Reconfigurable Computing
&
Its Use in Space Applications
in 20 minutes…
Dr. Brock J. LaMeres
Associate Professor
Department of Electrical & Computer Engineering
What is Reconfigurable Computing?
A System That Alters Its Hardware as its Normal Operating Procedure
• This can be done in real-time or at compile time.
• This can be done on the full-chip, or just on certain portions.
• Changing the hardware allows it to be optimized for the application at hand.
The typical approach to hardware is to
build everything that will ever be needed.
In Reconfigurable Computing, hardware is
instead altered and re-used.
Reconfigurable Computing
2
How is RC Different?
Today’s Computers are Based on a General-Purpose Processor
• The GP CPU is designed to do many things.
• This is the “Jack of All Trades, Master of None” approach.
General Purpose CPU Model
Reconfigurable Computing Model
Reconfigurable Computing
3
Who Cares?
Our Existing Computers Seem to be Working Well. Why Change?
• Our existing computers have benefited from 40 years of Moore’s Law.
• In the 1960’s, Gordon Moore predicted that the number of transistors on a chip would
double every ~24 months.
1959 - 1965
1971* - 2011
Gordon Moore, cofounder of Intel,
holding a vacuum
tube.
Note: First microprocessor introduced in 1971,
the Intel 4004 with 2300 transistors.
Reconfigurable Computing
4
Moore’s Law Rocks!
Why is Moore’s Law so Cool?
• Our general-purpose computer model has gotten smaller, faster, and more power
efficient every 24 months.
• This has allowed faster operation and more sophisticated software to be executed.
• The development tools evolve coinciding with the faster/smaller transistors.
• So we haven’t cared that most of the CPU sits idle since each new node is so much
better than the last, we always win!
1947 (1 transistor)
1971 (2300 transistors)
2012 (5B transistors)
Replica of the First Transistor
Intel 4004, 10um
Intel Xeon Phi, 22nm
(Source: AP Photo Paul Sakuma)
(Source: newsroom.intel.com)
(Source: newsroom.intel.com)
Reconfigurable Computing
5
How long can we do this?
When will Moore’s Law End?
• Most exponentials do come to an end.
Clock Speed
Power
Performance per Clock Cycle
But is transistor count
what we care about?
Reconfigurable Computing
6
Computation vs. Transistor Count
We Really Care About Computation
You Are
Here
Reconfigurable Computing
7
The Promise of RC
A Computer Always Needs an Application
• We discovered that space computers could be greatly enhanced by RC.
• They need to be light. RC reuses hardware, that saves mass.
Reconfigurable Computing
8
The Promise of RC
A Computer Always Needs an Application
• We discovered that space computers could be greatly enhanced by RC.
• They need to be light. RC reuses hardware, that saves mass.
Reconfigurable Computing
9
The Promise of RC
A Computer Always Needs an Application
• We discovered that space computers could be greatly enhanced by RC.
• They need to be light. RC reuses hardware, that saves mass.
• They need to be low power. RC eliminates unnecessary circuitry.
Reconfigurable Computing
10
The Promise of RC
A Computer Always Needs an Application
• We discovered that space computers could be greatly enhanced by RC.
• They need to be light. RC reuses hardware, that saves mass.
• They need to be low power. RC eliminates unnecessary circuitry.
Reconfigurable Computing
11
The Promise of RC
A Computer Always Needs an Application
• We discovered that space computers could be greatly enhanced by RC.
• They need to be light. RC reuses hardware, that saves mass.
• They need to be low power. RC eliminates unnecessary circuitry.
• They need to have high computation. RC can do that.
Reconfigurable Computing
12
The Promise of RC
A Computer Always Needs an Application
• We discovered that space computers could be greatly enhanced by RC.
• They need to be light. RC reuses hardware, that saves mass.
• They need to be low power. RC eliminates unnecessary circuitry.
• They need to have high computation. RC can do that.
Reconfigurable Computing
13
The Promise of RC
A Computer Always Needs an Application
• We discovered that space computers could be greatly enhanced by RC.
• They need to be light. RC reuses hardware, that saves mass.
• They need to be low power. RC eliminates unnecessary circuitry.
• They need to have high computation. RC can do that.
• They need to operate in the presence of harsh radiation.
Reconfigurable Computing
14
The Promise of RC
A Computer Always Needs an Application
• We discovered that space computers could be greatly enhanced by RC.
• They need to be light. RC reuses hardware, that saves mass.
• They need to be low power. RC eliminates unnecessary circuitry.
• They need to have high computation. RC can do that.
• They need to operate in the presence of harsh radiation.
Reconfigurable Computing
Can you repeat the
question???
15
Where Does Radiation Come From?
1) Cosmic Rays
2) Solar Particle Events
3) Trapped Radiation
Reconfigurable Computing
16
What Types of Radiation is There?
Radiation Categories
1. Ionizing Radiation
o
o
Sufficient energy to remove electrons from atomic orbit
Ex. High energy photons, charged particles
2. Non-Ionizing Radiation
o
o
Insufficient energy/charge to remove electrons from atomic orbit
Ex., microwaves, radio waves
Types of Ionizing Radiation
1. Gamma & X-Rays (photons)
o
Sufficient energy in the high end of the UV spectrum
2. Charged Particles
o
Electrons, positrons, protons, alpha, beta, heavy ions
3. Neutrons
o
No electrical charge but ionize indirectly through collisions
What Type are Electronics Sensitive To?
• Ionization which causes electrons to be displaced
• Particles which collide and displace silicon crystal
Reconfigurable Computing
17
What are the Effects?
1. Total Ionizing Dose (TID)
o Cumulative long term damage due to ionization.
o Primarily due to low energy protons and electrons due to
higher, more constant flux, particularly when trapped
o Problem #1 – Oxide Breakdown
» Threshold Shifts
» Leakage Current
» Timing Changes
Reconfigurable Computing
18
What are the Effects?
2. Single Event Effects (SEE)
o
o
o
o
Electron/hole pairs created by a single particle passing through semiconductor
Primarily due to heavy ions and high energy protons
Excess charge carriers cause current pulses
Creates a variety of destructive and non-destructive damage
“Critical Charge” = the amount of charge deposited to change the state of a gate
Reconfigurable Computing
19
But I’m Texting Right Now?
How can our computers function?
You Are
Here
Thank you atmosphere.
Thank you magnetosphere
Reconfigurable Computing
20
But there are computers in space?
Stuff is up there now, how does it function?
A-Side Computer
BAE Rad750
B-Side Computer
BAE Rad750
$200,000
$200,000
Thank you federal
government.
Reconfigurable Computing
21
But there are computers in space?
Rad-Hard Processors Can be Made that are SLOW and EXPENSIVE
• Rad-Hard computers tend to lag commercial versions in performance by 10+ years.
• They are also 100s-1000x more expensive.
Reconfigurable Computing
22
I Know You’re Going to Ask….
Shielding
• Shielding helps for protons and electrons <30MeV, but has diminishing returns after 0.25”.
• This shielding is typically inherent in the satellite/spacecraft design.
Shield Thickness vs. Dose Rate (LEO)
Reconfigurable Computing
23
How Does RC Help This?
Total Ionizing Dose
• TID actually diminishes as features get smaller.
• This is good because we want to use the smallest transistors to get the fastest
performance.
• Using off-the-shelf parts also reduces cost.
Single Event Effects
• SEE gets worse! But it isn’t permanent.
• So we just need a new computer architecture to handle it.
Reconfigurable Computing
24
What Technology is used for RC?
Field Programmable Gate Arrays (FPGA)
• Currently the most attractive option.
• SRAM-based FPGAs give the most flexibility
• Riding Moore’s Law feature shrinkage but achieving
computation in a different way.
Reconfigurable Computing
25
Enter MSU
FPGA-Based, Radiation Tolerant Computing System
• We have created a new computer architecture based on RC that provides tolerance to
SEE’s caused by radiation.
Reconfigurable Computing
26
Our Approach
What is needed for FPGA-Based Reconfigurable Computing?
1. SRAM-based FPGAs
o To support fast reconfiguration
2. Good TID Immunity
o FPGAs fabricated in 45nm or less processes have acceptable TID immunity for the majority of
missions.
The Final Piece is SEE Fault Mitigation due to High Energy Ionizing
Radiation
• SEEs will happen, nothing can stop this.
• A computer architecture that expects and response to faults is needed.
Reconfigurable Computing
27
Our Approach
A Many-Tile Architecture
• The FPGA is divided up into Tiles
• A Tile is a quantum of resources that:
o Fully contains a system (e.g., processor, accelerator)
o Can be programmed via partial reconfiguration (PR)
Fault Tolerance
1.
2.
3.
4.
TMR + Spares
Spatial Avoidance of Background Repair
Scrubbing
An External Radiation Sensor
16 MicroBlaze Soft Processors
on an Virtex-6
Reconfigurable Computing
28
Our Approach
1. TMR + Spares
• 3 Tiles run in TMR with the rest reserved as spares.
• In the event of a fault, the damaged tile is replaced
with a spare and foreground operation continues.
2. Spatial Avoidance & Repair
• The damaged Tile is “repaired” in the background via
Partial Reconfiguration.
• The repaired tile is reintroduced into the system as an
available spare.
3. Scrubbing
• A traditional scrubber runs in the background.
• Either blind or read-back.
• PR is technically a “blind scrub”, but of a particular
region of the FPGA.
4. External Sensor
• Provide information about radiation strikes that have occurred
but may not have caused a fault, yet....
Reconfigurable Computing
Shuttle Flight
Computer
(TMR + Spare)
29
Our Approach
Why do it this way?
With Spares, it basically becomes a flow-problem:
o If the repair rate is faster than the incoming fault rate, you’re safe.
o If the repair rate is slightly slower than the incoming fault rate,
spares give you additional time.
o The additional time can accommodate varying flux rates.
o Abundant resources on an FPGA enable dynamic scaling of the
number of spares. (e.g., build a bigger tub in real time)
Reconfigurable Computing
30
Let’s Get Started
Time for Research
Reconfigurable Computing
31
Technical Readiness Level (TRL-1)
Step 1 – Understand the Problem and See if RC Helps
• The Montana Space Grant Consortium funds an investigation into conducting radiation
tolerant computing research at MSU. The goal is to understand the problem, propose a
solution, and build relationships with scientists at NASA.
Clint Gauer (MSEE from MSU 2009) demo’s computer to MSFC Chief of Technology Andrew Keys
Timeline of Activity at MSU
Proof of
Concept
(2008-2010)
2008
2009
2010
2011
2012
2013
Reconfigurable Computing
2014
2015
2016
32
Technical Readiness Level (TRL-3)
Step 2 – Build a Prototype and Test in a Cyclotron
• NASA funds the development of a more functional prototype and testing
under bombardment by radiation at the Texas A&M Radiation Effects Facility.
Ray Weber (Ph.D., EE from MSU, 2014) prepares experiment.
Timeline of Activity at MSU
Proof of
Concept
(2008-2010)
2008
2009
Prototype Development
& Cyclotron Testing
(2010-2012)
2010
2011
2012
2013
Reconfigurable Computing
2014
2015
2016
33
Technical Readiness Level (TRL-5)
Step 3 – Demonstrate as Flight Hardware on High Altitude Balloons
• NASA funds the development of the computer into flight hardware for demonstration on
high altitude balloon systems, both in Montana and at NASA.
MSU
Computer
Justin Hogan (Ph.D., EE from MSU, 2014) prepares payload.
Timeline of Activity at MSU
Proof of
Concept
(2008-2010)
2008
2009
Prototype Development
& Cyclotron Testing
(2011-2013)
(2010-2012)
2010
High Altitude
Balloon Demos
2011
2012
2013
Reconfigurable Computing
2014
2015
2016
34
Technical Readiness Level (TRL-7)
Step 4 – Demonstrate as Flight Hardware on a Sounding Rocket
• NASA funds the demonstration of the computer system on sounding rocket.
• Payload is integrated and will fly on 10/20/14 at White Sands Missile Range.
Justin Hogan
and
Ray Weber
(MSU Ph.D. Grads)
at rocket training
boot camp in 2012.
Timeline of Activity at MSU
Proof of
Concept
(2008-2010)
2008
2009
Prototype Development
& Cyclotron Testing
Sounding
Rocket Demo
(2011-2013)
(2010-2012)
2010
High Altitude
Balloon Demos
2011
2012
(2012-2014)
2013
Reconfigurable Computing
2014
2015
2016
35
Technical Readiness Level (TRL-8)
Step 5 – Demonstrate on the International Space Station
• NASA funds the demonstration of computer system in International Space Station.
ISS Mockup at
Johnson Space Center
for Crew Training
Mission Control Room
for Apollo Program
Timeline of Activity at MSU
Proof of
Concept
(2008-2010)
2008
2009
Prototype Development
& Cyclotron Testing
Sounding
Rocket Demo
(2011-2013)
(2010-2012)
2010
High Altitude
Balloon Demos
2011
2012
(2012-2014)
2013
Reconfigurable Computing
2014
ISS
Demo
(2014-2015)
2015
2016
36
Technical Readiness Level (TRL-8)
Selfie with $60M Space Suit
Reconfigurable Computing
Technical Readiness Level (TRL-9)
Step 6 – Demonstrate as a Stand-Alone Satellite
• NASA funds the demonstration of the computer system in a Low Earth Orbit mission.
Timeline of Activity at MSU
Proof of
Concept
(2008-2010)
2008
2009
Prototype Development
& Cyclotron Testing
Sounding
Rocket Demo
(2011-2013)
(2010-2012)
2010
High Altitude
Balloon Demos
2011
2012
(2012-2014)
2013
Reconfigurable Computing
2014
ISS
Demo
(2014-2015)
2015
Satellite
Demo
(2014-2016)
2016
38
Technical Readiness Level (TRL-9)
Step 7 – Commercialize It
• License Agreement with 406 Aerospace, LLC, Bozeman, MT.
Reconfigurable Computing
39
Collaborators
Faculty & Scientists
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
Todd Kaiser, MSU Electrical & Computer Engineering Department
Ross Snider, MSU Electrical & Computer Engineering Department
Hunter Lloyd, MSU Computer Science Department
Robb Larson, MSU Mechanical & Industrial Engineering Department
Angela Des Jardins, MSU Physics Department & MSGC
Randy Larimer, MSU Electrical Engineering Department & MSGC
Berk Knighton, MSU Chemistry Department & MSGC
David Klumpar, MSU Physics Department & SSEL
Larry Springer, MSU Physics Department & SSEL
Ehson Mosleh, MSU Physics Department & SSEL
Gary Crum, NASA Goddard Space Flight Center
Thomas Flatley, NASA Goddard Space Flight Center
Leroy Hardin, NASA Marshall Space Flight Center
Kosta Varnavas, NASA Marshall Space Flight Center
Andrew Keys, NASA Marshall Space Flight Center
Robert Ray, NASA Marshall Space Flight Center
Leigh Smith, NASA Marshall Space Flight Center
Eric Eberly, NASA Marshall Space Flight Center
Alan George, University of Florida & NSF Center for High Performance Reconfig Comp.
Reconfigurable Computing
40
Students
MSU Students….
Reconfigurable Computing
41
Thank You For Not
Asking Questions
Reconfigurable Computing
42
References
Content
•
•
•
•
•
•
“Space Transportation Costs: Trends in Price Per Pound to Orbit 1990-2000. Fultron Inc Technical Report.,
September 6, 2002. Sammy Kayali, “Space Radiation Effects on Microelectronics”, JPL, [Available Online]:
http://parts.jpl.nasa.gov/docs/Radcrs_Final.pdf.
Holmes-Siedle & Adams, “Handbook of Radiation Effects”, 2nd Edition, Oxford Press 2002.
Thanh, Balk, “Elimination and Generation of Si-Si02 Interface Traps by Low Temperature Hydrogen Annealing”,
Journal of Electrochemical Society on Solid-State Science and Technology, July 1998.
Sturesson TEC-QEC, “Space Radiation and its Effects on EEE Components”, EPFL Space Center, June 9, 2009.
[Available Online]:
http://space.epfl.ch/webdav/site/space/shared/industry_media/07%20SEE%20Effect%20F.Sturesson.pdf
Lawrence T. Clark, Radiation Effects in SRAM: Design for Mitigation”, Arizona State University, [Available Online]:
http://www.cmoset.com/uploads/9B.1-08.pdf
K. Iniewski, “Radiation Effects in Semiconductors”, CRC Press, 2011.
Images
•
•
•
•
•
•
•
If not noted, images provided by www.nasa.gov or MSU
Displacement Image 1: Moises Pinada, http://moisespinedacaf.blogspot.com/2010_07_01_archive.html
Displacement Image 2/3: Vacancy and divacancy (V-V) in a bubble raft. Source: University of Wisconsin-Madison
SRAM Images: Kang and Leblebici, "CMOS Digital Integrated Circuits" 3rd Edition. McGraw Hill, 2003
SEB Images: Sturesson TEC-QEC, “Space Radiation and its Effects on EEE Components”, EPFL Space Center, June
9, 2009.
FPGA Images: www.xilinx.com, www.altera.com
RHBD Images: Giovanni Anelli & Alessandro Marchioro, “The future of rad-tol electronics for HEP”, CERN,
Experimental Physics Division, Microelectronics Group, [Available Online]:
Reconfigurable Computing
43