Mathematical constants¶

Defined in header <Kokkos_MathematicalConstants.hpp> which is included from <Kokkos_Core.hpp>

Attention

Recommendation: Since Kokkos 5.X requires C++20, users are encouraged to use the standard library constants (std::numbers::*) directly. The Kokkos::numbers namespace is maintained for backward compatibility and is implemented via using-declarations of the standard library constants.

Usage¶

auto const x = Kokkos::numbers::pi_v<float>;
auto const y = Kokkos::numbers::sqrt2_v<float>;

Provides access to the mathematical constants defined in the C++20 <numbers> header in the Standard Library.

All constants are defined in the Kokkos::numbers:: namespace.

Variable Templates¶

The following are variable templates defined for standard floating-point types (float, double, long double).

Template name	Math symbol	Description
`e_v`	\(e\)	Base of the natural logarithm
`log2e_v`	\(\log_{2}{e}\)	Base-2 logarithm of e
`log10e_v`	\(\log_{10}{e}\)	Base-10 logarithm of e
`pi_v`	\(\pi\)	Ratio of a circle’s circumference to its diameter
`inv_pi_v`	\(\frac{1}{\pi}\)	Inverse of pi
`inv_sqrtpi_v`	\(\frac{1}{\sqrt{\pi}}\)	Inverse of the square root of pi
`ln2_v`	\(\ln{2}\)	Natural logarithm of 2
`ln10_v`	\(\ln{10}\)	Natural logarithm of 10
`sqrt2_v`	\(\sqrt{2}\)	Square root of 2
`sqrt3_v`	\(\sqrt{3}\)	Square root of 3
`inv_sqrt3_v`	\(\frac{1}{\sqrt{3}}\)	Inverse of the square root of 3
`egamma_v`	\(\gamma\)	Euler-Mascheroni constant
`phi_v`	\(\varphi\)	Golden ratio constant \(\frac{1+\sqrt{5}}{2}\)

Convenience Constants (`double`)¶

For each variable template listed above, Kokkos provides an inline constexpr double constant without the _v suffix. These are shorthand for the double specialization.

Kokkos::numbers::pi is equivalent to Kokkos::numbers::pi_v<double>
Kokkos::numbers::e is equivalent to Kokkos::numbers::e_v<double>

Notes¶

Important

Portability: Passing mathematical constants by reference or taking their address in device code is not supported by some toolchains and hence not portable. (See known issues)

Note

Quadruple Precision: Support for quadruple precision floating-point __float128 can be enabled via -DKokkos_ENABLE_LIBQUADMATH=ON.

Example¶

KOKKOS_FUNCTION void example() {
    // Preferred C++20 usage
    constexpr auto pi_f = std::numbers::pi_v<float>;

    // Kokkos namespace usage (backward compatibility)
    constexpr auto pi = Kokkos::numbers::pi_v<float>;

    auto const x = Kokkos::sin(pi_f / 6);
}