🚀 go-pugleaf

In article <Bz5DIs.2uH@mentor.cc.purdue.edu> hrubin@mentor.cc.purdue.edu (Herman Rubin) writes:
 > In article <8311@charon.cwi.nl> dik@cwi.nl (Dik T. Winter) writes:
 > >In article <Bz1rFw.9s1@mentor.cc.purdue.edu> hrubin@pop.stat.purdue.edu (Herman Rubin) writes:
 > > >                                                     The hardware has
 > > > deteriorated, except for speed, for the number theorist; when Lehmer
 > > > wrote his criticism, virtually all computers had double-length integer
 > > > arithmetic.
 > >You asserted that before, and I asked you to name a few computers from the
 > >beginning of Fortran that had those capabilities.  I never saw a response.
 > >Can you respond now?
 > 
 > The IBM 704-709 series, which were the machines for which Fortran was written;
 > The Univac 1100 series; the CDC 1604 and 3600; and at least the great bulk of
 > the von Neumann type machines.  I am  not too familiar with the some of the
 > earlier machines, but the more-or-less standard hardware of that time had an
 > accumulator and a multiplier/quotient register; the actual additions were
 > done in the accumulator, and on multiplication the least significant part
 > was shifted into the MQ register, with the most significant part in the
 > accumulator, and on division, with the double-length dividend in the A/MQ
 > registers, the remainder was left in the accumulator as the quotient was
 > produced in the MQ register.	

While I now remember this was not the question I posed some time ago, I
see that you now say that virtually all computers did *not* have
double-length integer arithmetic.  What you say is that they had
n*n->2n and 2n/n->n.  I agree with that.  A new question: what current
machines do not have those operations?  (I know for multiply: Cray,
HP-PA version I, and Motorola 88000, do you know others?  I know also
that the newer version of HP-PA does provide the requested multiply.
There are a few more that do not provide the divide, but they do in most
cases not even provide HW integer division.  Sometimes designers have
reasons to omit something.  And, BTW, there are also old processors
that do not provide HW integer division!  And, as another BTW, the
processors I know that do multiply by a multiply-step instructions
provide the double length product: Sparc version 7 and earlier (version
8 has the multiply), AMD 29000 (later versions have the multiply) and
IBM PC/RT.)

My old question, which you never answered, was based on another assertion
by you.  You told us that in the early days of Fortran virtually all
processors had single*single->double fp multiply.  My question still stands:
name a few from that time.
-- 
dik t. winter, cwi, kruislaan 413, 1098 sj  amsterdam, nederland
home: bovenover 215, 1025 jn  amsterdam, nederland; e-mail: dik@cwi.nl