Base Conversion Report from Vern Paxson

[Vern Paxson]

David Hough wrote:

> Vern did a base conversion test program as a class project for Kahan this
> spring ...

He is being too modest.  I would like to clarify that David wrote the bulk
of the program.  I added an implementation of the algorithm for finding
"worst-case" inputs that was developed by Tim Peters and appeared on this
list a few months back.  I then ran the program on a number of different
machines and summarize in the paper results for tests of single and double
precision conversion.

I've made the program and the paper available via anonymous ftp to
ftp.ee.lbl.gov (128.3.112.20).  Retrieve testbase.tar.Z (140KB) and
testbase-report.ps.Z (117KB), using binary mode.

		Vern