Listing the "Blender Foundation" as copyright holder implied the Blender
Foundation holds copyright to files which may include work from many
developers.
While keeping copyright on headers makes sense for isolated libraries,
Blender's own code may be refactored or moved between files in a way
that makes the per file copyright holders less meaningful.
Copyright references to the "Blender Foundation" have been replaced with
"Blender Authors", with the exception of `./extern/` since these this
contains libraries which are more isolated, any changed to license
headers there can be handled on a case-by-case basis.
Some directories in `./intern/` have also been excluded:
- `./intern/cycles/` it's own `AUTHORS` file is planned.
- `./intern/opensubdiv/`.
An "AUTHORS" file has been added, using the chromium projects authors
file as a template.
Design task: #110784
Ref !110783.
A lot of files were missing copyright field in the header and
the Blender Foundation contributed to them in a sense of bug
fixing and general maintenance.
This change makes it explicit that those files are at least
partially copyrighted by the Blender Foundation.
Note that this does not make it so the Blender Foundation is
the only holder of the copyright in those files, and developers
who do not have a signed contract with the foundation still
hold the copyright as well.
Another aspect of this change is using SPDX format for the
header. We already used it for the license specification,
and now we state it for the copyright as well, following the
FAQ:
https://reuse.software/faq/
Goals of this refactor:
* Reduce memory consumption of `IndexMask`. The old `IndexMask` uses an
`int64_t` for each index which is more than necessary in pretty much all
practical cases currently. Using `int32_t` might still become limiting
in the future in case we use this to index e.g. byte buffers larger than
a few gigabytes. We also don't want to template `IndexMask`, because
that would cause a split in the "ecosystem", or everything would have to
be implemented twice or templated.
* Allow for more multi-threading. The old `IndexMask` contains a single
array. This is generally good but has the problem that it is hard to fill
from multiple-threads when the final size is not known from the beginning.
This is commonly the case when e.g. converting an array of bool to an
index mask. Currently, this kind of code only runs on a single thread.
* Allow for efficient set operations like join, intersect and difference.
It should be possible to multi-thread those operations.
* It should be possible to iterate over an `IndexMask` very efficiently.
The most important part of that is to avoid all memory access when iterating
over continuous ranges. For some core nodes (e.g. math nodes), we generate
optimized code for the cases of irregular index masks and simple index ranges.
To achieve these goals, a few compromises had to made:
* Slicing of the mask (at specific indices) and random element access is
`O(log #indices)` now, but with a low constant factor. It should be possible
to split a mask into n approximately equally sized parts in `O(n)` though,
making the time per split `O(1)`.
* Using range-based for loops does not work well when iterating over a nested
data structure like the new `IndexMask`. Therefor, `foreach_*` functions with
callbacks have to be used. To avoid extra code complexity at the call site,
the `foreach_*` methods support multi-threading out of the box.
The new data structure splits an `IndexMask` into an arbitrary number of ordered
`IndexMaskSegment`. Each segment can contain at most `2^14 = 16384` indices. The
indices within a segment are stored as `int16_t`. Each segment has an additional
`int64_t` offset which allows storing arbitrary `int64_t` indices. This approach
has the main benefits that segments can be processed/constructed individually on
multiple threads without a serial bottleneck. Also it reduces the memory
requirements significantly.
For more details see comments in `BLI_index_mask.hh`.
I did a few tests to verify that the data structure generally improves
performance and does not cause regressions:
* Our field evaluation benchmarks take about as much as before. This is to be
expected because we already made sure that e.g. add node evaluation is
vectorized. The important thing here is to check that changes to the way we
iterate over the indices still allows for auto-vectorization.
* Memory usage by a mask is about 1/4 of what it was before in the average case.
That's mainly caused by the switch from `int64_t` to `int16_t` for indices.
In the worst case, the memory requirements can be larger when there are many
indices that are very far away. However, when they are far away from each other,
that indicates that there aren't many indices in total. In common cases, memory
usage can be way lower than 1/4 of before, because sub-ranges use static memory.
* For some more specific numbers I benchmarked `IndexMask::from_bools` in
`index_mask_from_selection` on 10.000.000 elements at various probabilities for
`true` at every index:
```
Probability Old New
0 4.6 ms 0.8 ms
0.001 5.1 ms 1.3 ms
0.2 8.4 ms 1.8 ms
0.5 15.3 ms 3.0 ms
0.8 20.1 ms 3.0 ms
0.999 25.1 ms 1.7 ms
1 13.5 ms 1.1 ms
```
Pull Request: https://projects.blender.org/blender/blender/pulls/104629
This refactors how devirtualization is done in general and how
multi-functions use it.
* The old `Devirtualizer` class has been removed in favor of a simpler
solution. It is also more general in the sense that it is not coupled
with `IndexMask` and `VArray`. Instead there is a function that has
inputs which control how different types are devirtualized. The
new implementation is currently less general with regard to the number
of parameters it supports. This can be changed in the future, but
does not seem necessary now and would make the code less obvious.
* Devirtualizers for different types are now defined in their respective
headers.
* The multi-function builder works with the `GVArray` stored in `MFParams`
directly now, instead of first converting it to a `VArray<T>`. This reduces
some constant overhead, which makes the multi-function slightly
faster. This is only noticable when very few elements are processed though.
No functional changes or performance regressions are expected.
Goals:
* Better high level control over where devirtualization occurs. There is always
a trade-off between performance and compile-time/binary-size.
* Simplify using array devirtualization.
* Better performance for cases where devirtualization wasn't used before.
Many geometry nodes accept fields as inputs. Internally, that means that the
execution functions have to accept so called "virtual arrays" as inputs. Those
can be e.g. actual arrays, just single values, or lazily computed arrays.
Due to these different possible virtual arrays implementations, access to
individual elements is slower than it would be if everything was just a normal
array (access does through a virtual function call). For more complex execution
functions, this overhead does not matter, but for small functions (like a simple
addition) it very much does. The virtual function call also prevents the compiler
from doing some optimizations (e.g. loop unrolling and inserting simd instructions).
The solution is to "devirtualize" the virtual arrays for small functions where the
overhead is measurable. Essentially, the function is generated many times with
different array types as input. Then there is a run-time dispatch that calls the
best implementation. We have been doing devirtualization in e.g. math nodes
for a long time already. This patch just generalizes the concept and makes it
easier to control. It also makes it easier to investigate the different trade-offs
when it comes to devirtualization.
Nodes that we've optimized using devirtualization before didn't get a speedup.
However, a couple of nodes are using devirtualization now, that didn't before.
Those got a 2-4x speedup in common cases.
* Map Range
* Random Value
* Switch
* Combine XYZ
Differential Revision: https://developer.blender.org/D14628