Charm++ Extensions

Classes and functions used to make Charm++ easier and safer to use. More...

Classes
struct	Parallel::charmxx::MainChareRegistrationConstructor
	Class to mark a constructor as a constructor for the main chare. More...
struct	Parallel::charmxx::RegistrationHelper
	The base class used for automatic registration of entry methods. More...
struct	Parallel::charmxx::RegisterCharmMessage< MessageType >
	Class used for automatic registration of Charm++ messages. More...
struct	Parallel::charmxx::RegisterParallelComponent< ParallelComponent >
	Derived class for registering parallel components. More...
struct	Parallel::charmxx::RegisterChare< Chare, CkIndex >
	Derived class for registering chares. More...
struct	Parallel::charmxx::RegisterSimpleAction< ParallelComponent, Action, Args >
	Derived class for registering simple actions. More...
struct	Parallel::charmxx::RegisterThreadedAction< ParallelComponent, Action, Args >
	Derived class for registering threaded actions. More...
struct	Parallel::charmxx::RegisterReceiveData< ParallelComponent, ReceiveTag, UsingMessages >
	Derived class for registering receive_data functions. More...
struct	Parallel::charmxx::RegisterReductionAction< ParallelComponent, Action, ReductionType >
	Derived class for registering reduction actions. More...
struct	Parallel::charmxx::RegisterGlobalCacheMutate< Metavariables, GlobalCacheTag, Function, Args >
	Derived class for registering GlobalCache::mutate. More...
struct	Parallel::charmxx::RegisterInvokeIterableAction< ParallelComponent, Action, PhaseIndex, DataBoxIndex >
	Derived class for registering the invoke_iterable_action entry method. This is a local entry method that is only used for tracing the code with Charm++'s Projections. More...
struct	Parallel::charmxx::RegisterReducerFunction< ReducerFunctions >
	Class used for registering custom reduction function. More...

Typedefs
using	Parallel::charmxx::ReducerFunctions = CkReductionMsg* (*)(int, CkReductionMsg**)
	The type of a function pointer to a Charm++ custom reduction function.

Functions
template<class... Args>
std::string	Parallel::charmxx::get_template_parameters_as_string ()
	Returns the template parameter as a std::string.
template<typename Derived>
bool	Parallel::charmxx::register_func_with_charm ()
	Function that adds a pointer to a specific derived class to the charm_register_list.
template<ReducerFunctions F>
bool	Parallel::charmxx::register_reducer_function ()
	Function that stores a function pointer to the custom reduction function to be registered later.

Detailed Description

Classes and functions used to make Charm++ easier and safer to use.

Function Documentation

get_template_parameters_as_string()

template<class... Args>

std::string Parallel::charmxx::get_template_parameters_as_string

(

)

Returns the template parameter as a std::string.

Uses the PRETTY_FUNCTION compiler intrinsic to extract the template parameter names in the same form that Charm++ uses to register entry methods. This is used by the generated Singleton, Array, Group and Nodegroup headers, as well as in CharmMain.tpp.

register_func_with_charm()

template<typename Derived>

bool Parallel::charmxx::register_func_with_charm

(

)

Function that adds a pointer to a specific derived class to the charm_register_list.

Used to initialize the registrar bool of derived classes of RegistrationHelper. When the function is invoked it appends the derived class to the charm_register_list.

Note

The reason for not using a std::vector is that this did not behave correctly when calling push_back. Specifically, the final vector was always size 1, even though multiple elements were pushed back. The reason for that behavior was never tracked down and so in the future it could be possible to use a std::vector.

register_reducer_function()

template<ReducerFunctions F>

bool Parallel::charmxx::register_reducer_function

(

)

Function that stores a function pointer to the custom reduction function to be registered later.

Used to initialize the registrar bool of RegisterReducerFunction. When invoked it adds the function F of type ReducerFunctions to the list charm_reducer_functions_list.

Note

The reason for not using a std::vector<ReducerFunctions> is that this did not behave correctly when calling push_back. Specifically, the final vector was always size 1, even though multiple elements were pushed back. The reason for that behavior was never tracked down and so in the future it could be possible to use a std::vector.