https://bugs.documentfoundation.org/show_bug.cgi?id=154254

Mike Kaganski <[email protected]> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |RESOLVED
         Resolution|---                         |NOTABUG

--- Comment #1 from Mike Kaganski <[email protected]> ---
This is not a bug.

Floating-point calculations in Calc are imprecise by nature [1]. Every single
calculation may (and in most cases does) create a calculation error; sum of
these errors would usually be non-0.

However, there are ways to account for the error *inside a series of
calculations* - such as Kahan algorithm and its variations [2]. Note that it
can only increase precision when during the series of calculations, not only
the intermediate results are stored (which have errors every time), but also
separate values of errors, which are combined independently, and are added to
the final result of the main sum.

Such an algorithm had been implemented in Calc in version 7.2 (bug 137679). So
now the calculations that are performed together can benefit from that. But
calculations that store intermediate results separately, obviously loose the
error information.

[1] https://wiki.documentfoundation.org/Faq/Calc/Accuracy
[2] https://en.wikipedia.org/wiki/Kahan_summation_algorithm

-- 
You are receiving this mail because:
You are the assignee for the bug.

Reply via email to