src/ex/thread_pool.cpp

100.0% Lines (140 / 140) 100.0% Functions (29 / 29)

[ ] prev/next file

thread_pool.cpp

f(x) Functions (29)

Function Calls Lines Blocks

Line	Hits	Source Code
1		//
2		// Copyright (c) 2025 Vinnie Falco (vinnie.falco@gmail.com)
3		// Copyright (c) 2026 Michael Vandeberg
4		//
5		// Distributed under the Boost Software License, Version 1.0. (See accompanying
6		// file LICENSE_1_0.txt or copy at http://www.boost.org/LICENSE_1_0.txt)
7		//
8		// Official repository: https://github.com/boostorg/capy
9		//
10
11		#include <boost/capy/ex/thread_pool.hpp>
12		#include <boost/capy/continuation.hpp>
13		#include <boost/capy/detail/thread_local_ptr.hpp>
14		#include <boost/capy/ex/frame_allocator.hpp>
15		#include <boost/capy/test/thread_name.hpp>
16		#include <algorithm>
17		#include <atomic>
18		#include <condition_variable>
19		#include <cstdio>
20		#include <mutex>
21		#include <thread>
22		#include <vector>
23
24		/*
25		Thread pool implementation using a shared work queue.
26
27		Work items are continuations linked via their intrusive next pointer,
28		stored in a single queue protected by a mutex. No per-post heap
29		allocation: the continuation is owned by the caller and linked
30		directly. Worker threads wait on a condition_variable until work
31		is available or stop is requested.
32
33		Threads are started lazily on first post() via std::call_once to avoid
34		spawning threads for pools that are constructed but never used. Each
35		thread is named with a configurable prefix plus index for debugger
36		visibility.
37
38		Work tracking: on_work_started/on_work_finished maintain the atomic
39		outstanding_work_ counter. on_work_started is lock-free; the worker
40		that drives the count to zero takes mutex_ and re-reads the count
41		before deciding to stop, so the count and the stop decision stay
42		consistent even if work is started in between. join() blocks until
43		this counter reaches zero, then signals workers to stop and joins
44		threads.
45
46		Two shutdown paths:
47		- join(): waits for outstanding work to drain, then stops workers.
48		- stop(): immediately signals workers to exit; queued work is abandoned.
49		- Destructor: stop() then join() (abandon + wait for threads).
50		*/
51
52		namespace boost {
53		namespace capy {
54
55		//------------------------------------------------------------------------------
56
57		class thread_pool::impl
58		{
59		// Identifies the pool owning the current worker thread, or
60		// nullptr if the calling thread is not a pool worker. Checked
61		// by dispatch() to decide between symmetric transfer (inline
62		// resume) and post.
63		static inline detail::thread_local_ptr<impl const> current_;
64
65		// Intrusive queue of continuations: the next link is stored in
66		// continuation::reserved (typed continuation* round-tripped through
67		// void*). No per-post allocation: the continuation is owned by the caller.
68		continuation* head_ = nullptr;
69		continuation* tail_ = nullptr;
70
71	18319x	void push(continuation* c) noexcept
72		{
73	18319x	c->reserved = nullptr;
74	18319x	if(tail_)
75	4179x	tail_->reserved = c;
76		else
77	14140x	head_ = c;
78	18319x	tail_ = c;
79	18319x	}
80
81	18725x	continuation* pop() noexcept
82		{
83	18725x	if(!head_)
84	406x	return nullptr;
85	18319x	continuation* c = head_;
86	18319x	head_ = static_cast<continuation*>(head_->reserved);
87	18319x	if(!head_)
88	14140x	tail_ = nullptr;
89	18319x	return c;
90		}
91
92	32554x	bool empty() const noexcept
93		{
94	32554x	return head_ == nullptr;
95		}
96
97		std::mutex mutex_;
98		std::condition_variable work_cv_;
99		std::condition_variable done_cv_;
100		std::vector<std::thread> threads_;
101		std::atomic<std::size_t> outstanding_work_{0};
102		bool stop_{false};
103		bool joined_{false};
104		std::size_t num_threads_;
105		char thread_name_prefix_[13]{}; // 12 chars max + null terminator
106		std::once_flag start_flag_;
107
108		public:
109	406x	~impl() = default;
110
111		bool
112	694x	running_in_this_thread() const noexcept
113		{
114	694x	return current_.get() == this;
115		}
116
117		// Destroy abandoned coroutine frames. Must be called
118		// before execution_context::shutdown()/destroy() so
119		// that suspended-frame destructors touching services
120		// (e.g. cancelling registrations) run while those
121		// services are still valid.
122		void
123	406x	drain_abandoned() noexcept
124		{
125	584x	while(auto* c = pop())
126		{
127	178x	auto h = c->h;
128	178x	if(h && h != std::noop_coroutine())
129	127x	h.destroy();
130	178x	}
131	406x	}
132
133	406x	impl(std::size_t num_threads, std::string_view thread_name_prefix)
134	406x	: num_threads_(num_threads)
135		{
136	406x	if(num_threads_ == 0)
137	8x	num_threads_ = std::max(
138	4x	std::thread::hardware_concurrency(), 1u);
139
140		// Truncate prefix to 12 chars, leaving room for up to 3-digit index.
141	406x	auto n = thread_name_prefix.copy(thread_name_prefix_, 12);
142	406x	thread_name_prefix_[n] = '\0';
143	406x	}
144
145		void
146	18319x	post(continuation& c)
147		{
148	18319x	ensure_started();
149		{
150	18319x	std::lock_guard<std::mutex> lock(mutex_);
151	18319x	push(&c);
152		// Under the lock so the pool cannot drain, join, and
153		// destroy the condition variable mid-signal.
154	18319x	work_cv_.notify_one();
155	18319x	}
156	18319x	}
157
158		void
159	693x	on_work_started() noexcept
160		{
161	693x	outstanding_work_.fetch_add(1, std::memory_order_acq_rel);
162	693x	}
163
164		void
165	693x	on_work_finished() noexcept
166		{
167	693x	if(outstanding_work_.fetch_sub(
168	693x	1, std::memory_order_acq_rel) == 1)
169		{
170		// fetch_sub's result can be stale: a concurrent
171		// on_work_started() may raise the count before we take the
172		// lock, so re-read it here rather than trust the decrement.
173	330x	std::lock_guard<std::mutex> lock(mutex_);
174	330x	if(outstanding_work_.load(
175	330x	std::memory_order_acquire) == 0 && joined_ && !stop_)
176		{
177	184x	stop_ = true;
178	184x	done_cv_.notify_all();
179	184x	work_cv_.notify_all();
180		}
181	330x	}
182	693x	}
183
184		void
185	678x	join() noexcept
186		{
187		{
188	678x	std::unique_lock<std::mutex> lock(mutex_);
189	678x	if(joined_)
190	272x	return;
191	406x	joined_ = true;
192
193	406x	if(outstanding_work_.load(
194	406x	std::memory_order_acquire) == 0)
195		{
196	174x	stop_ = true;
197	174x	work_cv_.notify_all();
198		}
199		else
200		{
201	232x	done_cv_.wait(lock, [this]{
202	417x	return stop_;
203		});
204		}
205	678x	}
206
207	866x	for(auto& t : threads_)
208	460x	if(t.joinable())
209	460x	t.join();
210		}
211
212		void
213	408x	stop() noexcept
214		{
215		{
216	408x	std::lock_guard<std::mutex> lock(mutex_);
217	408x	stop_ = true;
218	408x	}
219	408x	work_cv_.notify_all();
220	408x	done_cv_.notify_all();
221	408x	}
222
223		private:
224		void
225	18319x	ensure_started()
226		{
227	18319x	std::call_once(start_flag_, [this]{
228	356x	threads_.reserve(num_threads_);
229	816x	for(std::size_t i = 0; i < num_threads_; ++i)
230	920x	threads_.emplace_back([this, i]{ run(i); });
231	356x	});
232	18319x	}
233
234		void
235	460x	run(std::size_t index)
236		{
237		// Build name; set_current_thread_name truncates to platform limits.
238		char name[16];
239	460x	std::snprintf(name, sizeof(name), "%s%zu", thread_name_prefix_, index);
240	460x	set_current_thread_name(name);
241
242		// Mark this thread as a worker of this pool so dispatch()
243		// can symmetric-transfer when called from within pool work.
244		struct scoped_pool
245		{
246	460x	scoped_pool(impl const* p) noexcept { current_.set(p); }
247	460x	~scoped_pool() noexcept { current_.set(nullptr); }
248	460x	} guard(this);
249
250		for(;;)
251		{
252	18601x	continuation* c = nullptr;
253		{
254	18601x	std::unique_lock<std::mutex> lock(mutex_);
255	18601x	work_cv_.wait(lock, [this]{
256	46888x	return !empty() \|\|
257	46888x	stop_;
258		});
259	18601x	if(stop_)
260	920x	return;
261	18141x	c = pop();
262	18601x	}
263	18141x	if(c)
264	18141x	safe_resume(c->h);
265	18141x	}
266	460x	}
267		};
268
269		//------------------------------------------------------------------------------
270
271	406x	thread_pool::
272		~thread_pool()
273		{
274	406x	impl_->stop();
275	406x	impl_->join();
276	406x	impl_->drain_abandoned();
277	406x	shutdown();
278	406x	destroy();
279	406x	delete impl_;
280	406x	}
281
282	406x	thread_pool::
283	406x	thread_pool(std::size_t num_threads, std::string_view thread_name_prefix)
284	406x	: impl_(new impl(num_threads, thread_name_prefix))
285		{
286	406x	this->set_frame_allocator(std::allocator<void>{});
287	406x	}
288
289		void
290	272x	thread_pool::
291		join() noexcept
292		{
293	272x	impl_->join();
294	272x	}
295
296		void
297	2x	thread_pool::
298		stop() noexcept
299		{
300	2x	impl_->stop();
301	2x	}
302
303		//------------------------------------------------------------------------------
304
305		thread_pool::executor_type
306	11818x	thread_pool::
307		get_executor() const noexcept
308		{
309	11818x	return executor_type(
310	11818x	const_cast<thread_pool&>(*this));
311		}
312
313		void
314	693x	thread_pool::executor_type::
315		on_work_started() const noexcept
316		{
317	693x	pool_->impl_->on_work_started();
318	693x	}
319
320		void
321	693x	thread_pool::executor_type::
322		on_work_finished() const noexcept
323		{
324	693x	pool_->impl_->on_work_finished();
325	693x	}
326
327		void
328	17632x	thread_pool::executor_type::
329		post(continuation& c) const
330		{
331	17632x	pool_->impl_->post(c);
332	17632x	}
333
334		std::coroutine_handle<>
335	694x	thread_pool::executor_type::
336		dispatch(continuation& c) const
337		{
338	694x	if(pool_->impl_->running_in_this_thread())
339	7x	return c.h;
340	687x	pool_->impl_->post(c);
341	687x	return std::noop_coroutine();
342		}
343
344		} // capy
345		} // boost
346