Gary, here's the paper I was telling you about a few days ago.

Everyone else: cool paper done by some guys at Microsoft research, the basic 
idea is that "everyone knows" languages like C/C++ where you manage the memory 
yourself are faster than languages where the memory is managed for you, but 
really if the memory is managed for you the runtime can do some cool tricks with 
rearranging where the memory's located while the program is running to improve 
caching and therefore get *enough of a speed boost to offset all the costs 
involved*. Very cool stuff, that'll hopefully start showing up in actual 
languages in the years to come.

-------- Original Message --------
Subject: Research on advanced caching in programming languages.
Date: Mon, 28 Apr 2008 12:02:34 -0400
From: Peter C. Chapin <[log in to unmask]>
To: Andrew Guertin <[log in to unmask]>

You asked about research done at Microsoft on moving objects around in a
running program in order to optimize the use of the cache. The attached
paper was given at PLDI'06 and pertains, I think, to what you were
talking about. The full citation is below. This is probably not the last
word on this work. There may be more recent papers on it by now.

Peter

@article{1134011,
   author = {Trishul M. Chilimbi and Ran Shaham},
   title = {Cache-conscious coallocation of hot data streams},
   journal = {SIGPLAN Not.},
   volume = {41},
   number = {6},
   year = {2006},
   issn = {0362-1340},
   pages = {252--262},
   doi = {http://doi.acm.org/10.1145/1133255.1134011},
   publisher = {ACM},
   address = {New York, NY, USA},
}