NAME¶

std::counting_semaphore::acquire - std::counting_semaphore::acquire

Synopsis¶

void acquire(); (since C++20)

Atomically decrements the internal counter by 1 if it is greater than 0; otherwise
blocks until it is greater than 0 and can successfully decrement the internal
counter.

Preconditions¶

(none)

Parameters¶

(none)

Exceptions¶

May throw std::system_error.

Example¶

The example visualizes a concurrent work of several randomized threads when no more
than N (N is the semaphore desired value) of the thread-functions are active, while
the other might wait on the semaphore.

// Run this code

#include <array>
#include <chrono>
#include <cstddef>
#include <iomanip>
#include <iostream>
#include <mutex>
#include <new>
#include <random>
#include <semaphore>
#include <thread>
#include <vector>

using namespace std::literals;

constexpr std::size_t max_threads{10U}; // change and see the effect
constexpr std::ptrdiff_t max_sema_threads{3}; // {1} for binary semaphore
std::counting_semaphore semaphore{max_sema_threads};
constexpr auto time_tick{10ms};

unsigned rnd()
{
static std::uniform_int_distribution<unsigned> distribution{2U, 9U}; // [delays]
static std::random_device engine;
static std::mt19937 noise{engine()};
return distribution(noise);
}

class alignas(std::hardware_destructive_interference_size) Guide
{
inline static std::mutex cout_mutex;
inline static std::chrono::time_point<std::chrono::high_resolution_clock> started_at;
unsigned delay{rnd()}, occupy{rnd()}, wait_on_sema{};

public:
static void start_time() { started_at = std::chrono::high_resolution_clock::now(); }

void initial_delay() { std::this_thread::sleep_for(delay * time_tick); }

void occupy_sema()
{
wait_on_sema =
static_cast<unsigned>(std::chrono::duration_cast<std::chrono::milliseconds>(
std::chrono::high_resolution_clock::now() - started_at -
delay * time_tick).count() / time_tick.count());
std::this_thread::sleep_for(occupy * time_tick);
}

void visualize(unsigned id, unsigned x_scale = 2) const
{
auto cout_n = [=](auto str, unsigned n)
{
for (n *= x_scale; n-- > 0; std::cout << str)
;
};
std::lock_guard lk{cout_mutex};
std::cout << '#' << std::setw(2) << id << ' ';
cout_n("░", delay);
cout_n("▒", wait_on_sema);
cout_n("█", occupy);
std::cout << '\n';
}

static void show_info()
{
std::cout << "\nThreads: " << max_threads << ", Throughput: " << max_sema_threads
<< " │ Legend: initial delay ░░ │ wait state ▒▒ │ sema occupation ██ \n"
<< std::endl;
}
};

std::array<Guide, max_threads> guides;

void workerThread(unsigned id)
{
guides[id].initial_delay(); // emulate some work before sema acquisition
semaphore.acquire(); // wait until a free sema slot is available
guides[id].occupy_sema(); // emulate some work while sema is acquired
semaphore.release();
guides[id].visualize(id);
}

int main()
{
std::vector<std::jthread> threads;
threads.reserve(max_threads);

Guide::show_info();
Guide::start_time();

for (auto id{0U}; id != max_threads; ++id)
threads.push_back(std::jthread(workerThread, id));
}

Possible output:¶

Default case: max_threads{10U}, max_sema_threads{3}

Threads: 10, Throughput: 3 │ Legend: initial delay ░░ │ wait state ▒▒ │ sema occupation ██

# 1 ░░░░██████
# 2 ░░░░████████
# 5 ░░░░░░██████████
# 8 ░░░░░░░░░░░░████████████
# 9 ░░░░░░░░░░░░██████████████
# 7 ░░░░░░░░░░░░▒▒▒▒████████████████
# 4 ░░░░░░░░░░░░░░▒▒▒▒▒▒▒▒▒▒▒▒████████
# 6 ░░░░░░░░░░░░░░▒▒▒▒▒▒▒▒▒▒██████████████████
# 3 ░░░░░░░░░░░░░░▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒████████████
# 0 ░░░░░░░░░░░░░░░░░░▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒██████████████

──────────────────────────────────────────────────────────────────────────────────────────────────────────────
"Enough for everyone" case (no wait states!): max_threads{10U}, max_sema_threads{10}

Threads: 10, Throughput: 10 │ Legend: initial delay ░░ │ wait state ▒▒ │ sema occupation ██

# 4 ░░░░██████
# 5 ░░░░░░████
# 3 ░░░░██████████
# 1 ░░░░██████████
# 8 ░░░░░░░░████████████
# 6 ░░░░░░░░░░░░░░░░██████
# 7 ░░░░░░░░░░░░░░░░██████
# 9 ░░░░░░░░░░░░░░░░██████████
# 0 ░░░░░░░░░░░░██████████████████
# 2 ░░░░░░░░░░░░░░░░░░████████████

──────────────────────────────────────────────────────────────────────────────────────────────────────────────
Binary semaphore case: max_threads{10U}, max_sema_threads{1}

Threads: 10, Throughput: 1 │ Legend: initial delay ░░ │ wait state ▒▒ │ sema occupation ██

# 6 ░░░░████
# 5 ░░░░▒▒▒▒████
# 4 ░░░░░░░░░░▒▒██████████
# 7 ░░░░░░░░░░▒▒▒▒▒▒▒▒▒▒▒▒████████████████
# 2 ░░░░░░░░░░░░▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒██████
# 3 ░░░░░░░░░░░░░░▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒████████████████
# 0 ░░░░░░░░░░░░░░▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒████████████
# 1 ░░░░░░░░░░░░░░░░▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒████████
# 8 ░░░░░░░░░░░░░░░░▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒██████
# 9 ░░░░░░░░░░░░░░░░░░▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒██████████████

See also¶

release increments the internal counter and unblocks acquirers
(public member function)
try_acquire tries to decrement the internal counter without blocking
(public member function)
tries to decrement the internal counter, blocking for up to a
try_acquire_for duration time
(public member function)
tries to decrement the internal counter, blocking until a point in
try_acquire_until time
(public member function)