Fw: Error in documentation for tan(), and poor implementation of sin() and cos()

[Paul Zimmermann]

       Hi Wilco,

> In all cases that is the shared range reducer which has bugs for huge inputs,
> so you could just directly check the range reducer. Using integer arithmetic
> like I did for float sin/cos results in a much simpler and faster range reducer.

instead of patching the range reducer, the ultimate solution is probably to
use some correctly-rounded implementation like the one from CORE-MATH.
Adhemerval, do you have performance figures for say sin?

Paul