Merge branch 'branch-24.12' into fix_similarity_issue

.pre-commit-config.yaml

@@ -55,7 +55,7 @@ repos:
               setup[.]cfg$
       - id: verify-alpha-spec
   - repo: https://github.com/rapidsai/dependency-file-generator
-    rev: v1.13.11
+    rev: v1.16.0
     hooks:
         - id: rapids-dependency-file-generator
           args: ["--clean"]

build.sh

@@ -43,7 +43,6 @@ VALIDARGS="
    --pydevelop
    --allgpuarch
    --skip_cpp_tests
-   --without_cugraphops
    --cmake_default_generator
    --clean
    -h
@@ -74,7 +73,6 @@ HELP="$0 [<target> ...] [<flag> ...]
    --pydevelop                - install the Python packages in editable mode
    --allgpuarch               - build for all supported GPU architectures
    --skip_cpp_tests           - do not build the SG test binaries as part of the libcugraph and libcugraph_etl targets
-   --without_cugraphops       - do not build algos that require cugraph-ops
    --cmake_default_generator  - use the default cmake generator instead of ninja
    --clean                    - clean an individual target (note: to do a complete rebuild, use the clean target described above)
    -h                         - print this text
@@ -107,7 +105,6 @@ BUILD_CPP_TESTS=ON
 BUILD_CPP_MG_TESTS=OFF
 BUILD_CPP_MTMG_TESTS=OFF
 BUILD_ALL_GPU_ARCH=0
-BUILD_WITH_CUGRAPHOPS=ON
 CMAKE_GENERATOR_OPTION="-G Ninja"
 PYTHON_ARGS_FOR_INSTALL="-m pip install --no-build-isolation --no-deps --config-settings rapidsai.disable-cuda=true"
 
@@ -170,9 +167,6 @@ fi
 if hasArg --skip_cpp_tests; then
     BUILD_CPP_TESTS=OFF
 fi
-if hasArg --without_cugraphops; then
-    BUILD_WITH_CUGRAPHOPS=OFF
-fi
 if hasArg cpp-mtmgtests; then
     BUILD_CPP_MTMG_TESTS=ON
 fi
@@ -268,7 +262,6 @@ if buildDefault || hasArg libcugraph || hasArg all; then
               -DBUILD_TESTS=${BUILD_CPP_TESTS} \
               -DBUILD_CUGRAPH_MG_TESTS=${BUILD_CPP_MG_TESTS} \
 	      -DBUILD_CUGRAPH_MTMG_TESTS=${BUILD_CPP_MTMG_TESTS} \
-	      -DUSE_CUGRAPH_OPS=${BUILD_WITH_CUGRAPHOPS} \
               ${CMAKE_GENERATOR_OPTION} \
               ${CMAKE_VERBOSE_OPTION}
         cmake --build "${LIBCUGRAPH_BUILD_DIR}" -j${PARALLEL_LEVEL} ${INSTALL_TARGET} ${VERBOSE_FLAG}
@@ -312,7 +305,7 @@ if buildDefault || hasArg pylibcugraph || hasArg all; then
     if hasArg --clean; then
         cleanPythonDir ${REPODIR}/python/pylibcugraph
     else
-        SKBUILD_CMAKE_ARGS="${SKBUILD_EXTRA_CMAKE_ARGS};-DUSE_CUGRAPH_OPS=${BUILD_WITH_CUGRAPHOPS}" \
+        SKBUILD_CMAKE_ARGS="${SKBUILD_EXTRA_CMAKE_ARGS}" \
             python ${PYTHON_ARGS_FOR_INSTALL} ${REPODIR}/python/pylibcugraph
     fi
 fi
@@ -322,7 +315,7 @@ if buildDefault || hasArg cugraph || hasArg all; then
     if hasArg --clean; then
         cleanPythonDir ${REPODIR}/python/cugraph
     else
-        SKBUILD_CMAKE_ARGS="${SKBUILD_EXTRA_CMAKE_ARGS};-DUSE_CUGRAPH_OPS=${BUILD_WITH_CUGRAPHOPS}" \
+        SKBUILD_CMAKE_ARGS="${SKBUILD_EXTRA_CMAKE_ARGS}" \
             python ${PYTHON_ARGS_FOR_INSTALL} ${REPODIR}/python/cugraph
     fi
 fi

ci/build_cpp.sh

@@ -17,4 +17,6 @@ rapids-logger "Begin cpp build"
 
 RAPIDS_PACKAGE_VERSION=$(rapids-generate-version) rapids-conda-retry mambabuild conda/recipes/libcugraph
 
+sccache --show-adv-stats
+
 rapids-upload-conda-to-s3 cpp

ci/build_python.sh

@@ -20,19 +20,26 @@ export RAPIDS_PACKAGE_VERSION=$(head -1 ./VERSION)
 
 rapids-logger "Begin py build"
 
+sccache --zero-stats
+
 # TODO: Remove `--no-test` flags once importing on a CPU
 # node works correctly
 rapids-conda-retry mambabuild \
   --no-test \
   --channel "${CPP_CHANNEL}" \
   conda/recipes/pylibcugraph
 
+sccache --show-adv-stats
+sccache --zero-stats
+
 rapids-conda-retry mambabuild \
   --no-test \
   --channel "${CPP_CHANNEL}" \
   --channel "${RAPIDS_CONDA_BLD_OUTPUT_DIR}" \
   conda/recipes/cugraph
 
+sccache --show-adv-stats
+
 # NOTE: nothing in nx-cugraph is CUDA-specific, but it is built on each CUDA
 # platform to ensure it is included in each set of artifacts, since test
 # scripts only install from one set of artifacts based on the CUDA version used

ci/build_wheel.sh

@@ -15,6 +15,10 @@ rapids-generate-version > ./VERSION
 
 cd "${package_dir}"
 
+sccache --zero-stats
+
+rapids-logger "Building '${package_name}' wheel"
+
 python -m pip wheel \
     -w dist \
     -v \
@@ -23,12 +27,14 @@ python -m pip wheel \
     --extra-index-url https://pypi.nvidia.com \
     .
 
+sccache --show-adv-stats
+
 # pure-python packages should be marked as pure, and not have auditwheel run on them.
 if [[ ${package_name} == "nx-cugraph" ]] || \
    [[ ${package_name} == "cugraph-dgl" ]] || \
    [[ ${package_name} == "cugraph-pyg" ]] || \
    [[ ${package_name} == "cugraph-equivariant" ]]; then
-    RAPIDS_PY_WHEEL_NAME="${package_name}_${RAPIDS_PY_CUDA_SUFFIX}" RAPIDS_PY_WHEEL_PURE="1" rapids-upload-wheels-to-s3 dist
+    RAPIDS_PY_WHEEL_NAME="${package_name}_${RAPIDS_PY_CUDA_SUFFIX}" RAPIDS_PY_WHEEL_PURE="1" rapids-upload-wheels-to-s3 python dist
 else
     case "${RAPIDS_CUDA_VERSION}" in
         12.*)
@@ -48,5 +54,5 @@ else
 
     mkdir -p final_dist
     python -m auditwheel repair -w final_dist "${EXCLUDE_ARGS[@]}" dist/*
-    RAPIDS_PY_WHEEL_NAME="${package_name}_${RAPIDS_PY_CUDA_SUFFIX}" rapids-upload-wheels-to-s3 final_dist
+    RAPIDS_PY_WHEEL_NAME="${package_name}_${RAPIDS_PY_CUDA_SUFFIX}" rapids-upload-wheels-to-s3 python final_dist
 fi

conda/recipes/cugraph/meta.yaml

@@ -75,10 +75,10 @@ requirements:
     - {{ pin_compatible('cuda-version', max_pin='x', min_pin='x') }}
     {% if cuda_major == "11" %}
     - cudatoolkit
-    - cuda-python >=11.7.1,<12.0a0
+    - cuda-python >=11.7.1,<12.0a0,<=11.8.3
     {% else %}
     - cuda-cudart
-    - cuda-python >=12.0,<13.0a0
+    - cuda-python >=12.0,<13.0a0,<=12.6.0
     {% endif %}
     - cudf ={{ minor_version }}
     - cupy >=12.0.0

cpp/CMakeLists.txt

@@ -54,17 +54,11 @@ option(BUILD_SHARED_LIBS "Build cuGraph shared libraries" ON)
 option(BUILD_CUGRAPH_MG_TESTS "Build cuGraph multigpu algorithm tests" OFF)
 option(CMAKE_CUDA_LINEINFO "Enable the -lineinfo option for nvcc (useful for cuda-memcheck / profiler" OFF)
 option(BUILD_TESTS "Configure CMake to build tests" ON)
-option(USE_CUGRAPH_OPS "Enable all functions that call cugraph-ops" ON)
 option(USE_RAFT_STATIC "Build raft as a static library" OFF)
 option(CUGRAPH_COMPILE_RAFT_LIB "Compile the raft library instead of using it header-only" ON)
 option(CUDA_STATIC_RUNTIME "Statically link the CUDA toolkit runtime and libraries" OFF)
-option(CUGRAPH_USE_CUGRAPH_OPS_STATIC "Build and statically link the cugraph-ops library" OFF)
-option(CUGRAPH_EXCLUDE_CUGRAPH_OPS_FROM_ALL "Exclude cugraph-ops targets from cuGraph's 'all' target" OFF)
-option(ALLOW_CLONE_CUGRAPH_OPS "Whether to attempt to clone cugraph-ops when a local version is not available" OFF)
 
 message(VERBOSE "CUGRAPH: CUDA_STATIC_RUNTIME=${CUDA_STATIC_RUNTIME}")
-message(VERBOSE "CUGRAPH: CUGRAPH_USE_CUGRAPH_OPS_STATIC=${CUGRAPH_USE_CUGRAPH_OPS_STATIC}")
-message(VERBOSE "CUGRAPH: CUGRAPH_EXCLUDE_CUGRAPH_OPS_FROM_ALL=${CUGRAPH_EXCLUDE_CUGRAPH_OPS_FROM_ALL}")
 
 ################################################################################
 # - compiler options -----------------------------------------------------------
@@ -115,13 +109,6 @@ if(CMAKE_BUILD_TYPE MATCHES Debug)
     list(APPEND CUGRAPH_CUDA_FLAGS -G -Xcompiler=-rdynamic)
 endif()
 
-if(NOT USE_CUGRAPH_OPS)
-    message(STATUS "Disabling functions that reference cugraph-ops")
-    list(APPEND CUGRAPH_C_FLAGS -DNO_CUGRAPH_OPS)
-    list(APPEND CUGRAPH_CXX_FLAGS -DNO_CUGRAPH_OPS)
-    list(APPEND CUGRAPH_CUDA_FLAGS -DNO_CUGRAPH_OPS)
-endif()
-
 ###################################################################################################
 # - find CPM based dependencies  ------------------------------------------------------------------
 
@@ -151,10 +138,6 @@ include(${rapids-cmake-dir}/cpm/cuco.cmake)
 rapids_cpm_cuco(BUILD_EXPORT_SET cugraph-exports INSTALL_EXPORT_SET cugraph-exports)
 include(cmake/thirdparty/get_raft.cmake)
 
-if(USE_CUGRAPH_OPS)
-  include(cmake/thirdparty/get_libcugraphops.cmake)
-endif()
-
 if(BUILD_TESTS)
   include(${rapids-cmake-dir}/cpm/gtest.cmake)
   rapids_cpm_gtest(BUILD_STATIC)
@@ -438,12 +421,6 @@ set(CUGRAPH_SOURCES
     src/mtmg/vertex_pairs_result_mg_v64_e64.cu
 )
 
-if(USE_CUGRAPH_OPS)
-    list(APPEND CUGRAPH_SOURCES
-        src/sampling/neighborhood.cu
-    )
-endif()
-
 add_library(cugraph ${CUGRAPH_SOURCES})
 
 set_target_properties(cugraph

cpp/include/cugraph/algorithms.hpp

@@ -23,16 +23,12 @@
 #include <cugraph/legacy/graph.hpp>
 #include <cugraph/legacy/internals.hpp>
 
-#include <rmm/resource_ref.hpp>
-
-#ifndef NO_CUGRAPH_OPS
-#include <cugraph-ops/graph/sampling.hpp>
-#endif
-
 #include <raft/core/device_span.hpp>
 #include <raft/core/handle.hpp>
 #include <raft/random/rng_state.hpp>
 
+#include <rmm/resource_ref.hpp>
+
 #include <optional>
 #include <tuple>
 
@@ -1678,72 +1674,6 @@ node2vec_random_walks(raft::handle_t const& handle,
                       weight_t p,
                       weight_t q);
 
-#ifndef NO_CUGRAPH_OPS
-/**
- * @brief generate sub-sampled graph as an adjacency list (CSR format) given input graph,
- * list of vertices and sample size per vertex. The output graph consists of the given
- * vertices with each vertex having at most `sample_size` neighbors from the original graph
- *
- * @deprecated This API will be deprecated.  uniform_neighbor_sample can be used instead.
- *
- * @tparam graph_t Type of input graph/view (typically, graph_view_t, non-transposed and
- * single-gpu).
- * @param handle RAFT handle object to encapsulate resources (e.g. CUDA stream, communicator, and
- * handles to various CUDA libraries) to run graph algorithms.
- * @param rng_state The RngState instance holding pseudo-random number generator state.
- * @param graph Graph (view )object to sub-sample.
- * @param ptr_d_start Device pointer to set of starting vertex indices for the sub-sampling.
- * @param num_start_vertices = number(vertices) to use for sub-sampling.
- * @param sampling_size = max number of neighbors per output vertex.
- * @param sampling_algo = the sampling algorithm (algo R/algo L/etc.) used to produce outputs.
- * @return std::tuple<rmm::device_uvector<typename graph_t::edge_type>,
- *                    rmm::device_uvector<typename graph_t::vertex_type>>
- * Tuple consisting of two arrays representing the offsets and indices of
- * the sub-sampled graph.
- */
-template <typename vertex_t, typename edge_t>
-std::tuple<rmm::device_uvector<edge_t>, rmm::device_uvector<vertex_t>>
-sample_neighbors_adjacency_list(raft::handle_t const& handle,
-                                raft::random::RngState& rng_state,
-                                graph_view_t<vertex_t, edge_t, false, false> const& graph_view,
-                                vertex_t const* ptr_d_start,
-                                size_t num_start_vertices,
-                                size_t sampling_size,
-                                ops::graph::SamplingAlgoT sampling_algo);
-
-/**
- * @brief generate sub-sampled graph as an edge list (COO format) given input graph,
- * list of vertices and sample size per vertex. The output graph consists of the given
- * vertices with each vertex having at most `sample_size` neighbors from the original graph
- *
- * @deprecated This API will be deprecated.  uniform_neighbor_sample can be used instead.
- *
- * @tparam graph_t Type of input graph/view (typically, graph_view_t, non-transposed and
- * single-gpu).
- * @param handle RAFT handle object to encapsulate resources (e.g. CUDA stream, communicator, and
- * handles to various CUDA libraries) to run graph algorithms.
- * @param rng_state The RngState instance holding pseudo-random number generator state.
- * @param graph Graph (view )object to sub-sample.
- * @param ptr_d_start Device pointer to set of starting vertex indices for the sub-sampling.
- * @param num_start_vertices = number(vertices) to use for sub-sampling.
- * @param sampling_size = max number of neighbors per output vertex.
- * @param sampling_algo = the sampling algorithm (algo R/algo L/etc.) used to produce outputs.
- * @return std::tuple<rmm::device_uvector<typename graph_t::edge_type>,
- *                    rmm::device_uvector<typename graph_t::vertex_type>>
- * Tuple consisting of two arrays representing the source and destination nodes of
- * the sub-sampled graph.
- */
-template <typename vertex_t, typename edge_t>
-std::tuple<rmm::device_uvector<vertex_t>, rmm::device_uvector<vertex_t>> sample_neighbors_edgelist(
-  raft::handle_t const& handle,
-  raft::random::RngState& rng_state,
-  graph_view_t<vertex_t, edge_t, false, false> const& graph_view,
-  vertex_t const* ptr_d_start,
-  size_t num_start_vertices,
-  size_t sampling_size,
-  ops::graph::SamplingAlgoT sampling_algo);
-#endif
-
 /**
  * @brief Finds (weakly-connected-)component IDs of each vertices in the input graph.
  *