mail headers and extended precision

[David G. Hough on validgh]

I never have any trouble figuring out who sent mail, although it seems to be 
a function both of the sender's mailer and the recipients.  For instance the
last few messages in the numeric-interest archive include

From: sun!Eng!dgh (David Hough)
From: sun!Eng!dgh (David Hough)
From: Tim Peters <uunet!ksr.com!tim>
From: uunet!research.att.com!sesv
From: sun!Eng!dgh (David Hough)
From: sun!Eng!dgh (David Hough)
From: Tim Peters <uunet!ksr.com!tim>
From: Tim Peters <uunet!ksr.com!tim>
From: sun!Eng!dgh (David Hough)
From: uunet!cims18.nyu.edu!mckenney (Alan M. McKenney)

which seems clear enough except for sesv.  /usr/ucb/mail on Suns identifies the
sender based on what's in /etc/passwd or NIS passwd.  Anyway...

>     Finally, at least on the Suns, I thought that the whole point of
> the FORTRAN compiler's -fstore option was to force a rounding step at
> each "=" in the source.

Actually -fstore is implemented by turning off register allocation, which eliminates
most of the optimization possibilities on extended-precision register architectures.

> shouldn't be a problem to run the 68881 so that all results are rounded
> to storage precision.

This is true as long as you don't care about performance.  Performance, however,
is the most important sales factor in workstations and larger systems.

>     Actually, the idea of automatically promoting intermediate
> results to a higher precision doesn't seem to me to be a win, anyway.

What about the extra exponent range?  Anyway the idea of extra precision is clear enough,
on 80x87 and 6888x systems,
if rounding to storage precision is enforced for assignments.  Automatically
computing in higher precision makes it possible for many programs to just work
without a lot of analysis, which could be done to make programs work
exploiting only storage precision, but usually isn't.