🚀 go-pugleaf

In article <3vhi5i$gut@newsbf02.news.aol.com>, donrawlins@aol.com (DonRawlins) writes:
|> dkomo writes..
|> 
|> >> As a hardware engineer I guess I take a little umbrage at these
|> remarks. 
|> >> I've always seen my design efforts as benign -- providing vast armies
|> of 
|> >> otherwise incompetent software "engineers" with employment. 
|> 
|> Remember that the speed increases in hardware are due primarily to the
|> fact that improved materials and fabrication processes allow those
|> "engineers" to pack progressively more logic into the same sized chip. 
|> Hence, what were once simple and elegant chip designs have now mushroomed
|> into monstrous multi-million transistor behemoths that run hot and, as we
|> saw with the Pentium, often contain bugs as bad or worse than production
|> quality software; despite the fact that these "hardware engineers" are
|> simply creating a device that can run a very simple instruction set -
|> thousands of times more simplistic than even the most basic of today's
|> software applications - faster than its predecessor.  Why not optimize
|> existing designs instead of adding millions of transistors of additional
|> logic?
Do you think that those millions of transistors are not an optimization?
Granted, there are other ways of optimizing hardware, for power consumption
or area being two of the other popular optimizations, but the one most 
people care about is speed.  These millions of transistors are used to
support such architectural improvements as cache, pipelining, branch 
prediction, out of order excution, and so on.  

|> And for a laugh take a look sometime at software written by these
|> "scientific" hardware experts.  The ability to deal with abstractions so
|> mandatory in software engineering is totally lacking in the hardware
|> community.  Rather, those folks are capable of operating only at very low
|> and simplistic levels where everything is extremely well defined and can
|> be easily simulated - again, implementing simple instructions that are
|> thousands of times less complicated than even the most primitive of
|> today's software.  Interaction with end users demanding progressively more
Have you ever seen the VAX instruction set?  Many lines of C code can be
directly translated into VAX instructions.  The reason DEC dumped it and are
now pushing the ALPHA line is because the ALPHA instuction set is easier to
make faster.  Instructions that do simpler things decode more quickly,
are easier to pipeline, don't require microcode, and are much easier to
complete out of order.  This really only affects the compiler anyway, so I don't
understand why this is a reason to whine.

|> complex applications in less and less time is also not a factor for
|> hardware people - nor are constantly changing product requirements.  The
|> importance of creating an interface that humans & not other machines can
|> use and appreciate is totally foreign to them.  After all, that can't be
|> subjected to a trivial scientific analysis through which one may assess
|> its "correctness".  Rather, heaven forbid, the uncommunicative 100% left
|> brained hardware engineer would actually need to deal face to face with
|> (gulp) humans.
This seems to ignore the fact that hardware designers are hardware designers,
not programmers.  I don't know any programmers who could design the simplest
digital circuit, let alone one as complicated as a cpu.  Hardware designers
are trained to be hardware designers, and programmers are trained to program.
 
|> IMHO software development is a high IQ activity that requires an equal mix
|> of both creative and analytic skills.  Hardware development is a mid-level
|> IQ activity that requires, almost exclusively, analytic skills.  Hence,
I see that you are an expert on neurophysics in addition to being an expert on
software and hardware.  :)

|> while hardware developers may whine about the state of the art in software
|> engineering, they certainly have nothing better to offer.  If they do, why
|> not try hard wiring the equivalent of Microsoft Excel into your next chip?
|>  Not so simple, is it?
Suppose somebody did design a processor to run Microsoft Excel.  Suppose it
ran Excel 10 times faster than any other processor.  More than likely, it would
run most other applications 10 times slower.  Do you find this a good tradeoff?
Do you want a different machine for every piece of software you run?   

|> Let me remind you also that without complex software that takes advantage
|> of faster processors all you hardware jocks would be looking for work
|> elsewhere :)
This is true.  Software and hardware work together.  Without hardware, there would
be no software and vice versa.  I however don't think there is any room for cheap
shots at either group of designers.   Before you make any more inane comments about
hardware, learn something about it.
-- 
----------------------------------------
Mark Vilas   mvilas@mipos2.intel.com
disclaimer:  "my opinions are my own and not those of Intel."