CSE 249A -- Cross-Layer System Architectures for Future Cloud and Mobile Platforms
Spring 2013, with Prof. Jason Mars

Course Description

The course will take an in-depth look at the current and future trends in the design of computer architectures and system platforms for two emerging computing domains, Cloud and Mobile Platforms. This course will be focused on deconstructing the seemingly high impact work that is fresh off the presses and understanding how these works may be bellweathers of things to come.

This is a paper reading course. We'll read 2-4 papers a week. Most will be presented by students. If you are enrolled for 2 units, you are expected to present on your turn, read the assigned papers, and participate in discussions when it is not your turn. If you are enrolled for 4 units, you are expected to also do a research-style project, to be turned in before Monday of finals week. We may have individual projects, or combine forces for a larger group research project -- to be determined.

Paper Questionnaire

The only assignments in this class will be to provide responses to a general questionnaire for each paper. This questionnaire is designed to take very little time to complete. The questionnaire below will be used for all of the paper readings.

• The Questionnaire is As Follows
• Paper Summary (~2-3 Sentences, ~1 sentence for the remaining questions below)
• What is the most important result in the work?
• What are the strengths of the paper?
• What are it's weaknesses?
• What is one thing you didn't understand?
• If this was your paper, what would your next paper be about?

• PIAZZA
• We will be using PIAZZA to share our responses to the questionnaire (described below) and to spark some online discussions/debates.
• URL: https://piazza.com/ucsd/spring2013/cse249a/home
• The initial class roster will receive invites, however it is your responsibility to make sure you are signed up. If you are not already on, click the link and sign up.
• I will put the papers and other materials on Piazza from time to time in PDF format.

• Submitting Your Responses
• To submit your response simply send me a private message on Piazza with your response.
• These should be sent out before class on the day we will discuss papers.
• After class, I will aggregate and post all of your response in a open post on Piazza.

Course Schedule

Subject to jitter

Date	Topic/Papers	Leader
Tues, Apr 2	Introducton	Jason
Thurs, Apr 4	"The Tail at Scale", CACM 2013 "Paragon: QoS-Aware Scheduling for Heterogeneous Datacenters", ASPLOS 2013	Dylan/Mike
Tues, Apr 9	"Bubble-Up: Increasing Utilization in Modern Warehouse Scale Computers via Sensible Co-locations", MICRO 2011 "Power Struggles: Revisiting the RISC vs. CISC Debate on Contemporary ARM and x86 Architectures", HPCA 2013	Daniel/Scott
Thurs, Apr 11	"MorphCore: An Energy-Efficient Microarchitecture for High Performance ILP and High Throughput TLP", MICRO 2012 "Composite cores: Pushing heterogeneity into a core", MICRO 2012	Trevor/Mark
Tues, Apr 16	No Class
Thurs, Apr 18	No Class
Tues, Apr 23	Project Brainstorm	Jason
Thurs, Apr 25	Project Proposal Presentation	All 4 Credit Students
Tues, Apr 30	"Discerning the Dominant Out-of-Order Performance Advantage: Is it Speculation or Dynamism?", ASPLOS 2013 "Illusionist: Transforming Lightweight Cores into Aggressive Cores on Demand", HPCA 2013	Curtis/Jon
Thurs, May 2	"Computational Sprinting", HPCA 2012 "Computational Sprinting on a Hardware/Software Testbed", ASPLOS 2013	Kurtis/Aravindh
Tues, May 7	"Omega: flexible, scalable schedulers for large compute clusters", EuroSys 2013 "CPI^2: CPU performance isolation for shared compute clusters", EuroSys 2013	MoZo/Jacky
Thurs, May 9	"High-Performance and Energy-Efficient Mobile Web Browsing on Big/Little Systems", HPCA 2013 "Unifying Primary Cache, Scratch, and Register File Memories in a Throughput Processor", MICRO 2012	Jessy/Yunqi
Tues, May 14	"Towards Energy-Proportional Datacenter Memory with Mobile DRAM", ISCA 2012	Trevor
Thurs, May 16	"Modeling Performance Variation Due to Cache Sharing", HPCA 2013 "Can Traditional Programming Bridge the Ninja Performance Gap for Parallel Computing Applications?"," ISCA 2012	??/??
Tues, May 21	No Class (Project Work Day)
Thurs, May 23	"The Yin and Yang of Power and Performance for Asymmetric Hardware and Managed Software", ISCA 2012	??
Tues, May 28	"KnightShift: Scaling the Energy Proportionality Wall Through Server-level Heterogeneity", MICRO 2013	??
Thurs, May 30	"Disintegrated Control for Energy-Efficient and Heterogeneous Memory Systems"," HPCA 2013	??
Tues, June 4	Project Presentations
Thurs, June 6	Project Presentations

• Paper Grab-Bag
• Towards Energy-Proportional Datacenter Memory with Mobile DRAM. ISCA 2012
• Scheduling Heterogeneous Multi-Cores through Performance Impact Estimation (PIE). ISCA 2012
• The Yin and Yang of Power and Performance for Asymmetric Hardware and Managed Software. ISCA 2012
• Staged Memory Scheduling: Achieving High Performance and Scalability in Heterogeneous Systems. ISCA 2012
• Probabilistic Shared Cache Management(PriSM). ISCA 2012
• Can Traditional Programming Bridge the Ninja Performance Gap for Parallel Computing Applications?. ISCA 2012
• Harmony: Collection and Analysis of Parallel Block Vectors. ISCA 2012
• Managing Distributed UPS Energy for Effective Power Capping in Data Centers. ISCA 2012
• Scale-Out Processors. ISCA 2012
• Parasol and GreenSwitch: Managing Datacenters Powered by Renewable Energy. ASPLOS 2013
• Power Containers: An OS Facility for Fine-Grained Power and Energy Management on Multicore Servers. ASPLOS 2013
• Computational Sprinting. HPCA 2012
• Computational Sprinting on a Hardware/Software Testbed. ASPLOS 2013
• Regularities Considered Harmful: Forcing Randomness to Memory Accesses to Reduce Row Buffer Conflicts for Multi-Core Multi-Bank Systems. ASPLOS 2013
• Stabilizer: Statistically Sound Performance Evaluation. ASPLOS 2013
• Discerning the Dominant Out-of-Order Performance Advantage: is it Dynamism or Speculation?. ASPLOS 2013
• Stochastic Superoptimization. ASPLOS 2013
• To Hardware Prefetch or Not to Prefetch? A Virtualized Environment Study and Core Binding Approach. ASPLOS 2013
• Traffic Management: A Holistic Approach to Memory Placement on NUMA Systems. ASPLOS 2013
• Demand-Based Coordinated Scheduling for SMP VMs. ASPLOS 2013
• Practical Automatic Loop Specialization. ASPLOS 2013
• Portable Performance on Heterogeneous Architectures. ASPLOS 2013
• TSO_Atomicity: Efficient Hardware Primitive for TSO-Preserving Region Optimizations. ASPLOS 2013
• Transactional Memory Architecture and Implementation for IBM System z. MICRO 2013
• Unifying Primary Cache, Scratch, and Register File Memories in a Throughput Processor. MICRO 2013
• KnightShift: Scaling the Energy Proportionality Wall Through Server-level Heterogeneity. MICRO 2013
• CoScale: Coordinating CPU and Memory System DVFS in Server Systems. MICRO 2013
• Predicting Performance Impact of DVFS for Realistic Memory Systems. MICRO 2013
• Systematic Energy Characterization of CMP/SMT Processor Systems via Automated Micro-Benchmarks. MICRO 2013
• AUDIT: Stress Testing the Automatic Way. MICRO 2013
• MorphCore: An Energy-Efficient Microarchitecture for High Performance ILP and High Throughput TLP. MICRO 2013
• Composite Cores: Pushing Heterogeneity into a Core. MICRO 2013
• Amoeba-Cache: Adaptive Blocks for Eliminating Waste in the Memory Hierarchy. MICRO 2013
• Improving Cache Management Policies Using Dynamic Reuse Distances. MICRO 2013
• Inferred Models for Dynamic and Sparse Hardware-Software Spaces. MICRO 2013
• Neural Acceleration for General-Purpose Approximate Programs. MICRO 2013
• High-performance and Energy-efficient Mobile Web Browsing on Big/Litttle Systems. HPCA 2013
• Enabling Distributed Generation Powered Sustainable High-Performance Data Center. HPCA 2013
• Store-Load-Branch (SLB) Predictor: A Compiler Assisted Branch Prediction for Data Dependent Branches. HPCA 2013
• ReCaP: A Region-Based Cure for the Common Cold (Cache). HPCA 2013
• Navigating Heterogeneous Processors with Market Mechanisms. HPCA 2013
• Application-to-Core Mapping Policies to Reduce Memory System Interference in Multi-Core Systems. HPCA 2013
• Modeling Performance Variation Due to Cache Sharing. HPCA 2013
• A Novel System Architecture for Web Scale Applications Using Lightweight CPUs and Virtualized I/O. HPCA 2013
• Hybrid Latency Tolerance for Robust Energy-Efficiency on 1000-Core Data Parallel Processors. HPCA 2013
• Optimizing Virtual Machine Scheduling in NUMA Multicore Systems. HPCA 2013
• Disintegrated Control for Energy-Efficient and Heterogeneous Memory Systems. HPCA 2013
• Illusionist: Transforming Lightweight Cores into Aggressive Cores on Demand. HPCA 2013
• ESESC: A Fast Multicore Simulator Using Time-Based Sampling. HPCA 2013
• Toward Energy-Efficient Manycores for Near-Threshold Computing. HPCA 2013
• A Multiple SIMD, Multiple Data (MSMD) Architecture: Parallel Execution of Dynamic and Static SIMD Fragments. HPCA 2013