MO801 / MC972 - Topics in Architecture and Hardware - Multicore Systems: Software and Hardware

Notices

18/06	List of deliveries already received.
24/04	Grid of the presentations of work 1 are already in the calendar at the end of this page.
17/04	List of abstracts already received. This file will be updated continuously until the end of the course.
17/03	Class in room 317.
27/02	Check the important dates on Postgraduate calendar and graduation calendar.

Objectives

Study concepts related to Multicore Computer Architectures. The Course will be divided into two parts, one focusing on Software and the other focusing on Hardware. The software part will contain various practical activities in the laboratory.

REFERENCES

See list of articles for abstracts and class notes.

Some useful items to carry out the work:

• PTthreads: POSIX Threads Programming (tutorial). Basic presentation about PThreads.
• OpenMP: official site e tutorial. OpenMP Basic Presentation.
• GProf: Super fast guide, slightly more detailed guide. Gprof manual (use man gprof).
• Vectorization: Examples, Intrinsecs documentation, another intrinsec document, two interesting extra documents: 1 e 2.

Especially in relation to the programming APIs, there are many pages with good tutorials. Do a Google search on them.

Activities

Each student must make 8 abstracts during the semester, a presentation on a specific topic and three individual papers (one theoretical and two practical).

Enrolled students only in the MC972 course, only 4 abstracts must be delivered.

Abstracts

All articles below must be read and summarized during the course. Delivery dates are on the calendar at the end of this page. A summary must have only one page. As a guide, a summary should contain: Overview of the subject of the article, contribution of the article, results obtained and a comment on the relevance of this work both in relation to the area and in relation to o to the course.

1. Moore, GE, "Cramming More Components Onto Integrated Circuits," Proceedings of the IEEE, vol.86, no.1, pp.82-85, Jan 1998.
2. Agarwal, Anant; Levy, Markus, "The KILL Rule for Multicore," Design Automation Conference, 2007. DAC '07. 44th ACM / IEEE, vol., No., Pp.750-753, 4-8 June 2007.
3. Fan, X., Weber, W., and Barroso, LA 2007. Power provisioning for a warehouse-sized computer. In Proceedings of the 34th Annual international Symposium on Computer Architecture (San Diego, California, USA, June 09 - 13, 2007). ISCA '07. ACM, New York, NY, 13-23.
4. Wenisch, TF, Ailamaki, A., Falsafi, B., and Moshovos, A. 2007. Mechanisms for store-wait-free multiprocessors. In Proceedings of the 34th Annual international Symposium on Computer Architecture (San Diego, California, USA, June 09 - 13, 2007). ISCA '07. ACM, New York, NY, 266-277.
5. Book. If it is not related to your work 1, summarize an article from the ISCA conference.
6. Book. If it is not related to your work 1, summarize an article from the ASPLOS conference.
7. Book. If it is not related to your work 1, summarize an article from the CGO conference.
8. Book. If it is not related to your work 1, summarize an article from the MICRO conference.

Delivery format: Send a summary by email, the subject must contain the discipline code (MO801), its RA and the summary number. Example: MO801 - 985306 - R1. Attach to your email your summary, in PDF format, with the name in the form RA-Numero_do_resumo.pdf. Example: 985306-R1.pdf. By doing this you will facilitate the identification of your email. Remember: there are 21 students who will send 8 emails, total: 168 emails with summaries!

Jobs

Each student must do the following tasks individually:

• Work 1 (theoretical). The first work has a free theme related to the discipline. It is divided into two parts, the written part and a presentation. Make the reservation of the theme and the date of the presentation in the wiki of the discipline. The text must follow the format standardized by SBC. The expected size is between 10 and 20 pages.
• Work 2 (practical): Parallelize a part of Imagemagick. The software is quite large, you must select a piece of it and work to optimize it. How to choose this piece? Make a profile of the code and define a test.
• Job 3 (practical): The third job is to parallelize a program. Choose a software, which can be from any area, including that of interest to you and implement a parallel version of it. Make a reservation at wiki of the discipline.

Presentations

Each student will make a presentation during the semester. The presentation will be on the theme of the chosen theoretical work.

Calendar

27/02	Presentation and Motivation of the Course
03/03	Reading and review of the first article. Comments.
05/03	Building a multicore architecture	ppt
10/03	Basic Concepts of Parallel Programming	ppt
12/03	Basic Concepts of Parallel Programming	ppt
17/03	Lab (PThreads and OpenMP). Room 317
19/03	Lab (PThreads and OpenMP). Room 317
24/03	Thread Problems	ppt
26/03	Deadline for booking topics for Work 1 and programs for Work 3. Vectorization.	ppt
31/03	There will be no MO801 / MC972 class
02/04	Cache Coherence	ppt
07/04	Architecture and Programming Course for Cell Processor (Leonardo Garcia - IBM)	pdf
09/04	Architecture and Programming Course for Cell Processor (Leonardo Garcia - IBM)	pdf
14/04
16/04	There was no class.
21/04	There will be no classes. (Unicamp calendar)
23/04	Profiling	log
28/04	Vectorization
30/04	Contact
05/05	Contact
07/05
12/05
14/05
19/05	Deadline for delivery of the Work 2. Presentations: Dynamic Parallelization of Code and Transactional Memories
21/05	Presentations: Energy saving in multi-processed embedded systems and Thread-Level Speculation (TLS)
26/05	There will be no class.
28/05	Presentations: Temperature-Aware Computer Systems and The Cilk Project
02/06	Presentations: Quality of service in CMPs and Branch Predictors
04/06	Presentations: Programming Languages, APIs and other tools for parallel programming and Comparison of memory models for multicolors
09/06	Presentations: Cache Coherence and Virtualization
11/06	Presentations: Intel Threading Building Blocks (TBB) and Auto Parallelization
16/06	Presentations: UltraSPARC T2 (Ni gara 2) and Scaling in multicolored architectures
18/06	Deadline for delivery of the Work 3. Presentations: Programming for GPU's and Parallelization of Execution Traces
23/06	Presentation of Work 3. Presentations: Simultaneous Multithreading (SMT) and Communication in Multicore Architectures
25/06	Presentation of Work 3
07/07	Exam (only for undergraduate students)