Talk:CSC 456 Fall 2013/8a wa: Difference between revisions

From Expertiza_Wiki
Jump to navigation Jump to search
No edit summary
 
Line 11: Line 11:
A good source for performance on other protocols is "Understanding performance issues on both single core and multi-core Architecture" from Computer Organization 2007; look it up in Google Scholar.
A good source for performance on other protocols is "Understanding performance issues on both single core and multi-core Architecture" from Computer Organization 2007; look it up in Google Scholar.


Have you changed besides inserting the paragraph on MESI?  I'm mainly interested in relevant performanc stats.
Have you changed besides inserting the paragraph on MESI?  I'm mainly interested in relevant performance stats.
 
== Comments on third draft ==
 
MOESI is lacking in description relative to the other protocols.  Can you provide a diagram and tell how it differs from MESI and MOSI?
 
On MESIF, please give the motivation for developing this new protocol.  What shortcoming(s) did it address in earlier protocols?  Please define "housing cache."  Is this different from the owning cache?
 
When you say that MESIF will outperform any bus-based protocol, please give evidence for this, preferably numerical.  (That is, how do we know that the number of hops is a major factor in determining protocol performance?)  Describe how the F state cuts down on bus traffic.

Latest revision as of 15:43, 19 November 2013

Comments on first draft

Glad to see that you have added info & a diagram on MESIF. Can you compare the protocols, e.g., search for "compare MESI MESIF" or "comparison MESI MESIF"? Same thing for comparing with MOSI and MOESI. It would be good if you could summarize quantitative comparisons, if available.

Comments on second draft

At the top or your page, please insert a link to the old version of this page.

Search Google Scholar for "MESIF performance." The second and third hits are good discussions. The performance advantages aren't measured in terms of time, but rather in number of messages required to perform certain operations. You may be possible to find comparisons of other protocols using the same metric.

A good source for performance on other protocols is "Understanding performance issues on both single core and multi-core Architecture" from Computer Organization 2007; look it up in Google Scholar.

Have you changed besides inserting the paragraph on MESI? I'm mainly interested in relevant performance stats.

Comments on third draft

MOESI is lacking in description relative to the other protocols. Can you provide a diagram and tell how it differs from MESI and MOSI?

On MESIF, please give the motivation for developing this new protocol. What shortcoming(s) did it address in earlier protocols? Please define "housing cache." Is this different from the owning cache?

When you say that MESIF will outperform any bus-based protocol, please give evidence for this, preferably numerical. (That is, how do we know that the number of hops is a major factor in determining protocol performance?) Describe how the F state cuts down on bus traffic.