Nick

@jzmaddock : Yes; [here's](https://randomascii.wordpress.com/2012/01/23/stupid-float-tricks-2/) some background on the trick.

@jzmaddock : Yeah, the fact that `agm` requires positive numbers (over the reals) simplifies the logic considerably. The wins for float128 are pretty huge: Without the fast float distance: ```...

@wgledbetter : Yes, we're interested. @pulver : Can your review the PR when it shows up?

@jzmaddock : Yeah, numerically evaluating the condition number is a bit of a problem. We could decrease the order of the derivative to get a quick win and probably get...

Ok, just checked with this diff: ``` diff --git a/example/airy_ulps_plot.cpp b/example/airy_ulps_plot.cpp index 695d49287..298eae417 100644 --- a/example/airy_ulps_plot.cpp +++ b/example/airy_ulps_plot.cpp @@ -6,12 +6,13 @@ #include #include #include +#include using boost::math::tools::ulps_plot; int main()...

@jzmaddock : Do any of these fixes look appealing to hou?

@mborland : OMG thank you; we've needed a prime sieve for so long.

@mborland : I would definitely change the name from `prime_functions.hpp` to `prime_sieve.hpp`; the first is not very memorable.

@mborland : IIRC, the Joy of Factoring also creates bitsets which set prime at bit i and has it zeroed at composites. Is this a useful API?

Would it be more ergonomic to have `prime_range` use a `back_inserter`, but `prime_sieve` behave more like `std::fill`? ```cpp std::vector primes(1000); // gimme 1000 primes prime_sieve(primes.begin(), primes.end()); ``` (I need to...