Difference between revisions of "cpp/atomic/atomic flag test and set"

Revision as of 14:10, 4 January 2012

<td>

Defined in header <atomic>

</td>

<td >

bool atomic_flag_test_and_set( volatile std::atomic_flag* p );

</td>

<td > (1) </td> <td > (since C++11) </td> </tr> <tr class="t-dcl ">

<td >

bool atomic_flag_test_and_set( std::atomic_flag* p );

</td>

<td > (2) </td> <td > (since C++11) </td> </tr> <tr class="t-dcl ">

<td >

bool atomic_flag_test_and_set_explicit( volatile std::atomic_flag* p,
std::memory_order order );

</td>

<td > (3) </td> <td > (since C++11) </td> </tr> <tr class="t-dcl ">

<td >

bool atomic_flag_test_and_set_explicit( std::atomic_flag* p,
std::memory_order order );

</td>

<td > (4) </td> <td > (since C++11) </td> </tr> Template:ddcl list end

Atomically changes the state of a Template:cpp pointed to by p to set (Template:cpp) and returns the value it held before.

@@ Line 58: / Line 58: @@
      return p->test_and_set(order);
 }
+}}
+===Example===
+{{example cpp
+ | A spinlock mutex can be implemented in userspace using an atomic_flag
+ | code=
+#include <thread>
+#include <vector>
+#include <iostream>
+#include <atomic>
+std::atomic_flag lock = ATOMIC_FLAG_INIT;
+void f(int n)
+{
+    for(int cnt = 0; cnt < 100; ++cnt) {
+        while(std::atomic_flag_test_and_set_explicit(&lock, std::memory_order_acquire))  // acquire lock
+             ; // spin
+        std::cout << "Output from thread " << n << '\n';
+        std::atomic_flag_clear_explicit(&lock, std::memory_order_release);               // release lock
+    }
+}
+int main()
+{
+    std::vector<std::thread> v;
+    for (int n = 0; n < 10; ++n) {
+        v.emplace_back(f, n);
+    }
+    for (auto& t : v) {
+        t.join();
+    }
+}
+ | output=
+Output from thread 2
+Output from thread 6
+Output from thread 7
+...<exactly 1000 lines>...
 }}

p	-	pointer to Template:cpp to access
order	-	the memory sycnhronization ordering for this operation