Sven Willenberger wrote:
> OK, that I do see; I guess I never noticed it on the other (i386)
> machine as the memory never exceeded the max amount allowed by the
> tunables. That raises a question though:
>
> Using identical data and identical queries, why would the amd64 system
> using postgresql 8.1.3 be using some 2/3s more memory to store the
> query results before output than the i386 system using postgresql
> 8.0.4?
>
> amd64 system:
> PID USERNAME THR PRI NICE SIZE RES STATE C TIME WCPU CMD
> 8193 svenw 1 5 0 1516M 1475M ttyin 1 0:15 0.00% psql
>
> i386 system:
> PID USERNAME THR PRI NICE SIZE RES STATE C TIME WCPU CMD
> 61263 svenw 1 5 0 972M 956M ttyin 2 0:27 0.00% psql
>
> These are the final stages after the file has flushed (this was run on
> the prior month's data which had fewer results). This would either
> point to an issue with a) amd64 memory allocation or b) palloc
> interacting with 64bit memory or c) some other change.
My suspicion is that it has to do with pointers being 8 bytes vs 4 bytes
in amd64.
--
Larry Rosenman
Database Support Engineer
PERVASIVE SOFTWARE. INC.
12365B RIATA TRACE PKWY
3015
AUSTIN TX 78727-6531
Tel: 512.231.6173
Fax: 512.231.6597
Email: Larry.Rosenman@pervasive.com
Web: www.pervasive.com