This is the mail archive of the mailing list for the GCC project.

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: ia64 XFmode patch rides again (need ia64-linux testing)

> On Tue, Oct 28, 2003 at 02:01:02AM +0100, Jan Hubicka wrote:
> > > On Mon, Oct 27, 2003 at 04:42:06PM -0800, Zack Weinberg wrote:
> > > > I have no preference at present.  You can look in
> > > > ia64.c:ia64_hpux_init_libfuncs (IIRC) for the HP-UX names.  Note that
> > > > GCC's floating point emulation library does not support IEEE quad, so
> > > > the question is moot until that gets resolved.  (Or does x86-64
> > > > support this in hardware?)
> > > 
> > > soft-fp in glibc does support it.
> > 
> > What is the name nangling of soft-fp?  Does it match the gcc one (*tf
> > functions we can't use) or does it use something new?  (then it would
> > make more sense to match soft-fp instead of HP-UX I guess)
> Whatever you choose.
> soft-fp is primarily lots of macros you can use.
> E.g. on sparc64 in libc/sysdeps/sparc/sparc64/soft-fp/ soft-fp is used
> to implement functions as defined by sparcv9 ABI (_Qp_* which use pointers
> to long double instead of long double arguments and return values).
> When/if sparc32 adds long double suport (I did a patch a year ago or when,
> but did not have time for it since then) in libm, it should follow
> the ABI with _Q_* functions. libc/soft-fp/ directory has __addtf3 style
> routines in it, so those could be included directly.  Or you can come with
> something else.

I think I need to come with something else to avoid conflicts with *tf?
functions produced for 128bit padded 80bit arithmetics.
Andreas, any preferences on glibc side?

> 	Jakub

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]