[extropy-chat] NEWS: optical processors

Eugen Leitl eugen at leitl.org
Wed Oct 29 18:28:58 UTC 2003


It's an analog device, and its parallelism is limited whatever you can
encode within a beam crossection/wavefront. Such conventional optical signal
processing is fundamentally limited -- this specialized device may do okay in
comparison to DSPs and ASICs (I'm missing the comparison with FPGAs), but
will look increasingly pale and paler with each subsequent iteration of
Moore's law (nevermind transition to reconfigurable logic in molecular
circuitry).

NLO logic has a niche in photonically switched networks though, and it can be
scaled down with SNOM and photonic crystal waveguides. So it will stay firmly
entrenched in a few niches.

On Wed, Oct 29, 2003 at 07:12:19PM +0100, Alfio Puglisi wrote:
> 
> Optical processors are finally making some steps out of the labs and
> commercial applications are expected in a short time frame. An Israeli
> startup has a 8,000 Gflop/sec prototype:
> 
> http://www.eetimes.com/story/OEG20011008S0024
> 
> What's interesting to me is that, according to the article, optical
> processors are inherently high-level, i.e. instead of gaining speed with
> RISC-like operation, you implement high-level functions directly in
> hardware (FFTs, pattern match, etc). Kinda like Mathlab with 1-clock
> functions.
> On the other hand, I could make thousands of matrix mult. per second,
> but it could take one minute to compute 2+2 if addition must be emulated.
> So the first devices will be more like specialized coprocessors and DSPs.

-- Eugen* Leitl <a href="http://leitl.org">leitl</a>
______________________________________________________________
ICBM: 48.07078, 11.61144 http://www.leitl.org
8B29F6BE: 099D 78BA 2FD3 B014 B08A  7779 75B0 2443 8B29 F6BE
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 198 bytes
Desc: not available
URL: <http://lists.extropy.org/pipermail/extropy-chat/attachments/20031029/d3de2133/attachment.bin>


More information about the extropy-chat mailing list