Hello everybody,
In the last release of CUDA, that is CUDA 4.1, my programs work well as they did in the past releases, but something weird is happening: it takes too long to start the CUDA program.
When debugging the code, I’ve seen that the problem occurred when the “cudaMallocArray()” function is called the very first time; It takes aproximately 45 seconds to pass through this function.
Can someone tell me why is this happening?
Thanks in advance.