home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.arch
- Path: sparky!uunet!iWarp.intel.com|ichips!ichips!glew
- From: glew@pdx007.intel.com (Andy Glew)
- Subject: Re: 128 bit floating point formats
- In-Reply-To: supnik@human.enet.dec.com's message of 12 Sep 92 00:42:02 GMT
- Message-ID: <GLEW.92Sep15181245@pdx007.intel.com>
- Sender: news@ichips.intel.com (News Account)
- Organization: Intel Corp., Hillsboro, Oregon
- References: <1746@sousa.ltn.dec.com>
- Date: Wed, 16 Sep 1992 02:12:45 GMT
- Lines: 24
-
- Bob Supnik >Supnik@human.enet.dec.com
-
- I have seen (at least) three formats for 128 bit (quad precision) IEEE
- floating point:
-
- ...
-
- 3. RS/6000: two double precision numbers used, with exponents offset by
- 2**(-53) (I believe), in effect 1 sign bit, 11 bit exponent, 105 bit
- fraction with implicit high order 1 (hidden bit)
-
-
- Q: I believe that the RS6000's double double format is not hardware supported.
- Don't they just emit the correct sequence of DP flops to get the QP answer?
- --
-
- Andy Glew, glew@ichips.intel.com
- Intel Corp., M/S JF1-19, 5200 NE Elam Young Pkwy,
- Hillsboro, Oregon 97124-6497
-
- This is a private posting; it does not indicate opinions or positions
- of Intel Corp.
-
- Intel Inside (tm)
-