https://issues.apache.org/SpamAssassin/show_bug.cgi?id=5965
--- Comment #10 from Mark Martinec <[EMAIL PROTECTED]> 2008-08-28 08:58:07 PST
---
> agreed, it's a hotspot. I did a quick check of performance of this morning's
> SVN code, based on 10000 accesses to a cached header; the new code seemed
> slightly slower, about 8% (for the get() method alone). imo that's OK, and
> can
> be optimized later.
One must take into account the actual hit/miss cache ratio.
Instrumenting the pms->get with two counters and letting it run for
a while on our production mail server gives 45% hits/attempts ratio.
While the hit code path is a bit slower, the miss path is much faster
(original code evaluates $_[0]->{c}->{$_[1]} multiple times),
which in my benchmark yields an overall improvement. With a
simulated and quick _get the improvement is 30%, yet in practice
the speedup is negligible because the actual _get is very slow
compared to the get().
--
Configure bugmail:
https://issues.apache.org/SpamAssassin/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.
