From: "Patrick LEGRAND" Newsgroups: comp.parallel.pvm References: <36140616.322A9323@scs.carleton.ca> Subject: Re: PVM Communication Time Measurement Message-Id: <6k_Q1.29$cf2.101351@newsfeed.slurp.net> Date: Fri, 2 Oct 1998 08:23:04 +0200 Hi, Try xpvm Patrick kenji imasaki a écrit dans le message <36140616.322A9323@scs.carleton.ca>... >Hello, > >I am trying to measure communication time using PVM and C++. I would >like to know if there are any tools or supporting software to measure >communication time? > >For example, I would like to get information as follows when a program >(master-slave type application) finishes its execution : > > 10 sec for whole execution, > 1 sec are spent on communication from master to slave1 > 1 sec are spent on communication from slave1 to master > 2(1+1) sec are spent on commutation between master and slave1 > 2 sec are spent on communication from master to slave2 > 2 sec are spent on communication from slave2 to master > 4(2+2) sec are spent on commutation between master and slave2 ) > 6 sec are spent on communication among master and slaves > >I prefer the overhead of this measurement is as small as possible and >modification to original programs is as small as possible as well. >These measurement should be stored in profile and can be analyzed by >any statistical tools. I don't think XPVM has this feature. > >I would appreciate any comments/suggestions. > >Thank you very much. >-- > /\_/\ > ( O O ) >/=====================---oOO--(_)--OOo---========================\ >| kenji@scs.carleton.ca http://www.scs.carleton.ca/~kenji | >\================================================================/ >We are born naked, wet and hungry. Then things get worse. > > >