Loop Optimization Techniques used by PyTorch

C++
1

Hi,

I am looking for any information on the loop optimization techniques like loop tiling, loop interchange, or loop rollout being employed by PyTorch convolution modules.

Thanks