The issue was introduced by a Threadripper2 commit back in
ce927e15e0. This boils down to threads inheriting affinity
from the parent thread. It is a question how this slipped
through the review (we definitely run benchmark round).
Quick fix could have been to always set CPU group affinity
in Cycles, and it would work for Windows. On other platforms
we did not have CPU groups API finished.
Ended up making Cycles aware of NUMA topology, so now we
bound threads to a specific NUMA node. This required adding
an external dependency to Cycles, but made some code there
shorter.
Previously we would try to guess what the main tablet device is, but this is
error prone. Now we keep a list of X11 devices and try to match events to
them. On the Blender side there are still some limitations in regards to using
multiple devices at the same time, but this should improve things already.
Fixes T59645.
The issue was caused by NaN valid of the average spring length being
stored in the file. This caused accumulation in the springs builder
to also deliver NaNs, which then caused solver itself to not do
anything.
Not sure why these values where never initialized prior to the
accumulation. Or even, why this runime data is stored in a DNA.
Some sanitizing is possible here, but needs to be done with care
to not disrupt Spring production.
The key indices were wrong: need to offset curve key index
by first curve key index. Also corrected calculation of the
interpolation step.
Annoyingly, can not reproduce this on a simple file, need
production rig. For the possible future look the following
file from Spring was used: 03_005_A.lighting.debug.blend
The original issue was that different platforms will use different
hash lengths, just because defaults on Git client were different.
Now we use explicit length for the hash, and length is the same as
is used for short hashes in Linux -- apparently they started to have
collisions with length of 11.
rBa520e7c85c83 defined T_OVERRIDE_CENTER(1 << 25)
which was already in use T_PROP_PROJECTED(1 << 25)
thus skipping center calculation
Fixes T58882, T59518
Reviewers: campbellbarton, brecht
Maniphest Tasks: T58882, T59518
Differential Revision: https://developer.blender.org/D4100
This has been unbelievably painful to understand... And solution is only
partially good actually, we may even want a single axis for all the
islands in that case? But for now this is giving much better results
already, compared to the random crazyness it used to produce.
Prefer legacy OpenGL library, for the compatibility and portability
reasons.
Also use proper OpenGL libraries to be linked against, so we can
change preference to GLVND.
There are some changes in API of OpenImageIO, but those are quite
simple to keep working with older and newer library versions.
Reviewers: brecht
Reviewed By: brecht
Differential Revision: https://developer.blender.org/D4064
It used to be used for some sort of ignoring automatically
generated bump nodes. But nowadays it causes one of the shaders
in Classroom demo file to be compiled wrong.
better have this vertex color layer cover the whole 0-1 range
thx @sergey for checking
Maniphest Tasks: T57994
Differential Revision: https://developer.blender.org/D3976
This is unfortunate, but the number of bugs in this configuration
keeps growing, and almost all of them are caused by bug in OpenCL
compiler.
The compiler is not likely to be fixed, since Apple declared OpenCL
deprecated.
This evil commit is aimed to keep officially supported features
of Blender in a good working and stable state.
