Talk:CSC 456 Fall 2013/7a ac

Comments on first draft

I suggest the following papers:

http://dl.acm.org.prox.lib.ncsu.edu/citation.cfm?id=1375591 Foundations of the C++ concurrency memory model

http://dl.acm.org.prox.lib.ncsu.edu/citation.cfm?id=1006236 Evaluating support for global address space languages on the Cray X1

http://dl.acm.org.prox.lib.ncsu.edu/citation.cfm?id=1018205 Modern concurrency abstractions for C#

Let me add these: http://dl.acm.org.prox.lib.ncsu.edu/citation.cfm?id=1178609 What do high-level memory models mean for transactions?

Replacing Locks by Higher-Level Primitives - CiteSeer (Search for this one; Google wouldn't give me a complete URL).

Comments on the second draft

The Cederman paper is very apropos. Would like to see you explain the mechanisms and summarize results.

The Apple article is also good, but a bit lower level than I would like you to focus.

The C++ synchronization article is very good, but I think it would be more difficult to adapt; it seems to rely heavily on code fragments.

Comments on the third draft

Please use a different font or typeface for names of operations/methods.

Please give examples of how the various mechanisms can be used to solve synchronization problems. These examples should be as short as possible, definitely less than 10 lines, and can use comments in place of the details of what work is done in critical sections.

OpenMP does not really belong in the list of synchronization types. It is an interface that has support for many different patterns of synchronization. It is more closely related to the language section. Along with OpenMP, you should consider MPI.

The language section consists of only a table. While the table is useful, you should cite sources for the blocks you fill in, and make comments about gaps in the table. (Is support for a particular mechanism hard to achieve in some language? Are there third-party libraries to fill the void, etc.)