hnswlib 0.3.0 → 0.5.2

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 7823039ef0a73f79011cbbb4dd8bb90f56549d82c93cee311725a2e728ab8809
-  data.tar.gz: be1846136d1d87c63d3aa49c404f9148c8d04d8c0ba4af65d6297096b5683150
+  metadata.gz: f313f103d9a4a6e6ae92f21cc92a4aaaecd3121e3a2c5be533f87954afb5ac7d
+  data.tar.gz: 3b7ec61621c905759f466df1ac76c18a76186a5ef5c84517ada5c1493524ec5c
 SHA512:
-  metadata.gz: 5ddd709e8bd403b959c7ab8bc21b17b6e264da0976a2d85a4acd7edeb057b7771d53fac337d3a6dfdd2718dcd499efd2b0f8633170241dd01269cd34753be10a
-  data.tar.gz: 11fd6ee5ca94949665fa30a106e926283c9e190030b47b6b001b0fd755ff5512f8b913b02bb79f1b51cb851c50867d0dca1a861fb2f09b6805e4d0b74bb033b5
+  metadata.gz: 12717a35c16fb2ef3f8884fb541559349410e43c9ba35c98de4fbf4a5faa6cd58ffba2cd26ac049ffaea855cf2c4e33f77ffefea023b901cbd8605ada3b63d5c
+  data.tar.gz: c134905c5c268313a8a6bfff69567a66ca62e3d2e1a9d133cd87592262994eaf3731ac3b59e1eaed212a9dfa514f2c91b9eceb11f420e6f988d15f9250dbe863

data/CHANGELOG.md

@@ -1,3 +1,21 @@
+## [0.5.2] - 2022-02-19
+
+- Update bundled hnswlib version to 0.6.2.
+
+## [0.5.1] - 2022-02-11
+
+- Update bundled hnswlib version to 0.6.1.
+- Update documentations.
+- Introduce conventional commits.
+
+## [0.5.0] - 2021-12-12
+
+- Update bundled hnswlib version to 0.6.0.
+
+## [0.4.0] - 2021-09-12
+
+- Add type declaration file.
+
 ## [0.3.0] - 2021-08-08
 
 - Rename `Hnswlib::Index` to `Hnswlib::HnswIndex` (for compatibility, `Hnswlib::Index` has been an alis for `Hnswlib::HnswIndex`).

data/README.md

@@ -25,6 +25,20 @@ Or install it yourself as:
 
     $ gem install hnswlib
 
+Note: hnswlib.rb gives no optimization options when building native extensions.
+If necessary, add the optimization options yourself during installation, as follows;
+
+```
+$ bundle config --local build.hnswlib "--with-cxxflags=-march=native"
+$ bundle install
+```
+
+Or:
+
+```
+$ gem install hnswlib -- --with-cxxflags=-march=native
+```
+
 ## Documentation
 
 * [Hnswlib.rb API Documentation](https://yoshoku.github.io/hnswlib.rb/doc/)
@@ -57,4 +71,4 @@ The gem is available as open source under the terms of the [Apache-2.0 License](
 
 Bug reports and pull requests are welcome on GitHub at https://github.com/yoshoku/hnswlib.rb.
 This project is intended to be a safe, welcoming space for collaboration,
-and contributors are expected to adhere to the [code of conduct](https://github.com/yoshoku/hnswlib.rb/blob/main/CODE_OF_CONDUCT.md).
+and contributors are expected to adhere to the [Contributor Covenant](http://contributor-covenant.org) code of conduct.

data/ext/hnswlib/hnswlibext.cpp

@@ -1,7 +1,7 @@
 /**
  * hnswlib.rb is a Ruby binding for the Hnswlib.
  *
- * Copyright (c) 2021 Atsushi Tatsuma
+ * Copyright (c) 2021-2022 Atsushi Tatsuma
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.

data/ext/hnswlib/hnswlibext.hpp

@@ -1,7 +1,7 @@
 /**
  * hnswlib.rb is a Ruby binding for the Hnswlib.
  *
- * Copyright (c) 2021 Atsushi Tatsuma
+ * Copyright (c) 2021-2022 Atsushi Tatsuma
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.

data/ext/hnswlib/src/bruteforce.h

@@ -92,13 +92,13 @@ namespace hnswlib {
         searchKnn(const void *query_data, size_t k) const {
             std::priority_queue<std::pair<dist_t, labeltype >> topResults;
             if (cur_element_count == 0) return topResults;
-            for (size_t i = 0; i < k; i++) {
+            for (int i = 0; i < k; i++) {
                 dist_t dist = fstdistfunc_(query_data, data_ + size_per_element_ * i, dist_func_param_);
                 topResults.push(std::pair<dist_t, labeltype>(dist, *((labeltype *) (data_ + size_per_element_ * i +
                                                                                     data_size_))));
             }
             dist_t lastdist = topResults.top().first;
-            for (size_t i = k; i < cur_element_count; i++) {
+            for (int i = k; i < cur_element_count; i++) {
                 dist_t dist = fstdistfunc_(query_data, data_ + size_per_element_ * i, dist_func_param_);
                 if (dist <= lastdist) {
                     topResults.push(std::pair<dist_t, labeltype>(dist, *((labeltype *) (data_ + size_per_element_ * i +

data/ext/hnswlib/src/hnswalg.h

@@ -19,7 +19,6 @@ namespace hnswlib {
         static const tableint max_update_element_locks = 65536;
         HierarchicalNSW() : visited_list_pool_(nullptr), data_level0_memory_(nullptr), linkLists_(nullptr), cur_element_count(0) { }
         HierarchicalNSW(SpaceInterface<dist_t> *s) {
-
         }
 
         HierarchicalNSW(SpaceInterface<dist_t> *s, const std::string &location, bool nmslib = false, size_t max_elements=0) {
@@ -30,7 +29,7 @@ namespace hnswlib {
                 link_list_locks_(max_elements), link_list_update_locks_(max_update_element_locks), element_levels_(max_elements) {
             max_elements_ = max_elements;
 
-            has_deletions_=false;
+            num_deleted_ = 0;
             data_size_ = s->get_data_size();
             fstdistfunc_ = s->get_dist_func();
             dist_func_param_ = s->get_dist_func_param();
@@ -57,8 +56,6 @@ namespace hnswlib {
 
             visited_list_pool_ = new VisitedListPool(1, max_elements);
 
-
-
             //initializations for special treatment of the first node
             enterpoint_node_ = -1;
             maxlevel_ = -1;
@@ -93,6 +90,7 @@ namespace hnswlib {
         size_t cur_element_count;
         size_t size_data_per_element_;
         size_t size_links_per_element_;
+        size_t num_deleted_;
 
         size_t M_;
         size_t maxM_;
@@ -113,20 +111,15 @@ namespace hnswlib {
         std::vector<std::mutex> link_list_update_locks_;
         tableint enterpoint_node_;
 
-
         size_t size_links_level0_;
         size_t offsetData_, offsetLevel0_;
 
-
         char *data_level0_memory_;
         char **linkLists_;
         std::vector<int> element_levels_;
 
         size_t data_size_;
 
-        bool has_deletions_;
-
-
         size_t label_offset_;
         DISTFUNC<dist_t> fstdistfunc_;
         void *dist_func_param_;
@@ -183,7 +176,7 @@ namespace hnswlib {
 
             while (!candidateSet.empty()) {
                 std::pair<dist_t, tableint> curr_el_pair = candidateSet.top();
-                if ((-curr_el_pair.first) > lowerBound) {
+                if ((-curr_el_pair.first) > lowerBound && top_candidates.size() == ef_construction_) {
                     break;
                 }
                 candidateSet.pop();
@@ -272,7 +265,7 @@ namespace hnswlib {
 
                 std::pair<dist_t, tableint> current_node_pair = candidate_set.top();
 
-                if ((-current_node_pair.first) > lowerBound) {
+                if ((-current_node_pair.first) > lowerBound && (top_candidates.size() == ef || has_deletions == false)) {
                     break;
                 }
                 candidate_set.pop();
@@ -548,7 +541,7 @@ namespace hnswlib {
                 }
             }
 
-            if (has_deletions_) {
+            if (num_deleted_) {
                 std::priority_queue<std::pair<dist_t, tableint  >> top_candidates1=searchBaseLayerST<true>(currObj, query_data,
                                                                                                            ef_);
                 top_candidates.swap(top_candidates1);
@@ -624,8 +617,6 @@ namespace hnswlib {
         }
 
         void loadIndex(const std::string &location, SpaceInterface<dist_t> *s, size_t max_elements_i=0) {
-
-
             std::ifstream input(location, std::ios::binary);
 
             if (!input.is_open())
@@ -640,7 +631,7 @@ namespace hnswlib {
             readBinaryPOD(input, max_elements_);
             readBinaryPOD(input, cur_element_count);
 
-            size_t max_elements=max_elements_i;
+            size_t max_elements = max_elements_i;
             if(max_elements < cur_element_count)
                 max_elements = max_elements_;
             max_elements_ = max_elements;
@@ -689,26 +680,19 @@ namespace hnswlib {
 
             input.seekg(pos,input.beg);
 
-
             data_level0_memory_ = (char *) malloc(max_elements * size_data_per_element_);
             if (data_level0_memory_ == nullptr)
                 throw std::runtime_error("Not enough memory: loadIndex failed to allocate level0");
             input.read(data_level0_memory_, cur_element_count * size_data_per_element_);
 
-
-
-
             size_links_per_element_ = maxM_ * sizeof(tableint) + sizeof(linklistsizeint);
 
-
             size_links_level0_ = maxM0_ * sizeof(tableint) + sizeof(linklistsizeint);
             std::vector<std::mutex>(max_elements).swap(link_list_locks_);
             std::vector<std::mutex>(max_update_element_locks).swap(link_list_update_locks_);
 
-
             visited_list_pool_ = new VisitedListPool(1, max_elements);
 
-
             linkLists_ = (char **) malloc(sizeof(void *) * max_elements);
             if (linkLists_ == nullptr)
                 throw std::runtime_error("Not enough memory: loadIndex failed to allocate linklists");
@@ -732,11 +716,9 @@ namespace hnswlib {
                 }
             }
 
-            has_deletions_=false;
-
             for (size_t i = 0; i < cur_element_count; i++) {
                 if(isMarkedDeleted(i))
-                    has_deletions_=true;
+                    num_deleted_ += 1;
             }
 
             input.close();
@@ -745,7 +727,7 @@ namespace hnswlib {
         }
 
         template<typename data_t>
-        std::vector<data_t> getDataByLabel(labeltype label)
+        std::vector<data_t> getDataByLabel(labeltype label) const
         {
             tableint label_c;
             auto search = label_lookup_.find(label);
@@ -758,7 +740,7 @@ namespace hnswlib {
             size_t dim = *((size_t *) dist_func_param_);
             std::vector<data_t> data;
             data_t* data_ptr = (data_t*) data_ptrv;
-            for (size_t i = 0; i < dim; i++) {
+            for (int i = 0; i < dim; i++) {
                 data.push_back(*data_ptr);
                 data_ptr += 1;
             }
@@ -766,19 +748,19 @@ namespace hnswlib {
         }
 
         static const unsigned char DELETE_MARK = 0x01;
-//        static const unsigned char REUSE_MARK = 0x10;
+        // static const unsigned char REUSE_MARK = 0x10;
         /**
          * Marks an element with the given label deleted, does NOT really change the current graph.
          * @param label
          */
         void markDelete(labeltype label)
         {
-            has_deletions_=true;
             auto search = label_lookup_.find(label);
             if (search == label_lookup_.end()) {
                 throw std::runtime_error("Label not found");
             }
-            markDeletedInternal(search->second);
+            tableint internalId = search->second;
+            markDeletedInternal(internalId);
         }
 
         /**
@@ -787,8 +769,31 @@ namespace hnswlib {
          * @param internalId
          */
         void markDeletedInternal(tableint internalId) {
-            unsigned char *ll_cur = ((unsigned char *)get_linklist0(internalId))+2;
-            *ll_cur |= DELETE_MARK;
+            assert(internalId < cur_element_count);
+            if (!isMarkedDeleted(internalId))
+            {
+                unsigned char *ll_cur = ((unsigned char *)get_linklist0(internalId))+2;
+                *ll_cur |= DELETE_MARK;
+                num_deleted_ += 1;
+            }
+            else
+            {
+                throw std::runtime_error("The requested to delete element is already deleted");
+            }
+        }
+
+        /**
+         * Remove the deleted mark of the node, does NOT really change the current graph.
+         * @param label
+         */
+        void unmarkDelete(labeltype label)
+        {
+            auto search = label_lookup_.find(label);
+            if (search == label_lookup_.end()) {
+                throw std::runtime_error("Label not found");
+            }
+            tableint internalId = search->second;
+            unmarkDeletedInternal(internalId);
         }
 
         /**
@@ -796,8 +801,17 @@ namespace hnswlib {
          * @param internalId
          */
         void unmarkDeletedInternal(tableint internalId) {
-            unsigned char *ll_cur = ((unsigned char *)get_linklist0(internalId))+2;
-            *ll_cur &= ~DELETE_MARK;
+            assert(internalId < cur_element_count);
+            if (isMarkedDeleted(internalId))
+            {
+                unsigned char *ll_cur = ((unsigned char *)get_linklist0(internalId))+2;
+                *ll_cur &= ~DELETE_MARK;
+                num_deleted_ -= 1;
+            }
+            else
+            {
+                throw std::runtime_error("The requested to undelete element is not deleted");
+            }
         }
 
         /**
@@ -858,8 +872,8 @@ namespace hnswlib {
                 }
 
                 for (auto&& neigh : sNeigh) {
-//                    if (neigh == internalId)
-//                        continue;
+                    // if (neigh == internalId)
+                    //     continue;
 
                     std::priority_queue<std::pair<dist_t, tableint>, std::vector<std::pair<dist_t, tableint>>, CompareByFirst> candidates;
                     size_t size = sCand.find(neigh) == sCand.end() ? sCand.size() : sCand.size() - 1; // sCand guaranteed to have size >= 1
@@ -990,7 +1004,7 @@ namespace hnswlib {
                         unmarkDeletedInternal(existingInternalId);
                     }
                     updatePoint(data_point, existingInternalId, 1.0);
-                    
+
                     return existingInternalId;
                 }
 
@@ -1134,7 +1148,7 @@ namespace hnswlib {
             }
 
             std::priority_queue<std::pair<dist_t, tableint>, std::vector<std::pair<dist_t, tableint>>, CompareByFirst> top_candidates;
-            if (has_deletions_) {
+            if (num_deleted_) {
                 top_candidates=searchBaseLayerST<true,true>(
                         currObj, query_data, std::max(ef_, k));
             }

data/ext/hnswlib/src/hnswlib.h

@@ -4,6 +4,9 @@
 #define USE_SSE
 #ifdef __AVX__
 #define USE_AVX
+#ifdef __AVX512F__
+#define USE_AVX512
+#endif
 #endif
 #endif
 #endif
@@ -12,15 +15,97 @@
 #ifdef _MSC_VER
 #include <intrin.h>
 #include <stdexcept>
+#include "cpu_x86.h"
+void cpu_x86::cpuid(int32_t out[4], int32_t eax, int32_t ecx) {
+    __cpuidex(out, eax, ecx);
+}
+__int64 xgetbv(unsigned int x) {
+    return _xgetbv(x);
+}
 #else
 #include <x86intrin.h>
+#include <cpuid.h>
+#include <stdint.h>
+void cpuid(int32_t cpuInfo[4], int32_t eax, int32_t ecx) {
+    __cpuid_count(eax, ecx, cpuInfo[0], cpuInfo[1], cpuInfo[2], cpuInfo[3]);
+}
+uint64_t xgetbv(unsigned int index) {
+    uint32_t eax, edx;
+    __asm__ __volatile__("xgetbv" : "=a"(eax), "=d"(edx) : "c"(index));
+    return ((uint64_t)edx << 32) | eax;
+}
+#endif
+
+#if defined(USE_AVX512)
+#include <immintrin.h>
 #endif
 
 #if defined(__GNUC__)
 #define PORTABLE_ALIGN32 __attribute__((aligned(32)))
+#define PORTABLE_ALIGN64 __attribute__((aligned(64)))
 #else
 #define PORTABLE_ALIGN32 __declspec(align(32))
+#define PORTABLE_ALIGN64 __declspec(align(64))
 #endif
+
+// Adapted from https://github.com/Mysticial/FeatureDetector
+#define _XCR_XFEATURE_ENABLED_MASK  0
+
+bool AVXCapable() {
+    int cpuInfo[4];
+
+    // CPU support
+    cpuid(cpuInfo, 0, 0);
+    int nIds = cpuInfo[0];
+
+    bool HW_AVX = false;
+    if (nIds >= 0x00000001) {
+        cpuid(cpuInfo, 0x00000001, 0);
+        HW_AVX = (cpuInfo[2] & ((int)1 << 28)) != 0;
+    }
+
+    // OS support
+    cpuid(cpuInfo, 1, 0);
+
+    bool osUsesXSAVE_XRSTORE = (cpuInfo[2] & (1 << 27)) != 0;
+    bool cpuAVXSuport = (cpuInfo[2] & (1 << 28)) != 0;
+
+    bool avxSupported = false;
+    if (osUsesXSAVE_XRSTORE && cpuAVXSuport) {
+        uint64_t xcrFeatureMask = xgetbv(_XCR_XFEATURE_ENABLED_MASK);
+        avxSupported = (xcrFeatureMask & 0x6) == 0x6;
+    }
+    return HW_AVX && avxSupported;
+}
+
+bool AVX512Capable() {
+    if (!AVXCapable()) return false;
+
+    int cpuInfo[4];
+
+    // CPU support
+    cpuid(cpuInfo, 0, 0);
+    int nIds = cpuInfo[0];
+
+    bool HW_AVX512F = false;
+    if (nIds >= 0x00000007) { //  AVX512 Foundation
+        cpuid(cpuInfo, 0x00000007, 0);
+        HW_AVX512F = (cpuInfo[1] & ((int)1 << 16)) != 0;
+    }
+
+    // OS support
+    cpuid(cpuInfo, 1, 0);
+
+    bool osUsesXSAVE_XRSTORE = (cpuInfo[2] & (1 << 27)) != 0;
+    bool cpuAVXSuport = (cpuInfo[2] & (1 << 28)) != 0;
+
+    bool avx512Supported = false;
+    if (osUsesXSAVE_XRSTORE && cpuAVXSuport) {
+        uint64_t xcrFeatureMask = xgetbv(_XCR_XFEATURE_ENABLED_MASK);
+        avx512Supported = (xcrFeatureMask & 0xe6) == 0xe6;
+    }
+    return HW_AVX512F && avx512Supported;
+}
 #endif
 
 #include <queue>
@@ -99,7 +184,6 @@ namespace hnswlib {
 
         return result;
     }
-
 }
 
 #include "space_l2.h"

data/ext/hnswlib/src/space_ip.h

@@ -10,15 +10,20 @@ namespace hnswlib {
         for (unsigned i = 0; i < qty; i++) {
             res += ((float *) pVect1)[i] * ((float *) pVect2)[i];
         }
-        return (1.0f - res);
+        return res;
 
     }
 
+    static float
+    InnerProductDistance(const void *pVect1, const void *pVect2, const void *qty_ptr) {
+        return 1.0f - InnerProduct(pVect1, pVect2, qty_ptr);
+    }
+
 #if defined(USE_AVX)
 
 // Favor using AVX if available.
     static float
-    InnerProductSIMD4Ext(const void *pVect1v, const void *pVect2v, const void *qty_ptr) {
+    InnerProductSIMD4ExtAVX(const void *pVect1v, const void *pVect2v, const void *qty_ptr) {
         float PORTABLE_ALIGN32 TmpRes[8];
         float *pVect1 = (float *) pVect1v;
         float *pVect2 = (float *) pVect2v;
@@ -61,13 +66,20 @@ namespace hnswlib {
 
         _mm_store_ps(TmpRes, sum_prod);
         float sum = TmpRes[0] + TmpRes[1] + TmpRes[2] + TmpRes[3];;
-        return 1.0f - sum;
-}
+        return sum;
+    }
 
-#elif defined(USE_SSE)
+    static float
+    InnerProductDistanceSIMD4ExtAVX(const void *pVect1v, const void *pVect2v, const void *qty_ptr) {
+        return 1.0f - InnerProductSIMD4ExtAVX(pVect1v, pVect2v, qty_ptr);
+    }
+
+#endif
+
+#if defined(USE_SSE)
 
     static float
-    InnerProductSIMD4Ext(const void *pVect1v, const void *pVect2v, const void *qty_ptr) {
+    InnerProductSIMD4ExtSSE(const void *pVect1v, const void *pVect2v, const void *qty_ptr) {
         float PORTABLE_ALIGN32 TmpRes[8];
         float *pVect1 = (float *) pVect1v;
         float *pVect2 = (float *) pVect2v;
@@ -119,7 +131,52 @@ namespace hnswlib {
         _mm_store_ps(TmpRes, sum_prod);
         float sum = TmpRes[0] + TmpRes[1] + TmpRes[2] + TmpRes[3];
 
-        return 1.0f - sum;
+        return sum;
+    }
+
+    static float
+    InnerProductDistanceSIMD4ExtSSE(const void *pVect1v, const void *pVect2v, const void *qty_ptr) {
+        return 1.0f - InnerProductSIMD4ExtSSE(pVect1v, pVect2v, qty_ptr);
+    }
+
+#endif
+
+
+#if defined(USE_AVX512)
+
+    static float
+    InnerProductSIMD16ExtAVX512(const void *pVect1v, const void *pVect2v, const void *qty_ptr) {
+        float PORTABLE_ALIGN64 TmpRes[16];
+        float *pVect1 = (float *) pVect1v;
+        float *pVect2 = (float *) pVect2v;
+        size_t qty = *((size_t *) qty_ptr);
+
+        size_t qty16 = qty / 16;
+
+
+        const float *pEnd1 = pVect1 + 16 * qty16;
+
+        __m512 sum512 = _mm512_set1_ps(0);
+
+        while (pVect1 < pEnd1) {
+            //_mm_prefetch((char*)(pVect2 + 16), _MM_HINT_T0);
+
+            __m512 v1 = _mm512_loadu_ps(pVect1);
+            pVect1 += 16;
+            __m512 v2 = _mm512_loadu_ps(pVect2);
+            pVect2 += 16;
+            sum512 = _mm512_add_ps(sum512, _mm512_mul_ps(v1, v2));
+        }
+
+        _mm512_store_ps(TmpRes, sum512);
+        float sum = TmpRes[0] + TmpRes[1] + TmpRes[2] + TmpRes[3] + TmpRes[4] + TmpRes[5] + TmpRes[6] + TmpRes[7] + TmpRes[8] + TmpRes[9] + TmpRes[10] + TmpRes[11] + TmpRes[12] + TmpRes[13] + TmpRes[14] + TmpRes[15];
+
+        return sum;
+    }
+
+    static float
+    InnerProductDistanceSIMD16ExtAVX512(const void *pVect1v, const void *pVect2v, const void *qty_ptr) {
+        return 1.0f - InnerProductSIMD16ExtAVX512(pVect1v, pVect2v, qty_ptr);
     }
 
 #endif
@@ -127,7 +184,7 @@ namespace hnswlib {
 #if defined(USE_AVX)
 
     static float
-    InnerProductSIMD16Ext(const void *pVect1v, const void *pVect2v, const void *qty_ptr) {
+    InnerProductSIMD16ExtAVX(const void *pVect1v, const void *pVect2v, const void *qty_ptr) {
         float PORTABLE_ALIGN32 TmpRes[8];
         float *pVect1 = (float *) pVect1v;
         float *pVect2 = (float *) pVect2v;
@@ -159,13 +216,20 @@ namespace hnswlib {
         _mm256_store_ps(TmpRes, sum256);
         float sum = TmpRes[0] + TmpRes[1] + TmpRes[2] + TmpRes[3] + TmpRes[4] + TmpRes[5] + TmpRes[6] + TmpRes[7];
 
-        return 1.0f - sum;
+        return sum;
     }
 
-#elif defined(USE_SSE)
+    static float
+    InnerProductDistanceSIMD16ExtAVX(const void *pVect1v, const void *pVect2v, const void *qty_ptr) {
+        return 1.0f - InnerProductSIMD16ExtAVX(pVect1v, pVect2v, qty_ptr);
+    }
+
+#endif
+
+#if defined(USE_SSE)
 
-      static float
-      InnerProductSIMD16Ext(const void *pVect1v, const void *pVect2v, const void *qty_ptr) {
+    static float
+    InnerProductSIMD16ExtSSE(const void *pVect1v, const void *pVect2v, const void *qty_ptr) {
         float PORTABLE_ALIGN32 TmpRes[8];
         float *pVect1 = (float *) pVect1v;
         float *pVect2 = (float *) pVect2v;
@@ -206,14 +270,24 @@ namespace hnswlib {
         _mm_store_ps(TmpRes, sum_prod);
         float sum = TmpRes[0] + TmpRes[1] + TmpRes[2] + TmpRes[3];
 
-        return 1.0f - sum;
+        return sum;
+    }
+
+    static float
+    InnerProductDistanceSIMD16ExtSSE(const void *pVect1v, const void *pVect2v, const void *qty_ptr) {
+        return 1.0f - InnerProductSIMD16ExtSSE(pVect1v, pVect2v, qty_ptr);
     }
 
 #endif
 
-#if defined(USE_SSE) || defined(USE_AVX)
+#if defined(USE_SSE) || defined(USE_AVX) || defined(USE_AVX512)
+    DISTFUNC<float> InnerProductSIMD16Ext = InnerProductSIMD16ExtSSE;
+    DISTFUNC<float> InnerProductSIMD4Ext = InnerProductSIMD4ExtSSE;
+    DISTFUNC<float> InnerProductDistanceSIMD16Ext = InnerProductDistanceSIMD16ExtSSE;
+    DISTFUNC<float> InnerProductDistanceSIMD4Ext = InnerProductDistanceSIMD4ExtSSE;
+
     static float
-    InnerProductSIMD16ExtResiduals(const void *pVect1v, const void *pVect2v, const void *qty_ptr) {
+    InnerProductDistanceSIMD16ExtResiduals(const void *pVect1v, const void *pVect2v, const void *qty_ptr) {
         size_t qty = *((size_t *) qty_ptr);
         size_t qty16 = qty >> 4 << 4;
         float res = InnerProductSIMD16Ext(pVect1v, pVect2v, &qty16);
@@ -222,11 +296,11 @@ namespace hnswlib {
 
         size_t qty_left = qty - qty16;
         float res_tail = InnerProduct(pVect1, pVect2, &qty_left);
-        return res + res_tail - 1.0f;
+        return 1.0f - (res + res_tail);
     }
 
     static float
-    InnerProductSIMD4ExtResiduals(const void *pVect1v, const void *pVect2v, const void *qty_ptr) {
+    InnerProductDistanceSIMD4ExtResiduals(const void *pVect1v, const void *pVect2v, const void *qty_ptr) {
         size_t qty = *((size_t *) qty_ptr);
         size_t qty4 = qty >> 2 << 2;
 
@@ -237,7 +311,7 @@ namespace hnswlib {
         float *pVect2 = (float *) pVect2v + qty4;
         float res_tail = InnerProduct(pVect1, pVect2, &qty_left);
 
-        return res + res_tail - 1.0f;
+        return 1.0f - (res + res_tail);
     }
 #endif
 
@@ -249,16 +323,37 @@ namespace hnswlib {
     public:
         InnerProductSpace() : data_size_(0), dim_(0) { }
         InnerProductSpace(size_t dim) {
-            fstdistfunc_ = InnerProduct;
-    #if defined(USE_AVX) || defined(USE_SSE)
+            fstdistfunc_ = InnerProductDistance;
+    #if defined(USE_AVX) || defined(USE_SSE) || defined(USE_AVX512)
+        #if defined(USE_AVX512)
+            if (AVX512Capable()) {
+                InnerProductSIMD16Ext = InnerProductSIMD16ExtAVX512;
+                InnerProductDistanceSIMD16Ext = InnerProductDistanceSIMD16ExtAVX512;
+            } else if (AVXCapable()) {
+                InnerProductSIMD16Ext = InnerProductSIMD16ExtAVX;
+                InnerProductDistanceSIMD16Ext = InnerProductDistanceSIMD16ExtAVX;
+            }
+        #elif defined(USE_AVX)
+            if (AVXCapable()) {
+                InnerProductSIMD16Ext = InnerProductSIMD16ExtAVX;
+                InnerProductDistanceSIMD16Ext = InnerProductDistanceSIMD16ExtAVX;
+            }
+        #endif
+        #if defined(USE_AVX)
+            if (AVXCapable()) {
+                InnerProductSIMD4Ext = InnerProductSIMD4ExtAVX;
+                InnerProductDistanceSIMD4Ext = InnerProductDistanceSIMD4ExtAVX;
+            }
+        #endif
+
             if (dim % 16 == 0)
-                fstdistfunc_ = InnerProductSIMD16Ext;
+                fstdistfunc_ = InnerProductDistanceSIMD16Ext;
             else if (dim % 4 == 0)
-                fstdistfunc_ = InnerProductSIMD4Ext;
+                fstdistfunc_ = InnerProductDistanceSIMD4Ext;
             else if (dim > 16)
-                fstdistfunc_ = InnerProductSIMD16ExtResiduals;
+                fstdistfunc_ = InnerProductDistanceSIMD16ExtResiduals;
             else if (dim > 4)
-                fstdistfunc_ = InnerProductSIMD4ExtResiduals;
+                fstdistfunc_ = InnerProductDistanceSIMD4ExtResiduals;
     #endif
             dim_ = dim;
             data_size_ = dim * sizeof(float);
@@ -279,5 +374,4 @@ namespace hnswlib {
     ~InnerProductSpace() {}
     };
 
-
 }

data/ext/hnswlib/src/space_l2.h

@@ -19,11 +19,46 @@ namespace hnswlib {
         return (res);
     }
 
+#if defined(USE_AVX512)
+
+    // Favor using AVX512 if available.
+    static float
+    L2SqrSIMD16ExtAVX512(const void *pVect1v, const void *pVect2v, const void *qty_ptr) {
+        float *pVect1 = (float *) pVect1v;
+        float *pVect2 = (float *) pVect2v;
+        size_t qty = *((size_t *) qty_ptr);
+        float PORTABLE_ALIGN64 TmpRes[16];
+        size_t qty16 = qty >> 4;
+
+        const float *pEnd1 = pVect1 + (qty16 << 4);
+
+        __m512 diff, v1, v2;
+        __m512 sum = _mm512_set1_ps(0);
+
+        while (pVect1 < pEnd1) {
+            v1 = _mm512_loadu_ps(pVect1);
+            pVect1 += 16;
+            v2 = _mm512_loadu_ps(pVect2);
+            pVect2 += 16;
+            diff = _mm512_sub_ps(v1, v2);
+            // sum = _mm512_fmadd_ps(diff, diff, sum);
+            sum = _mm512_add_ps(sum, _mm512_mul_ps(diff, diff));
+        }
+
+        _mm512_store_ps(TmpRes, sum);
+        float res = TmpRes[0] + TmpRes[1] + TmpRes[2] + TmpRes[3] + TmpRes[4] + TmpRes[5] + TmpRes[6] +
+                TmpRes[7] + TmpRes[8] + TmpRes[9] + TmpRes[10] + TmpRes[11] + TmpRes[12] +
+                TmpRes[13] + TmpRes[14] + TmpRes[15];
+
+        return (res);
+}
+#endif
+
 #if defined(USE_AVX)
 
     // Favor using AVX if available.
     static float
-    L2SqrSIMD16Ext(const void *pVect1v, const void *pVect2v, const void *qty_ptr) {
+    L2SqrSIMD16ExtAVX(const void *pVect1v, const void *pVect2v, const void *qty_ptr) {
         float *pVect1 = (float *) pVect1v;
         float *pVect2 = (float *) pVect2v;
         size_t qty = *((size_t *) qty_ptr);
@@ -55,10 +90,12 @@ namespace hnswlib {
         return TmpRes[0] + TmpRes[1] + TmpRes[2] + TmpRes[3] + TmpRes[4] + TmpRes[5] + TmpRes[6] + TmpRes[7];
     }
 
-#elif defined(USE_SSE)
+#endif
+
+#if defined(USE_SSE)
 
     static float
-    L2SqrSIMD16Ext(const void *pVect1v, const void *pVect2v, const void *qty_ptr) {
+    L2SqrSIMD16ExtSSE(const void *pVect1v, const void *pVect2v, const void *qty_ptr) {
         float *pVect1 = (float *) pVect1v;
         float *pVect2 = (float *) pVect2v;
         size_t qty = *((size_t *) qty_ptr);
@@ -106,7 +143,9 @@ namespace hnswlib {
     }
 #endif
 
-#if defined(USE_SSE) || defined(USE_AVX)
+#if defined(USE_SSE) || defined(USE_AVX) || defined(USE_AVX512)
+    DISTFUNC<float> L2SqrSIMD16Ext = L2SqrSIMD16ExtSSE;
+
     static float
     L2SqrSIMD16ExtResiduals(const void *pVect1v, const void *pVect2v, const void *qty_ptr) {
         size_t qty = *((size_t *) qty_ptr);
@@ -122,7 +161,7 @@ namespace hnswlib {
 #endif
 
 
-#ifdef USE_SSE
+#if defined(USE_SSE)
     static float
     L2SqrSIMD4Ext(const void *pVect1v, const void *pVect2v, const void *qty_ptr) {
         float PORTABLE_ALIGN32 TmpRes[8];
@@ -175,7 +214,17 @@ namespace hnswlib {
         L2Space() : data_size_(0), dim_(0) { }
         L2Space(size_t dim) {
             fstdistfunc_ = L2Sqr;
-        #if defined(USE_SSE) || defined(USE_AVX)
+    #if defined(USE_SSE) || defined(USE_AVX) || defined(USE_AVX512)
+        #if defined(USE_AVX512)
+            if (AVX512Capable())
+                L2SqrSIMD16Ext = L2SqrSIMD16ExtAVX512;
+            else if (AVXCapable())
+                L2SqrSIMD16Ext = L2SqrSIMD16ExtAVX;
+        #elif defined(USE_AVX)
+            if (AVXCapable())
+                L2SqrSIMD16Ext = L2SqrSIMD16ExtAVX;
+        #endif
+
             if (dim % 16 == 0)
                 fstdistfunc_ = L2SqrSIMD16Ext;
             else if (dim % 4 == 0)
@@ -184,7 +233,7 @@ namespace hnswlib {
                 fstdistfunc_ = L2SqrSIMD16ExtResiduals;
             else if (dim > 4)
                 fstdistfunc_ = L2SqrSIMD4ExtResiduals;
-        #endif
+    #endif
             dim_ = dim;
             data_size_ = dim * sizeof(float);
         }
@@ -279,4 +328,4 @@ namespace hnswlib {
     };
 
 
-}
+}

data/ext/hnswlib/src/visited_list_pool.h

@@ -2,6 +2,7 @@
 
 #include <mutex>
 #include <string.h>
+#include <deque>
 
 namespace hnswlib {
     typedef unsigned short int vl_type;

data/lib/hnswlib/version.rb

@@ -3,8 +3,8 @@
 # Hnswlib.rb provides Ruby bindings for the Hnswlib.
 module Hnswlib
   # The version of Hnswlib.rb you install.
-  VERSION = '0.3.0'
+  VERSION = '0.5.2'
 
   # The version of Hnswlib included with gem.
-  HSWLIB_VERSION = '0.5.2'
+  HSWLIB_VERSION = '0.6.2'
 end

data/sig/hnswlib.rbs

@@ -0,0 +1,69 @@
+module Hnswlib
+  VERSION: ::String
+  HSWLIB_VERSION: ::String
+
+  class HnswIndex
+    attr_reader metric: String
+
+    def initialize: (n_features: Integer n_features, max_item: Integer max_item, ?metric: ::String metric, ?m: ::Integer m, ?ef_construction: ::Integer ef_construction, ?random_seed: ::Integer random_seed) -> void
+    def add_item: (Integer i, Array[Float] v) -> bool
+    def get_item: (Integer i) -> Array[Float]
+    def remove_item: (Integer i) -> void
+    def get_nns_by_item: (Integer i, Integer n, ?include_distances: (true | false) include_distances) -> ([Array[Integer], Array[Float]] | Array[Integer])
+    def get_nns_by_vector: (Array[Float] v, Integer n, ?include_distances: (true | false) include_distances) -> ([Array[Integer], Array[Float]] | Array[Integer])
+    def resize_index: (Integer new_max_item) -> void
+    def set_ef: (Integer ef) -> void
+    def save: (String filename) -> void
+    def load: (String filename) -> void
+    def get_distance: (Integer i, Integer j) -> Float
+    def n_items: () -> Integer
+    def n_features: () -> Integer
+    def max_item: () -> Integer
+  end
+
+  Index: untyped
+
+  class L2Space
+    attr_accessor dim: Integer
+
+    def initialize: (Integer dim) -> void
+    def distance: (Array[Float] a, Array[Float] b) -> Float
+  end
+
+  class InnerProductSpace
+    attr_accessor dim: Integer
+
+    def initialize: (Integer dim) -> void
+    def distance: (Array[Float] a, Array[Float] b) -> Float
+  end
+
+  class BruteforceSearch
+    attr_accessor space: (::Hnswlib::L2Space | ::Hnswlib::InnerProductSpace)
+
+    def initialize: (space: (::Hnswlib::L2Space | ::Hnswlib::InnerProductSpace) space, max_elements: Integer max_elements) -> void
+    def add_point: (Array[Float] arr, Integer idx) -> bool
+    def current_count: () -> Integer
+    def load_index: (String filename) -> void
+    def max_elements: () -> Integer
+    def remove_point: (Integer idx) -> void
+    def save_index: (String filename) -> void
+    def search_knn: (Array[Float] arr, Integer k) -> [Array[Integer], Array[Float]]
+  end
+
+  class HierarchicalNSW
+    attr_accessor space: (::Hnswlib::L2Space | ::Hnswlib::InnerProductSpace)
+
+    def initialize: (space: (::Hnswlib::L2Space | ::Hnswlib::InnerProductSpace) space, max_elements: Integer max_elements, ?m: Integer m, ?ef_construction: Integer ef_construction, ?random_seed: Integer random_seed) -> void
+    def add_point: (Array[Float] arr, Integer idx) -> bool
+    def current_count: () -> Integer
+    def get_ids: () -> Array[Integer]
+    def get_point: (Integer idx) -> Array[Float]
+    def load_index: (String filename) -> void
+    def mark_deleted: (Integer idx) -> void
+    def max_elements: () -> Integer
+    def resize_index: (Integer new_max_elements) -> void
+    def save_index: (String filename) -> void
+    def search_knn: (Array[Float] arr, Integer k) -> [Array[Integer], Array[Float]]
+    def set_ef: (Integer ef) -> void
+  end
+end

metadata

@@ -1,14 +1,14 @@
 --- !ruby/object:Gem::Specification
 name: hnswlib
 version: !ruby/object:Gem::Version
-  version: 0.3.0
+  version: 0.5.2
 platform: ruby
 authors:
 - yoshoku
 autorequire:
 bindir: exe
 cert_chain: []
-date: 2021-08-08 00:00:00.000000000 Z
+date: 2022-02-19 00:00:00.000000000 Z
 dependencies: []
 description: Hnswlib.rb provides Ruby bindings for the Hnswlib.
 email:
@@ -18,15 +18,9 @@ extensions:
 - ext/hnswlib/extconf.rb
 extra_rdoc_files: []
 files:
-- ".github/workflows/build.yml"
-- ".gitignore"
-- ".rspec"
 - CHANGELOG.md
-- CODE_OF_CONDUCT.md
-- Gemfile
 - LICENSE.txt
 - README.md
-- Rakefile
 - ext/hnswlib/extconf.rb
 - ext/hnswlib/hnswlibext.cpp
 - ext/hnswlib/hnswlibext.hpp
@@ -37,9 +31,9 @@ files:
 - ext/hnswlib/src/space_ip.h
 - ext/hnswlib/src/space_l2.h
 - ext/hnswlib/src/visited_list_pool.h
-- hnswlib.gemspec
 - lib/hnswlib.rb
 - lib/hnswlib/version.rb
+- sig/hnswlib.rbs
 homepage: https://github.com/yoshoku/hnswlib.rb
 licenses:
 - Apache-2.0
@@ -62,7 +56,7 @@ required_rubygems_version: !ruby/object:Gem::Requirement
     - !ruby/object:Gem::Version
       version: '0'
 requirements: []
-rubygems_version: 3.2.22
+rubygems_version: 3.3.7
 signing_key:
 specification_version: 4
 summary: Ruby bindings for the Hnswlib.

data/.github/workflows/build.yml

@@ -1,20 +0,0 @@
-name: build
-
-on: [push, pull_request]
-
-jobs:
-  build:
-    runs-on: ubuntu-latest
-    strategy:
-      fail-fast: false
-      matrix:
-        ruby: [ '2.6', '2.7', '3.0' ]
-    steps:
-    - uses: actions/checkout@v2
-    - name: Set up Ruby ${{ matrix.ruby }}
-      uses: ruby/setup-ruby@v1
-      with:
-        ruby-version: ${{ matrix.ruby }}
-        bundler-cache: true
-    - name: Build and test with Rake
-      run: bundle exec rake

data/.gitignore

@@ -1,19 +0,0 @@
-/.bundle/
-/.yardoc
-/_yardoc/
-/coverage/
-/doc/
-/pkg/
-/spec/reports/
-/tmp/
-*.bundle
-*.so
-*.o
-*.a
-mkmf.log
-
-# rspec failure tracking
-.rspec_status
-
-*.ann
-/bin/

data/.rspec

@@ -1,3 +0,0 @@
---format documentation
---color
---require spec_helper

data/CODE_OF_CONDUCT.md

@@ -1,84 +0,0 @@
-# Contributor Covenant Code of Conduct
-
-## Our Pledge
-
-We as members, contributors, and leaders pledge to make participation in our community a harassment-free experience for everyone, regardless of age, body size, visible or invisible disability, ethnicity, sex characteristics, gender identity and expression, level of experience, education, socio-economic status, nationality, personal appearance, race, religion, or sexual identity and orientation.
-
-We pledge to act and interact in ways that contribute to an open, welcoming, diverse, inclusive, and healthy community.
-
-## Our Standards
-
-Examples of behavior that contributes to a positive environment for our community include:
-
-* Demonstrating empathy and kindness toward other people
-* Being respectful of differing opinions, viewpoints, and experiences
-* Giving and gracefully accepting constructive feedback
-* Accepting responsibility and apologizing to those affected by our mistakes, and learning from the experience
-* Focusing on what is best not just for us as individuals, but for the overall community
-
-Examples of unacceptable behavior include:
-
-* The use of sexualized language or imagery, and sexual attention or
-  advances of any kind
-* Trolling, insulting or derogatory comments, and personal or political attacks
-* Public or private harassment
-* Publishing others' private information, such as a physical or email
-  address, without their explicit permission
-* Other conduct which could reasonably be considered inappropriate in a
-  professional setting
-
-## Enforcement Responsibilities
-
-Community leaders are responsible for clarifying and enforcing our standards of acceptable behavior and will take appropriate and fair corrective action in response to any behavior that they deem inappropriate, threatening, offensive, or harmful.
-
-Community leaders have the right and responsibility to remove, edit, or reject comments, commits, code, wiki edits, issues, and other contributions that are not aligned to this Code of Conduct, and will communicate reasons for moderation decisions when appropriate.
-
-## Scope
-
-This Code of Conduct applies within all community spaces, and also applies when an individual is officially representing the community in public spaces. Examples of representing our community include using an official e-mail address, posting via an official social media account, or acting as an appointed representative at an online or offline event.
-
-## Enforcement
-
-Instances of abusive, harassing, or otherwise unacceptable behavior may be reported to the community leaders responsible for enforcement at yoshoku@outlook.com. All complaints will be reviewed and investigated promptly and fairly.
-
-All community leaders are obligated to respect the privacy and security of the reporter of any incident.
-
-## Enforcement Guidelines
-
-Community leaders will follow these Community Impact Guidelines in determining the consequences for any action they deem in violation of this Code of Conduct:
-
-### 1. Correction
-
-**Community Impact**: Use of inappropriate language or other behavior deemed unprofessional or unwelcome in the community.
-
-**Consequence**: A private, written warning from community leaders, providing clarity around the nature of the violation and an explanation of why the behavior was inappropriate. A public apology may be requested.
-
-### 2. Warning
-
-**Community Impact**: A violation through a single incident or series of actions.
-
-**Consequence**: A warning with consequences for continued behavior. No interaction with the people involved, including unsolicited interaction with those enforcing the Code of Conduct, for a specified period of time. This includes avoiding interactions in community spaces as well as external channels like social media. Violating these terms may lead to a temporary or permanent ban.
-
-### 3. Temporary Ban
-
-**Community Impact**: A serious violation of community standards, including sustained inappropriate behavior.
-
-**Consequence**: A temporary ban from any sort of interaction or public communication with the community for a specified period of time. No public or private interaction with the people involved, including unsolicited interaction with those enforcing the Code of Conduct, is allowed during this period. Violating these terms may lead to a permanent ban.
-
-### 4. Permanent Ban
-
-**Community Impact**: Demonstrating a pattern of violation of community standards, including sustained inappropriate behavior,  harassment of an individual, or aggression toward or disparagement of classes of individuals.
-
-**Consequence**: A permanent ban from any sort of public interaction within the community.
-
-## Attribution
-
-This Code of Conduct is adapted from the [Contributor Covenant][homepage], version 2.0,
-available at https://www.contributor-covenant.org/version/2/0/code_of_conduct.html.
-
-Community Impact Guidelines were inspired by [Mozilla's code of conduct enforcement ladder](https://github.com/mozilla/diversity).
-
-[homepage]: https://www.contributor-covenant.org
-
-For answers to common questions about this code of conduct, see the FAQ at
-https://www.contributor-covenant.org/faq. Translations are available at https://www.contributor-covenant.org/translations.

data/Gemfile

@@ -1,10 +0,0 @@
-# frozen_string_literal: true
-
-source 'https://rubygems.org'
-
-# Specify your gem's dependencies in hnswlib.gemspec
-gemspec
-
-gem 'rake', '~> 13.0'
-gem 'rake-compiler', '~> 1.1'
-gem 'rspec', '~> 3.0'

data/Rakefile

@@ -1,17 +0,0 @@
-# frozen_string_literal: true
-
-require 'bundler/gem_tasks'
-require 'rspec/core/rake_task'
-
-RSpec::Core::RakeTask.new(:spec)
-
-require 'rake/extensiontask'
-
-task build: :compile
-
-Rake::ExtensionTask.new('hnswlibext') do |ext|
-  ext.ext_dir = 'ext/hnswlib'
-  ext.lib_dir = 'lib/hnswlib'
-end
-
-task default: %i[clobber compile spec]

data/hnswlib.gemspec

@@ -1,35 +0,0 @@
-# frozen_string_literal: true
-
-require_relative 'lib/hnswlib/version'
-
-Gem::Specification.new do |spec|
-  spec.name          = 'hnswlib'
-  spec.version       = Hnswlib::VERSION
-  spec.authors       = ['yoshoku']
-  spec.email         = ['yoshoku@outlook.com']
-
-  spec.summary       = 'Ruby bindings for the Hnswlib.'
-  spec.description   = 'Hnswlib.rb provides Ruby bindings for the Hnswlib.'
-  spec.homepage      = 'https://github.com/yoshoku/hnswlib.rb'
-  spec.license       = 'Apache-2.0'
-
-  spec.metadata['homepage_uri'] = spec.homepage
-  spec.metadata['source_code_uri'] = spec.homepage
-  spec.metadata['changelog_uri'] = 'https://github.com/yoshoku/hnswlib.rb/blob/main/CHANGELOG.md'
-
-  # Specify which files should be added to the gem when it is released.
-  # The `git ls-files -z` loads the files in the RubyGem that have been added into git.
-  spec.files = Dir.chdir(File.expand_path(__dir__)) do
-    `git ls-files -z`.split("\x0").reject { |f| f.match(%r{\A(?:test|spec|features)/}) }.reject { |f| f.include?('dummy.rb') }
-  end
-  spec.bindir        = 'exe'
-  spec.executables   = spec.files.grep(%r{\Aexe/}) { |f| File.basename(f) }
-  spec.require_paths = ['lib']
-  spec.extensions    = ['ext/hnswlib/extconf.rb']
-
-  # Uncomment to register a new dependency of your gem
-  # spec.add_dependency "example-gem", "~> 1.0"
-
-  # For more information and examples about making a new gem, checkout our
-  # guide at: https://bundler.io/guides/creating_gem.html
-end