Grading Policy

Grade Breakdown:

  • 25% - 2 Exams 
  • 35% - Homework Assignments, Paper Critiques, Class Participation, Literature Survey
  • 40% - Course Project 

Exams (25%):

  • Two exams: 
    • Mid term - (TBD, Week 8)
    • Final - (5/8/12, 3:30-5:30 p.m.)

Assignments, Participation and Literature Survey (35%):

  • 10% - HW and programming assignments + Paper critique
  • 5% - Class participation
  • 20% - Literature survey

Assignments (10%):

Homeworks:

  • You are expected to turn in the HW's into eLearning by midnight of the day the HW is due.
  • HW's must be formatted as a pdf file to receive credit.
  • Any time after midnight is considered late.
  • Late assignments will have a 25% penalty per day. 
  • HWs more than 4 days late will not be accepted

Paper Critiques:

  • Read paper and write critique/answer questions due day paper is discussed in class.
  • Critique template posted on course website (TBD).
  • Paper critiques are due into eLearning before the beginning of the class period (for grading).
  • Paper critiques must also be formatted as a pdf to receive credit.
  • Students should bring either an electronic or paper copy of their critique to class each week for reference during the discussion.
  • Late paper critiques will not be accepted.

Class Participation (5%):

  • Come prepared to class
  • Your class participation grade will be based upon your ability to keep up with and add to the discussion during class.
  • Some readings will be due before each class, we discuss these readings during class.
  • Be on-time
  • Participate in discussions
  • No Disruptive behavior

Literature Survey (20%):

Survey Topic Proposal: 

  • 2% of grade
  • Due 1/30/12 - Via eLearning
  • Form groups of 2-4 students
  • Let me know if you have difficulty finding partners
  • Proposal should detail exactly who will work on which component of the literature survey.
  • A 1-2 paragraph literature survey topic proposal

Literature Survey Report:

  • 18% of grade
  • Due 2/20/12 - Via eLearning
  • Literature survey should be 6-8 pages including references
  • At least 5 non-webpage/wikipedia references
  • IEEE or ACM conference paper format
  • I will provide latex template (you are on your own if you want to use Word)
  • Example Topics: 
    • A detailed comparison of 2 or more modern superscalar microprocessors – Form a group – each person studies a different processor. Then, jointly write a report comparing and contrasting the 2 processors. The features you study must include issue width, out of ordering, branch prediction, caches, prefetching, etc. Create a table summarizing your findings.
    • Caches on Modern Microprocessors – Survey at least 6 different processors and write a report comparing and contrasting caches on different processors. Size, associativity, number of levels, architecture of the cache, cache assists, prefetching, cache coherency, etc. The report should include a table that you create summarizing the findings.
    • Branch Predictors in Modern Microprocessors – Survey at least 6 different processors and write a report comparing and contrasting the branch predictors in those processors. The report should include a table that you create summarizing the findings. 
    • Cache Consistency Models in Modern Microprocessors – Survey at least 4 different processors and write a report comparing and contrasting the memory consistency models in those processors. The report should include a table that you create summarizing the findings.  Be sure to discuss the implications for out-of-order execution within a processor and across processors in a shared memory system.
    • Cache Coherency in Modern Multiprocessors – Survey at least 4 different multi-processor systems and write a report comparing and contrasting the methods for maintaining. The report should include a table that you create summarizing the findings.
    • Inter-processor interconnects - Survey of current implemented interconnects in modern CMPs, state of research for many-core interconnect/networks-on-chip (NoCs).
    • New interconnect technologies (on-chip nanophotonics, RF interconnect architectures, Freespace Optics, etc.) - implications for network-on-chip architectures.
    • New "resistive" memory technologies (PCM RAM, Memristors, MRAM, etc) - implications for main memory and storage.
    • Techniques for high ILP in future process technologies (EDGE/TRIPS, core-fusion, wavescalar, RAW) - is ILP dead?
    • Power management strategies
    • Transactional memories
    • GP-GPU (General purpose computing using graphics processors)
    • SMT approaches, performance
    • Prefetching mechanisms
    • Specialized/novel microarchitectures (Anton, GPUs, physics accelerators etc)
    • Hardware support for debugging parallel programs
    • Reliability - microarchitectural approaches to address the challenge of reliability in future process tech.
    • Microarchitectural support for virtual machines 
    • Survey available architecture simulators, contrast detail, features and speed of simulation 
  • Start reading now!
  • Pick a topic, find partners, put together an outline and references identified (see calendar for due dates)
  • A literature survey is a particular form of research paper which requires more than just a summary of each of the works listed, it requires synthesis of the knowledge gained and comparison and contrasts drawn between the works examined. 
  • UCSD has a nice guide and example of how to write a good literature survey
  • At the end of the survey an appendix should be added to identify the contributions of each student in the group.
  • Good Sources for reading: 
    • Microprocessor Report (Paper copy available in library)
    • ACM and IEEE International Symposium on Computer Architecture (ISCA)
      • Available electronically via IEEE Xplore and the ACM Portal
    • ACM International Conference on Architectural Support for Programming Languages and Operating Systems (ASPLOS)
      • Available electronically via the ACM Portal
    • IEEE International Symposium on Microarchitecture (MICRO)
      • Available electronically via IEEE Xplore and the ACM Portal
    • IEEE International Symposium on High Performance Computer Architecture (HPCA)
      • Available electronically via IEEE Xplore
    • The Worldwide Computer Architecture Home Page at Wisc. is a very good source of information

Course Project (40%):

3% - Project Proposal (Due: 2/24/12)
2% - Project Status Report (Due: 3/19/12)
20% - Project Final Report (Due: 4/30/12, last day of class)
15% - Project Presentation (Due: 4/15/10 - 4/29/10)

  • Form groups of 2-4 students
  • May be same group as literature survey
  • Example projects:
    • Simulate a chip-multiprocessor (CMP) and analyze the impact of different cache coherency schemes on multi-threaded benchmark (PARSEC, Splash-2, etc) performance
    • Simulation of a superscalar processor and analyzing impact of design tradeoffs (instruction window size, issue width, functional unit makeup, etc.).
    • Implement various adaptive routing algorithms for on-chip interconnection network from the literature and compare performance
    • Characterize on-chip interconnect traffic under different single and multiprocessor workloads
    • Make an enhancement to a superscalar processor and analyze its performance impact.
    • Performance measurements of a modern processor on an actual machine using some state of the art benchmark or application and analyze bottlenecks in that processor.
    • Comparison of simulated versus real processor performance for a current modern processor, identifying where simulation is lacking.
    • Comparison of different modern processors using performance counter measurements
    • VHDL/Verilog Simulation of a simple superscalar processor (extend ECEN651 MIPS processor to be superscalar)
    • Implementation of a simple superscalar processor in an FPGA.
    • Adaptive last-level cache sharing, develop policies to effectively share the last-level cache in a CMP under multiprogram workloads.
    • Examine prefetching in modern multiprocessors, what are the trade-offs and benefits in a multiprogrammed system.
    • Examine reliability and wear-out in CMP design, propose microarchitectural enhancements to improve operating lifetime.
    • Reproducing results from a published paper from ISCA, ASPLOS, HPCA, or MICRO will be acceptable as a project. 
    • If you make an extension to what has been published, that will be excellent. 
    • More project ideas to come...

Project proposal:

  • Approx 1-2 pages
  • ACM/IEEE conference paper format
  • Will provide latex templates (you are on your own if you use MS Word)
  • Should include:
    • Objectives - What are you trying to find out? What’s the problem you are trying to solve?
    • Background and Motivation - What have others done in this area?
    • Relevance - Why do you think this project is important? What is the significance of this work?  Relate what you are doing to what others have done before.
    • Methodology - How will you accomplish your goals?  Which group members will do what tasks?
    • Expected Outcome - What do you expect would be the outcome of the project once it is completed.

Project status report:

  • 1-2 paragraphs detailing the work completed and the work left to do

Project report extending literature survey and project proposal:

  • Experimental results
  • Place results into greater context
  • Was this the expected outcome?  If not why?
  • What conclusions did you reach?
  • Expect 8-12 pages
  • ACM/IEEE conference paper format

Project presentation:

  • 15% of course grade
  • 10-15 minute presentation in class
  • Presentations that run over 15 minutes will be penalized 
  • Practice to make sure you don't run over!
  • 2-3 minutes for Q&A
  • 2 Weeks of class presentations (4/23/12 - 5/2/12)
  • Sign-up website: TBD
  • Groups that signup for the first two days of presentations (4/23/12 or 4/25/12) gain 3% extra credit on course grade
  • Each member should participate in the presentation
  • Sample generic presentation outline (expect ~7-15 slides total for a 15 minute presentation, spend 1-2 minutes per slide):
    • Title (1 slide) - Name of project, list of all participants.
    • Introduction (1-3 slides) - Introduce the overall topic.  Answers the question "Why should the audience care about your project?"   
    • Background and Prior Work (1-2 slides) - Summarize relevant parts of Lit Survey. What is the current state of research on this area?
    • Project Description (2-5 slides) - Discuss what you did and why.
    • Project Evaluation (2-5 slides) - Discuss the project evaluation results.  Start with at least one slide of methodolgy (how your project is evaluated).  Comment upon the results, what do they mean, what is important, was anything unexpected.
    • Conclusions (1-2 slides) - Summarize the project and highlight the important results.
  • You have the option of either emailing me (pgratz@gratz1.com) a copy of your presentation ( in .pdf (preferred), .ppt or .odp formats) 30 minutes before class or bringing a laptop to class and driving it from there.  Either way, presenters should arrive 15 minutes early to class and make sure the presentation works properly in the classroom.
Comments