From 587b72cc51f59d9144c8aadccd06818e064d839c Mon Sep 17 00:00:00 2001 From: Bradley Dice Date: Sun, 11 Oct 2026 23:27:54 +0000 Subject: [PATCH 1/3] Use itopk_size=128 in the core CAGRA multi-partition test sweep --- cpp/tests/neighbors/ann_cagra.cuh | 16 ++++++++-------- 1 file changed, 8 insertions(+), 8 deletions(-) diff --git a/cpp/tests/neighbors/ann_cagra.cuh b/cpp/tests/neighbors/ann_cagra.cuh index a662c32f03..4be2c8797c 100644 --- a/cpp/tests/neighbors/ann_cagra.cuh +++ b/cpp/tests/neighbors/ann_cagra.cuh @@ -2485,15 +2485,15 @@ inline std::vector generate_mp_inputs() split, graph_build_algo::NN_DESCENT, algo, - /*itopk_size*/ 64, + // With a single 10000-row partition, itopk_size=64 leaves + // the recall of the fp16/int8/uint8 datasets at ~0.98, with + // a tail below 0.97 since the NN-descent graph differs from + // build to build. 128 puts every case near 1.0 (minimum + // ~0.994 over thousands of builds), so the single-index + // 0.985 convention applies. + /*itopk_size*/ 128, metric, - // Lower than the single-index 0.985 convention. These - // low-redundancy single-partition / skewed configs dip - // further under filtering, where removing rows thins the - // survivors' graph connectivity; observed recall floors - // around 0.974, so 0.97 leaves headroom for run-to-run - // jitter while still catching real regressions. - /*min_recall*/ 0.97}); + /*min_recall*/ 0.985}); } } } From 39dc534726079f3bb217dc7a5e25a06d3dc76fe0 Mon Sep 17 00:00:00 2001 From: Bradley Dice Date: Sun, 11 Oct 2026 23:28:57 +0000 Subject: [PATCH 2/3] [DO NOT MERGE] Reduce CI to repeated CAGRA multi-partition runs on CUDA 13.3 (gb300, a100, h100, rtxpro6000) --- .github/workflows/pr.yaml | 723 +------------------------------------- ci/test_cpp.sh | 52 ++- 2 files changed, 46 insertions(+), 729 deletions(-) diff --git a/.github/workflows/pr.yaml b/.github/workflows/pr.yaml index a6438c1a8d..9034f96c98 100644 --- a/.github/workflows/pr.yaml +++ b/.github/workflows/pr.yaml @@ -11,450 +11,18 @@ jobs: pr-builder: needs: - build-details - - check-nightly-ci - - changed-files - - checks - - check-c-abi - - security-suite - conda-cpp-build - conda-cpp-tests - - conda-cpp-checks - - conda-python-build - - conda-python-tests - - rocky8-clib-standalone-build - - rocky8-clib-tests - - conda-java-build - - conda-java-tests - - conda-lucene-build - - conda-lucene-tests - - rust-build - - go-build - - docs-build - - docs-build-matrix - - wheel-build-libcuvs - - wheel-build-cuvs - - wheel-tests-cuvs - - devcontainer - - telemetry-setup uses: rapidsai/shared-workflows/.github/workflows/pr-builder.yaml@main permissions: contents: read if: always() with: needs: ${{ toJSON(needs) }} - telemetry-setup: - continue-on-error: true - runs-on: ubuntu-latest - permissions: - contents: read - env: - OTEL_SERVICE_NAME: 'pr-cuvs' - steps: - - name: Telemetry setup - if: ${{ vars.TELEMETRY_ENABLED == 'true' }} - uses: rapidsai/shared-actions/telemetry-dispatch-stash-base-env-vars@main build-details: uses: rapidsai/shared-workflows/.github/workflows/compute-build-details.yaml@main - check-nightly-ci: - runs-on: ubuntu-latest - permissions: - actions: read - contents: read - id-token: write - pull-requests: read - env: - GH_TOKEN: ${{ github.token }} - steps: - - name: Get PR Info - id: get-pr-info - uses: nv-gha-runners/get-pr-info@main - - name: Check if nightly CI is passing - uses: rapidsai/shared-actions/check_nightly_success/dispatch@main - with: - repo: ${{ github.repository }} - target-branch: ${{ fromJSON(steps.get-pr-info.outputs.pr-info).base.ref }} - max-days-without-success: 30 - changed-files: - needs: telemetry-setup - permissions: - actions: read - contents: read - packages: read - pull-requests: read - uses: rapidsai/shared-workflows/.github/workflows/changed-files.yaml@main - with: - files_yaml: | - build_docs: - - '**' - - '!**/*/REVIEW_GUIDELINES.md' - - '!.coderabbit.yaml' - - '!.devcontainer/**' - - '!.github/CODEOWNERS' - - '!.github/ISSUE_TEMPLATE/**' - - '!.github/PULL_REQUEST_TEMPLATE.md' - - '!.github/copy-pr-bot.yaml' - - '!.github/labeler.yml' - - '!.github/ops-bot.yaml' - - '!.github/release.yml' - - '!.github/workflows/build.yaml' - - '!.github/workflows/check-c-abi.yaml' - - '!.github/workflows/labeler.yml' - - '!.github/workflows/publish-rust.yaml' - - '!.github/workflows/security-suite.yml' - - '!.github/workflows/store-c-abi-baseline.yaml' - - '!.github/workflows/test.yaml' - - '!.github/workflows/trigger-breaking-change-alert.yaml' - - '!.github/workflows/update-c-abi-baseline.yaml' - - '!.pre-commit-config.yaml' - - '!.shellcheckrc' - - '!.yamllint.yaml' - - '!Dockerfile' - - '!SECURITY.md' - - '!ci/build_wheel*.sh' - - '!ci/check_style.sh' - - '!ci/release/update-version.sh' - - '!ci/run*.sh' - - '!ci/test*.sh' - - '!ci/validate_wheel.sh' - - '!thirdparty/LICENSES/**' - test_cpp: - - '.github/workflows/pr.yaml' - - 'RAPIDS_BRANCH' - - 'VERSION' - - 'build.sh' - - 'c/**' - - 'ci/build_cpp.sh' - - 'ci/get_test_data.sh' - - 'ci/test_cpp.sh' - - 'cmake/**' - - 'conda/recipes/libcuvs/**' - - 'cpp/**' - - 'dependencies.yaml' - - '!**/*/REVIEW_GUIDELINES.md' - - '!cpp/doxygen/**' - test_standalone_c: - - '.github/workflows/pr.yaml' - - 'RAPIDS_BRANCH' - - 'VERSION' - - 'build.sh' - - 'c/**' - - 'ci/standalone_c/**' - - 'ci/test_standalone_c.sh' - - 'cmake/**' - - 'cpp/**' - - 'dependencies.yaml' - - '!**/*/REVIEW_GUIDELINES.md' - - '!cpp/doxygen/**' - test_java: - - '**' - - '!**/*/REVIEW_GUIDELINES.md' - - '!.coderabbit.yaml' - - '!.devcontainer/**' - - '!.github/CODEOWNERS' - - '!.github/ISSUE_TEMPLATE/**' - - '!.github/PULL_REQUEST_TEMPLATE.md' - - '!.github/copy-pr-bot.yaml' - - '!.github/labeler.yml' - - '!.github/ops-bot.yaml' - - '!.github/release.yml' - - '!.github/workflows/build.yaml' - - '!.github/workflows/check-c-abi.yaml' - - '!.github/workflows/labeler.yml' - - '!.github/workflows/publish-rust.yaml' - - '!.github/workflows/security-suite.yml' - - '!.github/workflows/store-c-abi-baseline.yaml' - - '!.github/workflows/test.yaml' - - '!.github/workflows/trigger-breaking-change-alert.yaml' - - '!.github/workflows/update-c-abi-baseline.yaml' - - '!.pre-commit-config.yaml' - - '!.shellcheckrc' - - '!.yamllint.yaml' - - '!CODEOWNERS' - - '!Dockerfile' - - '!README.md' - - '!SECURITY.md' - - '!ci/build_docs.sh' - - '!ci/build_go.sh' - - '!ci/build_python.sh' - - '!ci/build_rust.sh' - - '!ci/build_wheel*.sh' - - '!ci/checkstyle.sh' - - '!ci/run_cuvs_pytests.sh' - - '!ci/test_cpp.sh' - - '!ci/test_python.sh' - - '!ci/test_wheel_cuvs.sh' - - '!ci/validate_wheel.sh' - - '!conda/environments/**' - - '!conda/recipes/cuvs/**' - - '!conda/recipes/cuvs-bench/**' - - '!conda/recipes/cuvs-bench-cpu/**' - - '!docs/**' - - '!examples/.clang-format' - - '!examples/.clang-tidy' - - '!examples/.clangd' - - '!examples/README.md' - - '!examples/build.sh' - - '!examples/c/**' - - '!examples/cmake/**' - - '!examples/cpp/**' - - '!examples/go/**' - - '!fern/**' - - '!go/**' - - '!img/**' - - '!notebooks/**' - - '!pyproject.toml' - - '!python/**' - - '!rust/**' - - '!thirdparty/LICENSES/**' - test_python_conda: - - '**' - - '!**/*/REVIEW_GUIDELINES.md' - - '!.coderabbit.yaml' - - '!.devcontainer/**' - - '!.github/CODEOWNERS' - - '!.github/ISSUE_TEMPLATE/**' - - '!.github/PULL_REQUEST_TEMPLATE.md' - - '!.github/copy-pr-bot.yaml' - - '!.github/labeler.yml' - - '!.github/ops-bot.yaml' - - '!.github/release.yml' - - '!.github/workflows/build.yaml' - - '!.github/workflows/check-c-abi.yaml' - - '!.github/workflows/labeler.yml' - - '!.github/workflows/publish-rust.yaml' - - '!.github/workflows/security-suite.yml' - - '!.github/workflows/store-c-abi-baseline.yaml' - - '!.github/workflows/test.yaml' - - '!.github/workflows/trigger-breaking-change-alert.yaml' - - '!.github/workflows/update-c-abi-baseline.yaml' - - '!.pre-commit-config.yaml' - - '!.shellcheckrc' - - '!.yamllint.yaml' - - '!ci/build_go.sh' - - '!ci/build_java.sh' - - '!ci/build_lucene.sh' - - '!ci/build_rust.sh' - - '!ci/check_style.sh' - - '!ci/release/update-version.sh' - - '!ci/run_ctests.sh' - - '!ci/test_cpp.sh' - - '!ci/test_java.sh' - - '!ci/test_lucene.sh' - - '!ci/test_standalone_c.sh' - - '!ci/test_wheel_cuvs.sh' - - '!ci/validate_wheel.sh' - - '!cpp/.clang-format' - - '!cpp/.clang-tidy' - - '!CODEOWNERS' - - '!Dockerfile' - - '!README.md' - - '!SECURITY.md' - - '!docs/**' - - '!examples/**' - - '!fern/**' - - '!go/**' - - '!img/**' - - '!java/**' - - '!notebooks/**' - - '!rust/**' - - '!thirdparty/LICENSES/**' - test_python_wheels: - - '**' - - '!**/*/REVIEW_GUIDELINES.md' - - '!.coderabbit.yaml' - - '!.devcontainer/**' - - '!.github/CODEOWNERS' - - '!.github/ISSUE_TEMPLATE/**' - - '!.github/PULL_REQUEST_TEMPLATE.md' - - '!.github/copy-pr-bot.yaml' - - '!.github/labeler.yml' - - '!.github/ops-bot.yaml' - - '!.github/release.yml' - - '!.github/workflows/build.yaml' - - '!.github/workflows/check-c-abi.yaml' - - '!.github/workflows/labeler.yml' - - '!.github/workflows/publish-rust.yaml' - - '!.github/workflows/security-suite.yml' - - '!.github/workflows/store-c-abi-baseline.yaml' - - '!.github/workflows/test.yaml' - - '!.github/workflows/trigger-breaking-change-alert.yaml' - - '!.github/workflows/update-c-abi-baseline.yaml' - - '!.pre-commit-config.yaml' - - '!.shellcheckrc' - - '!.yamllint.yaml' - - '!ci/build_cpp.sh' - - '!ci/build_docs.sh' - - '!ci/build_go.sh' - - '!ci/build_java.sh' - - '!ci/build_lucene.sh' - - '!ci/build_python.sh' - - '!ci/build_rust.sh' - - '!ci/build_standalone_c.sh' - - '!ci/check_style.sh' - - '!ci/release/update-version.sh' - - '!ci/run_ctests.sh' - - '!ci/test_cpp.sh' - - '!ci/test_java.sh' - - '!ci/test_lucene.sh' - - '!ci/test_python.sh' - - '!ci/test_standalone_c.sh' - - '!ci/validate_wheel.sh' - - '!CODEOWNERS' - - '!Dockerfile' - - '!README.md' - - '!SECURITY.md' - - '!conda/**' - - '!cpp/.clang-format' - - '!cpp/.clang-tidy' - - '!docs/**' - - '!examples/**' - - '!fern/**' - - '!java/**' - - '!go/**' - - '!img/**' - - '!notebooks/**' - - '!rust/**' - - '!thirdparty/LICENSES/**' - test_rust: - - '**' - - '!**/*/REVIEW_GUIDELINES.md' - - '!.coderabbit.yaml' - - '!.devcontainer/**' - - '!.github/CODEOWNERS' - - '!.github/ISSUE_TEMPLATE/**' - - '!.github/PULL_REQUEST_TEMPLATE.md' - - '!.github/copy-pr-bot.yaml' - - '!.github/labeler.yml' - - '!.github/ops-bot.yaml' - - '!.github/release.yml' - - '!.github/workflows/build.yaml' - - '!.github/workflows/check-c-abi.yaml' - - '!.github/workflows/labeler.yml' - - '!.github/workflows/publish-rust.yaml' - - '!.github/workflows/security-suite.yml' - - '!.github/workflows/store-c-abi-baseline.yaml' - - '!.github/workflows/test.yaml' - - '!.github/workflows/trigger-breaking-change-alert.yaml' - - '!.github/workflows/update-c-abi-baseline.yaml' - - '!.pre-commit-config.yaml' - - '!.shellcheckrc' - - '!.yamllint.yaml' - - '!CODEOWNERS' - - '!Dockerfile' - - '!README.md' - - '!SECURITY.md' - - '!ci/build_docs.sh' - - '!ci/build_go.sh' - - '!ci/build_java.sh' - - '!ci/build_lucene.sh' - - '!ci/build_python.sh' - - '!ci/build_wheel*.sh' - - '!ci/check_style.sh' - - '!ci/release/update-version.sh' - - '!ci/run_ctests.sh' - - '!ci/run_cuvs_pytests.sh' - - '!ci/test_cpp.sh' - - '!ci/test_java.sh' - - '!ci/test_lucene.sh' - - '!ci/test_python.sh' - - '!ci/test_wheel_cuvs.sh' - - '!ci/validate_wheel.sh' - - '!docs/**' - - '!examples/**' - - '!fern/**' - - '!go/**' - - '!img/**' - - '!java/**' - - '!notebooks/**' - - '!python/**' - - '!thirdparty/LICENSES/**' - test_go: - - '**' - - '!**/*/REVIEW_GUIDELINES.md' - - '!.coderabbit.yaml' - - '!.devcontainer/**' - - '!.github/CODEOWNERS' - - '!.github/ISSUE_TEMPLATE/**' - - '!.github/PULL_REQUEST_TEMPLATE.md' - - '!.github/copy-pr-bot.yaml' - - '!.github/labeler.yml' - - '!.github/ops-bot.yaml' - - '!.github/release.yml' - - '!.github/workflows/build.yaml' - - '!.github/workflows/check-c-abi.yaml' - - '!.github/workflows/labeler.yml' - - '!.github/workflows/publish-rust.yaml' - - '!.github/workflows/security-suite.yml' - - '!.github/workflows/store-c-abi-baseline.yaml' - - '!.github/workflows/test.yaml' - - '!.github/workflows/trigger-breaking-change-alert.yaml' - - '!.github/workflows/update-c-abi-baseline.yaml' - - '!.pre-commit-config.yaml' - - '!.shellcheckrc' - - '!.yamllint.yaml' - - '!CODEOWNERS' - - '!Dockerfile' - - '!README.md' - - '!SECURITY.md' - - '!ci/build_docs.sh' - - '!ci/build_java.sh' - - '!ci/build_lucene.sh' - - '!ci/build_python.sh' - - '!ci/build_rust.sh' - - '!ci/build_wheel*.sh' - - '!ci/check_style.sh' - - '!ci/release/update-version.sh' - - '!ci/run_ctests.sh' - - '!ci/run_cuvs_pytests.sh' - - '!ci/test_cpp.sh' - - '!ci/test_java.sh' - - '!ci/test_lucene.sh' - - '!ci/test_python.sh' - - '!ci/test_standalone_c.sh' - - '!ci/test_wheel_cuvs.sh' - - '!ci/validate_wheel.sh' - - '!docs/**' - - '!examples/.clang-format' - - '!examples/.clang-tidy' - - '!examples/.clangd' - - '!examples/README.md' - - '!examples/build.sh' - - '!examples/c/**' - - '!examples/cmake/**' - - '!examples/cpp/**' - - '!examples/java/**' - - '!fern/**' - - '!img/**' - - '!java/**' - - '!notebooks/**' - - '!python/**' - - '!rust/**' - - '!thirdparty/LICENSES/**' - checks: - needs: telemetry-setup - uses: rapidsai/shared-workflows/.github/workflows/checks.yaml@main - permissions: - contents: read - with: - enable_check_generated_files: false - ignored_pr_jobs: "telemetry-summarize" - check-c-abi: - needs: telemetry-setup - permissions: - actions: read - contents: read - issues: write - pull-requests: read - uses: ./.github/workflows/check-c-abi.yaml - security-suite: - permissions: - actions: read - contents: read - id-token: write # OIDC -> Vault -> nvcr.io image pull - security-events: write # publish redacted SARIF to code scanning - uses: ./.github/workflows/security-suite.yml conda-cpp-build: - needs: [build-details, checks] + needs: [build-details] permissions: actions: read contents: read @@ -468,8 +36,9 @@ jobs: build-datetime: ${{ needs.build-details.outputs.build-datetime }} node_type: cpu16 script: ci/build_cpp.sh + matrix_filter: 'map(select(.CUDA_VER == "13.3.0"))' conda-cpp-tests: - needs: [conda-cpp-build, changed-files] + needs: [conda-cpp-build] permissions: actions: read contents: read @@ -478,16 +47,16 @@ jobs: pull-requests: read secrets: inherit # zizmor: ignore[secrets-inherit] uses: rapidsai/shared-workflows/.github/workflows/conda-cpp-tests.yaml@main - if: fromJSON(needs.changed-files.outputs.changed_file_groups).test_cpp - # Split the C++ tests across several jobs for each CUDA/GPU configuration + # [DO NOT MERGE] One copy of each configuration strategy: fail-fast: false matrix: - shard: [1, 2, 3, 4] - num_shards: [4] + copy: [1] with: build_type: pull-request - script: "ci/test_cpp.sh ${{ matrix.shard }} ${{ matrix.num_shards }}" + script: "ci/test_cpp.sh" + matrix_type: nightly + matrix_filter: 'map(select(.CUDA_VER == "13.3.0" and (.GPU == "gb300" or .GPU == "a100" or .GPU == "h100" or .GPU == "rtxpro6000")))' cache-paths: .cache/cuda-jit cache-key-prefix: libcuvs-cudajit-v1 # Caches are immutable. Re-key when the JIT-relevant sources change so a completed run can @@ -518,279 +87,3 @@ jobs: cache-read-only: true cache-environment: | CUDA_CACHE_PATH=.cache/cuda-jit - conda-cpp-checks: - needs: conda-cpp-build - permissions: - actions: read - contents: read - id-token: write - packages: read - pull-requests: read - uses: rapidsai/shared-workflows/.github/workflows/conda-cpp-post-build-checks.yaml@main - with: - build_type: pull-request - symbol_exclusions: (void (thrust::|cub::)) - package_name: libcuvs - conda-python-build: - needs: [build-details, conda-cpp-build] - permissions: - actions: read - contents: read - id-token: write - packages: read - pull-requests: read - secrets: inherit # zizmor: ignore[secrets-inherit] - uses: rapidsai/shared-workflows/.github/workflows/conda-python-build.yaml@main - with: - build_type: pull-request - build-datetime: ${{ needs.build-details.outputs.build-datetime }} - script: ci/build_python.sh - # Build a conda package for each CUDA x ARCH x minimum supported Python version - matrix_filter: group_by({CUDA_VER, ARCH}) | map(min_by(.PY_VER | split(".") | map(tonumber))) - conda-python-tests: - needs: [conda-python-build, changed-files] - permissions: - actions: read - contents: read - id-token: write - packages: read - pull-requests: read - secrets: inherit # zizmor: ignore[secrets-inherit] - uses: rapidsai/shared-workflows/.github/workflows/conda-python-tests.yaml@main - if: fromJSON(needs.changed-files.outputs.changed_file_groups).test_python_conda - with: - build_type: pull-request - script: ci/test_python.sh - rocky8-clib-standalone-build: - needs: checks - permissions: - actions: read - contents: read - id-token: write - packages: read - pull-requests: read - uses: ./.github/workflows/clib-build.yaml - with: - build_type: pull-request - rocky8-clib-tests: - needs: [rocky8-clib-standalone-build, changed-files] - permissions: - actions: read - contents: read - id-token: write - packages: read - pull-requests: read - uses: ./.github/workflows/matrix-job.yaml - if: fromJSON(needs.changed-files.outputs.changed_file_groups).test_standalone_c - with: - build_type: pull-request - container-image-repository: "rapidsai/citestwheel" - script: "ci/test_standalone_c.sh" - conda-java-build: - needs: [conda-cpp-build, changed-files] - permissions: - actions: read - contents: read - id-token: write - packages: read - pull-requests: read - uses: ./.github/workflows/matrix-job.yaml - if: fromJSON(needs.changed-files.outputs.changed_file_groups).test_java || fromJSON(needs.changed-files.outputs.changed_file_groups).test_cpp - # Artifacts are not published from these jobs, so it's safe to run for multiple CUDA versions. - # If these jobs start producing artifacts, the names will have to differentiate between CUDA versions. - with: - build_type: pull-request - node_type: cpu8 - matrix_filter: map(select(.ARCH == "amd64")) - script: "ci/build_java.sh" - artifact-name-prefix: "cuvs-java" - file_to_upload: "java/cuvs-java/target/" - conda-java-tests: - needs: conda-java-build - permissions: - actions: read - contents: read - id-token: write - packages: read - pull-requests: read - uses: ./.github/workflows/matrix-job.yaml - with: - build_type: pull-request - script: "ci/test_java.sh" - conda-lucene-build: - # Depends on the Java job for the cuvs-java artifact that cuvs-lucene builds against. - needs: conda-java-build - permissions: - actions: read - contents: read - id-token: write - packages: read - pull-requests: read - uses: ./.github/workflows/matrix-job.yaml - # Artifacts are not published from these jobs, so it's safe to run for multiple CUDA versions. - # If these jobs start producing artifacts, the names will have to differentiate between CUDA versions. - with: - build_type: pull-request - node_type: cpu8 - matrix_filter: map(select(.ARCH == "amd64")) - script: "ci/build_lucene.sh" - artifact-name-prefix: "cuvs-lucene" - file_to_upload: "java/cuvs-lucene/target/" - conda-lucene-tests: - needs: conda-lucene-build - permissions: - actions: read - contents: read - id-token: write - packages: read - pull-requests: read - uses: ./.github/workflows/matrix-job.yaml - with: - build_type: pull-request - script: "ci/test_lucene.sh" - rust-build: - needs: [conda-cpp-build, changed-files] - permissions: - actions: read - contents: read - id-token: write - packages: read - pull-requests: read - uses: ./.github/workflows/matrix-job.yaml - if: fromJSON(needs.changed-files.outputs.changed_file_groups).test_rust || fromJSON(needs.changed-files.outputs.changed_file_groups).test_cpp - # Artifacts are not published from these jobs, so it's safe to run for multiple CUDA versions. - # If these jobs start producing artifacts, the names will have to differentiate between CUDA versions. - with: - build_type: pull-request - matrix_filter: map(select(.ARCH == "amd64")) - script: "ci/build_rust.sh" - go-build: - needs: [conda-cpp-build, changed-files] - permissions: - actions: read - contents: read - id-token: write - packages: read - pull-requests: read - uses: ./.github/workflows/matrix-job.yaml - if: fromJSON(needs.changed-files.outputs.changed_file_groups).test_go || fromJSON(needs.changed-files.outputs.changed_file_groups).test_cpp - # Artifacts are not published from these jobs, so it's safe to run for multiple CUDA versions. - # If these jobs start producing artifacts, the names will have to differentiate between CUDA versions. - with: - build_type: pull-request - matrix_filter: map(select(.ARCH == "amd64")) - script: "ci/build_go.sh" - docs-build-matrix: - permissions: - contents: read - uses: rapidsai/shared-workflows/.github/workflows/compute-matrix.yaml@main - with: - build_type: pull-request - matrix_name: docs-build - docs-build: - needs: [docs-build-matrix] - permissions: - actions: read - contents: read - id-token: write - packages: read - pull-requests: read - secrets: inherit # zizmor: ignore[secrets-inherit] - uses: rapidsai/shared-workflows/.github/workflows/custom-job.yaml@main - if: fromJSON(needs.changed-files.outputs.changed_file_groups).build_docs - strategy: - fail-fast: false - matrix: ${{ fromJSON(needs.docs-build-matrix.outputs.matrix) }} - with: - build_type: pull-request - node_type: "gpu-${{ matrix.GPU }}-${{ matrix.DRIVER }}-1" - arch: "${{ matrix.arch }}" - container_image: "rapidsai/ci-conda:26.12-cuda${{ matrix.CUDA_VER }}-${{ matrix.LINUX_VER }}-py${{ matrix.PY_VER }}" - script: "ci/build_docs.sh" - wheel-build-libcuvs: - needs: [build-details, checks] - permissions: - actions: read - contents: read - id-token: write - packages: read - pull-requests: read - secrets: inherit # zizmor: ignore[secrets-inherit] - uses: rapidsai/shared-workflows/.github/workflows/wheels-build.yaml@main - with: - build_type: pull-request - build-datetime: ${{ needs.build-details.outputs.build-datetime }} - node_type: cpu16 - script: ci/build_wheel_libcuvs.sh - # build for every combination of arch and CUDA version, but only for the latest Python - matrix_filter: group_by([.ARCH, (.CUDA_VER|split(".")|map(tonumber)|.[0])]) | map(max_by(.PY_VER|split(".")|map(tonumber))) - package-name: libcuvs - package-type: cpp - wheel-build-cuvs: - needs: [build-details, wheel-build-libcuvs] - permissions: - actions: read - contents: read - id-token: write - packages: read - pull-requests: read - secrets: inherit # zizmor: ignore[secrets-inherit] - uses: rapidsai/shared-workflows/.github/workflows/wheels-build.yaml@main - with: - build_type: pull-request - build-datetime: ${{ needs.build-details.outputs.build-datetime }} - node_type: cpu8 - script: ci/build_wheel_cuvs.sh - package-name: cuvs - package-type: python - # Build a wheel for each CUDA x ARCH x minimum supported Python version - matrix_filter: group_by({CUDA_VER, ARCH}) | map(min_by(.PY_VER | split(".") | map(tonumber))) - wheel-tests-cuvs: - needs: [wheel-build-cuvs, changed-files] - permissions: - actions: read - contents: read - id-token: write - packages: read - pull-requests: read - secrets: inherit # zizmor: ignore[secrets-inherit] - uses: rapidsai/shared-workflows/.github/workflows/wheels-test.yaml@main - if: fromJSON(needs.changed-files.outputs.changed_file_groups).test_python_wheels - with: - build_type: pull-request - script: ci/test_wheel_cuvs.sh - devcontainer: - permissions: - actions: read - contents: read - id-token: write - packages: read - pull-requests: read - secrets: inherit # zizmor: ignore[secrets-inherit] - needs: telemetry-setup - uses: rapidsai/shared-workflows/.github/workflows/build-in-devcontainer.yaml@main - with: - arch: '["amd64", "arm64"]' - cuda: '["13.3"]' - node_type: "cpu8" - env: | - SCCACHE_DIST_MAX_RETRIES=inf - SCCACHE_SERVER_LOG=sccache=debug - SCCACHE_DIST_FALLBACK_TO_LOCAL_COMPILE=false - build_command: | - sccache --zero-stats; - build-all -j0 --verbose 2>&1 | tee telemetry-artifacts/build.log; - sccache --show-adv-stats | tee telemetry-artifacts/sccache-stats.txt; - telemetry-summarize: - # This job must use a self-hosted runner to record telemetry traces. - runs-on: linux-amd64-cpu4 - needs: pr-builder - permissions: - actions: write - if: ${{ vars.TELEMETRY_ENABLED == 'true' && !cancelled() }} - continue-on-error: true - steps: - - name: Telemetry summarize - uses: rapidsai/shared-actions/telemetry-dispatch-summarize@main - env: - GH_TOKEN: ${{ github.token }} diff --git a/ci/test_cpp.sh b/ci/test_cpp.sh index 0e97a0ddef..4b5f1451e1 100755 --- a/ci/test_cpp.sh +++ b/ci/test_cpp.sh @@ -7,8 +7,6 @@ set -euo pipefail # Usage: ci/test_cpp.sh [SHARD NUM_SHARDS] # With SHARD and NUM_SHARDS (1 <= SHARD <= NUM_SHARDS), only every NUM_SHARDS-th libcuvs test is # run, starting from test number SHARD, so the tests can be split across several CI jobs. -SHARD=${1:-1} -NUM_SHARDS=${2:-1} . /opt/conda/etc/profile.d/conda.sh @@ -52,23 +50,49 @@ nvidia-smi # RAPIDS_DATASET_ROOT_DIR is used by test scripts RAPIDS_DATASET_ROOT_DIR=${RAPIDS_TESTS_DIR}/dataset export RAPIDS_DATASET_ROOT_DIR -./ci/get_test_data.sh --NEIGHBORS_ANN_VAMANA_TEST EXITCODE=0 trap "EXITCODE=1" ERR set +e -# Run Python build utilities tests (once, in the first shard) -if [[ "${SHARD}" == "1" ]]; then - rapids-logger "Run libcuvs Python build utilities tests" - pytest cpp/tests/python -fi - -# Run libcuvs gtests from libcuvs-tests package -rapids-logger "Run libcuvs tests (shard ${SHARD} of ${NUM_SHARDS})" -pushd "$CONDA_PREFIX"/bin/gtests/libcuvs -timeout -v --signal=SIGINT --kill-after=60s 100m ctest -j8 --output-on-failure -I "${SHARD},,${NUM_SHARDS}" -popd +# [DO NOT MERGE] Workstream Q: verify the multi-partition fix by repeating every AnnCagraMultiPartitionTest +# case for every data type, and the num_partitions=1 cases with a different dataset per repetition. The +# recall of every repetition is logged as a QREC line (run, test, recall). +rapids-logger "Workstream Q: AnnCagraMultiPartitionTest verification" +nvidia-smi --query-gpu=name,compute_cap,driver_version --format=csv +GT="$CONDA_PREFIX"/bin/gtests/libcuvs +QLOG="${RAPIDS_TESTS_DIR}/qdiag" +mkdir -p "${QLOG}" +NP1="" +for i in 0 1 2 3 4 5; do + NP1="${NP1}AnnCagraMultiPartitionTest/*.Search/${i}:AnnCagraMultiPartitionTest/*.FilteredSearch/${i}:" +done +ALLMP='AnnCagraMultiPartitionTest/*' + +run_q() { + local name=$1 + shift + rapids-logger "Q run: ${name}: $*" + "$@" > "${QLOG}/${name}.log" 2>&1 + echo "exit code: $?" + grep -E '^\[ FAILED \] Ann' "${QLOG}/${name}.log" | sort | uniq -c + awk -v run="${name}" ' + /^\[ RUN \] / { t = $4 } + /Recall = / { for (i = 1; i <= NF; i++) if ($i == "=") { r = $(i + 1); break } + printf "QREC %s %s %s\n", run, t, r }' "${QLOG}/${name}.log" > "${QLOG}/${name}.rec" + awk '{ n++; if (!(1 in mn) || $4 < mn[1]) mn[1] = $4 } + END { printf "QSUM %s n=%d min=%.3f\n", "'"${name}"'", n, mn[1] }' "${QLOG}/${name}.rec" +} + +for b in FLOAT HALF INT8 UINT8; do + run_q "all_${b}" "${GT}/NEIGHBORS_ANN_CAGRA_${b}_UINT32_TEST" --gtest_filter="${ALLMP}" --gtest_repeat=40 +done +for b in HALF INT8 UINT8; do + CUVS_QDIAG_SEED_SWEEP=1 run_q "np1_sweep_${b}" "${GT}/NEIGHBORS_ANN_CAGRA_${b}_UINT32_TEST" --gtest_filter="${NP1}" --gtest_repeat=150 +done + +rapids-logger "Q raw QREC lines" +cat "${QLOG}"/*.rec rapids-logger "Test script exiting with value: $EXITCODE" exit ${EXITCODE} From eaa8e39aee0b0e9c6b02e2622b8f0f674761cde2 Mon Sep 17 00:00:00 2001 From: Bradley Dice Date: Sun, 11 Oct 2026 23:29:02 +0000 Subject: [PATCH 3/3] [DO NOT MERGE] Optionally vary the multi-partition test dataset seed --- cpp/tests/neighbors/ann_cagra.cuh | 7 ++++++- 1 file changed, 6 insertions(+), 1 deletion(-) diff --git a/cpp/tests/neighbors/ann_cagra.cuh b/cpp/tests/neighbors/ann_cagra.cuh index 4be2c8797c..fe97a84378 100644 --- a/cpp/tests/neighbors/ann_cagra.cuh +++ b/cpp/tests/neighbors/ann_cagra.cuh @@ -43,6 +43,7 @@ #include #include #include +#include #include #include #include @@ -2435,7 +2436,11 @@ class AnnCagraMultiPartitionTest : public ::testing::TestWithParam(ps.n_rows) * ps.dim, stream_); search_queries.resize(static_cast(ps.n_queries) * ps.dim, stream_); - raft::random::RngState r(1234ULL); + // [DO NOT MERGE] CUVS_QDIAG_SEED_SWEEP=1 gives every test instance a different dataset seed. + static uint64_t seed_counter = 0; + uint64_t seed = 1234ULL; + if (std::getenv("CUVS_QDIAG_SEED_SWEEP") != nullptr) { seed = 1234ULL + (++seed_counter); } + raft::random::RngState r(seed); InitDataset(handle_, database.data(), ps.n_rows, ps.dim, ps.metric, r); InitDataset(handle_, search_queries.data(), ps.n_queries, ps.dim, ps.metric, r); raft::resource::sync_stream(handle_);