Comment #5 on issue 2741 by [email protected]: max_opt_count is
implemented using the wrong kind of limit
http://code.google.com/p/v8/issues/detail?id=2741
Indeed, I don't understand what problem you're describing. I also don't
understand what alternative you're suggesting.
You seem to think that --max-opt-count is making the optimizer too
aggressive, but that's not the case, as it doesn't control when functions
are optimized. On the contrary, it controls when V8 gives up after too many
failed optimization attempts.
Getting optimization decisions right is hard. Optimizing too early is bad,
optimizing too late is bad too. V8 uses a variety of signals to help it
make good decisions in a wide range of situations. Of course sometimes it
gets it wrong, but on the whole the current implementation seems to be
about right. I don't see evidence of a concrete problem in your statements.
The pdfjs benchmark is a bit flaky; when I run it repeatedly
--max-opt-count=20 seems to be slightly worse than =10, and knowing what
effect the flag has on this particular benchmark that observation makes
perfect sense.
You state that whether a function is optimized or not should depend on "how
often it's even going to be executed" -- true, that would be ideal.
Unfortunately, I haven't yet found a way to accurately predict the future.
Have you?
--
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/groups/opt_out.
