This is mostly a toy benchmark but I would like to hear suggestions on how results can be imroved. Eugene
If you really only want to compare the speed of the bayes engines, remove or zero-out the scores of all the non-bayes rules from SA.
If you've got a test-box for it, you can accomplish this by moving all the stuff out of /usr/share/spamassassin, and only moving back in the bayes stuff.
And then configure dspam to only use one of it's bayes engines.. You could even generate a table with all the combinations. SA (full), SA (bayes only), dspam (bayes), dspam (altbayes), dspam (bayes+altbayes)
