🚀 go-pugleaf

"robin"  wrote
> "Glen Herrmannsfeldt" writes:
> > According to his description, until 1963 IBM Fortran compilers would
> > generate double precision results for floating point multiply of two
single
> > precision numbers, at least when it was being summed into a double
precision
> > variable.  He then explains that IBM changed this in IBSYS 13,
anticipating
> > S/360.
> >
> > As far as I can tell, the Fortran-66 standard requires a single
precision
> > product when multiplying single precision numbers, though I don't know
how
> > much impact IBM had on Fortran-66.
> >
> > The reason why this comment is surprising to me is that S/360 (along
with
> > S/370 and ESA/390) do not have a single precision multipy instruction
with a
> > single precision product.  The product is always double precision.
>
> While this is true, it is also true that subsequent single-precision
> operations using the result will use only the single-precision
> product.

With the Fortran language as implemented on IBM computers in the 360 series
you specified all variables either implicit or explicit. Variables having
names
beginning with one of the characters "I" through "N" were INtegers by
default, the
rest were single precision real numbers by default.

How the multiplication was performed internally mattered very little for
code something
like A = B * C. According to the specifications for the Fortran compiler all
real
number calculations were carried out at maximum precision for intermediate
results;
rounding to the specific precision took place only when a result was stored
in the
target variable.

Therefore the result for:
Y = some second calculation on (the result of some first calculation on
(A,B,C) )

could very well be different from the result for:
Y := some first calculation on: (A, B, C)
followed by
Y = some second calculation on (Y)

because also the intermediate result was rounded in the second case, not in
the first.

Sven