[Benchmarking] custom vs. stock builds of MapServer

[Frank Warmerdam]

Daniel Morissette wrote:
> Frank Warmerdam wrote:
>>
>> Anyways, the point is that we do not seem to be enjoying any noticable
>> boost due to the way mapserver was built for the benchmarking test,
>> and people should be able to achieve similar results with the stock
>> FGS MapServer builds for intel linux.
>>
> 
> Frank,
> 
> Am I right in interpreting your numbers to mean that FGS is actually 
> about 10% faster than a native build? I would have expected FGS to be a 
> bit slower, not faster than a native build... odd.

Daniel,

I read the results as 'not significantly different'.  There was a modest
difference in the numbers (I think a bit less than 10%) but that could be
attributed to poor "hot start" values for the original test or other
statistically insignificant things.

> Or did we lose some performance between 5.4.2 and the latest 5.6 beta?

It has not been particularly apparent in other tests between 5.4 fastcgi
and 5.6 fastcgi (the only clear version to version tests) that there was
a meaningful difference.  I vaguely suspect the difference in speed might
relate to the benchmark binaries including extra libraries (ie. oracle,
sde, mrsid, etc) that aren't included in the fgs binaries.

A more careful study would be necessary to conclude there has been a
meaningful change between versions.

Best regards,
-- 
---------------------------------------+--------------------------------------
I set the clouds in motion - turn up   | Frank Warmerdam, warmerdam at pobox.com
light and sound - activate the windows | http://pobox.com/~warmerdam
and watch the world go round - Rush    | Geospatial Programmer for Rent