[Chandler-dev] Performance of first time view switch vs
D John Anderson
john at osafoundation.org
Tue Feb 13 12:07:59 PST 2007
Actually it's a little more complicated than this.
There are 3 situations:
1) we need to copy a new tree of blocks and need to attach new contents
2) we can reuse a cached copy but need to attach new contents
3) we can reuse a copy and don't need to attach new contents
1 is the slowest, 2 is faster and 3 is the fastest.
I changed the mix of situations to minimize the number of case 1,
which increases case 2.
The problem with the test is that it doesn't properly measure the
speed of each case separately, nor the mix of cases that actually
occur in practice.
On Feb 13, 2007, at 11:48 AM, Heikki Toivonen wrote:
> We had a major performance regression in the first time view
> but at the same time improving subsequent view switching time. The
> time view switching performance went from 1.01s to 3.06s, i.e. it got
> about 3 times slower.
> The same checkin also caused table scrolling to slow down on the
> test machine from 0.48s to 4.83s, i.e. about 9 times slower.
> John argues to we should not be measuring the first time switch
> We have continued to measure the first time performance for several
> * historical data
> * We don't want the first time to be too slow either.
> * It seems we have usage patterns were first time is actually not that
> uncommon. Many people have half a dozen or more collections that they
> overlaying with different calendars, and it seems every new
> is a "first time" performance. With that kind of usage you can
> expect to
> hit the "first time" case a couple of times a day for the first couple
> of weeks I'd guess.
> * There is also the case where people temporarily check different
> calendars, in which case they hit this.
> This is bug https://bugzilla.osafoundation.org/show_bug.cgi?id=7970
> So, what are we going to do?
> * Status quo
> * Switch to measuring second time performance
> * Measure both
> If we make any changes to current practice, then other tests would
> to change too.
> Heikki Toivonen
> _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
> Open Source Applications Foundation "chandler-dev" mailing list
More information about the chandler-dev