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

From	Bernd Paysan <bernd.paysan@gmx.de>
Newsgroups	comp.lang.forth
Subject	Re: Would green arrays produce something with a web browser? like a cheap appliance?
Date	2012-02-21 02:27 +0100
Organization	1&1 Internet AG
Message-ID	<jhurtk$g97$1@online.de> (permalink)
References	(15 earlier) <jh2ub4$npj$3@dont-email.me> <jh3gi3$4g4$1@online.de> <95a854dd-69c4-4cf9-8966-eadbaa57733c@l14g2000vbe.googlegroups.com> <jhph0e$nia$1@online.de> <7a5b4ac4-aaad-44bc-ab0f-b1f4f442a367@db5g2000vbb.googlegroups.com>

Show all headers | View raw

rickman wrote:
> This discussion is starting to get personal and I don't want to
> argue.  I'll just say that your numbers for the multiply are off by a
> minimum of a factor of two.

Are they?  The GA144 core can do a 18x1 multiply step, and AFAIK each 
takes two cycles (full carry chain).  To make that a 36x36 
multiplication, you need 4 times 18x18, i.e. 144 cycles for the 
multiplication steps alone.  Add accumulation and a bit of overhead, and 
you end up in "the order of magnitude" of 200, which I stated.

> But even if they were on target, they
> have no value in this conversation because they apply to one design
> niche.  Are you really going to try to itemize all the designs that
> the GA144 isn't good for?

No, I'm trying to map a real-world design I know of to the GA144, which 
is "more general purpose", and apparently good at audio-style 
applications.

> All audio doesn't require 35 bit
> multiplies.  I'm pretty sure the GA144 was intended to do audio since
> one of the apps some of the predecessor chips were touted for was home
> theater.

I know.  That's what the Zetex DDFA is touted for, as well - it is a 
special-purpose chip, built for a particular product segment, and by 
design a niche product. Nobody would argue that it is somewhere 
unbalanced, because it is designed to a particular target.

The 35 bit internal data path of the DDFA seems to be "absurd", but we 
wanted a high quality audio processor and did calculations how FIR 
filters add noise to the signal (yes, they do), and that was the 
headroom we ended up to make this additional noise be lower than the 
actual noise of the system.

As I said: You *first* do the analysis of your requirements, and *then* 
you build your product (where of course part of the analysis phase is 
building a prototype).  Dismissing some particular requirement which was 
derived by sound engineering, just because the GA144 looks bad on it, 
well, that's fanboi-ism.

> Anyway, do you think there is much use to continuing this
> conversation?  I think we have plowed this furrow enough and will just
> continue to see the thing differently.

I try to discuss, I didn't come to a final conclusion about the GA144 
yet.  If you don't want to discuss, because you "see things 
differently", feel free to stop.  This discussion started personally (I 
said "this is a research chip, not well thought-through"), and I have no 
problems with similar personal replies like "even a genius has a bad 
day" or such, because, yes, I do have bad days, and I can be convinced.

-- 
Bernd Paysan
"If you want it done right, you have to do it yourself"
http://bernd-paysan.de/

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

Thread

csiph-web