
CHAPTER 1 Introduction
(IEEE 1993). The Alewife project at MIT sought to minimize the hardware support
for shared memory (Agarwal et al. 1995), which was pushed further by researchers
at the University of Wisconsin (Wood et al. 1993). The Kendall Square Research
KSR1 (Frank, Burkhardt, and Rothnie 1993; Saavedra, Gains, and Carlton 1993)
went even further and allowed the home location of data in memory to migrate.
Alternatively, the Denelcor HEP attempted to hide the cost of remote memory
latency by interleaving many independent threads on each processor.
The 1990s have exhibited the beginnings of a dramatic convergence among these
various