Fix ValuePlug::getValue() performance regression
#5503
Merged
Conversation
This file contains 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
|
LGTM. I was feeling bad about not having caught this - but then you said that you actually had noted the problem, tested, and not noticed the case where it went bad, so I guess I shouldn't blame myself too much ... |
This has become performance critical since 055f07d inadvertently started performing unnecessary `runTimeCast<ComputeNode>()` for _every single_ call to `ValuePlug::getValue()`. We'll be fixing that separately, but it's still worth optimising the cast.
Commit 055f07d moved the `computeNode` assignment before the fast path for plugs with static values. This caused significant slowdown when the node was implemented in Python, due to the overhead of the `runTimeCast()`. The excessive Python-specific overhead from this was fixed in the preceding commit, but the additional fix here still yields a 6% improvement in `testStaticNumericValuePerformance()` for the remaining C++-only case. This improvement will become even more significant soon, because I'm sitting on another optimisation which removes the majority of the remaining overhead for this case (the reference counting). The assignment to `computeNode` within the `if` might be a bit subtle for the reader, but I tried various other ways of factoring it, and they all resulted in at least one more conditional block and some duplicated code. I think this is the clearest way of writing it, particularly with the new assertion that shows the intended result.
johnhaddon
force-pushed
the
regressionFixes
branch
from
e7d3e35 to
f9100a6
Compare
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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Outdated suggestions cannot be applied.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.
This fixes a performance regression introduced in 1.3.1.0 by 055f07d. That commit removed unnecessary overhead for any type-converting-inputs on plugs held by Python-derived nodes, but added equivalent overhead for the far more common case of unconnected inputs on Python-derived nodes. This had a noticeable impact on scene generation times (around 3x in the case I've been debugging).