Dale, Re: "Also note that according to these tables, it isn't just a matter of 8-way being as good as fully associative. These tables actually indicate a higher hit rate with 8-way set associativity than with full associativity. Without some type of an explanation of this discrepancy, I would tend to reject this data outright."
Well that would be convenient. Go ahead and throw out the data, since that would allow you to maintain your previous belief. Sorry to break it to you, though, but it is universally accepted that increasing cache size yields far better hit rates than increasing the associativity. Additionally, increasing associativity past 8-ways yields diminishing returns, and often adds overhead, since lookup time is increased.
FWIW, I apologize for calling you stupid, but I think you ought to keep an open mind to the design decisions of engineers with far more experience than you, rather than criticizing them based on your limited information.
wbmw
P.S. See my post to Tench about the 8192-ways of associativity. I should have included cache line size in the explanation - especially since I should have known my explanation would be scrutinized. |
