CS 433: Computer System Organization

Course Information for Fall 2021

Instructor: Sarita Adve

Introduction to Prof. Adve

PhD from Wisconsin in 1993, at Illinois since 1999

​

Main research area – architecture

−Cross-disciplinary work: languages, OS, applications, AR/VR…

​

Research interests and major projects

− Scalable specialization (Emphasis on Augmented and Virtual reality)

− Coherence and Consistency

− Heterogeneous memory systems

− Software-driven hardware reliability

− Approximate computing

− Cross-layer power management

Your introduction?

​

Due to an unforeseen COVID related situation, I have to teach remotely on zoom for a few lectures (~3 weeks). The meeting link is on the class web site and was also emailed to all registered students. We will announce on piazza when I will be back in-person.

​

I encourage you to go to our classroom 1404SC during our class time even while I am remote. Although you will have to watch the lecture on zoom on your laptop, this will allow you to get to know your classmates.

​

You can join a zoom lecture anonymously. Please get in touch with Prof. Adve in that case. Instructions for joining a meeting anonymously are here: https://wiki.illinois.edu/wiki/display/CSID/Attending+Sessions+Anonymously

​

All classes will be recorded.

​

Course Staff

• Instructor: Sarita Adve
• sadve@illinois.edu
• 217-333-8461
• Office hours:
• TTh 12:15-12:45pm Illinois time and by appointment on the lecture zoom meeting while remote
• TTh 12:30-1:00pm Illinois time and by appointment in 4104SC when switch to in-person
• Teaching Assistant: Omri Mor
• omrimor2@illinois.edu
• Office and office hours are on the web site
• Administrative Assistant: Kim Baker
• kabaker@illinois.edu, 217-300-6504

​

Website and Online Discussion

• Course home page: http://courses.grainger.illinois.edu/cs433/
• For lecture notes, homeworks, solutions sets, and other handouts

​

• Piazza: https://piazza.com/illinois/fall2021/cs433
• For all course announcements from course staff
• For online discussion, clarifications, etc. between staff and students
• This is the primary medium of communication between staff and students. It is your responsibility to check this site regularly.

Lectures

• Tuesdays and Thursdays, 11:00-12:15 pm Illinois time
• Meeting ID: 857 8571 8741
• Password: 737822

​

Pre-requisites

• CS 233 or equivalent
• You should be familiar with Chapters 1-5 and parts of Chapter 6 of “Computer Organization and Design: The Hardware Software Interface”, by Patterson and Hennessy, 6th edition, with specific emphasis on the topics on the next slide

​

Pre-requisites: Specific 233 topics (see previous slide)

• Instruction sets
• Familiarity with at least one RISC instruction set (e.g. MIPS, ARM, RISC V, Alpha, SPARC, ...)
• Addressing modes, arithmetic instructions, control instructions (jump, conditional branch), procedure calls and returns
• Basics of computer arithmetic
• Basic understanding for assessing performance
• Basic pipelined implementation
• Hazards (structural, data, control)
• Memory hierarchies - caches and virtual memory
• Concepts of ocality, miss vs. hit, miss and hit ratios, block/line, page, page fault, address translation, purpose of a TLB, associativity, block replacement policies (LRU, random, etc.), block finding (index, tag, etc.), write-through vs. write-back
• Basics of I/O - disks and buses
• Basic familiarity with shared memory and snooping based cache coherence and the SIMD model of computation

Course Material, Lecture Notes, Recordings

• John L. Hennessy and David Patterson, “Computer Architecture: A Quantitative Approach”, Elsevier, sixth edition
• The text will be supplemented with some additional material that will be provided in the lectures
• Most lecture notes will be available from the course web site the day before they are used in class
• Lecture recordings will be available from the web site, but these are intended for supplementary purposes
• If you are unable to attend synchronous lectures for a good reason (e.g., COVID related issues), please get in touch with me. Wanting to take a class that conflicts with this one is not a good reason - I cannot provide overrides for such conflicts

Course Topics

• Introduction – review of fundamental performance issues, power and reliability, cost vs. price, basic pipeline structure
• Instruction level parallelism – hardware and software techniques (e.g., dynamic scheduling, superscalar, static and dynamic branch prediction, VLIW, loop unrolling)
• Memory hierarchy – advanced concepts in caches, main memory, and virtual memory
• Multiprocessors/multicore – overview of different models, cache coherence with shared-memory systems/multicore (snoopy and directory solutions), synchronization, memory consistency models
• Data parallel architectures (e.g., SIMD, vectors, GPUs)
• Domain-specific architectures – depending on available time
• Storage systems, I/O – depending on available time
• Recent advances in architecture and future challenges – depending on available time

Assignments

• There will be several written assignments
• Usually due at 10pm Illinois time on due date
• An automatic extension of 10 minutes (i.e., until 10:10 p.m. Illinois time) is given without any further request. No further extensions for late submissions will be given except in the following cases:
• For exceptional reasons, an agreement to submit an assignment late may be reached with Professor Adve. Except for unforeseen medical or other emergencies, the agreement must be made within 48 hours after the assignment is handed out to the class. In all cases, you must get an email confirmation of the agreement. I am liberal with granting exceptions when requested in advance and responsibly (e.g., “excessive work this week due to a research deadline or three midterms” is a fine reason)
• Submit pdf through Illinois Compass 2g (https://compass2g.illinois.edu/)
• Preferably write in latex, word, etc., but pdf scans of hand-written solutions will be accepted

Exams

• One midterm exam:
• October 18, Monday, 7pm to 9pm Illinois time, Siebel 1404
• Will include material covered until and including the last class before the midterm
• Final exam:
• December 15, Wednesday, 8am to 11am, place will be announced by the campus office
• Will include material covered after the midterm and until last lecture
• If you will be remote for COVID related reasons, please contact me as soon as possible

Mini-projects

• Graduate students:
• Research how concepts presented in class are applied in a real system
• Present findings to class (towards end of semester)
• Work in small groups
• All students (grad and undergrad):
• All students are required to attend the mini-project presentations
• Brief questions from presentations will appear on the exam
• More details will be announced shortly (after we have covered some more material in the lectures)

Grading

• Undergraduate students: assignments 25%, midterm: 37%, final: 38%
• Graduate students: assignments: 20%, mini-projects: 14%, midterm: 33%, final: 33%
• Absolute scale. Minimum score for guaranteeing a grade:
• Undergrads: A+: 95%, A: 90%, A-: 85%, B+: 78%, B: 70%, B-: 65%, C+: 58%, C: 52%, C-: 45%, D: 35%, F: 0%
• Grads: A+: 97%, A: 92%, A-: 87%, B+: 80%, B: 75%, B-: 70%, C+: 62%, C: 55%, C-: 48%, D: 35%, F: 0%
• No D- and D+ grades will be alloted. Under exceptional circumstances, the minimum scores on some grades may be reduced, but they will not be increased

Regrade Requests

• If you think you have been unfairly graded on a homework or exam, you should petition the TA or Prof. Adve in writing within a week of distribution of the graded work. After a week, no regrade requests will be entertained

Honor Code (1 of 3)

• You may work with one partner for each assignment (different partners for different assignments are encouraged)
• You may discuss the problem and solution at length with your partner, but each partner must finally write up her or his own solution independently (on your own, with no more input from your partner) and submit an individual assignment
• Write the name of your partner clearly on the submitted work
• The partner will not get any credit for your submission
• Both partners must actively participate in the development of all solutions
• You may seek clarifications on the assignment problems from other students in the class as well, but you may not discuss the solutions with anyone other than your partner
• Exams are to be done individually. No books, notes, electronic devices, handouts, etc. may be used during the exams.

Honor Code (2 of 3)

Unless made available by the course staff, assignments, exams, and solution sets from previous offerings of the course or from other universities may not be used for this course. If you use such material, it is likely you will find the solutions to the problems in the assignments and the exams – we spend much time formulating the best possible set of problems and it is not possible to invent all new assignments for each course offering. Your use of such solutions will entirely defeat the goal of helping you learn the material and violate our honor code principles.

​

Honor Code (3 of 3)

I call the policies on the previous two slides and this slide the honor code because I would like to largely rely on your honor to enforce them – you are the only one to lose when you cheat. I consider both giving and receiving help beyond that allowed by the honor code policies to be forms of cheating, and take a violation of the honor code very seriously. My default is to give anyone found violating the above policies a zero on the entire section of the course where the policy is violated (where a section is all homeworks, the midterm, or the final); further, after the student’s scores are adjusted as above, the resulting grade will be reduced by a whole letter grade to determine the final grade. For example, if a student cheats on one homework problem, a zero score will be assigned on all homeworks. If this adjustment results in a total course score that would normally give a B- grade, then the final grade will be a C-. All cheating cases will be handled through the College of Engineering reporting system and will become part of the students’ permanent academic record. So please read these policies very carefully.

​