On Apr 25, 2009, at 6:02 PM, Marcus G. Daniels wrote:
Owen Densmore wrote:
Most of computing does not need to be exact .. a slight "error"
generally is not terrible and for imaging, audio, and so on simply
is not observable by a human.
And if what you need is a *lot* of random numbers [1], why do dozens
of cycles of exact arithmetic and memory lookups to make pseudo
random numbers, if you could instead just read a vector of physical
noise values from a CPU I/O port in a single cycle?
[1] http://en.wikipedia.org/wiki/Monte_Carlo_method
Hmm.. maybe it'll be a bit like graphics: initially graphics was all
done on the CPU. Then it became important enough to have its own co-
processor, a GPU. Maybe a RPU is next?
-- Owen
============================================================
FRIAM Applied Complexity Group listserv
Meets Fridays 9a-11:30 at cafe at St. John's College
lectures, archives, unsubscribe, maps at http://www.friam.org