Follow-up Comment #1, bug #47348 (group gsl): For the large double values above 2^52 (where the value after adding 0.5 cannot any longer be represented exactly because the number of mantissa bits is insufficient on standard IEEE 754 registers) one can patch this by replacing floor(x+0.5) by rint(x) (which gives correct results under gcc 7.5.0 with openSUSE 15.6 at least).
_______________________________________________________
Reply to this item at:
<https://savannah.gnu.org/bugs/?47348>
_______________________________________________
Message sent via Savannah
https://savannah.gnu.org/
signature.asc
Description: PGP signature
