Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!zaphod.mps.ohio-state.edu!swrinde!cs.utexas.edu!sun-barr!newstop!exodus!exodus-bb!khb From: khb@chiba.Eng.Sun.COM (Keith Bierman fpgroup) Newsgroups: comp.arch Subject: Re: IEEE arithmetic (Goldberg paper) Message-ID: Date: 9 Jun 91 20:44:01 GMT References: <9106072346.AA08023@ucbvax.Berkeley.EDU> Sender: news@exodus.Eng.Sun.COM Organization: Sun MegaSystems Lines: 52 In-reply-to: jbs@WATSON.IBM.COM's message of 7 Jun 91 23:50:01 GMT In article <9106072346.AA08023@ucbvax.Berkeley.EDU> jbs@WATSON.IBM.COM writes: > I don't believe interval arithmetic is used enough to justify > any hardware support. This falls into the "H.R." chicken and egg rathole. If interval arithmetic (or anything else new and claimed to be good) is 100x slower than the traditional approach, it will simply never get off the ground. The members of IEEE 754 chose to include rounding modes; at this point we should either craft arguments for deleting them from a future version of the standard or debate something else. Arguing for their inclusion in a system is now a moot point; they are part of the standard. >.... credentials... Kahan was given the ACM Turing award for his contributions to floating point (most notably the IEEE standard). I personally suspect this counts for more than some heated debates on usenet. It is pointless for us to post our respective resumes. >... misc. fortran vs. ieee interactions .... The ways various standards interact is usually unpleasant in edge cases (SVID vs. POSIX vs. ANSI C for example). One hopes that in time, sucessful standards will see evolution _towards_ mutual compatibility. In the meantime it is a series of "quality of implementation issues" for each vendor to cope with as best they can. X3J3 and X3J11 both steered away from trying to make their languages fully IEEE 754 savvy. Unfortunately, there does not yet seem to be a move afoot to correct this in their next efforts .... but "f90" is just out the door (standard-wise) so we have ample time to hope for the best from X3J3|WG5 next time ;> It is unclear to _me_ if X3J11 or X3J16 is really the best forum for trying to fix C the next time .... there is a group called NCEG which is taken on the challenge of crafting proposals and finding an eventual venue. As has been pointed out by others, there are very few computing environments which provide handles on the bulk of the IEEE754 knobs. This is unfortunate, as many of the benefits of IEEE754 are lost thereby. Thus providing those who wish to propose weaker fp standards with fodder of the sort "no one uses those features anyway" ... this is quite unfortunate. -- ---------------------------------------------------------------- Keith H. Bierman keith.bierman@Sun.COM| khb@chiba.Eng.Sun.COM SMI 2550 Garcia 12-33 | (415 336 2648) Mountain View, CA 94043