Exploring Compute-in-Memory Architecture Granularity for Structured Pruning of Neural Networks | IEEE Journals & Magazine | IEEE Xplore