Commit-Queue | +1 |
Inspect html for hidden footers to help with email filtering. To unsubscribe visit settings. |
const base::TimeDelta idle_time = base::TimeDelta::FromMillisecondsD(16);
Who guarantees that we actually have 16ms idle time?
I understand that idle tasks are not reliable but we also shouldn't create jank by just assuming that there's idle time in general.
Maybe the change around time (see below) is enough to make it trigger more reliably?
const bool time_to_run_exceeded = time_to_run > kMaxTimeToRun;
The mechanism is conservative and doesn't want this task to be run too far away from where the context disposal. We want the GC because the assumption is that lots of memory is dead, this only holds for the first GC after the disposal and not for any random time.
Given that the task is low priority, it's probably enough to just snapshot the GC count and make sure that no GC happened in between. This would allow the task to happen much later.
Inspect html for hidden footers to help with email filtering. To unsubscribe visit settings. |
const base::TimeDelta idle_time = base::TimeDelta::FromMillisecondsD(16);
Thinking a bit more about this: The scheduler should know how much time we have left in the frame? Can we make this queryable or pass it along somehow?
Inspect html for hidden footers to help with email filtering. To unsubscribe visit settings. |
📍 Job android-pixel9-pro-perf/speedometer3 complete.
See results at: https://pinpoint-dot-chromeperf.appspot.com/job/12e19354d10000
Inspect html for hidden footers to help with email filtering. To unsubscribe visit settings. |