Comment #8 on issue 4335 by [email protected]: large google maps
performance regression in chrome canary (version 46)
https://code.google.com/p/v8/issues/detail?id=4335
What's the status on this bug? Is it likely to be fixed before this
regression hits release channel?
In order to make sure I found all the places in google maps where we hit
this, I added some logging in ClearAllICsByKind which prints out callstacks
and cumulative time spent. For maps, the vast majority of the time was from
the one place I already mentioned (using an object to find the unique
elements from a group of 32-bit ids).
It found another interesting case as well - there's some formatting code
for constructing human-readable numbers that uses an object literal like
this:
var lookup = { ..., 1E7:{t:"00 million"},1E8:{t:"000 million"},1E9:{t:"0
billion"} ... }
this code seems almost designed to seek out this "poison range" (which it
finds in 1E9).
Since I had this logging on I went cruising around some other pages to see
who else will be affected by this.
The worst one I found was cnn.com - it spends 1.3s (SECONDS) of startup
time in ClearAllICsByKind. Looks like it has some (string!) article ids
which are used as properties in objects. And because v8 is clever, those
ids are treated as integers and stored as elements, which triggers this
issue.
--
You received this message because this project is configured to send all
issue notifications to this address.
You may adjust your notification preferences at:
https://code.google.com/hosting/settings
--
--
v8-dev mailing list
[email protected]
http://groups.google.com/group/v8-dev
---
You received this message because you are subscribed to the Google Groups "v8-dev" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
For more options, visit https://groups.google.com/d/optout.
