[postgis-devel] Prepared Geometry API

Paul Ramsey pramsey at cleverelephant.ca
Wed Oct 8 11:13:20 PDT 2008


You can test for yourself right now on 1.3-svn, just test

ST_Contains(*POLYGON, POINT)

against

ST_Contains(*POLYGON, POINT, ID)

The former is Mark Leslie's pure-PostGIS cached implementation, using
a memcmp to determine cache hits, the latter is the ID-based prepared
geometry implementation.

My tests showed no difference. If the memcmp overhead is 30% I'd
expect to see it show up, even over all the other noise (different
algorithms, GEOS vs pure PostGIS, etc).

P.

On Wed, Oct 8, 2008 at 11:03 AM, Chris Hodgson <chodgson at refractions.net> wrote:
> I recall Ben's original testing showing it to be a 30-40% loss of the
> increased performance gained by doing the caching. I'm eagerly awaiting your
> implementation and test results though, hopefully there was something wrong
> with those previous tests.



More information about the postgis-devel mailing list