Harnessing Checker Hierarchy for Reliable Microprocessors

Title of dissertation: HARNESSING CHECKER HIERARCHY FOR RELIABLE MICROPROCESSORS Joonhyuk Yoo Doctor of Philosophy, 2007 Dissertation directed by: Professor Manoj Franklin Department of Electrical and Computer Engineering Microprocessors are becoming increasingly susceptible to soft errors due to the current trends of semiconductor technology scaling. Traditional fault-tolerant multi-threading architectures provide good fault tolerance by re-executing all the computations. However, such a full re-execution significantly increases the demand on the processor resources, resulting in severe performance degradation. To address this problem, this dissertation presents Active Verification Management (AVM) approaches that utilize a checker hierarchy to increase its performance with a minimal effect on the overall reliability. Based on a simplified queueing model, AVM employs a filter checker which prioritizes the verification candidates so as to selectively do verification. This dissertation proposes three filter checkers— based on (1) result usage, (2) result bitwidth, and (3) result anomaly—that exploit correctness-criticality metrics and anomaly speculation. Binary Correctness Criticality (BCC) and Likelihood of Correctness Criticality (LoCC) are metrics that quantify whether an instruction is important for reliability or how likely an instruction is correctness-critical, respectively. Based on the BCC, a result-usage-based filter checker mitigates the verification workload by bypassing instructions that are unnecessary for correct execution. Computing the LoCC is accomplished by exploiting information redundancy of compressing computationally useful data bits. Numerical significance hints let the result-bitwidth-based filter checker guide a verification priority effectively before the re-execution process starts. Extensive measurements prove that the LoCC yields quite a wide distribution of values, indicating that it has the potential to differentiate diverse degrees of correctness-criticality. A result-anomaly-based filter checker exploits a value similarity property, which is defined by a frequent occurrence of partially identical values. Based on the biased distribution of similarity distance measure, this dissertation further investigates another application to exploit similar values for soft error tolerance with anomaly speculation. Extensive measurements show that the majority of instructions produce values that are different from the previous result value only in a few bits. Experimental results show that the proposed schemes accelerate the processor to be 180% faster than traditional fully-fault-tolerant processor, with a minimal impact on the overall soft error rate. With no AVM, congestion at the checker badly affects performance, to the tune of 57%, when compared to that of a nonfault-tolerant processor. With good marking by AVM, the performance of a reliable processor approaches that of a processor with no verification. These results explain that the proposed AVM has the potential to solve the verification congestion problem when perfect fault coverage is not needed. HARNESSING CHECKER HIERARCHY FOR RELIABLE MICROPROCESSORS by Joonhyuk Yoo Dissertation submitted to the Faculty of the Graduate School of the University of Maryland, College Park in partial fulfillment of the requirements for the degree of Doctor of Philosophy 2007 Advisory Committee: Professor Manoj Franklin, Chair/Advisor Professor Peter Petrov Professor Gang Qu Professor Charles Silio Professor Eun-Suk Seo c © Copyright by Joonhyuk Yoo 2007 Acknowledgments Many people generously contributed their love, prayer, time, and support toward the completion of this dissertation. It is my great honor to know them who made this dissertation possible and my life at the University of Maryland joyful. First and foremost, I would like to thank my advisor, Dr. Manoj Franklin for giving me an invaluable opportunity to work with him. It has been always pleasure with me to discuss problems and findings with him over the past three years. His devotion to research and advising has inspired me in the way I should go as a role model in my academic life. Without his instruction and encouragement, my graduate study could neither be enjoyable nor a meaningful part of my life. I would like to acknowledge the members of my dissertation committee, Dr. Peter Petrov, Dr. Gang Qu, and Dr. Charles Silio, for agreeing to serve on my committee and for sparing their time to review the manuscript. I am also grateful to Dr. Andre Tits and Dr. Gun-A Yoon for their counseling on changing my advisor and on moving forward when I was in despair three years ago. Especially, I would like to thank Dr. Eun-Suk Seo for giving me an opportunity to participate in the development of CDAQ software for NASA’s CREAM flight project launched in Antarctica. Due to her encouraging words and financial support, I was able to gain invaluable experience, improve my confidence, and everything turned out all right while working with her. I specially would like to thank my mom and sisters, who have always been my biggest supporters throughout my entire life, for their endless prayer and love. I ii am also grateful to my father-in-law, mother-in-law, brothers-in-law, and sisters-inlaw for their enormous encouragement and consistent support. Without their love, it would have been impossible for me to go through every step and arrive at this moment. During my stay in Maryland for the past seven years, I was so lucky to have many good people around me. I would like to thank my lab-mates in the System and Computer Architecture Laboratory of the University of Maryland for their help and discussion, and sharing ideas: Seungryul, Dongkeun, Hameed, Wanli, Xuanhua, Brinda, Rania, Mike, Sean, Kursad, Ohm, Jaeyong, Inseok, and all the other members. I also would like to acknowledge the Korean graduate students at Maryland and their family for making my life at the University of Maryland full of fun: Seungjong, Woochul, Donghoon, Soobum, Dongik, Younggu, Kyechong, Jusub, Chanho, Eunmi, Hojin, Seokjin, Kwangsik, Keunmin, Sangchul, Jookyung, Kyungil, Hyunmo, Kyungnam, Jongsu, Jiwhang, Yooah, Seungjoon, Minho, Minkyung, Jisun, Ilchul, Jiksoo and many others. In addition, I want to thank Anna and Vinay for sharing good memory as my office-mates. I also want to thank Sonny, Youngsoo, Prashant, Opher, Moohyun, Jihye, Larry, Peter, and other Cosmic Ray Physics Group members for their help and collaboration. I also thank Sangmin, Daesik, Yeonjin, Yoon, Jongso, Hyosang, Hyunchool, and all other Postechians in Maryland. My old friends over twenty years in Korea, Moongab and Euiyoon have always been supportive on me. In particular, I would like to thank Daehyun, Taeheum, Hyejung, Hyunwook, and Hyunmi for their sharing prayer and encouraging me to keep moving forward. iii I specially would like to thank my church family for standing by me and praying to God for me and my family in my suffering: Father Valentine Han, Mother Teresa Han, Sister Young Yim, and all other brothers and sisters in the Holy Cross Korean Episcopal Church. With their warm love and prayer, I was able to find a new life in Jesus Christ and finish my graduate study. Most significantly, I owe my deepest thanks to my dear Eunsook, Changmin, and Jimin. I am so thankful to God for being with them and for giving us such lovely kids, Changmin and Jimin, who have always been the spring of happiness and joyfulness in my family. Eunsook has always been my best friend, my companion in Christ Jesus, my trustful fellow counselor as well as my sweet wife. Any word cannot express how much I owe my gratitude to her and my kids. It is impossible to remember all, and I apologize to those I have inadvertently left out. Lastly, I thank God for giving me His grace to know Him during my Ph.D. study, showing me His endless love, and being always with me even though I walk through the valley of the shadow of death. Whenever anxiety was great within me, His consolation always brought joy to my soul. When there seemed to be no way in my study, God prepared a way for me. When trials seemed impossible and I could not face the day, the Lord extended His helping hand and made for me a way. I praise the Lord what He has done in my life.