Mixed-precision in-memory computing

By M. Le Gallo, Abu Sebastian, Roland Mathis, Matteo Manica, Heiner Giefers, Tomáš Tůma, C. Bekas, Alessandro Curioni, Evangelos Eleftheriou. Nature Electronics, 2017

420 citations recorded by the source corpus.

As complementary metal–oxide–semiconductor (CMOS) scaling reaches its technological limits, a radical departure from traditional von Neumann systems, which involve separate processing and memory units, is needed in order to extend the performance of today’s computers substantially. In-memory computing is a promising approach in which nanoscale resistive memory devices, organized in a computational memory unit, are used for both processing and memory. However, to reach the numerical accuracy typically required for data analytics and scientific computing, limitations arising from device variability and non-ideal device characteristics need to be addressed. Here we introduce the concept of mixed-precision in-memory computing, which combines a von Neumann machine with a computational memory unit. In this hybrid system, the computational memory unit performs the bulk of a computational task, while the von Neumann machine implements a backward method to iteratively improve the accuracy of the solution. The system therefore benefits from both the high precision of digital computing and the energy/areal efficiency of in-memory computing. We experimentally demonstrate the efficacy of the approach by accurately solving systems of linear equations, in particular, a system of 5,000 equations using 998,752 phase-change memory devices. A hybrid system that combines a von Neumann machine with a computational memory unit can offer both the high precision of digital computing and the energy/areal efficiency of in-memory computing, which is illustrated by accurately solving a system of 5,000 equations using 998,752 phase-change memory devices.

Subfields: Artificial Intelligence, Electrical and Electronic Engineering

View Paper: https://www.semanticscholar.org/paper/4308e439a232a56041e67b871472ad3b3a0726cf

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Electrical and Electronic Engineering

Citations in the corpus

In-memory computing on a photonic platform (2018, positive): “For example, an architecture comprising multiple scalar multiplying units could be exploited to calculate, in parallel, the result of the matrix-vector multiplications that underpin key processing operations in the area of ‘big data’ analytics 18 .”
A crossbar array of magnetoresistive memory devices for in-memory computing (2022, background): “Of the four NVMs that can be produced at volume—that is, resistive random-access memory 8–13 , phase-change random-access memory 14,15 , flash memory 16–19 and spin-transfer-torque magnetoresistive random-access memory (STT-MRAM, or MRAM for brevity) 20–22,25 , Here we fill this gap and implement…”
In-memory hyperdimensional computing (2019, methodology): “PCM has also been successfully employed in novel computing paradigms such as neuromorphic computing 36–40 and computational memory 20,22,41,42 , which makes it a good candidate for realizing the in-memory HDC system.”
Long short-term memory networks in memristor crossbar arrays (2018, methodology): “The parameter (weights or bias) gradients were accumulated as described in equation (21):”
A TinyML Platform for On-Device Continual Learning With Quantized Latent Replays (2021, methodology): “Most efforts in this direction employ a variety of inferenceonly acceleration techniques such as pruning [33] and byte and sub-byte integer quantization [4]; the use of large arrays of simple MAC units [34] or even mixed-signal techniques such as in-memory computing [35].”
This paper is included in the Proceedings of the 2022 USENIX Annual Technical Conference. (neutral): “…without TC. Haidar et al. [38] apply mixed-precision FP16–FP32/FP64 to a high-performance iterative refinement solvers and take advantage of TC. Gallo et al. [39] propose the concept of mixed-precision in-memory computing with a combination of a von Neumann machine and a computational memory unit.”
Towards spike-based machine intelligence with neuromorphic computing (2019, background): “Such approaches can be found in recent literature 137 , where low-precision memristors are used in combination with a high-precision digital processor.”
Hardware implementation of memristor-based artificial neural networks (2024, positive): “…training, whichaccumulatestheweightupdateinsoftwareandonlyupdatesthe memristor devices when the accumulated value surpasses the programming granularity, can greatly relax requirement for conductance update resolution and endurance and allow software-comparable accuracy to be achieved 192,193 .”
Memristive technologies for data storage, computation, encryption, and radio-frequency communication (2022, methodology): “A prominent example is solving systems of linear equations (74), which can be used in a wide range of applications such as regression (75) and solving partial differential equations (76).”
Towards Oxide Electronics: a Roadmap (2019, neutral): “demonstrated a mixed-precision computing approach [388].”
Memory devices and applications for in-memory computing (2020, background): “These research avenues could enlarge the application space of in-memory computing to encompass applications in scientific computing where high computational accuracy is required.”