Re: HP Solve issue #26 is posted... Message #5 Posted by Paul Dale on 3 Feb 2012, 6:14 p.m., in response to message #4 by Richard J. Nelson
Registers store floating point numbers in densely packed decimal format. We're not using the prefix encoding so we're not binary compatible with other such implementations -- not doing so avoids some large look up tables. Floating point numbers in registers are stored in decimal64 format, double precision numbers in decimal128. Statistical accumulations use both of these formats to avoid the large but similar numbers causing truncation problems (as mentioned in many HP calculator manuals).
Internally, computations are performed in base-1000 carrying 39 digits and a huge exponent range.
- Pauli
|