From: Michael H. <mw...@py...> - 2005-11-16 22:11:56
|
Raymond Toy <ray...@er...> writes: >>>>>> "Mario" == Mario S Mommer <mm...@co...> writes: > > Mario> I like the +I.0f+0 etc. style proposed by Christophe a little bit > Mario> better. For NaNs one could use something like N.0d+<char>, where > Mario> <char> is a one-char abbrev. of whatever additional attribute is > Mario> appropriate. > > The entire mantissa of the NaN could contain information. So a > one-char abbreviation isn't enough. You need all the bits. In some sense :) > However, I'm not aware of any processor that actually tells you what > it encodes in the bits, if anything. I didn't think the intent was so much for processors to use these bits as libraries -- I think the bits get carried through computations in a defined way, so you could potentially tell which routine produced the NaN. I don't think people actually do this, either, mind, though on Windows there are at least two printed representations for NaN... Oh, and most processor documentation will tell you the difference between a SNaN and a QNaN. Cheers, mwh -- The Oxford Bottled Beer Database heartily disapproves of the excessive consumption of alcohol. No, really. -- http://www.bottledbeer.co.uk/beergames.html (now sadly gone to the big 404 in the sky) |