Abstract
The University of Central Florida invention is a novel path-based computing paradigm on nanoscale memristor crossbars. The digital in-memory computing paradigm enables the computation of Boolean functions and, consequently, arithmetic operations.
In contrast with write-based digital in-memory computing paradigms, such as IMPLY, MAGIC, and FLOW, UCF’s path-based computing technology is a read-based digital, in-memory computing paradigm. This entails that the memristors must only be programmed once for a Boolean function evaluation. The absence of continuous programming of the memristors for evaluation, as in write-based digital in-memory computing, results in significant energy and latency improvements of 4.7X and 8.5X. It also elongates the device's lifetime due to the mitigation of device programming.
Partnering Opportunity
The research team is looking for partners to develop the technology further for commercialization.
Benefit
Lengthens device lifetime by mitigating WRITE operationsEnergy efficiency and low latencyProvides higher accuracy than analog in-memory computingMarket Application
Accelerating deep neural networks (DNNs)Computer chips that speed data-intensive systemsResearch institutionsPublications
PATH:
Evaluation of Boolean Logic using Path-based In-Memory Computing, paper accepted
at the 59th Design Automation Conference (DAC), July 10-14, 2022, San
Francisco, CA.
Brochure