While Cycles already supports using both CPU and GPU at the same time, there
currently is a large problem with it: Since the CPU grabs one tile per thread,
at the end of the render the GPU runs out of new work but the CPU still needs
quite some time to finish its current times.
Having smaller tiles helps somewhat, but especially OpenCL rendering tends to
lose performance with smaller tiles.
Therefore, this commit adds support for tile stealing: When a GPU device runs
out of new tiles, it can signal the CPU to release one of its tiles.
This way, at the end of the render, the GPU quickly finishes the remaining
tiles instead of having to wait for the CPU.
Thanks to AMD for sponsoring this work!
Differential Revision: https://developer.blender.org/D9324
The issue stems from the fact that scene arrays are not cleared when rendering is done. This was not really an issue before the introduction of the ownership system (rB429afe0c626a) as the id_map would recreate scene data arrays based on their new content. However, now that the id_maps do not have access to the scene data anymore the arrays are never created.
Another related issue is that the BlenderSync instance is never freed when the persistent data option is activated.
To fix this, we delete nodes created by the id_maps in their destructors, and delete the BlenderSync instance before creating a new one, so the id_maps destructors are actually called.
Reviewed By: brecht
Maniphest Tasks: T82129
Differential Revision: https://developer.blender.org/D9378
Rather than just printing a message and falling back to the CPU. For render
farms it's better to avoid a potentially slow render on the CPU if the intent
was to render on the GPU.
Ref T82193, D9086
The existing code for this was incomplete. Each instance can now have a set
of attributes stored separately from geometry attributes. Geometry attributes
take precedence over instance attributes.
Ref D2057
This avoids OpenCL inlining heavy volume interpolation code once for every
data type, which could cause a performance regression when we add a float4
data type in the next commit.
Ref D2057
This encapsulates Node socket members behind a set of specific methods;
as such it is no longer possible to directly access Node class members
from exporters and parts of Cycles.
The methods are defined via the NODE_SOCKET_API macros in `graph/
node.h`, and are for getting or setting a specific socket's value, as
well as querying or modifying the state of its update flag.
The setters will check whether the value has changed and tag the socket
as modified appropriately. This will let us know how a Node has changed
and what to update, which is the first concrete step toward a more
granular scene update system.
Since the setters will tag the Node sockets as modified when passed
different data, this patch also removes the various `modified` methods
on Nodes in favor of `Node::is_modified` which checks the sockets'
update flags status.
Reviewed By: brecht
Maniphest Tasks: T79174
Differential Revision: https://developer.blender.org/D8544
Code was assuming frustrum planes are symmetrical which is not the case
for shifting. This lead to a shrinking region if shift was negative (and
a growing region if shift was positive)
So instead of only keeping track of plane on one side (and mirroring
over in code) get the actual planes after shifting and use these
instead.
This code corrects this for ortho and perspective cameras, it does not
touch panoramic cameras.
Reviewed By: brecht
Maniphest Tasks: T69911
Differential Revision: https://developer.blender.org/D9342
Approximately 141 changes of capitalization to conform to MLA title style.
Differential Revision: https://developer.blender.org/D8392
Reviewed by Julian Eisel
Don't allocate a new buffer for refitting meshes, but update the existing one.
It's not clear from the API docs if this is required, but it appears to solve
the issue and should be more efficient.
Avoid accessing mesh emitter and hair at the same time. This is not ideal for
performance, but once we have a dedicated hair object this will resolve itself.
Differential Revision: https://developer.blender.org/D9322
This improves performance in scene synchronization when there are many
mesh, hair and volume objects. Sync time speedups in benchmarks:
barbershop 5.2x
bmw 1.3x
fishycat 1.5x
koro 1.0x
sponza 3.0x
victor 1.4x
wdas_cloud 0.9x
Implementation by Nicolas Lelong, and Jagannadhan Ravi (AMD).
Differential Revision: https://developer.blender.org/D9258
Only build avx/avx2 unit tests if supported by the compiler and
WITH_CYCLES_NATIVE_ONLY is off, otherwise the appropriate compiler flags
are not available.
Corrects incorrect usage of contraction for 'it is', when possessive 'its' was required.
Differential Revision: https://developer.blender.org/D9250
Reviewed by Campbell Barton
Caused by rBe65c78cd43aa.
Since above commit, only geometry and lights received the update,
previous to this check an instancer would receive that is well (in case
of 'is_updated_shading').
Now check for an instancer (checking OB_DUPLI via ob.is_instancer()) and
do an update then as well.
Reviewers: brecht
Maniphest Tasks: T81729
Differential Revision: https://developer.blender.org/D9222
With this patch the build system checks whether the "CUDA10_NVCC_EXECUTABLE" CMake
variable is set and if so will use that to build sm_30 kernels. Similarily for sm_8x kernels it
checks "CUDA11_NVCC_EXECUTABLE". All other kernels are built using the default CUDA
toolkit. This makes it possible to use either the CUDA 10 or CUDA 11 toolkit by default and
only selectively use the other for the kernels where its a hard requirement.
Reviewed By: brecht
Differential Revision: https://developer.blender.org/D9179
After tests were bundled in a single executable and cycles and libmv
created their own tests, the warnings on macOS have gone over 800.
The reason is setting `*_LIBRARIES` to names of the libraries
and later using `link_directories` to link them properly.
https://cmake.org/cmake/help/latest/command/link_directories.html
> Note This command is rarely necessary and should be avoided where
> there are other choices. Prefer to pass full absolute paths to
> libraries where possible, since this ensures the correct library
> will always be linked. The find_library() command provides the
> full path, which can generally be used directly in calls to
> target_link_libraries().
Warnings like the following popup for every target/executable,
for every library it links to.
```
ld: warning: directory not found for option
'-L/Users/me/blender-build/blender/../lib/darwin/jpeg/lib/Debug'
```
The patch completes a step towards removing `link_directories` as
mentioned in TODO at several places.
The patch uses absolute paths to link libraries and removes
all `*_LIBPATH`s except `PYTHON_LIBPATH` from
`platform_apple.cmake` file. (The corner case where it's used seems
like dead code. Python is no longer shipped with that file structure.)
Also, unused code for LLVM-3.4 has been removed.
Also, guards to avoid searching libraries in system directories have
been added.
`APPLE` platform now no longer needs `setup_libdirs`,
`cycles_link_directories`, and `link_directories`.
The number of warnings now is less than 100, most of them being
deprecation ones in dependencies.
This patch depended on {rBb746179d0add}, {rB2fdbe4d05011},
{rB402a4cadba49} and {rBd7f482f88ecb}.
Reviewed By: brecht
Differential Revision: https://developer.blender.org/D8855
NanoVDB is a platform-independent sparse volume data structure that makes it possible to
use OpenVDB volumes on the GPU. This patch uses it for volume rendering in Cycles,
replacing the previous usage of dense 3D textures.
Since it has a big impact on memory usage and performance and changes the OpenVDB
branch used for the rest of Blender as well, this is not enabled by default yet, which will
happen only after 2.82 was branched off. To enable it, build both dependencies and Blender
itself with the "WITH_NANOVDB" CMake option.
Reviewed By: brecht
Differential Revision: https://developer.blender.org/D8794
The current way of setting the compute device makes sense for local
use, but for headless rendering it it a massive pain to get Cycles
to use the correct device, usually involving entire Python scripts.
Therefore, this patch adds a simple command-line option to Blender
for specifying the type of device that should be used. If the option
is present, the settings in the user preferences and the scene are
ignored, and instead all devices matching the specified type are used.
Differential Revision: https://developer.blender.org/D9086
