1736338 - Barrier overhead from use of Heap<JSObject*> in ArrayBufferOrView

Status: RESOLVED FIXED

(Core :: JavaScript Engine, defect)

Description

[Jan de Mooij [:jandem]]

Attachment: Micro-benchmark

I noticed this in a profile last week but forgot where. Fortunately I can still reproduce this with a micro-benchmark. For the attached one I have this profile:

https://share.firefox.dev/3parIgP

We spend a lot of time in JS::HeapObjectPostWriteBarrier. I wonder if this is made worse because ArrayBufferView is passed by value and we end up copying it a lot.

Comment 1

[Jan de Mooij [:jandem]]

The micro-benchmark took about 1400 ms on my Mac before these changes, now it takes about 2000 ms, so it's a pretty severe regression for DOM bindings that use typed arrays.

Comment 2

[BugBot [:suhaib / :marco/ :calixte]]

The Bugbug bot thinks this bug is a defect, but please change it back in case of error.

Comment 3

[BugBot [:suhaib / :marco/ :calixte]]

Set release status flags based on info from the regressing bug 1720422

Comment 4

[Jon Coppeard (:jonco)]

We talked about this at the GC meeting last week. I think we came to the conclusion that the Heap<> wrapper is not necessary at the moment as these are currently all roots, and if they are used on the heap we can arrange it so you can Heap<ArrayBufferOrView>.

Comment 5

[Jan de Mooij [:jandem]]

Attachment: Bug 1736338 - Stop using Heap<> in ArrayBufferOrView. r?sfink!

This was triggering post-barriers. This is unnecessary because this type is currently
only used on the stack.

Comment 6

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/cb32ebc906a3
Stop using Heap<> in ArrayBufferOrView. r=sfink

Comment 7

[Iulian Moraru]

https://hg.mozilla.org/mozilla-central/rev/cb32ebc906a3

Comment 8

[Ryan VanderMeulen [:RyanVM]]

Given where we are in the cycle with no betas left for Fx94, let's let this ride the trains rather than trying to uplift directly into RC builds.