add win11 arm

Author: chrisoro

.github/workflows/pythonapp.yml

@@ -5,7 +5,6 @@ name: pyXCP
 
 on:
   push:
-    branches: [master, develop]
   pull_request:
     branches: [master, develop]
 
@@ -17,14 +16,19 @@ jobs:
     strategy:
       fail-fast: false
       matrix:
-        os:
-          - macos-latest
-          - ubuntu-24.04-arm
-          - ubuntu-latest
-          - windows-latest
+        include:
+          - os: macos-latest
+          - os: ubuntu-24.04-arm
+          - os: ubuntu-latest
+          - os: windows-11-arm
+            cibw_archs_windows: ARM64
+          - os: windows-latest
+            cibw_archs_windows: AMD64
     steps:
-      - uses: actions/checkout@v4
+      - uses: actions/checkout@v5
       - uses: pypa/[email protected]
+        env:
+          CIBW_ARCHS_WINDOWS: ${{ matrix.cibw_archs_windows || '' }}
       - uses: actions/upload-artifact@v4
         with:
           name: cibw-wheels-${{ matrix.os }}-${{ strategy.job-index }}
@@ -34,7 +38,7 @@ jobs:
     name: Build source distribution
     runs-on: ubuntu-latest
     steps:
-      - uses: actions/checkout@v3
+      - uses: actions/checkout@v5
       - name: Build sdist
         run: |
           pip install -U build
@@ -52,7 +56,7 @@ jobs:
     # alternatively, to publish when a GitHub Release is created, use the following rule:
     # if: github.event_name == 'release' && github.event.action == 'published'
     steps:
-      - uses: actions/download-artifact@v4
+      - uses: actions/download-artifact@v5
         with:
           path: dist
           merge-multiple: true

pyproject.toml

@@ -173,9 +173,6 @@ build = "cp3{10,11,12,13,14}-*"
 skip = ["*_i686", "*-musllinux*"]
 build-frontend = "build"
 
-[tool.cibuildwheel.windows]
-archs = ["AMD64"]
-
 [tool.pyright]
 include = ["pyxcp", "build_ext.py"]
 ignore = ["pyxcp/recorder/converter/**", "pyxcp/recorder/simdjson/**","pyxcp/recorder/mio/**", "pyxcp/recorder/lz4/**"]

pyxcp/daq_stim/scheduler.cpp

@@ -1,28 +1,52 @@
-
 #include "scheduler.hpp"
 
 #if defined(_WIN32)
+#include <cstdio>
+#include <stdexcept>
 
+// Timer callback function
 VOID CALLBACK TimerRoutine(PVOID lpParam, BOOLEAN TimerOrWaitFired) {
-    if (lpParam == NULL) {
-        printf("TimerRoutine lpParam is NULL\n");
+    if (lpParam == nullptr) {
+        printf("TimerRoutine: Invalid parameter (null)\n");
+        return;
+    }
+    
+    const auto* value = static_cast<const int*>(lpParam);
+    printf("Timer routine called. Parameter: %d\n", *value);
+    
+    if (TimerOrWaitFired) {
+        printf("Timer expired\n");
     } else {
-        // lpParam points to the argument; in this case it is an int
-
-        printf("Timer routine called. Parameter is %d.\n", *(int*)lpParam);
-        if (TimerOrWaitFired) {
-            printf("The wait timed out.\n");
-        } else {
-            printf("The wait event was signaled.\n");
-        }
+        printf("Wait event signaled\n");
     }
 }
 
+#endif // _WIN32
+
+// Vectorized square operation (separate from Windows-specific code)
+constexpr size_t VECTOR_SIZE = 4;
+
+void square_vector(float* ptr) {
+    if (ptr == nullptr) {
+        throw std::invalid_argument("Null pointer passed to square_vector");
+    }
+
+#if defined(__SSE__)
     #include <xmmintrin.h>
+    __m128 vec = _mm_loadu_ps(ptr);
+    vec = _mm_mul_ps(vec, vec);
+    _mm_storeu_ps(ptr, vec);
 
-void mul4_vectorized(float* ptr) {
-    __m128 f = _mm_loadu_ps(ptr);
-    f        = _mm_mul_ps(f, f);
-    _mm_storeu_ps(ptr, f);
-}
+#elif defined(__ARM_NEON)
+    #include <arm_neon.h>
+    float32x4_t vec = vld1q_f32(ptr);
+    vec = vmulq_f32(vec, vec);
+    vst1q_f32(ptr, vec);
+
+#else
+    // Scalar fallback
+    for (size_t i = 0; i < VECTOR_SIZE; ++i) {
+        ptr[i] *= ptr[i];
+    }
 #endif
+}