From: mailbox@dmitry-kazakov.de   
      
   On 2022-12-14 16:18, Richard Heathfield wrote:   
      
   >> BTW, averaging floats is a nasty problem too. A naive implementation   
   >> quickly loses precision.   
   >   
   > We're dealing with 'o'clock' and "HH:MM", and nowadays we have 64-bit   
   > integer types and there are even 128-bit integers mooching around   
   > looking for a reason to exist. You'd have to average a hell of a lot of   
   > times even to /need/ floats, let alone lose significant precision.   
      
   I never suggested float for averaging time stamps, I pointed out that   
   averaging is not a simple problem. E.g. try this one:   
      
       function Average (X : Float; N : Positive) return Float is   
          Sum : Float := 0.0;   
       begin   
          for Index in 1..N loop   
             Sum := Sum + X;   
          end loop;   
          return Sum / Float (N);   
       end Average;   
      
   The function does naive averaging. For simplicity it just sums up the   
   same number X N times and divides by N.   
      
       Average (1.0, 17_000_000)  = 0.986895   
       Average (1.0, 100_000_000) = 0.167772   
       Average (1.0, 200_000_000) = 0.838861   
      
   The issue is catastrophic precision loss of addition Sum + X when Sum >> X.   
      
   --   
   Regards,   
   Dmitry A. Kazakov   
   http://www.dmitry-kazakov.de   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)