Newsgroups: comp.parallel.pvm
From: nkramesh@asylum.cs.utah.edu (Nanda Ramesh)
Subject: Re: Shared memory comm. on HP machines, possible?
Organization: University of Utah
Date: 24 Jun 1994 16:18:58 GMT
Message-ID: <2uf11i$9s9@magus.cs.utah.edu>

In article <2uc4n3INN31l@duncan.cs.utk.edu>, hzhou@cs.utk.edu (Honbo Zhou) writes:
|> I have implemented a shared memory bewteen pvmd and local tasks (over uniprocessor
|> shared memory) based on PVM3.3b.
|> 
|> Unfortunately, I measured no performance gain. So it is never releasd. Many
|> people suggested the reason being expensive context switch even you don't
|> use the even more expensive semaphore...
|> 


I am a bit confused about the above statement. I thought you will not
be doing any context switching if you are using shared memory. 
Once you have mapped it into your address space you would access it
like private memory. I am assuming that you meant that you are not
using semaphores.

With sockets, as is the case now, you need minimum of 2 context switches
before the message gets across as the kernel gets involved.

Did I miss something? I would appeciate it if you could explain the 
reasoning of the above statement.

Thanks,
nanda

