I'm pretty sure that Navin said that we've changed the way we cache the folder DB (was that a different fix than the leak bug)?
Anyway, that is supposed to reveal that the 1st time loading a folder in a given session is a bit slower than the others (but I thought this was no longer the case?). I'm confused also, since I see numbers on runs like 10-26 where we had 1.46 (1.40 1.51 1.47) which really surprised me... This begs the question: Can we get either a PR_LOG setting, or even a -console dump() output of Folder Loading Speed (even if we only show the API?). This way, I might have some data points to prove/disprove the crazy things I sometimes see. Thanks - Stephen Scott Putterman wrote: > For the last 2 weeks, the 3 folder loading runs produce significantly > different results (this week is 1.32, 1.09, and .99) that are > increasingly better. Is this just timing differences or is there some > reason why the 3rd attempt is close to 30% better than the first? > Scrolling, Message Display, and Deleting have 3 times that are almost > identical. > > Scott > > Stephen Donner wrote: > >> Hi all - >> >> I've finished and committed to CVS the 11-9-2001 results for the >> Win32 (Windows 98) build. >> >> No changes from last week's numbers, at all. >> >> The other test machines (mac/linux) are still down (due to IS' >> re-architecture of DNS). >> >> http://www.mozilla.org/mailnews/win_performance_results.html >> >> - Stephen >> > >
