Architectural Support for Approximate Computing

• Grant number: 18/24177-0
• Support Opportunities: Scholarships in Brazil - Doctorate
• Effective date (Start): April 01, 2019
• Effective date (End): February 28, 2023
• Field of knowledge: Physical Sciences and Mathematics - Computer Science - Computer Systems
• Principal Investigator: Lucas Francisco Wanner
• Grantee: Isaías Bittencourt Felzmann
• Host Institution: Instituto de Computação (IC). Universidade Estadual de Campinas (UNICAMP). Campinas , SP, Brazil

Abstract

The failure of Dennard Scaling and difficulties in developing multicore architectures have negatively affected the energy efficiency of modern computing systems.Since many applications do not require perfect accuracy, the use of Approximate Computing techniques may trade the final expected quality of operations for energy cost reduction.These hardware level techniques usually optimize energy adjusting external parameters of the circuits, such as supply voltage and frequency, or replacing specific circuits, such as approximate arithmetic units.Approximate Computing has been proved advantageous when evaluated in isolation. However, its applicability in a computing architecture depends on the development of a control mechanism that allows tuning the accuracy level at which an application is run.This mechanism imposes design challenges, such as the latency needed to dynamically control circuit parameters and the need of replication of functional units, which may reduce potential energy savings.This work aims to propose and evaluate integration alternatives of Approximate Computing techniques into a computing architecture, including a control unit that allows operation in both accurate and approximate modes.The project will result in an "approximate" architecture, compatible with modern computing systems, that allows exploring potential gains in applications resilient to approximation and, at the same time, guarantee accuracy in critical and control operations.

Scientific publications

(8)

(References retrieved automatically from Web of Science and SciELO through information on FAPESP grants and their corresponding numbers as mentioned in the publications by the authors)

FELZMANN, ISAIAS; FABRICIO FILHO, JOAO; DE OLIVEIRA, JULIANE REGINA; WANNER, LUCAS; IEEE COMP SOC. Special Session: How much quality is enough quality? A case for acceptability in approximate designs. 2021 IEEE 39TH INTERNATIONAL CONFERENCE ON COMPUTER DESIGN (ICCD 2021), v. N/A, p. 4-pg., 2021-01-01. (18/24177-0)

FELZMANN, ISAIAS; FABRICIO FILHO, JOAO; WANNER, LUCAS. Risk-5: Controlled Approximations for RISC-V. IEEE TRANSACTIONS ON COMPUTER-AIDED DESIGN OF INTEGRATED CIRCUITS AND SYSTEMS, v. 39, n. 11, p. 4052-4063, NOV 2020. (18/24177-0)

FABRICIO FILHO, JOAO; FELZMANN, ISAIAS B.; AZEVEDO, RODOLFO J.; WANNER, LUCAS F.; IEEE. A Resilient Interface for Approximate Data Access. 2022 25TH INTERNATIONAL CONFERENCE ON INFORMATION FUSION (FUSION 2022), v. N/A, p. 8-pg., 2019-01-01. (18/24177-0)

FABRICIO FILHO, JOAO; FELZMANN, ISAIAS; WANNER, LUCAS; IEEE COMP SOC. Approximate Memory with Protected Static Allocation. 2022 IEEE 34TH INTERNATIONAL SYMPOSIUM ON COMPUTER ARCHITECTURE AND HIGH PERFORMANCE COMPUTING (SBAC-PAD 2022), v. N/A, p. 9-pg., 2022-01-01. (18/24177-0)

FELZMANN, ISAIAS; FABRICIO FILHO, JOAO; WANNER, LUCAS; IEEE. AxPIKE: Instruction-level Injection and Evaluation of Approximate Computing. PROCEEDINGS OF THE 2021 DESIGN, AUTOMATION & TEST IN EUROPE CONFERENCE & EXHIBITION (DATE 2021), v. N/A, p. 4-pg., 2021-01-01. (18/24177-0)

FABRICIO FILHO, JOAO; FELZMANN, ISAIAS; WANNER, LUCAS; HOCHBERGER, C; BAUER, L; PIONTECK, T. Transparent Resilience for Approximate DRAM. ARCHITECTURE OF COMPUTING SYSTEMS (ARCS 2021), v. 12800, p. 16-pg., 2021-01-01. (18/24177-0)

FILHO, JOAO FABRICIO; FELZMANN, ISAIAS B.; AZEVEDO, RODOLFO; WANNER, LUCAS F.. AxRAM: A lightweight implicit interface for approximate data access. FUTURE GENERATION COMPUTER SYSTEMS-THE INTERNATIONAL JOURNAL OF ESCIENCE, v. 113, p. 556-570, DEC 2020. (18/24177-0)

FABRICIO FILHO, JOAO; FELZMANN, ISAIAS; WANNER, LUCAS. SmartApprox: Learning-based configuration of approximate memories for energy-efficient execution. SUSTAINABLE COMPUTING-INFORMATICS & SYSTEMS, v. 34, p. 12-pg., 2022-04-01. (18/24177-0)

Academic Publications

(References retrieved automatically from State of São Paulo Research Institutions)

FELZMANN, Isaías Bittencourt. Suporte arquitetural para computação aproximada. 2023. Doctoral Thesis - Universidade Estadual de Campinas (UNICAMP). Instituto de Computação Campinas, SP.