If -fns is not specified, the nonstandard floating point mode is not enabled automatically. Standard IEEE 754 floating-point computation takes place, that is, underflows are gradual.
If only –fns is specified, –fns=yes is assumed.