Re: [microblaze-uclinux] OT?, Performance

[Matthew Rubenstein <email@xxxxxxxxxxxx>]

> What is hardware?  What is software?  Are they really distinguishable? 
> Does it matter?

	While FPGA development might drop some traditional limits of HW & SW,
allowing some techniques from either discipline to be applied to our
platforms, HW & SW are still easily distinguishable. HW is made of
atoms, while SW is made of bits. Even when HW changes to include
electrons themselves as the mechanical material (spintronics), or
possibly even photons (quantum entanglement), they're still made of
quarks/gluons: matter, with gravitational/inertial characteristics, such
as "identity": one thing in one place at one time. Even in the most
"continuum" ontogenal scheme, bits of information are much subtler
stuff, if even stuff at all. They are not bound by matter/energy
characteristics like identity, and the consequent conservation
constraints. HW is matter (and even energy), while SW is information
about HW.

	Explicating these differences is fun, but it also matters. As property,
HW is fairly well understood: if I give you some of mine, you have more,
and I have less. SW doesn't play by those rules, and can even invert
them: the "network effect" means information is more valuable when
corroborated, interconnected with identical or similar information.
Patents, although controversial at many boundary conditions (including
SW that simulates HW, on eg. Turing "universal machines" like a Pentium
or PPC), are a consensual device for the economics of invention of HW.
Copyrights are even more experimental, protecting economics of SW.
Protecting our investments in HW and SW require viewing our inventions
as one or the other, choosing our protections as most appropriate.

	This discussion is itself a boundary condition for lists like this one.
Working with MicroBlaze, a simulated processor, on proprietary, highly
reconfigurable FPGAs, running a version of Linux, is a hothouse for
intellectual property issues. We're inventing these distinctions, based
on our work with the differences. And our discussions contribute
significantly to the development of the new paths of invention,
economics and laws that lead from this watershed period when SW and HW
are so flexible that one can be mistaken for another. Take care how we
decide these issues, because they'll snowball as they descend from our
labs to the market - and we'll all be stuck with the consequences.


> On Wed, 2004-05-12 at 21:03, John Williams wrote:
> Yasushi SHOJI wrote:
> > Hi all,
> > 
> > # This must be OT for mb-uclinux ml.
> 
> Maybe not, the very fact that we are using linux on an FPGA makes these 
> philosophical questions worth asking (and possible to be answered!)
> 
> > say, if we have perfect C to RTL or HDL compiler, would it be
> > impractical to build not only web server but entire uclinux kernel and
> > userland purely from logic?
> > 
> > I have been wondering a few days.  what will stop us?
> 
> Now I *will* get off topic :)
> 
> I'm not convinced that we will ever be able to run the source for, say, 
> the boa webserver, through a translator, and get some detailed low level 
> RTL circuit to synthesise.  What does malloc() mean in hardware?
> 
> Well maybe that's not true, if you give me the boa source, I'll generate 
> a hardware description...  it's called a microprocessor operating on 
> stored opcodes, with a memory subsystem, and ...  :-)
> 
> What is hardware?  What is software?  Are they really distinguishable? 
> Does it matter?
> 
> John
> ___________________________
> microblaze-uclinux mailing list
> microblaze-uclinux@itee.uq.edu.au
> Project Home Page : http://www.itee.uq.edu.au/~jwilliams/mblaze-uclinux
> Mailing List Archive : http://www.itee.uq.edu.au/~listarch/microblaze-uclinux/
-- 


(C) Matthew Rubenstein

___________________________
microblaze-uclinux mailing list
microblaze-uclinux@itee.uq.edu.au
Project Home Page : http://www.itee.uq.edu.au/~jwilliams/mblaze-uclinux
Mailing List Archive : http://www.itee.uq.edu.au/~listarch/microblaze-uclinux/