• nickcw an hour ago

    Interesting reading about the difficulties with arctan.

    There is a better series for arctan than the Taylor series which converges for all x. You can see it here in the accelerated series section

    https://en.wikipedia.org/wiki/Arctangent_series

    I wrote about it in my calculating pi blog some time ago

    https://www.craig-wood.com/nick/articles/pi-machin/

    It also takes fewer operations which is nice.

    I thought it was invented by Euler but the Wikipedia article says Newton invented it and Euler popularized it.

    • HeavyStorm 2 hours ago

      Damn, blogspot still exists! And it doesn't render well in my mobile, a product of simpler times.

      • presz 4 hours ago

        You can't call it Ivy! That's an emacs package!

        • saghm 3 hours ago

          I genuinely can't tell if this is serious or not (or if the fact that it's impossible to tell is itself the point)

        • kayo_20211030 6 hours ago

          Super interesting. Thank you.

          • measurablefunc 2 hours ago

            By definition, they can not be implemented. You can approximate them to some finite precision but you can not really implement them.

            • Someone 29 minutes ago

              The typical computer algebra system (https://en.wikipedia.org/wiki/Computer_algebra_system) implements trigonometric functions just fine (with varying levels of sophistication)

              They can be implemented, but what you cannot always do is compute their values in a finite number.

              • measurablefunc 4 minutes ago

                Let me know when you find infinite amount of time on your hands to wait for the calculation.