home .. forth .. misc mail list archive ..

Re: PSC1000 -- Holy Doodle, at last!


vic plichota writes:

 > Please re-read my post, that info is in there.

The X fonts I have make text behind > difficult to read under XEmacs
(perhaps you should bounce posts the next time or paste text). 
Meanwhile I've found the info on the webpage, though.

 > It depends upon your definition;  certainly it has everything I need
 > to construct a serially-linked multiprocessor system that will
 > perform well.  Also remember that this is merely first-silicon:

Well, I'm talking about several (4-8) >>100 MBps serial links and a
switchable low-latency (few 10 ns) protocol. After the 8-node K7 I'm
planning to build in August, this stuff here
http://products.analog.com/products_html/list_gen_98_2_1.html 
looks quite interesting. Their EZ Lite eval kits are cheap, too.

 > PatSci is planning more sophisticated variants for the future.
 
Good to hear, and I hope these won't be vapourware.

 > My main point was the excellent -- in fact exemplary -- quality of
 > the doc'n.  TI has repeatedly set the benchmark for this, and PatSci
 > has at least equalled it.  None of this "Oh, ask Dr. Ting, and fuck
 > with the supply-voltage" horseshit...  very fine for discussion
 > amongst hobbyists, but very bad for building products.
 > 
 > I am NOT trying to ruin any other efforts (far from it!), but I must
 > acknowledge a "class act" when I see one -- puts me to shame, almost.

You make a very good point here. I was met with hoots of derision when 
I tried peddling the F21 using the web docs to developer friends of
mine. Having first class documentation cannot be valued high
enough (with the possible exception of shipping working silicon).