hrubin@pop.stat.purdue.edu (Herman Rubin) (05/17/91)
In article <15909@darkstar.ucsc.edu>, haynes@felix.ucsc.edu (99700000) writes: > > In article <12236@mentor.cc.purdue.edu> hrubin@pop.stat.purdue.edu (Herman Rubin) writes: > > > >Why do we have separate integer and floating units, especially without > >communication between them? I suggest those who push this horror look > >at how difficult conversion between them is. I have already pointed out > >that every trigonometic and exponential routine does, in some way, > >float/float -> integer, remainder. The integer is also used. .................... > Now that we have a floating point standard that requires almost always > normalizing I guess we can never go back to where we were in 1964. > (Burroughs - er, ah, Unisys - is still making machines with the old > number representation, but they must be mightly lonely.) They are not the only ones; the CYBER 205/ETA 10 and the CRAYs still have it. Even the 360 and its descendents, which do not have integer/ floating communication, do not (and in this case cannot)have forced normalization. Is the 1-bit gain from forced normalization worth the other problems? For example, without it, the RS/6000 would be capable of reasonable 52x52 -> 104 multiplication in its floating unit, with addition included, especially if a quick add-to-exponent operation were included. Oops, that would require communication between the integer and floating unit; so let's make that add-to-exponent immediate. Versatility cannot be easily added after the design, but it can be before. -- Herman Rubin, Dept. of Statistics, Purdue Univ., West Lafayette IN47907-1399 Phone: (317)494-6054 hrubin@l.cc.purdue.edu (Internet, bitnet) {purdue,pur-ee}!l.cc!hrubin(UUCP)
amos@SHUM.HUJI.AC.IL (amos shapir) (05/23/91)
[Quoted from the referenced article by hrubin@pop.stat.purdue.edu (Herman Rubin)] > >Is the 1-bit gain from forced normalization worth the other problems? Normalization is not done just for saving one bit. Its main purpose is to ensure a 1-1 correspondence between bit patterns and numeric values. Efficiency is also served: the format used by the IEEE754 standard also produces monotonous values, which makes comparison of FP numbers very easy. -- Amos Shapir amos@cs.huji.ac.il The Hebrew Univ. of Jerusalem, Dept. of Comp. Science. Tel. +972 2 585257 GEO: 35 11 46 E / 31 46 21 N
herrickd@iccgcc.decnet.ab.com (06/02/91)
In article <1207@shum.huji.ac.il>, amos@SHUM.HUJI.AC.IL (amos shapir) writes: > Efficiency is also served: the format used by the IEEE754 standard also > produces monotonous values, which makes comparison of FP numbers very easy. > This thread has not become monotonous, it is just beginning to be interesting. The word you wanted is monotonic. (Don't be offended, my Hebrew is very poor.) dan herrick herrickd@iccgcc.decnet.ab.com