關閉UEFI secure boot
現在比較新一點的主板預設可能會用UEFI secure boot，若沒有的話可以忽略這步驟。UEFI secure boot 會讓3rd party driver包括NVidia失效，所以最好關閉它，最簡單的作法就是到bios去設定，具體方式因bios而異。
NVidia Driver
優先透過apt-get安裝distribution-specific的驅動
sudo add-apt-repository ppa:graphics-drivers/ppa
sudo apt-get update
sudo apt-get install build-essential nvidia-375
如果行不通，再到官網去下載runfile安裝
驗證：
$ nvidia-smi
Mon Dec 19 21:03:02 2016
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 367.57 Driver Version: 367.57 |
|-------------------------------+----------------------+----------------------+
| 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 1070 Off | 0000:01:00.0 On | N/A |
| 28% 32C P8 7W / 151W | 312MiB / 8110MiB | 1% Default |
+-------------------------------+----------------------+----------------------+
+-----------------------------------------------------------------------------+
| Processes: GPU Memory |
| GPU PID Type Process name Usage |
|=============================================================================|
| 0 3200 G /usr/lib/xorg/Xorg 177MiB |
| 0 3927 G compiz 73MiB |
| 0 4233 G unity-control-center 1MiB |
| 0 4336 G ...onModel/Model0/DataReductionProxyUseQuic/ 57MiB |
+-----------------------------------------------------------------------------+
CUDA
官網提供兩種安裝方法，建議用.deb安裝，因為runfile安裝需要關閉x-server，比較麻煩。
安裝完之後，在.bashrc新增
export CUDA_HOME=/usr/local/cuda-8.0
export LD_LIBRARY_PATH=${CUDA_HOME}/lib64
PATH=${CUDA_HOME}/bin:${PATH}
export PATH
然後重開機
驗證安裝：
cd $CUDA_HOME
cd samples/1_Utilities/deviceQuery
sudo make
./deviceQuery
結果
/usr/local/cuda-8.0/samples/1_Utilities/deviceQuery/deviceQuery Starting...
CUDA Device Query (Runtime API) version (CUDART static linking)
Detected 1 CUDA Capable device(s)
Device 0: "GeForce GTX 1070"
CUDA Driver Version / Runtime Version 8.0 / 8.0
CUDA Capability Major/Minor version number: 6.1
Total amount of global memory: 8110 MBytes (8504279040 bytes)
(15) Multiprocessors, (128) CUDA Cores/MP: 1920 CUDA Cores
GPU Max Clock rate: 1772 MHz (1.77 GHz)
Memory Clock rate: 4004 Mhz
Memory Bus Width: 256-bit
L2 Cache Size: 2097152 bytes
Maximum Texture Dimension Size (x,y,z) 1D=(131072), 2D=(131072, 65536), 3D=(16384, 16384, 16384)
Maximum Layered 1D Texture Size, (num) layers 1D=(32768), 2048 layers
Maximum Layered 2D Texture Size, (num) layers 2D=(32768, 32768), 2048 layers
Total amount of constant memory: 65536 bytes
Total amount of shared memory per block: 49152 bytes
Total number of registers available per block: 65536
Warp size: 32
Maximum number of threads per multiprocessor: 2048
Maximum number of threads per block: 1024
Max dimension size of a thread block (x,y,z): (1024, 1024, 64)
Max dimension size of a grid size (x,y,z): (2147483647, 65535, 65535)
Maximum memory pitch: 2147483647 bytes
Texture alignment: 512 bytes
Concurrent copy and kernel execution: Yes with 2 copy engine(s)
Run time limit on kernels: Yes
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) >
deviceQuery, CUDA Driver = CUDART, CUDA Driver Version = 8.0, CUDA Runtime Version = 8.0, NumDevs = 1, Device0 = GeForce GTX 1070
Result = PASS
可以看到CUDA Capability Major是6.1，等等安裝tensorflow會用到。CUDA samples裡面還有許多有趣的範例，建議玩玩看。
參考：http://shamangary.logdown.com/posts/773013-install-torch7-cuda-cudnn-nvidia-driver
cuDNN
下載官方cuDNN包，並複製到cuda資料夾
(未必正確)
$ cd folder/extracted/contents
$ sudo cp -P include/cudnn.h /usr/include
$ sudo cp -P lib64/libcudnn* /usr/lib/x86_64-linux-gnu/
$ sudo chmod a+r /usr/lib/x86_64-linux-gnu/libcudnn*
安裝tensorflow
按照官網指示，介紹得很詳盡，基本沒啥問題。其中configure時的問與答，enable cuda選y，device compute capability填上剛剛deviceQuery查到的數值，其餘可用預設。
