Kevin said:
ChrisGibboGibson said:
Stefan Heinzmann wrote:
[snip]
Strange that you single out the inventiveness of the Wien bridge
oscillator with the bulb over the development of the transistor. Both
were developments that relied on previous work. That's not to
belittle their achievement, but Hewlett's thesis for example
contains a reference to a paper by Meacham that described the usage
of a lamp in a crystal oscillator. (I got this from Jim William's
book).
But if you are content with innovations of that scale, you could cite
some of Widlar's work, for example the bandgap reference.
Things I also find outstanding is the planar process, the DRAM, and
on a higher scale the invention of virtual memory.
DRAM yes, I hadn't thought of that one. Pure genius.
Virtual memory? An invention that would have no use whatsoever if PC
programmers learnt how to write software properly.
I disagree. There are many valid reasons why virtual memory is pretty
much indispensable for *good* code. For example, simply handling large
data files, e.g. Spice output files. These could be 100's MB. One simple
opens up the file, even if one doesn't have enough RAM. To write
specialised code to handle this sort of thing would be very messy
indeed.
The real world is always going to have more data than RAM available to
handle it, so its got nothing to do with poor software writing. Its just
the way it is.
Kevin Aylward
yup. The real problem with virtual memory is piss-poor implementations.
Mind you, I have seen a guy use a 32kb EEPROM and implement virtual
memory in a 8kB 8051 variant. Being E^2 based, the whole thing was
dreadfully vulnerable to power cuts, necessitating a wide array of
dubious strategies to try and prevent problems. This lead to a huge
project blowout, it took an extra year to get all of the code working.
Every time the code needed to be altered (bugfix, new algorithm
whatever) ALL the original code used to get attacked in an effort to
keep the most-often used code in flash..... greatly increasing the
likelihood of introducing new bugs (which they did, galore).
I was originally told it would add $5-$10 to the micro cost (almost
doubling it) to get more ROM, hence this dumb idea. Shaving a year off
the project would have been worth it (in all that time we never changed
the hardware
. Eventually it was found we could buy a 32k micro for
less than the cost of the original 2 parts. Turns out the twit
(ir-)responsible basically wanted to implement VM, and hadnt actually
got a volume quote - only a 1-off price, despite being told to do so.
And his cowardly manager did nothing - he should have been sacked (or at
least had formal disciplinary action, thereby setting the stage for a
dismissal the next time - we have quite good employment law in NZ). He
was the only person who understood the code though, and as such was
untouchable ("we cant take the risk of him leaving...")
Cheers
Terry
