Remove device_memory_resource and DMR bridge infrastructure

Delete device_memory_resource.hpp and device_memory_resource_view.hpp.
Strip DMR bridge code from cccl_adaptors.hpp, keeping shared_resource_cast
wrappers. Inline do_allocate/do_deallocate into allocate/deallocate in
stream_ordered_memory_resource. Convert benchmarks from shared_ptr<DMR> to
any_device_resource. Rewrite test mocks to satisfy CCCL concepts directly,
with copyable forwarding wrappers to work around basic_any type-erasure
limitations with GMock types. Replace reinterpret_cast stream constructions
with cuda_stream_view{}.

Author: bdice

cpp/include/rmm/detail/cccl_adaptors.hpp

@@ -24,7 +24,7 @@ namespace mr {
 /**
  * @brief A suballocator that emphasizes fragmentation avoidance and scalable concurrency support.
  *
- * Allocation (do_allocate()) and deallocation (do_deallocate()) are thread-safe. Also,
+ * Allocation and deallocation are thread-safe. Also,
  * this class is compatible with CUDA per-thread default stream.
  *
  * GPU memory is divided into a global arena, per-thread arenas for default streams, and per-stream

cpp/include/rmm/mr/arena_memory_resource.hpp

@@ -28,12 +28,10 @@ namespace mr {
  *
  * * Returns a pointer to an allocation of at least `bytes` usable immediately on
  *   `stream`. The stream-ordered behavior requirements are identical to
- *   `device_memory_resource::allocate`.
+ *   `allocate`.
  *
- * * This signature is compatible with `do_allocate` but adds the extra function
- *   parameter `arg`. The `arg` is provided to the constructor of the
- *   `callback_memory_resource` and will be forwarded along to every invocation
- *   of the callback function.
+ * * The `arg` is provided to the constructor of the `callback_memory_resource`
+ *   and will be forwarded along to every invocation of the callback function.
  */
 using allocate_callback_t = std::function<void*(std::size_t, cuda_stream_view, void*)>;
 
@@ -46,12 +44,10 @@ using allocate_callback_t = std::function<void*(std::size_t, cuda_stream_view, v
  * * Deallocates memory pointed to by `ptr`. `bytes` specifies the size of the allocation
  *   in bytes, and must equal the value of `bytes` that was passed to the allocate callback
  *   function. The stream-ordered behavior requirements are identical to
- *   `device_memory_resource::deallocate`.
+ *   `deallocate`.
  *
- * * This signature is compatible with `do_deallocate` but adds the extra function
- *   parameter `arg`. The `arg` is provided to the constructor of the
- *   `callback_memory_resource` and will be forwarded along to every invocation
- *   of the callback function.
+ * * The `arg` is provided to the constructor of the `callback_memory_resource`
+ *   and will be forwarded along to every invocation of the callback function.
  */
 using deallocate_callback_t = std::function<void(void*, std::size_t, cuda_stream_view, void*)>;
 

cpp/include/rmm/mr/callback_memory_resource.hpp

@@ -68,8 +68,6 @@ class cuda_async_view_memory_resource final {
   cuda_async_view_memory_resource& operator=(cuda_async_view_memory_resource&&) =
     default;  ///< @default_move_assignment{cuda_async_view_memory_resource}
 
-  // -- CCCL memory resource interface (hides device_memory_resource versions) --
-
   /**
    * @brief Allocates memory of size at least \p bytes.
    *

cpp/include/rmm/mr/cuda_async_view_memory_resource.hpp

@@ -34,8 +34,6 @@ class cuda_memory_resource final {
   cuda_memory_resource& operator=(cuda_memory_resource&&) =
     default;  ///< @default_move_assignment{cuda_memory_resource}
 
-  // -- CCCL memory resource interface (hides device_memory_resource versions) --
-
   /**
    * @brief Allocates memory of size at least \p bytes.
    *

cpp/include/rmm/mr/cuda_memory_resource.hpp

@@ -480,8 +480,6 @@ inline auto max_free_size(std::set<superblock> const& superblocks)
  *
  * The global arena is a shared memory pool from which other arenas allocate superblocks.
  *
- * @tparam Upstream Memory resource to use for allocating the arena. Implements
- * rmm::mr::device_memory_resource interface.
  */
 class global_arena final {
  public:
@@ -778,8 +776,6 @@ class global_arena final {
  * An arena is a per-thread or per-non-default-stream memory pool. It allocates
  * superblocks from the global arena, and returns them when the superblocks become empty.
  *
- * @tparam Upstream Memory resource to use for allocating the global arena. Implements
- * rmm::mr::device_memory_resource interface.
  */
 class arena {
  public:
@@ -957,8 +953,6 @@ class arena {
  *
  * This is useful when a thread is about to terminate, and it contains a per-thread arena.
  *
- * @tparam Upstream Memory resource to use for allocating the global arena. Implements
- * rmm::mr::device_memory_resource interface.
  */
 class arena_cleaner {
  public: