Classes
class	AdaptBins
	A class that bins particles in energy bins and allows for adaptive runtime rebinning. More...
struct	DeviceViewTraits
	Traits class to extract host and device view types from a given ViewType. More...
struct	DeviceViewTraits< true, ViewType >
	Specialization of DeviceViewTraits for when DualView is used. More...
struct	DeviceViewTraits< false, ViewType >
	Specialization of DeviceViewTraits for when DualView is not used. More...
class	Histogram
	Template class providing adaptive particle histogram binning with support for Kokkos Views and DualViews. More...
struct	CoordinateSelector
	Example struct used to access the binning variable for each particle. More...
struct	ArrayReduction
	A templated structure for performing array-based reductions in parallel computations. More...
struct	ReductionVariantHelper
	Primary template for generating std::variant types containing ArrayReduction specializations. More...
struct	ReductionVariantHelper< SizeType, IndexType, std::integer_sequence< IndexType, Sizes... > >
	Specialized template that expands std::integer_sequence to create a variant of ArrayReduction types. More...
struct	HostArrayReduction
	Host-only array reduction structure for dynamic array sizes in Kokkos::parallel_reduce. More...

Typedefs
template<typename SizeType, typename IndexType>
using	ReductionVariant = typename ReductionVariantHelper<SizeType, IndexType, std::make_integer_sequence<IndexType, maxArrSize<IndexType>>>::type
	Type alias for a std::variant containing all possible ArrayReduction types up to maxArrSize.

Enumerations
enum class	HistoReductionMode { Standard , ParallelReduce , TeamBased , HostOnly }

Functions
template<typename bin_index_type>
HistoReductionMode	determineHistoReductionMode (HistoReductionMode modePreference, bin_index_type binCount)
	Determines the appropriate histogram reduction mode based on user preference, bin count, and execution environment.
template<typename ViewType>
void	computeFixSum (const ViewType &input_view, const ViewType &post_sum_view)
	Computes the post- or prefix-sum of the input view and stores the result in the ...-sum view.
template<typename ValueType, typename SizeType, typename HashType>
bool	viewIsSorted (const Kokkos::View< ValueType * > view, HashType indices, SizeType npart)
	Checks if the elements in a Kokkos::View are sorted in non-decreasing order.
template<typename SizeType, typename IndexType, IndexType N>
ReductionVariant< SizeType, IndexType >	createReductionObjectHelper (IndexType binCount)
	Recursive helper function to create ArrayReduction objects with compile-time size matching.
template<typename SizeType, typename IndexType>
ReductionVariant< SizeType, IndexType >	createReductionObject (IndexType binCount)
	Factory function to create ArrayReduction objects with runtime-specified size.
template<typename T, unsigned Dim>
T	vnorm (const VField_t< T, Dim > &field, int p=2)
	Computes the (p)-norm of a vector field (for debugging purpose).

Variables
template<typename IndexType>
constexpr IndexType	maxArrSize = 5
	Maximum allowed array size for compile-time array reduction types.
template<typename SizeType, typename IndexType>
IndexType	HostArrayReduction< SizeType, IndexType >::binCountStatic = 10
	Static member initialization for the array size variable.

Class Documentation

◆ ParticleBinning::DeviceViewTraits

struct ParticleBinning::DeviceViewTraits

template<bool UseDualView, typename ViewType>
struct ParticleBinning::DeviceViewTraits< UseDualView, ViewType >

Traits class to extract host and device view types from a given ViewType.

This primary template is specialized based on the UseDualView boolean flag.

Template Parameters

UseDualView	Flag indicating whether Kokkos::DualView is used.
ViewType	The original view type (e.g., Kokkos::View or Kokkos::DualView).

Definition at line 20 of file BinHisto.h.

Inheritance diagram for ParticleBinning::DeviceViewTraits< UseDualView, ViewType >:

Collaboration diagram for ParticleBinning::DeviceViewTraits< UseDualView, ViewType >:

◆ ParticleBinning::DeviceViewTraits< true, ViewType >

struct ParticleBinning::DeviceViewTraits< true, ViewType >

template<typename ViewType>
struct ParticleBinning::DeviceViewTraits< true, ViewType >

Specialization of DeviceViewTraits for when DualView is used.

Definition at line 26 of file BinHisto.h.

Inheritance diagram for ParticleBinning::DeviceViewTraits< true, ViewType >:

Collaboration diagram for ParticleBinning::DeviceViewTraits< true, ViewType >:

Class Members
typedef typename t_dev	d_type typename ViewType::t_dev
typedef typename t_host	h_type typename ViewType::t_host

◆ ParticleBinning::DeviceViewTraits< false, ViewType >

struct ParticleBinning::DeviceViewTraits< false, ViewType >

template<typename ViewType>
struct ParticleBinning::DeviceViewTraits< false, ViewType >

Specialization of DeviceViewTraits for when DualView is not used.

Definition at line 35 of file BinHisto.h.

Inheritance diagram for ParticleBinning::DeviceViewTraits< false, ViewType >:

Collaboration diagram for ParticleBinning::DeviceViewTraits< false, ViewType >:

Class Members
typedef ViewType	d_type ViewType
typedef ViewType	h_type ViewType

◆ ParticleBinning::ReductionVariantHelper

struct ParticleBinning::ReductionVariantHelper

template<typename SizeType, typename IndexType, typename Sequence>
struct ParticleBinning::ReductionVariantHelper< SizeType, IndexType, Sequence >

Primary template for generating std::variant types containing ArrayReduction specializations.

This template is not used directly but serves as the base for the specialized version. It works in conjunction with std::integer_sequence to generate a variant containing ArrayReduction types with sizes from 1 to maxArrSize.

This is necessary to pre-compile types for histogram reduction on device.

Template Parameters

SizeType	The type used for array elements in the reduction operations.
IndexType	The type used for array indexing.
Sequence	A std::integer_sequence type that will be expanded in the specialization.

Definition at line 97 of file ParallelReduceTools.h.

Inheritance diagram for ParticleBinning::ReductionVariantHelper< SizeType, IndexType, Sequence >:

Collaboration diagram for ParticleBinning::ReductionVariantHelper< SizeType, IndexType, Sequence >:

◆ ParticleBinning::ReductionVariantHelper< SizeType, IndexType, std::integer_sequence< IndexType, Sizes... > >

struct ParticleBinning::ReductionVariantHelper< SizeType, IndexType, std::integer_sequence< IndexType, Sizes... > >

template<typename SizeType, typename IndexType, IndexType... Sizes>
struct ParticleBinning::ReductionVariantHelper< SizeType, IndexType, std::integer_sequence< IndexType, Sizes... > >

Specialized template that expands std::integer_sequence to create a variant of ArrayReduction types.

This specialization takes a std::integer_sequence and expands it into a std::variant containing ArrayReduction types with consecutive sizes. Each size in the sequence is incremented by 1 to create ArrayReduction<SizeType, IndexType, Sizes + 1>.

Template Parameters

SizeType	The type used for array elements in the reduction operations.
IndexType	The type used for array indexing.
Sizes...	Parameter pack of integer values from the std::integer_sequence.

Definition at line 111 of file ParallelReduceTools.h.

Inheritance diagram for ParticleBinning::ReductionVariantHelper< SizeType, IndexType, std::integer_sequence< IndexType, Sizes... > >:

Collaboration diagram for ParticleBinning::ReductionVariantHelper< SizeType, IndexType, std::integer_sequence< IndexType, Sizes... > >:

Class Members
typedef variant< ArrayReduction< SizeType, IndexType, Sizes+1 >... >	type std::variant<ArrayReduction<SizeType, IndexType, Sizes + 1>...>

Typedef Documentation

◆ ReductionVariant

template<typename SizeType, typename IndexType>

using ParticleBinning::ReductionVariant = typename ReductionVariantHelper<SizeType, IndexType, std::make_integer_sequence<IndexType, maxArrSize<IndexType>>>::type

Type alias for a std::variant containing all possible ArrayReduction types up to maxArrSize.

This alias creates a variant that can hold ArrayReduction objects with sizes from 1 to maxArrSize. It uses ReductionVariantHelper with std::make_integer_sequence to generate all variants type at compile time.

Template Parameters

SizeType	The type used for array elements in the reduction operations.
IndexType	The type used for array indexing and size specification.

Definition at line 126 of file ParallelReduceTools.h.

Enumeration Type Documentation

◆ HistoReductionMode

enum class ParticleBinning::HistoReductionMode

strong

Enumerator
Standard
ParallelReduce
TeamBased
HostOnly

Definition at line 11 of file ParallelReduceTools.h.

Function Documentation

◆ computeFixSum()

template<typename ViewType>

void ParticleBinning::computeFixSum	(	const ViewType &	input_view,
		const ViewType &	post_sum_view )

Computes the post- or prefix-sum of the input view and stores the result in the ...-sum view.

Template Parameters

SizeType The type of the elements in the input and ...-sum views.

Parameters

input_view	The input view containing the elements to be summed.
post_sum_view	The output view where the ...-sum results will be stored. It must have a size of input_view.extent(0) + 1.
postSum	If true, the post-sum is computed; otherwise, the prefix-sum is computed.

Exceptions

`ippl::Comm->abort();` if the size of the post_sum_view is not equal to input_view.extent(0) + 1.

Definition at line 140 of file BinningTools.h.

Referenced by ParticleBinning::Histogram< size_type, bin_index_type, value_type, true >::initPostSum().

Here is the caller graph for this function:

◆ createReductionObject()

template<typename SizeType, typename IndexType>

ReductionVariant< SizeType, IndexType > ParticleBinning::createReductionObject ( IndexType binCount )

Factory function to create ArrayReduction objects with runtime-specified size.

This function provides an interface for creating ArrayReduction objects where the array size is determined at runtime. It internally uses compile-time template recursion to match the runtime binCount with the template specialization.

This allows using "runtime sized arrays" as reducer objects inside a Kokkos kernel on device.

Template Parameters

SizeType	The type used for array elements in the reduction operations.
IndexType	The type used for array indexing.

Parameters

binCount The desired number of bins for the reduction array (must be $\leq$ maxArrSize).

Returns: ReductionVariant containing the ArrayReduction object with the specified size.

Exceptions

`std::out_of_range` If binCount exceeds maxArrSize or is less than 1.

Example usage (also see implementation of AdaptBins::initLocalHisto()):

std::visit(
    [&](auto& reducer_arr) {
        executeInitLocalHistoReduction(reducer_arr);
    }, createReductionObject<size_type, bin_index_type>(binCount)
);

Definition at line 186 of file ParallelReduceTools.h.

References createReductionObjectHelper().

Here is the call graph for this function:

◆ createReductionObjectHelper()

template<typename SizeType, typename IndexType, IndexType N>

ReductionVariant< SizeType, IndexType > ParticleBinning::createReductionObjectHelper ( IndexType binCount )

Recursive helper function to create ArrayReduction objects with compile-time size matching.

This function uses compile-time recursion to find and instantiate the correct ArrayReduction type based on the requested binCount. It recursively increments the template parameter N until it matches binCount, then returns the appropriate ArrayReduction instance.

Template Parameters

SizeType	The type used for array elements in the reduction operations.
IndexType	The type used for array indexing.
N	Current array size being tested in the recursive search.

Parameters

binCount The desired number of bins for the reduction array.

Returns: ReductionVariant containing the ArrayReduction object with size matching binCount.

Exceptions

`std::out_of_range` If binCount exceeds maxArrSize.

Note: This function should not be called directly. Use createReductionObject() instead and use std::visit to access the reducer.

Definition at line 148 of file ParallelReduceTools.h.

References createReductionObjectHelper(), and maxArrSize.

Referenced by createReductionObject(), and createReductionObjectHelper().

Here is the call graph for this function:

Here is the caller graph for this function:

◆ determineHistoReductionMode()

template<typename bin_index_type>

HistoReductionMode ParticleBinning::determineHistoReductionMode	(	HistoReductionMode	modePreference,
		bin_index_type	binCount )

Determines the appropriate histogram reduction mode based on user preference, bin count, and execution environment.

This function selects the optimal histogram reduction method. If the code is compiled with a host execution space (e.g., Serial or OpenMP), it forces the reduction mode to HistoReductionMode::HostOnly, disregarding the user's preference. Otherwise, if the user preference is HistoReductionMode::Standard, it automatically chooses between ParallelReduce or TeamBased based on the binCount. If a specific preference is provided (not Standard), that preference is respected.

Parameters

modePreference

The user's preferred reduction mode.

Standard to select a mode based on bin count.
ParallelReduce, TeamBased, or HostOnly to force a specific mode.

Returns

HistoReductionMode The selected histogram reduction mode:

HostOnly if the default execution space is a host space.
ParallelReduce if binCount is within maxArrSize<bin_index_type>.
TeamBased if binCount exceeds maxArrSize<bin_index_type>.
Otherwise, respects the specified modePreference.

Note: If compiled for a host-only execution environment, the returned mode will always be HistoReductionMode::HostOnly, regardless of modePreference.

See also: HistoReductionMode

Definition at line 33 of file BinningTools.h.

References HostOnly, maxArrSize, ParallelReduce, Standard, and TeamBased.

◆ viewIsSorted()

template<typename ValueType, typename SizeType, typename HashType>

bool ParticleBinning::viewIsSorted	(	const Kokkos::View< ValueType * >	view,
		HashType	indices,
		SizeType	npart )

Checks if the elements in a Kokkos::View are sorted in non-decreasing order.

Template Parameters

ValueType	The type of the elements in the Kokkos::View so be checked.
SizeType	The type used for indexing and size.

Parameters

view	The Kokkos::View containing the elements to be checked.
indices	Argsort of view.
npart	The number of elements in the view to be checked.

Returns: true if the elements are sorted in non-decreasing order, false otherwise.

Note: This function does not leverage short circuiting.

Definition at line 182 of file BinningTools.h.

◆ vnorm()

template<typename T, unsigned Dim>

T ParticleBinning::vnorm	(	const VField_t< T, Dim > &	field,
		int	p = 2 )

Computes the (p)-norm of a vector field (for debugging purpose).

Parameters

field	The input vector field.
p	The order of the norm (default is 2 for Euclidean norm).

Returns: The computed (p)-norm of the vector field.

Definition at line 322 of file ParallelReduceTools.h.

Variable Documentation

◆ HostArrayReduction< SizeType, IndexType >::binCountStatic

template<typename SizeType, typename IndexType>

IndexType ParticleBinning::HostArrayReduction< SizeType, IndexType >::binCountStatic = 10

Static member initialization for the array size variable.

This line provides the definition and default initialization for the static member binCountStatic. The default value of 10 serves as a place holder and should be overridden based on the required histogram size.

Template Parameters

SizeType	The type used for array elements.
IndexType	The type used for indexing and size specification.

Note: This must be defined outside the class template for proper static member initialization.; The value should be set appropriately before creating any HostArrayReduction instances.; This is only used in the host version of the code.

Definition at line 310 of file ParallelReduceTools.h.

◆ maxArrSize

template<typename IndexType>

IndexType ParticleBinning::maxArrSize = 5

constexpr

Maximum allowed array size for compile-time array reduction types.

This constexpr variable defines the upper limit for statically-sized array reductions that are used in parallel histogram reductions. For small histograms, this is prefereable to pure atomics. However, for large values, compile time is increased significantly.

Template Parameters

IndexType The type used for array indexing and size specification.

Note: 5 is empirically chosen, since in real world applications the team-based reduction is usually enough.

Definition at line 81 of file ParallelReduceTools.h.

Referenced by createReductionObjectHelper(), and determineHistoReductionMode().

Classes

Typedefs

Enumerations

Functions

Variables

Class Documentation

◆ ParticleBinning::DeviceViewTraits

◆ ParticleBinning::DeviceViewTraits< true, ViewType >

◆ ParticleBinning::DeviceViewTraits< false, ViewType >

◆ ParticleBinning::ReductionVariantHelper

◆ ParticleBinning::ReductionVariantHelper< SizeType, IndexType, std::integer_sequence< IndexType, Sizes... > >

Typedef Documentation

◆ ReductionVariant

Enumeration Type Documentation

◆ HistoReductionMode

Function Documentation

◆ computeFixSum()

◆ createReductionObject()

◆ createReductionObjectHelper()

◆ determineHistoReductionMode()

◆ viewIsSorted()

◆ vnorm()

Variable Documentation

◆ HostArrayReduction< SizeType, IndexType >::binCountStatic

◆ maxArrSize