On 21 August 2017 at 14:39, John Maddock via Boost wrote:

On 21/08/2017 06:33, Lakshay Garg via Boost wrote:

...

Hello all

The boost.math library stores quite a few constants for improving performance of various operations (example: boost/math/special_functions/detail/unchecked_factorial.hpp). I was wondering if marking these constants as constexpr would be desirable.

Please provide feedback / suggestions so that I can get started :)

PR's would be welcome as ever ;)

We did make everything in boost::math::constants:: constexpr but missed those - the 3 main candidates are the tables of prime numbers, bernoulli numbers and factorials. Some of them may need C++14 constexpr rather than the vanilla C++11 version. We have BOOST_CXX14_CONSTEXPR and BOOST_CONSTEXPR for marking this stuff up in a backwards compatible way BTW. Tests and docs will need updating as well. The unchecked_ functions can probably be noexcept as well.

Thanks! John.| |

​Thanks for the suggestions! I will send in a PR in a day or two.

On 21 August 2017 at 18:03, Lakshay Garg wrote:

On 21 August 2017 at 14:41, Lakshay Garg wrote:

...

On 21 August 2017 at 14:39, John Maddock via Boost

...

wrote:

...

On 21/08/2017 06:33, Lakshay Garg via Boost wrote:

...

Hello all

The boost.math library stores quite a few constants for improving performance of various operations (example: boost/math/special_functions/detail/unchecked_factorial.hpp). I was wondering if marking these constants as constexpr would be desirable.

Please provide feedback / suggestions so that I can get started :)

PR's would be welcome as ever ;)

We did make everything in boost::math::constants:: constexpr but missed those - the 3 main candidates are the tables of prime numbers, bernoulli numbers and factorials. Some of them may need C++14 constexpr rather than the vanilla C++11 version. We have BOOST_CXX14_CONSTEXPR and BOOST_CONSTEXPR for marking this stuff up in a backwards compatible way BTW. Tests and docs will need updating as well. The unchecked_ functions can probably be noexcept as well.

Thanks! John.

​@john, I was wondering if it would make sense to replace all "static const"s with BOOST_STATIC_CONSTEXPR?​

​^ by all I meant wherever it would compile successfully :)