Re: setsockopt() compat issue

[Robert Elz <kre%munnari.OZ.AU@localhost>]

    Date:        Mon, 20 Oct 2008 17:03:51 +0200
    From:        Johnny Billquist <bqt%softjar.se@localhost>
    Message-ID:  <48FC9DD7.9030501%softjar.se@localhost>

  | Don't that basically mean that an int should be 60 bits, while a pointer 
  | should be 18 bits?

That might be the way you'd do it, but as I recall (my memory is
pretty pathetic) most integer arithmetic used the "address" registers
(the 18 bit ones) - the 60 bit registers were basically float (though
also could be used for some integer operations I think - more that no-one
would really ever consider needing integers that big!   And that on that
system no-one really ever used integers anyway, other than as array
indices.)

But the point Simon was making, I think, was just that there really have been
architectures that have registers (true independent registers) with
different sizes unrelated to each other - not that picking a data type
model to use for that system would necessarily be hard (I think that in
their final days there might have been a Cyber C compiler, I never saw
it, just heard on the rumour mill...)

kre