Skip to content

wjh51333/Adder

Repository files navigation

Adder

⌨ team project for [2020-1] KNU capstone project2
🖱 mento professor YongTae Kim
🎥 Demo Video: https://youtu.be/1foM8boSv5A

Introduction

A Study & Proposal for a New Low-Power Energy Efficient Adder Algorithm for Reducing Power and Energy Consumption of Artificial Intelligence Processors
Evaluation of Proposed Approximate Adder Algorithm
  • Assessing algorithmic performance by building SW simulators using c/c++
  • Evaluate the accuracy of the operation on multiple criteria(ex.error rate)
  • Evaluation by applying algorithms that require additional computation(ex. Video processing)

Benefit

  • Develop a new low-power energy high-efficiency accelerator and apply it to various artificial intelligence algorithms including artificial intelligence processors to build an efficient artificial intelligence platform.
  • Additive operations are basic operations, so they are applied to various algorithms such as image/voice processing, communication, and diphrers that require them

Results

1. Design Low-power Approximate Floating Point Adder

We make accutrate floating point adder with C++ based on IEEE 754. As a result of 5 million tests, the accuracy is 100%. Then we applied it to LOA and ETA1. And based on these activities and other related studies, we designed low-power approximate floating point adder.

2. Design Approximate Adder with increased accuracy

3. Evaluation

About

Capstone Design

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published