Re: whole number constants

"glen herrmannsfeldt" <gah@[EMAIL PROTECTED]
> wrote in message
news:QOGdnTupY-HqlRvVnZ2dnUVZ_uydnZ2d@[EMAIL PROTECTED]
> Richard Maine wrote:
>
> > robin <robin_v@[EMAIL PROTECTED]
> wrote:
>
> >>A single precision value in storage is converted to double precision
> >>by appending (extending with) binary zeros.
>
> > Nothing in the Fortran standard specifies that. It is the most usual
> > implementation choice, but it is not required by the standard.
> (snip)
>
> Another case which resulted in a long discussion some years
> ago is the product of two single precision values.  On many
> machines such product has double (or higher) precision, but
> the standard specifies single precision.

The precision of the product is single, and would be
considered as such in any subsequent arithmetic for that
particular statement, and /or realized on assignment.
On the PC, extra precision bits are maintained in both SP and DP
arithmetic operations.  These extra (superfluous) bits are not
set to zero in arithmetic operations.

>  Many systems go
> through the extra work to put zeros in the low bits,

Such as?

> but
> I don't believe the standard requires that.  That is,
> the single precision product can be extended with the
> appropriate bits needed for a double precision product.

The function DPROD explicitly produces a DP result
from the product of two SP operands.