STOCHASTIC LOGIC Architectures for post-cmos switches

Similar documents
Novel Devices and Circuits for Computing

ESE570 Spring University of Pennsylvania Department of Electrical and System Engineering Digital Integrated Cicruits AND VLSI Fundamentals

A final review session will be offered on Thursday, May 10 from 10AM to 12noon in 521 Cory (the Hogan Room).

Digital Logic: Boolean Algebra and Gates. Textbook Chapter 3

CSE370: Introduction to Digital Design

EE141-Fall 2011 Digital Integrated Circuits

Introduction to CMOS VLSI Design Lecture 1: Introduction

Impact of parametric mismatch and fluctuations on performance and yield of deep-submicron CMOS technologies. Philips Research, The Netherlands

Case Studies of Logical Computation on Stochastic Bit Streams

Digital Logic. CS211 Computer Architecture. l Topics. l Transistors (Design & Types) l Logic Gates. l Combinational Circuits.

CARNEGIE MELLON UNIVERSITY DEPARTMENT OF ELECTRICAL AND COMPUTER ENGINEERING DIGITAL INTEGRATED CIRCUITS FALL 2002

Design of Optimized Quantum-dot Cellular Automata RS Flip Flops

Stochastic Computing: A Design Sciences Approach to Moore s Law

RAJASTHAN TECHNICAL UNIVERSITY, KOTA

EECS240 Spring Lecture 21: Matching. Elad Alon Dept. of EECS. V i+ V i-

Double Feynman Gate (F2G) in Quantumdot Cellular Automata (QCA)

Semiconductor Memories

STUDY AND IMPLEMENTATION OF MUX BASED FPGA IN QCA TECHNOLOGY

On the Reliable Performance of Sequential Adders for Soft Computing

The Robustness of Stochastic Switching Networks

DARPA/SRC STARnet. Avram Bar-Cohen Program Manager MTO. US-EU Workshop on 2D Layered Materials and Devices. April 23, 2015

Digital Integrated Circuits A Design Perspective. Semiconductor. Memories. Memories

Memory, Latches, & Registers

Synthetic and Natural Analog Computation in Living Cells

EE 447 VLSI Design. Lecture 5: Logical Effort

Outline Fault Simulation

Integrated Circuits & Systems

Design and Implementation of Carry Adders Using Adiabatic and Reversible Logic Gates

MTJ-Based Nonvolatile Logic-in-Memory Architecture and Its Application

DPA-Resistance without routing constraints?

Status. Embedded System Design and Synthesis. Power and temperature Definitions. Acoustic phonons. Optic phonons

ECE429 Introduction to VLSI Design

CMOS Inverter. Performance Scaling

Analysis of flip flop design using nanoelectronic single electron transistor

The Future of CMOS. David Pulfrey. CHRONOLOGY of the FET. Lecture Lilienfeld s patent (BG FET) 1965 Commercialization (Fairchild)

Dr. Nicola Nicolici COE/EE2DI4 Midterm Test #2 Nov 22, 2006

Floating Point Representation and Digital Logic. Lecture 11 CS301

CMPE12 - Notes chapter 1. Digital Logic. (Textbook Chapter 3)

Synthesis of Saturating Counters Using Traditional and Non-traditional Basic Counters

9/18/2008 GMU, ECE 680 Physical VLSI Design

Probabilistic and Approximate Computing. Dr. Laura Monroe

CMPEN 411 VLSI Digital Circuits. Lecture 03: MOS Transistor

Reliability Modeling of Nanoelectronic Circuits

UNIVERSITY OF CALGARY. Noise-Immune Digital Circuit Design. Based on Probabilistic Models. Golam Tangim A THESIS

Introduction to Computer Engineering. CS/ECE 252, Fall 2012 Prof. Guri Sohi Computer Sciences Department University of Wisconsin Madison

Lecture 24. CMOS Logic Gates and Digital VLSI II

Investigation of the Thermal Noise of MOS Transistors under Analog and RF Operating Conditions

A Novel LUT Using Quaternary Logic

Mark Redekopp, All rights reserved. Lecture 1 Slides. Intro Number Systems Logic Functions

CMPEN 411. Spring Lecture 18: Static Sequential Circuits

Administrative Stuff

PERFORMANCE ANALYSIS OF CLA CIRCUITS USING SAL AND REVERSIBLE LOGIC GATES FOR ULTRA LOW POWER APPLICATIONS

Combinational Logic Design Combinational Functions and Circuits

A Novel Design for Quantum-dot Cellular Automata Cells and Full Adders

Introduction to magnetic recording + recording materials

Thermal noise driven computing

Lecture 15: Scaling & Economics

Lecture 6: Logical Effort

Nanoscale CMOS Design Issues

Circuit for Revisable Quantum Multiplier Implementation of Adders with Reversible Logic 1 KONDADASULA VEDA NAGA SAI SRI, 2 M.

DESIGN OF QCA FULL ADDER CIRCUIT USING CORNER APPROACH INVERTER

WARM SRAM: A Novel Scheme to Reduce Static Leakage Energy in SRAM Arrays

EECS150 - Digital Design Lecture 26 - Faults and Error Correction. Types of Faults in Digital Designs

arxiv: v1 [cs.et] 13 Jul 2016

GMU, ECE 680 Physical VLSI Design

An Overview of Spin-based Integrated Circuits

Hw 6 and 7 Graded and available Project Phase 2 Graded Project Phase 3 Launch Today

Prime Clocks. Michael Stephen Fiske. 10th GI Conference on Autonomous Systems October 23, AEMEA Institute. San Francisco, California

Topics. Dynamic CMOS Sequential Design Memory and Control. John A. Chandy Dept. of Electrical and Computer Engineering University of Connecticut

CMPE12 - Notes chapter 2. Digital Logic. (Textbook Chapters and 2.1)"

KINGS COLLEGE OF ENGINEERING PUNALKULAM. DEPARTMENT OF ELECTRONICS AND COMMUNICATION ENGINEERING QUESTION BANK

Microelectronics Part 1: Main CMOS circuits design rules

A Low-Error Statistical Fixed-Width Multiplier and Its Applications

ESE 570: Digital Integrated Circuits and VLSI Fundamentals

Lecture 34: Portable Systems Technology Background Professor Randy H. Katz Computer Science 252 Fall 1995

Scaling of MOS Circuits. 4. International Technology Roadmap for Semiconductors (ITRS) 6. Scaling factors for device parameters

ECE/CS 250 Computer Architecture

Digital yet Deliberately Random: Synthesizing Logical Computation on Stochastic Bit Streams

E40M. Binary Numbers. M. Horowitz, J. Plummer, R. Howe 1

Logic BIST. Sungho Kang Yonsei University

Delay and Energy Consumption Analysis of Conventional SRAM

Lecture 22 Chapters 3 Logic Circuits Part 1

ECE 250 / CPS 250 Computer Architecture. Basics of Logic Design Boolean Algebra, Logic Gates

EE141- Fall 2002 Lecture 27. Memory EE141. Announcements. We finished all the labs No homework this week Projects are due next Tuesday 9am EE141

Stochastic Computational Models for Accurate Reliability Evaluation of Logic Circuits

ESE 570: Digital Integrated Circuits and VLSI Fundamentals

Section 3: Combinational Logic Design. Department of Electrical Engineering, University of Waterloo. Combinational Logic

Jahresbericht 2003 der Arbeitsgruppe Experimentalphysik Prof. Dr. Michael Farle

Chapter 5 MOSFET Theory for Submicron Technology

! Charge Leakage/Charge Sharing. " Domino Logic Design Considerations. ! Logic Comparisons. ! Memory. " Classification. " ROM Memories.

International Journal of Combined Research & Development (IJCRD) eissn: x;pissn: Volume: 7; Issue: 7; July -2018

194 IEEE TRANSACTIONS ON NANOTECHNOLOGY, VOL. 4, NO. 2, MARCH Yan Qi, Jianbo Gao, and José A. B. Fortes, Fellow, IEEE

Interconnect s Role in Deep Submicron. Second class to first class

Digital Integrated Circuits A Design Perspective

Quantum-dot cellular automata

EE241 - Spring 2007 Advanced Digital Integrated Circuits. Announcements

Lecture Notes 2 Charge-Coupled Devices (CCDs) Part I. Basic CCD Operation CCD Image Sensor Architectures Static and Dynamic Analysis

Test Pattern Generator for Built-in Self-Test using Spectral Methods

CPE100: Digital Logic Design I

SEMICONDUCTOR MEMORIES

Transcription:

STOCHASTIC LOGIC Architectures for post-cmos switches David S. Ricketts Electrical & Computer Carnegie Mellon University www.ece.cmu.edu/~ricketts Jehoshua (Shuki) Bruck Engineering Electrical Engineering California Institute of Technology bruck@caltech.com

Random at the Nanoscale Statistical Variation Probabilistic Stochastic ti.

Statistical Variation Random Dopant Fluctuations SiO 2 Gate wl M p1 M p2 Source Drain A. Brown et al., IEEE Trans. Nanotechnology, p. 195, 2002 1 2 M s1 M n1 M n2 M s2 Line Edge Roughness we data M mux Column mux M wr we data 1µm K. Shepard, U. Columbia Gate Oxide Variation At nanoscale, nothing is deterministic Year L eff 3σ T ox 3σ V T 3σ W 3σ H 3σ ρ 3σ 90 nm 47% 16% 13% 33% 36% 33% Momose et al, IEEE Trans. Electron Devices, 45(3), 1998 BOX Devices Wires [Sani Nassif, Proc. IEEE CICC, May 2001]

Probabilistic Random events during operation, dice are rolled continuously Example: Probabilistic bili CMOS (PCMOS), K. V. Palem, Rice Univ, et. al. VLSI-SoC: Research Trends in VLSI and Systems on Chip, Springer Boston,

Probabilistic (2) Markov Random Field (MRF) Logic, A. Zaslavsky, Brown Univ. Nepal, et. al. Designing Logic Circuits for Probabilistic Computation in the Presence of Noise RAZOR, T. Austin, T. Mudge, Univ Michigan D. Ernst, et. al., Razor: A Low-Power Pipeline Based on Circuit-Level Timing Speculation

Stochastic Logic Goal: Leverage random fluctuations of nanoscale switches to build a new paradigm in logic/computation. Random events during operation, dice are rolled continuously We don t look to fix randomness, but rather exploit it. We are investigating architectures/logic families that are based on stochastic logic using stochastic switches. We leverage the inherent random state of certain nanoscale devices.

Switch: Bistable Elements Basic idea of a switch * is a bistable element Transition between states t can be external, i.e. input, or internal, e.g. thermal energy. In nature, many physical systems have two stable, or meta-stable states with an energy barrier between them. Transitions occur randomly due to the thermal energy of the system Energy Barrier 0 1

Switch: Bistable Elements (2) Transition between states can be calculated from the barrier height and the thermal energy of the system ~ kt. Energy Barrier P exp E k T B E R 1/ exp flip f0 kt B E 0 1

Switch: Bistable Elements (3) Probability of state determined by symmetry of potential well Energy Barrier P exp E E 1 E 2 kbt 0 1

Example #1 Rotaxane Molecule Used for dense memory cross bar arrays Switching/relaxation R CHEMPHYSCHEM 2002, 3, 519 525 1 kt B G exp h RT : s days Eyring equation Phil. Trans. R. Soc. A (2007) 365, 1607 1625

Example #2 Data storage (disk drive) bit E KuV V 1 FePt Nanoparticle arrays E K V Low Moderate V 2 High V 3 < V 2 < V 1 V 3 20 nm d d 6.3 0.3 nm 0.0505 nm S. Sun, C. Murray, D. Weller, L. Folks, A. Moser, Science, 287, pp. 1989 2000 %) Magnetization ( 1 Stored Data Magnetization Decay V 3 V 2 Data Lost V 1 0 1 minute 1 year 10 years Log Time (s)

Stochastic Logic (J. Bruck) Shannon s work focused on deterministic switching circuits, circuits where each switch is associated with a Boolean variable defining whether the switch is closed. We instead focus on stochastic switching circuits, circuits where each pswitch is associated with a Bernoulli random variable defining the (independent) probability that the pswitch is closed. D. Wilhelm and J. Bruck, Stochastic switching circuit synthesis, IEEE Int.Sym. on Inf. Theory, Toronto, 2008. Deterministic switch Deterministic switch Switches are ON or OFF with a known probability (not necessarily 50%) We construct global probabilities based upon a logical connection of probabilistic switches (Pswitch) and deterministic i ti switches.

Stochastic Logic (2) Series/Parallel Pswitches

Stochastic Logic (3) Examples: Series/Parallel Pswitches 75% 37.5% 43.8% 1) Let circuit C1 be the single-pswitch circuit. 2) For bit Fi, i = 2 to n, let circuit Ci be: a) If Fi = 0, C1 in series with Ci 1, or b) If Fi = 1, C1 in parallel with Ci 1 1101.011 2 2 2 2.2 2 2 3 2 1 0 1 2 3

Stochastic Logic: Universal Probability Generator Generating a deterministic input to desired probabilistic states Could generate 2 n individual id probability generators and select one Inefficient exponential increase in switch count. Need an algorithm that creates the desired probabilities with the minimum hardware Need universal Pswitch network synthesizer

Universal Probability Generator Recursive architecture generates a minimal sized probability generator using 4n-2 switches

Universal Probability Generator (2) Example

Stochastic Logic Summary Goal: Leverage random fluctuations of nanoscale switches to build a new paradigm in logic/computation. Utilize thermal energy of nano-scale switches to generate switches that open/close randomly Probability of open/close is determined by energy states, i.e. double well geometry. Build stochastic ti switches Pswitches from these random devices. Are able to synthesize arbitrary probabilities given a known, but not designed, probability. Universal Probability Generator allows for minimal designs of arbitrary deterministic to probability logic networks.

An Application Example 1 Million RF temperature sensors distributed uniformly over Indiana, powered by solar energy scavenging circuits that operate sensors for 15 min per 24 hour period

An Application Example (2) How to ensure temperature readings throughout the state 24 hours a day? Need 1/100 of sensors on at all times. Need sensors that are on to be even distributed, so that no area is missed. What is the overhead of organization? Of communication? Build each sensor with a 1/100 probability bilit of being turned ON. Sensor network will inherently provide even coverage and time/energy organization. What if 15 min varied by weather, by adjusting random wake-up, energy adaptive networks could be implemented.

Outlook Applications General probability generators Energy/distribution management Stochastic computation in eoltionar evolutionary systems, sstemseg e.g. Stochastic switching as a survival strategy in fluctuating environments, Nature Genetics Efficient implementation of Pswithces at the nanoscale Development of more complex synthesis and computation theory/ algorithms.

QUESTIONS