home
***
CD-ROM
|
disk
|
FTP
|
other
***
search
/
ftp.cs.arizona.edu
/
ftp.cs.arizona.edu.tar
/
ftp.cs.arizona.edu
/
icon
/
newsgrp
/
group94b.txt
/
000101_icon-group-sender _Thu Nov 10 10:40:37 1994.msg
< prev
next >
Wrap
Internet Message Format
|
1995-02-09
|
3KB
Received: by cheltenham.cs.arizona.edu; Thu, 10 Nov 1994 13:03:45 MST
Date: Thu, 10 Nov 1994 10:40:37 +0600
From: jeffery@runner.jpl.utsa.edu (Clinton L. Jeffery)
Message-Id: <9411101640.AA07036@runner.utsa.edu>
To: Ryan.Walcott@columbiasc.ncr.com
Cc: icon-group@cs.arizona.edu
In-Reply-To: <Cz0AFI.67x@ncrcae.ColumbiaSC.NCR.COM> (message from Ryan Walcott on Wed, 9 Nov 1994 15:16:30 GMT)
Subject: Re: Optimizing Icon on a UNIX Platform
Content-Length: 2013
Errors-To: icon-group-errors@cs.arizona.edu
> We use ICON applications at AT&T Global Information Solutions to create
> statistical reports on data generated by benchmark runs. Currently these
> reports take several hours to run.
> Is there a way to optimize the applications for our platform so that the
> reports will execute faster? We are considering porting the applications
> to C and recompiling, but would like to avoid this if possible because
> they are very long and complicated.
There is no quick and easy solution to this, beyond studying the programs'
behavior, and using the Icon compiler for performance-critical applications
if it is feasible to do so. Hiring an Icon guru to look at your code might
be another possibility...
The "real" solution to this is for some of us to invest further in the
language. What we have so far is free, thanks to the investment by the
National Science Foundation and countless donated hours by the Griswolds
and various volunteers. Here are two further investments, one that we are
pursuing, and one that we haven't the resources to pursue at present:
(1) Invest in profiling and debugging tools. Taking the research on
execution monitoring and visualization that's already been done and
putting it in the hands of users is a lot of work, but we are in the
process of doing this. The core supporting code is present in the Icon
9.0 sources, but I think it will take awhile to do useful end-user
documentation and produce a polished Icon Program Library section on
profiling, debugging, and visualization. Some interesting profiling and
debugging tools have been produced by Icon users, which is another way
to invest.
(2) Invest further in the Icon compiler. With adequate resources, it could
be maintained and further optimized. Whether this happens appears like it
will depend on whether the industry folks and users choose to make it so.
You could do it yourself and contribute it to the community, or you could
commit resources to Icon Project for that express purpose.