问题[cuda](ubuntu)

我正在尝试遵循本指南，但它不起作用。https://docs.nvidia.com/cuda/wsl-user-guide/index.html#known-limitations

满足的系统要求：

cat /proc/version
Linux version 5.4.72-microsoft-standard-WSL2 (oe-user@oe-host) (gcc version 8.2.0 (GCC)) #1 SMP Wed Oct 28 23:40:43 UTC 2020

windows端，安装最新版本：https ://developer.nvidia.com/cuda/wsl

按照说明安装 cuda 工具包：

sudo apt-get install cuda-toolkit-11-2

尝试运行 BlackScholes 示例：

cd /usr/local/cuda/samples/4_Finance/BlackScholes
sudo make
./BlackScholes
[./BlackScholes] - Starting...
CUDA error at ../../common/inc/helper_cuda.h:779 code=100(cudaErrorNoDevice) "cudaGetDeviceCount(&device_count)"

英伟达-smi：

nvidia-smi
NVIDIA-SMI has failed because it couldn't communicate with the NVIDIA driver. Make sure that the latest NVIDIA driver is installed and running.

怎么了？

当我尝试将 CUDA 与 Blender 2.82.a+dfsg-1 一起使用时，我收到以下错误：

Compiling CUDA kernel ...
"nvcc" -arch=sm_86 --cubin "/usr/share/blender/scripts/addons/cycles/source/kernel/kernels/cuda/kernel.cu" -o "/home/bram/.cache/cycles/kernels/cycles_kernel_sm86_673C3180FADE8EDF0E80E61994C4DC92.cubin" -m64 --ptxas-options="-v" --use_fast_math -DNVCC -I"/usr/share/blender/scripts/addons/cycles/source" 
nvcc fatal   : Value 'sm_86' is not defined for option 'gpu-architecture'
CUDA kernel compilation failed, see console for details.

正如 nvidia-smi 报告的那样，CUDA 11.1 安装在我的系统上

$ nvidia-smi 
Sun Dec 20 19:34:48 2020       
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 455.38       Driver Version: 455.38       CUDA Version: 11.1     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|                               |                      |               MIG M. |
|===============================+======================+======================|
|   0  GeForce RTX 3070    Off  | 00000000:17:00.0  On |                  N/A |
|  0%   55C    P8    19W / 240W |    610MiB /  7979MiB |     12%      Default |
|                               |                      |                  N/A |
+-------------------------------+----------------------+----------------------+
                                                                               
+-----------------------------------------------------------------------------+
| Processes:                                                                  |
|  GPU   GI   CI        PID   Type   Process name                  GPU Memory |
|        ID   ID                                                   Usage      |
|=============================================================================|
|    0   N/A  N/A       968      G   /usr/lib/xorg/Xorg                 59MiB |
|    0   N/A  N/A      1598      G   /usr/lib/xorg/Xorg                247MiB |
|    0   N/A  N/A      1749      G   /usr/bin/gnome-shell               74MiB |
|    0   N/A  N/A      3855      G   ...e/Steam/ubuntu12_32/steam       17MiB |
|    0   N/A  N/A      3866      G   ./steamwebhelper                    3MiB |
|    0   N/A  N/A      3878      G   ...43331726878339,131072 --d       49MiB |
|    0   N/A  N/A      5384      G   /usr/lib/firefox/firefox            3MiB |
|    0   N/A  N/A      5463      G   /usr/lib/firefox/firefox            3MiB |
|    0   N/A  N/A      5686      G   nvidia-settings                     3MiB |
|    0   N/A  N/A      9586      G   /usr/lib/firefox/firefox            3MiB |
|    0   N/A  N/A      9642      G   blender                           120MiB |
|    0   N/A  N/A      9988      G   /usr/lib/firefox/firefox            3MiB |
+-----------------------------------------------------------------------------+

如何强制搅拌机使用不同的着色器模型？因为显然，它不喜欢sm_86。

官方软件包似乎不是最新的（https://packages.ubuntu.com/search?keywords=nvidia-cuda-toolkit）但在使用较新的硬件时使用最新的 cuda 似乎很重要。

Nvidia 正在提供软件包，但不能保证它们会很好地工作或将与 Ubuntu 的 nvidia 驱动程序管理集成。

是否努力在 PPA 中创建一个 ubuntu 包，这样我们就不必手动安装并单独解决冲突？（包会处理）

在安装 cuda 工具包之前。我的根目录上有大约 25 GB 的可用空间。在使用deb（网络安装程序）安装 cuda并在安装后手动重新启动后，我注意到根目录中只有大约 15 GB 的可用空间磁盘空间.由于 cuda 工具包通常只占用大约 5 GB 的空间，我认为安装程序可能无法自行清理（2*5GiB=10GiB）。建议我如何解决这个问题。

以下命令没有帮助 sudo apt-get autoclean sudo apt-get clean sudo apt-get autoremove
PS：我安装了 cuda 以便我可以使用我的 gpu 渲染搅拌机场景。是否有必要为此安装整个工具包？

我一直在尝试安装nvidia-cuda-toolkit，sudo apt install nvidia-cuda-toolkit它显示以下错误：

The following packages have unmet dependencies:  nvidia-cuda-toolkit :
Depends: nvidia-cuda-dev (= 10.1.243-3) but it is not going to be
installed E: Unable to correct problems, you have held broken
packages.

Running `sudo apt install nvidia-cuda-dev` yields:

The following packages have unmet dependencies:  nvidia-cuda-dev :
Depends: libcublas10 (= 10.1.243-3) but 10.2.2.214-1 is to be
installed
                   Recommends: libvdpau-dev but it is not going to be installed
                   Recommends: libnvcuvid1 but it is not installable E: Unable to correct problems, you have held broken packages.

我尝试从 NVIDIA 的官方网站安装它，但没有任何效果。

更新：输出sudo apt-cache show nvidia-cuda-dev：

Package: nvidia-cuda-dev Architecture: amd64 Version: 10.1.243-3
Priority: extra Section: multiverse/libdevel Source:
nvidia-cuda-toolkit Origin: Ubuntu Maintainer: Ubuntu Developers
<[email protected]> Original-Maintainer: Debian
NVIDIA Maintainers <[email protected]> Bugs:
https://bugs.launchpad.net/ubuntu/+filebug Installed-Size: 1100028
Depends: libcudart10.1 (= 10.1.243-3), libaccinj64-10.1 (=
10.1.243-3), libcuinj64-10.1 (= 10.1.243-3), libnvrtc10.1 (= 10.1.243-3), libnvtoolsext1 (= 10.1.243-3), libnvvm3 (= 10.1.243-3), libcublas10 (= 10.1.243-3), libcublaslt10 (= 10.1.243-3), libnvblas10
(= 10.1.243-3), libcufft10 (= 10.1.243-3), libcufftw10 (= 10.1.243-3),
libcurand10 (= 10.1.243-3), libcusolver10 (= 10.1.243-3),
libcusolvermg10 (= 10.1.243-3), libcusparse10 (= 10.1.243-3),
libnppc10 (= 10.1.243-3), libnppial10 (= 10.1.243-3), libnppicc10 (=
10.1.243-3), libnppicom10 (= 10.1.243-3), libnppidei10 (= 10.1.243-3), libnppif10 (= 10.1.243-3), libnppig10 (= 10.1.243-3), libnppim10 (=
10.1.243-3), libnppist10 (= 10.1.243-3), libnppisu10 (= 10.1.243-3), libnppitc10 (= 10.1.243-3), libnpps10 (= 10.1.243-3), libnvgraph10 (=
10.1.243-3), libnvjpeg10 (= 10.1.243-3), libcupti-dev (= 10.1.243-3), libnvidia-ml-dev (= 10.1.243-3), libthrust-dev Recommends: libcuda1
(>= 418.39) | libnvidia-tesla-cuda1 (>= 418.39) | libcuda.so.1 (>=
418.39) | libcuda-10.1-1, libgl1-mesa-dev | libgl-dev, libvdpau-dev, libnvcuvid1 Breaks: libcuda1 (<< 418.39) Filename:
pool/multiverse/n/nvidia-cuda-toolkit/nvidia-cuda-dev_10.1.243-3_amd64.deb
Size: 420308420 MD5sum: fbad3f84b31fe19a382ade63a57bc3ff SHA1:
ceb1f2e2bfdc6fafbeced70bbc707d3e9f83e6f5 SHA256:
85b188b198c53be4d24a9d706dce2252d1d4d69592691c90d4bfd0cabc04f720
Homepage: https://developer.nvidia.com/cuda-zone Description-en:
NVIDIA CUDA development files  The Compute Unified Device Architecture
(CUDA) enables NVIDIA  graphics processing units (GPUs) to be used for
massively parallel  general purpose computation.  .  This package
contains the development files: headers and libraries.
Description-md5: 23a17262479fe7daf1cae67727e949c7

我也尝试关注这篇 Medium 文章。

我按照此说明配置 Cuda，但这一步我尝试编写这些行

To configure the CUDA environment for all users (and applications) on your system create the file (use sudo and a text editor of your choice)

/etc/profile.d/cuda.sh

with the following content,

export PATH=$PATH:/usr/local/cuda/bin
export CUDADIR=/usr/local/cuda

但得到了"/etc/profile.d/cuda.sh" is a directory

我尝试运行 Cuda 示例并得到了

Error: target directory missing
Usage: cuda-install-samples-11.1.sh <target directory>
       Will append NVIDIA_CUDA-11.1_Samples to <target directory>

我试图在 bashrc 中编写以下内容

export CUDA_HOME=/usr/local/cuda
export LD_LIBRARY_PATH=${CUDA_HOME}/lib64
PATH=${CUDA_HOME}/bin:${PATH} 
export PATH 

编辑

我尝试使用此命令运行示例来测试 cuda

./cuda-install-samples-11.1.sh

但得到了

Error: target directory missing
  Usage: cuda-install-samples-11.1.sh <target directory>
   Will append NVIDIA_CUDA-11.1_Samples to <target directory>

编辑 2

我试图运行这个命令

./cuda-install-samples-11.1.sh /home/user/NVIDIA_CUDA-11.1_Samples

但得到了

bash: ./cuda-install-samples-11.1.sh: No such file or directory

dpkg -l | grep nvidia
ii libnvidia-cfg1-455:amd64 455.32.00-0ubuntu1 amd64 NVIDIA binary OpenGL/GLX configuration library
ii libnvidia-common-455 455.32.00-0ubuntu1 all Shared files used by the NVIDIA libraries
ii libnvidia-compute-455:amd64 455.32.00-0ubuntu1 amd64 NVIDIA libcompute package
ii libnvidia-decode-455:amd64 455.32.00-0ubuntu1 amd64 NVIDIA Video Decoding runtime libraries
ii libnvidia-encode-455:amd64 455.32.00-0ubuntu1 amd64 NVENC Video Encoding runtime library
ii libnvidia-extra-455:amd64 455.32.00-0ubuntu1 amd64 Extra libraries for the NVIDIA driver
ii libnvidia-fbc1-455:amd64 455.32.00-0ubuntu1 amd64 NVIDIA OpenGL-based Framebuffer Capture runtime library
ii libnvidia-gl-455:amd64 455.32.00-0ubuntu1 amd64 NVIDIA OpenGL/GLX/EGL/GLES GLVND libraries and Vulkan ICD
ii libnvidia-ifr1-455:amd64 455.32.00-0ubuntu1 amd64 NVIDIA OpenGL-based Inband Frame Readback runtime library
ii nvidia-compute-utils-455 455.32.00-0ubuntu1 amd64 NVIDIA compute utilities
ii nvidia-dkms-455 455.32.00-0ubuntu1 amd64 NVIDIA DKMS package
ii nvidia-driver-455 455.32.00-0ubuntu1 amd64 NVIDIA driver metapackage
ii nvidia-kernel-common-455 455.32.00-0ubuntu1 amd64 Shared files used with the kernel module
ii nvidia-kernel-source-455 455.32.00-0ubuntu1 amd64 NVIDIA kernel source package
ii nvidia-modprobe 455.32.00-0ubuntu1 amd64 Load the NVIDIA kernel driver and create device files
ii nvidia-prime 0.8.8.2 all Tools to enable NVIDIA’s Prime
ii nvidia-settings 455.32.00-0ubuntu1 amd64 Tool for configuring the NVIDIA graphics driver
ii nvidia-utils-455 455.32.00-0ubuntu1 amd64 NVIDIA driver support binaries
ii xserver-xorg-video-nvidia-455 455.32.00-0ubuntu1 amd64 NVIDIA binary Xorg driver

和

nvcc -V
nvcc: NVIDIA ® Cuda compiler driver
Copyright © 2005-2020 NVIDIA Corporation
Built on Tue_Sep_15_19:10:02_PDT_2020
Cuda compilation tools, release 11.1, V11.1.74
Build cuda_11.1.TC455_06.29069683_0

但

nvidia-smi
NVIDIA-SMI has failed because it couldn’t communicate with the NVIDIA driver. Make sure that the latest NVIDIA driver is installed and running.

the nvidia bug report starts off with a bunch of failed conftests:
 #error acpi_walk_namespace() conftest failed!
     ^
   /tmp/selfgz20040/NVIDIA-Linux-x86_64-455.23.05/kernel/conftest/functions.h:64:2: error: #error wait_on_bit_lock() conftest failed!
    #error wait_on_bit_lock() conftest failed!
     ^
   /tmp/selfgz20040/NVIDIA-Linux-x86_64-455.23.05/kernel/conftest/functions.h:69:2: error: #error radix_tree_replace_slot() conftest failed!
    #error radix_tree_replace_slot() conftest failed!

and then finishes with general errors:
./arch/x86/include/asm/bug.h:35:22: error: expected identifier or '(' before string constant
  asm_inline volatile("1:\t" ins "\n"    

还：

gcc -v
Using built-in specs.
COLLECT_GCC=gcc
COLLECT_LTO_WRAPPER=/usr/lib/gcc/x86_64-linux-gnu/5/lto-wrapper
Target: x86_64-linux-gnu
Configured with: ../src/configure -v --with-pkgversion='Ubuntu 5.5.0-12ubuntu1' --with-bugurl=file:///usr/share/doc/gcc-5/README.Bugs --enable-languages=c,ada,c++,go,d,fortran,objc,obj-c++ --prefix=/usr --program-suffix=-5 --enable-shared --enable-linker-build-id --libexecdir=/usr/lib --without-included-gettext --enable-threads=posix --libdir=/usr/lib --enable-nls --with-sysroot=/ --enable-clocale=gnu --enable-libstdcxx-debug --enable-libstdcxx-time=yes --with-default-libstdcxx-abi=new --enable-gnu-unique-object --disable-vtable-verify --enable-libmpx --enable-plugin --enable-default-pie --with-system-zlib --enable-objc-gc --enable-multiarch --disable-werror --with-arch-32=i686 --with-abi=m64 --with-multilib-list=m32,m64,mx32 --enable-multilib --with-tune=generic --enable-checking=release --build=x86_64-linux-gnu --host=x86_64-linux-gnu --target=x86_64-linux-gnu
Thread model: posix
gcc version 5.5.0 20171010 (Ubuntu 5.5.0-12ubuntu1) 

sudo lshw -C 视频 [sudo] password for iviti: *-display UNCLAIMED description: VGA compatible controller product: GP106M [GeForce GTX 1060 Mobile] vendor: NVIDIA Corporation physical id: 0 bus info: pci@0000:01:00.0 version: a1 width: 64 bits clock: 33MHz capabilities: pm msi pciexpress vga_controller bus_master cap_list configuration: latency=0 resources: memory:a3000000-a3ffffff memory:90000000-9fffffff memory:a0000000-a1ffffff ioport:4000(size=128) memory:a4080000-a40fffff *-display description: VGA compatible controller product: Intel Corporation vendor: Intel Corporation physical id: 2 bus info: pci@0000:00:02.0 version: 00 width: 64 bits clock: 33MHz capabilities: pciexpress msi pm vga_controller bus_master cap_list rom configuration: driver=i915 latency=0 resources: irq:158 memory:a2000000-a2ffffff memory:b0000000-bfffffff ioport:5000(size=64) memory:c0000-dffff

作为启动脚本的一部分，我正在尝试在 Ubuntu 18.04 VM 模板上自动安装 CUDA。一般来说，我对 bash 和云计算很陌生。当我运行安装 CUDA 的命令时，在安装过程中总是提示我通过输入相应的 ID 号从选项列表中选择国家和语言。

基本上，当我运行时：

sudo apt-get -y install cuda

我知道我必须手动输入“31”和“1”才能完成安装。作为一个相对的 n00b，我的问题是“我如何自动化这些输入，这样我就不必在每次启动这个 VM 模板的新副本时手动输入它们？”

我最初的方法是这样的：

printf "31\n1\n" | sudo apt-get -y install cuda

但这似乎不像我预期的那样工作。没有输入是自动填充的。

我的最终目标是在一个我不必接触的启动脚本中处理所有事情。

我感谢任何和所有帮助，如果我的问题已在其他地方得到回答，我深表歉意（如果有，我一直无法找到它，非常感谢被引导到它！）

有一个类似的问题，我一直在 askubuntu 上关注这个答案：https ://askubuntu.com/a/1158183/254943 。这个解决方案完全符合我的要求。可悲的是，在最近的更新之后，该解决方案停止工作，Xorg 现在再次使用 nvidia。

• | NVIDIA-SMI 450.66 驱动程序版本：450.66 CUDA 版本：11.0 |
• Linux XXX 5.4.0-48-generic #52-Ubuntu SMP Thu Sep 10 10:58:49 UTC 2020 x86_64 x86_64 x86_64 GNU/Linux
• 等离子壳 5.18.5 | Qt：5.12.8 | KDE 框架：5.68.0 | kf5-配置：1.0

$ cat  /usr/share/X11/xorg.conf.d/11-nvidia-prime.conf

# DO NOT EDIT. AUTOMATICALLY GENERATED BY gpu-manager

Section "OutputClass"
    Identifier "Nvidia Prime"
    MatchDriver "nvidia-drm"
    Driver "nvidia"
    Option "AllowEmptyInitialConfiguration"
    Option "IgnoreDisplayDevices" "CRT"
    # Option "PrimaryGPU" "Yes"   <<< commented out
    ModulePath "/x86_64-linux-gnu/nvidia/xorg"
EndSection


# added 
Section "OutputClass"
    Identifier "intel"
    MatchDriver "i915"
    Driver "modesetting"
    Option "PrimaryGPU" "yes"    
EndSection

我试图自己调试它，但我无法修复它。

$ nvidia-smi  # shows running process /usr/lib/xorg/Xorg

+-----------------------------------------------------------------------------+
| NVIDIA-SMI 450.66       Driver Version: 450.66       CUDA Version: 11.0     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|                               |                      |               MIG M. |
|===============================+======================+======================|
|   0  GeForce GTX 105...  Off  | 00000000:01:00.0 Off |                  N/A |
| N/A   50C    P8    N/A /  N/A |      6MiB /  4040MiB |      0%      Default |
|                               |                      |                  N/A |
+-------------------------------+----------------------+----------------------+
                                                                               
+-----------------------------------------------------------------------------+
| Processes:                                                                  |
|  GPU   GI   CI        PID   Type   Process name                  GPU Memory |
|        ID   ID                                                   Usage      |
|=============================================================================|
|    0   N/A  N/A      1257      G   /usr/lib/xorg/Xorg                  4MiB |
+-----------------------------------------------------------------------------+

我需要使用 CUDA 11.1 驱动程序进行机器学习。Steam 需要libnvidia-gl-450:i386安装驱动程序才能打开。但是，安装一个驱动程序会删除另一个驱动程序。我如何能够同时使用它们或者在使用 CUDA 驱动程序时能够玩 Steam 游戏？

我试过如何将 NVIDIA 和 CUDA 驱动程序安装到 Ubuntu 中？但它仅适用于 18.04。这里问了一个类似的问题，但没有任何答案。