Memory Resources

• cuda::experimental::memory_pool_properties
  • cuda::experimental::memory_pool_properties
    • initial_pool_size
    • release_threshold
    • allocation_handle_type
• cuda::experimental::device_memory_pool_ref
  • cuda::experimental::device_memory_pool_ref
    • default_queries
    • device_memory_pool_ref()
    • device_memory_pool_ref()
    • device_memory_pool_ref()
    • get_property
• cuda::experimental::pinned_memory_pool_ref
  • cuda::experimental::pinned_memory_pool_ref
    • default_queries
    • pinned_memory_pool_ref()
    • get_property
    • get_property
• cuda::experimental::legacy_pinned_memory_resource
  • cuda::experimental::legacy_pinned_memory_resource
    • default_queries
    • legacy_pinned_memory_resource()
    • allocate_sync()
    • deallocate_sync()
    • operator==()
    • get_property
    • get_property
• cuda::experimental::shared_resource
  • cuda::experimental::shared_resource
    • shared_resource()
    • shared_resource()
    • shared_resource()
    • ~shared_resource()
    • operator=()
    • operator=()
    • swap()
    • allocate_sync()
    • deallocate_sync()
    • allocate()
    • deallocate()
    • swap
    • operator==
    • operator!=
    • get_property
    • get_property

The <cuda/experimental/memory_resource.cuh> header provides:

• any_synchronous_resource and any_resource type erased memory resources similar to std::any. In contrast to resource_ref they own the contained resource.

• device_memory_resource A standard C++ interface for heterogeneous, stream-ordered memory allocation tailored to the needs of CUDA C++ developers. This design builds off of the success of the RAPIDS Memory Manager (RMM) project and evolves the design based on lessons learned.

• shared_resource a type erased reference counted memory resource. In contrast to any_resource it additionally provides shared ownership semantics.

<cuda/experimental/memory_resource.cuh> is not intended to replace RMM, but instead moves the definition of the memory allocation interface to a more centralized home in CCCL. RMM will remain as a collection of implementations of the cuda::mr interfaces.