-
Notifications
You must be signed in to change notification settings - Fork 0
Kokkos::MinMaxLoc
Specific implementation of ReducerConcept storing both the minimum and maximum values with corresponding indicies
Header File: Kokkos_Core.hpp
Usage:
MinMaxLoc<T,I,S>::value_type result;
parallel_reduce(N,Functor,MinMaxLoc<T,I,S>(result));.
template<class Scalar, class Space>
class MinMaxLoc{
public:
typedef MinMaxLoc reducer;
typedef MinMaxLocScalar<typename std::remove_cv<Scalar>::type,
typename std::remove_cv<Index>::type> value_type;
typedef Kokkos::View<value_type, Space> result_view_type;
KOKKOS_INLINE_FUNCTION
void join(value_type& dest, const value_type& src) const
KOKKOS_INLINE_FUNCTION
void join(volatile value_type& dest, const volatile value_type& src) const;
KOKKOS_INLINE_FUNCTION
void init( value_type& val) const;
KOKKOS_INLINE_FUNCTION
value_type& reference() const;
KOKKOS_INLINE_FUNCTION
result_view_type view() const;
KOKKOS_INLINE_FUNCTION
MinMaxLoc(value_type& value_);
KOKKOS_INLINE_FUNCTION
MinMaxLoc(const result_view_type& value_);
};-
reducer: The self type. -
value_type: The reduction scalar type (specialization of MinMaxLocScalar) -
result_view_type: AKokkos::Viewreferencing the reduction result
-
MinMaxLoc(value_type& result)Constructs a reducer which references a local variable as its result location.
-
MinMaxLoc(const result_view_type result)
Constructs a reducer which references a specific view as its result location.
-
void join(value_type& dest, const value_type& src) const;
- Store minimum with location of
srcanddestintodest. - Store maximum with location of
srcanddestintodest.
- Store minimum with location of
-
void join(volatile value_type& dest, const volatile value_type& src) const;
- Store minimum with location of
srcanddestintodest. - Store maximum with location of
srcanddestintodest.
- Store minimum with location of
-
void init( value_type& val) const;
Initialize
val.min_valusing the Kokkos::reduction_identity::min() method. The default implementation setsval=<TYPE>_MAX.Initialize
val.max_valusing the Kokkos::reduction_identity::max() method. The default implementation setsval=<TYPE>_MIN.Initialize
val.min_locusing the Kokkos::reduction_identity::min() method. The default implementation setsval=<TYPE>_MAX.Initialize
val.max_locusing the Kokkos::reduction_identity::min() method. The default implementation setsval=<TYPE>_MAX. -
value_type& reference() const;
Returns a reference to the result provided in class constructor.
-
result_view_type view() const;
Returns a view of the result place provided in class constructor.
-
MinMaxLoc<T,I,S>::value_typeis Specialization of MinMaxLocScalar on non-constTand non-constI -
MinMaxLoc<T,I,S>::result_view_typeisKokkos::View<T,S,Kokkos::MemoryTraits<Kokkos::Unmanaged>>. Note that the S (memory space) must be the same as the space where the result resides. - Requires:
Scalarhasoperator =,operator <andoperator >defined.Kokkos::reduction_identity<Scalar>::min()andKokkos::reduction_identity<Scalar>::max()are a valid expressions. - Requires:
Indexhasoperator =defined.Kokkos::reduction_identity<Scalar>::min()is a valid expressions. - In order to use MinMaxLoc with a custom type of either
ScalarorIndex, a template specialization of Kokkos::reduction_identity must be defined. See Using Built-in Reducers with Custom Scalar Types for details
Home:
- Introduction
- Machine Model
- Programming Model
- Compiling
- Initialization
- View
- Parallel Dispatch
- Hierarchical Parallelism
- Custom Reductions
- Atomic Operations
- Subviews
- Interoperability
- Kokkos and Virtual Functions
- Initialization and Finalization
- View
- Data Parallelism
- Execution Policies
- Spaces
- Task Parallelism
- Utilities
- STL Compatibility
- Numerics
- Detection Idiom