Rapid in-memory matrix multiplication using associative processor - Université Polytechnique des Hauts-de-France Accéder directement au contenu
Communication Dans Un Congrès Année : 2018

Rapid in-memory matrix multiplication using associative processor

Résumé

Memory hierarchy latency is one of the main problems that prevents processors from achieving high performance. To eliminate the need of loading/storing large sets of data, Resistive Associative Processors (ReAP) have been proposed as a solution to the von Neumann bottleneck. In ReAPs, logic and memory structures are combined together to allow inmemory computations. In this paper, we propose a new algorithm to compute the matrix multiplication inside the memory that exploits the benefits of ReAP. The proposed approach is based on the Cannon algorithm and uses a series of rotations without duplicating the data. It runs in O(n), where n is the dimension of the matrix. The method also applies to a large set of row by column matrix-based applications. Experimental results show several orders of magnitude increase in performance and reduction in energy and area when compared to the latest FPGA and CPU implementations.
Fichier non déposé

Dates et versions

hal-03383708 , version 1 (18-10-2021)

Identifiants

Citer

Mohamed Ayoub Neggaz, Hasan Erdem Yantir, Smail Niar, Fadi Kurdahi, Ahmed Eltawil. Rapid in-memory matrix multiplication using associative processor. IEEE-ACM Design, Automation & Test in Europe Conference & Exhibition (DATE'18), Mar 2018, Dresden, Germany. pp.985-990, ⟨10.23919/DATE.2018.8342152⟩. ⟨hal-03383708⟩
18 Consultations
0 Téléchargements

Altmetric

Partager

Gmail Facebook X LinkedIn More