svn commit: samba r15083 - in branches/SAMBA_3_0/source: locking smbd

[Jeremy Allison]

On Fri, Apr 14, 2006 at 01:08:51PM +1000, tridge at samba.org wrote:
> Jeremy,
> 
>  > Using talloc with destructors is nice and all, but in this
>  > case it's in a performace critical path and it *hurts* us.
>  > Go back to plain malloc/free with an explicit destructor
>  > call.
> 
> What sort of performance loss did you see, and what tool did you use
> to see it?

I'm used cachegrind to find small hotspots in the read/write
path. I have a micro-benchmark that shows current 3.0 (as of mondat)
as around 10% slower than 3.0.14a and I'm trying to track this
down. More as I work it out....

Jeremy.