Transcript On the Value Locality of Store Instructions

Let’s Build Brains
MICRO 2007 Panel
Computing Beyond Von Neumann
Mikko H. Lipasti
Associate Professor
Electrical and Computer Engineering
University of Wisconsin – Madison
http://www.ece.wisc.edu/~pharm
Brains (1)

Human brain  Von Neumann machine
Face recognition: <500ms
Neurons are slow:


Critical path is a handful of “gates”



Fundamentally different computational model
Made of shoddy, unreliable parts
“…neurons are noisy, unreliable devices, … the nervous system averages
over many cells to compensate for these shoddy components.”
-Christof Koch

We can build it. We have the technology.
Dec. 3, 2007
MICRO’-40 Panel: Computing Beyond Von Neumann
2
Brains (2)




Human neocortex:
 ~20B neurons, ~200T synapses
 Physiologically homogenous
 Hypothesis: runs common algorithm
Apply architecture 101?
 Abstraction layers
Let’s
 Hierarchy and replication
Build
Scientific understanding incomplete
 Several plausible hypotheses
Brains!
 Difficult to study brains in situ
Building leads to understanding [Anant’s keynote]
 Synthetic artifacts much easier to study
Dec. 3, 2007
MICRO’-40 Panel: Computing Beyond Von Neumann
3