https://issues.apache.org/SpamAssassin/show_bug.cgi?id=6444
--- Comment #15 from Mark Martinec <[email protected]> 2010-06-15 16:02:01 EDT --- (In reply to comment #7) > I did some benchmarking on a smallish database (started it few days ago, > thanks to enhancement in Bug 6447 it only accumulated 5000 tokens so far), Again my mistake (as in Bug 4400), that was 5000 messages, not tokens. Today (a few days later) our database grew to 12000 messages and 220.000 tokens, and I have analyzed our timing reports - while switching from the suggested tok_touch_all (my patch) back to the original tok_touch_all at noon. PostgreSQL server is 8.3.11 (from ports, running under FreeBSD). Tried with both indexing schemes (Bug 4400) - the indexing scheme makes hardly any difference. I can conclude that for our setup THE ORIGINAL tok_touch_all() works significantly faster! I will attach two diagrams to demonstrate it. For about half of the real-traffic messages the saving is about 30 ms. Compared to the total time spent in bayes (let alone compared to the total SA processing) this is not too drastic, but is quite noticeable nevertheless. I wonder, did a version of PostgreSQL cause this difference between our and Bradley's case? Which version was used there? -- Configure bugmail: https://issues.apache.org/SpamAssassin/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are the assignee for the bug.
