usearch 2.8.0 → 2.8.2

package/.github/workflows/release.yml

@@ -56,6 +56,7 @@ jobs:
     runs-on: ubuntu-22.04
     needs: versioning
     strategy:
+      fail-fast: false
       matrix:
         arch: ["linux/amd64", "linux/arm64"]
     steps:
@@ -110,6 +111,7 @@ jobs:
     runs-on: windows-2022
     needs: versioning
     strategy:
+      fail-fast: false
       matrix:
         arch: [x64, x86]
     steps:
@@ -158,6 +160,7 @@ jobs:
     runs-on: macos-12
     needs: versioning
     strategy:
+      fail-fast: false
       matrix:
         arch: [arm64, x86_64]
     steps:
@@ -196,6 +199,7 @@ jobs:
     runs-on: ${{ matrix.os }}
     needs: versioning
     strategy:
+      fail-fast: false
       matrix:
         os: [ubuntu-22.04, macOS-11, windows-2022]
     steps:
@@ -346,6 +350,7 @@ jobs:
     needs: versioning
     runs-on: ${{ matrix.os }}
     strategy:
+      fail-fast: false
       matrix:
         os: [ubuntu-22.04, macOS-11, windows-2022]
 
@@ -537,12 +542,14 @@ jobs:
     name: Build C Dependencies for C#
     needs: versioning
     strategy:
+      fail-fast: false
       matrix:
         os: [ubuntu-22.04, macos-12, windows-2022]
     runs-on: ${{ matrix.os }}
     defaults:
       run:
         shell: bash
+
     steps:
       - uses: actions/checkout@v3
         with:
@@ -554,6 +561,8 @@ jobs:
         run: |
           sudo apt update
           sudo apt install -y cmake build-essential libjemalloc-dev libomp-dev gcc-12 g++-12
+          export CC=gcc-12
+          export CXX=g++-12
           cmake -B build_artifacts -DCMAKE_BUILD_TYPE=Release -DUSEARCH_BUILD_TEST_CPP=0 -DUSEARCH_BUILD_TEST_C=0 -DUSEARCH_BUILD_LIB_C=1 -DUSEARCH_USE_OPENMP=1 -DUSEARCH_USE_SIMSIMD=1 -DUSEARCH_USE_JEMALLOC=1
           cmake --build build_artifacts --config Release
           mkdir -p "${{ github.workspace }}/csharp/lib/runtimes/linux-x64/native"
@@ -577,7 +586,6 @@ jobs:
           cmake --build build_artifacts --config Release
           mkdir -p "${{ github.workspace }}\csharp\lib\runtimes\win-x64\native"
           cp "${{ github.workspace }}\build_artifacts\libusearch_c.dll" "${{ github.workspace }}\csharp\lib\runtimes\win-x64\native"
-        shell: bash
 
       - name: Upload Artifacts
         uses: actions/upload-artifact@v3

package/CITATION.cff

@@ -5,7 +5,7 @@ authors:
   given-names: "Ash"
   orcid: "https://orcid.org/0000-0002-4882-1815"
 title: "USearch by Unum Cloud"
-version: 2.8.0
+version: 2.8.2
 doi: 10.5281/zenodo.7949416
 date-released: 2023-10-22
 url: "https://github.com/unum-cloud/usearch"

package/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "usearch"
-version = "2.8.0"
+version = "2.8.2"
 authors = ["Ash Vardanian <1983160+ashvardanian@users.noreply.github.com>"]
 description = "Smaller & Faster Single-File Vector Search Engine from Unum"
 edition = "2021"

package/README.md

@@ -392,7 +392,7 @@ doi = {10.5281/zenodo.7949416},
 author = {Vardanian, Ash},
 title = {{USearch by Unum Cloud}},
 url = {https://github.com/unum-cloud/usearch},
-version = {2.8.0},
+version = {2.8.2},
 year = {2023},
 month = oct,
 }

package/conanfile.py

@@ -7,7 +7,7 @@ required_conan_version = ">=1.53.0"
 class USearchConan(ConanFile):
 
     name = "usearch"
-    version = '2.8.0'
+    version = '2.8.2'
     license = "Apache-2.0"
     description = "Smaller & Faster Single-File Vector Search Engine from Unum"
     homepage = "https://github.com/unum-cloud/usearch"

package/csharp/nuget/nuget-package.props

@@ -1,7 +1,7 @@
 <?xml version="1.0" encoding="utf-8"?>
 <Project>
   <PropertyGroup>
-    <Version Condition="'$(Version)' == ''">2.8.0</Version>
+    <Version Condition="'$(Version)' == ''">2.8.2</Version>
 
     <Authors>Unum</Authors>
     <Company>Unum</Company>

package/include/usearch/index.hpp

@@ -11,7 +11,7 @@
 
 #define USEARCH_VERSION_MAJOR 2
 #define USEARCH_VERSION_MINOR 8
-#define USEARCH_VERSION_PATCH 0
+#define USEARCH_VERSION_PATCH 2
 
 // Inferring C++ version
 // https://stackoverflow.com/a/61552074

package/package.json

@@ -1,6 +1,6 @@
 {
     "name": "usearch",
-    "version": "2.8.0",
+    "version": "2.8.2",
     "description": "Smaller & Faster Single-File Vector Search Engine from Unum",
     "author": "Ash Vardanian (https://ashvardanian.com/)",
     "license": "Apache 2.0",

package/pyproject.toml

@@ -8,9 +8,13 @@ addopts = ["-ra", "--showlocals", "--strict-markers", "--strict-config"]
 xfail_strict = true
 filterwarnings = ["error"]
 
+# Avoid running tests, as everything is happening in a super slow container
+# We have already run all the relavent Python tests in `prerelease.yml`
+#   test-requires = ["pytest", "numpy"]
+#   test-command = "pytest {project}/python/scripts"
 [tool.cibuildwheel]
-test-requires = ["pytest", "numpy"]
-test-command = "pytest {project}/python/scripts"
+test-requires = []
+test-command = ""
 build-verbosity = 0
 skip = ["*musllinux*", "*i686*", "pp*", "cp36-*", "cp312-*"]
 
@@ -18,22 +22,20 @@ skip = ["*musllinux*", "*i686*", "pp*", "cp36-*", "cp312-*"]
 before-build = [
     "rm -rf {project}/CMakeCache.txt {project}/build {project}/build_debug {project}/CMakeFiles.txt {project}/_deps {project}/.pytest_cache",
     "mkdir -p build/usearch",
-    "git submodule update --init --recursive"
+    "git submodule update --init --recursive",
 ]
 manylinux-x86_64-image = "manylinux_2_28"
 manylinux-aarch64-image = "manylinux_2_28"
 
 archs = ["x86_64", "aarch64"]
-before-all = [
-    "yum install -y glibc-devel wget python3-devel"
-]
+before-all = ["yum install -y glibc-devel wget python3-devel"]
 repair-wheel-command = "auditwheel repair --lib-sdir . -w {dest_dir} {wheel}"
 
 [tool.cibuildwheel.macos]
 before-build = [
     "rm -rf {project}/CMakeCache.txt {project}/build {project}/build_debug {project}/CMakeFiles.txt {project}/_deps {project}/.pytest_cache",
     "mkdir -p build/usearch",
-    "git submodule update --init --recursive"
+    "git submodule update --init --recursive",
 ]
 archs = ["x86_64", "universal2", "arm64"]
 repair-wheel-command = "delocate-wheel --require-archs {delocate_archs} -w {dest_dir} -v {wheel}"
@@ -43,7 +45,7 @@ repair-wheel-command = "delocate-wheel --require-archs {delocate_archs} -w {dest
 before-build = [
     "rd /s /q {project}\\CMakeCache.txt {project}\\build {project}\\build_debug {project}\\CMakeFiles.txt {project}\\_deps {project}\\.pytest_cache || echo Done",
     "md build\\usearch",
-    "git submodule update --init --recursive"
+    "git submodule update --init --recursive",
 ]
 archs = ["x86", "AMD64"]
 skip = ["*win32*", "pp*", "cp36-*", "cp312-*"]

package/rust/lib.rs

@@ -14,6 +14,7 @@ pub mod ffi {
         Cos,
         Pearson,
         Haversine,
+        Divergence,
         Hamming,
         Tanimoto,
         Sorensen,

package/simsimd/.github/workflows/prerelease.yml

@@ -61,20 +61,29 @@ jobs:
 
   test_javascript:
     name: Test JavaScript
-    runs-on: ubuntu-latest
-    strategy:
-      matrix:
-        os: [ubuntu-22.04, macOS-11, windows-2022]
-        node-version: [18.x]
+    runs-on: ubuntu-22.04
+    env:
+      CC: gcc-12
+      CXX: g++-12
+
     steps:
       - uses: actions/checkout@v4
+        with:
+          ref: "main"
+      - run: git submodule update --init --recursive
+
+      - name: Update compilers
+        run: |
+          sudo apt update
+          sudo apt install -y cmake build-essential libjemalloc-dev libomp-dev gcc-12 g++-12
+
       - name: Set up Node.js
         uses: actions/setup-node@v3
         with:
-          node-version: "18.x"
+          node-version: 18
 
-      - name: Build locally
-        run: npm install
-
-      - name: Test
-        run: npm test
+      - name: Build and Test
+        run: |
+          npm install
+          npm ci
+          npm test

package/simsimd/.github/workflows/release.yml

@@ -18,18 +18,18 @@ permissions:
 jobs:
   versioning:
     name: Semantic Release
-    runs-on: ubuntu-latest
+    runs-on: ubuntu-22.04
     steps:
       - uses: actions/checkout@v3
         with:
           persist-credentials: false
       - uses: actions/setup-node@v3
-      - run: npm install --save-dev @semantic-release/exec @semantic-release/git conventional-changelog-eslint semantic-release && npx semantic-release
+      - run: npm install --save-dev --prefix ./package-ci @semantic-release/exec @semantic-release/git conventional-changelog-eslint semantic-release && npx --prefix ./package-ci semantic-release
 
   rebase:
     name: Rebase Dev. Branch
     needs: versioning
-    runs-on: ubuntu-latest
+    runs-on: ubuntu-22.04
     steps:
       - name: Checkout the latest code
         uses: actions/checkout@v3
@@ -109,17 +109,33 @@ jobs:
     name: Publish JavaScript
     needs: versioning
     runs-on: ubuntu-22.04
+    env:
+      CC: gcc-12
+      CXX: g++-12
+
     steps:
       - uses: actions/checkout@v3
         with:
           ref: "main"
       - run: git submodule update --init --recursive
-      - uses: actions/setup-node@v3
+
+      - name: Update compilers
+        run: |
+          sudo apt update
+          sudo apt install -y cmake build-essential libjemalloc-dev libomp-dev gcc-12 g++-12
+
+      - name: Set up Node.js
+        uses: actions/setup-node@v3
         with:
           node-version: 18
-      - run: npm install
-      - run: npm ci
-      - run: npm test
-      - uses: JS-DevTools/npm-publish@v2
+
+      - name: Build and Test
+        run: |
+          npm install
+          npm ci
+          npm test
+
+      - name: Publish
+        uses: JS-DevTools/npm-publish@v2
         with:
           token: ${{ secrets.NPM_TOKEN }}

package/simsimd/CMakeLists.txt

@@ -46,9 +46,9 @@ if(CMAKE_CXX_COMPILER_ID MATCHES "^(Apple)?Clang$")
   if(NOT APPLE)
     add_compile_options(-march=native)
   endif()
-  add_compile_options(-pedantic -ferror-limit=1 -ffast-math)
+  add_compile_options(-pedantic -ferror-limit=1)
 elseif(CMAKE_CXX_COMPILER_ID STREQUAL "GNU")
-  add_compile_options(-march=native -pedantic -fmax-errors=1 -ffast-math)
+  add_compile_options(-march=native -pedantic -fmax-errors=1)
 elseif(CMAKE_CXX_COMPILER_ID STREQUAL "IntelLLVM")
   add_compile_options(-w -ferror-limit=1)
 endif()

package/simsimd/README.md

@@ -34,7 +34,7 @@ Given 1000 embeddings from OpenAI Ada API with 1536 dimensions, running on the A
 
 ### Intel Sapphire Rapids
 
-On the Intel Sapphire Rapids platform, SimSIMD was benchmarked against auto-vectorized code using GCC 12. GCC handles single-precision `float` and `int8_t` well. However, it fails on `_Float16` arrays, which has been part of the C language since 2011.
+On the Intel Sapphire Rapids platform, SimSIMD was benchmarked against auto-vectorized code using GCC 12. GCC handles single-precision `float`, but might not be the best choice for `int8` and `_Float16` arrays, which has been part of the C language since 2011.
 
 |                 | GCC 12 `f32` | GCC 12 `f16` | SimSIMD `f16` | `f16` improvement |
 | :-------------- | -----------: | -----------: | ------------: | ----------------: |

package/simsimd/cpp/bench.cxx

@@ -3,8 +3,8 @@
 
 #include <benchmark/benchmark.h>
 
-#define SIMSIMD_RSQRT sqrtf
-#define SIMSIMD_LOG logf
+#define SIMSIMD_RSQRT(x) (1 / sqrtf(x))
+#define SIMSIMD_LOG(x) (logf(x))
 #include <simsimd/simsimd.h>
 
 namespace bm = benchmark;
@@ -78,7 +78,7 @@ void register_(std::string name, metric_at* distance_func, metric_at* baseline_f
     using pair_bytes_t = vectors_pair_gt<scalar_at, 1536 / sizeof(scalar_at)>;
 
     std::size_t seconds = 10;
-    std::size_t threads = std::thread::hardware_concurrency(); // 1;
+    std::size_t threads = 1; // std::thread::hardware_concurrency(); // 1;
     std::string name_dims = name + "_" + std::to_string(pair_dims_t{}.dimensions()) + "d";
     std::string name_bytes = name + "_" + std::to_string(pair_bytes_t{}.size_bytes()) + "b";
 
@@ -95,7 +95,8 @@ int main(int argc, char** argv) {
     bool compiled_with_sve = false;
     bool compiled_with_neon = false;
     bool compiled_with_avx2 = false;
-    bool compiled_with_avx512popcnt = false;
+    bool compiled_with_avx512vpopcntdq = false;
+    bool compiled_with_avx512vnni = false;
 
 #if defined(__ARM_FEATURE_SVE)
     compiled_with_sve = true;
@@ -107,7 +108,10 @@ int main(int argc, char** argv) {
     compiled_with_avx2 = true;
 #endif
 #if defined(__AVX512VPOPCNTDQ__)
-    compiled_with_avx512popcnt = true;
+    compiled_with_avx512vpopcntdq = true;
+#endif
+#if defined(__AVX512VNNI__)
+    compiled_with_avx512vnni = true;
 #endif
 
     // Log supported functionality
@@ -117,7 +121,8 @@ int main(int argc, char** argv) {
     std::printf("- Arm NEON support enabled: %s\n", flags[compiled_with_neon]);
     std::printf("- Arm SVE support enabled: %s\n", flags[compiled_with_sve]);
     std::printf("- x86 AVX2 support enabled: %s\n", flags[compiled_with_avx2]);
-    std::printf("- x86 AVX512VPOPCNTDQ support enabled: %s\n", flags[compiled_with_avx512popcnt]);
+    std::printf("- x86 AVX512VPOPCNTDQ support enabled: %s\n", flags[compiled_with_avx512vpopcntdq]);
+    std::printf("- x86 AVX512VNNI support enabled: %s\n", flags[compiled_with_avx512vnni]);
     std::printf("\n");
 
     // Run the benchmarks

package/simsimd/include/simsimd/binary.h

@@ -133,41 +133,55 @@ simsimd_sve_b8_jaccard(simsimd_b8_t const* a, simsimd_b8_t const* b, simsimd_siz
 
 #if SIMSIMD_TARGET_X86_AVX512
 
-__attribute__((target("avx512vpopcntdq,avx512vl,avx512bw,avx512f"))) //
+__attribute__((target("avx512vpopcntdq,avx512vl,avx512bw,avx512f,bmi2"))) //
 inline static simsimd_f32_t
 simsimd_avx512_b8_hamming(simsimd_b8_t const* a, simsimd_b8_t const* b, simsimd_size_t n_words) {
     __m512i differences_vec = _mm512_setzero_si512();
-    for (simsimd_size_t i = 0; i < n_words; i += 64) {
-
-        // Compute mask for tail elements
-        __mmask64 mask = (i + 64 <= n_words) ? 0xFFFFFFFFFFFFFFFF : (((1ull << (n_words - i)) - 1ull));
-        __m512i a_vec = _mm512_maskz_loadu_epi8(mask, a + i);
-        __m512i b_vec = _mm512_maskz_loadu_epi8(mask, b + i);
-        __m512i xor_vec = _mm512_xor_si512(a_vec, b_vec);
-
-        differences_vec = _mm512_add_epi64(differences_vec, _mm512_popcnt_epi64(xor_vec));
+    __m512i a_vec, b_vec;
+
+simsimd_avx512_b8_hamming_cycle:
+    if (n_words < 64) {
+        __mmask64 mask = _bzhi_u64(0xFFFFFFFFFFFFFFFF, n_words);
+        a_vec = _mm512_maskz_loadu_epi8(mask, a);
+        b_vec = _mm512_maskz_loadu_epi8(mask, b);
+        n_words = 0;
+    } else {
+        a_vec = _mm512_loadu_epi8(a);
+        b_vec = _mm512_loadu_epi8(b);
+        a += 64, b += 64, n_words -= 64;
     }
+    __m512i xor_vec = _mm512_xor_si512(a_vec, b_vec);
+    differences_vec = _mm512_add_epi64(differences_vec, _mm512_popcnt_epi64(xor_vec));
+    if (n_words)
+        goto simsimd_avx512_b8_hamming_cycle;
 
     simsimd_size_t differences = _mm512_reduce_add_epi64(differences_vec);
     return (simsimd_f32_t)differences;
 }
 
-__attribute__((target("avx512vpopcntdq,avx512vl,avx512bw,avx512f"))) //
+__attribute__((target("avx512vpopcntdq,avx512vl,avx512bw,avx512f,bmi2"))) //
 inline static simsimd_f32_t
 simsimd_avx512_b8_jaccard(simsimd_b8_t const* a, simsimd_b8_t const* b, simsimd_size_t n_words) {
     __m512i intersection_vec = _mm512_setzero_si512(), union_vec = _mm512_setzero_si512();
-    for (simsimd_size_t i = 0; i < n_words; i += 64) {
-
-        // Compute mask for tail elements
-        __mmask64 mask = (i + 64 <= n_words) ? 0xFFFFFFFFFFFFFFFF : (((1ull << (n_words - i)) - 1ull));
-        __m512i a_vec = _mm512_maskz_loadu_epi8(mask, a + i);
-        __m512i b_vec = _mm512_maskz_loadu_epi8(mask, b + i);
-        __m512i and_vec = _mm512_and_si512(a_vec, b_vec);
-        __m512i or_vec = _mm512_or_si512(a_vec, b_vec);
-
-        intersection_vec = _mm512_add_epi64(intersection_vec, _mm512_popcnt_epi64(and_vec));
-        union_vec = _mm512_add_epi64(union_vec, _mm512_popcnt_epi64(or_vec));
+    __m512i a_vec, b_vec;
+
+simsimd_avx512_b8_jaccard_cycle:
+    if (n_words < 64) {
+        __mmask64 mask = _bzhi_u64(0xFFFFFFFFFFFFFFFF, n_words);
+        a_vec = _mm512_maskz_loadu_epi8(mask, a);
+        b_vec = _mm512_maskz_loadu_epi8(mask, b);
+        n_words = 0;
+    } else {
+        a_vec = _mm512_loadu_epi8(a);
+        b_vec = _mm512_loadu_epi8(b);
+        a += 64, b += 64, n_words -= 64;
     }
+    __m512i and_vec = _mm512_and_si512(a_vec, b_vec);
+    __m512i or_vec = _mm512_or_si512(a_vec, b_vec);
+    intersection_vec = _mm512_add_epi64(intersection_vec, _mm512_popcnt_epi64(and_vec));
+    union_vec = _mm512_add_epi64(union_vec, _mm512_popcnt_epi64(or_vec));
+    if (n_words)
+        goto simsimd_avx512_b8_jaccard_cycle;
 
     simsimd_size_t intersection = _mm512_reduce_add_epi64(intersection_vec),
                    union_ = _mm512_reduce_add_epi64(union_vec);