Migrate stream APIs from rmm::cuda_stream_view to cuda::stream_ref

cpp/docs/DEVELOPER_GUIDE.md

@@ -173,7 +173,7 @@ Similar to a `rmm::device_vector`, allocates a contiguous set of elements in dev
 key differences:
 - As an optimization, elements are uninitialized and no synchronization occurs at construction.
 This limits the types `T` to trivially copyable types.
-- All operations are stream ordered (i.e., they accept a `cuda_stream_view` specifying the stream
+- All operations are stream ordered (i.e., they accept a `cuda::stream_ref` specifying the stream
 on which the operation is performed).
 
 ## Namespaces

cpp/examples/developers/graph_operations/graph_operations.cu

@@ -1,5 +1,5 @@
 /*
- * SPDX-FileCopyrightText: Copyright (c) 2024-2026, NVIDIA CORPORATION.
+ * SPDX-FileCopyrightText: Copyright (c) 2024-2026, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
  * SPDX-License-Identifier: Apache-2.0
  */
 
@@ -18,6 +18,7 @@
 #include <raft/core/handle.hpp>
 #include <raft/random/rng_state.hpp>
 
+#include <cuda/stream>
 #include <thrust/for_each.h>
 
 #include <iostream>
@@ -44,7 +45,7 @@ std::unique_ptr<raft::handle_t> initialize_mg_handle()
   rmm::mr::set_current_device_resource(resource);
 
   std::unique_ptr<raft::handle_t> handle =
-    std::make_unique<raft::handle_t>(rmm::cuda_stream_per_thread, nullptr, resource);
+    std::make_unique<raft::handle_t>(cuda::stream_ref{cudaStreamPerThread}, nullptr, resource);
 
   raft::comms::initialize_mpi_comms(handle.get(), MPI_COMM_WORLD);
   auto& comm           = handle->get_comms();

cpp/examples/developers/vertex_and_edge_partition/vertex_and_edge_partition.cu

@@ -1,5 +1,5 @@
 /*
- * SPDX-FileCopyrightText: Copyright (c) 2024-2026, NVIDIA CORPORATION.
+ * SPDX-FileCopyrightText: Copyright (c) 2024-2026, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
  * SPDX-License-Identifier: Apache-2.0
  */
 
@@ -15,6 +15,7 @@
 #include <raft/random/rng_state.hpp>
 
 #include <cuda/std/tuple>
+#include <cuda/stream>
 #include <thrust/for_each.h>
 
 #include <iostream>
@@ -41,7 +42,7 @@ std::unique_ptr<raft::handle_t> initialize_mg_handle()
   rmm::mr::set_current_device_resource(resource);
 
   std::unique_ptr<raft::handle_t> handle =
-    std::make_unique<raft::handle_t>(rmm::cuda_stream_per_thread, nullptr, resource);
+    std::make_unique<raft::handle_t>(cuda::stream_ref{cudaStreamPerThread}, nullptr, resource);
 
   raft::comms::initialize_mpi_comms(handle.get(), MPI_COMM_WORLD);
   auto& comm           = handle->get_comms();

cpp/examples/users/multi_gpu_application/mg_graph_algorithms.cpp

@@ -1,5 +1,5 @@
 /*
- * SPDX-FileCopyrightText: Copyright (c) 2024-2026, NVIDIA CORPORATION.
+ * SPDX-FileCopyrightText: Copyright (c) 2024-2026, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
  * SPDX-License-Identifier: Apache-2.0
  */
 
@@ -12,6 +12,8 @@
 #include <raft/core/handle.hpp>
 #include <raft/random/rng_state.hpp>
 
+#include <cuda/stream>
+
 #include <iostream>
 #include <string>
 
@@ -36,7 +38,7 @@ std::unique_ptr<raft::handle_t> initialize_mg_handle()
   rmm::mr::set_current_device_resource(resource);
 
   std::unique_ptr<raft::handle_t> handle =
-    std::make_unique<raft::handle_t>(rmm::cuda_stream_per_thread, nullptr, resource);
+    std::make_unique<raft::handle_t>(cuda::stream_ref{cudaStreamPerThread}, nullptr, resource);
 
   raft::comms::initialize_mpi_comms(handle.get(), MPI_COMM_WORLD);
   auto& comm           = handle->get_comms();

cpp/examples/users/single_gpu_application/sg_graph_algorithms.cpp

@@ -1,5 +1,5 @@
 /*
- * SPDX-FileCopyrightText: Copyright (c) 2024-2026, NVIDIA CORPORATION.
+ * SPDX-FileCopyrightText: Copyright (c) 2024-2026, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
  * SPDX-License-Identifier: Apache-2.0
  */
 
@@ -9,6 +9,8 @@
 #include <raft/core/handle.hpp>
 #include <raft/random/rng_state.hpp>
 
+#include <cuda/stream>
+
 #include <iostream>
 #include <string>
 
@@ -19,7 +21,7 @@ std::unique_ptr<raft::handle_t> initialize_sg_handle()
   rmm::mr::set_current_device_resource(resource);
 
   std::unique_ptr<raft::handle_t> handle =
-    std::make_unique<raft::handle_t>(rmm::cuda_stream_per_thread, nullptr, resource);
+    std::make_unique<raft::handle_t>(cuda::stream_ref{cudaStreamPerThread}, nullptr, resource);
   return std::move(handle);
 }
 

cpp/include/cugraph/dendrogram.hpp

@@ -1,5 +1,5 @@
 /*
- * SPDX-FileCopyrightText: Copyright (c) 2021-2026, NVIDIA CORPORATION.
+ * SPDX-FileCopyrightText: Copyright (c) 2021-2026, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
  * SPDX-License-Identifier: Apache-2.0
  */
 #pragma once
@@ -9,6 +9,8 @@
 #include <rmm/device_uvector.hpp>
 #include <rmm/resource_ref.hpp>
 
+#include <cuda/stream>
+
 #include <memory>
 #include <vector>
 
@@ -19,7 +21,7 @@ class Dendrogram {
  public:
   void add_level(vertex_t first_index,
                  vertex_t num_verts,
-                 rmm::cuda_stream_view stream_view,
+                 cuda::stream_ref stream_view,
                  rmm::device_async_resource_ref mr = rmm::mr::get_current_device_resource_ref())
   {
     level_ptr_.push_back(

cpp/include/cugraph/detail/decompress_edge_partition.cuh

@@ -1,5 +1,5 @@
 /*
- * SPDX-FileCopyrightText: Copyright (c) 2020-2026, NVIDIA CORPORATION.
+ * SPDX-FileCopyrightText: Copyright (c) 2020-2026, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
  * SPDX-License-Identifier: Apache-2.0
  */
 #pragma once
@@ -119,7 +119,7 @@ void decompress_edge_partition_to_fill_edgelist_majors(
       detail::decompress_to_edgelist_high_degree<<<update_grid.num_blocks,
                                                    update_grid.block_size,
                                                    0,
-                                                   handle.get_stream()>>>(
+                                                   handle.get_stream().get()>>>(
         edge_partition,
         edge_partition.major_range_first(),
         edge_partition.major_range_first() + (*segment_offsets)[1],
@@ -133,7 +133,7 @@ void decompress_edge_partition_to_fill_edgelist_majors(
       detail::decompress_to_edgelist_mid_degree<<<update_grid.num_blocks,
                                                   update_grid.block_size,
                                                   0,
-                                                  handle.get_stream()>>>(
+                                                  handle.get_stream().get()>>>(
         edge_partition,
         edge_partition.major_range_first() + (*segment_offsets)[1],
         edge_partition.major_range_first() + (*segment_offsets)[2],

cpp/include/cugraph/detail/utility_wrappers.hpp

@@ -1,5 +1,5 @@
 /*
- * SPDX-FileCopyrightText: Copyright (c) 2021-2026, NVIDIA CORPORATION.
+ * SPDX-FileCopyrightText: Copyright (c) 2021-2026, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
  * SPDX-License-Identifier: Apache-2.0
  */
 #pragma once
@@ -12,6 +12,8 @@
 
 #include <rmm/device_uvector.hpp>
 
+#include <cuda/stream>
+
 namespace CUGRAPH_EXPORT cugraph {
 namespace detail {
 
@@ -37,7 +39,7 @@ namespace detail {
  *
  */
 template <typename value_t>
-void uniform_random_fill(rmm::cuda_stream_view const& stream_view,
+void uniform_random_fill(cuda::stream_ref const& stream_view,
                          value_t* d_value,
                          size_t size,
                          value_t min_value,
@@ -58,7 +60,7 @@ void uniform_random_fill(rmm::cuda_stream_view const& stream_view,
 template <typename value_t>
 void transform_increment_ints(raft::device_span<value_t> values,
                               value_t value,
-                              rmm::cuda_stream_view const& stream_view);
+                              cuda::stream_ref const& stream_view);
 
 /**
  * @ingroup utility_wrappers_cpp
@@ -76,7 +78,7 @@ template <typename value_t>
 void transform_not_equal(raft::device_span<value_t> values,
                          raft::device_span<bool> result,
                          value_t compare,
-                         rmm::cuda_stream_view const& stream_view);
+                         cuda::stream_ref const& stream_view);
 
 /**
  * @ingroup utility_wrappers_cpp
@@ -94,7 +96,7 @@ void transform_not_equal(raft::device_span<value_t> values,
  * @param the maximum value occurring in the edge list
  */
 template <typename vertex_t>
-vertex_t compute_maximum_vertex_id(rmm::cuda_stream_view const& stream_view,
+vertex_t compute_maximum_vertex_id(cuda::stream_ref const& stream_view,
                                    vertex_t const* d_edgelist_srcs,
                                    vertex_t const* d_edgelist_dsts,
                                    size_t num_edges);
@@ -114,7 +116,7 @@ vertex_t compute_maximum_vertex_id(rmm::cuda_stream_view const& stream_view,
  * @param the maximum value occurring in the edge list
  */
 template <typename vertex_t>
-vertex_t compute_maximum_vertex_id(rmm::cuda_stream_view const& stream_view,
+vertex_t compute_maximum_vertex_id(cuda::stream_ref const& stream_view,
                                    rmm::device_uvector<vertex_t> const& d_edgelist_srcs,
                                    rmm::device_uvector<vertex_t> const& d_edgelist_dsts)
 {

cpp/include/cugraph/dynamic/memory_manager/block_array.hpp

@@ -1,5 +1,5 @@
 /*
- * SPDX-FileCopyrightText: Copyright (c) 2026, NVIDIA CORPORATION.
+ * SPDX-FileCopyrightText: Copyright (c) 2026, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
  * SPDX-License-Identifier: Apache-2.0
  */
 #pragma once
@@ -8,9 +8,8 @@
 #include <cugraph/utilities/dataframe_buffer.hpp>
 #include <cugraph/utilities/thrust_tuple_utils.hpp>
 
-#include <rmm/cuda_stream_view.hpp>
-
 #include <cuda/std/tuple>
+#include <cuda/stream>
 
 #include <cstddef>
 #include <type_traits>
@@ -31,7 +30,7 @@ class block_array_t {
 
   using buffer_type = dataframe_buffer_type_t<T>;
 
-  block_array_t(size_t elements_per_block, size_t num_blocks, rmm::cuda_stream_view stream)
+  block_array_t(size_t elements_per_block, size_t num_blocks, cuda::stream_ref stream)
     : bit_tree_(elements_per_block, num_blocks),
       block_storage_(allocate_dataframe_buffer<T>(num_blocks * elements_per_block, stream))
   {

cpp/include/cugraph/dynamic/memory_manager/block_array_manager.hpp

@@ -1,12 +1,12 @@
 /*
- * SPDX-FileCopyrightText: Copyright (c) 2026, NVIDIA CORPORATION.
+ * SPDX-FileCopyrightText: Copyright (c) 2026, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
  * SPDX-License-Identifier: Apache-2.0
  */
 #pragma once
 
 #include <cugraph/dynamic/memory_manager/block_array.hpp>
 
-#include <rmm/cuda_stream_view.hpp>
+#include <cuda/stream>
 
 #include <algorithm>
 #include <array>
@@ -55,7 +55,7 @@ class block_array_manager_t {
   block_array_manager_t(block_array_manager_t&&)            = default;
   block_array_manager_t& operator=(block_array_manager_t&&) = default;
 
-  block_access_data_t insert(size_t num_elements_per_block, rmm::cuda_stream_view stream)
+  block_access_data_t insert(size_t num_elements_per_block, cuda::stream_ref stream)
   {
     CUGRAPH_EXPECTS(
       num_elements_per_block <= max_elements_per_block_array_,