Built-in Fault-Tolerant Computing Paradigm for Resilient Large-Scale Chip Design: A Self-Test, Self-Diagnosis, and Self-Repair-Based Approach

318 Pages·2023·11.84 MB·English

by  Xiaowei Li, Guihai Yan, Cheng Liu

Xiaowei Li
Guihai Yan
Cheng Liu
Built-in
Fault-Tolerant
Computing Paradigm
for Resilient
Large-Scale Chip
Design
A Self-Test, Self-Diagnosis, and
Self-Repair-Based Approach
Built-in Fault-Tolerant Computing Paradigm
for Resilient Large-Scale Chip Design
Xiaowei Li • Guihai Yan (cid:129) Cheng Liu
Built-in Fault-Tolerant
Computing Paradigm
for Resilient Large-Scale
Chip Design
A Self-Test, Self-Diagnosis,
and Self-Repair-Based Approach
XiaoweiLi GuihaiYan
StateKeyLabofProcessors StateKeyLabofProcessors
InstituteofComputingTechnology,Chinese InstituteofComputingTechnology,Chinese
AcademyofSciences AcademyofSciences
Beijing,China Beijing,China
ChengLiu
StateKeyLabofProcessors
InstituteofComputingTechnology,Chinese
AcademyofSciences
Beijing,China
ISBN978-981-19-8550-8 ISBN978-981-19-8551-5 (eBook)
https://doi.org/10.1007/978-981-19-8551-5
©TheEditor(s)(ifapplicable)andTheAuthor(s),underexclusivelicensetoSpringerNatureSingapore
PteLtd.2023
Thisworkissubjecttocopyright.AllrightsaresolelyandexclusivelylicensedbythePublisher,whether
thewholeorpartofthematerialisconcerned,specificallytherightsoftranslation,reprinting,reuse
ofillustrations,recitation,broadcasting,reproductiononmicrofilmsorinanyotherphysicalway,and
transmissionorinformationstorageandretrieval,electronicadaptation,computersoftware,orbysimilar
ordissimilarmethodologynowknownorhereafterdeveloped.
Theuseofgeneraldescriptivenames,registerednames,trademarks,servicemarks,etc.inthispublication
doesnotimply,evenintheabsenceofaspecificstatement,thatsuchnamesareexemptfromtherelevant
protectivelawsandregulationsandthereforefreeforgeneraluse.
Thepublisher,theauthors,andtheeditorsaresafetoassumethattheadviceandinformationinthisbook
arebelievedtobetrueandaccurateatthedateofpublication.Neitherthepublishernortheauthorsor
theeditorsgiveawarranty,expressedorimplied,withrespecttothematerialcontainedhereinorforany
errorsoromissionsthatmayhavebeenmade.Thepublisherremainsneutralwithregardtojurisdictional
claimsinpublishedmapsandinstitutionalaffiliations.
ThisSpringerimprintispublishedbytheregisteredcompanySpringerNatureSingaporePteLtd.
The registered company address is: 152 Beach Road, #21-01/04 Gateway East, Singapore 189721,
Singapore
Toallthestudentsandcolleaguesin
IntegratedCircuitDesignGroupintheState
KeyLabofComputerArchitecture.
Preface
If your computer crashes, you can revive it by a reboot, an empirical solution that
usuallyturnsouttobeeffective.Therationalebehindthissolutionisthattransient
faults, either in hardware or software, can be fixed by refreshing the machine
state. Such a “silver bullet,” however, could be futile in the future because the
faults, especially those existing in the hardware such as Integrated Circuit (IC)
chips, cannot be eliminated by refreshing. What we need is a more sophisticated
mechanism to steer the system back on the right track. The “magic cure” is
the on-chip fault-tolerant mechanism, which relies on a suite of built-in design-
for-reliability logic, including fault detection, fault diagnosis, and fault recovery,
workinginaunifiedmanner.
With the shrinking semiconductor feature sizes and continuous scaling of the
IC designs, silicon defects caused by manufacture defects, radiation particles, or
progressively aging are almost inevitable and pose critical influence on both the
yieldandqualityofICproducts.Underthiscontext,wehavesuccessfullyapplied
on-chip fault-tolerant computing mechanism onto a set of different chip designs
includinggenericcircuits,general-purposeprocessors,network-on-chips,anddeep
learningprocessorsinthepastdecade,andgraduallyformulateasystematicbuilt-in
fault-tolerantcomputingparadigm,whichcanbeutilizedtoguideICdesignsagainst
thesetypicalsilicondefects.Inadditiontothebasicfaultdetection,faultdiagnosis,
and fault recovery, the proposed built-in fault-tolerant computing paradigm also
providesadditionalbenefits,suchasfacilitatinggracefulperformancedegradation,
mitigatingtheimpactofverificationblindspots,andimprovingthechipyield.
Inthisbook,wemainlyillustratethebuilt-infault-tolerantcomputingparadigm
with practical demonstrations on genetic circuits, general-purpose processors,
network-on-chips, and deep learning processors. The entire book consists of six
chapters. Chapter 1 presents the background of fault-tolerant chip designs and
overviewofthebuilt-infault-tolerantcomputingparadigm.Chapter2presentson-
line fault detection, on-chip path delay, and lifetime fault-tolerant pipeline design
for genetic circuits. Chapter 3 investigates the vulnerability of general-purpose
processorsundersilicondefectsandpresentsacoresalvagingapproach,particularly
formulti-coreprocessorarchitecture.Chapter4focusesonfault-tolerantnetwork-
vii
viii Preface
on-chip designs from distinct angles including topology reconfiguration, routing
design, and architecture design. Chapter 5 focuses on built-in fault-tolerant deep
learning processors fabricated with both conventional CMOS-based technology
and emerging ReRAM-based technology. Chapter 6 concludes this book with a
brief summary of the proposed built-in fault-tolerant computing paradigm and
a discussion of future fault-tolerant computing directions on large-scale VLSI
designs.
Themajorityofthecontentinvolvedinthisbookiscollectedfrompeer-reviewed
papersofGuihaiYan,ChengLiu,LeiZhang,WenLi,SongweiPei,SongjunPan,
BingzhangFu,YingWang,andHangLusupervisedbybothProf.XiaoweiLiand
Prof.HuaweiLiwholeadtheIntegratedCircuitDesignGroupinStateKeyLabof
Computer Architecture, and has already been published in the journals of TVLSI,
TCAD, TC, JCST, and Journal of China Science. Prof. Xiaowei Li organized this
book in general, Prof. Guihai Yan mainly worked on Chaps. 2 and 3. Prof. Cheng
LiuworkedonChaps.1,4,5,and6.Dr.JingyaWualsohelpedalottoeditthisbook.
Prof.HuaweiLiandProf.GuojieLuoreviewedthisbook.Prof.TimChengwrote
foreword for this book. All the efforts are indispenable for this book and greatly
appreciated.
Thetechniquespresentedinthisbookarepartlyselectedfromresearchfounded
by the National Key Research and Development Program of China under grant
2020YFB1600201,andtheNationalNaturalScienceFoundationofChina(NSFC)
undergrantNo.(62174162,62090024,61902375,U20A20202,61876173).
Beijing,China XiaoweiLi
May2022
Foreword
Hardware systems must have sufficient robustness to cope with failures resulting
fromvariousvariabilityandreliabilityconcerns.Thisrequirementnotonlyapplies
to safety-critical advanced systems in avionics and automotive applications but
also becomes a necessity for consumer electronics where cost has been a serious
constraint. For integrated circuits, device geometry shrinkage, very low power
supply levels, and ultra-high operating speeds have significantly reduced noise
margins and increased variations inprocess,device, and design parameters.These
continuing trends in technology scaling have resulted in lower reliability and
higher design uncertainty for highly integrated chips. Not just technology, the
environment,energy,thermalresources,andevenapplicationshavealsocontributed
to greater variations and more diverse sources of errors. Thus, high variability
and low reliability have become the predominant challenges for chip design and
manufacturing.
Whileverification,test,andfaulttolerancetechnologieshavebeenfoundational
disciplines for multiple decades for which the readers can find good textbooks for
theirrespectivebasicknowledge,principles,techniques,andsolutions,thesefields
allcontinuetoevolveandadvance,someofwhichhaveevenreinventedthemselves,
inordertackletheenormousvariabilityandreliabilitychallenges.Asaresult,new
and more effective and efficient solutions continue to emerge, replacing classical
approachesfordesigningandmanufacturingrobustandreliablehardware.
Forfaulttolerance,asuiteoftechniques,rangingfrombuilt-inredundancyand
online reconfiguration capability to tolerate errors, to built-in self-test/-diagnosis/-
repairtorecoverfromerrors,topost-fabricationtuning/adaptationcapability(either
off-lineoronline)tobypasserrors,toautomaticcompensationtoalleviatetheneg-
ative effect caused by variations, or to dynamic adaptation to mask environmental
noise and transient errors, have been developed; some of which have even been
advancedfromtheproof-of-conceptandprototypingstagestoactualdeployment.
Researchers at the Institute of Computing Technology of Chinese Academy of
Sciences have been among the most productive and impactful research groups in
addressingthetechnicalchallengesandcontributingnewsolutionsinthisarea.Over
the past decade, they have developed and employed a number of built-in and/or
ix
x Foreword
onlinefault-tolerantsolutions.Theirsolutionsareeithergeneric,broadlyapplicable
to digital designs and general-purpose processors, or specific to special-purpose
designsincludingnetwork-on-chipsanddeeplearningprocessors.Thisbookgives
in-depth and coherent explanations of these very interesting results. Particularly,
the solutions are introduced in a unified “3S” framework supporting a built-in
fault-tolerantcomputingparadigm,where“3S”standsforself-test,self-diagnosis,
and self-repair (or self-recovery). The description of each technique also includes
clarification of the key differences from the conventional counterparts which I am
sure the readers will find informative and insightful. It is commendable that the
authorshavedoneanoutstandingjobinproducingthisself-containedbookcovering
multiple aspects of built-in fault-tolerant design for resilientchips. Publishing this
bookalsoservesverywellformotivatingresearchgraduatestudentsandresearchers
togainthelatestresultsandinsightintothissubjectofsignificantimportance.
HongKongUniversity Kwang-Ting(Tim)Cheng(郑光廷)
ofScienceandTechnology
December12,2022
Contents
1 Introduction .................................................................. 1
1.1 TypicalOn-ChipFaults................................................. 1
1.1.1 ProcessVariation................................................ 2
1.1.2 ManufacturingDefects ......................................... 3
1.1.3 ChipAging...................................................... 4
1.1.4 SoftErrors....................................................... 5
1.1.5 IntermittentFaults .............................................. 6
1.1.6 EmergingTechnologiesInducedDefects...................... 8
1.2 ConventionalFault-TolerantChipDesignWisdom ................... 9
1.2.1 DesignforTest.................................................. 10
1.2.2 DesignforDiagnosis ........................................... 12
1.2.3 DesignforReliability........................................... 13
1.3 Built-InFault-TolerantComputingParadigm ......................... 14
1.3.1 Self-test.......................................................... 15
1.3.2 Self-diagnosis................................................... 16
1.3.3 Self-repair....................................................... 19
1.3.4 GeneralBenefits ................................................ 23
1.4 Summary ................................................................ 26
References..................................................................... 27
2 Fault-TolerantCircuits...................................................... 33
2.1 On-LineFaultDetection................................................ 33
2.1.1 ChallengesforOn-LineFaultDetection....................... 34
2.1.2 StabilityViolationBasedFaultDetection..................... 35
2.1.3 TimingConstrainsExploration ................................ 38
2.1.4 On-LineFaultDetectionArchitecture......................... 42
2.1.5 ExperimentResultAnalysis.................................... 48
2.1.6 Discussion....................................................... 56
2.2 On-ChipPathDelayMeasurement..................................... 58
2.2.1 PathDelayMeasurementandFaultTolerance ................ 58
2.2.2 PathDelayMeasurementCircuits ............................. 61
xi