Re: Would green arrays produce something with a web browser? like a cheap appliance?

From	Paul Rubin <no.email@nospam.invalid>
Newsgroups	comp.lang.forth
Subject	Re: Would green arrays produce something with a web browser? like a cheap appliance?
References	(15 earlier) <jh2ub4$npj$3@dont-email.me> <jh3gi3$4g4$1@online.de> <95a854dd-69c4-4cf9-8966-eadbaa57733c@l14g2000vbe.googlegroups.com> <7xvcn6tpub.fsf@ruckus.brouhaha.com> <31f4725f-8e32-4333-a2a7-2a06bf9a962b@ge5g2000vbb.googlegroups.com>
Date	2012-02-20 19:47 -0800
Message-ID	<7x4nukoniv.fsf@ruckus.brouhaha.com> (permalink)
Organization	Nightsong/Fort GNOX

Show all headers | View raw

rickman <gnuarm@gmail.com> writes:
> I'm not clear on what you have trouble believing?  I am sure there are
> any number of things that are hard to impossible to do on a GA144, but
> there are lots of SDRs that can be implemented on a GA144 with an
> appropriate IF input.  No, I don't think the ADCs will handle a 70 MHz
> IF with 20 MHz of bandwidth, but that is not the spec of every SDR.

OK, you know more about SDR than I do, and I can think of some legacy
analog modulation schemes that a GA144 could possibly handle.  With
digital modulation it may be more of a challenge, even in narrow bands.

> For the DSP the market (cell phones) didn't exist initially if I
> remember.

I thought DSP's were developed for use in the landline phone system long
before there were cell phones.  That market, and the market for signal
processing filters in general, formerly had to be served by boxes full
of expensive analog circuitry.

> That is largely because you aren't familiar with the GA144 I expect.
> I don't see any shortage of markets for it.  What markets is the GA144
> excluded from that the DSPs and MCUs address?  

> The only limitation that I see is that the entry level price is over
> $20 currently because you need to add a Flash and a RAM.  

The package is also pretty big, and the processor is slow unless you
have a way to put the parallelism to good use.  The 700 mhz base clock
isn't that helpful when you consider how many cycles it takes to get the
data from another node or from external memory, to where you can use it.
The 18 bit wordsize is also pretty awkward for "computer" applications
(it's probably less awkward for signal processing).

> there are lots of variations that can be done in the future.  The
> first DSP and MCU were real pieces of junk compared to what we have
> today.

I agree with you that future GA chips (if they are made) may be be much
more directly useable than the GA144.  But I think even the earliest
MCU's and DSP's enabled significant cost reductions over what could be
done without them.  It was before my time but I think the first MCU
application was a desk calculator using the Intel 4004.  The MCU
replaced dozens or 100's of chips used in earlier calculators.  The DSP
replaced boards full of complex and fragile analog filter circuits.
What does the GA144 replace?  

> You keep talking about what you can imagine.  I'm not totally clear on
> what you don't like about varaious things, but it is very clear to me
> you keep seeing the GA144 the same way you see a DSP or an MCU.  Until
> you see it as it is, I don't think you will find much use for it.  I
> would love to read what Jeff would have to say about it.  I have
> always learned a lot from reading Jeff's posts here.

I miss Jeff too, I was just reading his site last night in fact.  But I
remember discussing the GA144 with him and still not being that
persuaded.  One thing he claimed was that the Seaforth-based hearing aid
was equivalent to a DSP board that used 100 watts, requiring a
prospective user to carry a car battery around.  After further
discussion it turned out that the hearing aid had been prototyped with a
high-end DSP that used 1 watt, not 100.  The Seaforth chip used quite a
lot less than 1 watt, but it still wasn't clear to me that the functions
couldn't be implemented on a low powered DSP with comparable total
consumption than the 1 watt chip.

> I think the FPGA is the analog for the GA144 (or GA1024), much more so
> than MCUs or DSPs.  Designers often do DSP on FPGAs.  Does anyone
> worry that the FPGA isn't keeping LUTs busy, etc? 

Don't they?  I mean, there are small cheap FPGA's and big expensive
ones, and don't designers try to use the small cheap ones when they can
and the big expensive ones only if they must?

And certainly, if I had a giant FPGA (say one with a billion gates) I
would know right away what kinds of interesting new things I could do
with it (like implement a new CPU architecture, or a big
content-addressible memory, or something like DJ Bernstein's factoring
circuits) but it's harder for me to see what to do with a GA1024.  If
ideas for it are so easy to come up with, perhaps you could post a few.
Most applications I've thought of for a GA processor can either be done
in 10-20 nodes, or else can't be done competitively with other
approaches at all.  But I'm not a hardware guy so I agree with you that
maybe my vision is limited in this area.

Back to comp.lang.forth | Previous | Next — Previous in thread | Next in thread | Find similar

Thread

csiph-web