Near-Memory Computing Compiler for Neural Network Architectures
(2022) EITM01 20222Department of Electrical and Information Technology
- Abstract
- With an increased popularity of machine learning, both higher performance and more energy-efficient circuits are needed to meet the demands of increasing workloads. This master's thesis focuses on convolutional neural networks and implements a compiler that generates an accelerator architecture that can be tailored to performance needs. The implemented architecture utilizes near-memory computing to gain increased performance and higher energy efficiency. This report gives an overview of the implemented architecture. Area and performance results for an example use-case are presented and ideas for future improvements are listed.
Please use this url to cite or link to this publication:
http://lup.lub.lu.se/student-papers/record/9111229
- author
- Allfjord, Alex LU
- supervisor
- organization
- course
- EITM01 20222
- year
- 2022
- type
- H2 - Master's Degree (Two Years)
- subject
- report number
- LU/LTH-EIT 2023-911
- language
- English
- id
- 9111229
- date added to LUP
- 2023-02-27 11:09:17
- date last changed
- 2023-02-27 11:33:17
@misc{9111229, abstract = {{With an increased popularity of machine learning, both higher performance and more energy-efficient circuits are needed to meet the demands of increasing workloads. This master's thesis focuses on convolutional neural networks and implements a compiler that generates an accelerator architecture that can be tailored to performance needs. The implemented architecture utilizes near-memory computing to gain increased performance and higher energy efficiency. This report gives an overview of the implemented architecture. Area and performance results for an example use-case are presented and ideas for future improvements are listed.}}, author = {{Allfjord, Alex}}, language = {{eng}}, note = {{Student Paper}}, title = {{Near-Memory Computing Compiler for Neural Network Architectures}}, year = {{2022}}, }