Logic to for boids to avoid head-on collisions was in BLI_kdtree.
Move this into a callback which is now defined in boids.c
so the kdtree code can be kept generic.
The problem was that `balance` expected that all node children
are set to `KD_NODE_UNSET` by default.
However, this might not be the case when `balance` is called
more than once.
The balance function might change the order of nodes even
when no new point has been inserted.
This reverts the changes from ce927e1 to put the main and job threads on
node 0. The problem is that all threads created as children from these
threads will inherit the NUMA node and so will end up on the same node.
This can be fixed case-by-case by assigning the NUMA node for every child
thread, however this is difficult for external libraries and OpenMP, and
out of our control for plugins like external renderers.
Also ensure elements fit evenly into the chunk size
causing allocations to be slightly smaller in some cases.
In own tests reduces overall memory use by about ~4.5%
for high poly meshes in edit-mode.
This actually makes more sense than removing them in the
load-post callback. During load, the file might register
timers that would be removed immediately.
When there was no rotation the axis was zerod,
while not exactly a bug, it means changing the angle does nothing
and all axis-angle values are initialized with Y=1,
use this convention when resetting the axis too.
Was visible with certain configuration only, is a numeric
instability caused by degenerate ray direction.
Not sure the distribution is correct, just fixing crash
which was caused by usage of watertight intersection.
The old function was numerically very unstable for 2 reasons:
computing the square and then subtracting the results.
In the example in T60935 all precision was lost and it returned the distance 0
for all points.
I also removed the `depth` parameter since it wasn't used and computing
it would have made the code more complicated.
Reviewers: brecht, campbellbarton
Differential Revision: https://developer.blender.org/D4308
BF-admins agree to remove header information that isn't useful,
to reduce noise.
- BEGIN/END license blocks
Developers should add non license comments as separate comment blocks.
No need for separator text.
- Contributors
This is often invalid, outdated or misleading
especially when splitting files.
It's more useful to git-blame to find out who has developed the code.
See P901 for script to perform these edits.
BF-admins agree to remove header information that isn't useful,
to reduce noise.
- BEGIN/END license blocks
Developers should add non license comments as separate comment blocks.
No need for separator text.
- Contributors
This is often invalid, outdated or misleading
especially when splitting files.
It's more useful to git-blame to find out who has developed the code.
See P901 for script to perform these edits.
Free the BVH tree immediately along with the mesh, otherwise we might access
invalid mesh data.
Differential Revision: https://developer.blender.org/D4201
Not only were those often making doublons with already existing
BLI_math's stuff, but they were also used to hide implicit type
conversions...
As usual this adds some more exotic inlined vector functions (one of
the rare cases where I really miss C++ and its templates... ;) ).
Avoids nasty code all over where such math is required, and
compilers can easily deal with such situation.
Don't prefer questionable micro-optimization which comes with
a cost of nasty actual logic code.
