[libre-riscv-dev] fp special functions
Luke Kenneth Casson Leighton
lkcl at lkcl.net
Tue Aug 6 08:16:53 BST 2019
On Tue, Aug 6, 2019 at 1:46 PM Jacob Lifshay <programmerjake at gmail.com> wrote:
> that appears correct, do note that you need to use something more accurate
> than f64 to implement sin for f64 and the same for f32 and f16.
In the paper there was something about this. Certain instances need
double the precision in order to have the iteration stabilise.
The link is in the bugreport 127 and it is fascinating. Covers atan,
hyperbolic, linear interpolation, hypotenuse calculation and even
spherical to polar coordinates. Also inverses can be done because it
is an iterative algorithm.
More information about the libre-riscv-dev