COMP 322: Fundamentals of Parallel Programming (Spring 2016)
Prof. Vivek Sarkar, DH 3080
|Head TA:||Max Grossman|
Annepha Hurlock, email@example.com, DH 3080, 713-348-5186
Prasanth Chatarasi, Arghya Chatterjee, Yuhan Peng, Jonathan Sharman
|Co-Instructor:||Dr. Shams Imam||Undergraduate TAs:|
Prudhvi Boyapalli, Peter Elmers, Nicholas Hanson-Holtry, Ayush Narayan, Timothy Newton, Alitha Partono, Tom Roush, Hunter Tidwell, Bing Xue
https://piazza.com/class/iirz0u74egl2q9 (Piazza is the preferred medium for all course communications, but you can also send email to comp322-staff at rice dot edu if needed)
Herzstein Hall 210
MWF 1:00pm - 1:50pm (followed by office hours in Duncan Hall 3092 during 2pm - 3pm)
DH 1064 (Section A01), DH 1070 (Section A02)
Wednesday, 07:00pm - 08:30pm
A summary PDF file containing the course syllabus for the course can be found here. Much of the syllabus information is also included below in this course web site, along with some additional details that are not included in the syllabus.
The primary goal of COMP 322 is to introduce you to the fundamentals of parallel programming and parallel algorithms, by following a pedagogic approach that exposes you to the intellectual challenges in parallel software without enmeshing you in the jargon and lower-level details of today's parallel systems. A strong grasp of the course fundamentals will enable you to quickly pick up any specific parallel programming system that you may encounter in the future, and also prepare you for studying advanced topics related to parallelism and concurrency in courses such as COMP 422.
The desired learning outcomes fall into three major areas (course modules):
1) Parallelism: creation and coordination of parallelism (async, finish), abstract performance metrics (work, critical paths), Amdahl's Law, weak vs. strong scaling, data races and determinism, data race avoidance (immutability, futures, accumulators, dataflow), deadlock avoidance, abstract vs. real performance (granularity, scalability), collective & point-to-point synchronization (phasers, barriers), parallel algorithms, systolic algorithms.
2) Concurrency: critical sections, atomicity, isolation, high level data races, nondeterminism, linearizability, liveness/progress guarantees, actors, request-response parallelism, Java Concurrency, locks, condition variables, semaphores, memory consistency models.
3) Locality & Distribution: memory hierarchies, locality, cache affinity, data movement, message-passing (MPI), communication overheads (bandwidth, latency), MapReduce, accelerators, GPGPUs, CUDA, OpenCL.
To achieve these learning outcomes, each class period will include time for both instructor lectures and in-class exercises based on assigned reading and videos. The lab exercises will be used to help students gain hands-on programming experience with the concepts introduced in the lectures.
To ensure that students gain a strong knowledge of parallel programming foundations, the classes and homeworks will place equal emphasis on both theory and practice. The programming component of the course will mostly use the Habanero-Java Library (HJ-lib) pedagogic extension to the Java language developed in the Habanero Extreme Scale Software Research project at Rice University. The course will also introduce you to real-world parallel programming models including Java Concurrency, MapReduce, MPI, OpenCL and CUDA. An important goal is that, at the end of COMP 322, you should feel comfortable programming in any parallel language for which you are familiar with the underlying sequential language (Java or C). Any parallel programming primitives that you encounter in the future should be easily recognizable based on the fundamentals studied in COMP 322.
The prerequisite course requirements are COMP 182 and COMP 215. COMP 322 should be accessible to anyone familiar with the foundations of sequential algorithms and data structures, and with basic Java programming. COMP 321 is also recommended as a co-requisite.
There are no required textbooks for the class. Instead, lecture handouts are provided for each module as follows. The links to the latest versions on Owlspace are included below:
- Module 1 handout (Parallelism)
- Module 2 handout (Concurrency)
- Module 3 handout (Distribution and Locality)
You are expected to read the relevant sections in each lecture handout before coming to the lecture. We will also provide a number of references in the slides and handouts.
There are also a few optional textbooks that we will draw from quite heavily. You are encouraged to get copies of any or all of these books. They will serve as useful references both during and after this course:
- Fork-Join Parallelism with a Data-Structures Focus (FJP) by Dan Grossman (Chapter 7 in Topics in Parallel and Distributed Computing)
- Java Concurrency in Practice by Brian Goetz with Tim Peierls, Joshua Bloch, Joseph Bowbeer, David Holmes and Doug Lea
- Principles of Parallel Programming by Calvin Lin and Lawrence Snyder
- The Art of Multiprocessor Programming by Maurice Herlihy and Nir Shavit
Past Offerings of COMP 322
Assigned Videos (Quizzes due by Friday of each week)
Lecture 1: Task Creation and Termination (Async, Finish)
|Module 1: Section 1.1||worksheet1||lec1-slides|
Lecture 2: Computation Graphs, Ideal Parallelism
|Module 1: Sections 1.2, 1.3||Topic 1.2 Lecture, Topic 1.2 Demonstration, Topic 1.3 Lecture, Topic 1.3 Demonstration||worksheet2||lec2-slides|
|Fri||Jan 15||Lecture 3: Abstract Performance Metrics, Multiprocessor Scheduling||Module 1: Section 1.4||Topic 1.4 Lecture, Topic 1.4 Demonstration||worksheet3||lec3-slides|
|Lecture & demo quizzes for topics 1.1, 1.2, 1.3, 1.4|
No lecture, School Holiday (Martin Luther King, Jr. Day)
Lecture 4: Parallel Speedup and Amdahl's Law
|Module 1: Section 1.5||Topic 1.5 Lecture, Topic 1.5 Demonstration||worksheet4||lec4-slides|
Lecture 5: Future Tasks, Functional Parallelism
|Module 1: Section 2.1||Topic 2.1 Lecture, Topic 2.1 Demonstration||worksheet5||lec5-slides||Lecture & demo quizzes for topics 1.5, 2.1 (topic 1.6 is optional)|
Lecture 6: Memoization
|Module 1: Section 2.2||Topic 2.2 Lecture , Topic 2.2 Demonstration||worksheet6||lec6-slides|
Lecture 7: Finish Accumulators
|Module 1: Section 2.3||Topic 2.3 Lecture , Topic 2.3 Demonstration||worksheet7||lec7-slides|
Lecture 8: Data Races, Functional & Structural Determinism
|Module 1: Sections 2.5, 2.6||Topic 2.5 Lecture , Topic 2.5 Demonstration, Topic 2.6 Lecture , Topic 2.6 Demonstration||worksheet8||lec8-slides|
|Homework 1, Lecture & demo quizzes for topics 2.2, 2.3, 2.5, 2.6|
Lecture 9: Map Reduce
|Module 1: Section 2.4||Topic 2.4 Lecture , Topic 2.4 Demonstration||worksheet9||lec9-slides|
|Lecture 10: Java’s Fork/Join Library||FJP chapter: Sections 7.3 & 7.5||worksheet10||lec10-slides|
Lecture 11: Loop-Level Parallelism, Parallel Matrix Multiplication, Iteration Grouping (Chunking)
|Module 1: Sections 3.1, 3.2, 3.3||worksheet11||lec11-slides||Lecture & demo quizzes for topics 2.4, 3.1, 3.2, 3.3|
Lecture 12: Barrier Synchronization
|Module 1: Section 3.4||Topic 3.4 Lecture , Topic 3.4 Demonstration|
Lecture 13: Iterative Averaging Revisited, SPMD pattern
|Module 1: Sections 3.5, 3.6||Topic 3.5 Lecture, Topic 3.5 Demonstration, Topic 3.6 Lecture, Topic 3.6 Demonstration||worksheet13||Worksheet12.java|
Lecture 14: Data-Driven Tasks and Data-Driven Futures
|Module 1: Section 4.5||Topic 4.5 Lecture, Topic 4.5 Demonstration|
Homework 2, Lecture & demo quizzes for topics
Lecture 15: Phasers, Point-to-point Synchronization
|Module 1: Sections 4.2, 4.3||Topic 4.2 Lecture, Topic 4.2 Demonstration, Topic 4.3 Lecture, Topic 4.3 Demonstration|
Lecture 16: Pipeline Parallelism, Signal Statement, Fuzzy Barriers
|Module 1: Sections 4.4, 4.1||Topic 4.4 Lecture, Topic 4.4 Demonstration, Topic 4.1 Lecture, Topic 4.1 Demonstration,|
Lecture 17: Abstract vs. Real Performance
|Lecture & demo quizzes for topics 4.1, 4.2, 4.3, 4.4|
Lecture 18: Midterm Summary
Midterm Review (Interactice Q&A using PollEverywhere)
|Exam 1 held during lab time (7:00pm - 10:00pm)|
Lecture 19: Task Scheduling Policies
|Topic 4.6 Lecture, Topic 4.6 Demonstration||Lec19HelpFirstWorkStealing.java||Homework 3 Checkpoint-1, Lecture & demo quizzes for topic 4.6|
Feb 29- Mar 04
Lecture 20: Critical sections, Isolated construct, Parallel Spanning Tree algorithm (start of Module 2)
|Module 2: Sections 5.1, 5.2, 5.3, 5.6||Topic 5.1 Lecture, Topic 5.1 Demonstration, Topic 5.2 Lecture, Topic 5.2 Demonstration, Topic 5.3 Lecture, Topic 5.3 Demonstration|
Lecture 21: Atomic variables, Read-Write Isolation
|Module 2: Sections 5.4, 5.5||Topic 5.4 Lecture, Topic 5.4 Demonstration, Topic 5.5 Lecture, Topic 5.5 Demonstration, Topic 5.6 Lecture, Topic 5.6 Demonstration|
Lecture 22: Parallelism in Java Streams, Parallel Prefix Sums
Homework 3 Checkpoint-2, Lecture & demo quizzes for topics 5.1 to 5.6
Lecture 23: Java Threads, Java synchronized statement
|Topic 7.1 Lecture, Topic 7.2 Lecture|
Lecture 24: Java synchronized statement (contd), wait/notify
|Topic 7.3 Lecture|
Lecture 25: Concurrent Objects, Linearizability of Concurrent Objects
|Topic 7.4 Lecture|
Homework 3, Lecture quizzes for topics 7.1 - 7.4
Lecture 26: Linearizability (contd), Java locks
|Topic 7.3 Lecture (recap), Topic 7.4 Lecture (recap)|
(3 weeks, with one intermediate checkpoint)
Lecture 26: Safety and Liveness Properties
|Topic 7.5 Lecture|
Lecture 27: Actors
|Topic 6.1 Lecture, Topic 6.1 Demonstration, Topic 6.2 Lecture, Topic 6.2 Demonstration, Topic 6.3 Lecture, Topic 6.3 Demonstration|
Lecture & demo quizzes for topics 7.5, 6.1 - 6.3
Lecture 28: Actors (contd)
|Topic 6.4 Lecture, Topic 6.4 Demonstration, Topic 6.5 Lecture, Topic 6.5 Demonstration, Topic 6.6 Lecture, Topic 6.6 Demonstration|
Lecture 29: Dining Philosophers Problem
|Topic 7.6 Lecture|
|Lecture quiz for topics 6.4 - 6.6, 7.6|
Lecture 31: Eureka-style Speculative Task Parallelism
Homework 4 Checkpoint-1
Lecture 32: Task Affinity with Places (start of Module 3)
Lecture 33: Message Passing Interface (MPI)
(2 weeks, with 1-week automatic extension)
Lecture 34: Message Passing Interface (MPI, contd)
Lecture 35: PGAS languages
Lecture 36: Memory Consistency Models
Lecture 37: GPU Computing
Lecture 38: Fortress language
Lecture 39: Course Review (lectures 20-37), Last day of classes
|Homework 5 (automatic extension till April 29)|
Scheduled final exam
Async-Finish Parallel Programming
Abstract performance metrics with async & finish
DIY HJ-lib Programming, Futures, HJ-Viz
Finish Accumulators and Loop-Level Parallelism
Loop Chunking and Barrier Synchronization
Data-Driven Futures and Phasers
No lab this week — Exam 1
No lab this week — Spring Break
Isolated Statement and Atomic Variables
Actors and Selectors
Eureka-style Speculative Task Parallelism
Parallel Pretty Pictures
|13||Apr 20||Message Passing Interface (MPI)|
Grading, Honor Code Policy, Processes and Procedures
Grading will be based on your performance on five homeworks (weighted 40% in all), two exams (weighted 40% in all), weekly lab exercises (weighted 10% in all), and class participation including worksheets, in-class Q&A, Piazza participation, and online quizzes (weighted 10% in all).
The purpose of the homeworks is to train you to solve problems and to help deepen your understanding of concepts introduced in class. Homeworks are due on the dates and times specified in the course schedule. Please turn in all your homeworks using the subversion system set up for the class. Homework is worth full credit when turned in on time. No late submissions (other than those using slip days mentioned below) will be accepted.
As in COMP 321, all students will be given 3 slip days to use throughout the semester. When you use a slip day, you will receive up to 24 additional hours to complete the assignment. You may use these slip days in any way you see fit (3 days on one assignment, 1 day each on 3 assignments, etc.). If you use slip days, you must submit a SLIPDAY.txt file in your SVN homework folder before the actual submission deadline indicating the number of slip days that you plan to use. Other than slip days, no extensions will be given unless there are exceptional circumstances (such as severe sickness, not because you have too much other work). Such extensions must be requested and approved by the instructor (via e-mail, phone, or in person) before the due date for the assignment. Last minute requests are likely to be denied. If you do receive an extension from the instructor, please indicate this by placing an EXTENSION.txt file in your SVN homework folder before the actual submission deadline indicating the date that the extension was granted by the instructor as well as the length of the extension.
Labs must be checked off by a TA prior to the start of the lab the following week.
Worksheets are due by the beginning of the class after they are distributed, so that solutions to the worksheets can be discussed.
You will be expected to follow the Honor Code in all homeworks and exams. All submitted homeworks are expected to be the result of your individual effort. You are free to discuss course material and approaches to homework problems with your other classmates, the teaching assistants and the professor, but you should never misrepresent someone else’s work as your own. If you use any material from external sources, you must provide proper attribution (as shown here). Exams 1 and 2 test your individual understanding and knowledge of the material. Exams are closed-book, and collaboration on exams is strictly forbidden. Finally, it is also your responsibility to protect your homeworks and exams from unauthorized access.
Graded homeworks will be returned to you via email, and exams as marked-up hardcopies. If you believe we have made an error in grading your homework or exam, please bring the matter to our attention within one week.
Accommodations for Students with Special Needs
Students with disabilities are encouraged to contact me during the first two weeks of class regarding any special needs. Students with disabilities should also contact Disabled Student Services in the Ley Student Center and the Rice Disability Support Services.