mnorris11
diff --git a/‎.github/actions/build_cmake/action.yml‎
Lines changed: 34 additions & 4 deletions b/‎.github/actions/build_cmake/action.yml‎
Lines changed: 34 additions & 4 deletions
diff --git a/‎.github/workflows/build-pull-request.yml‎
Lines changed: 13 additions & 0 deletions b/‎.github/workflows/build-pull-request.yml‎
Lines changed: 13 additions & 0 deletions
diff --git a/‎CMakeLists.txt‎
Lines changed: 17 additions & 0 deletions b/‎CMakeLists.txt‎
Lines changed: 17 additions & 0 deletions
diff --git a/‎faiss/IndexBinaryHNSW.cpp‎
Lines changed: 11 additions & 134 deletions b/‎faiss/IndexBinaryHNSW.cpp‎
Lines changed: 11 additions & 134 deletions
@@ -32,6 +32,10 @@ inputs:
     description: 'Upload test artifacts. Prevents collisions when multiple jobs need to run build_cmake.'
     required: false
     default: 'true'
+  sanitizers:
+    description: 'Build with AddressSanitizer + UndefinedBehaviorSanitizer and run the tests under them.'
+    required: false
+    default: 'OFF'
 runs:
   using: composite
   steps:
@@ -201,7 +205,7 @@ runs:
     - name: Setup ccache
       uses: hendrikmuhs/ccache-action@v1
       with:
-        key: ${{ runner.os }}-${{ runner.arch }}-${{ inputs.opt_level }}-gpu${{ inputs.gpu }}-cuvs${{ inputs.cuvs }}-rocm${{ inputs.rocm }}-svs${{ inputs.svs }}
+        key: ${{ runner.os }}-${{ runner.arch }}-${{ inputs.opt_level }}-gpu${{ inputs.gpu }}-cuvs${{ inputs.cuvs }}-rocm${{ inputs.rocm }}-svs${{ inputs.svs }}-san${{ inputs.sanitizers }}
         max-size: 2G
         update-package-index: true
     - name: Setup macOS Metal environment
@@ -225,6 +229,8 @@ runs:
               -DFAISS_ENABLE_ROCM=${{ inputs.rocm }} \
               -DFAISS_OPT_LEVEL=${{ inputs.opt_level }} \
               -DFAISS_ENABLE_SVS=${{ inputs.svs }} \
+              -DFAISS_ENABLE_ASAN=${{ inputs.sanitizers }} \
+              -DFAISS_ENABLE_UBSAN=${{ inputs.sanitizers }} \
               -DFAISS_ENABLE_C_API=ON \
               -DPYTHON_EXECUTABLE=$CONDA/bin/python \
               -DCMAKE_BUILD_TYPE=Release \
@@ -251,6 +257,11 @@ runs:
     - name: C++ tests
       if: inputs.metal != 'ON'
       shell: bash
+      env:
+        # Leaks are expected (SWIG/BLAS/OpenMP own long-lived allocations), so
+        # only hunt for memory errors / UB. Harmless on non-sanitizer builds.
+        ASAN_OPTIONS: detect_leaks=0:abort_on_error=1
+        UBSAN_OPTIONS: print_stacktrace=1:halt_on_error=1
       run: |
         conda list --show-channel-urls
         export GTEST_OUTPUT="xml:$(realpath .)/test-results/googletest/"
@@ -269,7 +280,7 @@ runs:
       shell: bash
       run: python3 -m pytest faiss/gpu_metal/test/test_metal_python.py -v
     - name: C++ perf benchmarks
-      if: inputs.rocm == 'OFF' && inputs.metal != 'ON'
+      if: inputs.rocm == 'OFF' && inputs.metal != 'ON' && inputs.sanitizers != 'ON'
       shell: bash
       run: |
         conda list --show-channel-urls
@@ -319,12 +330,31 @@ runs:
         # Confirm torch links against ROCm 7.2 (matches the system /opt/rocm-7.2.0).
         python -c "import torch; print('torch', torch.__version__, 'hip', torch.version.hip)"
     - name: Python tests (CPU only)
-      if: inputs.gpu == 'OFF' && inputs.metal != 'ON'
+      if: inputs.gpu == 'OFF' && inputs.metal != 'ON' && inputs.sanitizers != 'ON'
       shell: bash
       run: |
         conda list --show-channel-urls
         pytest --junitxml=test-results/pytest/results.xml tests/test_*.py
         pytest --junitxml=test-results/pytest/results-torch.xml tests/torch_*.py
+    - name: Python tests (CPU only, sanitizers)
+      if: inputs.gpu == 'OFF' && inputs.metal != 'ON' && inputs.sanitizers == 'ON'
+      shell: bash
+      env:
+        ASAN_OPTIONS: detect_leaks=0:abort_on_error=1
+        UBSAN_OPTIONS: print_stacktrace=1:halt_on_error=1
+      run: |
+        conda list --show-channel-urls
+        # faiss.so / _swigfaiss*.so are built with -fsanitize=address, so a
+        # non-instrumented CPython must load the ASan (then UBSan) runtime first
+        # or the import aborts ("ASan runtime does not come first"). Preload them
+        # from the same conda toolchain that built faiss.
+        GCC=x86_64-conda-linux-gnu-gcc
+        export LD_PRELOAD="$($GCC -print-file-name=libasan.so):$($GCC -print-file-name=libubsan.so)"
+        # This job targets the HNSW build determinism bug; it is a dynamic-
+        # dispatch build, so the NONE/AVX2/AVX512 variants of these suites all
+        # run under the sanitizers. Broaden to tests/test_*.py once green.
+        pytest --junitxml=test-results/pytest/results.xml \
+          tests/test_graph_based.py tests/test_index_binary.py
     - name: Python tests (CPU + GPU)
       if: inputs.gpu == 'ON'
       shell: bash
@@ -349,7 +379,7 @@ runs:
       if: inputs.upload_artifacts == 'true'
       uses: actions/upload-artifact@v4
       with:
-        name: test-results-arch=${{ runner.arch }}-opt=${{ inputs.opt_level }}-gpu=${{ inputs.gpu }}-cuvs=${{ inputs.cuvs }}-rocm=${{ inputs.rocm }}-svs=${{ inputs.svs }}
+        name: test-results-arch=${{ runner.arch }}-opt=${{ inputs.opt_level }}-gpu=${{ inputs.gpu }}-cuvs=${{ inputs.cuvs }}-rocm=${{ inputs.rocm }}-svs=${{ inputs.svs }}-san=${{ inputs.sanitizers }}
         path: test-results
     - name: Check installed packages channel
       if: inputs.metal != 'ON'
 
@@ -124,6 +124,19 @@ jobs:
         uses: ./.github/actions/build_cmake
         with:
           opt_level: dd
+  linux-x86_64-DD-ASAN-UBSAN-cmake:
+    name: Linux x86_64 Dynamic Dispatch ASan+UBSan (cmake)
+    needs: [linux-x86_64-cmake, changes]
+    if: ${{ needs.changes.outputs.has_cpu_changes == 'true' || github.event_name != 'pull_request' }}
+    runs-on: faiss-aws-m7i.large
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v4
+      - name: Build and Test (cmake)
+        uses: ./.github/actions/build_cmake
+        with:
+          opt_level: dd
+          sanitizers: ON
   linux-x86_64-GPU-cmake:
     name: Linux x86_64 GPU (cmake)
     needs: [linux-x86_64-cmake, changes]
 
@@ -95,6 +95,9 @@ option(FAISS_ENABLE_PYTHON "Build Python extension." ON)
 option(FAISS_ENABLE_C_API "Build C API." OFF)
 option(FAISS_ENABLE_EXTRAS "Build extras like benchmarks and demos" ON)
 option(FAISS_USE_LTO "Enable Link-Time optimization" OFF)
+option(FAISS_ENABLE_ASAN "Build with AddressSanitizer (-fsanitize=address)." OFF)
+option(FAISS_ENABLE_UBSAN
+  "Build with UndefinedBehaviorSanitizer (-fsanitize=undefined)." OFF)
 if(APPLE AND CMAKE_SYSTEM_PROCESSOR STREQUAL "arm64")
   set(_FAISS_METAL_DEFAULT ON)
 else()
@@ -106,6 +109,20 @@ option(FAISS_ENABLE_SVS "Enable SVS (Intel(R) Scalable Vector Search) integratio
 set(FAISS_SVS_RUNTIME_VERSION "v0" CACHE STRING "Version of the SVS runtime API to use")
 set_property(CACHE FAISS_SVS_RUNTIME_VERSION PROPERTY STRINGS "v0")
 
+# Sanitizer builds (opt-in, driven by the ASan+UBSan CI job). Applied globally
+# so every faiss target plus the test binaries and the Python extension are
+# instrumented. Keep frame pointers for readable reports; make UBSan abort so
+# CI fails on the first violation instead of only logging it.
+if(FAISS_ENABLE_ASAN)
+  add_compile_options(-fsanitize=address -fno-omit-frame-pointer)
+  add_link_options(-fsanitize=address)
+endif()
+if(FAISS_ENABLE_UBSAN)
+  add_compile_options(-fsanitize=undefined -fno-omit-frame-pointer
+    -fno-sanitize-recover=all)
+  add_link_options(-fsanitize=undefined)
+endif()
+
 if(FAISS_ENABLE_GPU)
   if(FAISS_ENABLE_ROCM)
     enable_language(HIP)
 
@@ -43,136 +43,6 @@ namespace faiss {
  * add / search blocks of descriptors
  **************************************************************/
 
-namespace {
-
-void hnsw_add_vertices(
-        IndexBinaryHNSW& index_hnsw,
-        size_t n0,
-        size_t n,
-        const uint8_t* x,
-        bool verbose,
-        bool preset_levels = false) {
-    HNSW& hnsw = index_hnsw.hnsw;
-    size_t ntotal = n0 + n;
-    double t0 = getmillisecs();
-    if (verbose) {
-        printf("hnsw_add_vertices: adding %zd elements on top of %zd "
-               "(preset_levels=%d)\n",
-               n,
-               n0,
-               int(preset_levels));
-    }
-
-    int max_level = hnsw.prepare_level_tab(n, preset_levels);
-
-    if (verbose) {
-        printf("  max_level = %d\n", max_level);
-    }
-
-    auto& locks = index_hnsw.locks;
-    locks.prepare(ntotal);
-
-    // add vectors from highest to lowest level
-    std::vector<int> hist;
-    std::vector<int> order(n);
-
-    { // make buckets with vectors of the same level
-
-        // build histogram
-        for (size_t i = 0; i < n; i++) {
-            HNSW::storage_idx_t pt_id =
-                    static_cast<HNSW::storage_idx_t>(i + n0);
-            int pt_level = hnsw.levels[pt_id] - 1;
-            while (pt_level >= static_cast<int>(hist.size())) {
-                hist.push_back(0);
-            }
-            hist[pt_level]++;
-        }
-
-        // accumulate
-        std::vector<int> offsets(hist.size() + 1, 0);
-        for (size_t i = 0; i < hist.size() - 1; i++) {
-            offsets[i + 1] = offsets[i] + hist[i];
-        }
-
-        // bucket sort
-        for (size_t i = 0; i < n; i++) {
-            HNSW::storage_idx_t pt_id =
-                    static_cast<HNSW::storage_idx_t>(i + n0);
-            int pt_level = hnsw.levels[pt_id] - 1;
-            order[offsets[pt_level]++] = pt_id;
-        }
-    }
-
-    { // perform add
-        RandomGenerator rng2(789);
-
-        size_t i1 = static_cast<int>(n);
-
-        for (int pt_level = static_cast<int>(hist.size()) - 1;
-             pt_level >= int(!index_hnsw.init_level0);
-             pt_level--) {
-            size_t i0 = i1 - hist[pt_level];
-
-            if (verbose) {
-                printf("Adding %zu elements at level %d\n", i1 - i0, pt_level);
-            }
-
-            // random permutation to get rid of dataset order bias
-            for (size_t j = i0; j < i1; j++) {
-                std::swap(
-                        order[j],
-                        order[j + rng2.rand_int(static_cast<int>(i1 - j))]);
-            }
-
-#pragma omp parallel
-            {
-                std::unique_ptr<VisitedTable> vt = VisitedTable::create(ntotal);
-
-                std::unique_ptr<DistanceComputer> dis(
-                        index_hnsw.get_distance_computer());
-                bool do_display = verbose && omp_get_thread_num() == 0;
-                size_t prev_display = 0;
-
-#pragma omp for schedule(dynamic)
-                for (int64_t i = i0; i < i1; i++) {
-                    HNSW::storage_idx_t pt_id = order[i];
-                    dis->set_query(
-                            (float*)(x + (pt_id - n0) * index_hnsw.code_size));
-
-                    hnsw.add_with_locks(
-                            *dis,
-                            pt_level,
-                            pt_id,
-                            locks,
-                            *vt,
-                            index_hnsw.keep_max_size_level0 && (pt_level == 0));
-
-                    if (do_display && i - i0 > prev_display + 10000) {
-                        prev_display = i - i0;
-                        printf("  %zu / %zu\r", i - i0, i1 - i0);
-                        fflush(stdout);
-                    }
-                }
-            }
-            i1 = i0;
-        }
-        if (index_hnsw.init_level0) {
-            FAISS_ASSERT(i1 == 0);
-        } else {
-            FAISS_ASSERT((i1 - hist[0]) == 0);
-        }
-    }
-    if (verbose) {
-        printf("Done in %.3f ms\n", getmillisecs() - t0);
-    }
-    if (!index_hnsw.retain_locks) {
-        locks.clear();
-    }
-}
-
-} // anonymous namespace
-
 /**************************************************************
  * IndexBinaryHNSW implementation
  **************************************************************/
@@ -270,13 +140,20 @@ void IndexBinaryHNSW::add(idx_t n, const uint8_t* x) {
     storage->add(n, x);
     ntotal = storage->ntotal;
 
-    hnsw_add_vertices(
-            *this,
+    bool preset_levels = hnsw.levels.size() == static_cast<size_t>(ntotal);
+    hnsw_add_vertices_deterministic(
+            hnsw,
             n0,
             n,
-            x,
+            d,
+            init_level0,
+            keep_max_size_level0,
+            preset_levels,
             verbose,
-            hnsw.levels.size() == static_cast<size_t>(ntotal));
+            [this] { return get_distance_computer(); },
+            [this, x, n0](DistanceComputer& dc, HNSW::storage_idx_t pt_id) {
+                dc.set_query((const float*)(x + (pt_id - n0) * code_size));
+            });
 }
 
 void IndexBinaryHNSW::reset() {