Roberto Laudani

Optimizing ML Workload Partitioning between CPUs and CIM Accelerators for Heterogeneous Computing

Joel Klein, Rebecca Pelke, Roberto Laudani, Jan Moritz Joseph, Rainer Leupers

Computing-in-Memory (CIM) accelerators execute Matrix-Vector Multiplications (MVMs) in memory, making them a compelling solution for Machine Learning (ML) workloads. However, existing ML workload partitioning approaches for CIM accelerators do not fully account for Resistive Rand…Read more