Problems installing CUDA on 64 bit Ubuntu 8.04

Accelerated Computing CUDA CUDA Programming and Performance
1

Hello all.

I have installed the latest (Redhat 5.x 64 bit) drivers from the Nvidia website along with the SDK and toolkit. When I try to make the files in NVIDIA_CUDA_SDK I get the following error message:

/usr/bin/ld: skipping incompatible /usr/lib/gcc/x86_64-linux-gnu/4.2.3/…/…/…/…/lib/libcuda.so when searching for -lcuda
/usr/bin/ld: skipping incompatible /usr/lib/…/lib/libcuda.so when searching for -lcuda
/usr/bin/ld: skipping incompatible /usr/lib/gcc/x86_64-linux-gnu/4.2.3/…/…/…/libcuda.so when searching for -lcuda
/usr/bin/ld: skipping incompatible /usr/bin/…/lib/libcuda.so when searching for -lcuda
/usr/bin/ld: skipping incompatible /usr/lib64/libcuda.so when searching for -lcuda
/usr/bin/ld: skipping incompatible /usr/lib/libcuda.so when searching for -lcuda
/usr/bin/ld: cannot find -lcuda

I have also installed the:
sudo apt-get install libc6-dev-i386

I have been busting my head at this for a few days, and come to the conclusion: I need help :)

Please help and thanks for your time. Just ask if you need more information.

2

Ubuntu-8.04 is not currently supported with CUDA (only 7.04 & 7.10). Even if you resolve the problem above, you will very likely run into issues with the version of gcc that ships with 8.04.

A next release of CUDA (after 2.0) will support Ubuntu-8.04.

3

So what choices do I have. Reinstall 7.10 or ?

4

It sounds like you’ve somehow installed an incompatible version of the driver and the toolkit, so I would double check that everything’s okay there first. But as netllama said, gcc 4.3 doesn’t play nicely with CUDA sometimes and isn’t officially supported yet.

5

Shouldn’t this do the trick then:

sudo update-alternatives --install /usr/bin/gcc gcc /usr/bin/gcc-4.1 60 --slave /usr/bin/g++ g++ /usr/bin/g+±4.1

sudo update-alternatives --install /usr/bin/gcc gcc /usr/bin/gcc-4.2 40 --slave /usr/bin/g++ g++ /usr/bin/g+±4.2

never the less - same error. But I am reading this section as install 7.10 if I want to use CUDA?

6

It really should just work (I’m running Ubuntu 8.04 and CUDA 2.0b2 on my home machine); are you sure that you installed the 64-bit toolkit?

(gcc 4.3 fails pretty hard on the thread migration API sample and I believe compiling cutil)

7

This is the run down of what I did:

  1. Clean install of Ubuntu 8.04 - 64 bit

  2. Downloaded Redhat 5x - 64 bit drivers, toolkit and sdk (offical nvidia site).

  3. Installed the Nvidia drivers, then Toolkit then SDK

  4. Installed sudo apt-get install libc6-dev-i386

  5. Went to NVIDIA_CUDA_SDK and did make

Above error returned to me.

As a fix I tried:

sudo apt-get install build-essential libglut3-dev gcc-4.1 g+±4.1 -y

sudo update-alternatives --install /usr/bin/gcc gcc /usr/bin/gcc-4.1 60 --slave /usr/bin/g++ g++ /usr/bin/g+±4.1

sudo update-alternatives --install /usr/bin/gcc gcc /usr/bin/gcc-4.2 40 --slave /usr/bin/g++ g++ /usr/bin/g+±4.2

Same error.

8

Install the toolkit for Ubuntu 7.10. I’m wondering if it’s set some options that are trying to use RHEL’s paths instead of Ubuntu’s.

9

whoops double post

10

Didn’t solve the problem either. Thanks though.

11

Update:

I re-did the following:

  1. Clean install of 64 bit Ubuntu 8.04
  2. Download the Redhat Linux 5.x driver, toolkit and sdk
  3. Installed the drivers, toolkit and sdk
  4. apt-get install libc6-dev-i386
  5. apt-get install build-essential libglut3-dev gcc-4.1 g+±4.1 -y
  6. Did make and it compilled nicely…

NB:
Make sure that you copy the correct libcuda.so version from the toolkit you download (else your in a world of hurt):

sudo cp libcuda.so.177.13 /usr/local/cuda/lib/libcuda.so

12

I fixed GLU problems in kubuntu 8.04 when I found out that the 2.0b installer didn’t make the link:

/usr/lib32/libGLU.so → /usr/lib32/libGLU.so.1

As soon as I recreated that link, my GLU problems went away.

13

libGLU is not part of CUDA in any way, nor is it part of the NVIDIA display driver. If you’re missing a symlink, its due to an OS packaging problem.