Re: PATCH: PR target/15383: __float128 function name is used for long double

["H. J. Lu" <hjl at lucon dot org>]

On Wed, May 12, 2004 at 01:12:16PM -0700, Zack Weinberg wrote:
> 
> This patch completely misses the point.
> 
> Let me reiterate that we don't have an 128-bit floating point
> emulation library.  Until we do, __float128 is useless (except on
> ia64-hpux where the system C library includes one).  First come up
> with the library, then we can decide what names to use for its
> interface, and then we can modify ia64.c to set up those names for the
> libfuncs.

I do have a __float128 patch for glibc. It works on x86-64 and ia64. I
will try ia32 later. But I am not sure how/if it will be accepted
into glibc. What other alternative do we have for __float128 library?

BTW, I have to use the patch enclosed to have a working __float128 on
ia64 and provide backward compatibility.


H.J.

Attachment: gcc-3.4-ia64-float128-2.patch
Description: Text document