ported the newly added features

.github/workflows/build-artifacts.yml

@@ -28,14 +28,99 @@ on:
 permissions:
   contents: read
 
+env:
+  CUDA_VERSION: "12.9.1"
+  CUDA_ARCHS: "60;61;75;86;89"
+  FORCE_JAVASCRIPT_ACTIONS_TO_NODE24: "true"
+
 jobs:
+  build-windows:
+    name: Windows Win64
+    runs-on: windows-2022
+
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v5
+
+      - name: Install CUDA Toolkit
+        # v0.2.24 only knows about CUDA 12.9.0. Updating the action keeps our requested
+        # toolkit line at 12.9.1 instead of silently downgrading it just to satisfy the installer.
+        uses: Jimver/cuda-toolkit@v0.2.29
+        with:
+          cuda: ${{ env.CUDA_VERSION }}
+          method: network
+
+      - name: Verify CUDA toolchain
+        shell: pwsh
+        run: |
+          & "$env:CUDA_PATH\bin\nvcc.exe" --version
+          cmake --version
+          ninja --version
+
+      - name: Enter MSVC developer environment
+        shell: pwsh
+        run: |
+          $vswhere = "${env:ProgramFiles(x86)}\Microsoft Visual Studio\Installer\vswhere.exe"
+          $installationPath = & $vswhere -latest -products * -requires Microsoft.VisualStudio.Component.VC.Tools.x86.x64 -property installationPath
+          if (-not $installationPath) {
+            throw "Could not locate a Visual Studio installation with C++ tools."
+          }
+
+          $vsDevCmd = Join-Path $installationPath "Common7\Tools\VsDevCmd.bat"
+          if (-not (Test-Path $vsDevCmd)) {
+            throw "Could not locate VsDevCmd.bat at $vsDevCmd"
+          }
+
+          cmd /c "`"$vsDevCmd`" -arch=amd64 -host_arch=amd64 && set" |
+            ForEach-Object {
+              if ($_ -match '^(.*?)=(.*)$') {
+                "$($matches[1])=$($matches[2])" | Out-File -FilePath $env:GITHUB_ENV -Encoding utf8 -Append
+              }
+            }
+
+      - name: Configure
+        shell: pwsh
+        run: |
+          cmake -S . -B build/windows-win64 -G Ninja `
+            -DCMAKE_BUILD_TYPE=RelWithDebInfo `
+            -DIBKEYER_ENABLE_CUDA=ON `
+            -DCMAKE_CUDA_ARCHITECTURES="${env:CUDA_ARCHS}" `
+            -DCMAKE_CUDA_COMPILER="$env:CUDA_PATH\bin\nvcc.exe" `
+            -DCMAKE_CUDA_FLAGS=-allow-unsupported-compiler
+
+      - name: Build
+        shell: pwsh
+        run: cmake --build build/windows-win64 --config RelWithDebInfo --parallel
+
+      - name: Verify bundle
+        shell: pwsh
+        run: |
+          if (-not (Test-Path "IBKeymaster/IBKeyer.ofx.bundle/Contents/Win64/IBKeyer.ofx")) {
+            throw "Windows OFX bundle was not produced."
+          }
+
+      - name: Stage artifact
+        shell: pwsh
+        run: |
+          Remove-Item -Recurse -Force artifacts/windows-win64 -ErrorAction SilentlyContinue
+          New-Item -ItemType Directory -Force -Path artifacts/windows-win64/IBKeyer-Win64-portable | Out-Null
+          Copy-Item -Recurse "IBKeymaster/IBKeyer.ofx.bundle" "artifacts/windows-win64/IBKeyer-Win64-portable/"
+
+      - name: Upload artifact
+        uses: actions/upload-artifact@v6
+        with:
+          name: IBKeyer-Win64
+          path: artifacts/windows-win64/IBKeyer-Win64-portable
+          if-no-files-found: error
+          retention-days: 14
+
   build-macos-universal:
     name: macOS universal
     runs-on: macos-14
 
     steps:
       - name: Checkout
-        uses: actions/checkout@v4
+        uses: actions/checkout@v5
 
       - name: Configure
         shell: bash
@@ -64,7 +149,7 @@ jobs:
              artifacts/macos-universal/IBKeyer-macos-universal-portable/
 
       - name: Upload artifact
-        uses: actions/upload-artifact@v4
+        uses: actions/upload-artifact@v6
         with:
           name: IBKeyer-macos-universal
           path: artifacts/macos-universal/IBKeyer-macos-universal-portable
@@ -77,11 +162,11 @@ jobs:
     container:
       # GitHub's standard Linux runners do not guarantee an nvcc toolchain. Building inside an
       # NVIDIA CUDA devel image keeps the Linux artifact reproducible without needing a GPU.
-      image: nvidia/cuda:12.4.1-devel-ubuntu22.04
+      image: nvidia/cuda:12.9.1-devel-ubuntu22.04
 
     steps:
       - name: Checkout
-        uses: actions/checkout@v4
+        uses: actions/checkout@v5
 
       - name: Install build dependencies
         shell: bash
@@ -96,7 +181,11 @@ jobs:
 
       - name: Configure
         shell: bash
-        run: cmake -S . -B build/linux-x86_64 -DCMAKE_BUILD_TYPE=RelWithDebInfo
+        run: |
+          cmake -S . -B build/linux-x86_64 \
+            -DCMAKE_BUILD_TYPE=RelWithDebInfo \
+            -DIBKEYER_ENABLE_CUDA=ON \
+            -DCMAKE_CUDA_ARCHITECTURES="${CUDA_ARCHS}"
 
       - name: Build
         shell: bash
@@ -115,7 +204,7 @@ jobs:
              artifacts/linux-x86_64/IBKeyer-linux-x86_64-portable/IBKeyer.ofx.bundle/Contents/Linux-x86-64/
 
       - name: Upload artifact
-        uses: actions/upload-artifact@v4
+        uses: actions/upload-artifact@v6
         with:
           name: IBKeyer-linux-x86_64
           path: artifacts/linux-x86_64/IBKeyer-linux-x86_64-portable

CMakeLists.txt

@@ -47,7 +47,10 @@ if(IBKEYER_ENABLE_CUDA AND (WIN32 OR (UNIX AND NOT APPLE)))
   set(CMAKE_CUDA_STANDARD 17)
   set(CMAKE_CUDA_STANDARD_REQUIRED ON)
   if(NOT DEFINED CMAKE_CUDA_ARCHITECTURES)
-    set(CMAKE_CUDA_ARCHITECTURES 75 86 89)
+    # Pascal support needs sm_60/sm_61 in the fatbin, otherwise a "CUDA-enabled" build can still
+    # fail to launch on older but still very usable NVIDIA cards. Keeping this default aligned with
+    # CI artifacts avoids the trap where GitHub builds support Pascal but local builds silently do not.
+    set(CMAKE_CUDA_ARCHITECTURES 60 61 75 86 89)
   endif()
   if(WIN32)
     set(CMAKE_CUDA_FLAGS "${CMAKE_CUDA_FLAGS} -allow-unsupported-compiler")

CMakePresets.json

@@ -9,14 +9,14 @@
     {
       "name": "windows-vs",
       "displayName": "Windows VC + Ninja",
-        "generator": "Ninja",
-        "binaryDir": "${sourceDir}/build/windows-vc",
-        "cacheVariables": {
-          "CMAKE_BUILD_TYPE": "Release",
-          "IBKEYER_ENABLE_CUDA": "ON",
-          "CMAKE_CUDA_ARCHITECTURES": "75;86;89",
-          "CMAKE_CUDA_COMPILER": "C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v13.1/bin/nvcc.exe",
-          "CMAKE_CUDA_FLAGS": "-allow-unsupported-compiler",
+      "generator": "Ninja",
+      "binaryDir": "${sourceDir}/build/windows-vc",
+      "cacheVariables": {
+        "CMAKE_BUILD_TYPE": "Release",
+        "IBKEYER_ENABLE_CUDA": "ON",
+        "CMAKE_CUDA_ARCHITECTURES": "60;61;75;86;89",
+        "CMAKE_CUDA_COMPILER": "$env{CUDA_PATH}/bin/nvcc.exe",
+        "CMAKE_CUDA_FLAGS": "-allow-unsupported-compiler",
         "CMAKE_MAKE_PROGRAM": "C:/Program Files/Microsoft Visual Studio/18/Community/Common7/IDE/CommonExtensions/Microsoft/CMake/Ninja/ninja.exe"
       }
     },
@@ -27,7 +27,8 @@
       "binaryDir": "${sourceDir}/build/linux-ninja",
       "cacheVariables": {
         "CMAKE_BUILD_TYPE": "RelWithDebInfo",
-        "IBKEYER_ENABLE_CUDA": "ON"
+        "IBKEYER_ENABLE_CUDA": "ON",
+        "CMAKE_CUDA_ARCHITECTURES": "60;61;75;86;89"
       }
     },
     {