Enable MSVC cl.exe build on Windows ARM64

.github/workflows/windows.yml

@@ -94,7 +94,7 @@ jobs:
     - name: Setup compiler
       uses: ilammy/msvc-dev-cmd@v1
       with:
-        arch: amd64
+        arch: arm64
     - name: Setup Ninja
       run: |
         python3 -m pip install --upgrade pip setuptools wheel
@@ -107,3 +107,52 @@ jobs:
       run: cmake --build _build
     - name: Testing xsimd
       run: ./_build/test/test_xsimd
+    - name: Run benchmark
+      run: ./_build/benchmark/benchmark_xsimd
+    - name: Run example
+      run: ./_build/examples/mandelbrot
+
+  build-windows-arm64-clang:
+    name: 'MSYS2 CLANG64 arm64'
+    runs-on: windows-11-arm
+    defaults:
+      run:
+        shell: msys2 {0}
+    steps:
+      - name: Setup MSYS2 with Clang (ARM64)
+        uses: msys2/setup-msys2@v2
+        with:
+          # CLANG64 environment: uses clang/clang++ targeting Windows ARM64 natively.
+          # This is the ARM64-native MSYS2 Clang toolchain — not a cross-compiler.
+          msystem: CLANG64
+          update: true
+          path-type: minimal
+          pacboy: >-
+            cc:p
+            cmake:p
+            ninja:p
+
+      - name: Checkout xsimd
+        uses: actions/checkout@v4
+
+      - name: Configure
+        run: |
+          cmake -B _build \
+            -DBUILD_TESTS=ON \
+            -DDOWNLOAD_DOCTEST=ON \
+            -DBUILD_BENCHMARK=ON \
+            -DBUILD_EXAMPLES=ON \
+            -DCMAKE_BUILD_TYPE=Release \
+            -G Ninja
+
+      - name: Build
+        run: cmake --build _build
+
+      - name: Test xsimd
+        run: ./_build/test/test_xsimd
+
+      - name: Run benchmark
+        run: ./_build/benchmark/benchmark_xsimd
+
+      - name: Run mandelbrot example
+        run: ./_build/examples/mandelbrot

include/xsimd/arch/common/xsimd_common_memory.hpp

@@ -15,6 +15,7 @@
 #include <algorithm>
 #include <array>
 #include <complex>
+#include <cstdio>
 #include <stdexcept>
 
 #include "../../types/xsimd_batch_constant.hpp"
@@ -71,6 +72,9 @@ namespace xsimd
                 for (size_t i = 0; i < sizeof...(Is); ++i)
                     if ((bitmask >> i) & 1u)
                         std::swap(mask_buffer[inserted++], mask_buffer[i]);
+                // Fill remaining positions with the last valid index to avoid undefined behavior
+                for (size_t i = inserted; i < sizeof...(Is); ++i)
+                    mask_buffer[i] = mask_buffer[inserted > 0 ? inserted - 1 : 0];
                 return batch<IT, A>::load_aligned(&mask_buffer[0]);
             }
         }
@@ -85,7 +89,12 @@ namespace xsimd
             auto bitmask = mask.mask();
             auto z = select(mask, x, batch<T, A>((T)0));
             auto compress_mask = detail::create_compress_swizzle_mask<IT, A>(bitmask, std::make_index_sequence<size>());
-            return swizzle(z, compress_mask);
+            alignas(A::alignment()) IT mask_out[size];
+            compress_mask.store_aligned(&mask_out[0]);
+            alignas(A::alignment()) T z_out[size];
+            z.store_aligned(&z_out[0]);
+            auto res = swizzle(z, compress_mask);
+            return res;
         }
 
         // expand