[0.7.3] Update patch timing and CMake to adjust to more user envs (#535)

### What this PR does / why we need it?
To make compile adjust to more user envs.
To make patch timing better to adjust to more conditions.


### Does this PR introduce _any_ user-facing change?
No


### How was this patch tested?
Tested the same as #416 


Signed-off-by: Shuqiao Li <celestialli@outlook.com>

Author: celestialli

CMakeLists.txt

@@ -3,7 +3,7 @@ project(vllm_ascend_C)
 
 # include(CheckCXXcompilerFlag)
 # check_cxx_compiler_flag("-std=c++17", COMPILER_SUPPORTS_CXX17)
-
+set(CMAKE_CXX_STANDARD 17)
 
 include(${CMAKE_CURRENT_LIST_DIR}/cmake/utils.cmake)
 

setup.py

@@ -123,6 +123,10 @@ def configure(self, ext: CMakeExtension) -> None:
         cmake_args += [f"-DCMAKE_BUILD_TYPE={envs.CMAKE_BUILD_TYPE}"]
         # Default dump the compile commands for lsp
         cmake_args += ["-DCMAKE_EXPORT_COMPILE_COMMANDS=1"]
+        if envs.CXX_COMPILER is not None:
+            cmake_args += [f"-DCMAKE_CXX_COMPILER={envs.CXX_COMPILER}"]
+        if envs.C_COMPILER is not None:
+            cmake_args += [f"-DCMAKE_C_COMPILER={envs.C_COMPILER}"]
         if envs.VERBOSE:
             cmake_args += ["-DCMAKE_VERBOSE_MAKEFILE=ON"]
 

vllm_ascend/__init__.py

@@ -19,7 +19,6 @@
 def register():
     """Register the NPU platform."""
     # To ensure that the module is correctly replaced, add it at the beginning
-    import vllm_ascend.patch_config  # noqa: F401
     import vllm_ascend.patch_module  # noqa: F401
     return "vllm_ascend.platform.NPUPlatform"
 

vllm_ascend/device_allocator/camem.py

@@ -65,7 +65,7 @@ def find_loaded_library(lib_name) -> Optional[str]:
         init_module, python_create_and_map, python_unmap_and_release)
     lib_name = find_loaded_library("vllm_ascend_C")
     camem_available = True
-except ModuleNotFoundError as e:
+except ImportError as e:
     logger.error("Failed to import vllm_ascend_C:%s", e)
     init_module = None
     python_create_and_map = None

vllm_ascend/envs.py

@@ -27,7 +27,11 @@
     "LLMDATADIST_COMM_PORT":
     lambda: os.getenv("LLMDATADIST_COMM_PORT", "26000"),
     "LLMDATADIST_SYNC_CACHE_WAIT_TIME":
-    lambda: os.getenv("LLMDATADIST_SYNC_CACHE_WAIT_TIME", "5000")
+    lambda: os.getenv("LLMDATADIST_SYNC_CACHE_WAIT_TIME", "5000"),
+    "CXX_COMPILER":
+    lambda: os.getenv("CXX_COMPILER", None),
+    "C_COMPILER":
+    lambda: os.getenv("C_COMPILER", None)
 }
 
 
@@ -39,4 +43,4 @@ def __getattr__(name: str):
 
 
 def __dir__():
-    return list(env_variables.keys())
+    return list(env_variables.keys())

vllm_ascend/platform.py

@@ -63,6 +63,7 @@ class NPUPlatform(Platform):
     def pre_register_and_update(cls,
                                 parser: Optional[FlexibleArgumentParser] = None
                                 ) -> None:
+        import vllm_ascend.patch_config  # noqa: F401
         from vllm_ascend.quantization.quant_config import \
             AscendQuantConfig  # noqa: F401