Fix CUDA build with MSVC by enabling /Zc:preprocessor for nvcc host compilation on VS 16.5 or greater

CMakeLists.txt

@@ -198,6 +198,13 @@ if((USE_CUDA OR USE_TRT_RTX) AND CMAKE_CUDA_COMPILER)
     $<$<COMPILE_LANGUAGE:CUDA>:-Wno-deprecated-gpu-targets>
     $<$<AND:$<COMPILE_LANGUAGE:CUDA>,$<CXX_COMPILER_ID:MSVC>>:-Xcompiler=/wd4996>
   )
+
+  # nvcc compiles CUDA code with cl.exe, so we need to pass the /Zc:preprocessor option to avoid compilation errors
+  if(MSVC_VERSION GREATER_EQUAL 1925)
+    add_compile_options(
+      $<$<AND:$<COMPILE_LANGUAGE:CUDA>,$<CXX_COMPILER_ID:MSVC>>:-Xcompiler=/Zc:preprocessor>
+    )
+  endif()
 
   if(WIN32)
     add_library(onnxruntime-genai-cuda SHARED ${generator_cudalib_srcs} "${GENERATORS_ROOT}/dll/onnxruntime-genai.rc")