© Either my Pentium III is bugged or there is a problem...elsewhere.
© Precision of double floating-point format (non-extended)
© is 53 bits, including the implicit leading 1 bit of the
© significand. So, there should be no glitch as early as
© at 2*44, accuracy being guaranteed up to ¯1+2*53.
© Numeric encoding of floating-point data type is correct
© but comparison functions are seemingly flawed:
(2*44)=1+2*44
1
(2*44)=2+2*44
0
(2*45)=1+2*45
1
(2*45)=2+2*45
1
(2*45)=3+2*45
1
(2*45)=4+2*45
0