Improve cmake for benchmarks

Author: ZzEeKkAa

benchmarks/CMakeLists.txt

@@ -10,9 +10,11 @@ if(NOT WIN32)
     list(APPEND CMAKE_MODULE_PATH "${CMAKE_CURRENT_SOURCE_DIR}/cmake")
 endif()
 
-find_package(Python3 COMPONENTS Interpreter)
+find_package(Python3 REQUIRED
+  COMPONENTS Development.Module)
 find_package(Torch REQUIRED)
 find_library(TORCH_PYTHON_LIBRARY torch_python PATH "${TORCH_INSTALL_PREFIX}/lib")
+find_package(XeTLALibrary REQUIRED)
 
 if(USE_IPEX)
   string(APPEND CMAKE_CXX_FLAGS " -DUSE_IPEX")

benchmarks/cmake/FindXeTLALibrary.cmake

@@ -3,13 +3,15 @@
 include(FetchContent)
 
 if (NOT XeTLALibrary_FOUND)
+    # TODO: swithc ot FetchContent_MakeAvailable
+    cmake_policy(SET CMP0169 OLD)
 
     set(XeTLALibrary_SOURCE_DIR
             "${CMAKE_CURRENT_BINARY_DIR}/XeTLALibrary")
     message(STATUS "XeTLALibrary is not specified. Will try to download
                   XeTLA library from https://github.com/intel/xetla into
                   ${XeTLALibrary_SOURCE_DIR}")
-    file(READ xetla-library.conf XeTLALibrary_TAG)
+    file(READ xetla_kernel/xetla-library.conf XeTLALibrary_TAG)
     # Strip the potential trailing newline from tag
     string(STRIP "${XeTLALibrary_TAG}" XeTLALibrary_TAG)
     FetchContent_Declare(xetla-library

benchmarks/setup.py

@@ -2,7 +2,6 @@
 import re
 import shutil
 import subprocess
-import sysconfig
 import sys
 
 from setuptools import setup
@@ -37,40 +36,53 @@ def run(self):
         self.build_extension()
 
     def build_extension(self):
+        # configuration
+        build_type = "Debug"
+
         ninja_dir = shutil.which("ninja")
         # create build directories
         if not os.path.exists(self.build_temp):
             os.makedirs(self.build_temp)
-        # python directories
-        python_include_dir = sysconfig.get_path("platinclude")
         cmake_args = [
             "-G",
             "Ninja",  # Ninja is much faster than make
             "-DCMAKE_MAKE_PROGRAM=" +
             ninja_dir,  # Pass explicit path to ninja otherwise cmake may cache a temporary path
             f"-DCMAKE_PREFIX_PATH={torch.utils.cmake_prefix_path}{ipex_cmake_prefix_path}",
             f"-DUSE_IPEX={USE_IPEX_OPTION}",
-            "-DCMAKE_EXPORT_COMPILE_COMMANDS=ON",
-            "-DCMAKE_ARCHIVE_OUTPUT_DIRECTORY=" + self.extdir,
-            "-DCMAKE_LIBRARY_OUTPUT_DIRECTORY=" + self.extdir,
-            "-DPython3_EXECUTABLE:FILEPATH=" + sys.executable,
-            "-DCMAKE_VERBOSE_MAKEFILE:BOOL=ON",
-            "-DPYTHON_INCLUDE_DIRS=" + python_include_dir,
+            "-DCMAKE_INSTALL_PREFIX=" + self.extdir,
+            "-DPython3_ROOT_DIR:FILEPATH=" + sys.exec_prefix,
+            "-DCMAKE_VERBOSE_MAKEFILE=TRUE",
             "-DCMAKE_C_COMPILER=icx",
             "-DCMAKE_CXX_COMPILER=icpx",
+            "-DCMAKE_BUILD_TYPE=" + build_type,
+            "-S",
+            self.current_dir,
+            "-B",
+            self.build_temp,
         ]
 
-        # configuration
-        build_type = "Debug"
-        build_args = ["--config", build_type]
-        cmake_args += ["-DCMAKE_BUILD_TYPE=" + build_type]
         max_jobs = os.getenv("MAX_JOBS", str(2 * os.cpu_count()))
-        build_args += ["-j" + max_jobs]
+        build_args = [
+            "--build",
+            self.build_temp,
+            "-j" + max_jobs,
+        ]
+
+        install_args = [
+            "--build",
+            self.build_temp,
+            "--target",
+            "install",
+        ]
 
         env = os.environ.copy()
-        cmake_dir = self.build_temp
-        subprocess.check_call(["cmake", self.current_dir] + cmake_args, cwd=cmake_dir, env=env)
-        subprocess.check_call(["cmake", "--build", "."] + build_args, cwd=cmake_dir)
+        print(" ".join(["cmake"] + cmake_args))
+        subprocess.check_call(["cmake"] + cmake_args, env=env)
+        print(" ".join(["cmake"] + build_args))
+        subprocess.check_call(["cmake"] + build_args)
+        print(" ".join(["cmake"] + install_args))
+        subprocess.check_call(["cmake"] + install_args)
 
 
 cmake = CMakeBuild()
@@ -80,4 +92,4 @@ def build_extension(self):
     "triton_kernels_benchmark",
 ], package_dir={
     "triton_kernels_benchmark": "triton_kernels_benchmark",
-}, package_data={"triton_kernels_benchmark": ["xetla_kernel.so"]})
+}, package_data={"triton_kernels_benchmark": ["xetla_kernel.cpython-*.so"]})

benchmarks/xetla_kernel/CMakeLists.txt

@@ -1,7 +1,3 @@
-# XeTLA library is required.
-find_package(XeTLALibrary REQUIRED)
-set(CMAKE_CXX_STANDARD 20)
-
 set(XETLA_KERNEL_FLAGS ${XETLA_KERNEL_FLAGS} -fsycl)
 
 if (USE_AOT_DEVLIST)
@@ -26,8 +22,7 @@ else()
   set(XETLA_KERNEL_FLAGS ${XETLA_KERNEL_FLAGS} "${XETLA_OFFLINE_OPTIONS}")
 endif()
 
-add_library(xetla_kernel SHARED python_main.cpp)
-set_target_properties(xetla_kernel PROPERTIES PREFIX "")
+Python3_add_library(xetla_kernel MODULE WITH_SOABI python_main.cpp)
 target_compile_options(xetla_kernel PRIVATE "-fPIC")
 if(USE_IPEX)
   target_compile_options(xetla_kernel PRIVATE "-fsycl")
@@ -37,7 +32,6 @@ endif()
 target_compile_options(xetla_kernel PUBLIC "-DXETPP_NEW_XMAIN")
 target_link_options(xetla_kernel PRIVATE ${XETLA_KERNEL_FLAGS})
 target_link_libraries(xetla_kernel PUBLIC ${TORCH_LIBRARIES} ${TORCH_PYTHON_LIBRARY})
-target_include_directories(xetla_kernel PUBLIC "${PYTHON_INCLUDE_DIRS}")
 target_include_directories(xetla_kernel PUBLIC "${XeTLALibrary_INCLUDE_DIR}")
 
 if(USE_IPEX)
@@ -49,3 +43,5 @@ add_subdirectory(softmax)
 add_subdirectory(gemm)
 add_subdirectory(stream_k_gemm)
 add_subdirectory(flash_attention)
+
+install(TARGETS xetla_kernel LIBRARY DESTINATION .)