Why?
The CUDA programmer needs to consult the properties of the CUDA device quite frequently. This is commonly needed to decide the grid and block dimensions before launching a kernel. It is also required for other purposes, for example to learn about the size of the different memories in the CUDA device.
How?
One easy way to view the properties of a CUDA device is to compile and execute the devicequery
project. This ships in the GPU Computing SDK and can be found in the %NVSDKCOMPUTE_ROOT%/C/src
directory.
Tried with: CUDA 3.2