An In-Memory-Computing STT-MRAM Macro with Analog ReLU and Pooling Layers for Ultra-High Efficient Neural Network | IEEE Conference Publication | IEEE Xplore