John McFarlane

Is it possible to produce a quotient in the form of a fixed_point object that saves on run-time `idiv`? When the divisor is not known at compile time but will...

At least some versions of GCC seem to have `fmod` marked as `constexpr`. Try to detect this reliably and support compile-time `auto n = wide_integer{1.23};` for some targets.

Failing test can be found in #320.

I.e. assert that int > INT_MIN

Rounding free function failure

6

static constexpr auto a = cnl::elastic_number{0.3125}; static constexpr auto b = cnl::convert{}(a); static_assert(identical(cnl::elastic_number{0.5}, b),"Free rounding test1"); //Works static constexpr auto c = cnl::convert{}(a); static_assert(identical(cnl::elastic_number{0.25}, c),"Free rounding test2"); //Fails gives 0.75...

In ```c++ auto a = fixed_point(0x100000000LL); auto b = fixed_point(1); ASSERT_GT(a, b); ``` expression overflows. But only 16 (maybe 15) bits of range are required to scale operands such that...

Prove single-instruction specialization

2

E.g. an instruction that multiplies with saturation, then shifts right by 15 digits.

https://en.wikipedia.org/wiki/Scientific_notation#Other_bases http://en.cppreference.com/w/cpp/language/floating_literal

https://github.com/google/benchmark#preventing-optimisation

https://github.com/foonathan/type_safe