Newsgroups: comp.parallel.pvm,comp.parallel
From: byrde@patpserv.epfl.ch (Olivier Byrde)
Subject: Re: [Q]:Memory Access and Efficiency of parallel (PVM) programs
Organization: Ecole Polytechnique Federale de Lausanne
Date: Tue, 31 Oct 1995 18:00:50 GMT
Message-ID: <475o8i$49b@info.epfl.ch>

In article <475gdg$spf@diable.upc.es>,
Antonio Scotti <scotti@centauro.upc.es> writes:
|> Hello folk
|> let's say one has a parallel program written according to the
|> master/slave computational model.
|> This program needs to read data, say one or more big matrixes,
|> stored on disk, which need to be passed to the slaves too. 
|> Which of the following is more efficient and why?:
|> 
|> 1) Having the master reading the matrix(es) and multicast it(them)
|>    to all the slaves.
|> 
|> 2)Having each slave to access the memory to retrieve the data it needs? 
|> 
|> 3) .... any suggestions?

Hello Antonio,

It really depends on the hardware you are considering: how fast your
interconnection network is (bandwith & latency), what kind of I/O and
disks you have, etc.

Also, the amount of data is significant: the size of your matrices and
how they are stored can lead to very different answers.

Suggestion: give us more details!!

Olivier


