🚀 go-pugleaf

This article mentioned the "Wollongong" version ... having worked with the
original Wollongong version extensively, I know that indeed floating point
types in general do not work right in the version released by the
University of Wollongong. (I think PE fixed them.) The problems center
around the fact that the version he used when he ported the compiler only
supported one size of floating point word; there are all kinds of problems
related to floating point word length (e.g., constants are generated with
the wrong word length, activation records are created with the wrong size
(assuming a shorter word length than is actually used), etc.).  We tried to
fix that back when I was at Vanderbilt, and finally gave up because the
problems were so widespread (one of the people in our department had
programs for doing Hartree-Fock computations on atomic orbitals that would
reveal bugs in almost every FORTRAN compiler around, and she found new ones
as fast as we fixed the old ones).  We wrote to S.  I.  Feldman (the
Fortran compiler author) about the problems, and found out that in fact the
version of the Fortran compiler the U. of Wollongong people used was way
out of date at the time they released it (because there were also some
parser problems which Feldman said he'd fixed "several years ago").  Again,
I believe those problems don't exist in PE's released version.

I'll leave it to the compiler people at PE to comment on the current version,
but yes, if he's using the original Wollongong version, there are extensive
problems with floating point...
-- 
Full-Name:  J. Eric Roskos
UUCP:       ..!{decvax,ucbvax,ihnp4}!vax135!petsd!peora!jer
US Mail:    MS 795; Perkin-Elmer SDC;
	    2486 Sand Lake Road, Orlando, FL 32809-7642

"Not that the story need be long; but that it will take a long time
 to make it short."  -- H. D. Thoreau