Endianness based Entity representation #3788
Closed
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
james7132
added
A-ECS
Member
Author
|
Here's a updated set of benchmarks from my machine (running off of a AMD 5950x). 34 benchmarks saw a tangible improvement, 24 saw a tangible regression in performance. |
Contributor
|
Most pf the regressions don’t look too bad and the overall improvements are substantial especial on 2 or 3 of the benchmarks. I also think that those looking for determinism will be okay having to specify that themselves with regard to endianness of entities as this will not be the only concern they need to make sure they get right when seeking such determinism. I am in favour of this change. LGTM! |
alice-i-cecile
approved these changes
Jan 28, 2022
Member
alice-i-cecile
left a comment
alice-i-cecile
left a comment
There was a problem hiding this comment.
I like the look of the benchmarks, and the code changes are very well documented. Presumably this will play fine with niching?
Member
|
In particular, the improvements to the |
Contributor
|
What is niching @alice-i-cecile ? |
Member
The ability to store an |
Contributor
|
Ah yes, that is indeed quite awesome! As far as I can tell this PR should play well with niching. |
MinerSebas
reviewed
Jan 28, 2022
Member
Author
Niching only really has an effect on the assembly generated for constructing an Entity. |
james7132
force-pushed
the
aligned-entity
branch
from
913acba to
c5708d9
Compare
james7132
force-pushed
the
aligned-entity
branch
from
c5708d9 to
bd81fc8
Compare
Member
Author
|
Reran benchmarks now that we have |
Contributor
|
I'll leave the interpretation to you but here are the results from an M1 Max: |
Weibye
added
the
S-Adopt-Me
Member
Author
|
Closing in favor of #10558. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
6 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Objective
Fixes #2346. Optimize various functions on
Entity. This impacts the performance of all corners of ECS and the engine as a whole.Solution
Use
#[repr(C, align(8))]to makeEntitya pusedo-u64. This is reliant on the endianness of the target platform, so two different definitions of Entity are made, one for little endian and one for big endian. This significantly improves the assembly generated for many operations without adding to the complexity of the type's implementation (see https://godbolt.org/z/jcxY5G9r4). The one major cost is that accessinggenerationby itself requires one more instruction.Since the order of the fields is no longer consistent from platform to platform, a explicit implementation of
PartialOrdis required to enforce correctness. Asto_bitsbecomes a no-op when inlined, this comparison, while looking complex, boils down to a single u64 comparison.This PR aims to get the benefits of #2372 without introducing significant code complexity.