stapl::runtime Namespace Reference

STAPL Runtime System implementation classes and functions. More...

Namespaces
	this_context
	STAPL Runtime System RMI execution context management.

Classes
class	active_handle_response
	Function object to call the function operator of a distributed object with the given value. More...
class	aggregator
	Aggregator for point-to-point requests. More...
class	all_locations_rpc_aggregator
	Aggregator for RPC requests to all locations of a gang. More...
class	all_locs_construct_request
	Request to construct an object on all locally managed locations. More...
class	allgather_object
	Performs an allgather over all locations of the current gang. More...
class	allreduce_object
	Performs an allreduce over all locations of the current gang. More...
class	any
	any can hold an object of any type. The object can be retrieved only by casting to the correct type. More...
class	any_range
	Type erased range of const objects with associated size information. More...
struct	arbitrary
	Arbitrary mapping function. More...
class	arg_storage
	Storage for an object of type T that is retrieved by lvalue reference. More...
class	array_mailbox
	Mailbox for multiple threads that are contiguously numbered. More...
class	async_result
	Storage for a result that can be asynchronously assigned and retrieved. More...
class	async_results
	Storage for an array of results that can be asynchronously assigned and retrieved. More...
class	async_rmi_request
	RMI request that discards the return value and supports combining. More...
class	async_rpc_request
	Remote Procedure Call (RPC) request that discards return values. More...
class	backoff_metadata
	Backoff mechanism metadata. More...
class	barrier
	Performs a barrier over the processes specified by the topology object. More...
class	barrier_object
	Performs a barrier over all locations of the current gang. More...
class	bcast_aggregator
	Aggregator for ordered and unordered broadcast requests. More...
class	bind_rmi_result
	Establishes a communication tunnel to a member function of a distributed object. More...
class	binomial_tree_iterator
	Iterator to generate range of children ids required from make_binomial_tree. More...
class	block_registry
	Implements a registry that allows registration of ranges (blocks) of keys to be used by an entity that has a unique id. More...
struct	blocked
	Block mapping function. More...
class	bool_mutex
	Bool based mutex. More...
class	broadcast_object
	Performs a broadcast over all locations of the current gang. More...
class	buffer_iarchive
	Archive to unpack an object from a stream. More...
class	buffer_oarchive
	Archive to pack an object to a stream. More...
class	c_string
	Wraps a C string into a class. More...
class	cache_line_aligned_allocator
	An allocator that enforces that all allocated memory returned to the caller begins at the start of a cache line. More...
class	cache_line_aligned_storage
	Storage for a single instance of type T. Class has alignas directive to ensure cache line alignment. Used as storage type for cache_line_aligned_vector. More...
class	cache_line_aligned_vector
	Basic vector class based on std::vector which enforces cache line alignment. More...
class	call_promise_set_value
	Calls promise<rmi_handle::reference>::set_value() from location 0. More...
class	centralized_fuzzy_barrier
	Centralized fuzzy barrier. More...
struct	cloner
	Copy constructs a new T object on heap allocated memory. More...
class	collective
	Handle for distributed memory collective operations. More...
class	combinable_rmi_request
	Encapsulates an RMI request that supports combining for subsequent execution via the function operator. More...
class	combined_request_size
	Information about the size of combined requests. More...
class	common_gang_md
	Metadata that can be shared between multiple gangs. More...
class	communicator
	This class assists with sending messages between processes. More...
class	comparable_proxy
	Proxy object that allows type-erased objects or function pointers to be compared between them. More...
struct	comparator
	Function object to test if two objects are equal. More...
class	concurrency
	Initializes and controls threads for shared-memory concurrency. More...
struct	const_member_pointer
	Returns a member function pointer of type R (T::*)(Arg...) const. More...
class	construct_request
	Request to construct an object in a new gang over specific locations of an existing gang. More...
class	context
	Execution context of a task or request. More...
struct	context_id
	Id for an execution context. It is associated with context object. More...
class	continuation_request
	Calls function f with a Future created with t.... More...
class	copied_arg_storage
	Storage for a moved object that is retrieved by copy or by const reference. More...
class	copied_ref_storage
	Storage for a moved object that is retrieved by non-const reference. More...
struct	copyable
	Tag type for objects that for which is_copyable is std::true_type. More...
struct	create_object
	Calls the constructor of T with Args. More...
struct	create_shared_ptr
	Creates a new shared_ptr from the given object. More...
class	ctx_impl
	Wrapper to store a context and allow queueing it into a boost::intrusive::list. More...
struct	data_race_helper
	Helper class for detecting possible data races in a class static or stand-alone function. More...
struct	deferred_t
	Tag type to defer an operation. More...
class	delete_object_request
	Request to delete a distributed object. More...
class	distributor
	Gang id and metadata distributor. More...
class	exception
	Runtime base exception object. More...
class	executor_intermediate
	Intermediate class to allow adding entries with scheduling information. More...
class	executor_rmi_request
	Request for executing the given function on the destination location. More...
struct	external_caller
	Function object that calls a function passed from external_call(). More...
class	fence_md
	Fence metadata for intergang communication when a gang_md has no representative. More...
struct	fence_md_delete
	Destroys fence_md objects while sending the metadata to the correct process. More...
class	fence_section
	Enters a section of the code that is related to fence. More...
struct	fixed
	Fixed mapping function. More...
class	full_location
	Describes a location as a tuple of a gang id and a location id. More...
class	future_base
	Provides a common base for an asynchronously returned object. More...
class	futures_base
	Provides a common base for asynchronously returned arrays of objects. More...
class	gang_description
	Contains information about the locations of a gang. More...
class	gang_md
	Gang metadata that is shared by the locations that are on the same process. More...
class	gang_md_registry
	Gang metadata registry. More...
class	gang_retriever
	Retrieves the gang_md& associated with the gang id. More...
class	gang_switcher
	Switches to and from the given gang. More...
class	global_wf
	Function object to create a new distributed or mixed-mode environment. More...
class	handle_response
	Function object to assign an object to a distributed object via the Handle::set_value() function. More...
struct	has_define_type_impl
	Detects if T has a function T::define_type(stapl::typer&). More...
class	header
	Provides the enum and the different headers for communication. More...
struct	heap_tracker_init
	Nifty counter idiom to initialize the heap tracker and keep it alive until the last object file that uses it is finalized. More...
class	hierarchical_mutex
	Hierarchical synchronization primitive that does not emit memory fences. More...
struct	identity
	Identity mapping function. More...
struct	identity_member_pointer
	Returns a member function pointer of type R (T::*)(Arg...). More...
class	indexed_handle_response
	Function object to assign an object to a distributed object that accepts multiple objects via the Handle::set_value() function. More...
class	indexed_response
	Function object to assign an object to a known address that expects multiple objects. More...
class	instrument
	Provides instrumentation capabilities for runtime functions. More...
class	intrusive_mpsc_queue
	An intrusive lock-free MPSC (multiple producer, single consumer) queue. More...
class	intrusive_mpsc_queue_hook
	Hook for intrusive_mpsc_queue. More...
struct	is_appropriate_handle
	Returns std::true_type if Handle can be used to call PMF, otherwise std::false_type. More...
struct	is_contiguous_iterator
	If T is an iterator to a container that its elements are in contiguous space it is equivalent to std::true_type, for any other iterator it is std::false_type. More...
struct	is_copyable
	Typedefs member type to std::true_type if T is a copyable type, otherwise it typedefs it to std::false_type. More...
struct	is_known_polymorphic
	Returns if a type T has polymorphic_callable defined. More...
struct	is_movable
	Typedefs member type to std::true_type if T is a movable type, otherwise it typedefs it to std::false_type. More...
struct	is_reference_wrapper
	Returns if T is a std::reference_wrapper<> or boost::reference_wrapper. More...
struct	is_reference_wrapper_impl
	Returns if T is a std::reference_wrapper<> or boost::reference_wrapper. More...
struct	is_shared_ptr
	Returns if T is a std::shared_ptr<> or boost::shared_ptr. More...
struct	is_shared_ptr_impl
	Returns if T is a std::shared_ptr or boost::shared_ptr. More...
class	location_md
	Location metadata. More...
class	location_rpc_aggregator
	Aggregator for RPC requests to a single location. More...
class	location_rpc_executor
	Executes RPC requests for a specific location, preserving the requests that could not be executed for a later invocation. More...
class	location_rpc_request
	Encapsulates an RPC request directed to a location for subsequent execution via the function operator. More...
class	logical_clock
	Implements a logical clock for managing epochs. More...
class	mailbox
	Mailbox for multiple threads with arbitrary numbering. More...
struct	make_arg_storage
	Argument storage type creation metafunction. More...
struct	make_arguments< F, T... >
	Metafuction for creating a tuple of arg_storage_t to store the objects in list T. More...
class	malloc_allocator
	An allocator that uses std::malloc() directly. More...
class	managed_locations_info
	Container that associates parent gang to child gang location ids only for locally managed locations. More...
class	managed_locations_rpc_aggregator
	Aggregator for RPC requests to all locations in shared memory. More...
class	meets_requirements_oarchive
	Archive to determine if the object meets the packing requirements. More...
class	memory_allocator
	Provides interfaces to do pool allocation for objects. More...
class	message
	Communication buffer. More...
struct	message_delete
	Deleter for message objects. More...
class	message_handle
	This class is used to keep track of the handles associated with a message. More...
class	message_shared_ptr
	Intrusive message shared pointer. More...
class	message_slist
	Intrusive message list. More...
struct	movable
	Tag type for objects that for which is_movable is std::true_type. More...
class	moved_cref_storage
	Storage for a moved object that is retrieved by const reference. More...
class	moved_object_storage
	Storage for an object that is moved and retrieved either by copy or by rvalue reference. More...
class	moved_ref_storage
	Storage for a moved object that is retrieved by reference. More...
class	mpe_profiler
	Notifies MPE of a new instrumented section. More...
class	mpi_communicator
	Implements a communicator based on MPI-2. More...
class	nc_async_rmi_request
	RMI request that discards the return value. More...
class	nested_wf
	Function object to create a new threads-only environment. More...
struct	no_context_t
	Tag type to skip any operations related to a context. More...
struct	no_implicit_flush_t
	Tag type to mark that a flush is not implicit. More...
class	non_const_arg_storage
	Storage for a non-const, non-basic, non-distributed object that is retrieved by non-const lvalue reference to non-basic object. More...
class	non_const_ptr_storage
	Storage for a pointer to non-const, non-basic, non-distributed object that is retrieved by pointer to a non-const, non-basic object. More...
class	object_storage
	Storage for objects used in communication over shared memory. More...
class	object_virtual_address
	Distributed object virtual address. More...
struct	p_object_ptr_cast_impl
	Casts the void pointer to pointer to T. More...
class	p_object_ref_storage
	Storage for an lvalue reference to a distributed object. More...
class	packed_handle
	Packs rmi_handle::reference and rmi_handle::const_reference for use in requests. More...
class	packed_handle_epoch
	Packs rmi_handle::reference and rmi_handle::const_reference and the epoch for use in requests. More...
class	packed_object_storage
	Storage for object of type T used in communication. More...
struct	packed_value
	Handles unpacking objects stored in an arg_storage. More...
class	papi_clock
	Represents a clock based on PAPI_get_real_usec(). More...
class	papi_cycle_clock
	Represents a clock based on PAPI_get_real_cyc() and MHz. More...
class	poly_destroy
	Calls typer_traits::destroy() for polymorphic types. More...
class	poly_meets_requirements
	Calls the requested function from typer_traits for polymorphic types. More...
class	poly_pack
	Calls typer_traits::pack() for polymorphic types. More...
class	poly_packed_size
	Calls typer_traits::packed_size() for polymorphic types. More...
class	poly_unpack
	Calls for typer_traits::unpack() for polymorphic types. More...
class	pool
	Pool of memory chunks of a specific size. More...
struct	primitive_traits
	Traits for primitives. More...
struct	read_lock_guard
	Lock guard for read lock ownership. More...
struct	read_lock_t
	Tag type for read lock ownership. More...
class	ready_future
	future_base implementation for ready values. More...
class	ready_futures
	futures_base implementation for ready values. More...
class	reduce
	Distributed memory collective reduce. More...
class	reduce_rmi_handle
	Handle to wait for values from reduce_rmi(). More...
class	reduction
	Shared memory reduction using atomic counter. Last thread through does serial reduction and returns true, signaling value is ready for consumption. More...
class	ref_counted
	Provides a reference counting mechanism for objects that derive from it. More...
class	ref_counted_range
	Provides a range-based interface for ref_counted derived objects. More...
class	ref_counted_wrapper
	Wrapper over T to make it conformable with ref_counted. More...
class	request_queue
	Request queue. More...
class	response
	Function object to assign an object to a known address. More...
class	response_aggregator
	Aggregator for responses to requests. More...
class	restore_request
	Request for restoring an SPMD section. More...
class	rmi_delegate
	Delegate for an RMI request. More...
class	rmi_executor
	Executes RMI requests for a specific context, preserving the requests that could not be executed for a later invocation. More...
class	rmi_handle_base
	Distributed object base handle. More...
class	rmi_handle_extended_base
	Extended distributed object handle information. More...
class	rmi_handle_info
	Distributed object handle information. More...
class	rmi_request
	Encapsulates an RMI request for subsequent execution via the function operator. More...
class	rpc_aggregator
	Aggregator for RPC requests to a process or a range of processes. More...
struct	rpc_executor
	Executes RPC requests for a process. More...
class	rpc_request
	Encapsulates an RPC request directed to a process for subsequent execution via the function operator. More...
class	runqueue
	Queue for scheduling requests. More...
struct	runqueue_gang_id_comparator
	Less than equal comparator for runqueue::impl and gang ids. More...
class	runtime_error
	Runtime error exception object. More...
class	rval_storage
	Storage for an object that is retrieved by rvalue reference. More...
class	shared_object_factory
	Factory for creating object shared between threads. More...
class	shared_runqueue
	Runqueue shared among locations of a gang that are on the same process. More...
class	single_location_construct_request
	Request to construct an object in a new single location gang. More...
class	size_oarchive
	Archive to calculate the size an object needs for packing. More...
class	spmd_registry
	Registry for SPMD-registered objects. More...
class	stack
	Thread stack information object. More...
struct	supports_stapl_packing
	Returns std::true_type if typer_traits specialization exists for T or T::define_type(stapl::typer&) is defined. More...
struct	supports_stapl_packing_impl
	Returns std::true_type if typer_traits specialization exists for T or T::define_type(stapl::typer&) is defined. More...
class	sync_rmi_request
	RMI request that calls the given callback with the return value of the function. More...
class	task_queue< R(Args...)>
	A lock-free MPSC (multiple producer, single consumer) queue of tasks to be executed. More...
class	thin_spin_mutex
	Spin-based synchronization primitive that does not emit memory fences. More...
struct	thread_info_t
	Thread information. More...
class	timer
	Provides a timer mechanism. More...
class	tl_pools
	Thread local memory chunk pools. More...
class	topology
	Describes the topology of the processes of a gang. More...
struct	transport_qualifier
	Tags T with movable, copyable, or leaves it T. More...
class	try_rmi_request
	RMI request that will not execute if the distributed object has been destroyed and that discards the return value. More...
class	tunnel_aggregator
	RMI request aggregator for a single destination location. More...
struct	uint_rng
	Returns a suitable random number generator for the given unsigned integral type size. More...
class	value_handle
	Handle for receiving one result. More...
class	values_handle
	Handle for receiving multiple results. More...
struct	write_lock_t
	Tag type for write lock ownership. More...

Typedefs
typedef centralized_fuzzy_barrier	fuzzy_barrier
typedef std::uint32_t	location_id
	Location id type.
typedef std::uint32_t	gang_id
	Gang id type.
typedef std::uint16_t	nesting_level
	RMI nesting level type.
typedef std::uint8_t	magic_id
	Magic number that disambiguates between different deeply-nested RMIs.
typedef std::uint32_t	internal_affinity_tag
	Raw affinity tag type.
typedef std::uintptr_t	object_id
	Registered object id type. More...
typedef object_id	collective_id
	Collective operation id type.
using	message_ptr = std::unique_ptr< message, message_delete >
	message unique pointer. More...
using	message_hook_type = boost::intrusive::member_hook< message, message::slist_hook_type, &message::slist_hook >
	Intrusive message list hook. More...
using	message_raw_ptr_slist = typename boost::intrusive::make_slist< message, message_hook_type, boost::intrusive::cache_last< true >, boost::intrusive::constant_time_size< false > >::type
	Intrusive message list for message pointers. More...
template<typename T , typename R >
using	arg_storage_t = typename make_arg_storage< T, R >::type
	Argument storage type. More...
template<typename F , typename... T>
using	arguments_t = typename make_arguments< F, T... >::type
	A tuple of arg_storage_t to store the objects in list T.
template<typename T , typename BinaryOperation >
using	binary_operation_result_t = decltype(std::declval< BinaryOperation >()(std::declval< T >(), std::declval< T >()))
	Returns the result of binary operation BinaryOperation when passed T.
using	packed_handle_type = packed_handle_epoch
	Default handle type for requests.
template<std::size_t Len>
using	aligned_storage_t = typename std::aligned_storage< Len, STAPL_RUNTIME_DEFAULT_ALIGNMENT >::type
	Provides the member typedef type, which is a POD type suitable for use as uninitialized storage for any object whose size is at most Len and whose alignment requirement is a divisor of STAPL_RUNTIME_DEFAULT_ALIGNMENT.
using	list_hook_type = boost::intrusive::list_base_hook< boost::intrusive::link_mode< boost::intrusive::auto_unlink > >
	Intrusive list hook.
using	set_hook_type = boost::intrusive::set_base_hook< boost::intrusive::link_mode< boost::intrusive::auto_unlink > >
	Intrusive set hook.

Enumerations
enum	polling_policy_t { NO_THREAD = 0x0, SINGLE_THREAD, FIRST_THREAD, MASTER_THREAD }
	Communicator polling policy. More...

Functions
template<typename Handle >
context::epoch_type	find_target_epoch (context const &ctx, Handle const &h) noexcept
	Returns the epoch that a request from context ctx on handle h should execute in. More...
template<typename Handle >
gang_md *	find_target_gang_md (context &ctx, Handle const &h) noexcept
	Returns if possible the gang metadata associated with h, otherwise nullptr. More...
unsigned int	get_processes_per_node (MPI_Comm comm)
	Returns the number of processes per node, based on how many processes are co-scheduled with process 0. More...
std::ostream &	operator<< (std::ostream &, context const &)
constexpr bool	operator== (context_id const &x, context_id const &y) noexcept
constexpr bool	operator!= (context_id const &x, context_id const &y) noexcept
bool	operator< (context_id const &x, context_id const &y) noexcept
void	assert_fail (const char *s, const char *file, unsigned int line, const char *function)
void	assert_fail (const char *s, const char *function)
void	warning (const char *s, const char *function)
int	get_debug_level (void) noexcept
	Returns the debug level of the runtime.
executor_base &	get_executor (const gang_id gid)
	Returns the executor associated with the location of gang with id gid from the stack. More...
std::ostream &	operator<< (std::ostream &os, full_location const &l)
std::ostream &	operator<< (std::ostream &, gang_description const &)
constexpr bool	operator!= (gang_md const &x, gang_md const &y) noexcept
TauGroup_t const &	get_tau_group (void)
	Returns a TAU group for the runtime. More...
constexpr bool	operator!= (location_md const &x, location_md const &y) noexcept
message_ptr	make_message_ptr (message *const m)
	Returns a message_ptr from m. More...
template<typename ProcessRange , typename Callback , typename Function >
void	call_construct_all_impl (ProcessRange &&pids, Callback &&c, const gang_md::id gid, const context::epoch_type e, const gang_md::id new_gid, Function &&f)
	Calls construct_all_impl() on processes pids. More...
template<typename ProcessRange , typename Callback , typename T , typename Function >
void	call_construct_range_impl (ProcessRange &&pids, Callback &&c, const gang_md::id gid, location_range_wrapper< T > const &r, const context::epoch_type e, const gang_md::id new_gid, Function &&f)
	Calls construct_range_impl() on processes pids. More...
template<typename Callback , typename Function >
void	construct_all_impl (Callback &&c, gang_md &g, const context::epoch_type e, const gang_md::id new_gid, Function &&f)
	Creates a distributed object through f on all locations of gang g. More...
template<typename Callback , typename T , typename Function >
void	construct_range_impl (Callback &&c, gang_md &g, location_range_wrapper< T > const &r, const context::epoch_type e, const gang_md::id new_gid, Function &&f)
	Creates a distributed object through f on the subset locs of the locations of the gang associated with g. More...
template<typename Callback , typename T , typename Function >
void	fwd_construct_range_impl (Callback &&c, gang_md &g, location_range_wrapper< T > const &r, const context::epoch_type e, Function &&f)
	Target for forwarding the request to construct_range_impl(). More...
template<typename T , typename... Args>
auto	call_constructor (Args &&... args) -> decltype(std::bind(create_object< T >
	Creates a function object used to construct an instance T with args.
template<typename Callback , typename Function >
void	construct_single (Callback &&c, gang_md::id gid, gang_md *g, const location_md::id lid, const context::epoch_type e, Function &&f)
	Creates a distributed object through f in a new single location gang over location lid of gang g. More...
template<typename Callback , typename Function >
void	construct_neighbors (Callback &&c, gang_md &g, const context::epoch_type e, Function &&f)
	Creates a distributed object through f in a new gang over all the locations of the gang associated with ctx that are on shared memory. More...
template<typename Callback , typename Function >
void	construct_all (Callback &&c, const gang_md::id gid, gang_md *g, const context::epoch_type e, Function &&f)
	Creates a distributed object through f in a new gang over all locations of the gang of h. More...
template<typename Callback , typename T , typename Function >
void	construct_range (Callback &&c, gang_md::id gid, gang_md *g, location_range_wrapper< T > const &r, const context::epoch_type e, Function &&f)
	Creates a distributed object through f in a new gang over the range r of the locations of the gang associated with g. More...
template<typename F , typename Range , typename... T>
void	rpc (F &&f, Range &&pids, T &&... t)
	Remote Procedure Call (RPC) to process primitive. More...
template<typename Tuple , std::size_t... I, typename... T>
std::size_t	dynamic_size (index_sequence< I... >, T &&... t) noexcept
	Returns the dynamic size required for storing the objects t... in a arguments_t.
template<typename F , typename Tuple , std::size_t... I>
auto	invoke (F &&f, Tuple &args, void *const base, index_sequence< I... >) -> decltype(std::forward< F >(f)(std::get< I >(args).get(base)...))
	Calls f with the stored objects in args passed as arguments. More...
template<typename MemFun , typename Base , typename Tuple , std::size_t... I>
auto	invoke (MemFun const &pmf, Base &ref, Tuple &args, void *const base, index_sequence< I... >) -> decltype((ref.*pmf)(std::get< I >(args).get(base)...))
	Calls the member function pmf on ref with the stored objects in args passed as arguments. More...
template<typename MemFun , typename Base , typename Tuple , std::size_t... I>
auto	invoke (MemFun const &pmf, Base &ref, Tuple &args, void *const base, std::size_t &size, index_sequence< I... >) -> decltype((ref.*pmf)(std::get< I >(args).get(base, size)...))
	Calls the member function pmf on ref with the stored objects in args passed as arguments. More...
template<typename Tuple , std::size_t... I>
void	cleanup (Tuple &, index_sequence< I... >)
	Releases any resources associated with the stored objects if invoke() was not called. More...
header::epoch_type	get_message_epoch (message const &m) noexcept
	Returns the epoch m was sent from.
std::tuple< context_id, header::epoch_type, process_id >	get_message_info (message const &m, const location_id lid) noexcept
	Returns the context id, the epoch and the process id m was sent from as a std::tuple. More...
template<typename T , typename Handle >
T &	retrieve_object (Handle &&h, location_md &l) noexcept
	Returns a reference to the object that h refers to. More...
template<typename T , typename Handle >
T *	try_retrieve_object (Handle &&h, location_md *const l) noexcept
	Attempts to return a pointer to the object that h refers to. More...
template<typename T >
auto	define_type_cast (T const &t) noexcept -> decltype(define_type_provider< T >::apply(const_cast< T &>(t)))
template<typename T >
auto	define_type_cast (T volatile &t) noexcept -> decltype(define_type_provider< T >::apply(const_cast< T &>(t)))
template<typename T >
auto	define_type_cast (T const volatile &t) noexcept -> decltype(define_type_provider< T >::apply(const_cast< T &>(t)))
void	rmi_fence (context &)
	Ensures that all outstanding RMI requests have been completed. More...
void	rmi_fence (context &, std::function< bool(void)> f)
	Ensures that all outstanding RMI requests have been completed and the return value of f is true. More...
c_string	demangle (const char *)
	Returns a demangled version of the given mangled name.
c_string	demangle (c_string const &)
	Returns a demangled version of the given mangled name.
int	getpid (void) noexcept
	Returns the process id of the calling process.
std::size_t	get_total_physical_memory (void) noexcept
	Returns the total amount of physical memory in bytes.
std::size_t	get_used_physical_memory (void) noexcept
	Returns the amount of physical memory that is being used in bytes.
std::size_t	get_available_physical_memory (void) noexcept
	Returns the amount of physical memory available in bytes.
std::size_t	get_physical_memory_load (void) noexcept
	Returns the percentage of physical memory that is being used (0-100).
std::size_t	get_num_hardware_threads (void) noexcept
	Returns the number of hardware threads on the system.
constexpr std::size_t	aligned_size (const std::size_t n, const std::size_t alignment=STAPL_RUNTIME_DEFAULT_ALIGNMENT) noexcept
	Returns n adjusted with extra padding bytes to satisfy alignment of alignment.
constexpr std::false_type	define_type_exists (...)
	Returns std::false_type if T::define_type(stapl::typer&) does not exist or is not accessible.
template<typename T >
constexpr std::true_type	define_type_exists (T *t, decltype(runtime::define_type_cast(*t).define_type(std::declval< typer &>())) *=nullptr)
	Returns std::true_type if T::define_type(stapl::typer&) exists.
template<typename T , typename U >
T &	no_access_check_cast (U &u) noexcept
	Cast function to break access checking.
template<typename T , typename U >
T const &	no_access_check_cast (U const &u) noexcept
	Cast function to break access checking. More...
template<typename T >
T *	p_object_ptr_cast (void *p) noexcept
	Casts the void pointer to T, where T is a distributed object.
template<typename R , typename Tuple , typename F >
R	polymorphic_apply (F &&f, const type_id tid)
	Scans the tuple of types Tuple and calls f() for the type that matches the type id tid. More...
template<typename Range , typename T >
auto	find_index (Range const &r, T const &v) noexcept -> typename std::iterator_traits< decltype(std::begin(r))>::difference_type
	Finds index of the given element in a range. More...
template<typename Range >
bool	all_unique (Range const &r)
	Returns true if the range has unique elements, otherwise returns false.
template<typename UIntType >
UIntType	integral_ceil_log2 (UIntType i) noexcept
	Log2 algorithm for unsigned integral types. More...
template<typename UIntType , typename Size >
UIntType	set_bits (UIntType dst, UIntType v, Size msb, Size lsb) noexcept
	Sets the [msb, lsb] bits of dst to value v and returns the new dst.
template<typename UIntType , typename Size >
UIntType	read_bits (UIntType dst, Size msb, Size lsb) noexcept
	Reads the [msb, lsb] bits of dst.
template<typename T1 , typename T2 >
constexpr bool	operator== (cache_line_aligned_allocator< T1 > const &, cache_line_aligned_allocator< T2 > const &) noexcept
template<typename T1 , typename T2 >
constexpr bool	operator!= (cache_line_aligned_allocator< T1 > const &, cache_line_aligned_allocator< T2 > const &) noexcept
template<typename T >
std::vector< T >	make_vector (std::vector< T > v)
	Returns a std::vector<T>. More...
template<typename T , typename InputIterator >
std::vector< T >	make_vector (InputIterator first, InputIterator last)
	Returns a std::vector<T>. More...
template<typename T , typename U , typename = typename std::enable_if< !std::is_same< std::vector<T>, typename std::decay<U>::type >::value>::type>
std::vector< T >	make_vector (U &&u)
	Returns a std::vector<T>. More...
template<typename T1 , typename T2 >
constexpr bool	operator== (malloc_allocator< T1 > const &, malloc_allocator< T2 > const &) noexcept
template<typename T1 , typename T2 >
constexpr bool	operator!= (malloc_allocator< T1 > const &, malloc_allocator< T2 > const &) noexcept
template<typename T , typename Deleter >
void	intrusive_ptr_add_ref (ref_counted< T, Deleter > *p) noexcept
	Increases the reference count of the object pointed to by p. More...
template<typename T , typename Deleter >
void	intrusive_ptr_release (ref_counted< T, Deleter > *p) noexcept
	Decreases the reference count of the object pointed to by p and and deletes it if the count has reached 0. More...
template<typename T , typename... Args>
boost::intrusive_ptr< T >	make_ref_counted (Args &&... args)
	Constructs an object of type T that provides the interface and functionality of ref_counter using args as the parameter list for the constructor of T.
template<typename T >
std::vector< T >	split_string_to_vector (std::string const &s, const char *delims)
	Splits the given std::string to a vector of T, based on the delimiters. More...
template<typename IntType , typename Size >
std::tuple< IntType, IntType, decltype(boost::irange(std::declval< IntType >), std::declval< IntType >)))>	make_flat_tree (IntType myid, Size n)
	Returns the root, parent and children ids for the part of a flat tree that myid belongs to. More...
template<typename IntType , typename Size >
std::tuple< IntType, IntType, decltype(boost::irange(std::declval< IntType >), std::declval< IntType >)))>	make_binary_tree (IntType myid, Size n)
	Returns the root, parent and children ids for the part of a binary tree that myid belongs to. More...
template<typename IntType , typename Size >
std::tuple< IntType, IntType, boost::iterator_range< binomial_tree_iterator< IntType, Size > > >	make_binomial_tree (IntType myid, Size n)
	Returns the root, parent and children ids for the part of a binomial tree that myid belongs to. More...
template<typename... T>
void	unused (T &&...)
	Function to remove warnings on unused variables.
void	scheduling_point (context &)
	Yields the location to execute some requests.
template<typename Predicate >
bool	yield_until (context &ctx, Predicate &&pred)
	Yields until pred returns true. More...
template<typename Predicate >
bool	yield_until (const no_context_t, Predicate &&pred)
	Yields until pred returns true. More...
template<typename Predicate >
bool	yield_until (Predicate &&pred)
	Yields until pred returns true. More...
template<typename Predicate >
bool	yield_if_not (context &ctx, Predicate &&pred)
	Yields if pred does not return true. More...
template<typename Predicate >
bool	yield_if_not (const no_context_t, Predicate &&pred)
	Yields if pred does not return true. More...
template<typename Predicate >
bool	yield_if_not (Predicate &&pred)
	Yields if pred does not return true. More...
void	warning (const char *s, const char *file, unsigned int line, const char *function)
static void	fork_impl (const unsigned int parent_tid, thread_info_t const &parent_info, const unsigned int end_level, const unsigned int af_offset, std::function< void(unsigned int)> f)
	Recursive forking for implementing concurrency::fork(). More...
void	fence_md_update_sent (const gang_md::id init_gid, const gang_md::id gid, const nesting_level n, const unsigned int sent)
	Updates count for sent intergang requests. More...
void	fence_md_update_processed (const gang_md::id init_gid, const gang_md::id gid, const nesting_level n, const unsigned int processed)
	Updates count for processed intergang requests. More...
void	fence_md_update (const gang_md::id init_gid, const gang_md::id gid, const nesting_level n, const unsigned int sent, const unsigned int processed)
	Updates counts for sent and processed intergang requests. More...
void	release_gang_md (const gang_id gid)
	Releases the gang_md object associated with gid.
static std::pair< std::size_t, std::size_t >	allocation_sizes (std::size_t size) noexcept
	Returns the number of initial number of messages to allocate and the maximum number of messages allowed. More...
void	free_memory (void)
static	STAPL_RUNTIME_THREAD_LOCAL (tl_pools, pools) void free_memory(void)
	Thread-local pools.
std::ostream &	operator<< (std::ostream &os, object_virtual_address const &h)
std::ostream &	operator<< (std::ostream &os, rmi_handle_info const &h)
static void	notify_created (runqueue::impl &rq)
	Notifies that rq has been created.
stack &	get_thread_stack (void)
	Returns the thread-local stack.
static void	bcast_unordered (gang_md &, message_ptr, const process_id=invalid_process_id)
static void	deferred_process (gang_md &g, message_ptr m)
	Adds a message that was deferred because the gang metadata was not available.
static void	deferred_enqueue (const gang_md::id gid, message_ptr m)
	Defers the execution of the requests in m.
static	STAPL_RUNTIME_THREAD_LOCAL (backoff_metadata<>, backoff) static void bcast_unordered(gang_md &g
	Thread-local backoff metadata. More...
	if (m->is_forwarded()||pids.empty()||(pids.size()==1 &&pids[0]==id))
	if (id!=invalid_process_id &&id!=topo.root_id())
static void	bcast_unordered_init (gang_md &g, message_ptr m)
	Initializes a broadcast of m to all the locations of g.
static void	runqueue_add (message_ptr m)
	Adds m to its destination runqueue.
static void	process_incoming (const bool can_block)
	Process incoming messages.
template<typename Function >
void	rmi_fence_impl (context &ctx, Function &&f)
	RMI fence implementation. More...

Variables
constexpr read_lock_t	read_lock = { }
	Tag for read lock ownership.
constexpr write_lock_t	write_lock = { }
	Tag for write lock ownership.
constexpr gang_id	invalid_gang_id = std::numeric_limits<gang_id>::max()
	Invalid gang id.
constexpr object_id	invalid_object_id = std::numeric_limits<object_id>::max()
	Invalid object_id.
static struct stapl::runtime::heap_tracker_init	hpi
constexpr unsigned int	boost_serialization_flags
	Flags required for the boost::archive objects. More...
constexpr deferred_t	deferred = { }
	Tag to create gang_md objects with deferred id.
constexpr no_context_t	no_context = { }
	Tag to skip any operations related to a context.
constexpr no_implicit_flush_t	no_implicit_flush = { }
	Tag to mark that a flush is not implicit.
static std::unordered_map< collective::id, collective, boost::hash< collective::id > >	collective_registry
	collective object registry.
static std::mutex	collective_registry_mtx
	collective object registry mutex.
static mpi_communicator	mpi_comm
static std::thread::id	master_thread_id
	Master thread id.
static std::vector< unsigned int >	hierarchy
	Processing element hierarchy.
static std::vector< unsigned int >::size_type	max_num_levels = 0
	Number of available levels.
static unsigned int	max_num_threads = 0
	Maximum number of threads.
static unsigned int	process_affinity_width = 0
	Bits reserved for process affinity.
static std::vector< unsigned int >	affinity_width
	Bits reserved for each hierarchy level.
static hierarchical_mutex	process_mutex
	Master mutex.
static std::atomic< gang_md * >	global_gang_md {nullptr}
	Global gang metadata (gang with id 0)
static block_registry< gang_md::id, gang_md * >	registry
	Gang metadata registry.
static std::unordered_map< gang_md::id, fence_md * >	fence_md_registry
	Fence metadata registry.
static std::mutex	fence_md_registry_mtx
	Fence metadata registry mutex.
const auto	sizeof_location_md
const std::size_t	sizeof_message = sizeof(aligned_storage_t<sizeof(message)>)
	Aligned sizeof message.
static std::size_t	default_msg_capacity = 0
	Default message capacity.
static bool	enable_msg_pool = true
	Pool allocation enable flag.
static unsigned int	nppn = 0
	Number of processes per node.
static std::size_t	min_num_msgs = 0
	Minimum number of messages in the pool.
static double	max_mem_fraction = 0.0
	Maximum fraction of memory used for communication buffers.
static pool	msg_pool
	Message pool.
static std::new_handler	old_new_handler = nullptr
	std::new_handler before memory_allocator::initialize().
static bool	count_allocs = false
	true if run-time statistics for pool allocations enabled.
static unsigned int	aggregation_factor = 0
	Maximum number of RMI requests to aggregate before sending a message.
static bool	initialized = false
	true if the runtime is initialized, otherwise false.
static int	debug_level = 0
	Debug level.
static std::size_t	executor_window_size = 0
	Executor window size.
static std::size_t	executor_retire_chunk = 0
	Executor retire chunk.
static std::vector< unsigned int >	hierarchy_widths
	All the available hierarchy level widths.
static bool	consumed_1st_level = false
	true if the first level of parallelism is consumed
static static STAPL_RUNTIME_THREAD_LOCAL(stack, thread_stack) static gang_md unsigned int	owner_nlocs = 0
	Thread local stack. More...
static std::mutex	owner_mtx
static std::unordered_map< gang_md::id, message_slist >	deferred_requests
	Deferred requests per gang id.
static std::mutex	deferred_requests_mtx
	Deferred requests mutex.
static std::atomic< int >	overloaded {0}
	Positive number if the process is overloaded.
static polling_policy_t	polling_policy = FIRST_THREAD
	Polling policy.
static std::chrono::milliseconds	backoff_max_wait
	Maximum backoff wait in milliseconds.
static std::chrono::milliseconds	backoff_init_wait
	Initial backoff wait in milliseconds.
static message_ptr	m
static message_ptr const process_id	id
auto const &	pids = topo.children()
	else

Detailed Description

STAPL Runtime System implementation classes and functions.

Function Documentation

no_access_check_cast()

template<typename T , typename U >

T const& stapl::runtime::no_access_check_cast ( U const &  u )

noexcept

Cast function to break access checking.

polymorphic_apply()

template<typename R , typename Tuple , typename F >

R stapl::runtime::polymorphic_apply	(	F &&	f,
		const type_id	tid
	)

Scans the tuple of types Tuple and calls f() for the type that matches the type id tid.

Template Parameters

R	Return type of f.

allocation_sizes()

static std::pair<std::size_t, std::size_t> stapl::runtime::allocation_sizes ( std::size_t  size )

staticnoexcept

Returns the number of initial number of messages to allocate and the maximum number of messages allowed.

The calculation of the maximum number of messages allowed is based on a heuristic that uses the number of processes running on a node.

Variable Documentation

sizeof_location_md

const auto stapl::runtime::sizeof_location_md

Initial value:

=
  aligned_size(sizeof(location_md), runqueue::required_alignment())

owner_nlocs

static STAPL_RUNTIME_THREAD_LOCAL (stack, thread_stack) static gang_md unsigned int stapl::runtime::owner_nlocs = 0

static

Thread local stack.

Arbitrates the restore() primitive.

id

message_ptr const process_id stapl::runtime::id

Initial value:

{
  auto const& topo = g.get_topology()

else

stapl::runtime::else

Initial value:

{
    
    
    m->mark_forwarded()