Computer Architecture - HiPC keynote

  Attended Prof Arvind's keynote address on "Rethinking Computer Architecture 
  Research" at the International Conference on High Performance Computing' held
  in Bangalore (Dec19-22). The main focus of his talk was how to efficiently validate
  and realise any proposed architecture changes like cache coherency algorithms
  or reorder buffer implementations.

  In reasoning out why computer architecture research is in a bad shape and
  how most of the researchers are focussing just on incremental research like
  a performance gain of 2-4%, he took a shot at the computer architecture bible -
  Hennessy-Patterson's Computer Architecture: A Quantitative Approach  which,
  though being very successful at teaching computer architecture, has set the
  trend of Quantative analysis in research too. Most of the research papers these 
  days concentrate just on quantitative analysis and do not even mention
 implementation details or even feasibility. He then went out to describe his RTL implementation tool - bluespec
and how it simplifies preliminary verification and implementation modelling.

A few months back, I started studying caching / coherency techniques and faced similar problems.
Some of the algorithms were beasts for implementation and lot of papers never ever mentioned how their particular
scheme can be implemented. And you will be surprised by the number of papers published just on cache coherency.
Had a very frustrating time there :(

I am still looking for some good material on hardware implementation of caches (Got Any?)