Single kernel two sliding windows CUDA

Hi, I’m new to the CUDA framework, and I was wandering if it’s possible to implement using a kernel a single function that is applied to two sliding windows of different sizes to the same data/array?

Because this array is relatively big, I’d like to compute everything in the same kernel to avoid temporary large output arrays.

Could someone point to an example?

Thanks,
Mário Costa

Hello,

Welcome to the NVIDIA forums. I am going to move this topic over to the CUDA category for better visibility.

Cheers,
Tom

1 Like

Yes, it’s possible. I don’t have an example to point to, but it should be a trivial extension to the one-sliding-window case which is a common CUDA tutorial code. You can find a treatment of such in unit 2 of this online training series.

1 Like