Talk:CSC/ECE 506 Spring 2014/10c gk: Difference between revisions

From Expertiza_Wiki
Jump to navigation Jump to search
No edit summary
No edit summary
 
(8 intermediate revisions by 2 users not shown)
Line 4: Line 4:
2. Introduce the issue of SC performance on a mobile platform like Android(ARM architecture) <br>
2. Introduce the issue of SC performance on a mobile platform like Android(ARM architecture) <br>
3. Discuss in detail relaxed memory consistency models for mobile platforms like Android <br>
3. Discuss in detail relaxed memory consistency models for mobile platforms like Android <br>
4. Discuss how modern programming languages like C++/Java have built libraries using consistency models to provide atomic operations that can be used by low-level software developers effectively.<br>
4. Discuss how C++ provides low-level interfaces like atomics to synchronize memory accesses.<br>
5. Remove links to references that are no longer available<br>
5. Remove links to references that are no longer available<br>
http://www.ce.ewi.tudelft.nl/publicationfiles/915_463_bos.pdf<br>
http://www.ce.ewi.tudelft.nl/publicationfiles/915_463_bos.pdf<br>
Line 20: Line 20:


8. Some multi-choice exercises
8. Some multi-choice exercises
== Instructor comments ==
The planning document should have been submitted to Expertiza.  A reviewer would not have known to look on the talk page.
You were supposed to provide a link to the previous version of the page.
The title, "Introduction to Sequential Consistency Model" is confusing.  Why would you even need a heading here?  Generally, there should not be two headings in a row.  Just removing it would solve the problem.  There are some mechanical problems in this section: missing space after comma, "uniprocessors" for "uniprocessor", missing space before open paren.
I think the discussion of RCsc and RCpc is misplaced.  This becomes an issue only where synchronization ops need to be explicitly inserted, which usually is only after weak ordering is discussed.  RCpc presumes a discussion of processor consistency, which has not been provided yet.
The section on Relax Read-to-Read and Read-to-Write program orders discusses the models in terms of which platform they are used on.  It should discuss them in terms of the differences in models, and just mention as an example which platform uses which approach.
The section on Software based concurrency - C++ as an example just appears near the end of the page, with no apparent reason for including it, or placing it at this point in the discussion.
A quiz was added.  This is helpful.

Latest revision as of 14:06, 29 April 2014

We propose the following changes to the wiki chapter on relaxed consistency models:

1. Rewrite the introduction and explain the need for memory consistency models
2. Introduce the issue of SC performance on a mobile platform like Android(ARM architecture)
3. Discuss in detail relaxed memory consistency models for mobile platforms like Android
4. Discuss how C++ provides low-level interfaces like atomics to synchronize memory accesses.
5. Remove links to references that are no longer available
http://www.ce.ewi.tudelft.nl/publicationfiles/915_463_bos.pdf
http://www.cesr.ncsu.edu/solihin/Main.html
http://delivery.acm.org.www.lib.ncsu.edu:2048/10.1145/230000/224398/a61-kontothanassis.html?key1=224398&key2=6529911721&coll=ACM&dl=ACM&CFID=140829&CFTOKEN=60032635

6. Add more references to more recent scholarly articles
http://www.morganclaypool.com/doi/abs/10.2200/S00346ED1V01Y201104CAC016 http://delivery.acm.org/10.1145/1960000/1950858/p154-naeem.pdf?ip=149.173.1.43&id=1950858&acc=ACTIVE%20SERVICE&key=E4765D3E93FC8AF0%2EE4765D3E93FC8AF0%2E4D4702B0C3E38B35%2E4D4702B0C3E38B35&CFID=429804524&CFTOKEN=13003090&__acm__=1396380784_99942adbab461d58235c590a44d10a47 (2011)
http://delivery.acm.org/10.1145/1930000/1926393/p43-sevcik.pdf?ip=149.173.1.43&id=1926393&acc=ACTIVE%20SERVICE&key=E4765D3E93FC8AF0%2EE4765D3E93FC8AF0%2E4D4702B0C3E38B35%2E4D4702B0C3E38B35&CFID=429804524&CFTOKEN=13003090&__acm__=1396380834_dd1b4e9f590b79d54ad1a95ea4b47a4e (2011)
http://delivery.acm.org/10.1145/1960000/1950376/p67-devietti.pdf?ip=149.173.1.43&id=1950376&acc=ACTIVE%20SERVICE&key=E4765D3E93FC8AF0%2EE4765D3E93FC8AF0%2E4D4702B0C3E38B35%2E4D4702B0C3E38B35&CFID=429804524&CFTOKEN=13003090&__acm__=1396380973_155b28471dbd8b5126fb9fceee9246f0 (2011)
http://delivery.acm.org/10.1145/1760000/1755238/p8-naeem.pdf?ip=149.173.1.43&id=1755238&acc=ACTIVE%20SERVICE&key=E4765D3E93FC8AF0%2EE4765D3E93FC8AF0%2E4D4702B0C3E38B35%2E4D4702B0C3E38B35&CFID=429804524&CFTOKEN=13003090&__acm__=1396381165_1d9f9965ea71df09ccd8d7a6216f1925 (2010)

7. Add more diagrams and graphical analysis of system performance using the different models.

8. Some multi-choice exercises

Instructor comments

The planning document should have been submitted to Expertiza. A reviewer would not have known to look on the talk page.

You were supposed to provide a link to the previous version of the page.

The title, "Introduction to Sequential Consistency Model" is confusing. Why would you even need a heading here? Generally, there should not be two headings in a row. Just removing it would solve the problem. There are some mechanical problems in this section: missing space after comma, "uniprocessors" for "uniprocessor", missing space before open paren.

I think the discussion of RCsc and RCpc is misplaced. This becomes an issue only where synchronization ops need to be explicitly inserted, which usually is only after weak ordering is discussed. RCpc presumes a discussion of processor consistency, which has not been provided yet.

The section on Relax Read-to-Read and Read-to-Write program orders discusses the models in terms of which platform they are used on. It should discuss them in terms of the differences in models, and just mention as an example which platform uses which approach.

The section on Software based concurrency - C++ as an example just appears near the end of the page, with no apparent reason for including it, or placing it at this point in the discussion.

A quiz was added. This is helpful.