Those reference values are specifically for "double" with round-to-nearest.
M.
___
gnumeric-list mailing list
gnumeric-list@gnome.org
https://mail.gnome.org/mailman/listinfo/gnumeric-list
The crlibm values are carefully curated as the worst possible cases
for double rounding.
I don't know of anyone who has searched for the worst cases for long
double. It may not be computationally feasible with current
understanding and technology.
If you just want sample values, you can use
>> - how to deal with / construct tests for functions / ranges where 'long' and
>> 'double' version have justified different results?
( how to build tests which allow / demand improved results, but accept
'double accuracy results' when using double datatype? )
let me give it an
yes, I know, I understand,
my thoughts are:
1. How do we get tests that are suitable for 'long'?
2. do the current tests hinder improvements?
( suppose an error in the program or e.g. a library has led to a weak reference ... the error is not noticed because the references are not