This decreases BSDF_ROUGHNESS_SQ_THRESH so that the microfacet
roughness has a cutoff at much lower values and fixes a precision
issue in the bsdf_sample code that prevented this previously.
Pull Request: https://projects.blender.org/blender/blender/pulls/125919
I ran into this in a test scene - somehow the normalization here can result
in NaN (so presumably a zero vector). I don't think this has a notable
performance impact from some basic tests.
Pull Request: https://projects.blender.org/blender/blender/pulls/125930
the code snippet is supposed to compute the maximal `isect.t` in the
array, which is used to determine if subsequent intersections should be
added.
However, the previous implementation includes the old `isect.t` which is
going to be replaced, resulting an overestimation of `tmax_hits` and
thus missing closer intersections.
For BVH2, the issue is fixed by computing the `max_t` after a new entry
is inserted.
For Embree, the issue is fixed by finding the `second_largest_t` as well, and
compare that with the new insertion to find the new `max_t`.
Pull Request: https://projects.blender.org/blender/blender/pulls/125739
A phase function is normalized over the sphere, it is therefore
incorrect to sum two phase functions together when evaluating for NEE.
It should be a weighted sum with normalized weights, which, according to
`volume_shader_phase_pick()`, is `sample_weight / sum_sample_weight`.
Also corrects an error in `volume_shader_phase_pick()`.
Fix a NaN when rendering glossy materials that can appear due to a
division by zero in bsdf_D when rendering materials with low roughness.
Thank you to Weizhen for the fix after my incorrect
first attempt.
Pull Request: https://projects.blender.org/blender/blender/pulls/125756
Align Cycles SVM and EEVEE's rendering of the vector math node
in reflect mode with OSL when the normal vector is 0,0,0.
This is done by using safe_normalize rather than normalize on the
normal vector. Which also fixes a NaN in the reflect mode in this
specific configuration.
Pull Request: https://projects.blender.org/blender/blender/pulls/125688
This type of projection is often used e.g. in exhibitions that leverage big
curved screens.
Effectively, the frame is mapped onto a cylinder, with the x axis becoming the
longitude and y axis becoming the height.
Users can configure the min/max longitude, the min/max height and the radius of
the cylinder.
Co-authored-by: Lukas Stockner <lukas.stockner@freenet.de>
Pull Request: https://projects.blender.org/blender/blender/pulls/123046
After e3697710d0, if no UV map was found, then Cycles OSL would
generate UV coordinates for users. This was done to add UV coordinates
to lights, however it had the side effect of creating new UV
coordinates for other object types that don't have a UV map.
This lead to a rendering difference between OSL and SVM
when rendering meshes with no UV map, and objects with no
UV map, like curves.
This commit fixes this issue by adding a new "is_light" attribute to
Cycles OSL and using that to figure out if UV coordinates should be
generated for lights.
Pull Request: https://projects.blender.org/blender/blender/pulls/124673
The motivation is to be able to catch issues like #124705 early on,
by relying on asserts.
The not-so-obvious part of the change is the change in the order of
includes, which is needed for the types.h to have definition of the
kernel_assert().
Pull Request: https://projects.blender.org/blender/blender/pulls/124729
Update the string hashs in SSS OSL closure setup so they match the
strings being used by the SSS node.
This fixes two issues in OptiX OSL:
- SSS Random Walk would render as Random Walk Skin.
- Random Walk Skin wouldn't render at all.
Pull Request: https://projects.blender.org/blender/blender/pulls/124707
the object in volume stack should be used instead of `isect.object`.
NOTE: this solution does not work for overlapping volumes. But since
light linking of overlapping volumes did not work before, it should be
fine to implement this partial solution. We read the bottom of the stack
instead of the top to avoid looping through the entire stack.
Pull Request: https://projects.blender.org/blender/blender/pulls/124341
This PR adds a tag to prevent `kernel_data.integrator.seed` being baked into Metal pipelines as a specialisation constant when full kernel specialisation is enabled. This stops new pipelines from being continually compiled when animation is playing in live viewport mode.
Pull Request: https://projects.blender.org/blender/blender/pulls/124349
The function `direction_to_<some projection model>` computes the inverse of `<some projection model>_to_direction`.
Some of these functions had a bug where they mirror the x-axis, and some of them could be simplified.
I added round-trip tests for all of them.
This MR might change the behavior of the renderer when using equiangular_cubemap_face_to_direction:
I normalized the result vector. I looked at the usages and I think it's normalized later anyways, but someone else should probably verify that this doesn't cause issues.
Pull Request: https://projects.blender.org/blender/blender/pulls/123932
Setting this option to a value above zero replaces the lambertian Diffuse term
with the modified energy-preserving Oren-Nayar BSDF, which matches the OpenPBR
behavior.
Pull Request: https://projects.blender.org/blender/blender/pulls/123616
The function direction_to_fisheye_lens_polynomial computes the inverse of
fisheye_lens_polynomial_to_direction.
Previously the function worked almost correctly if all parameters except k_0
and k_1 were zero (in that case it was correct except for flipping the x-axis).
I replaced the fixed-point iteration (?) by Newton's method and implemented a
test to make sure it works correctly with a wider range of parameter sets.
Pull Request: https://projects.blender.org/blender/blender/pulls/123737
The function direction_to_fisheye_lens_polynomial computes the inverse of
fisheye_lens_polynomial_to_direction.
Previously the function worked almost correctly if all parameters except k_0
and k_1 were zero (in that case it was correct except for flipping the x-axis).
I replaced the fixed-point iteration (?) by Newton's method and implemented a
test to make sure it works correctly with a wider range of parameter sets.
Pull Request: https://projects.blender.org/blender/blender/pulls/123737
This is because with the addition of new features to Cycles, these GPUs
experienced significant performance regressions and bugs, all stemming
from bugs in the Metal GPU driver/compiler. The only reasonable way to
work around these issues was to disable parts of Cycles code on
these GPUs to avoid the driver/compiler bugs.
This resulted in increased development time maintaining these platforms
while being unable to deliver feature parity with other
GPU backends.
It has been decided that this development time is better spent
maintaining platforms that are still actively maintained by
hardware/software vendors, and so AMD and Intel GPU support will be
removed from the Metal backend for Cycles.
Pull Request: https://projects.blender.org/blender/blender/pulls/123551
when computing coefficients in volume, the volume density of the object
at the top of the stack is used, which leads to wrong result if
overlapping volumes have different scales.
This commit fixes the problem by pre-multiplying the volume density per
object when evaluating the shader.
Pull Request: https://projects.blender.org/blender/blender/pulls/123733
The original paper only considers the minimal distance of the cluster to
the ray, not the interval length, resulting in low weight for distant
lights that have large influence over a long distance.
This commit modifies the measure by considering `theta_b - theta_a` for
local lights and the ray length `t` for distant lights.
Pull Request: https://projects.blender.org/blender/blender/pulls/123537
Precompiled Cycles kernels make up a considerable fraction of the total size of
Blender builds nowadays. As we add more features and support for more
architectures, this will only continue to increase.
However, since these kernels tend to be quite compressible, we can save a lot
of storage by storing them in compressed form and decompressing the required
kernel(s) during loading.
By using Zstandard compression with a high level, we can get decent compression
ratios (~5x for the current kernels) while keeping decompression time low
(about 30ms in the worse case in my tests). And since we already require zstd
for Blender, this doesn't introduce a new dependency.
While the main improvement is to the size of the extracted Blender installation
(which is reduced by ~400-500MB currently), this also shrinks the download on
Windows, since .zip's deflate compression is less effective. It doesn't help on
Linux since we're already using .tar.xz there, but the smaller installed size
is still a good thing.
See #123522 for initial discussion.
Pull Request: https://projects.blender.org/blender/blender/pulls/123557
The Oren Nayer diffuse BSDF had a energy compensation term added in a
recent commit[1]. This energy compensation term used the colour input
in it's computation. The colour input was clamped in SVM, but not OSL,
resulting in differences between the two backends. This commit resolves
this issue by clamping the colour in the OSL script to match SVM.
[1] 5e40b9bb5c
Pull Request: https://projects.blender.org/blender/blender/pulls/123527
This patch implements a new Gabor noise node based on [1] but with the
improvements from [2] and the phasor formulation from [3].
We compare with the most popular existing implementation, that of OSL,
from the user's point of view:
- This implementation produces C1 continuous noise as opposed to the
non continuous OSL implementation, so it can be used for bump
mapping and is generally smother. This is achieved by windowing the
Gabor kernel using a Hann window.
- The Bandwidth input of OSL was hard-coded to 1 and was replaced with
a frequency input, which OSL hard codes to 2, since frequency is
more natural to control. This is even more true now that that Gabor
kernel is windowed as opposed to truncated, which means increasing
the bandwidth will just turn the Gaussian component of the Gabor
into a Hann window. While decreasing the bandwidth will eliminate
the harmonic from the Gabor kernel, which is the point of Gabor
noise.
- OSL had three discrete modes of operation for orienting the kernel.
Anisotropic, Isotropic, and a hybrid mode. While this implementation
provides a continuous Anisotropy parameter which users are already
familiar with from the Glossy BSDF node.
- This implementation provides not just the Gabor noise value, but
also its phase and intensity components. The Gabor noise value is
basically sin(phase) * intensity, but the phase is arguably more
useful since it does not suffer from the low contrast issues that
Gabor suffers from. While the intensity is useful to hide the
singularities in the phase.
- This implementation converges faster that OSL's relative to the
impulse count, so we fix the impulses count to 8 for simplicitly.
- This implementation does not implement anisotropic filtering.
Future improvements to the node includes implementing surface noise and
filtering. As well as extending the spectral control of the noise,
either by providing specialized kernels as was done in #110802, or by
providing some more procedural control over the frequencies of the
Gabor.
References:
[1]: Lagae, Ares, et al. "Procedural noise using sparse Gabor
convolution." ACM Transactions on Graphics (TOG) 28.3 (2009): 1-10.
[2]: Tavernier, Vincent, et al. "Making gabor noise fast and
normalized." Eurographics 2019-40th Annual Conference of the European
Association for Computer Graphics. 2019.
[3]: Tricard, Thibault, et al. "Procedural phasor noise." ACM
Transactions on Graphics (TOG) 38.4 (2019): 1-13.
Pull Request: https://projects.blender.org/blender/blender/pulls/121820
This multiscattering term comes from the OpenPBR specification and nicely
preserves energy while correctly modeling increased saturation at high
roughness.
Preparation for adding a diffuse roughness option to the Principled BSDF.
To me, the difference in output and computation seems small enough to
not need an enum for the old behavior.
Note that this also switches sampling to cosine-weighted, in my tests this
gives lower noise. I also checked doing MIS between cosine and uniform,
using the A term as a weight for how often to use cosine (since that term
is Lambertian diffuse), but always using cosine was better.
A nice consequence of that is that you don't get a huge noise jump when
going from 0.0 to 0.01 roughness.
Pull Request: https://projects.blender.org/blender/blender/pulls/123345
