On Wed, 26 Jan 2022, Maciej W. Rozycki wrote: > > The newer instruction semantics correspond to the functions fminimum_num > > and fmaximum_num, not fminimum and fmaximum (which are functions that > > treat both quiet and signaling NaNs like most libm functions do - any > > argument a NaN means the result is NaN). > > I got these wrong then, sorry. I got lost too: what is the difference > between `fmin'/`fmax' and `fminimum'/`fmaximum' then? It looks to me like > we have a zoo of selection functions now with very subtle differences > between each other.
fmin and fmax: * Treat quiet NaN as missing data and return the other argument (if a number). * Treat signaling NaN like most functions (raise invalid, return quiet NaN). fminimum and fmaximum: * Treat quiet NaN like most functions (return quiet NaN, no exceptions raised unless the other argument is signaling NaN). * Treat signaling NaN like most functions (raise invalid, return quiet NaN). fminimum_num and fmaximum_num: * Treat both quiet and signaling NaN as missing data and return the other argument (if a number). "invalid" is still raised if one argument is a signaling NaN (contrary to the normal rule that raising "invalid" means also returning a quiet NaN). -- Joseph S. Myers jos...@codesourcery.com
