Mark Klein
Tue Jan 26 06:53:00 GMT 1999

At 10:36 PM 1/25/99 -0700, Jeffrey A Law wrote:
>  In message < >you write:
>  > There is no millicode for long double support. Long double support is in
>  > the PRO ABI as the Quad Library. On HP, this is part of libc (look for
>  > _U_Qf* in libc.a). On HP, it exists, so there is no question about hard 
>  > vs. soft float on HP platforms, even if there is no FP coprocesser 
>  > present.
>Ah.  I think the same issues still apply though.

Then you're right ... I'm missing the point. On HP, I know that I can do
fmpy quad,quad for !TARGET_SOFT_FLOAT and _U_Qfmpy for TARGET_SOFT_FLOAT.
(I can even do _U_Qfmpy for both as it appears the HP compiler is doing).
What am I missing or misunderstanding?

>Even if it's a required library, I know of systems that do not have it.

That's my original question. I've no experience with non-HP PA-RISC and
didn't know if the entire PRO ABI got implemented on those non-HP machines.
This answers it.

>So, what I think we want is a target flag which enables long doubles.  
>Ports like mpe (and probably some hpux ports) will have that flag on,
>other ports will have it off.

-mquadlib work?
Mark Klein                                      DIS International, Ltd.                              415-892-8400
PGP Public Key Available			

More information about the Gcc mailing list