The issue was caused by a limitation of GetNumaNodeProcessorMask():
on systems with more than 64 processors, this parameter is set to the
processor mask for the node only if the node is in the same processor
group as the calling thread. Otherwise, the parameter is set to zero.
Patch from Max Dmitrichenko, thanks!
It's extremely slow to compile and run, so just disable it unless
WITH_CYCLES_KERNEL_ASAN is manually enabled. For Clang it's always
enabled since that appears to work ok.
This also limits the -fno-sanitize=vptr flag to the Cycles kernel, as it
was added specifically to work around an issue there.
Differential Revision: https://developer.blender.org/D5404
The partial disabling was causing issues with Clang and ASAN, and it seems we
don't need to restrict it to the kernel anymore now that we are no longer using
boost directly.
The issue is that wayland seems to impose a generic device naming scheme
when using Xwayland For example any table stylus will show up with the
following naming convention: xwayland-stylus:33
For this to work in blender, I had to modify how the identifier string
is extracted. I also renamed the two char pointers in the search
algorithm to be more logical.
Reviewed By: Brecht
Differential Revision: http://developer.blender.org/D5401
In a test file containing 1300 copies of the same shader, this reduces shader update time from 3.1 sec to 0.05 sec.
Thanks to @swerner for noticing this issue.
Reviewers: brecht, sergey, swerner
Subscribers: swerner
Differential Revision: https://developer.blender.org/D5376
During viewport rendering where the viewport samples are set to 0 the UI
showed 16777216 as number of samples. We should not show the number of
samples when the number of viewport samples are set to 0.
Differential Revision: https://developer.blender.org/D5301
The code was missing some checks for whether keyframe selection
went successfully and whether reconstruction has been successfully
initialized.
The interface still gives quite generic message, with the details
printed to the console. This can be addressed separately.
The issue was caused by un-initialized local storage for volume
intersection hits which are supposed to be stored in per-thread
KernelGlobals.
Fix is to make thread_shader() be the same as thread_render() in
respect of KernelGlobals.
Reviewers: brecht
Reviewed By: brecht
Differential Revision: https://developer.blender.org/D5230
The issue was in the optimization code path for opaque shadow rays
which was wrongly considering all primitives in the node to have
same visibility flags.
We want users to go to the current version for their current version
when possible if not point to latest.
/dev should really only be for development related work. End users
should not be browsing /dev unless they are reading about upcoming
features ahead of time.
Before this change doing something like `--verbose 10 --debug-cycles`
did not properly set verbosity, only using those arguments in an other
way around was leading to a correct verbosity level.
Am not even sure that it is possible to use fstrings at all when UI
translation is required (that is, is a sensible, reasonable way that
does not make things even more complicated than they already are), but
one thing is certain, this won't be trivial to get it working, so
definitively not a job for now.
Instead just do not use fstrings for UI translatable strings.
Switch to Gregory basis patches which are tangent continuous across their
boundaries.
Originally we've used BSpline basis patches to be more compatible with the
old subdivision code, but a lot of things changed anyway.
When compute preemption is available we schedule more work which is more
efficient. However the CUDA driver appears to be incorrectly reporting this as
unavailable, even though it should be supported starting with Windows 10 1803
and Pascal and Turing (10x0 and 20x0) graphics cards.
This reduces render time by about a 25% difference on our benchmark scenes. On
Linux compute preemption appears to be reported correctly.
Replace relative threshold for pseudo inverse in sharp remeshing modifier with
0.1 as proposed in the original paper.
Also change pseudo-inverse implementation that works with dynamic heap-allocated
matrix to static 3x3 version, for performance
Differential Revision: https://developer.blender.org/D5078
