Talk:CSC 456 Fall 2013/1d vb

Round 3 comments

Your running text should include citations.

It would be good to include a table of pipeline length in several processors, not just the 3 you have shown. You could look at the tables in the 1c submission (http://wiki.expertiza.ncsu.edu/index.php/CSC_456_Fall_2013/1c_wa) and look up pipeline parameters in those. With this greater data, you should be able to make more observations.

Comments from Round 2

The page could profit from an introduction. Maybe you could edit the writeup I gave you into a good introduction.

I'd change "Trends of Increasing Pipeline Length" to something like, "Factors Favoring Longer Pipeline Length". Ditto for decreasing.

The word "precise" in the second line is not quite correct. Rather than "small and precise", I'd rather you say "short". The idea is that the work is split into more and more distinct steps.

In the "Trends of Decreasing Pipeline Length", you could say that there are two major issues, and then enumerate them.

It would be better to explain the issues in your own words rather than to quote the cited works. Short quotes are fine, but they should not constitute the majority of the explanation.

It would be good to include a graph of pipeline length vs. time. In order to do that, though, you need more data points.

For the "Number of Pipelines" column, I'd like to see your sources, perhaps during class. I'm not sure you're comparing apples with apples.

As to the content, you should cite different papers, processor specs, etc. from different times. This will be considerable work, but it represents a lot of the value added of this work.

On the citations, you should give the title and author, and make the title link to the Web page. This is because readers will look at your reference list, and they should be able to determine what kinds of articles you are linking to without following the links.

Thanks for your help on this!

Comments from Round 1

Here are some suggestions for improving your page.

• Another limitation on pipeline depth is the overhead of latching, which is needed at the end of each stage, so partial results can be "pulsed" to the next stage.

• I did a Google Scholar search on "pipeline depth". An alternative is to search both the ACM Digital Library and IEEE Xplore.

• Power is another consideration; see http://dl.acm.org/citation.cfm?id=956566.

• It is important not just to give the results, but to explain how and why the authors have reached the conclusions that they did.

• The first several hits for this topic are from 2002-2003. Earlier work by Kunkel and Smith from 1997 is also worth citing, to show how the situation changed during the intervening years. I find later papers on the IBM Cell processor, about 2005. Then there are several references related to power from about 2008.

• In addition to theoretical and simulation results, it would also be good to look at trade publications for pipeline depths of widely used processors over the years.