[mapserver-dev] mapserver performance windows/linux

[Stephen Woodbridge]

Has anyone worked with valgrind profiling or callgrind, these tools 
typically do not require that you recompile using the -pg option and may 
return better profiling results. I use valgrind all the time for memory 
checking, but have not use the profiling tools.

-Steve W

Andy Colson wrote:
> I dont think this is a useful profile.  The program ran for 2.6 seconds, 
> and this profile says msClipPolylineRect took up 100% of the time, which 
> it measured 758 calls with a total time of 0.01 seconds.
> 
> So the profiler was able to account for 0.01 of 2.6 seconds.  I'm not 
> sure what that means, but the numbers don't add up.
> 
> -Andy
> 
> 
> JUDD.CHRISTOPHER wrote:
>> Andy,
>>
>> That looks pretty close to the numbers that I am getting. 
>> Steve, will that suffice for the profiling?  I guess it would be nice to
>> look at the differences between ms4w 2.3.1 and 3 beta 7.
>>
>> I'd like to do whatever I can to help identify this as a bug and try to
>> get it fixed.
>>
>> Thanks,
>> Chris
>>
>> -----Original Message-----
>> From: Andy Colson [mailto:andy at squeakycode.net] Sent: Thursday, 
>> September 17, 2009 3:27 PM
>> To: JUDD.CHRISTOPHER
>> Cc: Alan Boudreault; mapserver-dev at lists.osgeo.org
>> Subject: Re: [mapserver-dev] mapserver performance windows/linux
>>
>> JUDD.CHRISTOPHER wrote:
>>> That is right, I did find this same performance issue with 5.0, 5.2
>> and
>>> 5.4 on linux.
>>>
>>> I just tested ms4w_3.0_beta7 and it is taking about 3.7 seconds to
>> draw
>>> the same image that 2.3.1 did in 1.4 seconds.
>>>
>>> Someone else should probably do the same testing to make sure that I'm
>>> not off base. 
>>> How do I do the profiling?
>>> Thanks for your help,
>>> Chris
>>>
>>
>> On Linux its easy.  recompile with -pg, then run it once which will dump
>>
>> out a gmon.out, then use the gprof to view it.
>>
>> No idea on windows.
>>
>> I used shp2img because I dont have php script compiled.  I downloaded 
>> your test set and changed the mapfile extent to match what you had in 
>> the php script.
>>
>> shp2img takes 2.6 seconds to render, and my images came out a lot like 
>> yours.  (I'm also using mapserver 5.6)
>>
>> I tried it with both gd and agg and it didnt seem to make a 
>> difference.   But changing to ANGLE AUTO made it run in 0.8 seconds.
>>
>> Here are my gprof dumps.
>>
>> AGG, ANGLE FOLLOW:
>>
>> Flat profile:
>>
>> Each sample counts as 0.01 seconds.
>>    %   cumulative   self              self     total
>>   time   seconds   seconds    calls  ms/call  ms/call  name
>> 100.00      0.01     0.01      758     0.01     0.01  msClipPolylineRect
>>    0.00      0.01     0.00    82305     0.00     0.00  msRectOverlap
>>    0.00      0.01     0.00    67590     0.00     0.00 
>> intersectLabelPolygons
>>    0.00      0.01     0.00    36963     0.00     0.00
>> msIntersectSegments
>>    0.00      0.01     0.00    32460     0.00     0.00  msGetBit
>>    0.00      0.01     0.00    16988     0.00     0.00  msSHXReadSize
>>
>>
>>
>>
>>
>> AGG, ANGLE auto:
>>
>> Flat profile:
>>
>> Each sample counts as 0.01 seconds.
>>   no time accumulated
>>
>>    %   cumulative   self              self     total
>>   time   seconds   seconds    calls  Ts/call  Ts/call  name
>>    0.00      0.00     0.00    98521     0.00     0.00  msRectOverlap
>>    0.00      0.00     0.00    83806     0.00     0.00 
>> intersectLabelPolygons
>>    0.00      0.00     0.00    32460     0.00     0.00  msGetBit
>>    0.00      0.00     0.00    16988     0.00     0.00  msSHXReadSize
>>    0.00      0.00     0.00    15472     0.00     0.00  msSHXReadOffset
>>    0.00      0.00     0.00    14715     0.00     0.00  msSHPReadBounds
>>    0.00      0.00     0.00     4067     0.00     0.00  msFree
>>
>>
> 
> _______________________________________________
> mapserver-dev mailing list
> mapserver-dev at lists.osgeo.org
> http://lists.osgeo.org/mailman/listinfo/mapserver-dev