I’ve installed a 8800GT on the headnode of a Beowulf cluster. The X driver and toolkit
install went without a hitch. However, when I try to run the SDK installer I get
[root@abacus CUDA]# ./NVIDIA_CUDA_SDK_1.1_Linux.run
Verifying archive integrity… All good.
Uncompressing NVIDIA CUDA SDK…
AND IT JUST HANGS!! uname -a returns (I’ve edited out the machine name)
I may have solved this problem. When I executed the script using
sh -x ./NVIDIA_CUDA_SDK_1.1_Linux.run
It hung at the same place, but this time I simply pushed return (instead of Ctrl-C)
and it finished! Going back and looking at the output, I think it was waiting for
me to enter the install path or push enter for the default. For some reason the
question string was not being displayed so I did not know to do anything other
than wait. The CUDA folks may want to look into this.
Hans, this has been observed by other people on other Linuxes as well. It is not a bug of the installer. On SuSE for example, if you install a package called ‘buffer’ or similar, it buffers both console input and output, which is why the cursor prompt is invisible. Check whether some package is on your machine (rpm -qa | grep buffer)