Namespaces
Variants
Views
Actions

std::execution::scheduler

From cppreference.com
< cpp‎ | execution
Revision as of 05:39, 12 October 2024 by Space Mission (Talk | contribs)

(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
 
 
 
Defined in header <execution>
template< class Sch >

concept scheduler =
    std::derived_from<
        typename std::remove_cvref_t<Sch>::scheduler_concept,
        scheduler_t> &&
    /*queryable*/<Sch> &&
    requires(Sch&& sch)
    {
        {
            std::execution::schedule(std::forward<Sch>(sch))
        } -> std::execution::sender;
        {
            auto(
                std::execution::get_completion_scheduler<
                    std::execution::set_value_t>(
                        std::execution::get_env(
                            std::execution::schedule(
                                std::forward<Sch>(sch)))))
        } -> std::same_as<std::remove_cvref_t<Sch>>;
    } &&
    std::equality_comparable<std::remove_cvref_t<Sch>> &&
    std::copy_constructible<std::remove_cvref_t<Sch>>;

};
(1) (since C++26)
Helper tag type
struct scheduler_t {};
(2) (since C++26)

The concept scheduler is modeled by types that are schedulers, that is, lightweight handlers to execution resources such as thread pools that work with the C++ execution library.

[edit] Semantic requirements

Given a scheduler of type Sch and execution environment of type Env such that sender_in<schedule_result_t<Sch>, Env> is satisfied, then /*sender-in-of*/<schedule_result_t<Sch>, Env> is modeled.

The scheduler's copy constructor, destructor, equality comparison, or swap member functions must be non-throwing.

All of those member functions as well as the scheduler type's schedule function must be thread-safe.

Two schedulers are equal only if they represent the same execution resource.

For a given scheduler sch, the expression get_completion_scheduler<set_value_t>(get_env(schedule(sch))) compares equal to sch.

For a given scheduler sch, if the expression get_domain(sch) is well-formed, then the expression get_domain(get_env(schedule(sch))) is also well-formed and has the same type.

The destructor of a scheduler must not block pending completion of any receivers connected to the sender objects returned from schedule (the underlying resource may provide a separate API to wait for completion of submitted function objects)

[edit] Examples

simple wrapper for std::execution::run_loop that constantly polls run_loop's queue on a single dedicated thread. Demo using draft reference implementation: https://godbolt.org/z/146fY4Y91

#include <execution>
#include <iostream>
#include <thread>
 
class single_thread_context
{
    std::execution::run_loop loop_{};
    std::jthread thread_;
 
public:
    single_thread_context()
        : thread_([this] { loop_.run(); })
    {}
    single_thread_context(single_thread_context&&) = delete;
 
    ~single_thread_context()
    {
        loop_.finish();
    }
 
    std::execution::scheduler auto get_scheduler() noexcept
    {
        return loop_.get_scheduler();
    }
};
 
int main()
{
    single_thread_context ctx;
 
    std::execution::sender auto snd =
        std::execution::schedule(ctx.get_scheduler())
        | std::execution::then([]
            {
                std::cout << "Hello world! Have an int.\n";
                return 015;
            })
        | std::execution::then([](int arg) { return arg + 42; });
 
    auto [i] = std::this_thread::sync_wait(snd).value();
 
    std::cout << "Back in the main thread, result is " << i << '\n';
}

Output:

Hello world! Have an int.
Back in the main thread, result is 55

[edit] See also

prepares a task graph for execution on a given scheduler
(customization point object)[edit]