[Libre-soc-dev] daily kan-ban update 22apr2022
luke.leighton at gmail.com
Sat Apr 23 05:27:36 BST 2022
On April 23, 2022 2:33:33 AM UTC, Jacob Lifshay <programmerjake at gmail.com> wrote:
>Working on the goldschmidt division algorithm...
>I decided to change the algorithm to calculate the error bounds as
>it's generating the ops to run the div algorithm, and then throw an
>exception if the parameters are not accurate enough. that way, I can
>have a function that repeatedly calls the algorithm and varies the
>parameters until they work.
nice. so an exhaustive search provides the bounds on the (3?) parameters and they can be post-analysed to find patterns.
otherwise, yeah, it is necessary to always have one more iteration and test the difference.
i wonder if there might be a correlation where the difference can be used to determine whether one final iteration is required.
More information about the Libre-soc-dev