perf(ext/event): optimize Event constructor #20181
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
|
If we want to adhere to the specification, I will remove the changes to |
marcosc90
force-pushed
the
perf-event-constructor
branch
from
03a8c9e
to
2a506cc
Compare
|
After getting more familiar with the spec, I noticed we can safely remove from ~2m iter/s to ~27m iter/s |
|
Amazing results @marcosc90! |
|
Amazing result! |
littledivy
reviewed
Aug 17, 2023
Comment on lines
+415
to
+417
| // Not spec compliant. The spec defines it as [LegacyUnforgeable] | ||
| // but doing so has a big performance hit | ||
| ReflectDefineProperty(Event.prototype, "isTrusted", { |
littledivy
pushed a commit
to littledivy/deno
that referenced
this pull request
Aug 21, 2023
This PR optimizes `Event` constructor - ~Added a fast path for empty `eventInitDict`~ Removed `EventInit` dictionary converter - Don't make `isTrusted` a [LegacyUnforgeable](https://webidl.spec.whatwg.org/#LegacyUnforgeable) property. Doing so makes it non-spec compliant but calling `Object/Reflect.defineProperty` on the constructor is a big bottleneck. Node did the same a few months ago nodejs/node#46974. In my opinion, the performance gains are worth deviating from the spec for a browser-related property. **This PR** ``` cpu: 13th Gen Intel(R) Core(TM) i9-13900H runtime: deno 1.36.1 (x86_64-unknown-linux-gnu) benchmark time (avg) iter/s (min … max) p75 p99 p995 ------------------------------------------------------------------------------- ----------------------------- event constructor no init 36.69 ns/iter 27,257,504.6 (33.36 ns … 42.45 ns) 37.71 ns 39.61 ns 40.07 ns event constructor 36.7 ns/iter 27,246,776.6 (33.35 ns … 56.03 ns) 37.73 ns 40.14 ns 41.74 ns ``` **main** ``` cpu: 13th Gen Intel(R) Core(TM) i9-13900H runtime: deno 1.36.1 (x86_64-unknown-linux-gnu) benchmark time (avg) iter/s (min … max) p75 p99 p995 ------------------------------------------------------------------------------- ----------------------------- event constructor no init 380.48 ns/iter 2,628,275.8 (366.66 ns … 399.39 ns) 384.58 ns 398.27 ns 399.39 ns event constructor 480.33 ns/iter 2,081,882.6 (466.67 ns … 503.47 ns) 484.27 ns 501.28 ns 503.47 ns ``` ```js Deno.bench("event constructor no init", () => { const event = new Event("foo"); }); Deno.bench("event constructor", () => { const event = new Event("foo", { bubbles: true, cancelable: false }); }); ``` towards denoland#20167
littledivy
pushed a commit
that referenced
this pull request
Aug 21, 2023
This PR optimizes `Event` constructor - ~Added a fast path for empty `eventInitDict`~ Removed `EventInit` dictionary converter - Don't make `isTrusted` a [LegacyUnforgeable](https://webidl.spec.whatwg.org/#LegacyUnforgeable) property. Doing so makes it non-spec compliant but calling `Object/Reflect.defineProperty` on the constructor is a big bottleneck. Node did the same a few months ago nodejs/node#46974. In my opinion, the performance gains are worth deviating from the spec for a browser-related property. **This PR** ``` cpu: 13th Gen Intel(R) Core(TM) i9-13900H runtime: deno 1.36.1 (x86_64-unknown-linux-gnu) benchmark time (avg) iter/s (min … max) p75 p99 p995 ------------------------------------------------------------------------------- ----------------------------- event constructor no init 36.69 ns/iter 27,257,504.6 (33.36 ns … 42.45 ns) 37.71 ns 39.61 ns 40.07 ns event constructor 36.7 ns/iter 27,246,776.6 (33.35 ns … 56.03 ns) 37.73 ns 40.14 ns 41.74 ns ``` **main** ``` cpu: 13th Gen Intel(R) Core(TM) i9-13900H runtime: deno 1.36.1 (x86_64-unknown-linux-gnu) benchmark time (avg) iter/s (min … max) p75 p99 p995 ------------------------------------------------------------------------------- ----------------------------- event constructor no init 380.48 ns/iter 2,628,275.8 (366.66 ns … 399.39 ns) 384.58 ns 398.27 ns 399.39 ns event constructor 480.33 ns/iter 2,081,882.6 (466.67 ns … 503.47 ns) 484.27 ns 501.28 ns 503.47 ns ``` ```js Deno.bench("event constructor no init", () => { const event = new Event("foo"); }); Deno.bench("event constructor", () => { const event = new Event("foo", { bubbles: true, cancelable: false }); }); ``` towards #20167
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 PR optimizes
EventconstructorAdded a fast path for emptyRemovedeventInitDictEventInitdictionary converterisTrusteda LegacyUnforgeable property. Doing so makes it non-spec compliant but callingObject/Reflect.definePropertyon the constructor is a big bottleneck. Node did the same a few months ago lib: define Event.isTrusted in the prototype nodejs/node#46974. In my opinion, the performance gains are worth deviating from the spec for a browser-related property.This PR
main
towards #20167