Regenerate blackbody RGB curve fit to not clamp values, and extend down to
800K since it does now change below 965K.
Note that as before, blackbody is only defined in the range 800K to 12000K
and has a fixed value outside of that. But within that range there should
be no more unnecessary gamut clamping.
Found those missing casts while looking into a crash report made in
the Blender Chat. Was unable to reproduce the crash, but the casts
should totally be there to avoid integer overflow.
* Rename ambiguous rgb to scene_linear in some places
* Precompute matrices to directly go to scene instead of through XYZ
* Make function signatures more consistent
Between scene linear and sRGB, XYZ, linear Rec.709 and ACES2065-1.
And add some clarifications about color spaces in the docs.
Fixes T98267
Ref T68926
Differential Revision: https://developer.blender.org/D14989
This makes changes to the opensubdiv module to support additional vertex data
besides the vertex position, that is smootly interpolated the same way. This is
different than varying data which is interpolated linearly.
Fixes T96596: wrong generated texture coordinates with GPU subdivision. In that
bug lazy subdivision would not interpolate orcos.
Later on, this implementation can also be used to remove the modifier stack
mechanism where modifiers are evaluated a second time for orcos, which is messy
and inefficient. But that's a more risky change, this is just the part to fix
the bug in 3.2.
Differential Revision: https://developer.blender.org/D14973
Ensure the correct total/diffuse/transmission depth is set when evaluating
shaders for MNEE, consistent with regular light shader evaluation.
Differential Revision: https://developer.blender.org/D14902
Made tangent frame consistent across the surface regardless of the sample,
which was not the case with the previous algorithm. Previously, a tangent
frame would stay consistent for the same sample throughout the walk, but not
from sample to sample for the same triangle. This actually resulted in code
simplification.
Also includes additional fixes:
* Fixed an important bug that manifested itself with multiple lights in the
scene, where caustics had abnormally low amplitude: The final light pdf did
not include the light distribution pdf.
* Removed unnecessary orthonormal basis generation function, using cycles'
native one instead.
* Increased solver max iteration back to 64: It turns out we sometimes need
these extra iterations in cases where projection back to the surface takes
many steps. The effective solver iteration count, the most expensive part,
is actually much less than the raw iteration count.
Differential Revision: https://developer.blender.org/D14931
Assume geometry is always potentially animated, since we can't use our heuristic
to detect if the object is potentially animated by looking at modifiers on the
object.
The main original reason for this check was to avoid evaluating subdivision
surfaces for many static objects, which is not happening here anyway.
It was wrongly writing passes twice, for both the surface entry and exit points.
We can skip code for filtering closures, emission and holdout also, as these do
nothing with only a subsurface diffuse closure present.
evice_update_preprocess is supposed to detect modified attributes and flag the
device_vector for a copy through device_update_flags. However, since object
attributes are only created in device_update_attributes afterwards, they can't
be included in that check.
Change the function that actually updates the device_vector to tag it as
modified as soon as its content gets updated.
Differential Revision: https://developer.blender.org/D14815
A shader node setup accidentally used the bump normal as emission. Bump
mapping nodes are excluded from light shader evaluation to reduce kernel size
and register pressure, but in that case should write zero instead of leaving
memory uninitialized.
Thanks to Lukas for helping identify the cause.
The Multiscatter GGX code was missing the same-side checks for incoming and
outgoing directions w.r.t. to shading and geometry normal.
Should not be needed for the Glass variant since it intentionally has both
reflection and transmission.
Add "stageMetersPerUnit" render setting for USD files that have that set to
something other than the default (e.g. exported by Blender).
And fix a crash when an application creates a Hydra render pass on a thread
that does not have an OpenGL context current.
Long term, this should replace the XML format. This reuses the Hydra render
delegate implementation, and so supports the same features. The same command
line options and GUI work for both XML and USD also.
The implementation of this is still disabled, waiting for some refactoring of
USD library linking. However we want the Cycles code to be in sync between
repositories for the 3.2 release.
Ref T96731
* Leave code for building the render delegate against other applications and
their USD libraries to the Cycles repository, since this is not a great fit.
In the Blender repository, always use Blender's USD libraries now that they
include Hydra support.
* Hide non-USD symbols from the hdCycles shared library, to avoid library
version conflicts.
* Share Apple framework linking between the standalone app and plugin.
* Add cycles_hydra module, to be shared between the standalone app and plugin.
* Bring external libs code in sync with standalone repo, adding various missing
libraries.
* Move some cmake include directories to the top level cycles source folder
because we need to control their global order, to ensure we link against the
correct headers with mixed Blender libraries and external USD libraries.
* Float/double promotion warnings were mainly meant for avoiding slow
operatiosn in the kernel. Limit it to that to avoid hard to fix warnings
in Hydra.
* Const warnings in Hydra iterators.
* Unused variable warnings when building without glog.
* Wrong camera enum comparisons in assert.
* PASS_UNUSED is not a pass type, only for pass offsets.
Currently, the `eval` and `pdf` are not explicitly set to zero when a BSDF sample is invalid (e.g., below the upper hemisphere), when calling
`bsdf_sample` or `bsdf_eval`. It is assumed that `eval` and `pdf` are set to zero before these functions are called, which can cause problems if not.
This patch fixes this potential problem by explicitly setting `eval` and `pdf` to zero when the sampled direction is invalid.
I also added a sanity check if `eval` and `pdf` are valid (i.e., >= 0.f).
The check is activated when build in debug mode and with the `WITH_CYCLES_DEBUG` set to `ON`.
Reviewed By: brecht, sergey
Differential Revision: https://developer.blender.org/D14776
When converting from XYZ to RGB it can happen, in some sky models, that the resulting RGB values are negative.
Atm, this is not considered and the returned values for the sky model can be negative.
This patch clamps the returned RGB values to be `= 0.f`
Reviewed By: brecht, sergey
Differential Revision: https://developer.blender.org/D14777
This is a stripped down version of D14645 without the scene specialisation optimisations.
The two major changes in this patch are:
- Enables more aggressive inlining on Apple Silicon resulting in a 1.1x speedup and 10% reduction in spill, at the cost of longer pipeline build times
- Revival of shader binary archives through a new ShaderCache which is shared between MetalDevice instances using the same physical MTLDevice. This mitigates the extra compile times via explicit caching (rather than, as before, relying on the implicit system shader cache which can be purged without notice)
Reviewed By: brecht
Differential Revision: https://developer.blender.org/D14763
- Add missing doxy-section for Apply Parent Inverse Operator
- Use identity for None comparison in Python.
- Remove newline from operator doc-strings.
- Use '*' prefix multi-line C comment blocks.
- Separate filenames from doc-strings.
- Remove break after return.
The current code for computing tangents is not exactly fast.
This has been a long-standing issue, and recently came up again with T97378.
The main bottleneck is fetching the mesh data, since it's handled through a callback system and each vertex might have its data queried dozens of times.
I've tried a lot of things to optimize `mikktspace.c`, but unfortunately most weren't that useful:
- Vectorizing SVec3 gives a ~5% speedup, but I'm not sure if the additional ~70 lines of code are worth it
- Keeping an internal copy of the data instead of re-querying all the time helps a lot (~50-60% time reduction), but requires a lot of extra memory (~100 byte per face)
- Going C++ and replacing the internal quicksort with std::sort shows no difference
- Restructuring the entire file to be a header-only library so that the callbacks can be inlined gives ~10% reduction, but is a major change and deviation from the original library
In the end, two simple fixes that actually help remain:
- Don't re-query the number of faces in each loop iteration
- Don't bother looking for identical vertices if there's only one vertex with that hash
With this, time for the test case in T97378 goes from 6.64sec to 4.92sec. It's something I guess.
I feel like completely refactoring this library would not be a bad idea at some point, but for now it does the job...
Differential Revision: https://developer.blender.org/D14675
Remove need for shadow caustic caster geometry to have a UV layout. UVs were
useful to maintain a consistent tangent frame across the surface while
performing the walk. A consistent tangent frame is necessary for rough
surfaces where a normal offset encodes the sampled h, which should point
towards the same direction across the mesh.
In order to get a continuous surface parametrization without UVs, the
technique described in this paper was implemented:
"The Natural-Constraint Representation of the Path Space for Efficient
Light Transport Simulation" (Supplementary Material), SIGGRAPH 2014.
In addition to implementing this feature:
* Shadow caustic casters without smooth normals are now ignored (triggered
some refactoring and cleaning).
* Hit point calculation was refactored using existing utils functions,
simplifying the code.
* The max number of solver iterations was reduced to 32, a solution is
usually found by then.
* Added generalized geometry term clamping (transfer matrix calculation can
sometimes get unstable).
* Add stop condition to Newton solver for more consistent CPU and GPU result.
* Add support for multi scatter GGX refraction.
Fixes T96990, T96991
Ref T94120
Differential Revision: https://developer.blender.org/D14623
