The General Approach to Computing - A Ramble
mcguire at neurotica.com
Fri Apr 24 07:36:23 CDT 2009
On Apr 24, 2009, at 3:59 AM, Teo Zenios wrote:
>> Neither do I. On the other hand, I tend to think hardware
>> progress has been TOO fast, and that software, because of the
>> speed of the hardware, is immensely sloppy and inefficient, to a
>> degree I find amazing. Some years back, I was looking at a setup
>> for a Microsoft programming environment. It came on five CDs,
>> IIRC, and I couldn't help thinking of Turbo Pascal for CP/M, wich
>> came on a single floppy, and had an editor, compiler, and debugger
>> in about 56 K of programming space. The efficiency factor
>> difference is stunning. I believe that a bit more time at each
>> 'watermark' of hardware progress would result in tightening of
>> software efficiency to improve product performance, rather than
>> just waiting for the next generation of PC hardware to hide one's
>> poorly-written code. Again, that might just be me...
> Go get an embedded computer or program your phone of you like
> simple hardware and simple software.
Pardon me for jumping in, but...it's not a matter of "simple",
it's a matter of efficiency and good design. Warren hits the nail on
the head above. This is one of the biggest reasons why I don't, and
probably won't ever, delve into the Windows world. Most of the less-
clueful programmers work in that world (please don't flame, that's
very different from saying that all of them are less clueful) and
they tend to be extremely wasteful of resources.
I work in the embedded systems world, and now (again, thankfully)
in the world of scientific HPC, where every cycle counts. Again,
it's not a matter of simplicity of software (you should see some of
this stuff!)...it's a matter of taking performance seriously, not
being wasteful, and knowing what you're doing.
> I think hardware progress has stalled for a long time, instead of
> new architectures or new CPU designs we generally just have die
> shrinks, multiple cores, and new material to make the same old
> thing a bit faster (mostly spinning its wheels or bogging down in
> GUI hell).
Too true. I like to think of that as an architectural convergence
of sorts. We've learned a LOT about processor architecture since the
early 1970s, and many of today's architectures have become more
similar as a result. Ideas that seemed good at the time but failed
in practice have fallen by the wayside, the importance of register
count is now understood, etc etc.
Of course now we've got lots of people using one of the 1970s
throwback architectures with many of its disadvantages and bad
decisions still intact, but that's another matter.
> When intel tried a new way with the Itanium it was stillborn (has
> Intel designed anything besides the original x86 that did well?).
Yes, the most popular processor architecture of all time, in
number of chips shipped: The venerable 8051. The i960 did very well,
Itanium died without so much as a whimper, but Itanium2 seems to
be picking up steam. I gotta say, though...I have a high-end quad-
processor Itanium2 machine here, and it's not very fast. My central
computer at home, a Sun V480, glorious but two CPU-generations old,
even under a heavy load with lots of virtualized systems running on
it, is considerably faster at everything I've thrown at it.
Port Charlotte, FL
More information about the cctalk