I have read the cuda 3.0 docs which state that Cuda 3.0 will on run on fc10. Has any one gotten cuda to run on fc12 and are there any significant changes outside of the gcc 4.3 issue that occurs on fc11?
Thanks in advance!
Well I don’t know what to tell you about that, but i was wondering, can I have installed both 3.0 and 2.3 toolkits in different paths?
I can’t see how that would be a problem, but I am pretty new to cuda. I am pretty sure the tool kit is all self contained
As far as co-existence of 2.3 and 3.0 toolkits concerned: I fully switched to 3.0 as soon as it appeared for the first time, so I cannot guarantee, but I think with the version 3.0 NVIDIA switched to the installation mode where both header and library files are installed in the system directories too. So - I guess you could do with installing them under different paths, but be sure to provide, through your build system and LD_LIBRARY_PATH settings for the run time, that you pick right header/library files each time you build your code.
It will work fine, but remember that the 3.0 toolkit requires a 195.x beta driver to work correctly.
The best way to manage multiple versions of things like toolkits, libraries and compilers is to use the excellent modules system. Being able to type something like
module load cuda/2.3
module unload cuda/2.3
module load cuda/3.0
to seamlessly switch environments without fuss is peerless in my opinion.