Lost in readmes new linux user lost

Accelerated Computing CUDA CUDA Programming and Performance
1

Hi,

I have an installed C870 on Redhat 5.1 x86_64.

The latest Nvidia driver is installed (x86_64-169.09-pkg2.run).
Cuda toolkit 1.1 for RH5 x86_64 is also installed (/usr/local/cuda) as well as cuda sdk for linux (/root/NVIDIA_CUDA_SDK).

I red the release note available with sdk and i tried to specify the path of the cuda binaries the “path environment variable” but i don’t really know where to find that… The same problem happens when i try to specify cuda libraries to “LD_LIBRARY_PATH”.

The release note says that it can be done in .bash_profile via bash by adding :

PATH=$PATH:/usr/local/cuda/bin
LD_LIBRARY_PATH=$LD_LIBRARY_PATH::/usr/local/cuda/lib
export PATH
export LD_LIBRARY_PATH

I created the libcutil.a and libcutilD.a libraries in NIVIA_CUDA_SKD/lib.

I also tried to run one of the existing examples, but i’m not sure how to do that…

What should be done ???

Thanks

2

I finaly found what to do with the .bash_profile, it now looks like that :

#.bash_profile

#Get the aliases and functions
if [ -f ~/.bashrc ]; then
. ~/.bashrc
fi

User specific environment and startup programs

PATH=$PATH:/usr/local/cuda/bin
LD_LIBRARY_PATH=$LD_LIBRARY_PATH::/usr/local/cuda/lib
export PATH
export LD_LIBRARY_PATH
export USERNAME

I tried to “make” some examples, but it still tells me that command nvcc is not found…

3

It looks like you’re almost there, you just have to add the directory that contains ‘nvcc’ to your PATH variable:

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

4

Ok, it now works a bit better ^^

The .bash_file looks like that now :

#.bash_profile

#Get the aliases and functions

if [ -f ~/.bashrc ]; then

    . ~/.bashrc

fi

User specific environment and startup programs

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

LD_LIBRARY_PATH=$LD_LIBRARY_PATH::/usr/local/cuda/lib

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

export LD_LIBRARY_PATH

export USERNAME

I tried, after reboot, to “make” the project named matrixMul.

The terminal answers (approximately, 'cause my version is not in english) :

/usr/bin/ld: cannot find -lGLU

collect2: ld returned 1 execution state code

make: *** […/…/bin/release/matrixMul] Error 1

If you have any idea i’ll be glad to read them…

Thanks ^^

5

I’m not certain, but I think you need GLU / GLUT. Something installed them for me. A forum search should help.

6

You need to install mesa-libGLU-devel (or equivalent) which ships with the OS.

7

Me confused…

I took a look on what’s installed and it appears that some mesa-libGL* are already there…
But, indeed there are no mesa-libGLU-devel installed
I’ll try to find that !

Thanks a lot !

8

Thanks to netlama, I installed mesa-libGLU-devel, so i obtained libglut.so.3 !
Thanks to kristleifur, i searched a bit to know what to do with that.

I found the answer in this post :

[url=“http://forums.nvidia.com/index.php?showtopic=46575”]http://forums.nvidia.com/index.php?showtopic=46575[/url]

Thanks people !!!

9

I have another problem…

I used “make”, “make dbg=1”, “make emu=1” and “make emu=1 dbg=1”. Everything worked fine, but when I want to run one of the example located in /root/NVIDIA_CUDA_SDK/bin/linux/release/, terminal answers :

[root@tesla release]# ./matrixMul
./matrixMul: error while loading shared libraries: libcudart.so: cannot open shared object file: no such file or directory

I gave all access permission to cuda and its containt (/usr/local/cuda)
with chmod -R 777 cuda, same thing to sdk (/root/NVIDIA_CUDA_SDK).

I think it might be a path problem, but if it is so, i don’t really know how to correct that…

Any idea ?

Thanks ^^

10

You probably need to add /usr/local/cuda/lib to you LD_LIBRARY_PATH. For example in bash do:

export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/local/cuda/lib

11

Okay, when i type that just before using one of the examples located in /root/NVIDIA_CUDA_SDK/bin/linux/release/
most of them are working correctly.

But when i restart the computer, i have to type it again…
Is there a way to make that permanent ?
For example by creating a .bash_profile file with the bash command line you gave me ?

I tried to copy this command line in .bash_profile located in my home directory (/root) but it still asks me for this line after reboot…

Thanks already for the tip ^^

12

Hi !!

In order to understand a bit more about how to program the tesla card, i typed the example from the “NVIDIA CUDA Programming Guide v1.1” located page 69 (6.2 Source Code Listing).

I tried to make it but it answers me : No rule to male target

Then, i tried to compile it with nvcc, it happens that there are errors like “attribute” missing or undefined functions…

Is there more info available about this example ?