Parallel Pixie Dust or PPD is the name of the thread library within libjmmcg. More...

Namespaces
namespace	generic_traits

namespace	lock

namespace	no_locking

namespace	NTUtils

namespace	pool

namespace	pool_traits
	A namespace to hold various traits relating to selecting the specific specialisation of thread_pool they would like.

namespace	private_

namespace	pthreads

Classes
struct	api_lock_traits

struct	api_lock_traits< generic_traits::api_type::no_api, sequential_mode >

struct	api_lock_traits< generic_traits::api_type::posix_pthreads, heavyweight_threading >
	The various lock-types for POSIX-threads derived, platforms. More...

struct	api_lock_traits< generic_traits::MS_Win32, heavyweight_threading >

struct	api_lock_traits< platform_api, sequential_mode >
	The various lock-types for sequential use, with no threading. More...

struct	api_threading_traits

struct	api_threading_traits< generic_traits::api_type::no_api, sequential_mode >

class	atomic_ctr_gen
	A general-purpose, atomic counter class, similar to the atomic integer types in C++11 ch 29. More...

class	atomic_ctr_gen< V, api_lock_traits< API, sequential_mode > >
	A specialisation for sequential operation to optimise performance. More...

class	atomic_ctr_opt
	An atomic counter class for integer types, similar to the atomic integer types in C++11 ch 29. More...

class	basic_statistics
	Some basic statistics collected about the operation of the wrapping thread_pool. More...

struct	cliques
	A modifier that enables dividing the thread_pool into a number of equal-sized cliques of pool_threads. More...

class	control_flow_graph
	A class to generate a Control Flow Graph (CFG), if the support has been coded into the joinable, nonjoinable, etc modifiers. More...

class	funky_queue
	An adaptor to add thread-safety assistance, specifically for queues. More...

struct	jthread
	Make sure the kernel-thread is always joined in the dtor. More...

struct	model_traits

class	no_control_flow_graph
	A dummy class to not generate a Control Flow Graph (CFG), and hopefully and hopefully all calls will be optimised out, so there should be no cost. More...

class	no_signalling
	Atomically count the amount of work that there is to do, and provide access to the lock on the containing collection. More...

struct	no_statistics
	Just stub statistics - to support the interface. No actual statistics are collected. More...

class	noop_atomic_ctr_base
	An atomic counter class that actually does nothing, doesn't count at all. More...

struct	null_thread_worker_context
	A handy utility class for the "thread_wrapper" class, that is supposed to do nothing. More...

struct	numa_cpu_traits

struct	pool_aspects
	The fundamental way to specify the type of thread_pool that is required. More...

class	priority_queue
	A thread-safety-assisted adaptor for a priority_queue. More...

class	queue
	An adaptor to add thread-safety assistance, specifically for queues. More...

class	safe_colln
	An adaptor for a container that attempts to add some thread-safety to assist in making thread-safe programs. More...

class	signalling
	A flag to atomically signal if the container contains work or not and also count the amount of work that there is to do. More...

class	thread
	This class is used to implement a class that wraps using a thread, it is an alternative to using a thread pool to manage your threads, in a C++ std::thread-style manner. More...

struct	thread_os_traits
	This class is used to encapsulate the OS-specific threading traits, mutexes, etc for the "thread_pool<...>" and other classes. More...

class	thread_params

class	thread_params< generic_traits::api_type::posix_pthreads >
	Wrap up the PThreads details regarding creating a thread, so that it is normalised for the rest of the library. More...

class	thread_params< generic_traits::MS_Win32 >

class	thread_pool

class	thread_pool< pool_traits::work_distribution_mode_t::one_thread_distributes<>, pool_traits::size_mode_t::fixed_size, PTT >
	This pool has a limited, specified size, and uses a master to distribute the work to the worker threads. More...

class	thread_pool< pool_traits::work_distribution_mode_t::one_thread_distributes<>, pool_traits::size_mode_t::infinite, PTT >
	This pool has an unlimited size, and uses a master to distribute the work to the worker threads. More...

class	thread_pool< pool_traits::work_distribution_mode_t::one_thread_distributes<>, pool_traits::size_mode_t::sequential, P >
	This compatibility-pool has an unlimited size, and uses a master to execute the work. More...

class	thread_pool< pool_traits::work_distribution_mode_t::one_thread_distributes<>, pool_traits::size_mode_t::tracks_to_max, PTT >
	This pool has a maximum specified size to which it will grow and reduce from, and uses a master to distribute the work to the worker threads. More...

class	thread_pool< pool_traits::work_distribution_mode_t::worker_threads_get_work< pool_traits::work_distribution_mode_t::queue_model_t::pool_owns_queue >, pool_traits::size_mode_t::fixed_size, PTT >
	This pool has a specified size, and the worker pool_threads steal work from a centrally-held signalled_work_queue. More...

class	thread_pool< pool_traits::work_distribution_mode_t::worker_threads_get_work< pool_traits::work_distribution_mode_t::queue_model_t::pool_owns_queue >, pool_traits::size_mode_t::sequential, P >
	This compatibility-pool has an unlimited size, and the "worker thread"s directly execute the work. More...

class	thread_pool< pool_traits::work_distribution_mode_t::worker_threads_get_work< pool_traits::work_distribution_mode_t::queue_model_t::pool_owns_queue >, pool_traits::size_mode_t::tracks_to_max, PTT >
	This pool has a maximum specified size to which it will grow and reduce from, and the worker pool_threads steal work from a centrally-held signalled_work_queue. More...

class	thread_pool< pool_traits::work_distribution_mode_t::worker_threads_get_work< pool_traits::work_distribution_mode_t::queue_model_t::thread_owns_queue< pool_traits::work_distribution_mode_t::queue_model_t::stealing_mode_t::random > >, pool_traits::size_mode_t::fixed_size, PTT >
	This pool has a specified size, and the worker pool_threads steal each have a signalled_work_queue. More...

class	wrapper
	This class is used to implement a class that wraps using a thread, it is an alternative to using a thread pool to manage your threads. More...

Typedefs
using	sequential_mode = model_traits< generic_traits::creation_cost::sequential_creation, generic_traits::destruction_cost::sequential_destruction, generic_traits::synchronisation_cost::sequential_sync >
	I.e. single-threaded, no locks. More...

using	heavyweight_threading = model_traits< generic_traits::creation_cost::expensive_to_create, generic_traits::destruction_cost::expensive_to_destroy, generic_traits::synchronisation_cost::expensive_to_synchronise >

using	lightweight_threading = model_traits< generic_traits::creation_cost::cheap_to_create, generic_traits::destruction_cost::cheap_to_destroy, generic_traits::synchronisation_cost::cheap_to_synchronise >
	E.g. x86 or SPARC. More...

template<class V , class LkT >
using	atomic_ctr = typename std::conditional< std::is_integral< V >::value, atomic_ctr_opt< V, LkT >, typename std::conditional< std::is_pointer< V >::value, atomic_ctr_opt< V, LkT >, atomic_ctr_gen< V, LkT > >::type >::type
	A general-purpose, atomic-wrapper class, similar to the atomic types in C++11 ch 29, with some important differences. More...

Functions
template<generic_traits::api_type API>
tostream &__fastcall FORCE_INLINE	operator<< (tostream &os, thread_params< API > const &p)

template<class V , class LkT >
tostream &FORCE_INLINE	operator<< (tostream &os, atomic_ctr< V, LkT > const &a)

char const *	SIG_to_string (int signal_number) noexcept(true)

std::ostream &	operator<< (std::ostream &os, siginfo_t const &si) noexcept(true)

void	abort_handler (int signal_number, siginfo_t info, void ) noexcept(true)

void	dump_bt_to_cerr ()

Detailed Description

Parallel Pixie Dust or PPD is the name of the thread library within libjmmcg.

Why is it called "ppd"? It stands for "Parallel Pixie Dust", a term coined at the ACCU by, if I recall correctly, Hubert Matthews in around 2001, whilst I was in conversation with him about thread libraries and their ability to shield a programmer from the issues of threading.

Typedef Documentation

◆ atomic_ctr

template<class V , class LkT >

using jmmcg::LIBJMMCG_VER_NAMESPACE::ppd::atomic_ctr = typedef typename std::conditional< std::is_integral<V>::value, atomic_ctr_opt<V, LkT>, typename std::conditional< std::is_pointer<V>::value, atomic_ctr_opt<V, LkT>, atomic_ctr_gen<V, LkT> >::type >::type

A general-purpose, atomic-wrapper class, similar to the atomic types in C++11 ch 29, with some important differences.

Unlike std::atomic: #Works for any type: integer-types & pointers (to anything) are atomic, others use a mutex, so aren't: guaranteed. #Has a uniform interface. #Also note that this class reduces to a simple V-type, if it is single threaded, so costs nothing. #Note that this class is a valid value-type.

See also: std::stomic

Definition at line 520 of file atomic_counter.hpp.

◆ heavyweight_threading

using jmmcg::LIBJMMCG_VER_NAMESPACE::ppd::heavyweight_threading = typedef model_traits< generic_traits::creation_cost::expensive_to_create, generic_traits::destruction_cost::expensive_to_destroy, generic_traits::synchronisation_cost::expensive_to_synchronise >

Definition at line 270 of file thread_params_traits.hpp.

◆ lightweight_threading

using jmmcg::LIBJMMCG_VER_NAMESPACE::ppd::lightweight_threading = typedef model_traits< generic_traits::creation_cost::cheap_to_create, generic_traits::destruction_cost::cheap_to_destroy, generic_traits::synchronisation_cost::cheap_to_synchronise >

E.g. x86 or SPARC.

Definition at line 275 of file thread_params_traits.hpp.

◆ sequential_mode

using jmmcg::LIBJMMCG_VER_NAMESPACE::ppd::sequential_mode = typedef model_traits< generic_traits::creation_cost::sequential_creation, generic_traits::destruction_cost::sequential_destruction, generic_traits::synchronisation_cost::sequential_sync >

I.e. single-threaded, no locks.

Definition at line 265 of file thread_params_traits.hpp.

Function Documentation

◆ abort_handler()

void jmmcg::LIBJMMCG_VER_NAMESPACE::ppd::abort_handler	(	int	signal_number,
		siginfo_t *	info,
		void *
	)

inlinenoexcept

Definition at line 167 of file thread_api_traits_impl.hpp.

◆ dump_bt_to_cerr()

void jmmcg::LIBJMMCG_VER_NAMESPACE::ppd::dump_bt_to_cerr ( )

inline

Definition at line 208 of file thread_api_traits_impl.hpp.

◆ operator<<() [1/3]

std::ostream& jmmcg::LIBJMMCG_VER_NAMESPACE::ppd::operator<<	(	std::ostream &	os,
		siginfo_t const &	si
	)

inlinenoexcept

Definition at line 156 of file thread_api_traits_impl.hpp.

◆ operator<<() [2/3]

template<class V , class LkT >

tostream& FORCE_INLINE jmmcg::LIBJMMCG_VER_NAMESPACE::ppd::operator<<	(	tostream &	os,
		atomic_ctr< V, LkT > const &	a
	)

inline

Todo:: Implement using the advice given in "Standard C++ IOStreams and Locales" by A.Langer & K.Kreft, page 170.

Definition at line 534 of file atomic_counter.hpp.

◆ operator<<() [3/3]

template<generic_traits::api_type API>

tostream& __fastcall FORCE_INLINE jmmcg::LIBJMMCG_VER_NAMESPACE::ppd::operator<<	(	tostream &	os,
		thread_params< API > const &	p
	)

inline

Definition at line 249 of file thread_params_traits.hpp.

◆ SIG_to_string()

char const* jmmcg::LIBJMMCG_VER_NAMESPACE::ppd::SIG_to_string ( int signal_number )

inlinenoexcept

Definition at line 125 of file thread_api_traits_impl.hpp.

Namespaces

Classes

Typedefs

Functions

Detailed Description

Typedef Documentation

◆ atomic_ctr

◆ heavyweight_threading

◆ lightweight_threading

◆ sequential_mode

Function Documentation

◆ abort_handler()

◆ dump_bt_to_cerr()

◆ operator<<() [1/3]

◆ operator<<() [2/3]

◆ operator<<() [3/3]

◆ SIG_to_string()