From: krishna.myneni@ccreweb.org   
      
   On 6/2/24 07:11, Krishna Myneni wrote:   
   ...   
   > Using F*/ in Forth:   
   >   
   > 3.12e306 1.1e145 2.8e150 f*/ fs.   
   > 1.22571428571429e+301  ok   
   >   
      
   The alternative way to calculate this without overflow in Forth is to   
   add/subtract the logarithms, then take the antilog. However, let's check   
   the accuracy of this approach compared to the above:   
      
   3.12e306 flog 1.1e145 flog f+ 2.8e150 flog f- falog fs.   
   1.225714285714187e+301  ok   
      
   The base 10 significand of the result is   
      
   3.12e 1.1e f* 2.8e f/ fs.   
   1.225714285714286e+00  ok   
      
   The result from using F*/ to the same number of decimal digits is   
      
   3.12e306 1.1e145 2.8e150 f*/ fs.  \ Anton's version of F*/   
   1.225714285714286e+301  ok   
      
   The relative error of the logarithm approach to using F*/ is   
      
   1.225714285714286e+301 1.225714285714187e+301 f- 1.225714285714286e+301   
   f/ f.   
   8.07495e-14  ok   
      
   We appear losing between about 3 decimal digits of accuracy with the   
   logarithm approach.   
      
   --   
   Krishna   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)