-
Notifications
You must be signed in to change notification settings - Fork 94
/
CMakeLists.txt
111 lines (97 loc) · 6.41 KB
/
CMakeLists.txt
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
# 设置项目
cmake_minimum_required(VERSION 3.15.0) # 设置CMake的最低版本要求
cmake_policy(SET CMP0091 NEW) # 允许在CMake 3.10+中自动设置项目名作为二进制目录名
cmake_policy(SET CMP0146 OLD) # 忽略对find_package的过时警告
project(TensorRT-YOLO VERSION 5.1.0 LANGUAGES CXX CUDA) # 定义项目名称、版本和使用的编程语言（C++和CUDA）
# 设置 C++ 标准
set(CMAKE_CXX_STANDARD 17) # 设置C++标准为17
set(CMAKE_CXX_STANDARD_REQUIRED ON) # 要求C++标准必须满足17
# 添加编译规则
set(CMAKE_EXPORT_COMPILE_COMMANDS ON) # 生成编译数据库，便于代码分析工具使用
# 添加依赖项
## CUDA
find_package(CUDA REQUIRED) # 查找CUDA包
set(CMAKE_CUDA_ARCHITECTURES native) # 自动检测最佳CUDA架构
set(CUDA_PATH ${CUDA_TOOLKIT_ROOT_DIR}) # 缓存CUDA路径
## Pybind11
set(PYBIND11_FINDPYTHON ON) # 开启寻找Python
find_package(Python COMPONENTS Interpreter Development REQUIRED) # 查找Python解释器和开发组件
find_package(pybind11 CONFIG REQUIRED) # 查找pybind11包
# 添加编译选项
option(TENSORRT_PATH "TensorRT Path. Example: /usr/local/tensorrt" "") # 添加编译选项，用于指定TensorRT路径
# 检查 TensorRT 路径是否已设置
if(NOT TENSORRT_PATH)
message(FATAL_ERROR "TensorRT path is not set. Please specify the TensorRT path.") # 如果未设置，则报错
endif()
# 添加公共配置
function(configure_cuda_trt target)
# 添加CUDA定义、包含目录、链接库
target_compile_definitions(${target} PRIVATE ${CUDA_DEFINITIONS})
target_include_directories(${target} PRIVATE ${CUDA_INCLUDE_DIRS})
target_link_libraries(${target} PRIVATE ${CUDA_cudart_LIBRARY})
# 添加TensorRT的包含目录、库目录、链接库
target_include_directories(${target} PRIVATE ${TENSORRT_PATH}/include)
target_link_directories(${target} PRIVATE ${TENSORRT_PATH}/lib)
if(MSVC AND EXISTS ${TENSORRT_PATH}/lib/nvinfer_10.dll)
target_link_libraries(${target} PRIVATE nvinfer_10 nvinfer_plugin_10 nvonnxparser_10)
else()
target_link_libraries(${target} PRIVATE nvinfer nvinfer_plugin nvonnxparser)
endif()
endfunction()
function(add_compile_files target)
# 添加头文件搜索路径
include_directories(${PROJECT_SOURCE_DIR}/include)
file(GLOB_RECURSE SOURCES
${PROJECT_SOURCE_DIR}/source/deploy/core/*.cpp
${PROJECT_SOURCE_DIR}/source/deploy/utils/*.cpp
${PROJECT_SOURCE_DIR}/source/deploy/vision/*.cpp
${PROJECT_SOURCE_DIR}/source/deploy/vision/*.cu
)
# 添加源文件
target_sources(${target} PRIVATE ${SOURCES})
endfunction()
function(set_compile_options target)
if(MSVC)
target_compile_options(${target} PRIVATE $<$<CONFIG:Release>:-O2>) # 对于MSVC，在Release模式下使用-O2优化
set_property(TARGET ${target} PROPERTY MSVC_RUNTIME_LIBRARY "MultiThreaded$<$<CONFIG:Debug>:Debug>") # 设置MSVC运行时库
else()
target_compile_options(${target} PRIVATE $<$<COMPILE_LANGUAGE:CXX>:-O3 -flto=auto>) # 对于非MSVC，使用-O3优化和自动LTO
target_link_options(${target} PRIVATE $<$<COMPILE_LANGUAGE:CXX>:-O3 -flto=auto>) # 链接时同样使用-O3优化和自动LTO
endif()
endfunction()
# 添加子目录
add_subdirectory(${PROJECT_SOURCE_DIR}/plugin) # 添加插件目录
# 定义目标 deploy
add_library(deploy SHARED) # 创建共享库
add_compile_files(deploy) # 添加源文件
configure_cuda_trt(deploy) # 配置CUDA和TensorRT
set_compile_options(deploy) # 设置编译选项
set_target_properties(deploy PROPERTIES OUTPUT_NAME deploy) # 设置输出名称
if(MSVC)
set_target_properties(deploy PROPERTIES RUNTIME_OUTPUT_DIRECTORY_RELEASE ${CMAKE_SOURCE_DIR}/lib) # 设置Release模式下输出目录
set_target_properties(deploy PROPERTIES ARCHIVE_OUTPUT_DIRECTORY_RELEASE ${CMAKE_SOURCE_DIR}/lib)
set_target_properties(deploy PROPERTIES RUNTIME_OUTPUT_DIRECTORY_DEBUG ${CMAKE_SOURCE_DIR}/lib) # 设置Debug模式下输出目录
set_target_properties(deploy PROPERTIES ARCHIVE_OUTPUT_DIRECTORY_DEBUG ${CMAKE_SOURCE_DIR}/lib)
else()
set_target_properties(deploy PROPERTIES LIBRARY_OUTPUT_DIRECTORY ${CMAKE_SOURCE_DIR}/lib) # 设置库文件输出目录
endif()
# 定义目标 pydeploy
pybind11_add_module(pydeploy ${PROJECT_SOURCE_DIR}/source/deploy/pybind/deploy.cpp) # 创建pybind11模块
add_compile_files(pydeploy) # 添加源文件
configure_cuda_trt(pydeploy) # 配置CUDA和TensorRT
set_compile_options(pydeploy) # 设置编译选项
set_target_properties(pydeploy PROPERTIES OUTPUT_NAME pydeploy) # 设置输出名称
if(MSVC)
set_target_properties(pydeploy PROPERTIES LIBRARY_OUTPUT_DIRECTORY_RELEASE ${CMAKE_SOURCE_DIR}/tensorrt_yolo/libs) # 设置Release模式下输出目录
set_target_properties(pydeploy PROPERTIES LIBRARY_OUTPUT_DIRECTORY_DEBUG ${CMAKE_SOURCE_DIR}/tensorrt_yolo/libs) # 设置Debug模式下输出目录
else()
set_target_properties(pydeploy PROPERTIES LIBRARY_OUTPUT_DIRECTORY ${CMAKE_SOURCE_DIR}/tensorrt_yolo/libs) # 设置库文件输出目录
endif()
set_target_properties(pydeploy PROPERTIES
COMPILE_DEFINITIONS "CUDA_PATH=\"${CUDA_TOOLKIT_ROOT_DIR}\""
COMPILE_DEFINITIONS "TENSORRT_PATH=\"${TENSORRT_PATH}\""
) # 设置编译定义
configure_file(
${CMAKE_SOURCE_DIR}/tensorrt_yolo/c_lib_wrap.py.in
${CMAKE_SOURCE_DIR}/tensorrt_yolo/c_lib_wrap.py
) # 配置文件