XPost: comp.lang.c++   
   From: lynnmcguire5@gmail.com   
      
   On 2/6/2026 2:48 PM, David Jones wrote:   
   > Thomas Koenig wrote:   
   >   
   >> Radey Shouman  schrieb:   
   >>   
   >>> The size of the integer type was specified to be the same as the   
   >>> size of the real type.  The real type need not be 32-bit floating   
   >>> point, but that is certainly the way to bet.   
   >>   
   >> A famous counter-example: Cray's 64-bit REALs.   
   >   
   > And the Univac's 36 bit reals.   
      
   I miss the Univac 1108 36 bit reals and integers.  They were perfect for   
   1 ppm (1e-6) calculations, good for most of our calculations.   
   Unfortunately, we had to walk away from the Univac 1108 in 1981 ? due to   
   its 1 megaword addressing limit.  We had already moved to a Prime 450 in   
   1977 and a Prime 750 in 1980 ? for our software development.   
      
   Double precision in 64 bit reals takes you to 1e-15 calculations   
   generally.  That is actually a problem for some iterative calculations   
   as the delta variable goes too small and must be tested for routinely.   
      
   Lynn   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)