| Newsgroups |
gmane.comp.mozilla.performance |
| Organization |
http://groups.google.com |
| Message-ID |
<[email protected]> |
On Nov 13, 10:33 pm, Mike Beltzner <[email protected]> wrote:
> Hey everyone,
>
> I was wondering what we should be calling out in terms of performance
> for the Firefox 3 Beta 1 release notes. I've seen some really good
> figures here about memory usage, but was wondering what we're
> comfortable saying about:
>
> - memory usage
> - rendering speed
> - human performance (protection of data, clarity of text rendering,
> etc)
>
> I don't think we should look at absolute numbers, but rather talk in
> percent change to previous releases. Any help or guidance you can
> offer would be appreciated.
>
> I know that the evangelism team is looking for this stuff as well, so
> if I'm duplicating effort, I'll expect someone will let me know :)
>
> cheers,
> mike
Yes, we are very hot to take that information and spread it to the
masses. One thing that might be interesting is to take the code that
Stuart did and run it against FF2 and see what kinds of results that
we get. In particular heap usage, separate from heap fragmentation is
an interesting metric since it will tell us about both how the code is
doing in terms of allocating memory and is also likely to reveal leak
information as well. My understanding is that we've made a lot of
progress in these areas.
I'm going to try to attend the leak + performance meetings this week
and will try to start collecting information and figuring out what
might useful. I don't want to distract too much from the actual work
at hand but these are measurements that we should probably be tracking
anyway.
How do we want to measure rendering speed? Page load times? I don't
think that we have any metrics that point out response times, sadly.
i.e. the amount of time spent away from the idle loop, the amount of
time it takes a XUL button to respond to a click. Those are worth
their own discussion for sure and is probably a thing we want to
measure down the road. Interaction performance, if you will.
Additionally I agree that we should be measuring against ourselves for
now. As Hyatt pointed out in Shaver's weblog, other engines to
different things with memory so I think that we need to be somewhat
careful when doing measurements against other engines. Maybe at some
point but right now we should focus on our own metrics and set our own
goals.
I don't know about human performance - but I love the idea. I'll have
to think and see if I can figure out anything to add to that.