[Mapserver-dev] Performance loss between 3.5 and 4.2

Daniel Morissette dmorissette at dmsolutions.ca
Fri Sep 24 10:59:59 EDT 2004


Eric Boudaillier wrote:
> 
> I have done a simple test with shp2img, which is basically what I need,
> that shows the performance loss:
> 
> 762178 microseconds  home-made v4.2.3 statically linked
> 854505 microseconds  v3.6.7 binaries
> 848116 microseconds  v4.2 binaries
> 378156 microseconds  v3.5 msworkbench dll (compiled with MinGW)
> 
> The shapefile represents USA and is 18Mo.
> 
> Could someone explain these differences ?
> Could the compilation with MinGW give better result ?
> Or is 3.5 just faster ? If so, is there some inconvenient to use this
> version for my need ?

Lots of stuff has been added between 3.5 and 4.2 and there is always a 
chance that something could have lead to a performance loss, but we are 
not aware of any issue. Actually there was one that was reported back in 
the 3.6 days but it's been fixed in v3.6.1 and in this specific case 
v3.6 was made faster than 3.5:
http://mapserver.gis.umn.edu/bugs/show_bug.cgi?id=162

I would be tempted to suspect that some of the performance loss may be 
related to the build configuration. If you could compare two binaries 
built the same way and with about the same options (v3.5 vs 4.3) and 
there is still a significant performance difference then we would 
apreciate if you could submit a bug and attach a complete testcase to 
reproduce. Just like for bug 162, if the software got slower then we'd 
be interested to find out why and fix it.

> Then, where can I download a source package of 3.5 ?
> 

I was able to guess the following URL:
http://cvs.gis.umn.edu/dist/ms_3.5.tar.gz

You could also get the source from CVS using

cvs -d :pserver:cvs at cvs.gis.umn.edu:/data2/cvsroot checkout -r rel-3-5-0 
mapserver

Daniel
-- 
------------------------------------------------------------
  Daniel Morissette               dmorissette at dmsolutions.ca
  DM Solutions Group              http://www.dmsolutions.ca/
------------------------------------------------------------



More information about the mapserver-dev mailing list