Namespaces
Variants
Views
Actions

Difference between revisions of "Template:cpp/atomic/atomic/wait"

From cppreference.com
(as pointed out in Talk, outstanding_task_count can decrement on a thread before spawning loop increments it again)
m
 
(6 intermediate revisions by 3 users not shown)
Line 3: Line 3:
  
 
{{dcl begin}}
 
{{dcl begin}}
−
{{dcl rev begin | since=c++20}}
+
{{dcl|num=1|since=c++20|1=
−
{{dcl | 1=
+
void wait( {{#ifeq:{{{1|}}}|atomic_flag|bool|T}} old, std::memory_order order =
−
void wait( {{#ifeq:{{{1|}}}|atomic_flag|bool|T}} old,
+
          {{space as|{{#ifeq:{{{1|}}}|atomic_flag|bool|T}}}}          std::memory_order::seq_cst ) const noexcept;
−
          std::memory_order order = std::memory_order::seq_cst ) const noexcept;
+
 
}}
 
}}
−
{{dcl | 1=
+
{{dcl|num=2|since=c++20|1=
−
void wait( {{#ifeq:{{{1|}}}|atomic_flag|bool|T}} old,
+
void wait( {{#ifeq:{{{1|}}}|atomic_flag|bool|T}} old, std::memory_order order =
−
          std::memory_order order = std::memory_order::seq_cst ) const volatile noexcept;
+
          {{space as|{{#ifeq:{{{1|}}}|atomic_flag|bool|T}}}}          std::memory_order::seq_cst ) const volatile noexcept;
 
}}
 
}}
−
{{dcl rev end}}
 
 
{{dcl end}}
 
{{dcl end}}
  
 
Performs atomic waiting operations. Behaves as if it repeatedly performs the following steps:  
 
Performs atomic waiting operations. Behaves as if it repeatedly performs the following steps:  
−
 
+
* Compare {{#ifeq:{{{1|}}}|atomic_flag|{{c|this->test(order)}}|the [[cpp/language/object|value representation]] of {{c|this->load(order)}}}} with that of {{c|old}}.
−
* Compare {{#ifeq:{{{1|}}}|atomic_flag|{{c|this->test(order)}}|the [[cpp/language/object|value representation]] of {{c|this->load(order)}}}} with that of {{tt|old}}.
+
 
** If those are equal, then blocks until {{c|*this}} is notified by {{lc|notify_one()}} or {{lc|notify_all()}}, or the thread is unblocked spuriously.
 
** If those are equal, then blocks until {{c|*this}} is notified by {{lc|notify_one()}} or {{lc|notify_all()}}, or the thread is unblocked spuriously.
 
** Otherwise, returns.
 
** Otherwise, returns.
  
 
These functions are guaranteed to return only if value has changed, even if underlying implementation unblocks spuriously.
 
These functions are guaranteed to return only if value has changed, even if underlying implementation unblocks spuriously.
 +
 +
If {{c|order}} is one of {{c|std::memory_order::release}} and {{c|std::memory_order::acq_rel}}, the behavior is undefined.
  
 
===Parameters===
 
===Parameters===
 
{{par begin}}
 
{{par begin}}
−
{{par | old | the value to check the {{tt|{{{1|}}}}}'s object no longer contains}}
+
{{par|old|the value to check the {{tt|{{{1|}}}}}'s object no longer contains}}
−
{{par | order | the memory synchronization ordering for this operation: must not be {{lc|std::memory_order::release}} or {{lc|std::memory_order::acq_rel}}}}
+
{{par|order|memory order constraints to enforce}}
 
{{par end}}  
 
{{par end}}  
  
Line 35: Line 34:
 
This form of change-detection is often more efficient than simple polling or pure spinlocks.
 
This form of change-detection is often more efficient than simple polling or pure spinlocks.
  
−
Due to the [https://en.wikipedia.org/wiki/ABA_problem ABA problem], transient changes from {{tt|old}} to another value and back to {{tt|old}} might be missed, and not unblock.
+
Due to the {{enwiki|ABA problem}}, transient changes from {{c|old}} to another value and back to {{c|old}} might be missed, and not unblock.
  
 
{{#ifeq:{{{1|}}}|atomic_flag||The comparison is bitwise (similar to {{lc|std::memcmp}}); no comparison operator is used. Padding bits that never participate in an object's value representation are ignored.}}
 
{{#ifeq:{{{1|}}}|atomic_flag||The comparison is bitwise (similar to {{lc|std::memcmp}}); no comparison operator is used. Padding bits that never participate in an object's value representation are ignored.}}
  
 
===Example===
 
===Example===
−
{{example|
+
{{example
−
code=
+
|code=
−
{{#ifeq:{{{1|}}}|atomic|#include <atomic>
+
{{#ifeq:{{{1|atomic}}}|atomic|#include <atomic>
 
#include <chrono>
 
#include <chrono>
 
#include <future>
 
#include <future>
Line 56: Line 55:
 
     std::future<void> task_futures[16];
 
     std::future<void> task_futures[16];
 
     std::atomic<unsigned> outstanding_task_count{16};
 
     std::atomic<unsigned> outstanding_task_count{16};
−
 
+
   
−
     // Spawn several tasks which take different amounts of time, then
+
     // Spawn several tasks which take different amounts of
−
     // decrement the outstanding job count.
+
     // time, then decrement the outstanding task count.
 
     for (std::future<void>& task_future : task_futures)
 
     for (std::future<void>& task_future : task_futures)
−
    {
 
 
         task_future = std::async([&]
 
         task_future = std::async([&]
 
         {
 
         {
 
             // This sleep represents doing real work...
 
             // This sleep represents doing real work...
 
             std::this_thread::sleep_for(50ms);
 
             std::this_thread::sleep_for(50ms);
−
 
+
           
 
             ++completion_count;
 
             ++completion_count;
 
             --outstanding_task_count;
 
             --outstanding_task_count;
−
 
+
           
−
             // When the task count falls to zero, notify the waiter (main thread in this case).
+
             // When the task count falls to zero, notify
 +
            // the waiter (main thread in this case).
 
             if (outstanding_task_count.load() == 0)
 
             if (outstanding_task_count.load() == 0)
 
             {
 
             {
Line 76: Line 75:
 
             }
 
             }
 
         });
 
         });
−
     }
+
      
−
 
+
 
     all_tasks_completed.wait(false);
 
     all_tasks_completed.wait(false);
−
 
+
   
 
     std::cout << "Tasks completed = " << completion_count.load() << '\n';
 
     std::cout << "Tasks completed = " << completion_count.load() << '\n';
 
}|}}
 
}|}}
Line 88: Line 86:
 
===See also===
 
===See also===
 
{{dsc begin}}
 
{{dsc begin}}
−
{{dsc inc | cpp/atomic/atomic/dsc notify_one | {{{1|}}}}}
+
{{dsc inc|cpp/atomic/atomic/dsc notify_one|{{{1|}}}}}
−
{{dsc inc | cpp/atomic/atomic/dsc notify_all | {{{1|}}}}}
+
{{dsc inc|cpp/atomic/atomic/dsc notify_all|{{{1|}}}}}
−
{{#ifeq:{{{1|}}} | atomic_flag |
+
{{#ifeq:{{{1|}}}|atomic_flag|
−
{{dsc inc | cpp/atomic/dsc atomic_flag_notify_one}}
+
{{dsc inc|cpp/atomic/dsc atomic_flag_notify_one}}
−
{{dsc inc | cpp/atomic/dsc atomic_flag_notify_all}}
+
{{dsc inc|cpp/atomic/dsc atomic_flag_notify_all}}
 
|
 
|
−
{{dsc inc | cpp/atomic/dsc atomic_notify_one}}
+
{{dsc inc|cpp/atomic/dsc atomic_notify_one}}
−
{{dsc inc | cpp/atomic/dsc atomic_notify_all}}
+
{{dsc inc|cpp/atomic/dsc atomic_notify_all}}
 
}}
 
}}
 
{{dsc end}}
 
{{dsc end}}

Latest revision as of 06:13, 19 November 2023

Template:cpp/atomic//title Template:cpp/atomic//navbar

void wait( T old, std::memory_order order =
                      std::memory_order::seq_cst ) const noexcept;
 (1) (since C++20)
void wait( T old, std::memory_order order =
                      std::memory_order::seq_cst ) const volatile noexcept;
 (2) (since C++20)

Performs atomic waiting operations. Behaves as if it repeatedly performs the following steps:

  • Compare the value representation of this->load(order) with that of old.
    • If those are equal, then blocks until *this is notified by notify_one() or notify_all(), or the thread is unblocked spuriously.
    • Otherwise, returns.

These functions are guaranteed to return only if value has changed, even if underlying implementation unblocks spuriously.

If order is one of std::memory_order::release and std::memory_order::acq_rel, the behavior is undefined.

Contents

[edit] Parameters

old - the value to check the 's object no longer contains
order - memory order constraints to enforce

[edit] Return value

(none)

[edit] Notes

This form of change-detection is often more efficient than simple polling or pure spinlocks.

Due to the ABA problem, transient changes from old to another value and back to old might be missed, and not unblock.

The comparison is bitwise (similar to std::memcmp); no comparison operator is used. Padding bits that never participate in an object's value representation are ignored.

[edit] Example

Run this code
#include <atomic>
#include <chrono>
#include <future>
#include <iostream>
#include <thread>
 
using namespace std::literals;
 
int main()
{
    std::atomic<bool> all_tasks_completed{false};
    std::atomic<unsigned> completion_count{};
    std::future<void> task_futures[16];
    std::atomic<unsigned> outstanding_task_count{16};
 
    // Spawn several tasks which take different amounts of
    // time, then decrement the outstanding task count.
    for (std::future<void>& task_future : task_futures)
        task_future = std::async([&]
        {
            // This sleep represents doing real work...
            std::this_thread::sleep_for(50ms);
 
            ++completion_count;
            --outstanding_task_count;
 
            // When the task count falls to zero, notify
            // the waiter (main thread in this case).
            if (outstanding_task_count.load() == 0)
            {
                all_tasks_completed = true;
                all_tasks_completed.notify_one();
            }
        });
 
    all_tasks_completed.wait(false);
 
    std::cout << "Tasks completed = " << completion_count.load() << '\n';
}

Output: 

Tasks completed = 16

[edit] See also

(C++20)
 notifies at least one thread waiting on the atomic object
(public member function of Template:cpp/atomic//title) [edit]
(C++20)
 notifies all threads blocked waiting on the atomic object
(public member function of Template:cpp/atomic//title) [edit]
(C++20)
 notifies a thread blocked in atomic_wait
(function template) [edit]
(C++20)
 notifies all threads blocked in atomic_wait
(function template) [edit]
Retrieved from "https://en.cppreference.com/mwiki/index.php?title=Template:cpp/atomic/atomic/wait&oldid=163146"