Tom, On 12/17/05 10:47 AM, "Tom Lane" <tgl@xxxxxxxxxxxxx> wrote: > BTW, some experimentation suggests that in fact a star join is already > slower than the "regular" plan in 8.1. You can force a star-join plan > to be generated like this: Cool! We've got Paal's test case in the queue to run, it's taking us some time to get to it, possibly by next week we should be able to run some of these cases: 1) 8.1.1 btree with bitmap scan 2) 8.1.1 on-disk bitmap with direct AND operations 3) (2) with forced star transformation (materialize) We'll also be trying the same things with the CVS tip of Bizgres MPP, probably over X-mas. We should be able to handily beat Oracle's 3 second number. - Luke
