我已经遵循了一些关于如何在 Debian 9 中安装 CUDA 的教程。
到目前为止，让我使用的最好的一个nvcc是您可以在这个链接。
现在的问题是，我找不到该工具包。我已经尝试过使用find命令等，但什么也没有。有谁知道工具包在哪里吗？
因为，每当我nvcc使用 CUDA 运行编译一个简单的“Hello World”程序时，它都会出现错误，因为它找不到库。当我尝试安装示例时，它要求提供工具包路径，但我找不到它。
添加：
我使用以下方式安装了所有内容：
apt-get install nvidia-cuda-dev nvidia-cuda-toolkit nvidia-driver
之后，我跑了：
nvcc -V
要检查 nvcc 是否已安装，输出如下：
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2016 NVIDIA Corporation
Built on Sun_Sep__4_22:14:01_CDT_2016
我下载了 ubuntu 16.04 和 CUDA 8.0 的 .run 文件：
cuda_8.0.61_375.26_linux-运行
我跳过驱动程序的安装和工具包的安装，并直接跳转到示例安装
Do you accept the previously read EULA?
accept/decline/quit: accept
You are attempting to install on an unsupported configuration. Do you wish to continue?
(y)es/(n)o [ default is no ]: y
Install NVIDIA Accelerated Graphics Driver for Linux-x86_64 375.26?
(y)es/(n)o/(q)uit: n
Install the CUDA 8.0 Toolkit?
(y)es/(n)o/(q)uit: n
Install the CUDA 8.0 Samples?
(y)es/(n)o/(q)uit: y
Enter CUDA Samples Location
[ default is /root ]: /home/sergiobranco/cuda_samples
Enter Toolkit Location
[ default is /usr/local/cuda-8.0 ]:
Error: cannot find Toolkit in /usr/local/cuda-8.0
Enter Toolkit Location
[ default is /usr/local/cuda-8.0 ]: ??????????
问题是它要求提供工具包位置，但我不知道。我按 Enter 键，然后尝试安装示例，但出现错误：
Error: unsupported compiler: 6.3.0. Use --override to override this check.
Missing recommended library: libXmu.so
Error: cannot find Toolkit in /usr/local/cuda-8.0
===========
= Summary =
===========
Driver: Not Selected
Toolkit: Installation Failed. Using unsupported Compiler.
Samples: Cannot find Toolkit in /usr/local/cuda-8.0
Logfile is /tmp/cuda_install_3212.log
我已经使用了 --override 参数，但它失败了。
之后我尝试至少编译 cuda 给出的“第一个程序”之一：
#include <stdio.h>
__global__
void saxpy(int n, float a, float *x, float *y)
{
int i = blockIdx.x*blockDim.x + threadIdx.x;
if (i < n) y[i] = a*x[i] + y[i];
}
int main(void)
{
int N = 1<<20;
float *x, *y, *d_x, *d_y;
x = (float*)malloc(N*sizeof(float));
y = (float*)malloc(N*sizeof(float));
cudaMalloc(&d_x, N*sizeof(float));
cudaMalloc(&d_y, N*sizeof(float));
for (int i = 0; i < N; i++) {
x[i] = 1.0f;
y[i] = 2.0f;
}
cudaMemcpy(d_x, x, N*sizeof(float), cudaMemcpyHostToDevice);
cudaMemcpy(d_y, y, N*sizeof(float), cudaMemcpyHostToDevice);
// Perform SAXPY on 1M elements
saxpy<<<(N+255)/256, 256>>>(N, 2.0f, d_x, d_y);
cudaMemcpy(y, d_y, N*sizeof(float), cudaMemcpyDeviceToHost);
float maxError = 0.0f;
for (int i = 0; i < N; i++)
maxError = max(maxError, abs(y[i]-4.0f));
printf("Max error: %f\n", maxError);
cudaFree(d_x);
cudaFree(d_y);
free(x);
free(y);
}
但这是输出：
nvcc -ccbin clang-3.8 hello.c
nvcc warning : The 'compute_20', 'sm_20', and 'sm_21' architectures are deprecated, and may be removed in a future release (Use -Wno-deprecated-gpu-targets to suppress warning).
hello.c:3:1: error: unknown type name '__global__'
__global__
^
hello.c:4:1: error: expected identifier or '('
void saxpy(int n, float a, float *x, float *y)
^
hello.c:14:15: warning: implicitly declaring library function 'malloc' with type 'void *(unsigned long)' [-Wimplicit-function-declaration]
x = (float*)malloc(N*sizeof(float));
^
hello.c:14:15: note: include the header <stdlib.h> or explicitly provide a declaration for 'malloc'
hello.c:17:3: warning: implicit declaration of function 'cudaMalloc' is invalid in C99 [-Wimplicit-function-declaration]
cudaMalloc(&d_x, N*sizeof(float));
^
hello.c:25:3: warning: implicit declaration of function 'cudaMemcpy' is invalid in C99 [-Wimplicit-function-declaration]
cudaMemcpy(d_x, x, N*sizeof(float), cudaMemcpyHostToDevice);
^
hello.c:25:39: error: use of undeclared identifier 'cudaMemcpyHostToDevice'
cudaMemcpy(d_x, x, N*sizeof(float), cudaMemcpyHostToDevice);
^
hello.c:26:39: error: use of undeclared identifier 'cudaMemcpyHostToDevice'
cudaMemcpy(d_y, y, N*sizeof(float), cudaMemcpyHostToDevice);
^
hello.c:29:3: error: use of undeclared identifier 'saxpy'
saxpy<<<(N+255)/256, 256>>>(N, 2.0f, d_x, d_y);
^
hello.c:29:10: error: expected expression
saxpy<<<(N+255)/256, 256>>>(N, 2.0f, d_x, d_y);
^
hello.c:29:29: error: expected expression
saxpy<<<(N+255)/256, 256>>>(N, 2.0f, d_x, d_y);
^
hello.c:29:31: warning: expression result unused [-Wunused-value]
saxpy<<<(N+255)/256, 256>>>(N, 2.0f, d_x, d_y);
^
hello.c:29:34: warning: expression result unused [-Wunused-value]
saxpy<<<(N+255)/256, 256>>>(N, 2.0f, d_x, d_y);
^~~~
hello.c:29:40: warning: expression result unused [-Wunused-value]
saxpy<<<(N+255)/256, 256>>>(N, 2.0f, d_x, d_y);
^~~
hello.c:31:39: error: use of undeclared identifier 'cudaMemcpyDeviceToHost'
cudaMemcpy(y, d_y, N*sizeof(float), cudaMemcpyDeviceToHost);
^
hello.c:35:16: warning: implicit declaration of function 'max' is invalid in C99 [-Wimplicit-function-declaration]
maxError = max(maxError, abs(y[i]-4.0f));
^
hello.c:35:30: warning: implicitly declaring library function 'abs' with type 'int (int)' [-Wimplicit-function-declaration]
maxError = max(maxError, abs(y[i]-4.0f));
^
hello.c:35:30: note: include the header <stdlib.h> or explicitly provide a declaration for 'abs'
hello.c:35:30: warning: using integer absolute value function 'abs' when argument is of floating point type [-Wabsolute-value]
maxError = max(maxError, abs(y[i]-4.0f));
^
hello.c:35:30: note: use function 'fabsf' instead
maxError = max(maxError, abs(y[i]-4.0f));
^~~
fabsf
hello.c:35:30: note: include the header <math.h> or explicitly provide a declaration for 'fabsf'
hello.c:38:3: warning: implicit declaration of function 'cudaFree' is invalid in C99 [-Wimplicit-function-declaration]
cudaFree(d_x);
^
hello.c:40:3: warning: implicit declaration of function 'free' is invalid in C99 [-Wimplicit-function-declaration]
free(x);
^
11 warnings and 8 errors generated.
答案1
好吧，最后我能够安装所有东西并且工作正常。我将在这里发布关于如何在 debian 9 上完成此操作的完整教程：
第一步：
apt-get install nvidia-cuda-dev nvidia-cuda-toolkit nvidia-driver
要运行上面的命令，你应该检查这个链接为了更好地了解如何为您的主板正确执行此操作。
话虽如此，然后我下载以下运行文件CUDA 8.0
我还必须安装这些：
apt-get install libglu1-mesa libxi-dev libxmu-dev libglu1-mesa-dev
然后我必须将该工具包包含到我的 $PATH 中才能使其正常工作：
export PATH=$PATH:/usr/lib/nvidia-cuda-toolkit
那么你必须这样做：
sh /home/username/Downloads/cuda_8.0.61_375.26_linux.run --tar mxvf
cp InstallUtils.pm /usr/lib/x86_64-linux-gnu/perl-base/
export $PERL5LIB
现在您可以安装示例：
sh /home/username/Downloads/cuda_8.0.61_375.26_linux.run
当它询问工具包路径时，您应该输入：
/usr/lib/nvidia-cuda-toolkit
这是我的答案：
Do you accept the previously read EULA?
accept/decline/quit: accept
You are attempting to install on an unsupported configuration. Do you wish to continue?
(y)es/(n)o [ default is no ]: y
Install NVIDIA Accelerated Graphics Driver for Linux-x86_64 375.26?
(y)es/(n)o/(q)uit: n
Install the CUDA 8.0 Toolkit?
(y)es/(n)o/(q)uit: n
Install the CUDA 8.0 Samples?
(y)es/(n)o/(q)uit: y
Enter CUDA Samples Location
[ default is /root ]: /somewher
Enter Toolkit Location
[ default is /usr/local/cuda-8.0 ]: /usr/lib/nvidia-cuda-toolkit
现在应该可以毫无问题地安装示例了。然后您可以转到安装它们的文件夹并运行：
nvcc -ccbin clang++-3.8 somefile.cu -o somename
就这样吧。 。 。
如果你想安装 pycuda 你只需这样做：
apt-get install build-essential python-dev python-setuptools libboost-python-dev libboost-thread-dev -y
apt-get install python-pycuda