Recently [0] replaced back-traces from `execinfo.h` with ASAN's
`__asan_describe_address` since the linking options to hide symbols
cause the stack-traces only to include addresses (without functions).
Although using ASAN makes sense when enabled, in my tests the
stack-traces are sometimes empty. Using CMAKE_BUILD_TYPE=RelWithDebInfo
for e.g. wasn't showing a stack-trace for the leak fixed in [1].
A utility is now included to conveniently expand the addresses from
these stack traces (`tools/utils/addr2line_backtrace.py`).
Restore support for the execinfo stack-trace reporting, used when ASAN
is disabled.
[0]: 2e79ca3205
[1]: a9f0d19197
The sampling pattern is only a debug setting at this point
and should not be used without the debug UI enabled where
users can actually see and edit the value.
Pull Request: https://projects.blender.org/blender/blender/pulls/112606
This PR fixes T39823, the sole failing unit test when running with MetalRT. It does so by implementing and binding a missing intersection handler (`__anyhit__cycles_metalrt_volume_test_tri`) which is required for `scene_intersect_volume` (as used by `integrator_volume_stack_update_for_subsurface`) to work as intended. This scene exposed the error as it uses subsurface scattering on a sphere which is intersected by volume.
Pull Request: https://projects.blender.org/blender/blender/pulls/112876
This patch adds `BVHMetalBuildThrottler` which limits the amount of Metal BVH building work that runs concurrently on the GPU. Previously we submitted BVH build requests to the GPU as fast as possible, but in extreme cases this could fail when the device's working set size passes safe limits.
Pull Request: https://projects.blender.org/blender/blender/pulls/112821
This keeps the behavior similar to the Disney BRDF, where 0.5
is neutral and lower/higher values respectively decrease/increase
the dielectric specular. But it's more correct in that it's not
an arbitrary scale on Fresnel, but rather adjusting the IOR.
Ref #99447
Ref #112848
Pull Request: https://projects.blender.org/blender/blender/pulls/112552
This patch fixes the memory leak described in #107714 by adding an `@autoreleasepool` around Metal BVH builds. Certain NSObjects were being retained indefinitely, specifically ones which had been value-passed via an NSArray into acceleration structure descriptors.
Pull Request: https://projects.blender.org/blender/blender/pulls/112820
This patch adds a check to see whether we're actually using NanoVDB textures, and if not, removes `#define WITH_NANOVDB` when generating the scene-optimised kernels. This results in marginally faster render times (maybe 2 or 3%) for scenes that do not use NanoVDB. The generic kernels are unaffected, so this will not impact responsiveness on first render.
Pull Request: https://projects.blender.org/blender/blender/pulls/112822
This change makes it so the list interface in the properties panels looks
closer to things like shape keys, vertex groups and so on: there are two
buttons to add selected objects to the collection and remove active item
from the collection, as well as the "extra" drop down menu.
The add operator adds selected objects to the light linking collection
using the Include policy. For the light linking it means that the objects
are added as receivers that receive the light, and for the shadow linking
it means that objects are added as blockers which cast shadow from the
light.
The communication of the active list element is done via context property
similar to how it was done before. The difference is that these properties
are set on a parent of the list layout, which makes it so they are inherited
by the layout hierarchy needed to place the Remove button.
Pull Request: https://projects.blender.org/blender/blender/pulls/112713
Using SDL's initialization logic, this is mainly a change for XDG
as LIBDECOR already required a configure event before accessing
the underlying XDG window.
While I didn't notice functional changes with this change window
flickering on startup remains an issue with some compositors
(KDE & river). Debugging these issues is simpler when both windowing
decoration systems work in a similar way & window configuration
is guaranteed to have run before the window is returned.
(via #xdg_surface_ack_configure).
since the color is applied both at entry and exit, using the square root
of the color would make the perceived color closer to the desired one.
This also makes the transition smoother when changing the `Transmission`
value in the UI, and matches the behaviour of EEVEE.
Now that there are different Fresnel types and the reflectance can be tinted,
it is better to sample based on the actually used Fresnel type, instead of
the original Fresnel. This also avoids computing Fresnel multiple times.
Pull Request: https://projects.blender.org/blender/blender/pulls/112158
This PR enabled device features that are required for EEVEE-Next.
Enabling these features would allow an initial screen drawn by
EEVEE-Next.
Note:
* Changes to EEVEE-Next after this commit, might require other changes
* Not all features are working, but a basic diffuse cube without shadows
is.
* Vulkan views are not correct and leads to incorrect lighting.
Pull Request: https://projects.blender.org/blender/blender/pulls/112519
Since 34b4487844, attributes are always made mutable when
accessed from the RNA API. This can result in unnecessary copies, which
increases memory usage and reduces performance.
Cycles is the only user of the C++ RNA API, which we'd like to remove
in the future since it doesn't really make sense in the big picture.
Hydra is now a better alternative for external render engines.
To start that change and fix the unnecessary copies, this commit
moves to use Blender headers directly for accessing attribute and
other geometry data. This also removes the few places that still had
overhead from the RNA API after the changes ([0]) in 3.6. In a simple
test with a large grid, I observed a 1.76x performance improvement,
from 1.04 to 0.59 seconds to extract the mesh data to Cycles.
[0]: https://wiki.blender.org/wiki/Reference/Release_Notes/3.6/Cycles#Performance
Pull Request: https://projects.blender.org/blender/blender/pulls/112306
This was causing a warning when using OSL, since the OSL implementation
didn't implement the input.
Since the socket isn't really implemented on the Blender side anyways,
just get rid of it.
Also, the SVM code uses the shading normal while OSL used the geometric normal.
The SVM logic was changed a while ago to not adjust normals for curves,
but this wasn't applied to OSL as well, causing differences in SVM/OSL renders.
This has two main advantages: First, it allows to get rid of the extra closure
since the remaining float can just be moved to the main closure allocation.
Second, previously sd->N was completely unused and therefore unintialized,
which ended up causing issues for the Normal render pass.
The check for control characters didn't account for delete (127).
This wasn't noticeable in most cases as delete is mapped to delete text.
Pressing Shift-Delete would enter 127 control character in the
text-editor, 3D text & Python console. This happened X11 & Wayland,
I didn't check other platforms.
