Prototype of CPU-only execution for vLLM.

Co-authored-by: Kunshang Ji <kunshang.ji@intel.com>

CMakeLists.txt

@@ -0,0 +1,25 @@
+cmake_minimum_required(VERSION 3.15)
+message(STATUS "Building using CMake version: ${CMAKE_VERSION}")
+
+project(PageAttCPU)
+
+set(CMAKE_CXX_STANDARD 17)
+set(CMAKE_CXX_STANDARD_REQUIRED True)
+set(CMAKE_POSITION_INDEPENDENT_CODE ON)
+set(CMAKE_EXPORT_COMPILE_COMMANDS ON)
+
+add_compile_options(-mfma -mavx512f -mavx512bf16 -mavx512vl)
+
+set(PYTHON_LIB_PATH "/usr/local/lib/python3.10/dist-packages/")
+set(PYTHON_INCLUDE_PATH "/usr/include/python3.10")
+
+set(CMAKE_PREFIX_PATH ${CMAKE_PREFIX_PATH} ${PYTHON_LIB_PATH})
+set(CMAKE_MODULE_PATH ${CMAKE_MODULE_PATH} "${CMAKE_CURRENT_SOURCE_DIR}/FindModules")
+
+include_directories(${PYTHON_INCLUDE_PATH})
+include_directories("/usr/local/lib/python3.10/dist-packages/torch/include/")
+include_directories("/usr/local/lib/python3.10/dist-packages/torch/include/torch/csrc/api/include/")
+
+option(USE_STATIC_MKL "Prefer to link with MKL statically (Unix only)" OFF)
+
+add_subdirectory(csrc/cpu)