[笔记] Ubuntu 18.04安装cuda 10及cudnn 7流程

安装环境

• OS：Ubuntu 18.04 64 bit
• 显卡：NVidia GTX 1080

任务：安装 CUDA 10及cuDNN 7

工具下载

NVidia官网下载下列文件：

CUDA 10：cuda_10.0.130_410.48_linux.run
cnDNN 7.4：cudnn-10.0-linux-x64-v7.4.2.24.tgz

安装CUDA

$ sudo sh cuda_10.0.130_410.48_linux.run

先输入accept接受协议，然后按需回答问题即可。

注意：

• 如果当前显卡驱动版本高于CUDA安装包内的驱动，建议跳过这一步，保留原来的显卡驱动即可
• 为了节省空间，sample可以不装

接着将下面内容追加到~/.bashrc：

export PATH=/usr/local/cuda-10.0/bin:$PATH

验证CUDA

使用nvidia-smi查看驱动版本为415.23：

$ nvidia-smi
Thu Jan 24 18:00:52 2019       
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 415.23       Driver Version: 415.23       CUDA Version: 10.0     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|===============================+======================+======================|
|   0  GeForce GTX 1080    Off  | 00000000:01:00.0  On |                  N/A |
| 48%   35C    P8     8W / 180W |    200MiB /  8116MiB |      2%      Default |
+-------------------------------+----------------------+----------------------+
                                                                               
+-----------------------------------------------------------------------------+
| Processes:                                                       GPU Memory |
|  GPU       PID   Type   Process name                             Usage      |
|=============================================================================|
|    0      1462      G   /usr/lib/xorg/Xorg                            83MiB |
|    0      1600      G   /usr/bin/gnome-shell                         100MiB |
|    0      1794      G   /opt/teamviewer/tv_bin/TeamViewer             14MiB |
+-----------------------------------------------------------------------------+

CUDA版本为10.0：

$ nvcc --version
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2018 NVIDIA Corporation
Built on Sat_Aug_25_21:08:01_CDT_2018
Cuda compilation tools, release 10.0, V10.0.130

安装cuDNN

流程是：解压，拷贝，配置环境变量

# 解压
$ tar -zxvf cudnn-10.0-linux-x64-v7.4.2.24.tgz

# 拷贝
$ cd cudnn-10.0-linux-x64-v7.4.2.24
$ sudo cp cuda/include/cudnn.h /usr/local/cuda-10.0/include
$ sudo cp cuda/lib64/libcudnn* /usr/local/cuda-10.0/lib64

# 修改权限
$ sudo chmod a+r /usr/local/cuda-10.0/include/cudnn.h /usr/local/cuda-10.0/lib64/libcudnn*

将下面内容追加到~/.bashrc：

export PATH=/usr/local/cuda-10.0/bin:$PATH
export LD_LIBRARY_PATH=/usr/local/cuda-10.0/lib64:$LD_LIBRARY_PATH
export CUDNN_PATH="/usr/local/cuda-10.0/lib64/libcudnn.so"

验证cnDNN

下面命令不报错就OK。

$ echo -e '#include"cudnn.h"\n void main(){}' | nvcc -x c - -o /dev/null -lcudnn