How can I force float->double promotion for argument passing on PPC?

[Jeffrey A Law]

  In message < 00031400261500.01444@enzo.bigblue.local >you write:
  > Am Mon, 13 Mar 2000 schrieb Jeffrey A Law:
  > >In message < 00031322443600.01198@enzo.bigblue.local >you write:
  > >  > Am Mon, 13 Mar 2000 schrieb David Edelsohn:
  > >  > >I believe that this is fixed in the development sources on the
  > >  > >newppc-branch.  At one point there was concern about breaking backwar
  > d
  > >  > >compatbility, but that has been resolved.
  > >  > 
  > >  > No, the newppc-branch still shows the same behaviour.
  > >I do not think the behavior you want is not something we want to do -- unl
  > ess
  > >I'm misunderstanding something about this discussion.
  > 
  > Hmm, so you mean we should continue to violate the PPC SYSV ABI in this
  > respect? That's OK with me, but probably it should be stated somewhere.
It's in the manual.  The statement in the manual is a blanket statement about
the behavior of the compiler in this regard across all platforms.

  > 
  > On binary compatibility, it's currently broken anyway vs. doubles, please l
  > ook
  > at http://gcc.gnu.org/ml/gcc-patches/1999-11n/msg00616.html (btw, I'm seeki
  > ng
  > approval for this, mainline+branch).
  > 
  > >Please see the GCC manual's discussion of float vs double promotion for
  > >arguments.
  > 
  > I can't find it, any hints on where to look?
Try bugs, non-bugs, incomapatibilities or something like that.  It discusses
float vs double promotions explicitly.  It is a blanket statement about how
GCC should work regardless of how other compilers should work.

jeff