Re: Meeting the Scope and Purpose of P754

[Nick Maclaren <nmm1@xxxxxxxxxxxxx>]

Vincent Lefevre <vincent@xxxxxxxxxx> wrote:
> On 2007-07-14 09:51:55 +0200, Florent de Dinechin wrote:
>
> > I believe it could be done in a way that would not go against the C99 
> > standard. C99 allows the compiler to use any precision larger than the 
> > declared one, but does not mandate to do so AFAIK (Guillaume or Vincent 
> > will correct me on this one).
>
> First, the double type does not necessarily correspond to the IEEE-754
> double precision (unless __STDC_IEC_559__ is 1). Then the compiler may
> use any precision larger than the one corresponding to the FP type
> (FLT_EVAL_METHOD gives information about that). But if you use a cast
> or an assignment (see 5.1.2.3#12, 6.3.1.5#2 and 6.3.1.8#2[52]), then
> the value must be converted to be exactly representable in its semantic
> type.
>
> This is what the C99 standard mandates, but in practice, compilers
> often generate buggy code on x86 processors (in particular under
> Linux, but I recall that MS-Windows can be affected too).

Yes, but it manages to make even that seriously ambiguous.  6.3.1.5
says:

       If  the  value
       being  converted  is  in  the  range  of  values that can be
       represented but cannot be represented exactly, the result is
       either  the  nearest  higher  or nearest lower representable
       value, chosen in an implementation-defined manner.   If  the
       value  being  converted  is outside the range of values that
       can be represented, the behavior is undefined.

As usual with the C standard, there is no hint anywhere of which
section takes precedence over which, and so it is perfectly reasonable
to interpret this as overriding Annex F.

There is an even more horrible problem, which is that C99 does
not define what it means by type in the specification of casts and
assignments.  Again, it is reasonable to interpret it as meaning
syntactic type, so the value need not be converted provided that
it is treated as if it were a value of another type.  The C90
standard implied that was the case; the C99 one implies that it is
not, but the only semi-definite statement is in a merely informative
example.


Regards,
Nick Maclaren,
University of Cambridge Computing Service,
New Museums Site, Pembroke Street, Cambridge CB2 3QH, England.
Email:  nmm1@xxxxxxxxx
Tel.:  +44 1223 334761    Fax:  +44 1223 334679