"Jim C. Nasby" <decibel@decibel.org> writes:
> On Sun, Jan 30, 2005 at 08:50:49PM +1100, Neil Conway wrote:
>> Do you have any thoughts on how to improve it?
> See http://archives.postgresql.org/pgsql-general/2003-06/msg01072.php
> and http://archives.postgresql.org/pgsql-general/2003-06/msg01089.php
All of this falls down on twin problems: (1) we have no portable way to
find out how much RAM is really available, and (2) the planner has to
make a cost estimate for the sort in advance of actual runtime --- which
requires making some assumption about the value of sort_mem that will be
used.
regards, tom lane