[ExI] The tyranny of context free grammars

Tomasz Rola rtomek at ceti.pl
Wed Dec 15 18:04:53 UTC 2010


On Tue, 7 Dec 2010, Alan Grimes wrote:

> Ryan Rawson wrote:
> > was merely pointing out the downfall of a context-ful grammar, since
> > the original poster makes it seem like "if only we shed the shackles
> > of CFGs we'd be better for it".  There is a dark side of that though,
> > and perl is one of them... I've written a few thousand lines of perl,
> > it's great for hacking things out, but trying to re-discover what you
> > coded a few weeks/months/years ago can be... difficult.
> 
> Not quite.

Let's say there is strong human factor in this - some people so much 
dislike Perl that they have to relearn it every time they have a caprice 
to "try it once more" (I seem to fall into this category).

> I was pointing out that the paradigm of the chomsky heirarchy is
> limiting in that it has failed to provide us with the tools we need to
> design software for CAM architectures which would seem to enable, with
> today's technology, several orders of magnitude more performance per
> transistor than is possible with von-neuman machines because millions of
> bits could be computed per clock cycle instead of a few thousand...
> 
> Of course there could be problems with the idea of the cellular automata
> as being an optimal computing architecture. In any event, computronium
> is not just a fabrication issue, it's also a computer science issue.

While von Neumann architecture is not perfect, I think it was the easiest 
__usable__ thing to be built in the past and it will continue to be such 
in the quite a long future. Thanks to increasing performance, one can 
research, emulate, and maybe find, a better computing architecture. After 
emulation of the concept proves to be good, one can design 
"transistor-level" implementation (VHDL or whatever is suitable for ASIC 
synthesis) and submit it to manufacturer specialising in ASIC 
custom-building. Or one can push this low-level design into FPGA.

I'm not sure cellular automata is better alternative for v-N. I guess it 
is one of the easiest to fabricate, but programming it is, IMHO, tricky in 
best case. And I don't want to think about debugging it. I may yet learn 
the opposite after reading Wolfram's "New Kind of Science" but I don't 
expect this much, even though I expect quite a lot :-).

Anyway, I may change my thinking when I see Linux ported to CA-CPU :-). 
Before this happens, CAs are probably going to be confined to coprocessors 
and other such helper hardware. And to be frank, I would rather invest my 
time in something else, even though ATM I am not quite sure what exactly 
(the subject is broad and needs a lot of exploring).

Regards,
Tomasz Rola

--
** A C programmer asked whether computer had Buddha's nature.      **
** As the answer, master did "rm -rif" on the programmer's home    **
** directory. And then the C programmer became enlightened...      **
**                                                                 **
** Tomasz Rola          mailto:tomasz_rola at bigfoot.com             **



More information about the extropy-chat mailing list