735894 - Fire the image discard timeout more often

Status: RESOLVED FIXED

(Core :: ImageLib, defect)

Description

[Justin Lebar (not reading bugmail)]

The main motivation for this is areweslimyet.com.  (We're still working out a few bugs before we un-protect it, but I'm happy to give out the password privately.)

AWSY waits for 30s before taking some measurements.  But images are currently discarded 20-40s after they're unlocked.  (This is true even with the rewrite of discarding in bug 732820.)

It's kind of lame to make a change primarily for a benchmark, but I don't see much possible harm in firing the discard timer once every 10s instead of once every 20s.  We'd leave 20s as the discard timeout; we'd just check for timed-out images twice as frequently, so images would be discarded 20-30s after they're unlocked.

Comment 1

[Justin Lebar (not reading bugmail)]

Attachment: Patch v1

Comment 2

[Joe Drew (not getting mail)]

This isn't free, of course: we'll use more battery.

Comment 3

[Justin Lebar (not reading bugmail)]

(In reply to Joe Drew (:JOEDREW!) from comment #2)
> This isn't free, of course: we'll use more battery.

I imagine that 1000 extra events use less energy than decoding one large JPEG image -- that is to say, we'd have to fire a lot of extra events before this matters.  But perhaps we can think of a pathological case where we'd fire a lot of extra events, or perhaps an event is more expensive than I imagine...

I'm OK trying to shim a fix into the config file used by AWSY, rather than changing this pref here, if you're concerned.

Comment 4

[Joe Drew (not getting mail)]

Well, of course if we actually re-use those decoded images rather than redecoding them, we'd save that decoding energy too. :)

I'm not that conerned about this value. It was not arrived at scientifically. We should just go in to this with eyes open.

Comment 5

[Joe Drew (not getting mail)]

Comment on attachment 606000 [details] [diff] [review]
Patch v1

Review of attachment 606000 [details] [diff] [review]:
-----------------------------------------------------------------

I had initially misread this bug as changing the timeout. However, this just makes us a little more accurate towards our timeout deadlines, which I am fine with.

Comment 6

[Justin Lebar (not reading bugmail)]

https://hg.mozilla.org/integration/mozilla-inbound/rev/803853bf2a55

Comment 7

[Marco Bonardo [:mak]]

https://hg.mozilla.org/mozilla-central/rev/803853bf2a55