Talk:MESI protocol

Page contents not supported in other languages.
From Wikipedia, the free encyclopedia

Error[edit]

Illinois Protocol is not MESI, it is a different thing. Read Jim Handy "The Cache Memory Book" Second Edition. 4.3.6. Please remove redirection. —Preceding unsigned comment added by 194.67.67.138 (talk) 15:19, 24 August 2009 (UTC)[reply]

Great page, just one doubt; in the Table1.3 example, step3 says no access to main memory whereas there is a flush involved because P3 is trying to read modified data by P1. — Preceding unsigned comment added by 106.200.204.65 (talk) 18:44, 18 July 2017 (UTC)[reply]

State transitions[edit]

  • BusUpgrd - half of busRdX: throw away your copy.
  • BusRdX - asks for a copy with intent to modify: 1. give me data 2. throw away your copy.
  • BusRd - asks for copy with no intent to modify.
  • Flush - writes to memory.
  • BusWB - updates memory (synonym for Flush).


In table 1.1, is it possible that line 2 (Invalid receives PrWr) contains a mistake? "If other Caches have copy, they send value, otherwise fetch from Main Memory" - probably doesn't belong there. — Preceding unsigned comment added by 77.138.251.232 (talk) 19:21, 21 September 2019 (UTC)[reply]

Diagram[edit]

A state diagram would be very valuable here.

Diagram Error[edit]

The path from I to E should have the label PrRd / BusRd(~S) (S-bar) indicating not-shared. This diagram is correct: http://pg-server.csc.ncsu.edu/mediawiki/images/4/4d/MESI.jpg -pflents — Preceding unsigned comment added by 130.86.104.236 (talk) 20:30, 5 December 2011 (UTC)[reply]

The path from S to M should be labeled PrWr / BusUpgr : if the cache is in the shared state for a given block, then a PrWr on this block will necessarily be a hit. — Preceding unsigned comment added by 141.115.77.18 (talk) 07:34, 12 September 2017 (UTC)[reply]

Friendly Thoughts[edit]

I just want to say that this page is AWESOME. —Preceding unsigned comment added by 216.27.163.78 (talk) 05:57, 29 April 2009 (UTC)[reply]

Alternatives[edit]

> So MESI in practise doesn't quite work - not a problem if you're single threaded, but definitely a problem if not.

This last sentence of the article is very confusing. Would it be more correct to say that MESI performance on multicore processors suffers and name any currently known alternatives?