Merge pull request #250 from ashvardanian/main-dev

Fix: SVE L2 and Arm CI

.github/workflows/prerelease.yml

@@ -225,16 +225,29 @@ jobs:
           CIBW_PLATFORM: linux
           CIBW_ARCHS: x86_64
 
+  # These are maintained by Arm, not GitHub, so we need to include
+  # extra steps, like installing Docker and starting the service
   build_wheels_linux_arm:
     name: Build Python Wheels (Linux ARM)
-    runs-on: ubuntu-24.04-arm # These are maintained by Arm, not GitHub
+    runs-on: ubuntu-24.04-arm
     strategy:
       matrix:
         python-version: ["37", "38", "39", "310", "311", "312", "313"]
     needs: [test_python]
     steps:
       - name: Checkout
         uses: actions/checkout@v4
+      - name: Install Docker CE from Official Repository
+        run: |
+          sudo apt-get update
+          sudo apt-get install -y ca-certificates curl gnupg lsb-release
+          sudo mkdir -p /etc/apt/keyrings
+          curl -fsSL https://download.docker.com/linux/ubuntu/gpg | sudo gpg --dearmor -o /etc/apt/keyrings/docker.gpg
+          echo "deb [arch=$(dpkg --print-architecture) signed-by=/etc/apt/keyrings/docker.gpg] https://download.docker.com/linux/ubuntu $(lsb_release -cs) stable" | sudo tee /etc/apt/sources.list.d/docker.list > /dev/null
+          sudo apt-get update
+          sudo apt-get install -y docker-ce docker-ce-cli containerd.io docker-compose-plugin
+          sudo systemctl start docker
+          docker version
       - name: Install Python
         uses: actions/setup-python@v5
         with:

.github/workflows/release.yml

@@ -101,9 +101,11 @@ jobs:
           path: wheelhouse/*.whl
           overwrite: true
 
+  # These are maintained by Arm, not GitHub, so we need to include
+  # extra steps, like installing Docker and starting the service
   build_wheels_linux_arm:
     name: Build Python Wheels (Linux ARM)
-    runs-on: ubuntu-24.04-arm # If you have access to Arm runners
+    runs-on: ubuntu-24.04-arm
     needs: versioning
     strategy:
       matrix:
@@ -112,6 +114,17 @@ jobs:
       - uses: actions/checkout@v4
         with:
           ref: "main"
+      - name: Install Docker CE from Official Repository
+        run: |
+          sudo apt-get update
+          sudo apt-get install -y ca-certificates curl gnupg lsb-release
+          sudo mkdir -p /etc/apt/keyrings
+          curl -fsSL https://download.docker.com/linux/ubuntu/gpg | sudo gpg --dearmor -o /etc/apt/keyrings/docker.gpg
+          echo "deb [arch=$(dpkg --print-architecture) signed-by=/etc/apt/keyrings/docker.gpg] https://download.docker.com/linux/ubuntu $(lsb_release -cs) stable" | sudo tee /etc/apt/sources.list.d/docker.list > /dev/null
+          sudo apt-get update
+          sudo apt-get install -y docker-ce docker-ce-cli containerd.io docker-compose-plugin
+          sudo systemctl start docker
+          docker version
       - name: Install Python
         uses: actions/setup-python@v5
         with:
@@ -233,7 +246,7 @@ jobs:
       - name: Upload wheels
         uses: actions/upload-artifact@v4
         with:
-          name: cibw-wheels-windows
+          name: cibw-wheels-windows-${{ matrix.architecture }}
           path: wheelhouse/*.whl
           overwrite: true
 

include/simsimd/spatial.h

@@ -970,7 +970,7 @@ SIMSIMD_PUBLIC void simsimd_l2sq_bf16_sve(simsimd_bf16_t const *a_enum, simsimd_
         svfloat32_t a_minus_b_low_vec = svsub_f32_x(pg_low_vec, a_low_vec, b_low_vec);
         svfloat32_t a_minus_b_high_vec = svsub_f32_x(pg_high_vec, a_high_vec, b_high_vec);
         d2_low_vec = svmla_f32_x(pg_vec, d2_low_vec, a_minus_b_low_vec, a_minus_b_low_vec);
-        d2_high_vec = svmla_f32_x(pg_vec, d2_high_vec, a_minus_b_low_vec, a_minus_b_low_vec);
+        d2_high_vec = svmla_f32_x(pg_vec, d2_high_vec, a_minus_b_high_vec, a_minus_b_high_vec);
         i += svcnth();
     } while (i < n);
     simsimd_f32_t d2 = svaddv_f32(svptrue_b32(), d2_low_vec) + svaddv_f32(svptrue_b32(), d2_high_vec);

pyproject.toml

@@ -71,7 +71,9 @@ musllinux-aarch64-image = "musllinux_1_2"
 
 # On CentOS we have to use `yum`.
 # The healthy version would be: `apt-get update && apt-get install -y libc6-dev wget python3-dev`.
-before-all = ["yum update -y && yum install -y glibc-devel wget python3-devel"]
+# We can't even use direct YUM variant: `yum update -y && yum install -y glibc-devel wget python3-devel`
+# as downloading from `extras` fails on Arm instances!
+before-all = ["yum update -y --disablerepo=extras && yum install -y --disablerepo=extras glibc-devel wget python3-devel"]
 
 # With `musl` builds, we obviously don't need the `glibc` and can't use `yum`.
 # This may also be handy for using custom dependencies for different Python versions: