`oiiotool` diff generation is single-threaded and can become a
bottleneck on fast to render tests.
This PR generates diffs in parallel using the `multiprocessing` module.
Overlay tests (local): 90s -> 30s
buildbot +gpu:
macOS: 2020s -> 1808s
Linux: 1901s -> 1327s
Pull Request: https://projects.blender.org/blender/blender/pulls/134938
This shared state between original data and depsgraphs was added in
98a0bcd425. Other physics systems also share
the pointcache, but not the simulation state to this extent, which leads
to this kind of crash.
The mutex lock is not a great solution, you don't really want both render and viewport to be filling the same cache in parallel. However
this kind of problem also exists in other physics systems, and solving
that is certainly beyond the scope of 4.4, and probably needs to wait
for a bigger physics rewrite. In general the recommendation is to bake
everything before rendering.
Pull Request: https://projects.blender.org/blender/blender/pulls/134779
The issue was that in the macro used, the wrong ID was passed into
the rename function. Since nodetrees are a separate embedded ID,
the rename function couldn't resolve the RNA path and thus failed.
This broke keys and drivers when renaming nodes.
Not only for the compositor, but all node trees.
This was caused by 1c7c1829b6
Pull Request: https://projects.blender.org/blender/blender/pulls/134936
The previous fix 8f00f068ad
doesn't work as the printf buffer gets recreated.
Ensure render boundaries at lower level and do the printf
flush manually.
Subdivision shaders currently fail to compile using Metal as it doesn't recognize
packed_float3 as an internal data type. This PR includes packed_float3 as an
internal data type.
Without this `blender --debug-gpu-compile-shaders` will fail as it includes a namespace.
```
ERROR (gpu.shader): subdiv_normals_accumulate Compute Shader:
|
| source/blender/gpu/metal/mtl_shader_generator.mm:971:9: Error: no type named 'packed_float3' in 'MTLShaderComputeImpl'; did you mean simply 'packed_float3'?
|
| device MTLShaderComputeImpl::packed_float3* normals[[buffer(MTL_storage_buffer_base_index+4)]],
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
| packed_float3
|
| /System/Library/PrivateFrameworks/GPUCompiler.framework/Versions/32023/Libraries/lib/clang/32023.196/include/metal/metal_packed_vector:145:58: Note: 'packed_float3' declared here
|
| typedef __attribute__((__packed_vector_type__(3))) float packed_float3;
| ^
```
Pull Request: https://projects.blender.org/blender/blender/pulls/134925
This adds three new functions to find the first 0 or 1 bit in an arbitrarily
long bit span:
```cpp
blender::bits::find_first_0_index(BitSpan) -> std::optional<int64_t>
blender::bits::find_first_1_index(BitSpan) -> std::optional<int64_t>
blender::bits::find_first_1_index_expr(Expr, BitSpans...) -> std::optional<int64_t>
```
The two first ones are implemented in terms of the third. The `*_expr` variant
allows e.g. finding the first set bit when ORing two bit spans together without
computing the entire intermediate result first. Or it can be used to find the
first index where two bit spans are different.
Pull Request: https://projects.blender.org/blender/blender/pulls/134923
This happend for meshes without UVs.
The call to `texpaint_request_active_uv` in
`DRW_mesh_batch_cache_get_surface_edges` is from 9c010c44f4.
Looks like this is not needed (assumed to be copy paste error).
Pull Request: https://projects.blender.org/blender/blender/pulls/134856
In the Vertex Weights panel of the 3D Viewport sidebar, allow for
normalization of vertex groups when any number of vertex groups are
locked.
As with other normalization operators that handle locked vertex groups,
locked weights are summed, and whatever space is left between that sum
and 1.0 is used to normalize the unlocked weights.
Pull Request: https://projects.blender.org/blender/blender/pulls/134535
Make `action_channelbag_ensure()` and `action_fcurve_ensure()` return a
reference. Earlier it returned pointer that should never be `nullptr`,
but in certain cases (which would indicate a bug, as the function's
preconditions aren't met) could technically still be a `nullptr`.
Instead, the preconditions are checked with `BLI_assert()`, and a
release build will simply assume that they are met. This simplifies the
code of the functions themselves, as well as the callers.
For reference: this was discussed in !134866.
No functional changes.
Pull Request: https://projects.blender.org/blender/blender/pulls/134918
Over time the `FCurveDescriptor` class has grown, and some concern was
raised (!134866) about the performance impact of passing it by value.
It's now passed by const reference instead.
No functional changes. I did find some non-`const` uses that are now
also `const`.
Pull Request: https://projects.blender.org/blender/blender/pulls/134913
Subdiv shader library contains an optional part that shaders can include.
As we are migrating each shader at a time there are now multiple PRs doing
the same thing. Best to add the changes to main already so it doesn't add
confusion during review.
Pull Request: https://projects.blender.org/blender/blender/pulls/134906
This commit adds a test for the light falloff node on many different
objects. Examples include point lights, sun lights, mesh lights,
background lights, and direct visibility to camera.
Ref: blender/blender-test-data!56
This commit contains the following changes:
* Reduces scope of variables where possible, joining declaration and
assignment where necessary.
* Uses const for parameters and local variables where possible.
* Uses static_cast or reinterpret_cast where possible.
Pull Request: https://projects.blender.org/blender/blender/pulls/130487
* Reduces scope of variables
* Joins declaration and assignment where possible
* Uses static_cast and reinterpret_cast instead of C-style cast
* Uses const where possible
* Uses reference instead of pointer where possible
Pull Request: https://projects.blender.org/blender/blender/pulls/134816
The attribute handling code in the kernel is currently highly duplicated since
it needs to handle five different data types and we couldn't use templates
back then.
We can now, so might as well make use of it and get rid of ~1000 lines.
There are also some small fixes for the GPU OSL code:
- Wrong derivative for .w component when converting float2/float3->float4
- Different conversion for float2->float (CPU averages, GPU used to take .x)
- Removed useless code for converting to float2, not used by OSL
Pull Request: https://projects.blender.org/blender/blender/pulls/134694
This change brings the following improvements on the user level
- Support of GPUs with gfx12 architecture
- New HIP-RT library which in addition to the gfx12 support brings
various bug-fixes.
The known limitation of gfx12 is that OpenImageDenoiser does not yet
support this GPU architecture. This means that while Cycles will use the
full advantage of the gfx12 (including hardware accelerated ray-tracing),
denoising will only be possible on CPU, or secondary gfx11 or below GPU.
This is something that requires a change in OIDN and it is to late to do
it for Blender 4.4, but it is something to look forward for Blender 4.5.
The gfx12 changes for the pre-compiled kernels is rather trivial,
so it comes together (in the same PR) as the bigger HIP-RT change.
On the development side this change brings the following improvements:
- One step compile and link (much simpler CMake rules)
- Embedding BVH binaries in hiprt dll (which makes it easier to package
and load, without relying on special path configuration)
Co-authored-by: Sahar Kashi <sahar.kashi@amd.com>
Co-authored-by: Sergey Sharybin <sergey@blender.org>
Co-authored-by: Brecht Van Lommel <brecht@blender.org>
Pull Request: https://projects.blender.org/blender/blender/pulls/133129
This adds support for Overlay tests.
There are some differences with how we handle tests for other engines:
- The renders are captured using `bpy.ops.render.opengl()`, but this
won't work on our GPU build bots.
- A single blend file can run multiple tests by outputting a txt list
with the test names.
- Each overlay test blend file requires a matching script file with
the same name inside `tests/python/overlay/`.
- To reproduce a specific test state you can run
`blender "(...)/tests/data/overlay/<test>.blend" -P "(...)/tests/python/overlay/<test>.py" -- --test <test-number>`.
Note:
The current test permutations are WIP, so reference images are not
committed to the data repo for now.
Pull Request: https://projects.blender.org/blender/blender/pulls/133879
The `BKE_fcurve_bezt_shrink()` function had an artificial limitation to
only allow shrinking the `bezt` array. That limitation is now removed,
and therefore the function renamed to `BKE_fcurve_bezt_resize()`.
A note was added to the documentation that newly added array elements
should be initialized by the caller.
Pull Request: https://projects.blender.org/blender/blender/pulls/134864
Split the majority of `animrig::action_fcurve_ensure()` into a new
function `action_channelbag_ensure()`. This ensures that the Action has
a layer, keyframe strip, action slot, and channelbag for the given ID.
`animrig::action_fcurve_ensure()` now just calls into that function, and
then ensures that there is an F-Curve in that channelbag.
No functional changes.
Pull Request: https://projects.blender.org/blender/blender/pulls/134864
Followup to 48e26c3afe, and discussions in !134771 about keeping
'C-style' and 'C++ template type-safe style' implementations of our
guardedalloc separated. And it makes `MEM_freeN<T>` code simpler.
Also skip type-checking in `MEM_freeN<T>` only with MSVC, as clang-cl on
windows-arm64 does work fine with DNA structs using
`DNA_DEFINE_CXX_METHODS`.
Pull Request: https://projects.blender.org/blender/blender/pulls/134861
Hydra's viewport and final engines conflict on whether the rendering loop
is internal or external. The old approach works for Hydra Delegates that
converge with a single sample (hdStorm) but causes engines relying on sample
accumulation (hdEmrbee) to become stuck.
This minimal change retains compatibility with delegates like Storm while
replicating the behavior of the viewport engine in the final engine.
Pull Request: https://projects.blender.org/blender/blender/pulls/134804
The Grease Pencil extrusion operator didn't transfer the vertex group names
from the old `CurvesGeometry` to the new one, resulting in the `gather_attributes`
function not transferring the vertex group weights correctly. This PR adds
`BKE_defgroup_copy_list` to the operator to remedy this.
Pull Request: https://projects.blender.org/blender/blender/pulls/134695
In contrast to libamdhip64, for which blender searches in multiple directories,
libhiprt64 only had limited support for non-standard installation directories.
With this commit, the library lookup for HIPRT is handled in a similar way in
the original HIP/ROCm module. This also fixes issues for downstream Linux
distributions where HIPRT support was enabled during build time but was not
available at runtime.
Co-authored-by: Torsten Keßler <t.kessler@posteo.de>
Pull Request: https://projects.blender.org/blender/blender/pulls/131046
