[svn.haxx.se] · SVN Dev · SVN Users · SVN Org · TSVN Dev · TSVN Users · Subclipse Dev · Subclipse Users · this month's index

Re: Performance benchmarks-1.5 results show significant regression in 1.6

From: C. Michael Pilato <cmpilato_at_collab.net>
Date: Thu, 31 Mar 2011 11:13:27 -0400

On 03/31/2011 11:04 AM, Mark Phippard wrote:
> A user supplied me with results of running the benchmarks with 1.5,
> 1.6 and 1.7 on Windows. I have updated the wiki with those results.
>
> There was one area of the tests that I thought were interesting,
> because there is a massive regression in 1.6. The tests are with a
> working copy that has folders with a couple thousand icons in it. I
> mainly wanted to look for problems with a lot of files in one folder
> as people on users@ had been complaining. You can see those results
> here:
>
> https://ctf.open.collab.net/sf/go/wiki2115
>
> Here are the average checkout/switch times:
>
> Checkout Switch
> 1.5: 0:30 0:18
> 1.6: 7:40 7:15
> 1.7 2:14 1:46
>
> So while 1.7 has made nice improvements over 1.6, there is still a
> huge regression. I would blame it on tree conflicts, but would this
> show up in a checkout (maybe because it shares code with update)?
> Might be something for someone to look at.

Well, I just made switch slower yesterday, and knowingly so. At least, I
made it do some extra up-front work to verify that the switch isn't
potentially unwanted. My changes won't affect the time spent doing the
actual work of the switch (and you can pass --ignore-ancestry if you know
what you're doing and skip my extra logic altogether).

-- 
C. Michael Pilato <cmpilato_at_collab.net>
CollabNet   <>   www.collab.net   <>   Distributed Development On Demand

Received on 2011-03-31 17:14:02 CEST

This is an archived mail posted to the Subversion Dev mailing list.