vecmem 1.18.0
|
Memory resource that wraps page-locked CUDA host allocation. More...
#include <vecmem/memory/cuda/host_memory_resource.hpp>
Public Member Functions | |
VECMEM_CUDA_EXPORT | host_memory_resource () |
Default constructor. | |
VECMEM_CUDA_EXPORT | ~host_memory_resource () override |
Destructor. | |
Private Member Functions | |
Function(s) implementing @c vecmem::memory_resource | |
VECMEM_CUDA_EXPORT void * | do_allocate (std::size_t, std::size_t) override |
Allocate page-locked host memory. | |
VECMEM_CUDA_EXPORT void | do_deallocate (void *p, std::size_t, std::size_t) override |
De-allocate a previously allocated page-locked memory block. | |
VECMEM_CUDA_EXPORT bool | do_is_equal (const memory_resource &other) const noexcept override |
Compares *this for equality with other . | |
Memory resource that wraps page-locked CUDA host allocation.
This is an allocator-type memory resource that allocates CUDA host memory, which is page-locked by default to allow faster transfer to the CUDA devices.