Re: Transactionless reporter performance results

Greg Hudson wrote:
> Old code: 22.0 seconds
> New code: 7.4 seconds

Very nice!

> I'm not sure why there is a ~14-second variance in some of the trials.
> I'd say it has to do with BDB logfile creation, but the new code
> shouldn't result in any logfiles being created.

It could be delayed writes by the OS. You could try to create the
repository with the nosync option, and change the client test sequence to

rm -rf wc ; cp -r wcsave wc ; cd wc ; sync ; time svn up ; cd ..

and see if it helps.

/Tobias

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Sat Jan 10 09:51:59 2004