ring_buffer: Use std::hardware_destructive_interference_size to determine alignment size for avoiding false sharing
MSVC 19.11 (A.K.A. VS 15.3)'s C++ standard library implements P0154R1 (http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2016/p0154r1.html) which defines two new constants within the <new> header, std::hardware_destructive_interference_size and std::hardware_constructive_interference_size. std::hardware_destructive_interference_size defines the minimum recommended offset between two concurrently-accessed objects to avoid performance degradation due to contention introduced by the implementation (with the lower-bound being at least alignof(max_align_t)). In other words, the minimum offset between objects necessary to avoid false-sharing. std::hardware_constructive_interference_size on the other hand defines the maximum recommended size of contiguous memory occupied by two objects accessed wth temporal locality by concurrent threads (also defined to be at least alignof(max_align_t)). In other words the maximum size to promote true-sharing. So we can simply use this facility to determine the ideal alignment size. Unfortunately, only MSVC supports this right now, so we need to enclose it within an ifdef for the time being.
This commit is contained in:
parent
b33ce787b7
commit
c51f8563a6
|
@ -9,6 +9,7 @@
|
||||||
#include <atomic>
|
#include <atomic>
|
||||||
#include <cstddef>
|
#include <cstddef>
|
||||||
#include <cstring>
|
#include <cstring>
|
||||||
|
#include <new>
|
||||||
#include <type_traits>
|
#include <type_traits>
|
||||||
#include <vector>
|
#include <vector>
|
||||||
#include "common/common_types.h"
|
#include "common/common_types.h"
|
||||||
|
@ -102,8 +103,15 @@ public:
|
||||||
private:
|
private:
|
||||||
// It is important to align the below variables for performance reasons:
|
// It is important to align the below variables for performance reasons:
|
||||||
// Having them on the same cache-line would result in false-sharing between them.
|
// Having them on the same cache-line would result in false-sharing between them.
|
||||||
alignas(128) std::atomic<std::size_t> m_read_index{0};
|
// TODO: Remove this ifdef whenever clang and GCC support
|
||||||
alignas(128) std::atomic<std::size_t> m_write_index{0};
|
// std::hardware_destructive_interference_size.
|
||||||
|
#if defined(_MSC_VER) && _MSC_VER >= 1911
|
||||||
|
alignas(std::hardware_destructive_interference_size) std::atomic_size_t m_read_index{0};
|
||||||
|
alignas(std::hardware_destructive_interference_size) std::atomic_size_t m_write_index{0};
|
||||||
|
#else
|
||||||
|
alignas(128) std::atomic_size_t m_read_index{0};
|
||||||
|
alignas(128) std::atomic_size_t m_write_index{0};
|
||||||
|
#endif
|
||||||
|
|
||||||
std::array<T, granularity * capacity> m_data;
|
std::array<T, granularity * capacity> m_data;
|
||||||
};
|
};
|
||||||
|
|
Reference in New Issue