This feature takes some inspiration from
"RenderMan: An Advanced Path Tracing Architecture for Movie Rendering" and
"A Hierarchical Automatic Stopping Condition for Monte Carlo Global Illumination"
The basic principle is as follows:
While samples are being added to a pixel, the adaptive sampler writes half
of the samples to a separate buffer. This gives it two separate estimates
of the same pixel, and by comparing their difference it estimates convergence.
Once convergence drops below a given threshold, the pixel is considered done.
When a pixel has not converged yet and needs more samples than the minimum,
its immediate neighbors are also set to take more samples. This is done in order
to more reliably detect sharp features such as caustics. A 3x3 box filter that
is run periodically over the tile buffer is used for that purpose.
After a tile has finished rendering, the values of all passes are scaled as if
they were rendered with the full number of samples. This way, any code operating
on these buffers, for example the denoiser, does not need to be changed for
per-pixel sample counts.
Reviewed By: brecht, #cycles
Differential Revision: https://developer.blender.org/D4686
When running with debug enabled ('-d' argument),
warnings are printed for add-ons which are not yet updated.
Reminder to name things based on what they do,
not the technologies they use :)
This sampling pattern is particularly suited to adaptive sampling, and will
be used for that upcoming feature.
Based on "Progressive Multi-Jittered Sample Sequences" by Per Christensen,
Andrew Kensler and Charlie Kilpatrick.
Ref D4686
This fixes denoising being delayed until after all rendering has finished. Instead, tile-based
denoising is now part of the "RENDER" task again, so that it is all in one task and does not
cause issues with dedicated task pools where tasks are serialized.
Reviewed By: brecht
Differential Revision: https://developer.blender.org/D6940
This makes the memory allocation for the denoiser state use the memory allocator in Cycles, which
will evict textures to host memory when there is not enough space on the device. This means the
allocation for the denoiser state won't just fail if there is no more space and instead more space is
made for it to work. Also simplifies code somewhat.
This patch adds a new user-configurable option to change at which sample viewport
denoising should kick in. Setting it to zero retains previous behavior (start immediately), while
other values will defer denoising until the particular sample has been reached. Default is now
at one, to avoid the weirdness that is AI denoising at small resolutions.
Reviewed By: brecht
Differential Revision: https://developer.blender.org/D6906
This simplifies compositors setups and will be consistent with Eevee render
passes from D6331. There's a continuum between these passes and it's not clear
there is much advantage to having them available separately.
Differential Revision: https://developer.blender.org/D6848
With the OptiX viewport denoiser active, an OptiX device was added to the device list even when
rendering and denoising on different devices (e.g. CPU or CUDA) in background rendering. This
fixes it, so the OptiX device is only added when actually needed, as in only when OptiX denoising is
actually active in background rendering.
The latest versions of Embree support Catmull-Rom splines
which use less memory than the previously used Hermite splines.
The representation is also much closer to Cycles own data structures
and can hopefully be unified in the future for more memory savings.
Memory savings using Victor benchmark scene:
Compared to previous Embree: ~400MB
Compared to Cycles' native BVH: ~1GB
The node would render black in this case (but should use the
'active_render' layer choosen in the object data properties -- this is
now in line to how this is handled for e.g. UVs)
This introduces ATTR_STD_VERTEX_COLOR and uses this thoughout, if no
particular layer is specified in the node.
Maniphest Tasks: T73938
Differential Revision: https://developer.blender.org/D6887
cycles_util_avxf_avx2_test failed on the cross test, since
it wasn't immediately clear why, the test was disabled.
After looking into it, this test when build for AVX2 is
generating FMA instructions where the intermediate results
have "infinite" precision [1] leading to slightly different
results.
This diff re-enables the cross test and allows for a small error
in the results.
[1] https://www.felixcloutier.com/x86/vfmadd132ps:vfmadd213ps:vfmadd231ps
Differential Revision: https://developer.blender.org/D6873
Reviewers: brecht
This extends FindOpenShadingLanguage.cmake to also look for the location of
stdosl.h and adds the path to the invocation of oslc to deal with the headers
being in different locations a little better.
Differential Revision: https://developer.blender.org/D6865
Reviewers: brecht
This node provides the ability to rotate a vector around a `center` point using either `Axis Angle` , `Single Axis` or `Euler` methods.
Reviewed By: #cycles, brecht
Differential Revision: https://developer.blender.org/D3789
Rendering with multiple CUDA devices but denoising with OptiX caused parts of the image to go
missing at the start while the resolution was scaled. This is because the copy operation in
`MultiDevice::map_neighbor_tiles` which slices the copy across all devices would slice based on the
full resolution and not the scaled one and therefore copy incorrect data between devices.
Since this is not the recommended way of using viewport denoising anyway, simply avoid those
incorrect copies for now by disabling denoising while the resolution is scaled. Doing both rendering
and denoising with OptiX is not affected by this, since it avoids those copies altogether anyway.
This modifies the common CUDA implementation for adaptive kernel compilation slightly to support both CUBIN and PTX output (the latter which is then used in the OptiX device). It also fixes adaptive kernel compilation on Windows.
Reviewed By: brecht
Differential Revision: https://developer.blender.org/D6851
* Direction mode X, Y and Z to align with axes rather than diagonal or
spherical as previously. X is the new default, existing files will
use diagonal or spherical for compatibility.
* Phase offset to offset the wave along its direction, for purposes like
animation and distortion.
https://developer.blender.org/D6382
This adds some extra functions recently added to the float Maths Node.
Not all functions have been ported over in this patch.
Also:
+ Tidy up menu
+ Change node color to match other vector nodes, this helps distinguish vector and float nodes in the tree
+ Move shared OSL functions to new header node_math.h
Reviewed By: brecht
Differential Revision: https://developer.blender.org/D6713
Sometimes the viewport buffer size is zero for a frame, which caused the denoising task to also try to
launch CUDA kernels with a launch size of zero, which in turn failed with a CUDA error. This patch
prevents launches from occuring in this case, similar to how it is handled in `copy_to_display_buffer`.
MSVC does not have -march=native, so the kernel gets built without AVX2 and
BVH8 support. The code assumed it to be available and crashed
Differential Revision: https://developer.blender.org/D6082
