alpaka::rand::distribution::UniformReal< T_Result, T_Interval > Struct Template Reference

Select a floating-point value from a uniform interval. More...

#include <UniformReal.hpp>

Inheritance diagram for alpaka::rand::distribution::UniformReal< T_Result, T_Interval >:

Public Types
using	Base = internal::UniformRealBase<T_Result, T_Interval>
Public Types inherited from alpaka::rand::distribution::internal::UniformRealBase< T_Result, interval::CO >
using	Interval_type
using	result_type

Public Member Functions
constexpr	UniformReal (T_Interval interval=T_Interval{})
constexpr	UniformReal (T_Result min, T_Result max, T_Interval interval=T_Interval{})
template<concepts::UniformRandomEngine T_Engine>
constexpr auto	operator() (T_Engine &engine) const -> T_Result
	Selects a value from a uniform distribution over the configured (min, max) interval, respecting the specified interval bounds.
Public Member Functions inherited from alpaka::rand::distribution::internal::UniformRealBase< T_Result, interval::CO >
constexpr	UniformRealBase (T_Floating min, T_Floating max, T_Interval)

Static Public Member Functions
template<std::integral T_Value>
static consteval void	checkValueConformity ()

Private Member Functions
template<concepts::UniformVectorEngine T_Engine>
constexpr auto	engineDispatch (T_Engine &engine) const -> T_Result
	Dispatch for vector engines (uniform bit generators that return a vector (e.g Philox4x32x10Vector) to enable efficient double precision uniform_real generation (reducing the number of invocations).
template<concepts::UniformStdEngine T_Engine>
constexpr auto	engineDispatch (T_Engine &engine) const -> T_Result
	Dispatch for std engines and alpaka abbreviations conforming to the std::uniform_random_bit_generator concept (e.g.
constexpr auto	scaleInterval (T_Result const &normalizedVal) const -> T_Result
	For open bounds, the normalized value in (0, 1) may (still) hit the lower or upper bound due to floating-point rounding (after scaling is applied).

Additional Inherited Members
Protected Attributes inherited from alpaka::rand::distribution::internal::UniformRealBase< T_Result, interval::CO >
T_Floating const	m_max
T_Floating const	m_min
T_Floating const	m_range

Detailed Description

template<std::floating_point T_Result, concepts::Interval T_Interval = interval::CO>
struct alpaka::rand::distribution::UniformReal< T_Result, T_Interval >

Select a floating-point value from a uniform interval.

This generator produces floating-point values of type T_Result drawn from a uniform interval [a, b) or (a, b], depending on the interval type specified via Interval_v: default case is CO ->[a,b). The interface mirrors std::uniform_real_distribution, and can be invoked with any engine adhering to the 'UniformRandomEngine' concept, which includes std uniform engines.

Supported result types: float, double Supported engine result widths: 32-bit and 64-bit unsigned integers.

Note

This distribution is subject to a small non-uniform bias; see UniformReal::operator() for details.

Definition at line 258 of file UniformReal.hpp.

Member Typedef Documentation

Base

template<std::floating_point T_Result, concepts::Interval T_Interval = interval::CO>

using alpaka::rand::distribution::UniformReal< T_Result, T_Interval >::Base = internal::UniformRealBase<T_Result, T_Interval>

Definition at line 269 of file UniformReal.hpp.

Constructor & Destructor Documentation

UniformReal() [1/2]

template<std::floating_point T_Result, concepts::Interval T_Interval = interval::CO>

alpaka::rand::distribution::UniformReal< T_Result, T_Interval >::UniformReal ( T_Interval interval = T_Interval{} )

inlineexplicitconstexpr

Definition at line 271 of file UniformReal.hpp.

UniformReal() [2/2]

template<std::floating_point T_Result, concepts::Interval T_Interval = interval::CO>

alpaka::rand::distribution::UniformReal< T_Result, T_Interval >::UniformReal ( T_Result min, T_Result max, T_Interval interval = T_Interval{} )

inlineconstexpr

Definition at line 275 of file UniformReal.hpp.

Member Function Documentation

checkValueConformity()

template<std::floating_point T_Result, concepts::Interval T_Interval = interval::CO>

template<std::integral T_Value>

consteval void alpaka::rand::distribution::UniformReal< T_Result, T_Interval >::checkValueConformity ( )

inlinestaticconsteval

Definition at line 263 of file UniformReal.hpp.

engineDispatch() [1/2]

template<std::floating_point T_Result, concepts::Interval T_Interval = interval::CO>

template<concepts::UniformVectorEngine T_Engine>

auto alpaka::rand::distribution::UniformReal< T_Result, T_Interval >::engineDispatch ( T_Engine & engine ) const -> T_Result

inlineconstexprprivate

Dispatch for vector engines (uniform bit generators that return a vector (e.g Philox4x32x10Vector) to enable efficient double precision uniform_real generation (reducing the number of invocations).

Definition at line 319 of file UniformReal.hpp.

engineDispatch() [2/2]

template<std::floating_point T_Result, concepts::Interval T_Interval = interval::CO>

template<concepts::UniformStdEngine T_Engine>

auto alpaka::rand::distribution::UniformReal< T_Result, T_Interval >::engineDispatch ( T_Engine & engine ) const -> T_Result

inlineconstexprprivate

Dispatch for std engines and alpaka abbreviations conforming to the std::uniform_random_bit_generator concept (e.g.

Philox4x32x10)

Definition at line 306 of file UniformReal.hpp.

operator()()

template<std::floating_point T_Result, concepts::Interval T_Interval = interval::CO>

template<concepts::UniformRandomEngine T_Engine>

auto alpaka::rand::distribution::UniformReal< T_Result, T_Interval >::operator() ( T_Engine & engine ) const -> T_Result

inlineconstexpr

Selects a value from a uniform distribution over the configured (min, max) interval, respecting the specified interval bounds.

**

Input: a random engine conforming to the UniformRandomEngine concept (currently accepts stdlib uniform engines and alpaka engines included in the alpaka::rand::engine namespace) Output: a floating-point value sampled from the configured distribution.

Note

: This distribution introduces a slight numerical bias due to floating-point rounding effects and the use of a 1 / MAX integer-to-floating-point conversion methods

See also

Goualard, F. (2020). Generating Random Floating-Point Numbers by Dividing Integers: A Case Study. https://doi.org/10.1007/978-3-030-50417-5_2 or Allen B. Downey Generating Pseudo-random Floating-Point Values https://allendowney.com/research/rand/. Additionally, using an interval with an open bound (OC,CO,OO) may introduce yet another small non-uniform bias –

scaleInterval().

Definition at line 296 of file UniformReal.hpp.

scaleInterval()

template<std::floating_point T_Result, concepts::Interval T_Interval = interval::CO>

auto alpaka::rand::distribution::UniformReal< T_Result, T_Interval >::scaleInterval ( T_Result const & normalizedVal ) const -> T_Result

inlineconstexprprivate

For open bounds, the normalized value in (0, 1) may (still) hit the lower or upper bound due to floating-point rounding (after scaling is applied).

To enforce adherence to the requested interval, the result is shifted to the next representable value using std::nextafter. WARNING: std::nextafter is not constexpr in cpp20 -> this might cause problems on some devices (regarding missing device annotations) and requires often unnecessary runtime evaluation – future maintainers should consider creating a constexpr adaptation of std::nextafter

Note

This introduces a(-/another) small non-uniform bias. The current implementation is inherently non-uniform due to integer-to-floating-point mapping

See also

https://doi.org/10.1007/978-3-030-50417-5_2

Definition at line 350 of file UniformReal.hpp.

---

The documentation for this struct was generated from the following file:

• /home/docs/checkouts/readthedocs.org/user_builds/alpaka3/checkouts/latest/include/alpaka/rand/distribution/UniformReal.hpp