> If that extra bit weren't there, the 80-bit number will round-to-even
> and therefore round down, back to 1, which isn't what is wanted.
It's been a few years since I read the spec, but as I remember it, the
IEEE sticky bit takes care of situations like this. If anything
non-zero gets shifted off the right of the mantissa, the sticky bit
remains turned on to ensure correct rounding.
> I think this is about the best one can do on an x86 architecture.
Testing in Emacs Lisp and GCL shows that the machine correctly
implements the sticky bit, so what is happening in C?
-s