rapidsai · rapids-bot · Dec 14, 2021 · Oct 18, 2021 · Oct 19, 2021 · Oct 21, 2021
diff --git a/CHANGELOG.md b/CHANGELOG.md
@@ -1,3 +1,7 @@
+# cuGraph 22.02.00 (Date TBD)
+
+Please see https://github.com/rapidsai/cugraph/releases/tag/v22.02.00a for the latest changes to this development branch.
+
 # cuGraph 21.12.00 (Date TBD)
 
 Please see https://github.com/rapidsai/cugraph/releases/tag/v21.12.00a for the latest changes to this development branch.

diff --git a/conda/environments/cugraph_dev_cuda11.0.yml b/conda/environments/cugraph_dev_cuda11.0.yml
@@ -6,14 +6,14 @@ channels:
 - conda-forge
 dependencies:
 - cudatoolkit=11.0
-- cudf=21.12.*
-- libcudf=21.12.*
-- rmm=21.12.*
-- librmm=21.12.*
+- cudf=22.02.*
+- libcudf=22.02.*
+- rmm=22.02.*
+- librmm=22.02.*
 - dask>=2021.09.1
 - distributed>=2021.09.1
-- dask-cuda=21.12.*
-- dask-cudf=21.12.*
+- dask-cuda=22.02.*
+- dask-cudf=22.02.*
 - nccl>=2.9.9
 - ucx-py=0.23.*
 - ucx-proc=*=gpu

diff --git a/conda/environments/cugraph_dev_cuda11.2.yml b/conda/environments/cugraph_dev_cuda11.2.yml
@@ -6,14 +6,14 @@ channels:
 - conda-forge
 dependencies:
 - cudatoolkit=11.2
-- cudf=21.12.*
-- libcudf=21.12.*
-- rmm=21.12.*
-- librmm=21.12.*
+- cudf=22.02.*
+- libcudf=22.02.*
+- rmm=22.02.*
+- librmm=22.02.*
 - dask>=2021.09.1
 - distributed>=2021.09.1
-- dask-cuda=21.12.*
-- dask-cudf=21.12.*
+- dask-cuda=22.02.*
+- dask-cudf=22.02.*
 - nccl>=2.9.9
 - ucx-py=0.23.*
 - ucx-proc=*=gpu

diff --git a/conda/environments/cugraph_dev_cuda11.4.yml b/conda/environments/cugraph_dev_cuda11.4.yml
@@ -6,14 +6,14 @@ channels:
 - conda-forge
 dependencies:
 - cudatoolkit=11.4
-- cudf=21.12.*
-- libcudf=21.12.*
-- rmm=21.12.*
-- librmm=21.12.*
+- cudf=22.02.*
+- libcudf=22.02.*
+- rmm=22.02.*
+- librmm=22.02.*
 - dask>=2021.09.1
 - distributed>=2021.09.1
-- dask-cuda=21.12.*
-- dask-cudf=21.12.*
+- dask-cuda=22.02.*
+- dask-cudf=22.02.*
 - nccl>=2.9.9
 - ucx-py=0.23.*
 - ucx-proc=*=gpu

@@ -31,6 +31,7 @@ requirements:
     - ucx-py 0.23
     - ucx-proc=*=gpu
     - cudatoolkit {{ cuda_version }}.*
+    - rmm {{ minor_version }}.*
   run:
     - python x.x
     - libcugraph={{ version }}

diff --git a/cpp/CMakeLists.txt b/cpp/CMakeLists.txt
@@ -15,7 +15,7 @@
 #=============================================================================
 
 cmake_minimum_required(VERSION 3.20.1 FATAL_ERROR)
-file(DOWNLOAD https://raw.githubusercontent.com/rapidsai/rapids-cmake/branch-21.12/RAPIDS.cmake
+file(DOWNLOAD https://raw.githubusercontent.com/rapidsai/rapids-cmake/branch-22.02/RAPIDS.cmake
     ${CMAKE_BINARY_DIR}/RAPIDS.cmake)
 include(${CMAKE_BINARY_DIR}/RAPIDS.cmake)
 
@@ -27,7 +27,7 @@ include(rapids-find)
 
 rapids_cuda_init_architectures(CUGRAPH)
 
-project(CUGRAPH VERSION 21.12.00 LANGUAGES C CXX CUDA)
+project(CUGRAPH VERSION 22.02.00 LANGUAGES C CXX CUDA)
 
 if(CMAKE_CUDA_COMPILER_ID STREQUAL "NVIDIA" AND
    CMAKE_CUDA_COMPILER_VERSION VERSION_LESS 11.0)

diff --git a/cpp/cmake/thirdparty/get_raft.cmake b/cpp/cmake/thirdparty/get_raft.cmake
@@ -42,6 +42,6 @@ set(CUGRAPH_BRANCH_VERSION_raft "${CUGRAPH_VERSION_MAJOR}.${CUGRAPH_VERSION_MINO
 # To use a different RAFT locally, set the CMake variable
 # RPM_raft_SOURCE=/path/to/local/raft
 find_and_configure_raft(VERSION    ${CUGRAPH_MIN_VERSION_raft}
-                        FORK       rapidsai
-                        PINNED_TAG branch-${CUGRAPH_BRANCH_VERSION_raft}
+                        FORK       divyegala
+                        PINNED_TAG imp-21.10-handle_stream
                         )
diff --git a/cpp/doxygen/Doxyfile b/cpp/doxygen/Doxyfile
@@ -2075,7 +2075,7 @@ SKIP_FUNCTION_MACROS   = YES
 # the path). If a tag file is not located in the directory in which doxygen is
 # run, you must also specify the path to the tagfile here.
 
-TAGFILES               = rmm.tag=https://docs.rapids.ai/api/librmm/21.12
+TAGFILES               = rmm.tag=https://docs.rapids.ai/api/librmm/22.02
 
 # When a file name is specified after GENERATE_TAGFILE, doxygen will create a
 # tag file that is based on the input files it reads. See section "Linking to

@@ -266,7 +266,7 @@ void copy_v_transform_reduce_key_aggregated_out_nbr(
                    handle.get_stream());
     }
 
-    handle.get_stream_view().synchronize();  // cuco::static_map currently does not take stream
+    handle.sync_stream();  // cuco::static_map currently does not take stream
 
     kv_map_ptr.reset();
 
@@ -284,7 +284,7 @@ void copy_v_transform_reduce_key_aggregated_out_nbr(
       thrust::make_tuple(map_keys.begin(), get_dataframe_buffer_begin(map_value_buffer)));
     kv_map_ptr->insert(pair_first, pair_first + map_keys.size());
   } else {
-    handle.get_stream_view().synchronize();  // cuco::static_map currently does not take stream
+    handle.sync_stream();  // cuco::static_map currently does not take stream
 
     kv_map_ptr.reset();
 

@@ -637,8 +637,7 @@ typename GraphViewType::edge_type compute_num_out_nbrs_from_frontier(
       auto& col_comm = handle.get_subcomm(cugraph::partition_2d::key_naming_t().col_name());
       auto const col_comm_rank = col_comm.get_rank();
 
-      rmm::device_uvector<vertex_t> frontier_vertices(local_frontier_sizes[i],
-                                                      handle.get_stream_view());
+      rmm::device_uvector<vertex_t> frontier_vertices(local_frontier_sizes[i], handle.get_stream());
       device_bcast(col_comm,
                    local_frontier_vertex_first,
                    frontier_vertices.data(),
@@ -1096,7 +1095,7 @@ void update_frontier_v_push_if_out_nbr(
                       d_tx_buffer_last_boundaries.data(),
                       d_tx_buffer_last_boundaries.size(),
                       handle.get_stream());
-    handle.get_stream_view().synchronize();
+    handle.sync_stream();
     std::vector<size_t> tx_counts(h_tx_buffer_last_boundaries.size());
     std::adjacent_difference(
       h_tx_buffer_last_boundaries.begin(), h_tx_buffer_last_boundaries.end(), tx_counts.begin());

@@ -433,7 +433,7 @@ class VertexFrontier {
         h_indices.data(), d_indices.data(), d_indices.size(), handle_ptr_->get_stream());
       raft::update_host(
         h_counts.data(), d_counts.data(), d_counts.size(), handle_ptr_->get_stream());
-      handle_ptr_->get_stream_view().synchronize();
+      handle_ptr_->sync_stream();
 
       size_t offset{0};
       for (size_t i = 0; i < h_indices.size(); ++i) {

@@ -52,7 +52,7 @@ extern "C" cugraph_error_code_t cugraph_type_erased_device_array_create(
     size_t n_bytes = n_elems * (::data_type_sz[dtype]);
 
     auto ret_value = new cugraph::c_api::cugraph_type_erased_device_array_t(
-      n_elems, n_bytes, dtype, raft_handle->get_stream_view());
+      n_elems, n_bytes, dtype, raft_handle->get_stream());
 
     *array = reinterpret_cast<cugraph_type_erased_device_array_t*>(ret_value);
     return CUGRAPH_SUCCESS;

@@ -116,12 +116,10 @@ class EcgLouvain : public cugraph::legacy::Louvain<graph_type> {
 
   void initialize_dendrogram_level(vertex_t num_vertices) override
   {
-    this->dendrogram_->add_level(0, num_vertices, this->handle_.get_stream_view());
+    this->dendrogram_->add_level(0, num_vertices, this->handle_.get_stream());
 
-    get_permutation_vector(num_vertices,
-                           seed_,
-                           this->dendrogram_->current_level_begin(),
-                           this->handle_.get_stream_view());
+    get_permutation_vector(
+      num_vertices, seed_, this->dendrogram_->current_level_begin(), this->handle_.get_stream());
   }
 
  private:
@@ -147,7 +145,7 @@ void ecg(raft::handle_t const& handle,
                   "Invalid input argument: clustering is NULL, should be a device pointer to "
                   "memory for storing the result");
 
-  rmm::device_uvector<weight_t> ecg_weights_v(graph.number_of_edges, handle.get_stream_view());
+  rmm::device_uvector<weight_t> ecg_weights_v(graph.number_of_edges, handle.get_stream());
 
   thrust::copy(handle.get_thrust_policy(),
                graph.edge_data,

@@ -66,7 +66,7 @@ extract(raft::handle_t const& handle,
         vertex_t radius)
 {
   auto v                = csr_view.get_number_of_vertices();
-  auto user_stream_view = handle.get_stream_view();
+  auto user_stream_view = handle.get_stream();
   rmm::device_vector<size_t> neighbors_offsets(n_subgraphs + 1);
   rmm::device_vector<vertex_t> neighbors;
 
@@ -77,7 +77,7 @@ extract(raft::handle_t const& handle,
   reached.reserve(n_subgraphs);
   for (vertex_t i = 0; i < n_subgraphs; i++) {
     // Allocations and operations are attached to the worker stream
-    rmm::device_uvector<vertex_t> local_reach(v, handle.get_internal_stream_view(i));
+    rmm::device_uvector<vertex_t> local_reach(v, handle.get_next_usable_stream(i));
     reached.push_back(std::move(local_reach));
   }
 
@@ -89,8 +89,8 @@ extract(raft::handle_t const& handle,
 
   for (vertex_t i = 0; i < n_subgraphs; i++) {
     // get light handle from worker pool
-    raft::handle_t light_handle(handle, i);
-    auto worker_stream_view = light_handle.get_stream_view();
+    raft::handle_t light_handle(handle.get_next_usable_stream(i));
+    auto worker_stream_view = light_handle.get_stream();
 
     // BFS with cutoff
     // consider adding a device API to BFS (ie. accept source on the device)
@@ -132,7 +132,7 @@ extract(raft::handle_t const& handle,
   }
 
   // wait on every one to identify their neighboors before proceeding to concatenation
-  handle.wait_on_internal_streams();
+  handle.sync_stream_pool();
 
   // Construct neighboors offsets (just a scan on neighborhod vector sizes)
   h_neighbors_offsets[0] = 0;
@@ -148,7 +148,7 @@ extract(raft::handle_t const& handle,
 
   // Construct the neighboors list concurrently
   for (vertex_t i = 0; i < n_subgraphs; i++) {
-    auto worker_stream_view = handle.get_internal_stream_view(i);
+    auto worker_stream_view = handle.get_next_usable_stream(i);
     thrust::copy(rmm::exec_policy(worker_stream_view),
                  reached[i].begin(),
                  reached[i].end(),
@@ -160,7 +160,7 @@ extract(raft::handle_t const& handle,
   }
 
   // wait on every one before proceeding to grouped extraction
-  handle.wait_on_internal_streams();
+  handle.sync_stream_pool();
 
 #ifdef TIMING
   hr_timer.stop();

@@ -43,12 +43,11 @@ class Leiden : public Louvain<graph_type> {
     this->timer_start("update_clustering_constrained");
 
     rmm::device_uvector<vertex_t> next_cluster_v(this->dendrogram_->current_level_size(),
-                                                 this->handle_.get_stream_view());
-    rmm::device_uvector<weight_t> delta_Q_v(graph.number_of_edges, this->handle_.get_stream_view());
-    rmm::device_uvector<vertex_t> cluster_hash_v(graph.number_of_edges,
-                                                 this->handle_.get_stream_view());
+                                                 this->handle_.get_stream());
+    rmm::device_uvector<weight_t> delta_Q_v(graph.number_of_edges, this->handle_.get_stream());
+    rmm::device_uvector<vertex_t> cluster_hash_v(graph.number_of_edges, this->handle_.get_stream());
     rmm::device_uvector<weight_t> old_cluster_sum_v(graph.number_of_vertices,
-                                                    this->handle_.get_stream_view());
+                                                    this->handle_.get_stream());
 
     vertex_t const* d_src_indices    = this->src_indices_v_.data();
     vertex_t const* d_dst_indices    = graph.indices;
@@ -105,7 +104,7 @@ class Leiden : public Louvain<graph_type> {
       }
     }
 
-    this->timer_stop(this->handle_.get_stream_view());
+    this->timer_stop(this->handle_.get_stream());
     return cur_Q;
   }
 
@@ -134,8 +133,7 @@ class Leiden : public Louvain<graph_type> {
       //
       //  Initialize every cluster to reference each vertex to itself
       //
-      this->dendrogram_->add_level(
-        0, current_graph.number_of_vertices, this->handle_.get_stream_view());
+      this->dendrogram_->add_level(0, current_graph.number_of_vertices, this->handle_.get_stream());
 
       thrust::sequence(this->handle_.get_thrust_policy(),
                        this->dendrogram_->current_level_begin(),