Clustering & Minor Patches in JS, Rust, & Java SDKs

.github/workflows/prerelease.yml

@@ -11,6 +11,8 @@ env:
   PYTHONUTF8: 1
   PYTHON_VERSION: 3.11
   DOTNET_VERSION: 7.0.x
+  ANDROID_NDK_VERSION: 26.3.11579264
+  ANDROID_SDK_VERSION: 21
 
 # Sets permissions of the GITHUB_TOKEN to allow deployment to GitHub Pages
 permissions:
@@ -101,11 +103,15 @@ jobs:
         run: npm test
 
         # Rust
+      - name: Set up Rust
+        run: |
+          rustup update stable
+          rustup default stable
+          rustc -vV
+      - name: Build Rust
+        run: cargo build
       - name: Test Rust
-        uses: actions-rs/toolchain@v1
-        with:
-          toolchain: stable
-          override: true
+        run: cargo test
 
         # Java
       - name: Setup Java
@@ -193,6 +199,17 @@ jobs:
       - name: Test Python
         run: pytest
 
+        # Rust
+      - name: Set up Rust
+        run: |
+          rustup update stable
+          rustup default stable
+          rustc -vV
+      - name: Build Rust
+        run: cargo build
+      - name: Test Rust
+        run: cargo test
+
         # C#
       - name: Setup .NET ${{ env.DOTNET_VERSION }}
         uses: actions/setup-dotnet@v3
@@ -260,6 +277,17 @@ jobs:
       - name: Test ObjC/Swift
         run: swift test
 
+        # Rust
+      - name: Set up Rust
+        run: |
+          rustup update stable
+          rustup default stable
+          rustc -vV
+      - name: Build Rust
+        run: cargo build
+      - name: Test Rust
+        run: cargo test
+
         # C#
       - name: Setup .NET ${{ env.DOTNET_VERSION }}
         uses: actions/setup-dotnet@v3
@@ -318,6 +346,17 @@ jobs:
       - name: Test JavaScript
         run: npm test
 
+        # Rust
+      - name: Set up Rust
+        run: |
+          rustup update stable
+          rustup default stable
+          rustc -vV
+      - name: Build Rust
+        run: cargo build
+      - name: Test Rust
+        run: cargo test
+
         # C#
       - name: Setup .NET ${{ env.DOTNET_VERSION }}
         uses: actions/setup-dotnet@v3
@@ -421,3 +460,65 @@ jobs:
         run: |
           test -e build_artifacts/libusearch_c.so
           test -e build_artifacts/libusearch_sqlite.so
+
+  test_ubuntu_android_ndk:
+    name: Android NDK Build
+    runs-on: ubuntu-22.04
+    strategy:
+      matrix:
+        include:
+          - processor: armv7a
+            abi: armeabi-v7a
+            target: armv7-linux-androideabi
+          - processor: aarch64
+            abi: arm64-v8a
+            target: aarch64-linux-android
+    steps:
+      - uses: actions/checkout@v4
+      - run: git submodule update --init --recursive
+
+      - name: Install NDK ndk
+        run: |
+          ${ANDROID_HOME}/cmdline-tools/latest/bin/sdkmanager --install "ndk;${{ env.ANDROID_NDK_VERSION }}"
+
+      - name: Build C/C++
+        run: |
+          cmake -B build_artifacts \
+            -D CMAKE_BUILD_TYPE=RelWithDebInfo \
+            -D CMAKE_EXPORT_COMPILE_COMMANDS=1 \
+            -D CMAKE_TOOLCHAIN_FILE=${ANDROID_HOME}/ndk/${{ env.ANDROID_NDK_VERSION }}/build/cmake/android.toolchain.cmake \
+            -D CMAKE_ANDROID_STL_TYPE=c++_static \
+            -D ANDROID_PLATFORM=${{ env.ANDROID_SDK_VERSION }} \
+            -D ANDROID_ABI=${{ matrix.abi }} \
+            -D USEARCH_BUILD_LIB_C=1 \
+            -D USEARCH_BUILD_TEST_CPP=0 \
+            -D USEARCH_BUILD_BENCH_CPP=0
+
+          cmake --build build_artifacts --config RelWithDebInfo
+
+        # We can't run the produced builds, but we can make sure they exist
+      - name: Test artifacts presense
+        run: |
+          test -e build_artifacts/libusearch_c.so
+
+        # Rust
+      - name: Set up Rust
+        run: |
+          rustup update stable
+          rustup default stable
+          rustup target add ${{ matrix.target }}
+          rustc -vV
+
+      - name: Set up Rust Env
+        run: |
+          TOOLCHAIN=${ANDROID_HOME}/ndk/${{ env.ANDROID_NDK_VERSION }}/toolchains/llvm/prebuilt/linux-x86_64/bin/
+          NDK_CLANG=$(find ${TOOLCHAIN} -name "${{ matrix.processor }}*${{ env.ANDROID_SDK_VERSION }}-clang")
+          echo "CC_${{ matrix.target }}=${NDK_CLANG}" >> ${GITHUB_ENV}
+          echo "CXX_${{ matrix.target }}=${NDK_CLANG}++" >> ${GITHUB_ENV}
+          echo "AR_${{ matrix.target }}=${TOOLCHAIN}/llvm-ar" >> ${GITHUB_ENV}
+          echo "CARGO_${{ matrix.target }}=${NDK_CLANG}" >> ${GITHUB_ENV}
+          echo "CARGO_${{ matrix.target }}=${TOOLCHAIN}/llvm-ar" >> ${GITHUB_ENV}
+
+      - name: Build Rust
+        run: |
+          cargo build --target ${{ matrix.target }}

.vscode/settings.json

@@ -157,6 +157,7 @@
     "ivdep",
     "jaccard",
     "Jemalloc",
+    "kmeans",
     "Kullback",
     "Leibler",
     "libjemalloc",
@@ -173,6 +174,7 @@
     "nlevels",
     "Numba",
     "numpy",
+    "NVME",
     "objc",
     "OPENMP",
     "preprocess",
@@ -198,6 +200,7 @@
     "uninitialize",
     "unumusearch",
     "usearch",
+    "usecase",
     "usecases",
     "Vardanian",
     "vectorize",

BENCHMARKS.md

@@ -10,11 +10,11 @@ All major HNSW implementation share an identical list of hyper-parameters:
 
 The default values vary drastically.
 
-|  Library  | Connectivity | EF @ A | EF @ S |
-| :-------: | :----------: | :----: | :----: |
-| `hnswlib` |      16      |  200   |   10   |
-|  `FAISS`  |      32      |   40   |   16   |
-| `USearch` |      16      |  128   |   64   |
+| Library   | Connectivity | EF @ A | EF @ S |
+| :-------- | -----------: | -----: | -----: |
+| `hnswlib` |           16 |    200 |     10 |
+| `FAISS`   |           32 |     40 |     16 |
+| `USearch` |           16 |    128 |     64 |
 
 Below are the performance numbers for a benchmark running on the 64 cores of AWS `c7g.metal` "Graviton 3"-based instances.
 The main columns are:
@@ -26,27 +26,27 @@ The main columns are:
 ### Different "connectivity"
 
 | Vectors    | Connectivity | EF @ A | EF @ S | __Add__, QPS | __Search__, QPS | __Recall @1__ |
-| :--------- | :----------: | :----: | :----: | :----------: | :-------------: | ------------: |
-| `f32` x256 |      16      |  128   |   64   |    75'640    |     131'654     |         99.3% |
-| `f32` x256 |      12      |  128   |   64   |    81'747    |     149'728     |         99.0% |
-| `f32` x256 |      32      |  128   |   64   |    64'368    |     104'050     |         99.4% |
+| :--------- | -----------: | -----: | -----: | -----------: | --------------: | ------------: |
+| `f32` x256 |           16 |    128 |     64 |       75'640 |         131'654 |         99.3% |
+| `f32` x256 |           12 |    128 |     64 |       81'747 |         149'728 |         99.0% |
+| `f32` x256 |           32 |    128 |     64 |       64'368 |         104'050 |         99.4% |
 
 ### Different "expansion factors"
 
 | Vectors    | Connectivity | EF @ A | EF @ S | __Add__, QPS | __Search__, QPS | __Recall @1__ |
-| :--------- | :----------: | :----: | :----: | :----------: | :-------------: | ------------: |
-| `f32` x256 |      16      |  128   |   64   |    75'640    |     131'654     |         99.3% |
-| `f32` x256 |      16      |   64   |   32   |   128'644    |     228'422     |         97.2% |
-| `f32` x256 |      16      |  256   |  128   |    39'981    |     69'065      |         99.2% |
+| :--------- | -----------: | -----: | -----: | -----------: | --------------: | ------------: |
+| `f32` x256 |           16 |    128 |     64 |       75'640 |         131'654 |         99.3% |
+| `f32` x256 |           16 |     64 |     32 |      128'644 |         228'422 |         97.2% |
+| `f32` x256 |           16 |    256 |    128 |       39'981 |          69'065 |         99.2% |
 
 ### Different vectors "quantization"
 
 | Vectors      | Connectivity | EF @ A | EF @ S | __Add__, QPS | __Search__, QPS | __Recall @1__ |
-| :----------- | :----------: | :----: | :----: | :----------: | :-------------: | ------------: |
-| `f32` x256   |      16      |  128   |   64   |    87'995    |     171'856     |         99.1% |
-| `f16` x256   |      16      |  128   |   64   |    87'270    |     153'788     |         98.4% |
-| `f16` x256 ✳️ |      16      |  128   |   64   |    71'454    |     132'673     |         98.4% |
-| `i8` x256    |      16      |  128   |   64   |   115'923    |     274'653     |         98.9% |
+| :----------- | -----------: | -----: | -----: | -----------: | --------------: | ------------: |
+| `f32` x256   |           16 |    128 |     64 |       87'995 |         171'856 |         99.1% |
+| `f16` x256   |           16 |    128 |     64 |       87'270 |         153'788 |         98.4% |
+| `f16` x256 ✳️ |           16 |    128 |     64 |       71'454 |         132'673 |         98.4% |
+| `i8` x256    |           16 |    128 |     64 |      115'923 |         274'653 |         98.9% |
 
 As seen on the chart, for `f16` quantization, performance may differ depending on native hardware support for that numeric type.
 Also worth noting, 8-bit quantization results in almost no quantization loss and may perform better than `f16`.
@@ -58,9 +58,12 @@ Within this repository you will find two commonly used utilities:
 - `cpp/bench.cpp` the produces the `bench_cpp` binary for broad USearch benchmarks.
 - `python/bench.py` and `python/bench.ipynb` for interactive charts against FAISS.
 
+### C++ Benchmarking Utilities
+
 To achieve best highest results we suggest compiling locally for the target architecture.
 
 ```sh
+git submodule update --init --recursive
 cmake -USEARCH_BUILD_BENCH_CPP=1 -DUSEARCH_BUILD_TEST_C=1 -DUSEARCH_USE_OPENMP=1 -DUSEARCH_USE_SIMSIMD=1 -DCMAKE_BUILD_TYPE=RelWithDebInfo -B build_profile
 cmake --build build_profile --config RelWithDebInfo -j
 build_profile/bench_cpp --help
@@ -146,11 +149,20 @@ build_profile/bench_cpp \
     --cos
 ```
 
-
 > Optional parameters include `connectivity`, `expansion_add`, `expansion_search`.
 
 For Python, jut open the Jupyter Notebook and start playing around.
 
+### Python Benchmarking Utilities
+
+Several benchmarking suites are available for Python: approximate search, exact search, and clustering.
+
+```sh
+python/scripts/bench.py --help
+python/scripts/bench_exact.py --help
+python/scripts/bench_cluster.py --help
+```
+
 ## Datasets
 
 BigANN benchmark is a good starting point, if you are searching for large collections of high-dimensional vectors.

CONTRIBUTING.md

@@ -200,7 +200,7 @@ cibuildwheel --platform macos                   # works only on MacOS
 cibuildwheel --platform windows                 # works only on Windows
 ```
 
-You may need root previligies for multi-architecture builds:
+You may need root privileges for multi-architecture builds:
 
 ```sh
 sudo $(which cibuildwheel) --platform linux

build.gradle

@@ -2,6 +2,7 @@ import org.gradle.internal.jvm.Jvm
 
 plugins {
     id 'java-library'
+    id 'c'
     id 'cpp'
     id 'maven-publish'
     id 'signing'
@@ -66,6 +67,15 @@ model {
                         srcDirs "include", "fp16/include", "simsimd/include", "${Jvm.current().javaHome}/include"
                     }
                 }
+                c {
+                    source {
+                        srcDirs "simsimd/c/"
+                        include "**/*.c"
+                    }
+                    exportedHeaders {
+                        srcDirs "simsimd/include"
+                    }
+                }
             }
             binaries.withType(StaticLibraryBinarySpec) {
                 buildable = false
@@ -83,6 +93,11 @@ model {
                     cppCompiler.args "-I${Jvm.current().javaHome}/include/win32"
                     cppCompiler.args '/std:c++11'
                 }
+                cppCompiler.args '-DUSEARCH_USE_FP16LIB=1'
+                cppCompiler.args '-DUSEARCH_USE_SIMSIMD=1'
+                cppCompiler.args '-DSIMSIMD_DYNAMIC_DISPATCH=1'
+                cppCompiler.args '-DSIMSIMD_NATIVE_BF16=0'
+                cppCompiler.args '-DSIMSIMD_NATIVE_F16=0'
             }
         }
     }