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	17834x	void push(continuation* c) noexcept
72		{
73	17834x	c->reserved = nullptr;
74	17834x	if(tail_)
75	4118x	tail_->reserved = c;
76		else
77	13716x	head_ = c;
78	17834x	tail_ = c;
79	17834x	}
80
81	18222x	continuation* pop() noexcept
82		{
83	18222x	if(!head_)
84	388x	return nullptr;
85	17834x	continuation* c = head_;
86	17834x	head_ = static_cast<continuation*>(head_->reserved);
87	17834x	if(!head_)
88	13716x	tail_ = nullptr;
89	17834x	return c;
90		}
91
92	31661x	bool empty() const noexcept
93		{
94	31661x	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	388x	~impl() = default;
110
111		bool
112	676x	running_in_this_thread() const noexcept
113		{
114	676x	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	388x	drain_abandoned() noexcept
124		{
125	573x	while(auto* c = pop())
126		{
127	185x	auto h = c->h;
128	185x	if(h && h != std::noop_coroutine())
129	134x	h.destroy();
130	185x	}
131	388x	}
132
133	388x	impl(std::size_t num_threads, std::string_view thread_name_prefix)
134	388x	: num_threads_(num_threads)
135		{
136	388x	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	388x	auto n = thread_name_prefix.copy(thread_name_prefix_, 12);
142	388x	thread_name_prefix_[n] = '\0';
143	388x	}
144
145		void
146	17834x	post(continuation& c)
147		{
148	17834x	ensure_started();
149		{
150	17834x	std::lock_guard<std::mutex> lock(mutex_);
151	17834x	push(&c);
152		// Under the lock so the pool cannot drain, join, and
153		// destroy the condition variable mid-signal.
154	17834x	work_cv_.notify_one();
155	17834x	}
156	17834x	}
157
158		void
159	675x	on_work_started() noexcept
160		{
161	675x	outstanding_work_.fetch_add(1, std::memory_order_acq_rel);
162	675x	}
163
164		void
165	675x	on_work_finished() noexcept
166		{
167	675x	if(outstanding_work_.fetch_sub(
168	675x	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	312x	std::lock_guard<std::mutex> lock(mutex_);
174	312x	if(outstanding_work_.load(
175	312x	std::memory_order_acquire) == 0 && joined_ && !stop_)
176		{
177	171x	stop_ = true;
178	171x	done_cv_.notify_all();
179	171x	work_cv_.notify_all();
180		}
181	312x	}
182	675x	}
183
184		void
185	642x	join() noexcept
186		{
187		{
188	642x	std::unique_lock<std::mutex> lock(mutex_);
189	642x	if(joined_)
190	254x	return;
191	388x	joined_ = true;
192
193	388x	if(outstanding_work_.load(
194	388x	std::memory_order_acquire) == 0)
195		{
196	164x	stop_ = true;
197	164x	work_cv_.notify_all();
198		}
199		else
200		{
201	224x	done_cv_.wait(lock, [this]{
202	396x	return stop_;
203		});
204		}
205	642x	}
206
207	830x	for(auto& t : threads_)
208	442x	if(t.joinable())
209	442x	t.join();
210		}
211
212		void
213	390x	stop() noexcept
214		{
215		{
216	390x	std::lock_guard<std::mutex> lock(mutex_);
217	390x	stop_ = true;
218	390x	}
219	390x	work_cv_.notify_all();
220	390x	done_cv_.notify_all();
221	390x	}
222
223		private:
224		void
225	17834x	ensure_started()
226		{
227	17834x	std::call_once(start_flag_, [this]{
228	338x	threads_.reserve(num_threads_);
229	780x	for(std::size_t i = 0; i < num_threads_; ++i)
230	884x	threads_.emplace_back([this, i]{ run(i); });
231	338x	});
232	17834x	}
233
234		void
235	442x	run(std::size_t index)
236		{
237		// Build name; set_current_thread_name truncates to platform limits.
238		char name[16];
239	442x	std::snprintf(name, sizeof(name), "%s%zu", thread_name_prefix_, index);
240	442x	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	442x	scoped_pool(impl const* p) noexcept { current_.set(p); }
247	442x	~scoped_pool() noexcept { current_.set(nullptr); }
248	442x	} guard(this);
249
250		for(;;)
251		{
252	18091x	continuation* c = nullptr;
253		{
254	18091x	std::unique_lock<std::mutex> lock(mutex_);
255	18091x	work_cv_.wait(lock, [this]{
256	45582x	return !empty() \|\|
257	45582x	stop_;
258		});
259	18091x	if(stop_)
260	884x	return;
261	17649x	c = pop();
262	18091x	}
263	17649x	if(c)
264	17649x	safe_resume(c->h);
265	17649x	}
266	442x	}
267		};
268
269		//------------------------------------------------------------------------------
270
271	388x	thread_pool::
272		~thread_pool()
273		{
274	388x	impl_->stop();
275	388x	impl_->join();
276	388x	impl_->drain_abandoned();
277	388x	shutdown();
278	388x	destroy();
279	388x	delete impl_;
280	388x	}
281
282	388x	thread_pool::
283	388x	thread_pool(std::size_t num_threads, std::string_view thread_name_prefix)
284	388x	: impl_(new impl(num_threads, thread_name_prefix))
285		{
286	388x	this->set_frame_allocator(std::allocator<void>{});
287	388x	}
288
289		void
290	254x	thread_pool::
291		join() noexcept
292		{
293	254x	impl_->join();
294	254x	}
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	11800x	thread_pool::
307		get_executor() const noexcept
308		{
309	11800x	return executor_type(
310	11800x	const_cast<thread_pool&>(*this));
311		}
312
313		void
314	675x	thread_pool::executor_type::
315		on_work_started() const noexcept
316		{
317	675x	pool_->impl_->on_work_started();
318	675x	}
319
320		void
321	675x	thread_pool::executor_type::
322		on_work_finished() const noexcept
323		{
324	675x	pool_->impl_->on_work_finished();
325	675x	}
326
327		void
328	17165x	thread_pool::executor_type::
329		post(continuation& c) const
330		{
331	17165x	pool_->impl_->post(c);
332	17165x	}
333
334		std::coroutine_handle<>
335	676x	thread_pool::executor_type::
336		dispatch(continuation& c) const
337		{
338	676x	if(pool_->impl_->running_in_this_thread())
339	7x	return c.h;
340	669x	pool_->impl_->post(c);
341	669x	return std::noop_coroutine();
342		}
343
344		} // capy
345		} // boost
346