Form a "usually reliable source" (at IBM) (exact quotation), "IBM has stated that it will be putting support for decimal floating-point in hardware in future processors. That hardware will implement the decimal formats and arithmetic that were agreed by the 754r committee in 2003 and which are in the current IEEE 754r draft."
I was also told to "watch this spot" for future announcements (but no indications WHEN). NOTE WELL: to follow-up on my original note, it really, REALLY, is important for those concerned with this issue to vote (hopefully NO) on the outstanding proposal to CHANGE the current draft specification to a "binary" format - and to do so before Sep 28. Again, if you have any question (if my original note was not worded well enough), please feel free to contact me "off-list" at: wmklein <at> ix.netcom.com > -----Original Message----- > From: IBM Mainframe Assembler List > [mailto:[EMAIL PROTECTED] On Behalf Of Ray Mullins > Sent: Monday, September 19, 2005 10:58 AM > To: [EMAIL PROTECTED] > Subject: Re: IBM and float decimal > > Interesting. I've heard there are two levels of precision, > which I guess > would correspond to 4 and 8 bytes. > > Now to go do some Google searching. > <snip>> > > > -----Original Message----- > > From: IBM Mainframe Assembler List > > [mailto:[EMAIL PROTECTED] On Behalf Of glen > > herrmannsfeldt > > Sent: Monday 19 September 2005 00:13 > > To: [EMAIL PROTECTED] > > Subject: IBM and float decimal > > > > Ray Mullins <[EMAIL PROTECTED]> wrote: > > > > > I can't vote, but I've got a question... > > > > > I've heard rumors about a new IBM numeric format - sort of like > > > "decimal floating point" - combining the best of float + PD. It's > > > supposedly in the > > > z9 hardware (or will be). How does this play in this whole > > discussion? > > > > There has been discussion of the proposed float decimal > > standard for some time, and I believe some papers (maybe > > electronic only) written about it. It is pretty close in > > efficiency for the available range to float binary, where the > > extra range of a base 10 exponent makes up for the reduction > > in the fraction due to decimal normalization. > > (The fraction is stored three digits per 10 bits, with the > > remaining digits in normal BCD form, and a base 10 exponent.) > > > > I have not heard any stories of it appearing on any machine, though. > > ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
