[RELEASE] wholegraph v24.06

.github/workflows/build.yaml

@@ -28,7 +28,7 @@ concurrency:
 jobs:
   cpp-build:
     secrets: inherit
-    uses: rapidsai/shared-workflows/.github/workflows/conda-cpp-build.yaml@branch-24.04
+    uses: rapidsai/shared-workflows/.github/workflows/conda-cpp-build.yaml@branch-24.06
     with:
       build_type: ${{ inputs.build_type || 'branch' }}
       branch: ${{ inputs.branch }}
@@ -38,7 +38,7 @@ jobs:
   python-build:
     needs: [cpp-build]
     secrets: inherit
-    uses: rapidsai/shared-workflows/.github/workflows/conda-python-build.yaml@branch-24.04
+    uses: rapidsai/shared-workflows/.github/workflows/conda-python-build.yaml@branch-24.06
     with:
       build_type: ${{ inputs.build_type || 'branch' }}
       branch: ${{ inputs.branch }}
@@ -49,7 +49,7 @@ jobs:
     if: github.ref_type == 'branch'
     needs: [python-build]
     secrets: inherit
-    uses: rapidsai/shared-workflows/.github/workflows/custom-job.yaml@branch-24.04
+    uses: rapidsai/shared-workflows/.github/workflows/custom-job.yaml@branch-24.06
     with:
       arch: "amd64"
       branch: ${{ inputs.branch }}
@@ -62,15 +62,15 @@ jobs:
   upload-conda:
     needs: [cpp-build, python-build]
     secrets: inherit
-    uses: rapidsai/shared-workflows/.github/workflows/conda-upload-packages.yaml@branch-24.04
+    uses: rapidsai/shared-workflows/.github/workflows/conda-upload-packages.yaml@branch-24.06
     with:
       build_type: ${{ inputs.build_type || 'branch' }}
       branch: ${{ inputs.branch }}
       date: ${{ inputs.date }}
       sha: ${{ inputs.sha }}
   wheel-build-pylibwholegraph:
     secrets: inherit
-    uses: rapidsai/shared-workflows/.github/workflows/wheels-build.yaml@branch-24.04
+    uses: rapidsai/shared-workflows/.github/workflows/wheels-build.yaml@branch-24.06
     with:
       build_type: ${{ inputs.build_type || 'branch' }}
       branch: ${{ inputs.branch }}
@@ -80,7 +80,7 @@ jobs:
   wheel-publish-pylibwholegraph:
     needs: wheel-build-pylibwholegraph
     secrets: inherit
-    uses: rapidsai/shared-workflows/.github/workflows/wheels-publish.yaml@branch-24.04
+    uses: rapidsai/shared-workflows/.github/workflows/wheels-publish.yaml@branch-24.06
     with:
       build_type: ${{ inputs.build_type || 'branch' }}
       branch: ${{ inputs.branch }}

.github/workflows/pr.yaml

@@ -21,41 +21,41 @@ jobs:
       - wheel-build-pylibwholegraph
       - wheel-test-pylibwholegraph
     secrets: inherit
-    uses: rapidsai/shared-workflows/.github/workflows/pr-builder.yaml@branch-24.04
+    uses: rapidsai/shared-workflows/.github/workflows/pr-builder.yaml@branch-24.06
   checks:
     secrets: inherit
-    uses: rapidsai/shared-workflows/.github/workflows/checks.yaml@branch-24.04
+    uses: rapidsai/shared-workflows/.github/workflows/checks.yaml@branch-24.06
     with:
       enable_check_generated_files: false
   conda-cpp-build:
     needs: checks
     secrets: inherit
-    uses: rapidsai/shared-workflows/.github/workflows/conda-cpp-build.yaml@branch-24.04
+    uses: rapidsai/shared-workflows/.github/workflows/conda-cpp-build.yaml@branch-24.06
     with:
       build_type: pull-request
       node_type: cpu16
   conda-cpp-tests:
     needs: conda-cpp-build
     secrets: inherit
-    uses: rapidsai/shared-workflows/.github/workflows/conda-cpp-tests.yaml@branch-24.04
+    uses: rapidsai/shared-workflows/.github/workflows/conda-cpp-tests.yaml@branch-24.06
     with:
       build_type: pull-request
   conda-python-build:
     needs: conda-cpp-build
     secrets: inherit
-    uses: rapidsai/shared-workflows/.github/workflows/conda-python-build.yaml@branch-24.04
+    uses: rapidsai/shared-workflows/.github/workflows/conda-python-build.yaml@branch-24.06
     with:
       build_type: pull-request
   conda-python-tests:
     needs: conda-python-build
     secrets: inherit
-    uses: rapidsai/shared-workflows/.github/workflows/conda-python-tests.yaml@branch-24.04
+    uses: rapidsai/shared-workflows/.github/workflows/conda-python-tests.yaml@branch-24.06
     with:
       build_type: pull-request
   docs-build:
     needs: conda-python-build
     secrets: inherit
-    uses: rapidsai/shared-workflows/.github/workflows/custom-job.yaml@branch-24.04
+    uses: rapidsai/shared-workflows/.github/workflows/custom-job.yaml@branch-24.06
     with:
       build_type: pull-request
       arch: "amd64"
@@ -64,14 +64,14 @@ jobs:
   wheel-build-pylibwholegraph:
     needs: checks
     secrets: inherit
-    uses: rapidsai/shared-workflows/.github/workflows/wheels-build.yaml@branch-24.04
+    uses: rapidsai/shared-workflows/.github/workflows/wheels-build.yaml@branch-24.06
     with:
       build_type: pull-request
       script: ci/build_wheel.sh
   wheel-test-pylibwholegraph:
     needs: wheel-build-pylibwholegraph
     secrets: inherit
-    uses: rapidsai/shared-workflows/.github/workflows/wheels-test.yaml@branch-24.04
+    uses: rapidsai/shared-workflows/.github/workflows/wheels-test.yaml@branch-24.06
     with:
       build_type: pull-request
       script: ci/test_wheel.sh

.github/workflows/test.yaml

@@ -16,23 +16,23 @@ on:
 jobs:
   conda-cpp-tests:
     secrets: inherit
-    uses: rapidsai/shared-workflows/.github/workflows/conda-cpp-tests.yaml@branch-24.04
+    uses: rapidsai/shared-workflows/.github/workflows/conda-cpp-tests.yaml@branch-24.06
     with:
       build_type: nightly
       branch: ${{ inputs.branch }}
       date: ${{ inputs.date }}
       sha: ${{ inputs.sha }}
   conda-pytorch-tests:
     secrets: inherit
-    uses: rapidsai/shared-workflows/.github/workflows/conda-python-tests.yaml@branch-24.04
+    uses: rapidsai/shared-workflows/.github/workflows/conda-python-tests.yaml@branch-24.06
     with:
       build_type: nightly
       branch: ${{ inputs.branch }}
       date: ${{ inputs.date }}
       sha: ${{ inputs.sha }}
   wheel-tests-pylibwholegraph:
     secrets: inherit
-    uses: rapidsai/shared-workflows/.github/workflows/wheels-test.yaml@branch-24.04
+    uses: rapidsai/shared-workflows/.github/workflows/wheels-test.yaml@branch-24.06
     with:
       build_type: nightly
       branch: ${{ inputs.branch }}

VERSION

@@ -1 +1 @@
-24.04.00
+24.06.00

ci/build_docs.sh

@@ -22,7 +22,7 @@ rapids-print-env
 rapids-logger "Downloading artifacts from previous jobs"
 
 CPP_CHANNEL=$(rapids-download-conda-from-s3 cpp)
-export RAPIDS_VERSION_NUMBER="24.04"
+export RAPIDS_VERSION_NUMBER="24.06"
 export RAPIDS_DOCS_DIR="$(mktemp -d)"
 
 rapids-mamba-retry install \

ci/test_python.sh

@@ -49,6 +49,7 @@ PACKAGES="pylibwholegraph"
 rapids-mamba-retry install \
   --channel "${CPP_CHANNEL}" \
   --channel "${PYTHON_CHANNEL}" \
+  'mkl<2024.1.0' \
   "${PACKAGES}"
 
 rapids-logger "Check GPU usage"

conda/environments/all_cuda-118_arch-x86_64.yaml

@@ -24,8 +24,8 @@ dependencies:
 - graphviz
 - ipykernel
 - ipython
-- libraft-headers==24.4.*
-- librmm==24.4.*
+- libraft-headers==24.6.*
+- librmm==24.6.*
 - nanobind>=0.2.0
 - nbsphinx
 - nccl

conda/environments/all_cuda-122_arch-x86_64.yaml

@@ -25,8 +25,8 @@ dependencies:
 - graphviz
 - ipykernel
 - ipython
-- libraft-headers==24.4.*
-- librmm==24.4.*
+- libraft-headers==24.6.*
+- librmm==24.6.*
 - nanobind>=0.2.0
 - nbsphinx
 - nccl

conda/recipes/libwholegraph/conda_build_config.yaml

@@ -19,11 +19,8 @@ doxygen_version:
 nccl_version:
   - ">=2.9.9"
 
-gtest_version:
-  - ">=1.13.0"
+c_stdlib:
+  - sysroot
 
-gmock_version:
-  - ">=1.13.0"
-
-sysroot_version:
+c_stdlib_version:
   - "2.17"

conda/recipes/libwholegraph/meta.yaml

@@ -1,4 +1,4 @@
-# Copyright (c) 2019-2023, NVIDIA CORPORATION.
+# Copyright (c) 2019-2024, NVIDIA CORPORATION.
 
 {% set version = environ['RAPIDS_PACKAGE_VERSION'].lstrip('v') + environ.get('VERSION_SUFFIX', '') %}
 {% set minor_version = version.split('.')[0] + '.' + version.split('.')[1] %}
@@ -49,7 +49,7 @@ requirements:
     - cuda-version ={{ cuda_version }}
     - cmake {{ cmake_version }}
     - ninja
-    - sysroot_{{ target_platform }} {{ sysroot_version }}
+    - {{ stdlib("c") }}
   host:
     {% if cuda_major == "11" %}
     - cudatoolkit
@@ -59,8 +59,6 @@ requirements:
     {% endif %}
     - cuda-version ={{ cuda_version }}
     - doxygen {{ doxygen_version }}
-    - gmock {{ gtest_version }}
-    - gtest {{ gtest_version }}
     - libraft ={{ minor_version }}
     - libraft-headers ={{ minor_version }}
     - librmm ={{ minor_version }}
@@ -134,8 +132,6 @@ outputs:
         {% else %}
         - cuda-cudart
         {% endif %}
-        - gmock {{ gtest_version }}
-        - gtest {{ gtest_version }}
     about:
       home: https://rapids.ai/
       license: Apache-2.0

conda/recipes/pylibwholegraph/conda_build_config.yaml

@@ -16,5 +16,8 @@ cmake_version:
 scikit_build_core_version:
   - ">=0.7.0"
 
-sysroot_version:
+c_stdlib:
+  - sysroot
+
+c_stdlib_version:
   - "2.17"

conda/recipes/pylibwholegraph/meta.yaml

@@ -54,7 +54,7 @@ requirements:
     - cmake {{ cmake_version }}
     - ninja
     - doxygen =1.8.20
-    - sysroot_{{ target_platform }} {{ sysroot_version }}
+    - {{ stdlib("c") }}
   host:
     - cuda-version ={{ cuda_version }}
     {% if cuda_major == "11" %}

cpp/CMakeLists.txt

@@ -1,5 +1,5 @@
 #=============================================================================
-# Copyright (c) 2018-2023, NVIDIA CORPORATION.
+# Copyright (c) 2018-2024, NVIDIA CORPORATION.
 #
 # Licensed under the Apache License, Version 2.0 (the "License");
 # you may not use this file except in compliance with the License.
@@ -14,7 +14,7 @@
 # limitations under the License.
 #=============================================================================
 
-set(RAPIDS_VERSION "24.04")
+set(RAPIDS_VERSION "24.06")
 set(WHOLEGRAPH_VERSION "${RAPIDS_VERSION}.00")
 
 cmake_minimum_required(VERSION 3.23.1 FATAL_ERROR)
@@ -213,7 +213,8 @@ endif()
 
 # optionally build tests
 if(BUILD_TESTS AND CMAKE_PROJECT_NAME STREQUAL PROJECT_NAME)
-    include(./cmake/thirdparty/get_gtest.cmake)
+	include(${rapids-cmake-dir}/cpm/gtest.cmake)
+	rapids_cpm_gtest(BUILD_STATIC)
     include(CTest) # calls enable_testing()
 
     add_subdirectory(tests)

cpp/Doxyfile

@@ -38,7 +38,7 @@ PROJECT_NAME           = "WholeGraph C API"
 # could be handy for archiving the generated documentation or if some version
 # control system is used.
 
-PROJECT_NUMBER         = 24.04
+PROJECT_NUMBER         = 24.06
 
 # Using the PROJECT_BRIEF tag one can provide an optional one line description
 # for a project that appears at the top of each page and should give viewer a

cpp/include/wholememory/wholememory.h

@@ -387,6 +387,12 @@ wholememory_error_code_t wholememory_store_to_file(wholememory_handle_t wholemem
                                                    size_t file_entry_size,
                                                    const char* local_file_name);
 
+/**
+ * @param comm : WholeMemory Comm
+ * @return : bool
+ */
+bool wholememory_is_intranode_communicator(wholememory_comm_t comm);
+
 bool wholememory_is_build_with_nvshmem();
 #ifdef WITH_NVSHMEM_SUPPORT
 wholememory_error_code_t wholememory_get_nvshmem_reference(

cpp/src/graph_ops/append_unique_func.cuh

@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2019-2023, NVIDIA CORPORATION.
+ * Copyright (c) 2019-2024, NVIDIA CORPORATION.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -316,7 +316,7 @@ void graph_append_unique_func(void* target_nodes_ptr,
     <<<num_blocks, kAssignThreadBlockSize, 0, stream>>>(value_id, bucket_count_ptr);
   WM_CUDA_CHECK(cudaGetLastError());
   wholememory_ops::wm_thrust_allocator thrust_allocator(p_env_fns);
-  thrust::exclusive_scan(thrust::cuda::par(thrust_allocator).on(stream),
+  thrust::exclusive_scan(thrust::cuda::par_nosync(thrust_allocator).on(stream),
                          bucket_count_ptr,
                          bucket_count_ptr + num_bucket_count,
                          (int*)bucket_prefix_sum_ptr);

cpp/src/wholegraph_ops/sample_comm.cuh

@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2019-2023, NVIDIA CORPORATION.
+ * Copyright (c) 2019-2024, NVIDIA CORPORATION.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -57,4 +57,41 @@ __global__ void sample_all_kernel(wholememory_gref_t wm_csr_row_ptr,
     }
   }
 }
+
+__device__ __forceinline__ int log2_up_device(int x)
+{
+  if (x <= 2) return x - 1;
+  return 32 - __clz(x - 1);
+}
+template <typename IdType>
+struct ExpandWithOffsetFunc {
+  const IdType* indptr;
+  IdType* indptr_shift;
+  int length;
+  __host__ __device__ auto operator()(int64_t tIdx)
+  {
+    indptr_shift[tIdx] = indptr[tIdx % length] + tIdx / length;
+  }
+};
+
+template <typename WMIdType, typename DegreeType>
+struct ReduceForDegrees {
+  WMIdType* rowoffsets;
+  DegreeType* in_degree_ptr;
+  int length;
+  __host__ __device__ auto operator()(int64_t tIdx)
+  {
+    in_degree_ptr[tIdx] = rowoffsets[tIdx + length] - rowoffsets[tIdx];
+  }
+};
+
+template <typename DegreeType>
+struct MinInDegreeFanout {
+  int max_sample_count;
+  __host__ __device__ auto operator()(DegreeType degree)
+  {
+    return min(static_cast<int>(degree), max_sample_count);
+  }
+};
+
 }  // namespace wholegraph_ops