src/ex/thread_pool.cpp

100.0% Lines (140/0/140) 100.0% List of functions (29/0/29)

[ ] 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	20136x	void push(continuation* c) noexcept
72		{
73	20136x	c->reserved = nullptr;
74	20136x	if(tail_)
75	1528x	tail_->reserved = c;
76		else
77	18608x	head_ = c;
78	20136x	tail_ = c;
79	20136x	}
80
81	20540x	continuation* pop() noexcept
82		{
83	20540x	if(!head_)
84	404x	return nullptr;
85	20136x	continuation* c = head_;
86	20136x	head_ = static_cast<continuation*>(head_->reserved);
87	20136x	if(!head_)
88	18608x	tail_ = nullptr;
89	20136x	return c;
90		}
91
92	39383x	bool empty() const noexcept
93		{
94	39383x	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	404x	~impl() = default;
110
111		bool
112	692x	running_in_this_thread() const noexcept
113		{
114	692x	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	404x	drain_abandoned() noexcept
124		{
125	614x	while(auto* c = pop())
126		{
127	210x	auto h = c->h;
128	210x	if(h && h != std::noop_coroutine())
129	159x	h.destroy();
130	210x	}
131	404x	}
132
133	404x	impl(std::size_t num_threads, std::string_view thread_name_prefix)
134	404x	: num_threads_(num_threads)
135		{
136	404x	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	404x	auto n = thread_name_prefix.copy(thread_name_prefix_, 12);
142	404x	thread_name_prefix_[n] = '\0';
143	404x	}
144
145		void
146	20136x	post(continuation& c)
147		{
148	20136x	ensure_started();
149		{
150	20136x	std::lock_guard<std::mutex> lock(mutex_);
151	20136x	push(&c);
152		// Under the lock so the pool cannot drain, join, and
153		// destroy the condition variable mid-signal.
154	20136x	work_cv_.notify_one();
155	20136x	}
156	20136x	}
157
158		void
159	691x	on_work_started() noexcept
160		{
161	691x	outstanding_work_.fetch_add(1, std::memory_order_acq_rel);
162	691x	}
163
164		void
165	691x	on_work_finished() noexcept
166		{
167	691x	if(outstanding_work_.fetch_sub(
168	691x	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	328x	std::lock_guard<std::mutex> lock(mutex_);
174	328x	if(outstanding_work_.load(
175	328x	std::memory_order_acquire) == 0 && joined_ && !stop_)
176		{
177	186x	stop_ = true;
178	186x	done_cv_.notify_all();
179	186x	work_cv_.notify_all();
180		}
181	328x	}
182	691x	}
183
184		void
185	674x	join() noexcept
186		{
187		{
188	674x	std::unique_lock<std::mutex> lock(mutex_);
189	674x	if(joined_)
190	270x	return;
191	404x	joined_ = true;
192
193	404x	if(outstanding_work_.load(
194	404x	std::memory_order_acquire) == 0)
195		{
196	166x	stop_ = true;
197	166x	work_cv_.notify_all();
198		}
199		else
200		{
201	238x	done_cv_.wait(lock, [this]{
202	425x	return stop_;
203		});
204		}
205	674x	}
206
207	862x	for(auto& t : threads_)
208	458x	if(t.joinable())
209	458x	t.join();
210		}
211
212		void
213	406x	stop() noexcept
214		{
215		{
216	406x	std::lock_guard<std::mutex> lock(mutex_);
217	406x	stop_ = true;
218	406x	}
219	406x	work_cv_.notify_all();
220	406x	done_cv_.notify_all();
221	406x	}
222
223		private:
224		void
225	20136x	ensure_started()
226		{
227	20136x	std::call_once(start_flag_, [this]{
228	354x	threads_.reserve(num_threads_);
229	812x	for(std::size_t i = 0; i < num_threads_; ++i)
230	916x	threads_.emplace_back([this, i]{ run(i); });
231	354x	});
232	20136x	}
233
234		void
235	458x	run(std::size_t index)
236		{
237		// Build name; set_current_thread_name truncates to platform limits.
238		char name[16];
239	458x	std::snprintf(name, sizeof(name), "%s%zu", thread_name_prefix_, index);
240	458x	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	458x	scoped_pool(impl const* p) noexcept { current_.set(p); }
247	458x	~scoped_pool() noexcept { current_.set(nullptr); }
248	458x	} guard(this);
249
250		for(;;)
251		{
252	20384x	continuation* c = nullptr;
253		{
254	20384x	std::unique_lock<std::mutex> lock(mutex_);
255	20384x	work_cv_.wait(lock, [this]{
256	58749x	return !empty() \|\|
257	58749x	stop_;
258		});
259	20384x	if(stop_)
260	916x	return;
261	19926x	c = pop();
262	20384x	}
263	19926x	if(c)
264	19926x	safe_resume(c->h);
265	19926x	}
266	458x	}
267		};
268
269		//------------------------------------------------------------------------------
270
271	404x	thread_pool::
272		~thread_pool()
273		{
274	404x	impl_->stop();
275	404x	impl_->join();
276	404x	impl_->drain_abandoned();
277	404x	shutdown();
278	404x	destroy();
279	404x	delete impl_;
280	404x	}
281
282	404x	thread_pool::
283	404x	thread_pool(std::size_t num_threads, std::string_view thread_name_prefix)
284	404x	: impl_(new impl(num_threads, thread_name_prefix))
285		{
286	404x	this->set_frame_allocator(std::allocator<void>{});
287	404x	}
288
289		void
290	270x	thread_pool::
291		join() noexcept
292		{
293	270x	impl_->join();
294	270x	}
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	11816x	thread_pool::
307		get_executor() const noexcept
308		{
309	11816x	return executor_type(
310	11816x	const_cast<thread_pool&>(*this));
311		}
312
313		void
314	691x	thread_pool::executor_type::
315		on_work_started() const noexcept
316		{
317	691x	pool_->impl_->on_work_started();
318	691x	}
319
320		void
321	691x	thread_pool::executor_type::
322		on_work_finished() const noexcept
323		{
324	691x	pool_->impl_->on_work_finished();
325	691x	}
326
327		void
328	19451x	thread_pool::executor_type::
329		post(continuation& c) const
330		{
331	19451x	pool_->impl_->post(c);
332	19451x	}
333
334		std::coroutine_handle<>
335	692x	thread_pool::executor_type::
336		dispatch(continuation& c) const
337		{
338	692x	if(pool_->impl_->running_in_this_thread())
339	7x	return c.h;
340	685x	pool_->impl_->post(c);
341	685x	return std::noop_coroutine();
342		}
343
344		} // capy
345		} // boost
346