Description
General-purpose software must become more parallel in order to scale to microprocessors that are providing more, instead of faster, cores. However, programmers using high-performance, general-purpose languages such as Java and C++ are basically stuck using shared memory with locks, making it hard to achieve both parallelism and correctness. Software and hardware systems and dynamic and static program analyses can provide better support for achieving correct parallelism, but existing approaches have serious limitations or drawbacks.
In this course, we'll explore -- and perhaps improve on -- the state of the art in using program analysis and runtime systems to achieve parallel systems that are reliable and scalable. Students will read papers, discuss in small groups, write critiques, and lead in-class discussions. Students will undertake a project focused on the research topics explored in the course.
There are no formal prerequisites, but it may help to have some background in programming languages, runtime systems, program analysis, compilers, and/or parallelism/concurrency (and potentially software engineering, architecture, and/or operating systems).
In this course, we'll explore -- and perhaps improve on -- the state of the art in using program analysis and runtime systems to achieve parallel systems that are reliable and scalable. Students will read papers, discuss in small groups, write critiques, and lead in-class discussions. Students will undertake a project focused on the research topics explored in the course.
There are no formal prerequisites, but it may help to have some background in programming languages, runtime systems, program analysis, compilers, and/or parallelism/concurrency (and potentially software engineering, architecture, and/or operating systems).
General Information
Topics and plan
We'll read and discuss about 2 papers each week. Some of those papers will be from the following list:
http://web.cse.ohio-state.edu/~mikebond/5349/papers.html
We'll focus on the RoadRunner infrastructure for dynamic analysis of concurrent programs (http://www.cs.williams.edu/~freund/rr/), but it's possible to use other infrastructures (e.g., hardware simulators) for projects.
http://web.cse.ohio-state.edu/~mikebond/5349/papers.html
We'll focus on the RoadRunner infrastructure for dynamic analysis of concurrent programs (http://www.cs.williams.edu/~freund/rr/), but it's possible to use other infrastructures (e.g., hardware simulators) for projects.
Discussing and critiquing papers
The following links (critiques and policies) provide a lot of general info. Here's some specific info for this semester:
- Everyone except discussion leader: send critiques as a private Piazza post to Mike by Monday at noon. Mike will make them publicly viewable shortly after noon.
- Discussion leader: send your discussion slides, notes, and/or questions to Mike by Monday at 2 pm, so he can give you feedback. You're welcome to ask questions or ask for feedback much sooner than that, e.g., if you have questions after reading the papers but before starting to work on discussion topics.
http://web.cse.ohio-state.edu/~mikebond/5349/critiques.html
- Everyone except discussion leader: send critiques as a private Piazza post to Mike by Monday at noon. Mike will make them publicly viewable shortly after noon.
- Discussion leader: send your discussion slides, notes, and/or questions to Mike by Monday at 2 pm, so he can give you feedback. You're welcome to ask questions or ask for feedback much sooner than that, e.g., if you have questions after reading the papers but before starting to work on discussion topics.
http://web.cse.ohio-state.edu/~mikebond/5349/critiques.html
Course policies
Course calendar
The official source for class times, class plans, and office hours will be the course calendar:
https://calendar.google.com/calendar/embed?src=v6lifs6181s08vhl38h7be3ulg%40group.calendar.google.com&ctz=America%2FNew_York
https://calendar.google.com/calendar/embed?src=v6lifs6181s08vhl38h7be3ulg%40group.calendar.google.com&ctz=America%2FNew_York
Acknowledgments
Thanks to Luis Ceze, Cormac Flanagan, Steve Freund, Dan Grossman, Shan Lu, Kathryn McKinley, Feng Qin, Nasko Rountev, and Craig Zilles, as well as others whom I'm probably missing, for providing advice, feedback, inspiration, and other help with this course.
| Name | Office Hours | |
|---|---|---|
Michael Bond | When? Where? |
General Resources
Nothing has been added to the General Resources section, yet. Stay tuned!