Scott P. Jones
Scott P. Jones
I see it's not in the number, something would need to restrict EBASE to 2 or 10. Could you have BinaryUnum{ESZ, UINT} & DecimalUnum{ESZ, UINT}?
I think it's just more complexity, that simply isn't needed. I think the only reason to use a base other than 2 is for conversion issues with human input/output.
Are the 3, 7, 16, 20 bits for exponents what are recommended in the book? 16 - ubit - sign - 3 exp = 11 bits for fraction on Unum16?...
I doubt anything you do is thoughtless or arbitrary! I hope you didn't think I was implying otherwise! (really need that "Scott's comments" repository for peer review before release! :grinning:)
OK, good information. I am more interested in how I can pack unums most efficiently for storage, in such a way that it is easy also to unpack them reasonably...
John: Does the esize need to be stored as esize-1? If it were esize, you could have 0 exponent bits (i.e. for storing simple integers without using up another bit....
@JohnLGustafson Thanks for the info. I was used to the decimal numbers in M/Mumps (now CachéObjectScript), which didn't represent Inf or NaN, because they always gave a MaxNumber or DivideByZero...
Yes, that's what I meant (of course, you expressed it better!)
> what bothers me about decimal floats is that very few floats come from or go to human-readable form @JohnLGustafson About decimal floats, it really is application specific. For the...
Another thing I'm very interested in, is how to convert both binary and decimal unums into keys that would be suitable for sorting correctly, and in such a way that...