note the address change, and a message from Earl Killian

[dgh]

1) Please note that the mailing list that I explicitly invited you to use
	...sun!dgh!numeric-interest
   used to be called nceg-interest; I changed the name to avoid any
   confusion that NCEG activities were the only subject or that this was
   an "official" mailing list of NCEG or anybody else.

2) My workstation's address is sometimes reported as dgh.eng.sun.com and
   sometimes not; the mail software seems to be vacillating on this point.
   I think sun!dgh!numeric-interest will work indefinitely.

3) The following message from Earl got sidetracked:

Date: Thu, 4 Jan 90 17:05:41 PST
From: earlamips.com (Earl Killian)
To: dfpaattunix.sf.att.com
Cc: nceg-interestadgh.Eng.Sun.COM
Subject:  what are the advantages of Fortran over C?

   From: dfpaattunix.sf.att.com
   Date: Thu, 4 Jan 90 16:37 EST

   >Another advantage of Fortran is that it has do loops.  C has only
   >while loops ("for" is just syntatic sugar for while).  It is possible
   >to generate more efficient code for
   >	do i = 1,n
   >than for
   >	for (i = 1; i <= n; i++)
   >because of the more limited semantics of do loops.

   I completely disagree.  C optimizers have been turning such loops into
   the same sort of zippy instruction sequences for years now.  I admit
   that the analysis necessary to determine that a C loop matches the
   behavior of a count up/down loop is trickier, but it is quite do-able.

I would guess then that the C optimizers that you refer to are buggy
because such transforms probably fail when n = 2147483647.  The
programmer probably knew that n != 2147483647, but the compiler in
most cases cannot.