http://issues.apache.org/SpamAssassin/show_bug.cgi?id=3109





------- Additional Comments From [EMAIL PROTECTED]  2006-05-29 19:36 -------
Created an attachment (id=3530)
 --> (http://issues.apache.org/SpamAssassin/attachment.cgi?id=3530&action=view)
shortcircuit.cf as used in jm's test

I got a chance to do a follow-up on the performance of this code.   I redid my
rules files to not include any network rules at all, which require the "stop
everything and wait for completion" code to kick in. Instead, just using the
"shuffle priorities to ensure all local rules complete in the right order"
algorithm, and just very simple 2 shortcircuit rules, here's what I get:

sc=spam: 161 messages, total time 108.4, avg 0.673291925465839
sc=no: 1365 messages, total time 4170.9, avg 3.0556043956044
sc=ham: 129 messages, total time 485.6, avg 3.76434108527132
all: 1655 messages, total time 4764.9, avg 2.87909365558913

reminder: here's what I had beforehand using svn trunk:

all: 1980 messages, total time 9672.09999999999, avg 4.88489898989899

so, on average, it now completes scans in *on average* across all messages a
whopping 58% of the runtime.   That is a MAJOR speedup!

Most of this is coming from the spams taking only 0.673 seconds to scan on
average.

This is probably the largest speedup in the open source SpamAssassin codebase
in years. Nice ;)




------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.

Reply via email to