Evaluating Spatial
Computation For Whole Program Execution using the Tartan Architecture
Tuesday March 21, 2006
Hamerschlag Hall D-210
4:30 pm
Mahim
Mishra
Carnegie Mellon University
Spatial Computing (SC) has been shown to be an
energy-efficient scaling-tolerant model for implementing kernels. In
this talk I explore the feasibility of building a general-purpose
processing system based on the SC model: I describe Tartan, an
architecture which integrates a general-purpose processor with a
reconfigurable fabric (RF), and evaluate its performance and energy
efficiency when executing entire (large) applications. I also present
Tartan's compilation toolflow, which can automatically partition an
application and generate a processor instruction stream and an RF
configuration. The results indicate that a hierarchical RF architecture,
designed around a scalable interconnect, is instrumental in harnessing
the benefits of spatial computation in such a hybrid system. The
interconnect uses static configuration and routing at the lower levels,
and a packet-switched, dynamically-routed network at the top level.
Tartan is most energy-efficient when almost all of the application is
mapped to the RF, indicating the need for the RF to support most
general-purpose programming constructs. Initial investigations reveal
that such a system can provide, on average, an order of magnitude
increase in energy efficiency compared to an aggressive superscalar core
on single-threaded workloads, while delivering acceptable performance.
Mahim is a PhD candidate in the Computer Science Department at CMU,
where he is part of the Phoenix project led by Prof. Seth Goldstein. His
current research focuses on the architecture of the Tartan
reconfigurable fabric, with particular emphasis on the design of the
compute units and memory hierarchy, and on the fabric's defect tolerance
aspects. His other research interests include processor architecture,
high-level synthesis, CAD algorithms, VLSI test methods, and simulation
and modeling. When not chained to his desk in the interior of Wean Hall,
Mahim likes to play tennis, listen to classic rock and drink beer.
Before joining CMU, Mahim obtained a BS in Computer Science and
Engineering from the Indian Institute of Technology, Kanpur.
|