Fellow Paranoid,

This disk is a collection of programs intended to test the arithmetic
of your computer.  The original version in BASIC was written by
Professor W. Kahan at the University of California at Berkeley.
The FORTRAN version was created by students at Berkeley under Prof.
Kahan's direction.  The Pascal version is due to Dr. Brian Wichmann
at the National Physical Laboratory in England.  The C version was
begun by Thos Sumner at the University of California at San Francisco
and completed by Dr. David Gay of the Bell Telephone Laboratories.

Each version is intended to give the same results under any given
circumstances.  In each case, the purpose is to discover and reveal
any defects in the operation of the computer arithmetic.  These tests
are not expected to be comprehensive, but rather indicative.  That is,
while many flaws will be found, if present, no guarantee can be made
about detecting all flaws actually present in the arithmetic.

This is the second general release of Paranoia software.  The changes
from the earlier version are given in the CHANGES file on this disk.
Dr. David Gay is owed a debt of gratitude for his extensive efforts
in completing the C version and revising the Pascal and FORTRAN ones.
Many others have given freely of their time and effort as well, but
now it is your turn.  Use the tests.  If you have reason to suspect
that your results may include some computer or compiler version not
previously reported, please return yours in machine readable form.

Send results by electronic mail to:
   Telemail:   RKarpinski
   UUCP:       ucsfcgl!cca.uscf!dick@Berkeley
   Compuserve: 70215,1277
   BITNET:     dick@ucsfcca

Send results on IBM PC, Macintosh, Sage/Stride, or 8 inch DEC format
diskettes to:
   Richard Karpinski
   6521 Raymond Street
   Oakland, CA 94609

Call me as needed at (415) 658-3797 (answering machine) or at work
at UCSF (415) 476-4529 (from 12-7 pm, usually).  

This disk is available in 360k IBM PC DOS format for $15 from the
above address.  A second disk is also available, which contains the
double precision version of the FORTRAN program as well as the
results from running several versions on several machines.  Write
me at any time for the current list of results available.  Enclose
a self-addressed stamped envelope for the free list.  The second
and subsequent disks will also be available at $15.  

All of these disks may be freely copied and passed to others for
their use.  Please tell us about any changes you had to make to
run the programs on your machine.  Please make the printed results
show who made what changes, and mark them with consistent comments
in the source.  The purpose of this request is to make the results
more easily comparable.

Do not be surprised to discover serious difficulties with your
computer's arithmetic.  Most arithmetics designed before the
IEEE 754 Binary Floating-point standard have half a dozen minor
or serious flaws.  However, while these should not be surprising,
they may well be distressing, depending on your specific needs.

   Try running the programs.

   Tell us about the changes you made.

   Send a disk with the results you got.
 
   Pass the programs on to whomever can use them.

Together, we may uncover important problems to be addressed by 
hardware and software companies, world-wide.  It could help.

Dick Karpinski       24 Feb 1986
