revap types, step 1

[Steve Kargl]

On Thu, Aug 26, 2004 at 01:06:30PM -0700, Richard Henderson wrote:
> On Thu, Aug 26, 2004 at 11:06:04AM -0700, Steve Kargl wrote:
> 
> > Incidently, the behavior of the -r8 option is outside the scope
> > of the Fortran standard.  It typically means that all REAL variables
> > are promoted to DOUBLE PRECISION.  It does not necessarily
> > include the promotion of all DOUBLE PRECISION variables to 
> > a QUAD precision. 
> 
> Really?  In which case maybe we don't have to do anything.
> With no special options, GCC does not currently support a
> target that doesn't have both a 1 word and a 2 word float.
> 
> I can certainly change the patch such that DOUBLE PRECISION
> doesn't adjust based on -r8.  I'm not a Fortran user though,
> so someone will have to tell me what's most useful.
> 

I'll defer to Paul and Toon.  AFAICT, g77 does not have an
option equivalent to -r8, so this isn't a compatibility issue.
Here's the man page entries for f2c and NAG's F95 compiler.

f2c
  -r8  Promote REAL to DOUBLE PRECISION, COMPLEX to DOUBLE
	   COMPLEX.

NAG
  -r8  Double  the size of default REAL/COMPLEX, and on machines for
       which quadruple-precision REAL/COMPLEX are available,  double
       the  size  of  DOUBLE PRECISION.  REAL/COMPLEX specified with
       explicit KIND numbers or byte lengths are  unaffected  -  but
       since  the  KIND  intrinsic  returns the correct values, COM-
       PLEX(KIND(0d0)) on a machine  with  quad-precision  floating-
       point will correctly select quad-precision COMPLEX.

So, if a target supports QUAD, then I think we should promote
DOUBLE PRECISION to QUAD.  If a target does not support QUAD,
then we ignore promotion of DOUBLE PRECISION.

-- 
Steve