Error handling in CUDA

We have not checked the availability of GPU devices or memory for our CUDA programs. It may happen that, when you run your CUDA program, the GPU device is not available or is out of memory. In that case, you may find it difficult to understand the reason for the termination of your program. Therefore, it is a good practice to add error handling code in CUDA programs. In this section, we will try to understand how we can add this error handling code to CUDA functions. When the code is not giving the intended output, it is useful to check the functionality of the code line-by-line or by adding a breakpoint in the program. This is called debugging. CUDA provides debugging tools that can help. So, in the following section, we will see some debugging tools that are provided by Nvidia with CUDA.

...