(B>Actually, the results are completely expected once you know what's
(B>exactly is going on. I found it weird that my sorts were also slowing
(B>down with more sort memory until Tom or Bruce or someone pointed out to
(B>me that my stats said my sorts were swapping.
(Bthis way my first expectation but since the machine was newly booted and
(Bhad 1/2 gig of ram (nothing running but PostgreSQL) I couldn't believe
(Bin that theory ...
(BMaybe but I couldn't verify that ...
(BOf course swapping is worse than anything else.
(B>This strikes me, too, as another area where mmap might allow the system
(B>to do a better job with less tuning. Basically, the sort is getting
(B>split into a bunch of smaller chunks, each of which is individually
(B>sorted, and then you merge at the end, right? So if all those individual
(B>chunks were mmaped, the system could deal with paging them out if and
(B>when necessary, and for the sorts you do before the merge, you could
(B>mlock() the area that you're currently sorting to make sure that it
(BAs far as I have seen in the source code they use Knuth's tape
(Balgorithm. It is based on dividing, sorting, and merging together.
(B>If the VM system accepts hints, you might also get some further
(B>optimizations because you can tell it (using madvise()) when you're
(B>doing random versus sequential access on a chunk of memory.
(Bit is an interesting topic. the result of the benchmark is very clock
(Bspeed depedent (at least in case my of my data structure).
(BCybertec Geschwinde &. Schoenig
(BLudo-Hartmannplatz 1/14; A-1160 Wien
(BTel.: +43/1/913 68 09 oder +43/664/233 90 75
(BURL: www.postgresql.at, www.cybertec.at, www.python.co.at, www.openldap.at
(B---------------------------(end of broadcast)---------------------------
(BTIP 3: if posting/reading through Usenet, please send an appropriate
(Bsubscribe-nomail command to [EMAIL PROTECTED] so that your
(Bmessage can get through to the mailing list cleanly