One paper accepted to IEEE CAL (Journal, SCIE)

One paper has been accepted for publication in IEEE Computer Architecture Letters. Our paper presents a processing-in-memory (PIM) architecture for accelerating GeMV-softmax compute pipeline, called SoftmaxPIM. In order to accelerate GeMV-softmax compute pipeline, SoftmaxPIM integrates lightweight near-bank softmax units based on an approximation-based softmax formulation alongside with the SIMD units on PIM dies of an HBM stack.