Is this something that can be dealt with? Would it make it into the beta, or
the release candidate for 1.0?
> On Wed, 2003-10-15 at 09:17, Julian Foad wrote:
> > The cost of opening N similar temporary files with
> > svn_io_open_unique_file is proportional to N squared, and
> with a high
> > constant of proportionality. Its cost scales linearly with
> the number
> > of temp files of the same pattern that exist already, so the total
> > cost of opening N such temp files is quadratic. If the application
> > creates say 200 temp files of the same pattern, it will perform
> > (200+199+...+1) = 20100 iterations of its loop which contains:
>
> That might explain why imports seem to get slower and slower
> as they go
> on. (And why memory usage was so high back when we were opening those
> temp files using a long-lived pool.) I'd always figured it was an FS
> problem.
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Wed Oct 15 19:05:39 2003