CUDA

安装失败，或者重新安装，清理残余文件

a)`.deb`安装失败的

sudo apt-get--purge remove nvidia*

b).run安装失败的

sudo /usr/local/cuda-8.0/bin/uninstall_cuda_8.0.pl

sudo /usr/bin/nvidia-uninstall

c) 在 a) 或 b) 方法清除后，安装还是失败

sudo apt-get autoremove --purge nvidia-*   
#把nvidia驱动清个干干净净

sudo reboot

1. PRE-INSTALLATION ACTION

1.1 Verify you have a CUDA-Capable GPU

lspci | grep -i nvidia

result show:

01:00.0 VGA compatible controller: NVIDIA Corporation Device 1b81 (rev a1)

01:00.1 Audio device: NVIDIA Corporation Device 10f0 (rev a1)

1.2 Verify you have a Supported Version of Linux

uname -m &&cat /etc/*release

result show:

x86_64

DISTRIB_ID=Ubuntu

DISTRIB_RELEASE=14.04

DISTRIB_CODENAME=trusty

DISTRIB_DESCRIPTION="Ubuntu 14.04.5 LTS"

.....

1.3 Verify the System Has GCC Installed

gcc --version

result show:

gcc (Ubuntu 4.8.4-2ubuntu1~14.04.3) 4.8.4

1.4 Verify the System has the Correct Kernel Headers and Development Packages Installed

uname -r

result show:

4.4.0-66-generic

安装对应的kernels header和开发包：

sudo apt-get install linux-headers-$(uname -r)

1.5 Download the NVIDIA CUDA Toolkit

这里下载最新.run版本

下载完后，用MD5 检验，如果序号不和，得重新下载

md5sum cuda_8.0.27_linux.run

2. RUNFILE INSTALLATION

2.1 Disabling Nouveau

lsmod | grep nouveau

如果有内容输出，则需禁掉nouveau

sudo gedit /etc/modprobe.d/blacklist-nouveau.conf

添加如下内容：

blacklist nouveau
options nouveau modeset=0

执行

sudo update-initramfs –u

验证是否成功禁用

lsmod | grep nouveau

重启电脑

2.2 Reboot Into Text Mode

重启后，进入登录界面的时候，不要登录进入桌面(否则可能会失败，若不小心进入，请重启电脑)，直接按

Ctrl+Alt+F1进入文本模式（命令行界面），登录账户。

关闭图形化界面

sudo service lightdm stop

切换到cuda_8.0.27_linux.run的目录

sudo sh cuda_8.0.27_linux.run

!Note:安装的时候，要让你先看一堆文字（EULA)，我们直接不停的按空格键到100%;

遇到提示是否安装openGL ，选择NO,其他的可以一路accept, yes或回车

安装成功后，会显示installed，否则会显示failed

重启图形化界面

sudo service lightdm start

登录时能进入桌面，不会一直在重复登录，成功已近大半。

!Note:如果出现重复登陆情况，请卸载cuda,然后重装。

原因：是OpenGL与NVIDIA 冲突。

卸载：在登陆界面时，按Ctrl + Alt + f1,进入TUI
sudo /usr/local/cuda-8.0/bin/uninstall_cuda_8.0.pl
sudo /usr/bin/nvidia-uninstall
sudo reboot

2.3 Device Node Verification

ls /dev/nvidia*

可能出现a), b), c)，d)三种结果，请对号入座。前方高能！

a)若结果显示

/dev/nvidia0  /dev/nvidiactl  /dev/nvidia-uvm

或显示出类似的信息，应该有三个（包含一个类似/dev/nvidia-nvm的），则安装成功

b)如果运气有点背，结果是这样的

ls: cannot access /dev/nvidia*: No such file or directory

或是这样的，只出现

/dev/nvidia0  /dev/nvidiactl

把下面的.sh文件随便命个名Nka.sh

#!/bin/bash
/sbin/modprobe nvidia
if [ "$?" -eq 0 ]; then
  # Count the number of NVIDIA controllers found.
  NVDEVS=`lspci | grep -i NVIDIA`
  N3D=`echo "$NVDEVS" | grep "3D controller" | wc -l`
  NVGA=`echo "$NVDEVS" | grep "VGA compatible controller" | wc -l`

  N=`expr $N3D + $NVGA - 1`
  for i in `seq 0 $N`; do
    mknod -m 666 /dev/nvidia$i c 195 $i
  done

  mknod -m 666 /dev/nvidiactl c 195 255

else
  exit 1
fi

/sbin/modprobe nvidia-uvm

if [ "$?" -eq 0 ]; then
  # Find out the major device number used by the nvidia-uvm driver
  D=`grep nvidia-uvm /proc/devices | awk '{print $1}'`

  mknod -m 666 /dev/nvidia-uvm c $D 0
else
  exit 1
fi

然后执行

sudo chmod +x Nka.sh
sudo ./Nka.sh

ls /dev/nvidia*

添加启动脚本自动加载

sudo gedit /etc/rc.local

如果你是第一次打开这个文件，它应该是空的(除了一行又一行的#注释项外)。这文件的第一行是

#!/bin/sh -e

把

-e

去掉（这步很重要，否则它不会加载这文本的内容）

然后把Nka.s的内容除了#!/bin/bash外复制到其中，(before exit 0)

保存退出。

下次重启时，你应该能直接看到/dev目录下的三个nvidia的文件

c) 如果人品实在不好（我就遇过几次。。。），结果是这样的

modprobe: ERROR: could not insert 'nvidia_uvm': Operation not permitted

当出现这种情况时，可能是驱动打起架来了。

sudo apt-get autoremove --purge nvidia-* 
#把nvidia驱动清个干干净净

sudo reboot         
#一定记得重启，不然你会后悔的!

然后

sudo ./Nka.sh
ls /dev/nvidia*

这时，应该可以见到

/dev/nvidia0 /dev/nvidiactl /dev/nvidia-uvm

3 POST-INSTALLATION ACTIONS

3.1 Environment Setup

sudo gedit /etc/profile

在文件最后添加

export PATH=/usr/local/cuda-8.0/bin:$PATH

export LD_LIBRARY_PATH=/usr/local/cuda-8.0/lib64:$LD_LIBRARY_PATH

如果启动tensorflow报错

Couldn't open CUDA library libcuda.so.

#手动查找添加进 LD_LIBRARY_PATH
ldconfig -p | grep libcuda

最后

source /etc/profile

3.2 Verify the Installation

3.2.1 Verify the Driver Version

nvidia-smi

cat /proc/driver/nvidia/version

NVRM version: NVIDIA UNIX x86_64 Kernel Module 361.77 Sun Jul 17 21:18:18 PDT 2016

GCC version: gcc version 4.8.4 (Ubuntu 4.8.4-2ubuntu1~14.04.3)

3.2.2 Verify CUDA Toolkit

nvcc -V

nvcc: NVIDIA (R) Cuda compiler driver

Built on Tue_Jan_10_13:22:03_CST_2017

Cuda compilation tools, release 8.0, V8.0.61

!Note: 如果是这样的：
The program 'nvcc' is currently not installed. You can install itbytyping:sudo apt-get install nvidia-cuda-toolkit
莫方，确认下/etc/profil的配置环境是否正确

即使什么都没改，可能忘了这一步,或是之前执行了，但过了有段时间，且又还没重启电脑。因为source /etc/profile是临时生效，重启电脑才是永久生效

source /etc/profile

再执行(应该就有显示了）
nvcc -V

3.2.3 Complie sample

cd进NVIDIA_CUDA-8.0_Samples目录，执行

make

运行完后，编译结果会放在NVIDIA_CUDA-8.0_Samples目录下的bin目录

3.2.3 Running the Binaries

cd进bin目录里面的里面的里面，知道看到一堆可执行文件（菱形的图标），大概是 ~/NVIDIA_CUDA-8.0_Samples/bin/x86_64/linux/release

执行

./deviceQuery

./deviceQuery Starting...
CUDA Device Query (Runtime API) version (CUDART static linking)
Detected 1 CUDA Capable device(s)
Device 0: "GeForce GT 720M"
CUDA Driver Version / Runtime Version          
CUDA Capability Major/Minor version number:    2.1
Total amount of global memory:                 1985 MBytes (2081226752 bytes)
( 2) Multiprocessors, ( 48) CUDA Cores/MP:     96 CUDA Core
GPU Max Clock rate:                            1250 MHz (1.25 GHz)
Memory Clock rate:                             800 Mhz
Memory Bus Width:                              64-bi
L2 Cache Size:                                 131072 bytes
Maximum memory pitch:                          2147483647 byte
exture alignment:                             512 bytes
Concurrent copy and kernel execution:          Yes with 1 copy engine(s)
Run time limit on kernels:                     No
Integrated GPU sharing Host Memory:            No
Support host page-locked memory mapping:       Yes
Alignment requirement for Surfaces:            Yes
Device has ECC support:                        Disabled
Device supports Unified Addressing (UVA):      Yes
Device PCI Domain ID / Bus ID / location ID:   0 / 1 / 0
Compute Mode:< Default (multiple host threads can use ::cudaSetDevice() with device simultaneously) >
eviceQuery, CUDA Driver = CUDART, CUDA Driver Version = 8.0, CUDA Runtime Version = 8.0, NumDevs = 1, Device0 = GeForce GT 720M

Result = PASS

或之类的东东，且最后是 Result = PASS 若失败 Result = FAIL

执行

./bandwidthTest

[CUDA Bandwidth Test] - Starting...
Running on...

 Device 0: GeForce GT 720M
 Quick Mode

 Host to Device Bandwidth, 1 Device(s)
 PINNED Memory Transfers
   Transfer Size (Bytes)    Bandwidth(MB/s)
   33554432            3220.9

 Device to Host Bandwidth, 1 Device(s)
 PINNED Memory Transfers
   Transfer Size (Bytes)    Bandwidth(MB/s)
   33554432            3271.9

 Device to Device Bandwidth, 1 Device(s)
 PINNED Memory Transfers
   Transfer Size (Bytes)    Bandwidth(MB/s)
   33554432            9772.8

Result = PASS

NOTE: The CUDA Samples are not meant for performance measurements. Results may vary when GPU Boost is enabled.

成功cuda

4.cudnn8.0

tar -zxvf cudnn-8.0-linux-x64-v*.tgz
cd cuda
sudo cp lib64/* /usr/local/cuda/lib64/
sudo cp include/cudnn.h /usr/local/cuda/include/

(no need)更新软连接：

cd /usr/local/cuda/lib64

sudo rm -rf libcudnn.so libcudnn.so.5          #删除原有动态文件
sudo ln -s libcudnn.so.5.1.5 libcudnn.so.5 
sudo ln -s libcudnn.so.5 libcudnn.so

之前安装了某个cuDNN的版本，想替换为新的：

解决方法是删除掉原先的，重新装新的cuDNN

删除旧的

cd /usr/local/cuda/lib64
sudo rm libcudnn*

然后再按照上述方法安装新的

ubuntu14.04-CUDA8

CUDA

安装失败，或者重新安装，清理残余文件

a)`.deb`安装失败的

1. PRE-INSTALLATION ACTION

1.1 Verify you have a CUDA-Capable GPU

1.2 Verify you have a Supported Version of Linux

1.3 Verify the System Has GCC Installed

1.4 Verify the System has the Correct Kernel Headers and Development Packages Installed

1.5 Download the NVIDIA CUDA Toolkit

2. RUNFILE INSTALLATION

2.1 Disabling Nouveau

2.2 Reboot Into Text Mode

2.3 Device Node Verification

3 POST-INSTALLATION ACTIONS

3.1 Environment Setup

3.2 Verify the Installation

3.2.1 Verify the Driver Version

3.2.2 Verify CUDA Toolkit

3.2.3 Complie sample

3.2.3 Running the Binaries

4.cudnn8.0

results matching ""

No results matching ""

CUDA

安装失败，或者重新安装，清理残余文件

a).deb安装失败的

1. PRE-INSTALLATION ACTION

1.1 Verify you have a CUDA-Capable GPU

1.2 Verify you have a Supported Version of Linux

1.3 Verify the System Has GCC Installed

1.4 Verify the System has the Correct Kernel Headers and Development Packages Installed

1.5 Download the NVIDIA CUDA Toolkit

2. RUNFILE INSTALLATION

2.1 Disabling Nouveau

2.2 Reboot Into Text Mode

2.3 Device Node Verification

3 POST-INSTALLATION ACTIONS

3.1 Environment Setup

3.2 Verify the Installation

3.2.1 Verify the Driver Version

3.2.2 Verify CUDA Toolkit

3.2.3 Complie sample

3.2.3 Running the Binaries

4.cudnn8.0

results matching ""

No results matching ""

a)`.deb`安装失败的