#include <RandPhilox.hpp>

Inheritance diagram for alpaka::rand::Philox4x32x10Vector:

Public Types
using	EngineParams = engine::PhiloxParams< 4, 32, 10 >

using	EngineVariant = engine::PhiloxVector< EngineParams >

template<typename TScalar >
using	ResultContainer = typename EngineVariant::template ResultContainer< TScalar >

using	ResultInt = std::uint32_t

using	ResultVec = decltype(std::declval< EngineVariant >()())

Public Member Functions
ALPAKA_FN_HOST_ACC	Philox4x32x10Vector (std::uint32_t const seed=0, std::uint32_t const subsequence=0, std::uint32_t const offset=0)

ALPAKA_FN_HOST_ACC constexpr auto	max () -> ResultInt

ALPAKA_FN_HOST_ACC constexpr auto	min () -> ResultInt

ALPAKA_FN_HOST_ACC auto	operator() () -> ResultVec

Detailed Description

Most common Philox engine variant, outputs a 4-vector of floats

This is a variant of the Philox engine generator which outputs a vector containing 4 floats. The counter size is $4 \times 32 = 128$ bits. Since the engine returns the whole generated vector, it is up to the user to extract individual floats as they need. The benefit is smaller state size since the state does not contain the intermediate results. The total size of the state is 192 bits = 24 bytes.

Ref.: J. K. Salmon, M. A. Moraes, R. O. Dror and D. E. Shaw, "Parallel random numbers: As easy as 1, 2, 3," SC '11: Proceedings of 2011 International Conference for High Performance Computing, Networking, Storage and Analysis, 2011, pp. 1-12, doi: 10.1145/2063384.2063405.

Definition at line 87 of file RandPhilox.hpp.

Member Typedef Documentation

◆ EngineParams

using alpaka::rand::Philox4x32x10Vector::EngineParams = engine::PhiloxParams<4, 32, 10>

Definition at line 90 of file RandPhilox.hpp.

◆ EngineVariant

using alpaka::rand::Philox4x32x10Vector::EngineVariant = engine::PhiloxVector<EngineParams>

Definition at line 91 of file RandPhilox.hpp.

◆ ResultContainer

template<typename TScalar >

using alpaka::rand::Philox4x32x10Vector::ResultContainer = typename EngineVariant::template ResultContainer<TScalar>

Definition at line 108 of file RandPhilox.hpp.

◆ ResultInt

using alpaka::rand::Philox4x32x10Vector::ResultInt = std::uint32_t

Definition at line 110 of file RandPhilox.hpp.

◆ ResultVec

using alpaka::rand::Philox4x32x10Vector::ResultVec = decltype(std::declval<EngineVariant>()())

Definition at line 111 of file RandPhilox.hpp.

Constructor & Destructor Documentation

◆ Philox4x32x10Vector()

ALPAKA_FN_HOST_ACC alpaka::rand::Philox4x32x10Vector::Philox4x32x10Vector	(	std::uint32_t const	seed = `0`,
		std::uint32_t const	subsequence = `0`,
		std::uint32_t const	offset = `0`
	)

inline

Initialize a new Philox engine

Parameters

seed	Set the Philox generator key
subsequence	Select a subsequence of size 2^64
offset	Number of numbers to skip form the start of the subsequence.

Definition at line 99 of file RandPhilox.hpp.

Member Function Documentation

◆ max()

ALPAKA_FN_HOST_ACC constexpr auto alpaka::rand::Philox4x32x10Vector::max ( ) -> ResultInt

inlineconstexpr

Definition at line 118 of file RandPhilox.hpp.

◆ min()

ALPAKA_FN_HOST_ACC constexpr auto alpaka::rand::Philox4x32x10Vector::min ( ) -> ResultInt

inlineconstexpr

Definition at line 113 of file RandPhilox.hpp.

◆ operator()()

ALPAKA_FN_HOST_ACC auto alpaka::rand::Philox4x32x10Vector::operator() ( ) -> ResultVec