Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!sdd.hp.com!zaphod.mps.ohio-state.edu!sol.ctr.columbia.edu!emory!hubcap!eugene
From: eugene@wilbur.nas.nasa.gov (Eugene N. Miya)
Newsgroups: comp.parallel
Subject: Re: fine->medium->coarse grain
Message-ID: <9210@hubcap.clemson.edu>
Date: 5 Jun 90 12:38:30 GMT
Sender: fpst@hubcap.clemson.edu
Lines: 21
Approved: parallel@hubcap.clemson.edu

There is no accepted definition of grain.

It's a loose (as opposed to a tightly)-coupled concept (yet another
vague concept).
A feeling, like parallelism.  It's yet another elephant.  I've seen Bell's
papers and was at the talk he gave at the History of Workstations conference
where he gave this paper in Palo Alto.  Grain existed way before this
like Rumbaugh's Coarse Grain Dataflow IEEE TOC paper.  Robbie Babb's
LGDF (Large Grained Data Flow), etc.  From my programming language [PL]
background, we would say that a lot of "scope" is used in large grain.
I do not think synchronization is the only issue.  Dataflow is usually
regarded as fine grain and procedure "level" work is coarse grain.
With dataflow you have the concept of "Throttling" execution (See
Manchester and SISAL papers).  It gets into scheduling and OS rather
than traditional PL topics.  This in turn gets into data consistency
and other issues, and we swim around in circles.....

Just another blindman.

--e. nobuo miya, NASA Ames Research Center, eugene@orville.nas.nasa.gov
  {uunet,mailrus,other gateways}!ames!eugene