blob: 9a39d27915bb84c0cf59653395c88028e666066e (
plain)
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
|
# Maintainer: Daniel Bermond <dbermond@archlinux.org>
# NOTE:
# In order to build the package, you need to manually download the TensorRT
# file from NVIDIA's website (registration required). Place the downloaded
# file in the PKGBUILD directory and run makepkg.
# Download website:
# https://developer.nvidia.com/tensorrt/
pkgbase=tensorrt
pkgname=('tensorrt' 'python-tensorrt' 'tensorrt-doc')
pkgver=8.4.2.4
_ossver=22.08
_cudaver=11.6
_cudnnver=8.4
_protobuf_ver=3.17.3
_pybind11_ver=2.9.2
_graphsurgeon_ver=0.4.6
_uffver=0.6.9
pkgrel=1
pkgdesc='A platform for high-performance deep learning inference on NVIDIA hardware'
arch=('x86_64')
url='https://developer.nvidia.com/tensorrt/'
license=('custom:NVIDIA-SLA' 'Apache')
makedepends=('git' 'cmake' 'cuda' 'cudnn' 'python' 'python-build' 'python-installer' 'python-onnx'
'python-setuptools' 'python-wheel')
source=("local://TensorRT-${pkgver}.Linux.${CARCH}-gnu.cuda-${_cudaver}.cudnn${_cudnnver}.tar.gz"
"git+https://github.com/NVIDIA/TensorRT.git#tag=${_ossver}"
'protobuf-protocolbuffers'::'git+https://github.com/protocolbuffers/protobuf.git'
'cub-nvlabs'::'git+https://github.com/NVlabs/cub.git'
'git+https://github.com/onnx/onnx-tensorrt.git'
'git+https://github.com/onnx/onnx.git'
'git+https://github.com/pybind/pybind11.git'
'git+https://github.com/google/benchmark.git'
"https://github.com/google/protobuf/releases/download/v${_protobuf_ver}/protobuf-cpp-${_protobuf_ver}.tar.gz"
'010-tensorrt-use-local-protobuf-sources.patch'
'020-tensorrt-fix-python.patch'
'TensorRT-SLA.txt')
noextract=("protobuf-cpp-${_protobuf_ver}.tar.gz")
sha256sums=('530d38d631c86a289df465ff0f31b6653b6338aecfa68c2f4c363fac5f81d7b6'
'SKIP'
'SKIP'
'SKIP'
'SKIP'
'SKIP'
'SKIP'
'SKIP'
'51cec99f108b83422b7af1170afd7aeb2dd77d2bcbb7b6bad1f92509e9ccf8cb'
'ea25bb1b188d53cbfbec35d242ab2a2fa8d6009c547c9f5f67bc2f1ad127ceac'
'ab18bcaebc59620a0b97b67c4bfb7646d9310fffbeb15b8441552c600b0212c9'
'a3e33bbee568e03e7f32481df930c78edc95b49b6b667b37db57cedb1967faed')
prepare() {
# tensorrt git submodules
git -C TensorRT submodule init
git -C TensorRT config --local submodule.parsers/onnx.url "${srcdir}/onnx-tensorrt"
git -C TensorRT config --local submodule.third_party/protobuf.url "${srcdir}/protobuf-protocolbuffers"
git -C TensorRT config --local submodule.third_party/cub.url "${srcdir}/cub-nvlabs"
git -C TensorRT submodule update
# onnx-tensorrt git submodule
git -C TensorRT/parsers/onnx submodule init
git -C TensorRT/parsers/onnx config --local submodule.third_party/onnx.url "${srcdir}/onnx"
git -C TensorRT/parsers/onnx submodule update
# onnx git submodules
git -C TensorRT/parsers/onnx/third_party/onnx submodule init
git -C TensorRT/parsers/onnx/third_party/onnx config --local submodule.third_party/pybind11.url "${srcdir}/pybind11"
git -C TensorRT/parsers/onnx/third_party/onnx config --local submodule.third_party/benchmark.url "${srcdir}/benchmark"
git -C TensorRT/parsers/onnx/third_party/onnx submodule update
git -C pybind11 config --local advice.detachedHead false
# protobuf
mkdir -p build/third_party.protobuf/src
cp -af "protobuf-cpp-${_protobuf_ver}.tar.gz" build/third_party.protobuf/src
patch -d TensorRT -Np1 -i "${srcdir}/010-tensorrt-use-local-protobuf-sources.patch"
patch -d TensorRT -Np1 -i "${srcdir}/020-tensorrt-fix-python.patch"
}
build() {
export CXXFLAGS+=' -ffat-lto-objects'
cmake -B build -S TensorRT \
-DBUILD_ONNX_PYTHON:BOOL='ON' \
-DBUILD_SAMPLES:BOOL='OFF' \
-DCMAKE_BUILD_TYPE:STRING='None' \
-DCMAKE_INSTALL_PREFIX:PATH='/usr' \
-DTRT_LIB_DIR="${srcdir}/TensorRT-${pkgver}/lib" \
-DGPU_ARCHS='53 60 61 62 70 72 75 80 86' \
-DPROTOBUF_VERSION="$_protobuf_ver" \
-DCUDA_VERSION="$_cudaver" \
-DCUDNN_VERSION="$_cudnnver" \
-Wno-dev
make -C build
# python bindings
local _pyver
_pyver="$(python -c 'import sys; print("%s.%s" %sys.version_info[0:2])')"
local -x PYTHON_MAJOR_VERSION="${_pyver%%.*}"
local -x PYTHON_MINOR_VERSION="${_pyver#*.}"
local -x TARGET_ARCHITECTURE="$CARCH"
local -x TRT_OSSPATH="${srcdir}/TensorRT"
local -x CUDA_ROOT='/opt/cuda'
local -x ROOT_PATH="${srcdir}/TensorRT"
local -x EXT_PATH="$srcdir"
local -x TRT_NONOSS_ROOT="${srcdir}/TensorRT-${pkgver}"
git -C pybind11 checkout v${_pybind11_ver}
cd TensorRT/python
./build.sh
# python tools (pytorch-quantization currently fails to build)
local _dir
for _dir in onnx-graphsurgeon Polygraphy #pytorch-quantization
do
cd "${srcdir}/TensorRT/tools/${_dir}"
python -m build --wheel --no-isolation
done
}
package_tensorrt() {
depends=('cuda' 'cudnn')
make -C build DESTDIR="$pkgdir" install
install -D -m755 "TensorRT-${pkgver}/bin"/* -t "${pkgdir}/usr/bin"
install -D -m644 "TensorRT-${pkgver}/lib/libnvinfer_builder_resource.so.${pkgver%.*}" -t "${pkgdir}/usr/lib"
install -D -m644 build/libnv{caffeparser,infer_plugin}_static.a -t "${pkgdir}/usr/lib"
cp -dr --no-preserve='ownership' "TensorRT-${pkgver}/include" "${pkgdir}/usr"
cp -dr --no-preserve='ownership' "TensorRT-${pkgver}/lib"/libnv{infer,parsers}{.so*,_static.a} "${pkgdir}/usr/lib"
ln -s "libnvinfer_builder_resource.so.${pkgver%.*}" "${pkgdir}/usr/lib/libnvinfer_builder_resource.so.${pkgver%%.*}"
ln -s "libnvinfer_builder_resource.so.${pkgver%%.*}" "${pkgdir}/usr/lib/libnvinfer_builder_resource.so"
install -D -m644 TensorRT-SLA.txt "${pkgdir}/usr/share/licenses/${pkgname}/LICENSE"
install -D -m644 "TensorRT-${pkgver}/doc/Acknowledgements.txt" "${pkgdir}/usr/share/licenses/${pkgname}/ACKNOWLEDGEMENTS"
install -D -m644 TensorRT/NOTICE -t "${pkgdir}/usr/share/licenses/${pkgname}"
}
package_python-tensorrt() {
pkgdesc+=' (python bindings and tools)'
license=('custom' 'Apache')
depends=('python' 'python-numpy' 'python-pycuda' 'tensorrt')
optdepends=('python-onnx: for onnx_graphsurgeon python module'
'python-onnxruntime: for onnx_graphsurgeon and polygraphy python modules'
'python-protobuf: for polygraphy and uff python modules'
'python-tensorflow-cuda: for graphsurgeon, polygraphy and uff python modules and convert-to-uff tool')
local _dir
for _dir in "TensorRT-${pkgver}"/{graphsurgeon,uff} \
TensorRT/{python/build/dist,tools/{onnx-graphsurgeon,Polygraphy}/dist}
do
cd "${srcdir}/${_dir}"
python -m installer --destdir="$pkgdir" *.whl
done
local _pyver
_pyver="$(python -c 'import sys; print("%s.%s" %sys.version_info[0:2])')"
install -d -m755 "${pkgdir}/usr/share/licenses/${pkgname}"
ln -s "../../../lib/python${_pyver}/site-packages/tensorrt-${pkgver}.dist-info/LICENSE.txt" \
"${pkgdir}/usr/share/licenses/${pkgname}/LICENSE-tensorrt"
ln -s "../../../lib/python${_pyver}/site-packages/graphsurgeon-${_graphsurgeon_ver}.dist-info/LICENSE.txt" \
"${pkgdir}/usr/share/licenses/${pkgname}/LICENSE-graphsurgeon"
ln -s "../../../lib/python${_pyver}/site-packages/uff-${_uffver}.dist-info/LICENSE.txt" \
"${pkgdir}/usr/share/licenses/${pkgname}/LICENSE-uff"
install -D -m644 "${srcdir}/TensorRT-SLA.txt" "${pkgdir}/usr/share/licenses/${pkgname}/LICENSE-NVIDIA-SLA"
install -D -m644 "${srcdir}/TensorRT-${pkgver}/doc/Acknowledgements.txt" "${pkgdir}/usr/share/licenses/${pkgname}/ACKNOWLEDGEMENTS"
install -D -m644 "${srcdir}/TensorRT/NOTICE" -t "${pkgdir}/usr/share/licenses/${pkgname}"
}
package_tensorrt-doc() {
pkgdesc+=' (documentation)'
arch=('any')
license=('custom:NVIDIA-SLA')
install -d -m755 "${pkgdir}/usr/share/doc/${pkgbase}"
cp -dr --no-preserve='ownership' "TensorRT-${pkgver}/doc"/{cpp,python} "${pkgdir}/usr/share/doc/${pkgbase}"
install -D -m644 TensorRT-SLA.txt "${pkgdir}/usr/share/licenses/${pkgname}/LICENSE"
install -D -m644 "TensorRT-${pkgver}/doc/Acknowledgements.txt" "${pkgdir}/usr/share/licenses/${pkgname}/ACKNOWLEDGEMENTS"
}
|