From: Jim Tuccillo Newsgroups: comp.parallel.mpi Subject: Re: Performance of overlapped communication and computation Date: Fri, 08 Jan 1999 15:08:45 -0500 Organization: Posted via RemarQ, http://www.remarQ.com - Discussions start here! Message-Id: <369665CD.794B@radia1.com> References: <76tkr7$efc$1@lux.doc.ic.ac.uk> <76tn46$ni3$1@pegasus.csx.cam.ac.uk> <36962A8B.2781@radia1.com> <775guv$f81$1@lux.doc.ic.ac.uk> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit I measured the number of ticks at 3% greater with no communication. The communication is overlapped in the sense that the total time is what you would expect ( 40 seconds ) but some CPU cycles are "stolen" from the computational loop. 3% of 40 seconds is about 1.2 seconds which corresponds roughly to the 150 MB/s that the interconnect fabric will deliver. There is no free lunch here. Having the communication appear to be completely overlapped ( but less work gets done in the computational loop than if there was no communication ) is no different then having the communication appear to not be overlapped ( but the same amount of work gets done in the computational loops as when there is no communication ). IS this a correct summary of what you have seen so far ? Regards, Jim William Knottenbelt wrote: > > Hi Jim > > Jim Tuccillo (jjt@radia1.com) wrote: > : On the SP, the communication is completely overlapped. See below. > > [gloriously low wait times snipped] > > : [0] all done, tick = 17933592, time = 40.0098 > : [1] all done, tick = 17889158, time = 40.0091 > > Wow, now that's more like it! It'd be interested to see what the tick > count is when you run the program without any sending/receiving (so we > can get an idea of the CPU overhead associated with the background > sending/receiving). > > Cheers > -- > William Knottenbelt > Department of Computing > Imperial College > 180 Queens Gate > South Kensington > LONDON SW7 2BZ -- ========================================================================= Jim Tuccillo jjt@radia1.com IBM tuccillo@us.ibm.com 415 Loyd Road voice: 770.487.6694, fax: same, call first Peachtree City, GA 30269 http://www.radia1.com/jjt1 =========================================================================