There is not enough time before the release to improve Random Walk to handle
all cases this was used for, so restore it for now.
Since there is no more path splitting in cycles-x, this can increase noise in
non-flat areas for the sample number of samples, though fewer rays will be traced
also. This is fundamentally a trade-off we made in the new design and why Random
Walk is a better fit. However the importance resampling we do now does help to
reduce noise.
Differential Revision: https://developer.blender.org/D12800
Only count volume objects after shader optimization.
Allows to discard objects which don't have effective volume
BSDF connected to the shader output (i.e. constant folded,
or non-volume BSDF used by mistake).
Solves memory regression reported in T92014.
There is still possibility to improve memory even further
for cases when there are a lot of non-intersecting volume
objects, but that requires a deeper refactor of update
process. Will happen as a followup development.
Differential Revision: https://developer.blender.org/D12797
The longer-term goal is to separate host-only scene update
from device update: make it possible to make kernel features
depend on actual scene state and flags.
This change makes it so shaders are compiled before kernel
load, making checks like "has_volume" available at the
kernel features calculation state.
No functional changes are expected at this point.
Differential Revision: https://developer.blender.org/D12795
Enabling or disabling motion blur requires rebuilding the BVH of affected geometry and
uploading modified vertices to the device (since without motion blur the transform is
applied to the vertex positions, whereas with motion blur this is done during traversal).
Previously neither was happening when persistent data was enabled, since the relevant
node sockets were not tagged as modified after toggling motion blur.
The change to blender_object.cpp makes it so `geom->set_use_motion_blur()` is always
called (regardless of motion blur being toggled on or off), which will tag the geometry
as modified if that value changed and ensures the BVH is updated.
The change to hair.cpp/mesh.cpp was necessary since after motion blur is disabled,
the transform is applied to the vertex positions of a mesh, but those changes were not
uploaded to the device. This is fixed now that they are tagged as modified.
Maniphest Tasks: T90666
Differential Revision: https://developer.blender.org/D12781
Previously the storage here was optimized to avoid indirections in BVH2
traversal. This helps improve performance a bit, but makes performance
and memory usage of Embree and OptiX BVHs a bit worse also. It also adds
code complexity in other parts of the code.
Now decouple triangle and curve primitive storage from BVH2.
* Reduced peak memory usage on all devices
* Bit better performance for OptiX and Embree
* Bit worse performance for CUDA
* Simplified code:
** Intersection.prim/object now matches ShaderData.prim/object
** No more offset manipulation for mesh displacement before a BVH is built
** Remove primitive packing code and flags for Embree and OptiX
** Curve segments are now stored in a KernelCurve struct
* Also happens to fix a bug in baking with incorrect prim/object
Fixes T91968, T91770, T91902
Differential Revision: https://developer.blender.org/D12766
Make volume stack allocated conditionally, potentially based on the
actual nested level of objects in the scene.
Currently the nested level is estimated by number of volume objects.
This is a non-expensive check which is probably enough in practice
to get almost perfect memory usage and performance.
The conditional allocation is a bit tricky.
For the CPU we declare and define maximum possible volume stack,
because there are only that many integrator states on the CPU.
On the GPU we declare outer SoA to have all volume stack elements,
but only allocate actually needed ones. The actually used volume
stack size is passed as a pre-processor, which seems to be easiest
and fastest for the GPU state copy.
There seems to be no speed regression in the demo files on RTX6000.
Note that scenes with high nested level of volume will now be slower
but correct.
Differential Revision: https://developer.blender.org/D12759
Implement an overscan support for tiles, so that adaptive sampling can
rely on the pixels neighbourhood.
Differential Revision: https://developer.blender.org/D12599
For example, crash when attempting to use OptiX denoiser on systems
without OptiX-capable device.
Perform check that scene update happened without errors.
Note that `et_error` makes progress to cancel, so the code was
simplified a bit.
* Add OutputDriver, replacing function callbacks in Session.
* Add PathTraceTile, replacing tile access methods in Session.
* Add more detailed comments about how this driver should be implemented.
* Add OIIOOutputDriver for Cycles standalone to output an image.
Differential Revision: https://developer.blender.org/D12627
* Split GPUDisplay into two classes. PathTraceDisplay to implement the Cycles side,
and DisplayDriver to implement the host application side. The DisplayDriver is now
a fully abstract base class, embedded in the PathTraceDisplay.
* Move copy_pixels_to_texture implementation out of the host side into the Cycles side,
since it can be implemented in terms of the texture buffer mapping.
* Move definition of DeviceGraphicsInteropDestination into display driver header, so
that we do not need to expose private device headers in the public API.
* Add more detailed comments about how the DisplayDriver should be implemented.
The "driver" terminology might not be obvious, but is also used in other renderers.
Differential Revision: https://developer.blender.org/D12626
This change fixes an issue when scene has a shadow catcher and film is
configured to be transparent. Starting viewport render and making the
background non-transparent will cause bad memory access (wrong render
and possibly crash).
Film passes depends on transparency of background, so check for this.
Demo file: F10650585
Differential Revision: https://developer.blender.org/D12666
We were writing large 2048x2048 tiles into EXR files, which appears to cause
integer overflow inside the OpenEXR library when there are multiple passes. Now
use smaller tiles in the image file, while still rendering large tiles.
This adds the requirement that the render tile size must be a multiple of 128
or be smaller than 128, this is adjusted automatically.
Goal is to add the length attribute to the Hair Info node, for better control over color gradients or similar along the hair.
Reviewed By: #eevee_viewport, brecht
Differential Revision: https://developer.blender.org/D10481
Making object which uses volume shader invisible will mark the shader
as not having a volume, forcing re-compilation of the shader to bring
it back to a consistent state.
The compilation is happening as part of scene update, which needs to
know kernel features. So there is a feedback loop.
Use more relaxed way of knowing whether there is a volume in the
shader for the kernel features, which doesn't require shader to be
compiled first.
Solves issues from the report, but potentially causes extra memory
allocated if the volume part of graph is fully optimized out. This
downside is solvable, but would need to split scene update into two
steps (the one which requires on kernel, and the one which does not).
It will be an interesting project to tackle, but for a bug fix is
better to use simpler solution.
There was an optimization to remove duplicate storage of normals as attributes
when using normal maps. However for named attributes like this we still need to
store the attribute.
Don't request normal attribute from the normal map node now, instead of skipping it
in the geometry code.
Samples count pass is normalized to the overall number of samples.
This means that we need to store actual value of the samples in the
tile buffer file.
A bit annoying to pull all those settings to BufferParams and need
to find a more generic solution, but for now this is easiest and a
quickest solution.
Differential Revision: https://developer.blender.org/D12597
This includes much improved GPU rendering performance, viewport interactivity,
new shadow catcher, revamped sampling settings, subsurface scattering anisotropy,
new GPU volume sampling, improved PMJ sampling pattern, and more.
Some features have also been removed or changed, breaking backwards compatibility.
Including the removal of the OpenCL backend, for which alternatives are under
development.
Release notes and code docs:
https://wiki.blender.org/wiki/Reference/Release_Notes/3.0/Cycleshttps://wiki.blender.org/wiki/Source/Render/Cycles
Credits:
* Sergey Sharybin
* Brecht Van Lommel
* Patrick Mours (OptiX backend)
* Christophe Hery (subsurface scattering anisotropy)
* William Leeson (PMJ sampling pattern)
* Alaska (various fixes and tweaks)
* Thomas Dinges (various fixes)
For the full commit history, see the cycles-x branch. This squashes together
all the changes since intermediate changes would often fail building or tests.
Ref T87839, T87837, T87836
Fixes T90734, T89353, T80267, T80267, T77185, T69800
When rendering the test scene in T79190 which has only emissive surfaces a division by zero occurs. This is a simple patch to remove this.
Reviewed By: brecht
Maniphest Tasks: T79190
Differential Revision: https://developer.blender.org/D11682
This adds a setting to enable data caching, and another one to set the
maximum cache size in megabytes.
When caching is enabled we load the data for the entire animation in
memory, as we already do, however, if the data exceeds the memory limit,
render is aborted.
When caching is disabled, we simply load the data for the current frame
in memory.
Ref D10197
Reviewed By: brecht
Differential Revision: https://developer.blender.org/D11163
As subdivision objects are first class citizens in Alembic, to
differentiate them with non-subdivided polygon meshes, the Alembic
Procedural automatically sets up subdivision properties on the generated
Cycles Mesh.
However, for real-time playback subdivision is far too slow, so this
modifies the detection of a MeshSeqCache modifier used to activate the
procedural to allow for a Subsurf modifier right after the cache one. If
present, the procedural will tag the object for subdivision, if absent, the
object will be treated as a regular mesh.
This is a temporary measure for until subdivision surface settings are part
of the Mesh datablock (see T68891).
Reviewed By: brecht
Differential Revision: https://developer.blender.org/D11162
The main reason for this is to speed up updates by avoid unnecessary
copies as the Generated coordinates are a copy of the vertices.
Creating this attribute may become optional in the future, with UI
parameters to select which attribute to use from the Alembic archive as
reference.
Vertex normals are needed for normals maps and therefore are packed and send
to the device alongside the other float3 attributes. However, we already pack
and send vertex normals through `DeviceScene.tri_vnormal`.
This removes the packing of vertex normals from the attributes buffer, and
reuses `tri_vnormal` in the kernel for normals lookup for normal maps, which
reduces memory usage a bit, and speeds up device updates.
This also fixes potential missing normals updates following rB12a06292af86,
since the need for vertex normals for normals maps was overlooked.
Reviewed By: brecht
Differential Revision: https://developer.blender.org/D12237
Standard attributes are not added to the attributes requests when
shaders only have displacement. This is because nodes are only
considering the case when the surface socket is connected.
To support this, added `Shader.has_surface_link()` which checks for both
cases (`has_surface` and `has_displacement`) and replaces all checks on
`Shader.has_surface`.
Reviewed By: brecht
Differential Revision: https://developer.blender.org/D12240
Caused by 4f64fa4f86.
Was a bad backport from the Cycles X branch: the fact that CPU and GPU
has different reset code paths was not taken into account.
This is a backport of recent development in the Cycles X branch.
Fixes possible dead-lock in viewport rendering when exiting at an
exact bad moment (couldn't reproduce in master branch, but in the
cycles-x branch it was happening every now and then).
Differential Revision: https://developer.blender.org/D12154
Makes it consistent with the guidelines and the Cycles X branch, and
allows to backport fix for the viewport update from the branch. Will
cause a merge conflict, which should be simple accept-ours in the
branch.
The goal: allow to easily use AO approximation in scenes which combines
both small and large scale objects.
The idea: use per-object AO distance which will allow to override world
settings. Instancer object will "propagate" its AO distance to all its
instances unless the instance defines own distance (this allows to
modify AO distance in the shot files, without requiring to modify props
used in the shots.
Available from the new Fats GI Approximation panel in object properties.
Differential Revision: https://developer.blender.org/D12112
Motion attributes expects mesh to have non-zero number of motion steps,
which was violated in the case when fluid mesh had motion blur disabled.
This is a bit of annoying fix, because of the order of updates. More
ideal solution would be to handle cached and fluid velocities in the
sync_mesh_motion() which ensures all the dependencies between settings.
WITH_CYCLES_DEBUG was used for rendering BVH debugging passes. But since we
mainly use Embree an OptiX now, this information is no longer important.
WITH_CYCLES_DEBUG_NAN will enable additional checks for NaNs and invalid values
in the kernel, for Cycles developers. Previously these asserts where enabled in
all debug builds, but this is too likely to crash Blender in scenes that render
fine regardless of the NaNs. So this is behind a CMake option now.
Fixes T90240
