Question: Bit-exact floats on different computers?

[theagentd]

@Roquen

Looking quickly, the follow should all be good of the fp routines:
sqrt,ceil,floor,rint,round,abs,min,max,ulp,signum,copySign,next{Up,Down,After},scalb

Are all functions in StrictMath OK in that sense as well?

EDIT: Apparently that’s the whole point of StrictMath. According to this: https://gist.github.com/ijuma/840120, the only things slower are sin, cos, tan, exp and log.

[Roquen]

Yeah StrictMath is bit-exact, but so is rolling your own approx defed with strictfp. The compiler is aware of many of these functions and sqrt is an intrinsic. The rule is the function must be “properly rounded”, “correctly rounded” isn’t good enough to insure bit-exact.