forked from NVIDIA/TensorRT
-
Notifications
You must be signed in to change notification settings - Fork 0
/
CMakeLists.txt
196 lines (161 loc) · 6.91 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
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
#
# Copyright (c) 2021, NVIDIA CORPORATION. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
#
cmake_minimum_required(VERSION 3.13 FATAL_ERROR)
include(cmake/modules/set_ifndef.cmake)
include(cmake/modules/find_library_create_target.cmake)
set_ifndef(TRT_LIB_DIR ${CMAKE_BINARY_DIR})
set_ifndef(TRT_OUT_DIR ${CMAKE_BINARY_DIR})
file(STRINGS "${CMAKE_CURRENT_SOURCE_DIR}/include/NvInferVersion.h" VERSION_STRINGS REGEX "#define NV_TENSORRT_.*")
foreach(TYPE MAJOR MINOR PATCH BUILD)
string(REGEX MATCH "NV_TENSORRT_${TYPE} [0-9]" TRT_TYPE_STRING ${VERSION_STRINGS})
string(REGEX MATCH "[0-9]" TRT_${TYPE} ${TRT_TYPE_STRING})
endforeach(TYPE)
foreach(TYPE MAJOR MINOR PATCH)
string(REGEX MATCH "NV_TENSORRT_SONAME_${TYPE} [0-9]" TRT_TYPE_STRING ${VERSION_STRINGS})
string(REGEX MATCH "[0-9]" TRT_SO_${TYPE} ${TRT_TYPE_STRING})
endforeach(TYPE)
set(TRT_VERSION "${TRT_MAJOR}.${TRT_MINOR}.${TRT_PATCH}" CACHE STRING "TensorRT project version")
set(TRT_SOVERSION "${TRT_SO_MAJOR}" CACHE STRING "TensorRT library so version")
message("Building for TensorRT version: ${TRT_VERSION}, library version: ${TRT_SOVERSION}")
if(NOT DEFINED CMAKE_TOOLCHAIN_FILE)
find_program(CMAKE_CXX_COMPILER NAMES $ENV{CXX} g++)
endif()
set(CMAKE_SKIP_BUILD_RPATH True)
project(TensorRT
LANGUAGES CXX CUDA
VERSION ${TRT_VERSION}
DESCRIPTION "TensorRT is a C++ library that facilitates high performance inference on NVIDIA GPUs and deep learning accelerators."
HOMEPAGE_URL "https://github.com/NVIDIA/TensorRT")
if(CMAKE_INSTALL_PREFIX_INITIALIZED_TO_DEFAULT)
set(CMAKE_INSTALL_PREFIX ${TRT_LIB_DIR}/../ CACHE PATH "TensorRT installation" FORCE)
endif(CMAKE_INSTALL_PREFIX_INITIALIZED_TO_DEFAULT)
option(BUILD_PLUGINS "Build TensorRT plugin" ON)
option(BUILD_PARSERS "Build TensorRT parsers" ON)
option(BUILD_SAMPLES "Build TensorRT samples" ON)
set(CMAKE_CXX_STANDARD 11)
set(CMAKE_CXX_STANDARD_REQUIRED ON)
set(CMAKE_CXX_EXTENSIONS OFF)
set(CMAKE_CXX_FLAGS "-Wno-deprecated-declarations ${CMAKE_CXX_FLAGS} -DBUILD_SYSTEM=cmake_oss")
############################################################################################
# Cross-compilation settings
set_ifndef(TRT_PLATFORM_ID "x86_64")
message(STATUS "Targeting TRT Platform: ${TRT_PLATFORM_ID}")
############################################################################################
# Debug settings
set(TRT_DEBUG_POSTFIX _debug CACHE STRING "suffix for debug builds")
if (CMAKE_BUILD_TYPE STREQUAL "Debug")
message("Building in debug mode ${DEBUG_POSTFIX}")
endif()
############################################################################################
# Dependencies
set(DEFAULT_CUDA_VERSION 11.3.1)
set(DEFAULT_CUDNN_VERSION 8.2)
set(DEFAULT_PROTOBUF_VERSION 3.0.0)
set(DEFAULT_CUB_VERSION 1.8.0)
# Dependency Version Resolution
set_ifndef(CUDA_VERSION ${DEFAULT_CUDA_VERSION})
message(STATUS "CUDA version set to ${CUDA_VERSION}")
set_ifndef(CUDNN_VERSION ${DEFAULT_CUDNN_VERSION})
message(STATUS "cuDNN version set to ${CUDNN_VERSION}")
set_ifndef(PROTOBUF_VERSION ${DEFAULT_PROTOBUF_VERSION})
message(STATUS "Protobuf version set to ${PROTOBUF_VERSION}")
find_package(Threads REQUIRED)
if (BUILD_PLUGINS OR BUILD_PARSERS)
include(third_party/zlib.cmake)
include(third_party/protobuf.cmake)
endif()
if(NOT CUB_ROOT_DIR)
set(CUB_ROOT_DIR ${CMAKE_CURRENT_SOURCE_DIR}/third_party/cub CACHE STRING "directory of CUB installation")
endif()
## find_package(CUDA) is broken for cross-compilation. Enable CUDA language instead.
if(NOT DEFINED CMAKE_TOOLCHAIN_FILE)
find_package(CUDA ${CUDA_VERSION} REQUIRED)
endif()
include_directories(
${CUDA_INCLUDE_DIRS}
${CUDNN_ROOT_DIR}/include
)
find_library(CUDNN_LIB cudnn HINTS
${CUDA_TOOLKIT_ROOT_DIR} ${CUDNN_ROOT_DIR} PATH_SUFFIXES lib64 lib)
find_library(CUBLAS_LIB cublas HINTS
${CUDA_TOOLKIT_ROOT_DIR} PATH_SUFFIXES lib64 lib lib/stubs)
find_library(CUBLASLT_LIB cublasLt HINTS
${CUDA_TOOLKIT_ROOT_DIR} PATH_SUFFIXES lib64 lib lib/stubs)
if(BUILD_PARSERS)
configure_protobuf(${PROTOBUF_VERSION})
endif()
find_library_create_target(nvinfer nvinfer SHARED ${TRT_LIB_DIR})
find_library_create_target(nvuffparser nvparsers SHARED ${TRT_LIB_DIR})
find_library(CUDART_LIB cudart HINTS ${CUDA_TOOLKIT_ROOT_DIR} PATH_SUFFIXES lib lib64)
find_library(RT_LIB rt)
set(CUDA_LIBRARIES ${CUDART_LIB})
############################################################################################
# CUDA targets
if (DEFINED GPU_ARCHS)
message(STATUS "GPU_ARCHS defined as ${GPU_ARCHS}. Generating CUDA code for SM ${GPU_ARCHS}")
separate_arguments(GPU_ARCHS)
else()
list(APPEND GPU_ARCHS
35
53
61
70
75
)
string(REGEX MATCH "aarch64" IS_ARM "${TRT_PLATFORM_ID}")
if (IS_ARM)
# Xavier (SM72) only supported for aarch64.
list(APPEND GPU_ARCHS 72)
endif()
if (CUDA_VERSION VERSION_GREATER_EQUAL 11.0)
# Ampere GPU (SM80) support is only available in CUDA versions > 11.0
list(APPEND GPU_ARCHS 80)
else()
message(WARNING "Detected CUDA version is < 11.0. SM80 not supported.")
endif()
message(STATUS "GPU_ARCHS is not defined. Generating CUDA code for default SMs: ${GPU_ARCHS}")
endif()
set(BERT_GENCODES)
# Generate SASS for each architecture
foreach(arch ${GPU_ARCHS})
if (${arch} GREATER_EQUAL 70)
set(BERT_GENCODES "${BERT_GENCODES} -gencode arch=compute_${arch},code=sm_${arch}")
endif()
set(GENCODES "${GENCODES} -gencode arch=compute_${arch},code=sm_${arch}")
endforeach()
# Generate PTX for the last architecture in the list.
list(GET GPU_ARCHS -1 LATEST_SM)
set(GENCODES "${GENCODES} -gencode arch=compute_${LATEST_SM},code=compute_${LATEST_SM}")
if (${LATEST_SM} GREATER_EQUAL 70)
set(BERT_GENCODES "${BERT_GENCODES} -gencode arch=compute_${LATEST_SM},code=compute_${LATEST_SM}")
endif()
set(CMAKE_CUDA_FLAGS "${CMAKE_CUDA_FLAGS} -Xcompiler -Wno-deprecated-declarations")
############################################################################################
# TensorRT
if(BUILD_PLUGINS)
add_subdirectory(plugin)
else()
find_library_create_target(nvinfer_plugin nvinfer_plugin SHARED ${TRT_OUT_DIR} ${TRT_LIB_DIR})
endif()
if(BUILD_PARSERS)
add_subdirectory(parsers)
else()
find_library_create_target(nvcaffeparser nvparsers SHARED ${TRT_OUT_DIR} ${TRT_LIB_DIR})
find_library_create_target(nvonnxparser nvonnxparser SHARED ${TRT_OUT_DIR} ${TRT_LIB_DIR})
endif()
if(BUILD_SAMPLES)
add_subdirectory(samples)
endif()