Ok hi guys,
Ive installed the display driver as instructed to do so by the nvidia cuda website, and then i installed the SDK no problems aswell. But as im using Fedora 7, which toolkit can i use? I have read in the documentation that cuda is compatible with other linux distros, but im not sure which tool kit i can use. any pointers? thanks!!
Just pick one that is close. Start with one of the red hat ones: if those dont work, keep trying others. You’ll know they don’t work, because you will get errors related to missing shared libraries that exist on the arch they were compiled on but Fedora 7 probably has a newer version of.
ok, so i just tried the RedHat 5.x tool kit and it did install ok and now i get this error:
“nvcc: command not found”
This is returned when trying to make the makefile from the bandwidth test. I have added the PATH and LD_LIBRARY_PATH environment variables like it describes in the release notes. Im assuming this allows the OS to find the nvcc call right?
Anyways, does this error sound familiar to anyone?
You need to add the path to nvcc to your $PATH. If you installed in the default location, that would be /usr/local/cuda/bin.
Fedora 7 will be officially supported in the next CUDA release.
ah ha! thats awesome news for me then. but until then:
warning: libtlshook.so, needed by /usr/local/cuda/lib/libcudart.so, not found (try using -rpath or -rpath-link)
/usr/local/cuda/lib/libcudart.so: undefined reference to `__tlsHookClearTlsValue’
/usr/local/cuda/lib/libcudart.so: undefined reference to `__tlsHookDeleteIndex’
/usr/local/cuda/lib/libcudart.so: undefined reference to `__tlsHookSetTlsValue’
/usr/local/cuda/lib/libcudart.so: undefined reference to `__tlsHookNewIndex’
thats what im getting now, is it not finding that libcudart.so file? Sorry for being such a n00b. HAvnt used linux in the last year or so and this CUDA business is a bit new for me aswell.
Thanks for clearing that up for me. I think i was using the wrong directory as it worked straight away when i did it that time.
forget that i was being stupid! i was typing usr/local/cuda/lib instead of /usr/local/cuda/lib !! opps! :s thanks for your help guys!
OK so unfortunaly i have to revisit this post. Every time i switch my machine oni have to provide the paths to my $PATH for nvcc every time. Its really annoying. Is this just the way it works or is there a way of making this a permanent entry?
If you are using bash, you can add the command into “.bashrc”. Now every new shell you create will have the correct path.
If you are using a tcsh you have to add the command to the .tcshrc file.
sorry for the late reply,
Im using a bash shells, and ive had a look in the .bashrc file but im not really sure where or what to put in there.
Bit of a linux learner aswell. Im getting thre tho! ive had a look around on google and what not, but am un-sure of what to write there. Any pointers would be handy
thanks
baffa
export PATH=$PATH:/usr/local/cuda/bin
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/local/cuda/lib
