home *** CD-ROM | disk | FTP | other *** search
- >>>>> "Scott" == Scott Wurcer <Scott.Wurcer@analog.com> writes:
-
- >> Yes, that's exactly right. Floating point could even be a lot
- >> faster than it is, but we just don't have time to work on
- >> optimizing SANE right now.
-
- Scott> How does executor handle the 96 bit extended format ? I
- Scott> have noticed differences on some computations. Namely ones
- Scott> that return -0 from SANE and "?" under E/L. I haven't had
- Scott> time to document the exact computations though.
-
- The "96 bit extended format" is really 80-bit IEEE with 16 bits of
- zeros thrown in. The 80387 also supports 80-bit floating point so we
- just use it to do the computations.
-
- -Mat
-
-