Create an action runner build for MSVC.

.github/action-scripts/configure-vcvars.ps1

@@ -0,0 +1,20 @@
+$msvcPath = "$(vswhere -products * -latest -property installationPath)"
+
+Write-Output "Loading VC from: $msvcPath\VC\Auxiliary\Build"
+
+Push-Location "$msvcPath\VC\Auxiliary\Build"
+cmd /c "vcvars64.bat & set" |
+    foreach {
+        if ($_ -match "=") {
+            $v = $_.split("="); set-item -force -path "ENV:\$($v[0])"  -value "$($v[1])"
+            echo "$($v[0])=$($v[1])" >> $ENV:GITHUB_ENV
+        }
+    }
+Pop-Location
+
+# Stupid, but makes cl.exe happy
+$global:CC_FP = $(get-command cl).Source.Replace("\","/")
+
+Write-Host "`nVisual Studio Command Prompt variables set." -ForegroundColor Yellow
+Write-Host "Use `$CC_FP as shortcut for Cmake: $CC_FP" -ForegroundColor Yellow
+

.github/action-scripts/fetch-cuda.ps1

@@ -0,0 +1,9 @@
+
+Invoke-WebRequest -Uri "https://developer.download.nvidia.com/compute/cuda/12.1.0/network_installers/cuda_12.1.0_windows_network.exe" -OutFile "./cuda_network.exe"
+Start-Process -Wait -FilePath ./cuda_network.exe -ArgumentList "-s nvcc_12.1 cudart_12.1"
+
+$ENV:PATH="$ENV:PATH;C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v12.1\bin"
+$ENV:CUDA_PATH="C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v12.1"
+
+echo "PATH=$ENV:PATH"      >> $ENV:GITHUB_ENV
+echo "CUDA_PATH=$ENV:CUDA_PATH" >> $ENV:GITHUB_ENV

.github/action-scripts/fetch-lit.ps1

@@ -0,0 +1,7 @@
+$LIT_PATH = "$((gi $((gcm python).path)).Directory.FullName)\Scripts"
+
+python -m pip install wheel
+python -m pip install lit
+
+$ENV:PATH="$ENV:PATH;$LIT_PATH"
+echo "PATH=$ENV:PATH" >> $ENV:GITHUB_ENV

.github/action-scripts/fetch-ninja.ps1

@@ -0,0 +1,6 @@
+# Fetch Ninja, this is a powershell step because of Expand-Archive
+Invoke-WebRequest -Uri "https://github.com/ninja-build/ninja/releases/download/v1.11.1/ninja-win.zip" -OutFile "./ninja.zip"
+Expand-Archive ./ninja.zip
+
+$ENV:PATH="$ENV:PATH;$(pwd)\ninja"
+echo "PATH=$ENV:PATH" >> $ENV:GITHUB_ENV

.github/workflows/main.yml

@@ -0,0 +1,47 @@
+on:
+  pull_request:
+    branches:
+    - main
+
+permissions:
+  contents: read
+  pull-requests: read
+
+# This allows a subsequently queued workflow run to interrupt previous runs
+concurrency:
+  group: '${{ github.workflow }} @ ${{ github.event.pull_request.head.label || github.head_ref || github.ref }}'
+  cancel-in-progress: true
+
+jobs:
+  Lit-NoExecutor:
+    runs-on: windows-latest
+    steps:
+      - uses: actions/checkout@master
+      - uses: actions/setup-python@master
+
+      - name: Install Lit
+        run: .github\action-scripts\fetch-lit.ps1
+
+      - name: Fetch CUDA Toolkit
+        run: .github\action-scripts\fetch-cuda.ps1
+
+      - name: Fetch Ninja
+        run: .github\action-scripts\fetch-ninja.ps1
+
+      - name: Set vcvars
+        run: .github\action-scripts\configure-vcvars.ps1
+
+      - name: Configure CMake project
+        run: |
+          echo "CUDA_PATH: $ENV:CUDA_PATH"
+          cmake -S ./ -B ../build -DLIBCUDACXX_ENABLE_LIBCUDACXX_TESTS=ON -DCMAKE_CXX_COMPILER=cl -DCMAKE_CUDA_COMPILER=nvcc -G Ninja
+
+      - name: Build libcudacxx tests
+        run: |
+          $ENV:LIBCUDACXX_SITE_CONFIG="$(pwd)/../build/test/lit.site.cfg"
+          lit --no-progress-bar -o ./lit_test_results.json -Dcompute_archs=90 -Dexecutor="NoopExecutor()" -Dstd=c++20 -j3 .upstream-tests/test
+
+      - uses: actions/upload-artifact@v3
+        with:
+          name: lit_test_results.json
+          path: ./lit_test_results.json