Tech Support | Generator Login | Careers | Contact Us
PARTNERS

  Overview

  Operating Systems

  HW Proto/Emulation

  JTAG Probes

  SOC Design

  Audio SW Packages

  Application Software

  Silicon Sources

  EDA Tools

  ESL/Virtual Platforms

  University Program

  + Licensing Policy

  + Application Form

  + License Agreement

  + Renewal Agreement

  + Program Info Form

  + Contact Info & FAQs

The University of Adelaide

Adelaide, Australia

Research using Tensilica's Technology

 

We joined the Tensilica University program in May 2006 for a specific research project. Our aim is to study a low power implementation of the RSA cryptosystem based on the Xtensa processor. The implementation uses the residue number system (RNS) as we hypothesis that we will be able to exploit the arithmetic parallelism exposed by this scheme to build a fast implementation using relatively small hardware extensions to the Xtensa. The Xtensa is an ideal tool for this research as it allows us to quickly evaluate the cost and benefit of changes in the architecture and instruction set.

Since, the last extension to our Xtensa license, we have used Tensilica Instruction Extension language to describe extensions to the processor for residue arithmetic. We found that an RSA signature generation uses 67.7% fewer clock cycles, excluding cashe misses, with the RNS version compared with a non-RNS baseline. The extra hardware requires 42.7% more gates than the base processor core.

The details of our work (see link below) of our conference paper "An RNS-Enhanced Microprocessor Implementation of Public Key Cryptography." We presented it at the Asilomar Conference on Signals, Systems, and Computers in November 2007.

"An RNS-Enhanced Microprocessor Implementation of Public Key Cryptography"

The Xtensa tools have been excellent for this research.

The next stage of the project is to investigate the power usage of the residue arithmetic hardware. Tensilica's Xenergy tool will be an excellent starting point, as it determines the power usage for specific applications. This allows us to form a more accurate impression of the actual power used when performing the RSA signature generation. We are hoping that it can also give an idea of the power usage of the hardware specified with the TIE, then there is some scope for optimizing the TIE for low power consumption.

Dr. Braden Phillips
Lecturer
Schools of Electrical & Electronic Engineering
Braden.Phillips@adelaide.edu.au

Zhining Lim
PhD Candidate
zlim@eleceng.adelaide.edu.au

The University of Adelaide's URL link

School of Electrical and Electronic Engineering link


Tensilica Xtensions Network