Some of our math functions have two implementations: A more efficient one for runtime (e.g. using PTX instructions), and a less-efficient one which is constexpr. I keep them separate both by having two header files, and by having a constexpr_ namespace so that nobody calls the slower implementation by mistake at runtime.
However, in this StackOverflow question, several methods are suggested for detecting and choosing, at compile time, which variation of the function to run. Perhaps we could use that to just have a singel ug
Some of our math functions have two implementations: A more efficient one for runtime (e.g. using PTX instructions), and a less-efficient one which is
constexpr. I keep them separate both by having two header files, and by having aconstexpr_namespace so that nobody calls the slower implementation by mistake at runtime.However, in this StackOverflow question, several methods are suggested for detecting and choosing, at compile time, which variation of the function to run. Perhaps we could use that to just have a singel ug