed files with velocity data even when the frame is empty
for some reason. We need to explicity compare the data size with the
vertex size, and refuse to load the attribute if there is a data size
mismatch.
The dangling pointer caused errors further down the line.
The solution is to simply delete an internal link when one
of the corresponding sockets is removed (just like normal
links are removed as well).
We can't include `BLI_utildefines.h` in `RNA_types.h` since Cycles includes
that, but duplicates some of the util defines. So you'd have duplicated
definitions.
Fixes a bug introduced in rB5dedb39d447b. `mesh_original` is not set if the
mesh has no generative modifiers, in which case we can use `mesh_final`, which
would seem to be consistent with the rest of the particle code. An alternative
approach would be to make sure that `mesh_original` is always set in
`deformVerts`.
Differential Revision: https://developer.blender.org/D13754
Must take into account SD_OBJECT_TRANSFORM_APPLIED to determine if the normal
was already in world space.
Differential Revision: https://developer.blender.org/D13639
The root of the issue is caused by Cycles ignoring OpenGL limitation on
the maximum resolution of textures: Cycles was allocating texture of the
final render resolution. It was exceeding limitation on certain GPUs and
driver.
The idea is simple: use multiple textures for the display, each of which
will fit into OpenGL limitations.
There is some code which allows the display driver to know when to start
the new tile. Also added some code to allow force graphics interop to be
re-created. The latter one ended up not used in the final version of the
patch, but it might be helpful for other drivers implementation.
The tile size is limited to 8K now as it is the safest size for textures
on many GPUs and OpenGL drivers.
This is an updated fix with a workaround for freezing with the NVIDIA
driver on Linux.
Differential Revision: https://developer.blender.org/D13385
This patch fixes a couple of new Metal kernel compilation errors: 1) a kernel parameter count overflow, and 2) missing address space qualifiers.
Reviewed By: brecht
Differential Revision: https://developer.blender.org/D13763
Simplify signature of `BKE_lib_override_library_resync` and make it a
shallow wrapper around new internal `lib_override_library_resync` that
can then be easily extended for other internal needs.
Not functional changes expected here.
`DEG_OBJECT_ITER_FOR_RENDER_ENGINE_BEGIN` creates temporary objects that
correspond to duplicates or instances.
These temporary objects can share same pointers with the original object
as in the case of Bounding Box.
Bound Box of temporary objects is marked dirty in
`BKE_object_replace_data_on_shallow_copy` since `ob->data` is different.
This causes the original Bounding Box, calculated for the evaluated
geometry, to be lost.
The solution in this commit is to change the boundbox reference of the
temporary objects, so the boundbox of the non-temporary object (with
the data curve) is not marked dirty.
Differential Revision: https://developer.blender.org/D13581
It it rather an old experiment now which didn't pay off.
The initial idea was to have main and jobs threads on fast
nodes of TR2 processors. This didn't really work reliably
because in Blender we need to be able to create nested
threads without their affinity set. This is not how some of
OS are creating nested threads, and we don't always have
access to child threads to reset their affinity.
So overall complexity of the initial idea implementation
became too much compared to the performance gain.
Request from studio, to help identify quickly libs that need update.
NOTE: Currently only outputing INFO log in console, display of this info
in the outliner will come in a separate commit.
No need for it now since all the threading queries and
scheduling is done via TBB.
Should be no functional changes as all the removed code
is supposed to be unused.
Query TBB for the maximum allowed concurrency, which is free from a bug
in own concurrency detection code. One thing to keep in mind is that now
Cycles is limited by the number of threads in the TBB areana from which
Session is created. This isn't a problem for Blender since we do not limit
arena on Blender side. Could be something to watch out for in other Cycles
integrations.
Differential Revision: https://developer.blender.org/D13658
Add a data boundary check in the flipping code.
This code now also communicates the number of mipmap levels
it processed with an intent to avoid GPU texture from using
more levels than there are in the DDS data.
Differential Revision: https://developer.blender.org/D13755
This was missing from rB3e92b4ed2408eacd126c0.
Before only the Separate Geometry node was fixed, because that
node was used in the file from the bug report. The same issue
existed in the Delete Geometry node as well though.
Regression in 7972785d7b that caused
Python callback arguments to be de-referenced twice - potentially
accessing freed memory. Making a new-file with a circle-select
tool active triggered this (for example).
Now arguments aren't de-referenced when Blender it's self has already
removed the callback handle.
Fix IMB_flip[xy] to handle cases where integer overflow might occur when
given sufficiently large image dimensions.
All of these fixes were of a similar class where the intermediate
sub-expression would overflow silently. Widen the types as necessary.
Differential Revision: https://developer.blender.org/D13744
Assert that only the file name component is passed in
since special handling for UDIM should only be applied to the file name.
Also remove an unnecessary NULL check on the filename argument.
For sake of consistencey with other node tree types, create its own cmake module.
This change helps keep `bf_nodes` focused on generic nodes files.
Texture nodes are end of life and hopefully for Blender 4.0 they can be removed.
It is not expected that these will see the updates that the other nodes are getting.
This change also helps isolate the end of life files, we may move some texture
specific node tree execution code out of `node_exec` and into a `node_texture_exec` files.
Differential Revision: https://developer.blender.org/D13743
Sculpt Smooth in Surface mode (as opposed to Laplacian) needs a cache
initialized on first time. In anchored stroke mode with spherical falloff
this was skipped though (because this starts of with no PBVH nodes and
an early return checks for this) and `first_time` was set to false before
cache initialization.
Now move the cache initalization to happen earlier (same as the cache
initialization for automasking).
Maniphest Tasks: T94635
Differential Revision: https://developer.blender.org/D13746
The `lines_adjacency` IBO build in the GPU subdivision case was missing
edges at the boundaries of open meshes. As it is used for the shadow
pass, the shadows were then not clipped properly.
This would also make X-Ray mode render differently in those cases.
To fix this, we can simply reuse the buffer finalization routine from the
non-subdivision case, as such edges are handled there.
This relation is intended to ensure that the properties of the IK
constraint are ready by the time the IK solver tree is built. This
however can cause spurious dependency cycles, because there is only
one init tree node for the whole armature, and the relation actually
implies dependency on all properties of the bone.
This patch reduces spurious dependencies by only creating the relation
if any properties of the IK constraint specifically are animated.
Differential Revision: https://developer.blender.org/D13714
When weight painting the bone overlay is extremely intrusive,
effectively requiring either extensive use of hiding individual
bones, or disabling the whole bone overlay between selections.
This addresses the issue by adding a bone opacity slider that
is used for the 'wireframe' armature drawing mode. It directly
controls the uniform opacity as a straightforward option.
Differential Revision: https://developer.blender.org/D11804
If multiple bones have a custom property with the same name,
depsgraph didn't distinguish between them, potentially leading
to spurious cycles.
This patch moves ID_PROPERTY operation nodes for bone custom
properties from the parameters component to individual bone
components, thus decoupling them.
Differential Revision: https://developer.blender.org/D13729
Oversight in {rB9cb5f0a2282a}.
Above commit made an entry in `rna_Space_refine()`, but the entry in
`rna_Space_refine_reverse()` was missing (and this is what python uses
for the Space callbacks).
Maniphest Tasks: T94685
Differential Revision: https://developer.blender.org/D13751
The crash is caused as the data is only for the first frame, but the mesh
changes topology, so reading the data in subsequent frames causes a
buffer overflow. To fix this, we check that the data size matches the
mesh's vertex count.
Remove `const` from pass-by-value parameters in function declarations.
The variables passed as parameters can never be modified by the function
anyway, so declaring them as `const` is meaningless. Having the
declaration there could confuse, especially as it suggests it does have
a meaning, training people to write meaningless code.
Remove `const` from pass-by-value parameters in function declarations.
The variables passed as parameters can never be modified by the function
anyway, so declaring them as `const` is meaningless. Having the
declaration there could confuse, especially as it suggests it does have
a meaning, training people to write meaningless code.
Error in ffc4c126f5,
which moved doc-strings from implementation into headers.
Some changes in BKE_animsys.h needed to done manually as there
were already doc-strings in both the header and implementation
(with overlapping information).
When making these changes some doc-strings were removed unintentionally.
Thanks for @sybren for the heads up.
MSVC2017 and early 2019 versions are under
the impression struct OGLRender is non trivial
type due to the ThreadCondition field, not
entirely sure why, but it is what it is.
Differential Revision: https://developer.blender.org/D13742
Reviewed by: JacquesLucke
Just disable these tests on macOS for now as fixing seems hard, and we want to
be able to cross-compile and test x86_64 on Arm machines on the buildbot.
An important check to reject edge linehits when a vertex of that edge
was already hit was accidentally removed in
rB6e77afe6ec7b6a73f218f1fef264758abcbc778a
From what I can tell these are left over from Blender Internal render.
Test still pass locally, also tested a couple eevee scenes.
Reviewed By: JacquesLucke, brecht
Differential Revision: https://developer.blender.org/D13732
When tiled rendering was used the render result was
allocated at the end of every view layer render as
opposite of an intended end of all rendering.
Modify the render_result_end so that it only ensures
pixels are allocated if pixels are actually copied
over.
Overrides that are not created as part of an override hierarchy should
not be handled through (auto)resync at all. users are responsible to
hanlde those updates if they need it.
This is achieved by flagging overrides created outside of a hierarchical
process accordingly, and skipping them during resync process.
The current preview generation is more confusing than useful.
Therefore it is better to disable it until better preview generation
methods are found.
Differential Revision: https://developer.blender.org/D13728
The crash was caused by using `modify_geometry_sets` to modify
instances, which does not generally work unfortunately.
The intended behavior was wrong anyway. In instances mode,
only top level instances should be deleted.
Also removed the old error handling because it doesn't look like it
ever worked. all_is_error remained false all the time.
Furthermore, updating it was not thread safe.
Differential Revision: https://developer.blender.org/D13736
Take the Use Modifier Stack setting into account when connecting hair, and
fix wrong results results when using deforming modifiers also.
Differential Revision: https://developer.blender.org/D13704
Enables the `bpy.ops.cycles.denoise_animation()` operator again and modifies it to support
temporal denoising with OptiX. This requires renders that were done with both the "Vector"
and "Denoising Data" passes.
Differential Revision: https://developer.blender.org/D11442
This adds support to render PointCloud motion blur from a standard
"velocity" attribute.
This implementation is similar to that of the Mesh geometry, and
perhaps some code could be deduplicated through a more generic API.
`mesh_need_motion_attribute` was renamed `object_need_motion_attribute`
as it does not really require a mesh and moved to `util.h` so that
it can be shared.
This fixes T94622.
Reviewed By: brecht
Maniphest Tasks: T94622
Differential Revision: https://developer.blender.org/D13719
The DWAB compression was disabled in the d59721c2c3 due to
a bug in the OpenEXR library which is now resolved.
Re-enable the DWAB compression for OpenEXR output. It is a
simple change, and DWAB often behaves better than DWAA.
Differential Revision: https://developer.blender.org/D13713
If a mirror object is used in a mirror modifier, sculptmode did not take
this into account (and instead always clipped on the sculpt objects
local axis).
Now take this into account by storing a matrix in the preparation
function `sculpt_init_mirror_clipping` and use that later in
`SCULPT_clip`.
Maniphest Tasks: T94564
Differential Revision: https://developer.blender.org/D13711
This case wasn't handled in rBf5ce243a56a22d718 correctly.
Now `object_get_evaluated_geometry_set` just returns a geometry
set that contains the collection instance for collection instance objects.
Just an oversight in rBe9607f45d85d.
Now add notifier that toolsettings changed.
Maniphest Tasks: T94366
Differential Revision: https://developer.blender.org/D13723
Previously operations for the math node when connecting to
outputs weren't added. It also used a different method to
check whether the link would be valid.
Do not temporarily change U.pixelsize while creating object previews
in object_preview_render. It does nothing to the render, but the change
in line width can affect other UI drawing since it is done in a thread.
see D13717 for for details.
Differential Revision: https://developer.blender.org/D13717
Reviewed by Julian Eisel
Along with the general changes to CPP this commit does the following
- Use static casts where possible
- Use new CPP MEM library functions instead of cast
- Use listbase macros where possible
- Declare variables where initialized
Reviewed By: HooglyBoogly, JacquesLucke
Differential Revision: https://developer.blender.org/D13718
Dragging from a color socket would hit an assert in a debug build.
The node does not have a color mode currently, so use the vector mode
instead when connecting to a color socket.
std::min was used without including the algorithm
header. Seems to be implicitly included by
something in newer MSVC versions and GCC, however
vs16.4 needed a little help here.
Since 2.8, background images are tied to cameras (in 2.79 these were
tied to a View3D I think).
Code in `BKE_library_id_can_use_idtype` wasnt taking this relation
between `Camera` and `Image` into account, thus leading to ID deletion/
unlinking not working properly -- in particular `libblock_remap_data`
not doing its thing (and leaving the camera as a user of the image),
then things went downhill from there...
Now make the "Camera-can-use-an-Image" relation clear in
`BKE_library_id_can_use_idtype`.
Maniphest Tasks: T94544
Differential Revision: https://developer.blender.org/D13722
Currently the crop higher limits are inclusive too which contradicts
the documentation as it says that if Left and Right are both 50, it
will result in a zero-sized image. And the result is one pixel out of
the crop gizmo, which is another hint that this is not intended.
In "Full Frame" experimental mode it's two pixels short because of
a misuse of `BLI_rcti_isect_pt` as it considers max limits inclusive.
Reviewed By: jbakker
Maniphest Tasks: T90830
Differential Revision: https://developer.blender.org/D12786
The frame indicator/controller is not displayed when in the Graph or Dopesheet view of the Movie Clip Editor
To solve this we could call the function ED_time_scrub_draw_current_frame in clip_draw_dopesheet_main and graph_region_draw in space_clip.c
Reviewed By: jbakker
Maniphest Tasks: T91160
Differential Revision: https://developer.blender.org/D12659
Enable unity builds for `bf_nodes_shader`, gives about a 2.7x speedup
of total compile times when just building `bf_nodes_shader`.
On my machine, this equates to saving about 30 seconds.
Differential Revision: https://developer.blender.org/D13720
Function `blend_color_softlight_float` used math different to compositor and
produced result that had abrupt value changes.
Use math based on modified screen blend mode as compositor does.
This flag is only used a few small cases, so instead
of setting the flag for every node only set the
required flag for the nodes that require it.
Mostly the flag is used to set `ntype.flag = NODE_PREVIEW`
For nodes that should have previews by default which
is only some compositor nodes and some texture nodes.
The frame node also sets the `NODE_BACKGROUND` flag.
All other nodes were setting a flag of 0 which has no purpose.
Reviewed By: JacquesLucke
Differential Revision: https://developer.blender.org/D13699
This was originally written by Ankit Meel as a GSoC 2020 project.
Howard Trickey added some tests and made some corrections/modifications.
See D13046 for more details.
This commit inserts a new menu item into the export menu called
"Wavefront OBJ (.obj) - New".
For now the old Python exporter remains in the menu, along with
the Python importer, but we plan to remove it soon (leaving the
old addon bundled with Blender but not enabled by default).
Compare the start of the range to zero to figure out whether the
indices for the instances to keep starts at zero. Also rename the
selection argument, since it made it seem like the selected indices
should be removed rather than kept.
The activation of the text button is a bit special, since it happens during
drawing, the layout isn't computed yet then. Comparable cases where the button
is added on top don't use the layout system, so this didn't become an issue
until now. Trigger a delayed call to `UI_but_ensure_in_view()`.
This completes 1a721c5dbe by versioning old files to correct the
region type. The "tools" region type is relatively standard for this type
of region and doesn't require any changes to the theme, unlike
the "nav bar" type, which would have been a reasonable choice.
Calculates the angle in radians between two faces that meet at an edge.
0 to PI in either direction with flat being 0 and folded over on itself PI.
If there are not 2 faces on the edge, the angle will be 0.
For valid edges, the angle is the same as the 'edge angle' overlay.
For the Face and Point domain, the node uses simple interpolation to calculate a value.
Differential Revision: https://developer.blender.org/D13366
Occured because "PATH_RAY_SHADOW_CATCHER_BACKGROUND" is expressed as an unsigned
integer, because too large for a signed integer, but the "PathRayFlag" enum type defaulted to a
signed integer still.
Allow overriding simple properties of cloth simulations, colliders
and force fields. Vertex group and shape key selectors in cloth are
still not overridable since they are tied to mesh data.
Force fields have a number of physical fields shared between multiple
RNA fields. Until they are decoupled, they will produce redundant
overrides, and cannot have different hard range limits.
Differential Revision: https://developer.blender.org/D13710
Regression introduced in rB098008f42d8127d9b60717c7059d3c55a3bfada7
Previously the selected geometry was ignored along with the hidden one.
The mentioned commit caused neither the hidden nor the selected one to be ignored.
But hidden geometry needs to be ignored.
Continuation of the D13404 which finished the design of not having
geometry-level nodes dependent on object-level.
Differential Revision: https://developer.blender.org/D13405
Weirdly enough, our 'mono' font already had it, but not the main one.
Copied from DeJaVu sans font.
CC @Tamuna who started the translation for that language.
This implements the design detailed in T92696 to support virtual
filenames for UDIM textures. Currently, the following 2 substitution
tokens are supported:
| Token | Meaning |
| ----- | ---- |
| <UDIM> | 1001 + u-tile + v-tile * 10 |
| <UVTILE> | Equivalent to u<u-tile + 1>_v<v-tile + 1> |
Example for u-tile of 3 and v-tile of 1:
filename.<UDIM>_ver0023.png --> filename.1014_ver0023.png
filename.<UVTILE>_ver0023.png --> filename.u4_v2_ver0023.png
For image loading, the existing workflow is unchanged. A user can select
one or more image files, belonging to one or more UDIM tile sets, and
have Blender load them all as it does today. Now the <UVTILE> format is
"guessed" just as the <UDIM> format was guessed before.
If guessing fails, the user can simply go into the Image Editor and type
the proper substitution in the filename. Once typing is complete,
Blender will reload the files and correctly fill the tiles. This
workflow is new as attempting to fix the guessing in current versions
did not really work, and the user was often stuck with a confusing
situation.
For image saving, the existing workflow is changed slightly. Currently,
when saving, a user has to be sure to type the filename of the first
tile (e.g. filename.1001.png) to save the entire UDIM set. The number
could differ if they start at a different tile etc. This is confusing.
Now, the user should type a filename containing the appropriate
substitution token. By default Blender will fill in a default name using
the <UDIM> token but the user is free to save out images using <UVTILE>
if they wish.
Differential Revision: https://developer.blender.org/D13057
It is common to have fields that contain a constant value. Before this
commit, such constants were represented by operation nodes which
don't have inputs. Having a special node type for constants makes
working with them a bit cheaper.
It also allows skipping some unnecessary processing when evaluating
fields, because constant fields can be detected more easily.
This commit also generalizes the concept of field node types a bit.
Currently, a node either supports lazyness during execution (like the Switch
node), or it doesn't. If it does support lazyness, then every input is computed
lazily. However, usually not all inputs actually have to be computed lazily.
E.g. the boolean switch input is always required, while the other inputs
should be computed lazily.
Better support for such sockets can avoid unnecessary round trips through
the node execution function.
Exposes compare operations via rna emums.
This uses the rna enum to build the search list using
named operations linked to socket A.
This also weights the Math Node comparison operations lower
for geometry node trees.
Differential Revision: https://developer.blender.org/D13695
The logic used to be:
"if collection doesn't have child collection, check if ob is from this one"
The correct logic should be:
"if collection child does not have this ob, then check this collection".
In the past that worked because the `GPUMaterial` referenced the
`ImageUser` from the image node. However, that design was incompatible
with the recent node tree update refactor (rB7e712b2d6a0d257d272e).
Also, in general it is a bad idea to have references between data that is
owned by two different data blocks.
This incompatibility was resolved by copying the image user from the node
to the `GPUMaterial` (rB28df0107d4a8). Unfortunately, eevee depended
on this reference, because the image user on the node was update when the
frame changed. Because the image user was copied, the image user in the
`GPUMaterial` did not receive the frame update anymore.
This frame update is added back by this commit. The main change is that
the image user iterator now also iterates over image users in `GPUMaterial`s
on material and world data blocks. An issue is that these materials don't
exist on the original data blocks and that caused the check in
`build_animation_images` in the depsgraph to give the wrong answer.
Therefore the check is extended.
Right now the check is not optimal, because it results in more depsgraph
nodes than are necessary. This can be improved when it becomes cheaper
to check if a node tree contains any references to a video texture.
The node tree update refactor mentioned before makes it much easier
to construct this kind of run-time data from the bottom up, instead of
scanning the entire node tree recursively every time some information
is needed.
This just skips the entire algorithm when there are cycles.
In the future, cycles could be handled more gracefully in the
algorithm, but for now that's not worth it and is not necessary
to fix the bug.
As @hooglyboogly suggested in D13680, this patch adds weighting
to the search results. Dragging from a vector/rgba socket weights
the Vector Math node higher than a float Math node, and vice versa.
Reviewed By: HooglyBoogly
Differential Revision: https://developer.blender.org/D13691
When the material is used in several objects, the filter by material is not working as expected because the internal pointers are different due eval version.
Now, the original version of the material is compared to keep same address.
This is analagous to 6a71b2af66 which did the same
thing for mesh data. Two differences are that here the coordinates
are simply `float3`, and we account for the radius if it's available.
Here I observed a similar performance increase, from 50ms
average to 10ms average, with 16 million points, a 5x speedup.
The calculation is about 1.4 times faster when no radius is used, down
to 7.3ms average. Before, the calculation was only 1.2 times faster.
This was caused by a mistake in eb0eb54d96, which removed
the clearing of the curve edit mode pointers that are set when creating
the temporary data for the conversion. If they are not cleared, the
generic ID free function will also free the edit mode data, which is
wrong when the source curve is in edit mode.
At the time of allocating the buffer with vertices in context, we don't
know exactly how many vertices are affected, but we do know that it is
less than or equal to twice the number of vertices killed.
This just moves the relations update to a lower level function that is used
by other functions. Eventually, the special case for this node should be
generalized.
Ever since the instance domain was added, this was exposed, it just
didn't do anything. This patch implements the instances domain in the
delete and separate geometry nodes, where it acts on the top-level
instances.
We act on a mutable instances input, with the idea that eventually
copy on write attribute layers will make this less expensive. It also
allows us to keep the instance references in place and to do less
work in some situations.
Ref T93554
Differential Revision: https://developer.blender.org/D13565
This function node creates a running total of a given Vector, Float, or
Int field.
Inputs:
- Value: The field to be accumulated
- Group Index: The values of this input are used to aggregate the input
into separate 'bins', creating multiple accumulations.
Outputs:
- Leading and Trailing: Returns the running totals starting
at either the first value of each accumulations or 0 respectively.
- Total: Returns the total accumulation at all positions of the field.
There's currently plenty of duplicate work happening when multiple outputs
are used that could be optimized by a future refactor to field inputs.
Differential Revision: https://developer.blender.org/D12743
A few typos in 17770192fb lead to an incorrect count of custom
data layers in the test meshes. We only want to consider layers that are
not anonymous, and there was a copy and paste mistake.
`GeometrySet::compute_boundbox_without_instances` may not initialize min
max in some cases such as meshes without vertices.
This can result in a Bounding Box with impossible dimensions
(min=FLT_MAX, max=-FLT_MAX).
So repeat the same solution seen in `BKE_object_boundbox_calc_from_mesh`
and set boundbox values to zero.
Reviewed By: HooglyBoogly
Differential Revision: https://developer.blender.org/D13664
All these function paramaters are set to NULL so they arent necessary.
Reviewed By: HooglyBoogly, JacquesLucke
Differential Revision: https://developer.blender.org/D13686
Some projects need more than 8-bit RGBA off-screen, so add the ability to
accept color format and defaults to RGBA8 so existing code should not be
affected.
Currently supported formats:
- RGBA8 (default)
- RGBA16
- RGBA16F
- RGBA32F
Reviewed By: mano-wii
Differential Revision: https://developer.blender.org/D13650
Blend factor was used to adjust alpha of background image, which is not
correct. This was done in fdee84fd56 where another change was, that
background alpha is copied into result, which is correct.
Apply blend factor to foreground image alpha channel.
The array modifier does not necessarily tag normals dirty.
If it doesnt, normals are recalculated "internally" using the offset ob
transform. This was happening for the array items, but not for the caps.
Now do the same thing for caps.
Maniphest Tasks: T94422
Differential Revision: https://developer.blender.org/D13681
I had assumed that the span's size was the same as the length variable.
In the future, separate lengths could be removed in favor of using
lengths directly from spans.
Fixes two instances of `-Wunused-but-set-variable`
There are several more of these but these were low hanging
and noisy with one being in a header functions.
Unity build saves 5 seconds off the total build time when compiling `bf_nodes_function`.
Total build times went from 25s to 20s (20% reduction),
tested with ninja on linux running i5 8250U.
Using the output declarations is incorrect because there is a
declaration for each type. Instead loop over the names directly,
since it will make it easier to add an integer mode that only
supports some of the outputs.
The calculation to find the factor between two evaluated points assumed
that the points were not at the same location. This assumption is some-
what reasonable, since we might expect `lower_bound` to skip those
point anyway. However, the report found a case where the first two
evaluated points were coincident, and there is no strong reason not
to make this safe, so add a check for 0 length before the division.
The region used to be type "Channels", but the standard for this
type is "Tools", which is what the file browser uses. This follows
the changes in rB01df48a98394, which also make the region more
"standard."
Some effect functions looped over alternating lines, previously with
different factors. Since only one factor is used, code can be
simplified by looping all lines in one for loop.
There should be no functional changes.
* Assert about source ID of an overridden pointer property not being a
liboverride was not necessary, just skip in that case.
* Tag actual 'real' ID owner for resync, and not (potentially) an embedded one.
2 factor variables were passed to effects, but they were hard coded to
have same value.
Remove duplicate variable from arguments, rename single argument to
`fac`. Inverted factor variables were renamed to `mfac`. Any other
factor related variables are prefixed with `temp_`.
There should be no functional changes.
If the override system creates an override record for the cache
name (no idea why though), it trashes the disk cache on file load.
The reason is that it tries to rename cache files in update handler
when assigning the name, and BLI_rename deletes the target file even
if both names are the same.
This is a safe fix that simply aborts the pointless rename attempt.
Differential Revision: https://developer.blender.org/D13679
This evaluator is used in order to evaluate subdivision at render time, allowing for
faster renders of meshes with a subdivision surface modifier placed at the last
position in the modifier list.
When evaluating the subsurf modifier, we detect whether we can delegate evaluation
to the draw code. If so, the subdivision is first evaluated on the GPU using our own
custom evaluator (only the coarse data needs to be initially sent to the GPU), then,
buffers for the final `MeshBufferCache` are filled on the GPU using a set of
compute shaders. However, some buffers are still filled on the CPU side, if doing so
on the GPU is impractical (e.g. the line adjacency buffer used for x-ray, whose
logic is hardly GPU compatible).
This is done at the mesh buffer extraction level so that the result can be readily used
in the various OpenGL engines, without having to write custom geometry or tesselation
shaders.
We use our own subdivision evaluation shaders, instead of OpenSubDiv's vanilla one, in
order to control the data layout, and interpolation. For example, we store vertex colors
as compressed 16-bit integers, while OpenSubDiv's default evaluator only work for float
types.
In order to still access the modified geometry on the CPU side, for use in modifiers
or transform operators, a dedicated wrapper type is added `MESH_WRAPPER_TYPE_SUBD`.
Subdivision will be lazily evaluated via `BKE_object_get_evaluated_mesh` which will
create such a wrapper if possible. If the final subdivision surface is not needed on
the CPU side, `BKE_object_get_evaluated_mesh_no_subsurf` should be used.
Enabling or disabling GPU subdivision can be done through the user preferences (under
Viewport -> Subdivision).
See patch description for benchmarks.
Reviewed By: campbellbarton, jbakker, fclem, brecht, #eevee_viewport
Differential Revision: https://developer.blender.org/D12406
Before one could only create a new group input using the link drag search.
With this patch it becomes possible to create a Group Input node for an
existing input.
Differential Revision: https://developer.blender.org/D13674
This adds `blender::is_same_any_v` which is the almost the same as
`std::is_same_v`. The difference is that it allows for checking multiple
types at the same time.
Differential Revision: https://developer.blender.org/D13673
The edge is linked twice from differen calls during line art calculation
Probably caused by a merge and both calls stayed for some reason.
This would lead to edge link overflowing its limit of 2^16 items.
The crash happens when opening a panel (added in rB43f5e761a66e87fed664a199cda867639f8daf3e)
when no CacheFile is set in the modifier.
To fix this, check that the CacheFile pointer is not null before attempting to draw anything.
Most of the comment block is similar to the text in the source
code documentation wiki. It's helpful to have some text in
a header file too, so it's closer for programmers already looking
at the code.
This also uses more consistent syntax and wording in the comments
about the attribute API in `GeometryComponent`.
Ref T93753
Differential Revision: https://developer.blender.org/D13661
This patch adds the breakdown (or tween) functionality to the graph editor.
The factor defines the linear interpolation from left key to right key.
Reviewed by: Sybren A. Stüvel
Differential Revision: https://developer.blender.org/D9375
Ref: D9375
This node is a bit special in that it uses two internal sockets
for a hack for Eevee; see rBffd5e1e6acd296a187e7af016f9d7f8a9f209f87
As a result, the `SOCK_UNAVAIL` flag is exposed to socket builder API.
Reviewed By: JacquesLucke, fclem
Differential Revision: https://developer.blender.org/D13496
This patch adds the blend to neighbor operator to the Graph editor.
The operator acts like the blend to neighbor operator for a pose context, just working on keyframes.
Reviewed by: Sybren A. Stüvel
Differential Revision: https://developer.blender.org/D9374
Ref: D9374
The issue was caused by rB7e712b2d6a0d257d272ed35622b41d06274af8df
and the fact that `GPUMaterialTexture` contains an `ImageUser *` which
references the `ImageUser` on e.g. `NodeTexImage`.
Since the node tree update refactor, it is possible that the node tree changes
without changing the actual material. Therefore, either the renderer should
check if the node tree has changed or it should not store pointers to data in
node storage. The latter approach is implemented in this patch.
Differential Revision: https://developer.blender.org/D13663
Loops come last in the struct's definition, use the same order when
initializing the common masks in customdata.c (they were switched
with the poly masks).
Typically a node group should only have a single Group Output node.
However, currently Blender already supports having multiple group outputs,
one of which is active. This wasn't handled correctly by geometry nodes.
Differential Revision: https://developer.blender.org/D13611
rB05f900e3466b45a19e13bea6dd641e4f7b8b46e9 removed unused button functions,
but since that commit the `uiDefIconTextButBit()` static function sits
unused as well. It's now been removed.
This looks like a copy and paste error from the original commit.
The virtual array output used the number of mesh polygons instead
of the number of edges.
These were part of the older buttons API that shouldn't be used in
more places at this point. Most layouts should be built with the regular
layout system API and RNA properties. This sort of button can still be
created though, since these were just shortcuts anyway.
This adds interface panels to organize the Cache File UI parameters for
modifiers and constraints into related components: velocity, time, and
render procedural.
Properties relating to the three aforementioned components are separated
from `uiTemplateCacheFile` into their own functions (e.g.
`uiTemplateCacheFileVelocity` for the velocity one), which are in turn
called from the specific panel creation routines of the modifiers and
constraints (for constraints, the functions are exposed to the RNA).
`uiTemplateCacheFile` now only shows the properties for the file path,
and in the case of constraints, the scale property.
The properties that are only defined per modifier (like the velocity
scale), are shown in the proper modifier layout panel if applicable.
Reviewed By: sybren
Differential Revision: https://developer.blender.org/D13652
When using graph slider operators like D9374
it showed a warning when no keys were selected.
However since that stops the modal operation it should be an Error.
Also the message was misleading
since it could error for different reasons than stated.
Reviewed by: Sybren A. Stüvel
Differential Revision: https://developer.blender.org/D13655
Ref: D13655
The `BKE_animdata_driver_path_hack()` function has had almost no effect
since rB51b796ff1528, and basically boils down to:
```
return base_path ? base_path : RNA_path_from_ID_to_property(ptr, prop);
```
Since `base_path` was `NULL` in the majority of cases, it's just been
replaced by a direct call to `RNA_path_from_ID_to_property()`. The
conditional now just appears in one remaining case.
This relates to T91387.
Reviewed By: mont29
Differential Revision: https://developer.blender.org/D13646
Do not allow 3DView operator to run on the liboverride of an
instantiating Empty object. And tweak behavior in the Outliner
operations too.
Related to T94226.
Note that this remains fairly exotic, bad idea not recommended cases,
such complex inter-dependencies between different libraries inside a
same liboverride hierarchy is just not possible to handle properly.
rBbd3bd776c893 broke compilation here due to missing type declaration
for basic types as the source file is not including this header. In any
case, it is the responsibility of header files to include headers for
types used by value in function parameters or struct definitions.
If the input mesh had no materials already, the new material would
become the only material on the mesh, meaning the material was
added to all of the faces, instead of just the selected faces.
The mesh primitive nodes in geometry nodes already add an empty
slot by default, so this only affects outside geometry.
The fix is just adding an empty slot before the new slot, so the
non-selected material indices can still point to an empty slot.
Differential Revision: https://developer.blender.org/D13654
It's better to calculate the size of a spline before creating it, and this
should simplify refactoring to a data structure that stores all point
attribute contiguously (see T94193). The mesh to curve conversion is
simplified slightly now, it creates the curve output after gathering all
of the result vertex indices. This should be more efficient too, since
it only grows an index vector for each spline, not a whole spline.
Previously only the float version of the node was connected to.
This adds connection operations for vector sockets, and exposes
the "Steps" socket properly when it's selected.
Caused by capturing local variables by reference in a function that
outlives the scope it was created in. Also use a more generic function
for the first two inputs.
I don't think this has been visible, since I only ran into it after
changing other code that affected this. However, some attributes
can keep a reference to the source component to use when tagging
caches dirty (like the position attribute tagging the normals dirty).
Here, the component was created inside a function, then the attributes
were used afterwards.
Also add some comments warning about this in the header file.
When COM_EXPORT_GRAPHVIZ is enabled, DebugInfo::graphviz
uses a char[1000000] as local variable. When this function
is called this is allocated on the stack, which has a size
of just 1MB on mac and may cause a stack overflow.
This patch allocates the memory on the heap and frees
the memory at the end of the function.
Reviewed By: LazyDodo
Differential Revision: https://developer.blender.org/D13628
This should be faster because it avoids reallocating the internal
vectors when the size is known beforehand, but it may also help
a potential refactor to a different data structure (see T94193).
Toggling node mute doesn't cause node trees to reevaluate after
rB7e712b2d6a0d257. Toggling a link mute still works though. To fix this,
the operator tags the node and node with a new update tag function
(that uses an existing tag internally).
Differential Revision: https://developer.blender.org/D13653
`SOCK_IN_USE` is now set in `update_socket_used_tags` in
`node_tree_update.cc` when a node tree is changed.
It doesn't need to run every single redraw. Removing this
results in a small speedup of 0.4 ms when drawing a tree
with about 4000 nodes (from about 70 ms total).
Differential Revision: https://developer.blender.org/D13645
This replaces the single-threaded calculation of mesh min and max
positions with a `parallel_reduce` loop. Since the bounding box
of a mesh is retrieved quite often (at the end of each evaluation,
currently 2(?!) times when leaving edit mode, etc.), this makes for a
quite noticeable speedup actually.
On my Ryzen 3700x and a 4.2 million vertex mesh, I observed
a 4.4x performance increase, from 14 ms to 4.4 ms.
I added some methods to `float3` so they would be inlined, but
they're also a nice addition, since they're used often anyway.
Differential Revision: https://developer.blender.org/D13572
Fix is similar to how CollectionObject with NULL object pointers are handled.
Using one of the 'free' pad bytes in Object_Runtime struct instead of a
gset (or other external way to detect object duplicates), as this is
several times faster.
NOTE: This makes remapping slightly slower again (adds 10 extra seconds
to file case in T94059).
General improvements of remapping time complexity, especially when
remapping a lot of IDs at once, is a separate topic currently
investigated in D13615.
This commit fixes the crash itself, however this can still lead to a
same collection 'owning' the same object several time.
Issue here was a bad assumption in layer resync code, that would lead to
removing valid objects from the viewlayer's `object_bases_hash` in
`BKE_layer_collection_sync`, when deleting no-more-used bases, in case
of bases duplicate.
The cast to size_t leads to a build issue on 32
bit archs. cursor_delim_type_utf8 expects an int
so an additional cast to size_t is not required.
Reported by user frispete on devtalk.
The problem was the number of points for each edge of the box was wrong and the wheelmouse effect was anulated.
Also fixed the value displayed in the status bar to keep consistency with subdivision value.
Reviewed By: lichtwerk
Maniphest Tasks: T93999
Differential Revision: https://developer.blender.org/D1363
When copying a full node tree, we can avoid an O(n^2) loop finding a
unique name for every node if we assume they already have unique names.
That is a reasonable assumption, since unique names are verified
elsewhere when adding a new node.
Copying a node tree with about 4000 nodes took 42 ms before,
now it takes 6 ms.
Differential Revision: https://developer.blender.org/D13644
These pointers point to the new nodes when duplicating,
and their even used to point to "original" nodes for
"localized" trees. They're just a bad design decision
that make code confusing and buggy.
Instead, node copy functions now optionally add to a map
of old to new socket pointers. The case where the compositor
abused these pointers as "original" pointers are handled
by looking up the string node names.
Differential Revision: https://developer.blender.org/D13518
Previously, the values passed to a multi-input socket were stored
in the order that they arrived in. Then, when the values are accessed,
they are sorted depending on the link order.
Now, the ordering is determined in the beginning before execution starts.
Every value is assigned to the right index directly, avoiding the sort
in the end. This makes the ordering more explicit.
This errored out in two scenarios:
- current frame not in strips framerange (this was reported)
- no strips selected at all
Now handle these cases properly in the operator and give appropriate
report info.
Maniphest Tasks: T94295
Differential Revision: https://developer.blender.org/D13642
Caused by {rBb0d9e6797fb8}
For the header (both Graph Editor case in general `headerTranslation` as
well as `headerTimeTranslate`) we are interested in deltas values
(not absolute values).
Since culprit commit, `snapFrameTransform` was not working with deltas
anymore, but we have to compensate for this.
For the Graph Editor, this only worked "by accident" in rB7192e57d63a5,
since `ival` is still zero at this point.
So now, reacquire the delta right after the snap operation.
Also use a more appropriate center value in the translate operator.
Maniphest Tasks: T94191
Differential Revision: https://developer.blender.org/D13641
This commit makes freeing a node tree about 25 to 30 times faster.
Freeing a node tree happens whenever it is edited. Freeing a node
tree with about 4000 nodes went from 30-50ms to about 2 ms.
This was so slow before because for every node that was freed
when freeing the node tree, `node_free_node` looped over all
other nodes to detach frames, and then looped over all links to
remove any links connected to the node. That was all pointless
work because everything else is about to be freed anyway.
Instead, move that "detaching" behavior to the dedicated function
for removing a single node, and to the "local" version of the free
function to be safe, since I know less about what that version expects.
Differential Revision: https://developer.blender.org/D13636
On Windows, encode file paths as UTF-16 before trying to open the file
for reading/writing.
This introduces a new class `blender::fstream`, which wraps
`std::fstream` and provides this UTF-16 encoding. This class should also
be used in other areas, like the Alembic importer/exporter.
Manifest Task: T93960
Reviewed By: JacquesLucke
Differential Revision: https://developer.blender.org/D13633
Goals of this refactor:
* More unified approach to updating everything that needs to be updated
after a change in a node tree.
* The updates should happen in the correct order and quadratic or worse
algorithms should be avoided.
* Improve detection of changes to the output to avoid tagging the depsgraph
when it's not necessary.
* Move towards a more declarative style of defining nodes by having a
more centralized update procedure.
The refactor consists of two main parts:
* Node tree tagging and update refactor.
* Generally, when changes are done to a node tree, it is tagged dirty
until a global update function is called that updates everything in
the correct order.
* The tagging is more fine-grained compared to before, to allow for more
precise depsgraph update tagging.
* Depsgraph changes.
* The shading specific depsgraph node for node trees as been removed.
* Instead, there is a new `NTREE_OUTPUT` depsgrap node, which is only
tagged when the output of the node tree changed (e.g. the Group Output
or Material Output node).
* The copy-on-write relation from node trees to the data block they are
embedded in is now non-flushing. This avoids e.g. triggering a material
update after the shader node tree changed in unrelated ways. Instead
the material has a flushing relation to the new `NTREE_OUTPUT` node now.
* The depsgraph no longer reports data block changes through to cycles
through `Depsgraph.updates` when only the node tree changed in ways
that do not affect the output.
Avoiding unnecessary updates seems to work well for geometry nodes and cycles.
The situation is a bit worse when there are drivers on the node tree, but that
could potentially be improved separately in the future.
Avoiding updates in eevee and the compositor is more tricky, but also less urgent.
* Eevee updates are triggered by calling `DRW_notify_view_update` in
`ED_render_view3d_update` indirectly from `DEG_editors_update`.
* Compositor updates are triggered by `ED_node_composite_job` in `node_area_refresh`.
This is triggered by calling `ED_area_tag_refresh` in `node_area_listener`.
Removing updates always has the risk of breaking some dependency that no
one was aware of. It's not unlikely that this will happen here as well. Adding
back missing updates should be quite a bit easier than getting rid of
unnecessary updates though.
Differential Revision: https://developer.blender.org/D13246
Outliner would frame the armature object instead of the bone if the bone
was on a hidden armature layer.
Similar to issues reported in e.g. T58068 and T80464, this is due to the
fact that `BKE_pose_channel_active` always checks for the armature layer
(and returns NULL if a bone is not on a visible armature layer).
Now propose to make this layer check **optional** (and e.g. from the
Outliner be more permissive). This also introduces
`BKE_pose_channel_active_if_layer_visible` which just wraps
`BKE_pose_channel_active` with the check being ON.
Maniphest Tasks: T92930
Differential Revision: https://developer.blender.org/D13154
Initialise the BKE callback system in
`BlendfileLoadingBaseTest::SetUpTestCase()`. This allows certain tests
to run in debug mode (when `BLI_assert` is enabled).
This happens because in `SEQ_time_update_sequence` function
`SEQ_get_meta_by_seqbase` returns uninitialized value. This isn't nice,
but it shouldn't happen in first place. Problem is, that
`SEQ_edit_strip_split` does move strips into detached `ListBase`, so
other functions can't see them anymore. Detached `ListBase` is used
solely to preserve relationships during duplication.
Move strips to original `ListBase` immediately after duplication and
return `NULL` if `SEQ_get_meta_by_seqbase` can't find meta strip.
Splitting itself can still rely on fact, that number of original and
duplicated strips is same and they are placed next to each other in
exactly same order at the end of original `ListBase`.
In this commit I changed many loops to range-based for loops.
I also removed some of the redundant iterator variables, using
indexing inside the loop instead. Generally an optimizing compiler
should have no problem doing the smartest thing in that situation,
and this way makes it much easier to tell where data is coming from.
I only changed the loops I was confident about, so there is still more
that could be done in the future.
Differential Revision: https://developer.blender.org/D13637
This was visible outside of camera view and was not respecting the
"Depth of Field" checkbox on the Camera properties.
Now return early if DoF should not be visible.
Maniphest Tasks: T94262
Differential Revision: https://developer.blender.org/D13631
Sculpt.c is now three files:
* Sculpt.c: main API methods and the brush stroke operator
* Sculpt_brushes.c: Code for individual brushes.
* Sculpt_ops.c: Sculpt operators other than the brush stroke operator.
TODO: split brush stroke operator into a new file (sculpt_stroke.c?).
The duplicated edges were caused by 'oversubdivided' edges, i.e. edges
where some of the vertices on them are only connected to two polygons.
The fix finds these vertices and 'dissolves' them so that only one edge
is created.
For most 'normal' meshes this shouldn't occurr, or only very little, so
the performance impact of this change should be neglegible. In practice
this is also avoidable by triangulating the mesh first.
Differential Revision: https://developer.blender.org/D13445
This fix provides better conditions for asserts in `NURBSpline::knots`
method accounting for cyclic NURBS curves.
Differential Revision: https://developer.blender.org/D13620
On a Ryzen 3700x, this ended up 2.5x faster than before. More
benchmarking details are included in the differential revision.
For smaller grids, all this should do is increase the
code size a bit, and add a few more if statements.
Differential Revision: https://developer.blender.org/D13617
There are two issues in USD code that break building it with gcc-11,
one (in `demangle.cpp`) was already fixed upstream, the other (in
`singularTask.h`) is still pending (reported upstream, see
https://github.com/PixarAnimationStudios/USD/issues/1721).
CC #platforms_builds_tests_devices project.
Root path variables for those libraries is now using the 'standard' naming
scheme.
With tweaks/cleanups from @mont29.
Reviewed By: mont29
Differential Revision: https://developer.blender.org/D13591
This was basically not implemented, do this via
`ED_autokeyframe_property` in a new dedicated function in
special_aftertrans_update.
Maniphest Tasks: T93570
Differential Revision: https://developer.blender.org/D13608
Differentiate doc-strings from title/section text.
Also use explicit doxygen references to struct members
so it's not ambiguous which member is being referenced.
Note that these changes aren't complete (some files weren't touched).
This moves `MOD_weld.cc` to C++, fixing compiler warnings
coming from the change. It also goes a little bit further and converts
the code to use C++ data structures: `Span`, `Array`, and `Vector`.
This makes the code more shorter and easier to reason about, and
makes memory maneagement more automatic.
Differential Revision: https://developer.blender.org/D13618
When dragging from the inputs of function nodes, other function
nodes wouldn't connect, because their socket declaration field types
weren't set correctly. Instead, they relied on code properly checking
the *node* declaration's `is_function_node()` method. However,
that increases complexity and requires passing the node instead of
just the socket in more places. Instead, set the proper field types
in the socket declaration during building.
Differential Revision: https://developer.blender.org/D13600
Before d56bbfea7b, nodes were updated (size calculated and
buttons added) in reverse order. Instead, now calculate the size of
frame nodes after all other nodes. Separating the drawing further
may be a good step to removing the O(n^2) loop later on.
This adds memory barriers to use with `GPU_memory_barrier` to ensure that
writes to a vertex or index buffer issued before the barrier are
completed after it, so they can be safely read later by another shader.
`GPU_BARRIER_VERTEX_ATTRIB_ARRAY` should be used for vertex buffers (`GPUVertBuf`),
and `GPU_BARRIER_ELEMENT_ARRAY` should be used for index buffers (`GPUIndexBuf`).
Reviewed By: fclem
Differential Revision: https://developer.blender.org/D13595
Add a function that returns a list of keyframe segments
A segment being a continuous selection of keyframes
Will be used by future operators in the graph editor
Reviewed by: Sybren A. Stüvel
Differential Revision: https://developer.blender.org/D13531
Ref: D13531
In the context of the dragdrop tooltip, the event referenced to the window
is out of date and contains invalid `mval` values.
Avoid using `win->eventstate` as much as possible.
The RNA setter now ensures that driver variables are uniquely named
(within the scope of the driver).
Versioning code has been added to ensure this uniqueness. The last
variable with the non-unique name retains the original name; this
ensures that the driver will still evaluate to the same value as before
this fix.
This also introduces a new blenlib function `BLI_listbase_from_link()`,
which can be used to find the entire list from any item within the list.
Manifest Task: T94116
Reviewed By: mont29, JacquesLucke
Maniphest Tasks: T94116
Differential Revision: https://developer.blender.org/D13594
The node group alpha theme was used for the overlay drawing in the node
editor. Since this was removed (919e513fa8) the alpha channel doesn't
need to be exposed anymore.
Reported as part of T93654.
The style guide mentions that unsigned integers shouldn't be used to
show that a value won't be negative. Many places don't follow this
properly yet. The modifier used to cast an array of `uint` to `int` in
order to pass it to `BLI_kdtree_3d_calc_duplicates_fast`. That is no
longer necessary.
Differential Revision: https://developer.blender.org/D13613
Using the `MEM_*` API from C++ code was a bit annoying:
* When converting C to C++ code, one often has to add a type cast on
returned `void *`. That leads to having the same type name three times
in the same line. This patch reduces the amount to two and removes the
`sizeof(...)` from the line.
* The existing alternative of using `OBJECT_GUARDED_NEW` looks a out
of place compared to other allocation methods. Sometimes
`MEM_CXX_CLASS_ALLOC_FUNCS` can be used when structs are defined
in C++ code. It doesn't look great but it's definitely better. The downside
is that it makes the name of the allocation less useful. That's because
the same name is used for all allocations of a type, independend of
where it is allocated.
This patch introduces three new functions: `MEM_new`, `MEM_cnew` and
`MEM_delete`. These cover the majority of use cases (array allocation is
not covered).
The `OBJECT_GUARDED_*` macros are removed because they are not
needed anymore.
Differential Revision: https://developer.blender.org/D13502
The patch fixes some misalignments in the nodes' sockets/options
recently introduced in 26d2caee3b, while maintaining the original
fix for T92268.
The original fix made the top padding always of the same size; while
that works when the first row of the other node is `Socket | Socket`,
it doesn't for other more common cases, `like Socket | Node Option`,
where the text results misaligned.
Differential Revision: https://developer.blender.org/D13451
The eraser checks the current, previous and next point (and sets pc0,
pc1 & pc2 corresponding to that for futher occlusion/brush/clipping
checks). For the very first point, it sets pc0 to pc1 [which makes sense,
there is no previous point, so we should assume the previous segment is
"visible" as soon as the first point is], but does so *before* pc1 is
even calculated. This makes following occlusion/brush/clipping checks
work with zero values [which leads to no earsing in most cases].
Now *first* calculate pc1, *then* set pc0 to pc1.
Maniphest Tasks: T94137
Differential Revision: https://developer.blender.org/D13593
The correct type should be set by invoke already, changing it to a non-
matching type (e.g. trying to remap Mesh users with a Camera block) does
not really make sense afaict, reason being that we would be presented
with the "Invalid old/new ID pair" message in such case anyways (code
checks GS(old_id->name) == GS(new_id->name)).
This alone wouldnt be a pressing issue, but since doing this with an
object ID type crashes atm., it seems to make sense to clean this up now
(of course the crash should be looked into, but this is for a separate
patch -- if that is solved, we could also think about adding the "Remap
Users" entry back in the context menu for objects as well [which was
removed in rB17bd5c9d4b1e for some reason]).
Part of T93799.
Differential Revision: https://developer.blender.org/D13512
The versioning code was accidentally put not at the very bottom.
That lead to a situation where it wasn't run on some files that happened
to be within a specific short time frame.
Since the versioning code is idempotent, it can just run again on existing
files. Therefore, this commit just moves it back to the bottom so that it
is executed on all files again.
Broken Commit: rB5b61737a8f41688699fd1d711a25b7cea86d1530.
Change the spelling of the QuickTime output video container item from
"Quicktime" to "QuickTime"
Differential Revision: https://developer.blender.org/D10929
Reviewed by Harley Acheson
Now that point clouds can be rendered with cycles, it makes sense
to allow assigning a material to them. Note that like volumes, they
only support a single material though.
This add support for rendering of the point cloud object in Blender, as a native
geometry type in Cycles that is more memory and time efficient than instancing
sphere meshes. This can be useful for rendering sand, water splashes, particles,
motion graphics, etc.
Points are currently always rendered as spheres, with backface culling. More
shapes are likely to be added later, but this is the most important one and can
be customized with shaders.
For CPU rendering the Embree primitive is used, for GPU there is our own
intersection code. Motion blur is suppored. Volumes inside points are not
currently supported.
Implemented with help from:
* Kévin Dietrich: Alembic procedural integration
* Patrick Mourse: OptiX integration
* Josh Whelchel: update for cycles-x changes
Ref T92573
Differential Revision: https://developer.blender.org/D9887
I need this for a refactor I'm looking into for bounding boxes.
It may be helpful in the future when using `CurveEval` in more places.
Differential Revision: https://developer.blender.org/D13596
Fully get rid of `BKE_collection_object_find` in
`lib_override_group_tag_data_object_to_collection_init`, even if only
used a few times this function was still noticeable in profiling data.
Now instead loop over collections' objects to build required
object-to-collections mapping.
Adds an extra 5-10% speed-up compared to previous commit rB0624fad0f3ff.
Related to T94059.
Operator was erroneously starting edge_slide operation.
Revert part of the changes in rB3fab16fe8eb4 as obedit_type was being
confused with object_mode.
`BKE_collection_object_find` has extremely bad performances (very high
time complexity). While ideally this should be fixed in that API, for
now cache its results once at the beginning of the resync/creation
process.
This makes loading of complex production files with a lot of
liboverrides to resync three to four times faster.
Thanks to @brecht for the profiling in T94059.
Not finding subitem when its name and index are invalid/unset is
expected behavior, and does happen when e.g. inserting a new constraint
or modifier at the begining of the stack.
When selecting the current undo step there is no need to do anything.
Fix and minor refactor to de-duplicate refreshing after running
undo/redo & undo history.
Both {key Alt} editing behavior as well as `Copy To Selected` were not
working on geometry nodes modifiers (even if these matched exactly -
having the same nodegroup - on multiple objects)
Reason is that code checks pointer equality on the discovered properties
[geometry nodes modifier properties are stored as ID properties], but
these are not the same across objects (since these are fetched from
NodesModifierSettings - which are different on different objects).
note: if general custom properties are "API defined" on existing classes,
this was working, we are getting the exact property for different IDs in
this case
Now be more permissive with ID properties not defined on classes in
general and dont check pointer equality for them. For ID properties on
specific IDs (not the ones defined on classes) this //might// be undesired
(havent spotted issues though, even if equally named ID properies with
different types existed -- this then simply does nothing).
For geometry nodes modifiers, new code also checks if the nodegroups are
the same [since generic naming "Input_XXX" is shared for all modifiers --
and starting to copy over things to unrelated modifiers is not desired
here].
Fixes T93983.
Maniphest Tasks: T93983
Differential Revision: https://developer.blender.org/D13573
This resolves an old TODO to deduplicate code in copy_to_selected_button
& ui_selectcontext_begin.
This is also in hindsight of adding id-property support [incl. Geometry
Nodes modifier properties] for this in the next commit.
No behavior change expected here.
ref T93983 & D13573
`<some_id>.keyframe_insert()` now sends a notifier that animation data
was changed, so that animation-related editors can properly refresh.
Since this function is quite high-level (if necessary it creates the
Action and FCurves), I thought this would be a suitable location for the
notifier. If high keyframing speed is required, it is still recommended
to use `FCurveKeyframePoints.insert(options={'FAST'})` instead.
Saving with only a filename (from Python) wasn't being prevented,
while it would successfully write the file to the working-directory,
path remapping and setting relative paths wouldn't work afterwards
as `Main.filepath` would have no directory component.
Disallow this since it's a corner case which only ever occurs
when path names without any directories are used from Python,
the overhead of expanding the working-directory for all data saving
operations isn't worthwhile.
The following changes have been made:
- bpy.ops.wm.save_mainfile() without a filepath argument
fails & reports and error when the file hasn't been saved.
Previously it would write to "untitled.blend" and set the
`G.main->filepath` to this as well.
- bpy.ops.wm.save_mainfile(filepath="untitled.blend")
fails & reports and error as the filename has no directory component.
- `BLI_path_is_abs_from_cwd` was added to check if the path would
attempt to expand to the CWD.
Revert part of the fix from 073669dd85
that initialized the file-path on first save as it's no longer needed.
Also remove relbase argument to BLI_path_normalize as the destination
file paths shouldn't use relative locations.
Currently, most node buttons are defined in `drawnode.cc` however,
this is inconvenient because it requires editing many files when adding new nodes.
The goal is to minimize the number of files needed to add or update a node.
This commit moves most of the node layout functions for composite nodes into their respected
`source/blender/nodes/composite/nodes` file.
In the future, these functions will be simplified to `node_layout` once files have their own namespace.
See {D13466} for more information.
Reviewed By: HooglyBoogly
Differential Revision: https://developer.blender.org/D13523
In almost all cases there is no difference between `G.relbase_valid`
and checking `G.main->filepath` isn't an empty string.
In many places a non-empty string is already being used instead of
`G.relbase_valid`.
The only situation where this was needed was when saving from
`wm_file_write` where they temporarily became out of sync.
This has been replaced by adding a new member to `BlendFileWriteParams`
to account for saving an unsaved file for the first time.
Reviewed By: brecht
Ref D13564
Some nodes didn't check the type of the link's socket for filtering.
Do this with a combination of manually calling the node tree's validate
links function and using the helper function for declarations.
Also clean up a few cases that added geometry sockets manually
when they can use the simpler helper function.
This patch removes no-op node editor preview code (`PR_NODE_RENDER`)
and most calls to `BKE_node_preview_init_tree`. The only remaining call is
in the compositor.
- Shader nodes previews don't seem to do anything.
- In-node previews for the texture node system doesn't work either.
This is a first step to refactoring to remove `preview_xsize`,
`preview_ysize`, and `prvr` from nodes in DNA, aligned with
the general goal of removing runtime/derived data from data
structs.
Differential Revision: https://developer.blender.org/D13578
This commit adds a search menu when links are dragged above empty
space. When releasing the drag, a menu displays all compatible
sockets with the source link. The "main" sockets (usually the first)
are weighted above other sockets in the search, so they appear first
when you type the name of the node.
A few special operators for creating a reroute or a group input node
are also added to the search.
Translation is started after choosing a node so it can be placed
quickly, since users would likely adjust the position after anyway.
A small "+" is displayed next to the cursor to give a hint about this.
Further improvements are possible after this first iteration:
- Support custom node trees.
- Better drawing of items in the search menu.
- Potential tweaks to filtering of items, depending on user feedback.
Thanks to Juanfran Matheu for developing an initial patch.
Differential Revision: https://developer.blender.org/D8286
Previously we used a `CellValue` class to hold the data for a cell,
and called a function to fill it whenever necessary. This is an
unnecessary complication when we have virtual generic arrays
and most data is already easily accessible that way anyway.
This patch removes `CellValue` and uses `fn::GVArray` to provide
access to data instead.
In the future, if rows have different types within a single column,
we can use a `GVArray` of `blender::Any` to interface with the drawing.
Along with that, the use of virtual arrays made it easy to do a
few other cleanups:
- Use selection domain interpolations from rB5841f8656d95
for the mesh selection filter.
- Change the row filter to only calculate for necessary indices.
Differential Revision: https://developer.blender.org/D13478
The constructor of MemoryProxy initializes 3 of 4 instances variables.
If a MemoryProxy is constructed and MemoryProxy::free is called
on this instance, buffer_ is undefined and 'delete buffer_;' causes errors.
Although this misuse pattern does not exist in the current codebase
it already tripped up the Address Sanitizer on various occasions
while debugging unrelated problems.
Reviewed By: jbakker
Differential Revision: https://developer.blender.org/D13569
This adds a bool field selection input to the Attribute Statistics node.
This is useful for running calculations on a subset of the input field data
rather that then whole set.
Differential Revision: https://developer.blender.org/D13520
bda9e4238a changed smooth-view
not to redraw when there were no changes made.
Redrawing is needed for repositioning the cursor.
Subscribe to changes to the 3d cursor to ensure all view ports
are updated (not just the current one).
Double the allocation size when the limit is reached instead of
increasing by a fixed number.
Also re-allocate to the exact size once complete instead of over
allocating. This gives a minor speedup in my tests ~19% faster
tessellation for ~1million faces.
This broke "test_undo.view3d_multi_mode_select" test in
"lib/tests/ui_simulate" and is likely exposed by recent changes to
bounding box calculation.
The missing check for DL_INDEX4 dates back to code from 2002 which
intended to check this but was checking for DL_INDEX3 twice
which got removed as part of a cleaned up.
This could be hidden from memory checking tools as meta-balls
over-allocate vertex arrays.
This is basically the same as rBee4ed99866fbb7ab04, the fix is
simply to check if the spline has evaluated points when deciding
the offsets into the result points array.
Fix an error if "File Preview Type" is "Auto" and there is no screen.
See D13574 for details.
Differential Revision: https://developer.blender.org/D13574
Reviewed by Julian Eisel
Whenever an exception happens in VR session code, we cancel the entire
session. Alongside that, we removed the "surface" item used to draw into
an offscreen context. This would mess up the iterator of the surface
draw loop.
Similar to 7afd84df40.
Whenever an exception happens in VR session code, we cancel the entire
session. Alongside that, we removed the "surface" item used to draw into
an offscreen context. But this may still be stored as active surface,
leading to a use-after-free when deactivating this active surface, for
example.
This replaces lost functionality from the old GN Attribute Map Range node.
This also adds vector support to the shader version of the node.
Notes:
This breaks forward compatibility as this node now uses data storage.
Reviewed By: HooglyBoogly, brecht
Differential Revision: https://developer.blender.org/D12760
Code is simpler when the uiBlocks used during drawing are simply
stored in an array. Additionally, looping can be simpler when we use
an vector to hold a temporary copy of the tree's linked list of nodes.
This patch also slightly changes how uiBlocks are "named" in
`node_uiblocks_init`. Now it uses the node name instead of the
pointer, which is helpful so we rely less on the node's address.
Differential Revision: https://developer.blender.org/D13540
This commit refactors the way the socket lists for group nodes,
and group input/output nodes are verified to match the group's
interface.
Previously the `bNodeSocket.new_sock` pointer was used to
temporarily mark the new sockets. This made the code confusing
and more complicated than necessary.
Now the old socket list is saved, and sockets are moved directly from
the old list to a new list if they match, or a new socket is created
directly in the new list.
This change is split from D13518, which aims to remove the `new_node`
and `new_sock` pointers. In the future this code might be removed
entirely in favor of using node socket declarations.
Differential Revision: https://developer.blender.org/D13543
This patch refactors the instance-realization code and adds new functionality.
* Named and anonymous attributes are propagated from instances to the
realized geometry. If the same attribute exists on the geometry and on an
instance, the attribute on the geometry has precedence.
* The id attribute has special handling to avoid creating the same id on many
output points. This is necessary to make e.g. the Random Value node work
as expected afterwards.
Realizing instance attributes has an effect on existing files, especially due to the
id attribute. To avoid breaking existing files, the Realize Instances node now has
a legacy option that is enabled for all already existing Realize Instances nodes.
Removing this legacy behavior does affect some existing files (although not many).
We can decide whether it's worth to remove the old behavior as a separate step.
This refactor also improves performance when realizing instances. That is mainly
due to multi-threading. See D13446 to get the file used for benchmarking. The
curve code is not as optimized as it could be yet. That's mainly because the storage
for these attributes might change soonish and it wasn't worth optimizing for the
current storage format right now.
```
1,000,000 x mesh vertex: 530 ms -> 130 ms
1,000,000 x simple cube: 1290 ms -> 190 ms
1,000,000 x point: 1000 ms -> 150 ms
1,000,000 x curve spiral: 1740 ms -> 330 ms
1,000,000 x curve line: 1110 ms -> 210 ms
10,000 x subdivided cylinder: 170 ms -> 40 ms
10 x subdivided spiral: 180 ms -> 180 ms
```
Differential Revision: https://developer.blender.org/D13446
We often had to use two `FieldEvaluator` instances to first evaluate
the selection and then the remaining fields. Now both can be done
with a single `FieldEvaluator`. This results in less boilerplate code in
many cases.
Performance is not affected by this change. In a separate patch we
could improve performance by reusing evaluated sub-fields that are
used by the selection and the other fields.
Differential Revision: https://developer.blender.org/D13571
Dead-lock when VR viewport drawing and depsgraph updates would fight for
the draw-manager GL lock. This didn't usually cause issues because the
depsgraph would be evaluated at this point already, except in rare
exceptions like after file writing.
Fix this by ensuring the XR surface gets its depsgraph updated after
handling notifiers, which is where regular windows also do the depsgraph
updating.
- Nest compositor pages under the compositor module
- Nest GUI, DNA/RNA & externformats modules under Blender.
- Remove modules from intern which no longer exist.
- Add intern modules (atomic, eigen, glew-mx, libc_compat, locale,
numaapi, rigidbody, sky, utfconv).
- Use 'intern_' prefix for intern modules since some of the modules
use generic terms such as locale & atomic.
Audio PTS was reset for each new file. This caused misalignment of video
and audio streams. In Blender, these files can't be loaded, other
players will fail to align audio and video.
Since timestamps are reset intentionally, reset also video stream
timestamps.
There were other bugs:
After timestamp was reset for audio, write_audio_frames started
encoding from timeline start until target frame, so each split video
had more audio than it should.
Also audio for last frame before splitting was written into new file.
Differential Revision: https://developer.blender.org/D13280
Since 88c02bf826 FFmpeg handles are freed if image is not displayed.
This change did not work correctly if strips are inside meta strip,
because overlap did not consider meta strip boundary, only strips inside
of meta strip.
Pass frame range to `sequencer_all_free_anim_ibufs`, if strip is inside
of meta strip, frame range is reduced to fit meta strip boundary, but if
meta strip is being edited, range must be set to +/-`MAXFRAME`,
otherwise playback performance would be too bad.
The difference between G.save_over and G.relbase_valid was minor.
There is one change in functionality. When saving the default-startup
file from an already loaded blend file - future save actions will
continue to write to the originally loaded file instead of prompting
the user to select a location to save the file.
This change makes saving the startup file behave the same way
"Save a Copy" does.
Reviewed By: brecht
Ref D13556
This fixes crash T94022 when selecting live viewport render with both GPU & CPU devices selected. It is caused by incorrect `KernelBVHLayout` assignment. Similar to `BVH_LAYOUT_MULTI_OPTIX` for Optix, this patch adds a `BVH_LAYOUT_MULTI_METAL` to correctly redirect to the correct Metal BVH layout type.
Reviewed By: brecht
Differential Revision: https://developer.blender.org/D13561
There were a few unused enum values: `CU_CARDINAL` and `CU_BSPLINE`
This commit cleans them up from code as they were not used for
anything meaningful.
Differential Revision: https://developer.blender.org/D13554
This builds off of rBf951aa063f7, adding a weight parameter which can
be used to change the order of items when they have the same match
score. In the future, if string searching gets a C++ API, we could
use an optional parameter for the weight, since it is not used yet.
This will be used for the node link drag search menu (D8286).
Differential Revision: https://developer.blender.org/D13559
This patch adds a randomize factor for the start/end lengths in the Length modifier.
Reviewed By: #grease_pencil, antoniov, pepeland, HooglyBoogly
Differential Revision: https://developer.blender.org/D12928
his new modifier is equals to the existing mesh modifier but adapted to grease pencil.
The underlying functions used to calculate the shrink are the same used in meshes.
{F11794101}
Reviewed By: pepeland, HooglyBoogly
Differential Revision: https://developer.blender.org/D13192
This adds the remaining bits to enable Metal on macOS. There are still
performance optimizations and other improvements planned, but it should
now be ready for early testing.
This is currently only enabled on in Arm builds for M1 GPUs. It is not
yet working on AMD or Intel GPUs.
Ref T92212
Differential Revision: https://developer.blender.org/D13503
Resolving the path to a missing pose-bone (for example),
was not raising an error as it should have.
Regression introduced in f9ccd26b03,
which didn't update collection lookup logic to fail in the case the
key of a collection wasn't found.
This implements an optimization pass for multi-function procedures.
It optimizes memory reuse by moving destruct instructions up.
For more details see the in-code comment.
In very large fields with many short lived intermediate values, this change
can improve performance 3-4x. Furthermore, in such cases, peak memory
consumption is reduced significantly (e.g. 100x lower peak memory usage).
Differential Revision: https://developer.blender.org/D13548
This came with {rBf8a0e102cf5e}.
The panel was meant specifically for the gradient tool, but since it was
given the ".weighpaint" context, it would also draw as part of generic
header toolsettings drawing.
Now remove this context on purpose and only draw this specifically from
the gradient tools ToolDef.
Maniphest Tasks: T93169
Differential Revision: https://developer.blender.org/D13268
Use "filepath" which is the current convention for naming full paths.
- Main use "name" which isn't obviously a file path.
- BlendFileData & FileGlobal used "filename" which is often
used for the name component of a path (without the directory).
These arguments must be non-null for useful functionality,
there is no need for paranoid checks.
The return value in case of invalid input for BPY_run_string_as_number
was also wrong (casting -1 to a bool, when false was expected).
Point cloud is two separate words, this just changes comments in
a few places where we were inconsistent. A small wording change
to another comment is also included.
This reduces the number of separate memory allocations done
by the multi-function procedure executor (which is used by the
field evaluation).
Now a linear memory allocator is used to allocate all intermediate
values. Furthermore, more buffers are reused when possible. This
reduces the total amount of allocated memory and improves
cache efficiency because the values are more likely to be in cache
already.
The performance improvement of this patch are most noticable
when few elements are processed by many functions. The situation
will improve even more with D13548, because then buffers can actually
be reused in practice. I measured up to 20% faster field evaluation
in extreme cases with this change.
This patch exposes the vector handle options as buttons
and aligns the UI between CurveMap and CurveProfile more closely.
- CurveMap point editing is on a single row like CurveProfile
- Tools menu is moved to the right hand side on both widgets
- Emboss curve map buttons
Reviewed By: HooglyBoogly
Differential Revision: https://developer.blender.org/D10980
Replaces `HOME` environment variable usage for user
directories like in D12802.
Reviewed By: #platform_macos, brecht
Differential Revision: https://developer.blender.org/D13212
Early return in some places also.
De-duplicate getSystemDir and getUserDir also.
Reviewed By: #platform_macos, brecht
Differential Revision: https://developer.blender.org/D13211
Calling `foreach_field_input` on a highly nested field (we do that
often) has an exponential running time in the number of nodes.
That is because the same node may be visited many times.
This made Blender freeze on some setups that should work just fine.
Now every field keeps track of its inputs all the time. That replaces
the exponential algorithm with constant time access.
This allows to use fewer evaluations (30 msec down to 23 for me) while giving more accurate results (3x-10x less relative absolute error) compared to classic ray marching.
Not a massive difference, but meh, it's better. For Cycles the speedup doesn't really matter much, but I also have a patch for Eevee support.
I've also tried Gauss-Legendre and Gauss-Lobatto - the latter was always worse, while the former was slightly better at 2deg elevation but notably worse on 15deg.
Unfortunately the same approach can't be used for the integration along the primary ray, since there we also need the accumulated transmission so far at every integration point, not just the total result.
Differential Revision: https://developer.blender.org/D13521
The const argument makes sense because these are the "source"
sockets, even though a const cast is necessary at one point.
The name "interface_socket" is an improvement over "stemp"
because the latter sounds like "temporary", or it confuses
the old socket template system with a node group's interface.
I assume this `butr` rectangle was used more in the past,
but currently its value is set and used less than 10 lines apart,
so it's trivial to remove 16 bytes from every node. The other
rectangles are also runtime data and could be removed, but
they are more difficult.
This reverts commit 943aed0de3 and f76e04bf4d
The latter caused a test failure: `sequencer_render_transform`
Reverting since the fix is not obvious (to me), and people are
away for the weekend.
Most of the fields in Mesh had no comments, or outdated misleading
comments. For example, "BMESH ONLY" referred to the BMesh project,
not the data structure. Given how much these structs are used, it should
save a lot of time to have proper comments.
I also rearranged the fields in mesh to have a more logical order. Now
the most important fields come first. In the process I was able to
remove 19 bytes of unnecessary padding (31->12). I just had to
change a `short` flag to `char`.
Differential Revision: https://developer.blender.org/D13454
The defrag shader make sure the free heap is free of holes. Making
the allocation more straightforward.
Since we now only reference the pages using the tiles, we introduce
a debug shader that produces an image with page data in a visual way.
This replaces the debug 8 option.
This also fixes some bug that were still present in the pipeline.
This patch replaces `round_fl_to_int` with `floor` and adjusts the
maximum value accordingly. The call to `round_fl_to_int` is problematic
here because it messes with the probability distribution at the edges
of the value range, meaning the first and last values were only half
as common as all other values. Since `round_fl_to_int` does
`floor(val + 0.5)`, it will not introduce misbehavior in edge cases.
Differential Revision: https://developer.blender.org/D13474
Improvement of the IMB_transform function by separating the UVWrapping
method from the color interpolation function. This would allow us to
perform uv wrapping operations separate from the interpolation function.
Currently this wrapping is only supported when interpolating colors.
This change would allow us to perform it on non-color image buffers.
Reduce the inner loop of IMB_transform by extracting writing to an
output buffer in a template. This reduces a branch in the inner loop and
would allow different number of channels in the future.
Some doc-strings were skipped because of blank-lines between
the doc-string and the symbol and needed to be moved manually.
- Added space below non doc-string comments to make it clear
these aren't comments for the symbols directly below them.
- Use doxy sections for some headers.
Ref T92709
When loading huge images (30k) blender crashed with a buffer overflow.
The reason is that determine the length of a buffer was done in 32bit
precision and afterwards stored in 64 bit precision.
This patch adds a new function to do the correct calculation in 64bit.
It should be added to other sections in blender as well. But that should
be tested on a per case situation.
When rendering volume surfaces in unbounded worlds the volume stepping can produce large values. If used with a magic texture node the values can results in a Inf float which when used in a sin or cos produces a NaN.
To fix this the input values are mapped into the periodic range of the sin and cos functions (-2*PI 2*PI) this stops the possibility of a Inf occurring and thus the NaN. It also improves the accuracy and smoothness of the result due to loss of precision when large values are summed with smaller ones effectively removing the parts of the smaller number (i.e. those in the -2PI to 2PI range) that result in variation of the output of sin and cos.
Reviewed By: brecht
Maniphest Tasks: T92036
Differential Revision: https://developer.blender.org/D12821
- Added space below non doc-string comments to make it clear
these aren't comments for the symbols directly below them.
- Use doxy sections for some headers.
Ref T92709
In the `replace_mesh`/`replace_curve` etc. methods, the component
was retrieved with write access. Retrieving with write access will
duplicate the data if the component has another user. This means that
the replaced geometry data was often duplicated just to be deleted
a moment later.
I expect this would have a large impact in performance in some specific
situations when dealing with large geometry. In a scene with many small
meshes though, I didn't observe a significant difference.
This also makes replacing a geometry set's data with the same data
that's already in the set safe. It would be valid to assert for that
case instead, but this seems safer.
Differential Revision: https://developer.blender.org/D13530
I'm planning to make these functions slightly more complicated,
and it makes sense to return early when checking one of the parameters
for null anyway.
Sample offset was not accounted for in the adaptive sampling code and caused
issues, like immediately applied adaptive filtering, with non-zero values.
Differential Revision: https://developer.blender.org/D13510
This patch adds support for merging images rendered with adaptive sampling to
the merge operator (which is currently only exposed in the Python API).
To do this an sample count buffer is created for each render layer from the
sample count pass if it exists, or from the metadata otherwise. This is then
used for averaging passes.
Differential Revision: https://developer.blender.org/D13457
Use structured binding and for-each loop, remove reduntant type casts, use
find_if instead of loop.
Differential Revision: https://developer.blender.org/D13456
This node outputs the current scene time in seconds or in frames.
Use of this node eliminates the need to use drivers to control values
in the node tree that are driven by the scene time.
Frame is a float value to provide for subframe rendering for motion
blur.
Differential Revision: https://developer.blender.org/D13455
The corresponding vertex group was renamed properly, but the armature
influence was broken for that bone.
Caused by {rB3b6ee8cee708}.
Since above commit, vertex group names are stored on object data (mesh/
lattice/gpencil) and if we update these, we have to inform dependency
graph to have immediate effect.
Maniphest Tasks: T93892
Differential Revision: https://developer.blender.org/D13526
This was an issue with the mixed list of external assets and assets from
the current file. When closing the File Browser to select the custom
preview image, the assets from the current file would be cleared for
reread, to make sure we display up-to-date file data. That is because
the workspace of the temporary File Browser was deleted, causing a
change in the file data (main data-base). The reread would happen in a
background thread, meaning it might not finish before the custom preview
operator runs and queries the active asset. So the preview operator
would get the wrong active asset from context.
Two fixes were needed:
* Make sure current file data is reread before the operator runs, by
doing this partial rereading on the main thread.
* Ensure the asset list (in fact file list) order stays consistent over
rereads. If multiple assets with the same name were shown, the
operator might also have gotten the wrong asset, also leading to a
crash.
Additionally the file operation handler should probably poll before
executing, to fail gracefully at least (not crash).
A crash happened when `instance_id_attribute` further down in the
function was null. This issue was probably introduced when the
id attribute starting using generic attribute handling.
* Correct URL for documentation (was changed recently).
* Add comments.
* Reevaluate and update which functions are public, protected or
private.
* Reorder functions and classes to be more logical and readable.
* Add helper class for the public item API so individual functions it
uses can be made protected/private (the helper class is a friend).
Also allows splitting API implementation from the C-API.
* Move internal layout builder helper class to the source file, out of
the header.
* More consistent naming.
* Add alias for item-container, so it's more clear how it can be used.
* Use const.
* Remove unnecessary forward declaration.
Adds some basic high-level explanations for editor/UI level asset APIs.
Also moves one such comment from the source file to the header file,
so it's in the same file as other API comments.
Autocomplete entires keep track of the length of the prefix
in `name_prefix_offset`. However, the name matching logic
was comparing the string including the prefix which resulted
in tab-completion not working (when the user didn't also type
in the prefix, typically two whitespaces).
This is fixed by passing in a char pointer after the end of
the prefix.
Additionally, some searchbox logic is moved. Previously,
`ui_searchbox_apply` would clear the entry which would mean
that `ui_searchbox_find_index` would never succeed. Now the
search box is only cleared if no match was found.
Differential Revision: https://developer.blender.org/D13483
- Added space below non doc-string comments to make it clear
these aren't comments for the symbols directly below them.
- Use doxy sections for some headers.
- Minor improvements to doc-strings.
Ref T92709
Adds hysteresis to the resolution divider algorithm to avoid having the resolution bounce around when on the boundary of two resolutions.
Reviewed By: brecht, leesonw
Differential Revision: https://developer.blender.org/D12385
Float precision issues cause the `ED_view3d_win_to_3d_on_plane` to return
a value even when the view ray is parallel to the plane.
A more general solution might be desired in this case, as other areas that
use `ED_view3d_win_to_3d_on_plane` might have the same problem.
For now, just work around the problem for the snap cursor.
Fixes a memory leak introduced by D13225.
Caused by not freeing the hash-map in some cases.
Differential Revision: https://developer.blender.org/D13432
The issue here was that after the seek table check, the underlying file wasn't
rewound to the start, so the code that checks for the BLENDER header
immediately reaches EOF and fails.
Since Blender always writes files with a seek table, this bug isn't triggered
by files saved in Blender itself. However, files compressed in external tools
generally don't have a seek table.
This separate the handling of directional lights (sun) into their
own loops. This will help reduce register pressure and remove some
pollution of the local light culling.
All sun lights are packed at the start of the light array.
If a transform operator is executed as modal and has a "value" set, that value works as an offset.
However few transform modes were supporting this.
Include more transform modes as supported.
This node is a field input that outputs a separate index for each mesh island.
The indices are based on the order of the lowest-numbered vertex in each island.
Authoring help from @hooglyboogly
Differential Revision: https://developer.blender.org/D13504
Caused by {rB3b6ee8cee708}
The raw data is copied here correctly
[`BKE_gpencil_stroke_weights_duplicate` in
`BKE_gpencil_stroke_duplicate`] but the vertex groups names are missing.
Prior to above commit is was enough to have `ED_object_add_duplicate`
(this seemingly took care of duplicating object's defbase).
Now vertex groups names sit on the `bGPdata` rather than the `Object`,
and since the separate operation creates **new** `bGPdata` we have to
copy vertex groups names - and active index - over [via
`BKE_defgroup_copy_list`].
Maniphest Tasks: T93728
Differential Revision: https://developer.blender.org/D13509
This patch suppresses OS version warnings and hides currently unsupported Metal GPUs when enumerating devices.
Reviewed By: brecht
Differential Revision: https://developer.blender.org/D13506
eed48a7322 caused the `SnapObjectParams` to be stored in the `SnapObjectContext`.
As this pointer is always passing in stack memory, so it seems error prone to keep a reference to this in `SnapObjectContext` since failure to set this will reference undefined stack memory.
So avoid this by moving params out of `SnapObjectContext`.
Differential Revision: https://developer.blender.org/D13401
Though the interfacing with `rctf` becomes slightly more complicated,
this should be more helpful as more of this code usese `float2` instead
of two separate floats.
We now scan the depth buffer after the prepass to tag the needed
shadow tiles.
This is much more precise than the bound box tagging which is now
reserved for transparent objects.
This also:
- fix pixel radius size.
- add a dedicated info buffer to avoid having one unused tile.
This reverts commit 5e37f70307.
It is leading to freezing of the entire desktop for a few seconds when stopping
3D viewport rendering on my Linux / NVIDIA system.
This allows real world cameras to be modeled by specifying the coordinates of a
4th degree polynomial that relates a pixels distance (in mm) from the optical
center on the sensor to the angle (in radians) of the world ray that is
projected onto that pixel.
This is available as part of the panoramic lens type, however it can also be
used to model lens distortions in projective cameras for example.
Differential Revision: https://developer.blender.org/D12691
This included is needed for the `ATTR_NONNULL` macro used in the header.
As found in a recent c --> c++ if the includes get ordered in a different order
this could result in an error.
Re commits rBc20098e6ec6adee874a12e510aa4a56d89f92838
This adds a bit more information to `GeometrySet` and each of the
geometry components. There is probably still more that can be written,
but this includes the most important information that I could think of.
I'd like to include some more general information about the
attribute API in a separate patch.
Differential Revision: https://developer.blender.org/D13501
The root of the issue is caused by Cycles ignoring OpenGL limitation on
the maximum resolution of textures: Cycles was allocating texture of the
final render resolution. It was exceeding limitation on certain GPUs and
driver.
The idea is simple: use multiple textures for the display, each of which
will fit into OpenGL limitations.
There is some code which allows the display driver to know when to start
the new tile. Also added some code to allow force graphics interop to be
re-created. The latter one ended up not used in the final version of the
patch, but it might be helpful for other drivers implementation.
The tile size is limited to 8K now as it is the safest size for textures
on many GPUs and OpenGL drivers.
Differential Revision: https://developer.blender.org/D13385
* Don't link embree / OSL when WITH_CYCLES is disabled
* Simplify lite config by disabling Cycles as a whole using this
* Remove code handling the removed WITH_CYCLES_NETWORK option
Somehow only a part of rBf4f8b6dde32b0438e0b97a6d8ebeb89802987127 ended up in
Cycles X, causing the issue that commit fixed, "OPTIX_ERROR_INVALID_VALUE" when the
system is out of memory, to show up again.
This adds the missing changes to fix that problem.
Maniphest Tasks: T93620
Differential Revision: https://developer.blender.org/D13488
Creates a new Edge Neighbors node which outputs a field
containing the number of faces connected to each edge.
Differential Revision: https://developer.blender.org/D13493
This patch adds the Metal host-side code:
- Add all core host-side Metal backend files (device_impl, queue, etc)
- Add MetalRT BVH setup files
- Integrate with Cycles device enumeration code
- Revive `path_source_replace_includes` in util/path (required for MSL compilation)
This patch also includes a couple of small kernel-side fixes:
- Add an implementation of `lgammaf` for Metal [Nemes, Gergő (2010), "New asymptotic expansion for the Gamma function", Archiv der Mathematik](https://users.renyi.hu/~gergonemes/)
- include "work_stealing.h" inside the Metal context class because it accesses state now
Ref T92212
Reviewed By: brecht
Maniphest Tasks: T92212
Differential Revision: https://developer.blender.org/D13423
This adds a new Geometry to Instance node that turns every
connected input geometry into an instance. Those instances
can for example be used in the Instance on Points node.
Differential Revision: https://developer.blender.org/D13500
The type conversions do not depend on other files in the nodes
module. Furthermore we want to use the conversions in the
geometry module without creating a dependency to the
nodes module there.
The `node_storage` functions to retrieve const and mutable structs
from a node are generated by a short macro that can be placed at the
top of each relevant file. I use this in D8286 to make code snippets
in the socket declarations much shorter, but I thought it would be
good to use it consistently everywhere else too.
The functions are also useful to avoid copy and paste errors,
like the one corrected in the cylinder node in this commit.
Differential Revision: https://developer.blender.org/D13491
Use the World viewport color when rendering pose library previews.
The World's viewport color is chosen instead of the World shading nodes,
as the latter would require rendering with `OB_RENDER` (instead of
`OB_SOLID`), which would take considerably longer.
Manifest Task: T93467
Reviewed By: sybren
Differential Revision: https://developer.blender.org/D13470
This reverts to following commits:
* rB5cad004d716da02f511bd34983ac7da820308676
* rB97e3a2d935ba9b21b127eda7ca104d4bcf4e48bd
* rBf60b95b5320f8d6abe6a629fe8fc4f1b94d0d91c
* rB0bd3cad04edf4bf9b9d3b1353f955534aa5e6740
* rBf72cc47d8edf849af98e196f721022bacf86a5e7
* rB3f7014ecc9d523997062eadd62888af5fc70a2b6
* rB0578921063fbb081239439062215f2538a31af4b
* rBc20098e6ec6adee874a12e510aa4a56d89f92838
* rBd5efda72f501ad95679d7ac554086a1fb18c1ac0
The original move to c++ that the other commits depended upon had some issues
that should be fixed before committing it again. The issues were reported in
T93797, T93809 and T93798.
We should also find a better rule for not using c-style casts going forward,
although that wouldn't have been reason enough to revert the commits.
Introducing something like a `MEM_new<T>` and `MEM_delete<T>`
function might help with the the most common case of casting the return
type of `MEM_malloc`.
Going forward, I recommend first committing the changes that don't
require converting files to c++. Then convert the shading node files
in smaller chunks. Especially don't mix fairly low risk changes like
moving some simple nodes, with higher risk changes.
Error in d5efda72f5. Was changing an iteration that would free items
to an iterator that is not safe for use in such cases.
There still seem to be significant issues with the rendering, but that's
a separate issue to be fixed.
Issue is that external engine uses the gpu info. but overwrote the
instance data. The draw manager would then detect instance data and
required the engine type to have a instance free callback.
The solution is to save some space in the engine data to hold an empty
and unused instance_data attribute to comply with `ViewportEngineData`
struct.
In the original design draw engines had to copy with a limitation that
they were not allowed to reuse complex data structures between drawing
calls. Data that could be reused were limited to:
- GPUFramebuffers
- GPUTextures
- Memory that could be removed calling MEM_freeN (storage list)
- DRWPass
This is fine when the storage list contains arrays or structs but when
more complex data types (vectors, maps) etc wasn't possible.
This patch adds instance_data that can be reused between drawing calls.
The instance_data is controlled by the draw engine and doesn't need to
be limited as described above.
When an engines stores instance_data it must implement the
`DrawEngineType.instance_free` callback to free the data.
The patch originates from eevee rewrite. But was added to master as the
image engine rewrite also has a need for it.
Reviewed By: fclem
Differential Revision: https://developer.blender.org/D13425
Dirty regions when painting are not using rcti. Meaning less
understandable code. Found issue when refactoring the image_gpu partial
update. In a future change gpu partial update API will be using rcti
also what makes the code even cleaner.
Reviewed By: campbellbarton
Differential Revision: https://developer.blender.org/D13260
- Added space below non doc-string comments to make it clear
these aren't comments for the symbols directly below them.
- Use doxy sections for some headers.
- Minor improvements to doc-strings.
Ref T92709
The line that sets the factor_prop in graph_slider_ops.c
has been left in the common invoke function by accident.
Reviewed by: Sybren A. Stüvel
Differential Revision: https://developer.blender.org/D13477
Ref: D13477
This improves both code finding, for example "color ramp" now has its own file.
And now each node has its own namespace so function names can be simplified
similar to rBfab39440e94
This commit also makes all file names use snake case instead of camel case.
Reviewed By: HooglyBoogly
Differential Revision: https://developer.blender.org/D13482
This flag is currently only used for vector sockets
so the function is limited to the vector builder.
The flag is only used by two shader nodes at the moment
and this is needed to port them over to the new socket declaration API.
Reviewed By: JacquesLucke
Differential Revision: https://developer.blender.org/D13490
Most of our field inputs are currently specific to geometry. This patch introduces
a new `GeometryFieldInput` that reduces the overhead of adding new geometry
field input.
Differential Revision: https://developer.blender.org/D13489
Creates 4 new nodes which provide topology information
for the mesh. Values are interpolated from the primary
domain in each case using basic attribute interpolation.
Vertex Neighbors
- Vertex Count
- Face Count
Face Neighbors
- Vertex Count
- Neighboring Face Count
Edge Vertices
- Vertex Index 1
- Vertex Index 2
- Position 1
- Position 2
Face Area
- Face Area
Differential Revision: https://developer.blender.org/D13343
The resample node didn't handle the case of when a spline didn't have
any evaluated points. For poly and Bezier splines we should never hit
this case, but it is expected when the number of NURBS control points
is smaller than its order, so we have to handle the case here.
It's not that obvious what to do in this case, there are a few options:
- Remove the bad splines from the result
- Generate empty splines for those inputs
- Skip resampling the bad splines, copy them to the result
- Arbitrarily generate single point splines
I chose option three, just skipping the "bad" splines. Since the node
already has a selection input, this can be described by just extending
that. "Splines with no evaluated points are implicitly deselected."
The first option would probably be valid too though.
Differential Revision: https://developer.blender.org/D13434
This commit adds the shader socket type to the new socket builder api.
As a test, this commit also converts the Add Shader node to the new API
Reviewed By: JacquesLucke
Differential Revision: https://developer.blender.org/D13485
- Create a new `bf_nodes_shader` library
- Enable unity builds for `bf_nodes_shader`, gives abount a 2.7x speed up for compile times
Reviewed By: JacquesLucke
Differential Revision: https://developer.blender.org/D13484
Use fixed-length string to convert `char[2]` to `std::string`. Otherwise
`strlen()` is called, which is problematic as the `char[2]` is not
zero-terminated.
Fix typos in the variables of the Xxf86vm libray to fix link failure
with bfd. These variables are defined in platform_unix.cmake.
Reviewed By: zeddb
Differential Revision: https://developer.blender.org/D12911
Earlier code assumed that the active strip was on the active track. This
commit detects when this assumption doesn't hold, and adds a more thorough
search of the active strip.
Caused by {rB3b6ee8cee708}
Above commit was trying to get the vertexgroup from the mesh that is
passed into `deformVertsEM` (but that can be NULL).
When can it be NULL, when is is non-NULL?
`editbmesh_calc_modifiers` only passes in a non-NULL mesh to
`deformVertsEM` under certain conditions:
- a non-deform-only modifier is handled currently
- a non-deform-only modifier preceeds the current modifier
- a deform-only modifier preceeds the current modifier (and the current
one depends on normals)
So the passed-in mesh cannot be relied on, now get the vertex group from
the context object data (like it was before the culprit commit).
Related commit: rB8f22feefbc20
Maniphest Tasks: T93611
Differential Revision: https://developer.blender.org/D13487
`ED_view3d_win_to_3d_on_plane` does not use the `clip_start` and
`clip_end` values of the scene, so the `do_clip` option can be misleading
especially in the orthographic view where the `clip_start` is negative.
For now, don't use the `do_clip` option in orthographic view.
Follow up on rB1df8abff257030ba79bc23dc321f35494f4d91c5
This puts all static functions in geometry node files into a new
namespace. This allows using unity build which can improve
compile times significantly
- The namespace name is derived from the file name.
That makes it possible to write some tooling that checks the names later on.
The filename extension (cc) is added to the namespace name as well.
This also possibly simplifies tooling but also makes it more obvious that this namespace is specific to a file.
- In the register function of every node, I added a namespace alias namespace `file_ns = blender::nodes::node_shader_*_cc`;.
This avoids some duplication of the file name and may also simplify tooling, because this line is easy to detect.
The name `file_ns` stands for "file namespace" and also indicates that this namespace corresponds to the current file.
In the future some nodes will be split up to separate files and given their own namespace
This will allow function names to be simplified similar to rBfab39440e94
Reviewed By: HooglyBoogly
Differential Revision: https://developer.blender.org/D13480
As a followup to 338c1060d5, apply the same change to the node
drawing callback. This helps to simplify code when the complexity
of a callback isn't necessary right now.
Until now the LOD selection was based on distance from camera.
Now it is based on receiver distance ratio. We compute the world
size of one view pixel along with the world size of one shadow texel.
By knowing one point distance to the light or to the view, we can
compute the pixel density ratio and deduce the corresponding LOD.
We use this to compute the min LOD during the visibility selection phase
and the "mean" LOD for usage tagging by BBoxes.
The tagging LOD is a crude approximation as it only uses the BBox
center.
Currently there are a few callbacks on `bNodeType` that do the same
thing for every node type except reroutes and frame nodes. Having a
callback for basic things complicates code and makes it harder to
understand, and reroutes and frames are special cases in larger way.
Arguably frame nodes shouldn't even be drawn like regular nodes,
given that it adds a case of O(N^2) looping through all nodes.
"Unrolling" the callbacks makes it easier to see what's happening,
and therefore easier to optimize.
Differential Revision: https://developer.blender.org/D13463
This will be useful in the future to use the new socket builder API
Aditional changes:
- Declare variables where initialized
- Do not use relative includes
Differential Revision: https://developer.blender.org/D13465
`GeometrySet` contains at most one component of each type.
Previously, a map was used to make sure that each component
type only exists once. The overhead of a map (especially with
inline storage) is rather large though. Since all component types
are known at compile time and the number of types is low,
a simple `std::array` works as well.
Some benefits of using `std::array` here:
* Looking up the component of a specific type is a bit faster.
* The size of `GeometrySet` becomes much smaller from 192 to 40 bytes.
* Debugging a `GeometrySet` in many tools becomes simpler because
one can easily see which components exists and which don't
The problem is that `current_mask` can become `selection.size()`.
The loop could also be refactored to break out of it when the end
of the selection is reached. This can be done separately.
This makes every shadow setup pass aware of the LOD chain of the tilemap
for each cubemap face.
In the free phase, we mask any LOD page that is completely covered by
higher LOD. This avoir commiting memory twice or more per area.
In the allocation phase, we check for the last valid LOD and set it
in the LOD 0 meta data. We also store the actual page location in LOD0 but
do not mark it as allocated as the LOD tile has the ownership of the page.
This patch leaves a out a few nodes:
- Group Nodes
- Image input node
- File output node
- Switch View
- Cryptomatte
These nodes above are a bit more complicated and should be worked on individually.
Differential Revision: https://developer.blender.org/D13266
This patch extracts the modal functions from GRAPH_OT_decimate
and makes them generic so they can be reused by future operators
Reviewed by: Sybren A. Stüvel
Differential Revision: https://developer.blender.org/D9326
Ref: D9326
This removes the light count limit for the forward shaded object. This
also provides a more efficient way of computing the culling directly on
the GPU. Moreover, this avoids doing multiple lighting passes for high
light counts in the deferred pipeline, improving performance.
This helps to tell when a pointer is expected to be null, and avoid
overly verbose code when dereferencing. This commit also includes
a few other cleanups in this area:
- Use const in a few places
- Use `float2` instead of `float[2]`
- Remove some unnecessary includes and old code
The change can be continued further in the future.
`IMB_transform` is used in VSE. It had a required crop parameter
for cropping the source buffer. This is not always needed.
In the image engine we want to use the use the `IMB_transform`
with wrap repeat. Both options are mutual exclusive and due
to performance reasons the wrap repeat is only available when
performing a nearest interpolation.
Instead of using RPT_ERROR, use RPT_WARNING which will not raise an
exception to Python. This broke some scripts (including FBX import)
which already check for a None return value.
Ref D13458
Reviewed By: campbellbarton
This continue the effort to implement virtual shadow mapping.
This includes:
- Spot cone culling of tile.
- Tile vs. view frustum tagging.
- Shadowmap Page allocation / freeing.
- Rendering to 4K buffer only tiles that needs it.
- Copying to shadow atlas.
This debug buffer is automatically bound if a shader is including
`common_debug_lib.glsl`. One buffer is created for each shading group
using such a shader.
The shader can then use the functions from that file to draw debug
lines. There is a hardcoded limit of line one buffer can contain. Make
sure to only output lines for a few threads at most.
Under the hood this uses a vertex buffer bound as SSBO that contains
the number of verts and all the positions and colors packed into 1 vec4.
We render by just rendering the whole buffer.
All unused vertices are initialized with NaN positions and will not be
drawn.
This is a total refactor of how shadows are handled.
We use Virtual shadow maps with different Level of details to
ensure a somewhat evenly distributed precision.
The shadow test is a really crude shadow test that will be
improved in further commit.
There is a pool of 4096 Tilemaps that are distributed between
shadowed ligths. These tilemaps are 16x16 each and reference
shadow map pages that are allocated in an atlas. Pages are only
allocated if needed (i.e: visible for rendering an object).
Page management is done on GPU using compute shaders to reduce
CPU task.
On CPU only one draw pass per updated tilemaps is issued.
This reduces the memory requirement of shadowmapping large scenes
with many lights.
Denoising make use of more memory to store and reproject the result of
previous frame to reduce noise. This only works for viewport.
There is a final bilateral filter for cleaning up noise even more.
Screen space Raytracing is supported by alpha blended surfaces.
However only opaque surfaces will be visible to the rays. This means
Alpha blended surfaces cannot reflect or refract themselves.
Denoising is not possible on alpha blended surfaces. Many samples
are needed for noise free results.
Since the cost of tracing can be very high, raytracing will only be
enabled on demand, on a per-material basis.
This simply reuse the reflection raytracing pipeline but with another
ray distribution. Only direct lighting, distant lighting and emissive
light are visible to diffuse rays.
Subsurface effect is not visible but transmittance effect is visible
to diffuse rays.
Indirect diffuse light is processed by the SSS filter.
The new pipeline is now cleaner and allows for deferred refraction.
The refractions are more accurate but are not denoised for now. More
research needs to be done in this area.
There is no feedback buffer for now, so reflections of metallic surfaces
will appear black.
The same restriction on refractive materials still holds true. They will
not appear in screen space tracing of other non refractive surfaces.
However, refractive surfaces (non-blended) can now reflect themselves
and the other surfaces with screen space reflections.
Half res tracing is not implemented back yet.
This is to automate the generation of reuse sample tables and maybe more
in the future. This is not designed to make compilation way longer than
expected.
Same as SSS this has been rewritten to support varying SSS radius.
Instead of relying on shadowmap hack to improve the transmittance
artifact (previously called translucency) we exposed a min thickness
output that will reduce the maximum of light bleeding that can happen
at the shading point. This is far from perfect but at least it is
tweakable.
The effect is now cheaper and the option to enable it is now gone.
It can always be artificially disabled by making the thickness bigger
than the sss radius.
The effect is always enabled for all SSS surfaces and will even be
applied on forward shaded object (alpha blend mode).
This only adds the output but the output is not yet used.
This thickness output is meant to control the aspect of subsurface,
refraction, absorption and volume shaders.
The value expected is the mean thickness inside the object at the
shading point. The source can be a vertex color or a texture map baked
from a raytracer.
This new implementation follows the technique described in
"Efficient screen space subsurface scattering Siggraph 2018".
Compared to the old implementation it fixes a lot of issues at
the cost of it being slower. This fixes:
- Light leaking between different objects.
- Light leaking between different surfaces with different depths.
- SSS radii are now "texturable" per pixel. No SSS surfaces limits.
- Noise should be lower.
- Precomputation is only done once for all SSS surfaces which lowers the
per material storage and precomputation time.
Implementation is also simpler as it is only a one pass processing.
We differ from the reference presentation by not precomputing the
RGB weights per samples. We actually compute them on the fly in order
to support varying SSS radii.
Notes:
- SSS IOR and SSS anisotropy are not supported.
- Object level light leak prevention might not work for high number of
objects in the scene (> 1024). In this case light leak might occur.
Adding or deleting (hidding) objects in the scene might change which
objects can leak.
This was caused by the StructArrayBuffer wrapper not being tagged as NonMovable.
The UBO was in fact being freed at creation time in debug build, but the
pointer was kept as valid in the copied wrapper.
Changing the higher level structure to not use the copy constructor to avoid this.
This is a needed change for the viewport compositor. The compositor
needs to draw to `dtxl->color` to have correct overlay / background
composition.
The solution here is to have a separate buffer that keeps the first
sample we blend from. This increases VRAM usage but it is the most
elegant option.
This integer divide by zero was evaluated to 0 on all platform but apple,
where it yields 1. The world lighting would then sample the 1 sample of the
first grid instead of its own sample.
This was caused by the blend mode that was used even with full opacity.
This caused issues with the viewport was resized and the color of the
framebuffer becomes undefined, leading to undefined values in the blend
equation.
Another fix would be to clear the viewport color on resize inside the
GPUViewport.
This is a necessary step for EEVEE's new arch. This moves more data
to the draw manager. This makes it easier to have the render or draw
engines manage their own data.
This makes more sense and cleans-up what the GPUViewport holds
Also rewrites the Texture pool manager to be in C++.
This also move the DefaultFramebuffer/TextureList and the engine related
data to a new `DRWViewData` struct. This struct manages the per view
(as in stereo view) engine data.
There is a bit of cleanup in the way the draw manager is setup.
We now use a temporary DRWData instead of creating a dummy viewport.
Differential Revision: https://developer.blender.org/D11966
This change the gbuffer layout to use more of the hardware to converting
data back and forth. Normals are encoded as two 16 bits components and
colors as R11G11B10F format.
This was motivated by the need of better quality normals. The issue is
that this increase the GBuffer size consequently. In order to balance
this we chose to merge the refraction and Diffuse/SSS data to use the
same buffer. This means we need to stochastically chose one of these
layers (so noise appear). Given that Glass BSDFs are rarely mixed
with Diffuse BSDFs, we think this is a good tradeoff.
The functions need to be declared before main as prototypes.
The appended libs will use the resources (textures, UBOs) defined at
global scope.
This removes a bit of code duplication and some long macros.
Instead of appending using `BLENDER_REQUIRE`, shaders can now ask for
libs to be added after the shader's `main()` by using the
`BLENDER_REQUIRE_POST` pragma.
Use viewspace instead of world space to compute pixel projection.
This fix issues when camera is far from origin and float precision would
produce artifacts.
This port the facing "flat" normal trick used by the gpencil engine
to EEVEE as well as the thickness mode.
The objects parameters are passed via the objectInfos UBO to avoid
much boiler plate code. However if this UBO grows too much we might have
to split it.
The normal trick for planar surfaces is quite simple to port to the
vertex shader even if it is less efficient.
However to compute it we need the objects bounds. This is passed as a
scale only through the orco factors. This will needs a bit of cleaning
at some points, with boundbox computed at object level.
Nothing much different compared to the previous implementation.
The transparent BSDF and principled BSDF now detects when the material
is potentially transparent to select the best way to render it.
This makes is possible to have AA and correct blending of the
forward rendered spheres.
However, to avoid distorded spheres we need to not support Lookdev
in panoramic projection mode.
Also remove support for LookDev when using render border for now.
This differs a bit from old implementation.
- Instead of manually adjusting the viewport we correctly place the
sphere in the vertex shader.
- Rendering happens after TAA accumulation: This is because we now
support panoramic cameras and TAA would distort the spheres.
This expose the capability of having no light and no probe (except the
world one) for specific views / code path.
The caller just need to pass 0 as extent to the `set_view()` function.
This is usefull for lookdev.
This does not include reference spheres rendering.
The approach is a bit different than before.
Now we use a `bNodeTree` to control the rendering of lookdev. This
generates a `GPUMaterial` that is stored per `Instance`. This way
rendering lookdev is just updating the temp light cache using this
material as world material. Removing the use of custom shader.
This introduces a small hack in order to bind the studiolight hdri after
the nodetree glsl parsing.
The background display however is still using a custom shader in order
to sample the world cubemap with different roughness.
The view space option of the studiolight is now faster by using a
transform before shading instead of rebaking the lightprobe constantly.
This should not have any particular impact on render time.
When evaluating surfaces, the deferred passes needs to sample the
depth buffer. But it also test against the stancil buffer.
Moreover the sampler needs to be a 2D sampler which is not the case
for cubemaps and texture2Darrays.
To overcome this we simply copy the gbuffer depth to another
temp texture using framebuffer blitting.
Some things differs from old implementation.
- Object visibility is filtered correctly without using a visibility
callback (which is to be removed).
The implementation is also more high level using less low level tricks.
A dedicated LightProbeView is created for each lightprobe cubeface to
render using all pipeline (deferred and forward).
There is still a few things not working.
Only world probe is supported for now.
The new implementation diverge from the original by randomly
selecting one lightprobe instead of sampling them all.
This speeds-up rendering a bit.
This is a small convenience. This let the render engine use this
default world if scene has no world.
World is black to keep the same behavior as before.
Shading groups are now created by the material_array_get functions
instead of passing a reference to be filled later. This avoids having
to wait later to maybe create a sub shading group.
This also simplifies different geomety type handling.
This adds a new closure selection method.
- In a first pass, weights are accumulated per output type (diffuse,
reflection, refraction).
- A random threshold is then generated before evaluating the BSDF nodes
again.
- During the evaluation pass the random threshold is decremented until
it reaches 0. At this moment the current BSDF is sampled.
For this to work, I splited the evaluation and the weighting in two
functions for all BSDF. The `*_eval` nodes are generated as dangling
nodes from the graph and only serialized after the rest of the graph.
Recalc flag on Material ID being unavailable to render engine, this
adds a simple way to detect material update by detecting shader creation
or update.
This constructs a "mirror" nodetree that feeds the closure "shader"
nodes with their respective final weight.
The tree is mirrored using simple math nodes. This is quite messy but
this is the only way to proceed without introducing special nodes.
The other issue with this method is that inputs are all uniforms even
for unplugged socket on temporary math nodes with add bloat to the
shader uniform buffer structure.
Only the part relevant to the weighting is duplicated. Other connexions
with the shading tree are reuse.
All shader nodes are updated to receive a `Weight` hidden parameter.
The original shader mixing tree is preserve to let the choice of using
either way to weight the output.
For now this is only done for the output nodes. This will need to be
extended to Closure to RGBA sub-tree.
This is the first step towards the new evaluation scheme of EEVEE
closures.
This commit contains:
- Removal of GPU_SOURCE_BUILTIN type, prefering global instead. This
avoid many boilerplate code since most of the old builtins are now
datas that are always present (i.e: view matrices, normals).
- Rewritting of codegen in C++ to use `std::stringstream`.
- Added a callback to let engine decide what to do with codegen code.
This remove a lot of needs for defines because of code order
dependency. The engine can insert the nodetree code in custom ways
to create advance effects (i.e: add displacement or vertex lighting).
Engine now returns final shader strings.
- Closure nodes evaluation replacment is a placeholder for now.
This is a port of the old material grouping. This is a bit more
clean as we use containers for each passes and other structures.
Nodetree is generated without major error for simple materials but
it is not yet used as closures are not outputed.
This adds the transparency and volume handling in the deferred
render pipeline.
Implementation is still unfinished.
To have better naming convention, I renamed object shader to surface.
This introduce a fat Gbuffer layout that groups closure data in groups
of similar BSDF. The goal is to have at least one sample for each
group to avoid too much code complexity and expected worse performance.
There is a lot of room for buffer reuse to reduce memory usage but it is
not considered a priority for now.
Add a smooth transition to avoid flickering of stochastic effects such
as soft shadows.
This use a simple blend method to progressively reveal the render
after some low sample count to avoid most of the flickering.
Parameters are hardcoded for now.
We use a new RNG to avoid correlation artifacts between Anti-Aliasing
and Shadow samples (see T68594).
The new sequence is a leap halton sequence. This makes it good with
low number of samples and yield less correlation issues.
Another change is that we directly jitter the projection matrix instead
of rotating the view matrix. This is improving convergence time and
avoid passing a second matrix to the shader.
However this case lead to discontinuity artifacts at face boders.
We might want to revert to the old rotation method for this
reason even if convergence is slower.
Now the shadows are linked to a `Light` object. The `Light` object is
linked to an `ObjectKey` to ensure persistence and deletion tracking.
The Uniform data are packed so that there is 1 `ShadowPunctualData`
per light in a `LightBatch`. This means there is only a shadowmap
limit to the number of `Shadow` in a scene.
Difference with previous implementation:
- Better texture space usage of cone and area light shadow.
- Shadows are packed in an atlas. Reducing requirements for future
features.
- Sampling is simpler because shadow matrix does everything.
This follows closely the implementation of 2.5D tiled light
culling described in the presentation:
"Improved Culling for Tiled and Clustered Rendering"
from Michal Drobot
http://advances.realtimerendering.com/s2017/2017_Sig_Improved_Culling_final.pdf
I chose the tile + Z binning approach for its high depth range support
and low CPU overhead & low memory consumption compared to the cluster
based culling. The cons is that the culling is a bit less precise in
some aspect but it is quite balanced.
The culling is done by the `Culling` object which is templated to easily
be reused for light probes cullg.
The Z-binning process is described starting from slide 20 in the
reference pdf.
I also implemented a debug pass to visualize false negative (light
culled when they shouldn't) and light evaluation density.
This is useful to detect failure case and hotspot. This could be exposed
as a developper only render pass in the future.
Some optimization of the reference implementation requires extensions
not yet added to GPU module and will be added later.
This has the basis of clustered light culling but does not yet do
it. The lights are only culled by frustum.
Its the same as if there was only one Cell for the entire Viewport.
This also wrap GPUFrameBuffer & GPUTexture inside eevee:Framebuffer
and eevee:Texture to improve managment.
Another cleanup was to put all members of `Instance` public to
avoid much complexity in accessing the data with modules
dependencies.
Also split velocity View related data to `class Velocity` and
rename previous `Velocity` to `VelocityModule`
Support infinite light count by dividing rendering into chucks of
LIGHT_MAX. Forward passes are just rendered again and deferred passes
(not implemented yet) will just have to have multiple light evaluation
passes.
This is almost the same thing as old implementation.
Differences:
- We clamp the motion vectors to their maximum when sampling the velocity buffer.
- Velocity rendering (and data manager) is separated from motion blur. This allows
outputing the motion vector render pass and in the future use motion vectors to
reproject older frames.
- Vector render pass support (only if motion blur is disabled, just like cycles).
- Velocity tiles are computed in one pass (simpler code, less CPU overhead, less
VRAM usage, maybe a bit slower but imperceivable (< 0.3ms)).
- Two velocity passes are outputed, one for motion blur fx (applied per shading view)
and one for the vector pass. This could be optimized further in the future.
- No current support for deformation & hair (to come).
Bonus addition, support for shutter curve.
Compared to the old implementation, the per time step sync function
is lighter and localized. Also it does not require a full engine
"reboot" in order to work.
Also modifies camera setup to be compatible with future camera motion
blur.
Bonus addition, support for shutter curve.
Compared to the old implementation, the per time step sync function
is lighter and localized. Also it does not require a full engine
"reboot" in order to work.
Pretty much identical to the previous implementation. With the exception
of a temporary noise function and some simplification of the CoC
computation. This also fixes issues with the Ortho depth of field.
Most of the files were modified to comply to new shader codestyle.
This also adds partial support of panoramic cameras (bokeh and
anamorphic is still buggy).
This cleansup a lot of confusion / complexity in the setup code.
Setup is closer to what cycles does now.
Also duplicates some buggy behavior of Cycles for now until this
is fixed.
This move view resolution handling to the `Camera` class that will
in the future clip and trim each view in panoramic projection.
There is a new `CameraView` that contains the `DRWView` and subview.
This way each `ShadingView` is associated to a unique `CameraView`.
ShadingView` & `CameraView` are all allocated & defined at creation time
but only the one activated by `Camera` will be rendered.
This option will make accumulation happen in a pre exposed logarithm
color space. This reduces the importance of bright pixels in the pixel
filter which will result in less aliasing in theses areas.
There is a few cases where one might want to disable this option to
match cycles better.
Render mode is really close to what the viewport render does.
Film output is done by resolving the data to the next (double buffered)
framebuffer and read back.
This also includes a bit of cleaning about naming of init() and sync()
functions.
This commit adds the Film class that handles accumulation of color and
non-color data using arbitrary projection and filter size.
A weighted accumulation (sum) is done into a data buffer with an
additional weight buffer. The sum being per pixel, it allows the input
textures that are not aligned with the output pixel grid.
Panoramic projection works by rendering a cubemap (6 views) of the scene
at the camera position. The Film filter pass then gather the pixels
using the correct Panoramic projection ensuring correct Anti-Aliasing.
For Non-color data (depth, normals) we only keep the closest value to
the target pixel center (simulating a filter size of 0).
Color data is accumulated in a log space to improve AntiAliasing output.
This is hardcoded for now.
Larger filters have poor performance but are very fast to converge.
Code Wise: This commit rename some modules to avoid possible confusion
and have better meaning. Use namespace instead of prefixes.
Added a new eevee_shared.hh file to share structure and enum definitions
between GLSL and C++.
Same idea as previous commit. This cleans-up the interface and put all
viewport related data inside the `DRWData` struct.
The draw manager is responsible for freeing it. That is the main point
of this all. In the future, we can have custom freeing method for each
engine.
This also move the DefaultFramebuffer/TextureList and the engine related
data to a new `DRWViewData` struct. This struct manages the per view
(as in stereo view) engine data.
There is a bit of cleanup in the way the draw manager is setup.
We now use a temporary DRWData instead of creating a dummy viewport.
setBUILD_CMAKE_ARGS=%BUILD_CMAKE_ARGS% -G "Visual Studio %BUILD_VS_VER%%BUILD_VS_YEAR%%BUILD_GENERATOR_POST%"%BUILD_PLATFORM_SELECT%%TESTS_CMAKE_ARGS%%CLANG_CMAKE_ARGS%%ASAN_CMAKE_ARGS%%PYDEBUG_CMAKE_ARGS%
Some files were not shown because too many files have changed in this diff
Show More
Reference in New Issue
Block a user
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
