Cuda SDK backward compatibility

Hi guys, there seems to be a problem with an application that fails to be launched on Nvidia Fermi C2050. The point is that on the machine Cuda SDK 3.2 is installed, while my application was built with SDK 2.2. Are there any compatibility issues going on here?

Hi guys, there seems to be a problem with an application that fails to be launched on Nvidia Fermi C2050. The point is that on the machine Cuda SDK 3.2 is installed, while my application was built with SDK 2.2. Are there any compatibility issues going on here?

Yes there are. You will need to have the 2.2 runtime library somewhere the executable can find it on the Fermi machine. After that, it should probably work, but be aware that launch times will be slower because there will be JIT compilation of the PTX into something that will run on the Fermi.

Yes there are. You will need to have the 2.2 runtime library somewhere the executable can find it on the Fermi machine. After that, it should probably work, but be aware that launch times will be slower because there will be JIT compilation of the PTX into something that will run on the Fermi.