We explore a novel general-purpose CPU design that tightly integrates analog in-memory computing accelerators in processor pipelines execution pipelines. The accelerator is able to execute accelerating matrix-vector multiplications, which dominate deep neural network applications, in constant time, while presenting a high degree of flexibility, allowing to seamlessly accommodate diverse workloads.
Related Publications
Using Algorithmic Transformations and Sensitivity Analysis to Unleash Approximations in CNNs at the Edge | ||||||
Ponzina, Flavio; Ansaloni, Giovanni; Peon Quiros, Miguel; Atienza Alonso, David | ||||||
2022-07-19 | MDPI Micromachines - Special Issue "Hardware-Friendly Machine Learning and Its Applications" | |||||
Error Resilient In-Memory Computing Architecture for CNN Inference on the Edge | ||||||
Rios, Marco Antonio; Ponzina, Flavio; Ansaloni, Giovanni; Levisse, Alexandre Sébastien Julien; Atienza Alonso, David | ||||||
2022-06-07 | Proceedings of the Great Lakes Symposium on VLSI 2022 |