[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: small integers, numerical efficiency

  Date: Thu, 29 Oct 92 15:56:46 EST
  From: "David A. Moon" <moon@cambridge.apple.com>
  To: John Lewis <zilla@ccrl.nj.nec.com>
  Subject: Re: small integers, numerical efficiency
  Cc: info-dylan@cambridge.apple.com

  What I said was that I think a Dylan implementation would be better off 
  leaving out floating point entirely than putting in floating point but not 
  doing it efficiently.
  Then from a language point of view, we could decide that few implementations 
  will want to bother doing floating point efficiently, and therefore remove it 
  from the language, or we could make it optional in the language, or we could 
  leave the language the way it is and some implementations could announce that 
  they only support a subset.  Personally I prefer making it optional over 
  either extreme.  Floating point is not the only thing in the Dylan book that 
  ought to be optional.
Does "optional" mean one would not be able to use floats in a portable way? 

  Also, C's integer numeric system is half-baked, but its floating-point numeric 
  system is not particularly bad.  I say enough bad things about C that you 
  don't need to misquote me disparaging a part of C that doesn't bother me!

Perhaps we should order the numerical classes on page 147 by frequency of use.
Then we could use this to focus our attention on performance, or which
classes to prune from the language.