Proteus : Enabling High-Performance Processing-Using-DRAM with Dynamic Bit-Precision, Adaptive Data Representation, and Flexible Arithmetic

Geraldo F. Oliveira † Mayank Kabra † Yuxin Guo ‡ Kangqi Chen † A. Giray Yağlıkçı † Melina Soysal † Mohammad Sadrosadati † Joaquin O. Bueno ⋆ Saugata Ghose ∇ Juan Gómez-Luna § Onur Mutlu † † ETH Zürich ‡ Cambridge University ⋆ Universidad de Córdoba ∇ Univ. of Illinois Urbana-Champaign § NVIDIA Resea · 2025

Read Proteus : Enabling High-Performance Processing-Using-DRAM with Dynamic Bit-Precision, Adaptive Data Representation, and Flexible Arithmetic by Geraldo F. Oliveira † Mayank Kabra † Yuxin Guo ‡ Kangqi Chen † A. Giray Yağlıkçı † Melina Soysal † Mohammad Sadrosadati † Joaquin O. Bueno ⋆ Saugata Ghose ∇ Juan Gómez-Luna § Onur Mutlu † † ETH Zürich ‡ Cambridge University ⋆ Universidad de Córdoba ∇ Univ. of Illinois Urbana-Champaign § NVIDIA Resea (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2501-17466.txt.

Description

Geraldo Oliveira and ten co-authors present a hardware design that computes inside memory chips, in a 2025 paper. Modern memory, called DRAM, can be tricked into doing bulk logic on the data it stores, which promises speed at low energy. The authors found three flaws in current designs. Fixed number formats waste work on useless bits, execution waits for huge batches, and high-precision math runs slowly.

Their framework, Proteus, watches the data as it runs. It trims the number of bits when values are narrow, and it runs independent pieces of one operation across several memory arrays at once. It also picks the best number format for each instruction without bothering the programmer. The aim is to cut the delay of in-memory operations while keeping their energy edge.

Catalog record

CREATORGeraldo F. Oliveira † Mayank Kabra † Yuxin Guo ‡ Kangqi Chen † A. Giray Yağlıkçı † Melina Soysal † Mohammad Sadrosadati † Joaquin O. Bueno ⋆ Saugata Ghose ∇ Juan Gómez-Luna § Onur Mutlu † † ETH Zürich ‡ Cambridge University ⋆ Universidad de Córdoba ∇ Univ. of Illinois Urbana-Champaign § NVIDIA Resea
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE82 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2501-17466

Browse the full library · About the Tower of Babel