run time variability

*To*: mathgroup at smc.vnet.net*Subject*: [mg65264] run time variability*From*: John Dickmann <jqd at mit.edu>*Date*: Thu, 23 Mar 2006 06:58:22 -0500 (EST)*References*: <200603170513.AAA09771@smc.vnet.net> <4EF26559-3634-461D-9E5E-6522C06A6751@mimuw.edu.pl> <200603181140.GAA09775@smc.vnet.net> <200603190819.DAA29881@smc.vnet.net>*Sender*: owner-wri-mathgroup at wolfram.com

I have a question about run time variablity in Mathematica. I am running the same program with different inputs, counting the computational time for each input. I get highly variable run times with different inputs and am wondering if there is something in the computing envirionment that is the cause and not just the complexity of the input. I am counting paths through networks of various sizes, consisting of between 35-70 nodes and 130-300 edges (70, 300 the largest; 35, 130 the smallest). run times vary from 10 minutes to 28 hours, with 27 and 36 minutes for the two data points in the middle. there does not seem to be any pattern timing and results are below: edges nodes time(sec) #paths sec/path 158 49 627 493,180 0.0013 131 36 1662 2,578,174 0.0006 167 48 2212 3,601,321 0.0006 304 73 101106 2,685,255 0.0377 the code is inefficient and my problem computationally intensive, but my key concern here is the wide variablity in performance. Could it be the specific topologies of the cases that causes the variablity? thanks, john

**References**:**A Reap Sow question***From:*János <janos.lobb@yale.edu>

**Re: A Reap Sow question***From:*János <janos.lobb@yale.edu>

**Re: Re: A Reap Sow question***From:*Andrzej Kozlowski <akoz@mimuw.edu.pl>