cache_aligned_resource

[memory_allocation.cache_aligned_resource]

A cache_aligned_resource is a general-purpose memory resource class, which acts as a wrapper to another memory resource to ensure that all allocations are aligned on cache line boundaries to avoid false sharing.

See the cache_aligned_allocator template class section for more information about false sharing avoidance.

// Defined in header <oneapi/tbb/cache_aligned_allocator.h>

namespace oneapi {
namespace tbb {
    class cache_aligned_resource {
    public:
        cache_aligned_resource();
        explicit cache_aligned_resource( std::pmr::memory_resource* );

        std::pmr::memory_resource* upstream_resource() const;

    private:
        void* do_allocate(size_t n, size_t alignment) override;
        void do_deallocate(void* p, size_t n, size_t alignment) override;
        bool do_is_equal(const std::pmr::memory_resource& other) const noexcept override;
    };
} // namespace tbb
} // namespace oneapi

Member Functions

cache_aligned_resource()

Constructs a cache_aligned_resource over std::pmr::get_default_resource().

explicit cache_aligned_resource(std::pmr::memory_resource *r)

Constructs a cache_aligned_resource over the memory resource r.

std::pmr::memory_resource *upstream_resource() const

Returns the pointer to the underlying memory resource.

void *do_allocate(size_t n, size_t alignment) override

Allocates n bytes of memory on a cache-line boundary, with alignment not less than requested. The allocation may include extra memory for padding. Returns pointer to the allocated memory.

void do_deallocate(void *p, size_t n, size_t alignment) override

Deallocates memory pointed to by p and any extra padding. Pointer p must be obtained with do_allocate(n, alignment). The memory must not be deallocated beforehand.

bool do_is_equal(const std::pmr::memory_resource &other) const noexcept override

Compares upstream memory resources of *this and other. If other is not a cache_aligned_resource, returns false.