Path: utzoo!attcan!utgpu!jarvis.csri.toronto.edu!rutgers!apple!bloom-beacon!tut.cis.ohio-state.edu!ukma!gatech!hubcap!Eugene
From: eugene@eos.arc.nasa.gov (Eugene Miya)
Newsgroups: comp.parallel
Subject: Too much concern for efficiency
Message-ID: <5507@hubcap.clemson.edu>
Date: 17 May 89 15:18:44 GMT
Sender: fpst@hubcap.clemson.edu
Lines: 34
Approved: parallel@hubcap.clemson.edu

Parallelism considered harmful...

I am not exceedingly pro-Linda, and I read most postings with
interest, and I hope that one day we will have something...what
I don't know.

But, I think there are times we are too concern with efficiency.
True, it is one of the major reasons for doing things in parallel.
But I think we are sometimes blindsighted to do things faster and more
efficiently.

I have rewritten several codes to use parallelism and found
numerous little "gotcha's."  They ran, they just didn't run quite
the same way.  When I attend conferences of people
presenting papers on parallelism, I am surprised at the numbers
of people who never comapared anything more than the runs times
(just about always faster right?).

Quoting from Kerighan and Plauger:
Make it right before you make it faster.
Keep it right when you make it faster.
Make it clear before you make it faster.
Lastly, don't sacrifice clarity for small gains in "efficiency."
K&P say a lot about the problems of pre-mature optimization.
Let's learn from the software mistakes of the past.

Michael Scott asks:
>How long do we have to wait?
How long can you wait?  The problem isn't as simple as overnight
otherwise it would have been done by now.
Some research fruits take 20 years to find applications.
We computer people tend to be a bit more impatient. 8)

--eugene miya