1820133 - Speedometer test triggers more GC than expected

Status: NEW

(Core :: JavaScript: GC, defect, P3)

Description

[Jon Coppeard (:jonco)]

There are a lot of DOCSHELL GCs triggered in the speedometer benchmark, as seen in the following profile:

https://profiler.firefox.com/public/x253vp24nw63y9jdmb8dtwnv0b8er7hq3jgwcc8/calltree/?globalTrackOrder=a0w9&hiddenGlobalTracks=1w68&hiddenLocalTracksByPid=6308-0w4~10620-0~10156-0~2612-0~1880-0~12872-0~8760-0~4876-0w3~10072-0w2&thread=j&transforms=df-27&v=8

Fortunately these mostly fall outside the periods measured and so don't affect the score much.

They don't seem to be collecting much though. It might be better to relax whatever is triggering these.

Comment 1

[Olli Pettay [:smaug][[e-mail]]]

The behavior is quite different these days. Bug 1856574 changed how often we trigger major GC and bug 1856577 affected minor GCs.
I'd like to run major GC and CC more often, so that we could collect data sooner (but somehow that doesn't seem to work well atm).