Re: How many different processors do you use?

[muddy_buddy@my-deja.com]

In article <7jk7hk$36s$1@nnrp1.deja.com>,
  Robert Dewar <robert_dewar@my-deja.com> wrote:
> In article <7jjij7$qci$1@nnrp1.deja.com>,
>   muddy_buddy@my-deja.com wrote:
> > 1. The Ada comittee went and created their own syntak, and
> > increased the learning curve.
>
> Committee? what committee?
I read someones' website and your right, but that was the
belief.  There certainly was an approval commitee that took
too long.

>
> > 2. The Reagen adminstration was so anti-gov that they didn't
> > fund a quality free or at least cheap Ada development system
> > for education, and small companies.
>
> This claim is not even vagely related to reality.

It was a major factor. Look at Sun efforts to get Java accepted.
They created a fair development system and gave it away.  This allowed
many schools, companies and others to quickly release up grades.  The
Reagen Admin. decide that the free market would produce tools, and
they did according to other posts, just 10 years too late.  If the
government had funded the NYU or some other effort and produced a
easy to use and hack complier and debugger and gave it away.  Ada
would have replaced the nearly worthless PASCAL language as a teaching
tool. (PASCAL is a good teaching lang., but didn't support I/O etc.)
That very idea was suggested to the Government's Ada Backers(by me
at least) and they quote Bonzo's free market concepts as why their
weren't allowed to do that.  I was on a program that considered funding
an Ada Complier and debugger for the 320C25 and it only cost a million
or so.  That would be chicken feed compare to what the Ada effort
failure has costed.


>
> > An Ada development system could cost upto 10,000 $ for
> > a bad product.
>
> Perhaps, but certainly there were excellent compiler costing
> much less than this figure. In my experience, cost was a small
> factor in most situations.
It certainly wasn't a small factor for us, and it prevented Ada from
gaining any widespread use as a lab tool development language, and
prevented electrical engineers from having access.  10k is a whole
lot of money for a University, and out of sight for a hacker.  BTW
in 1987, what good set of Ada tools for a DSP were there that costs
much less than $10,000.



>
> > 3. Ada's requirements on compiliers made them more expensive,
> > and very hard to do on DSP, and small embedded Processors.
>
> Dubious claim. Certainly most of the cost of an Ada compiler is
> entirely processor independent.
>
> > This meant that Chip manufactures were reluctant to fund
> > development.
>
> Actually many manufacturers DID fund development in Ada 83 days

For larger chips perhaps, but embedded processors and Small DSP were
not supported.  As for it being processor independent, I would agree
it should be, but every time we tired to pay for it the complier
companies wanted a whole lot of money.


> >
> > 4. The high costs of using Ada prevented widespread use
> > outside the defense industry.
>
> This is a claim without any data to back it up
That is not true, I provided several of examples of why Ada costs
prevented its use anywhere other than where it was required.


>
> > 5. As said before, the government blind insistance that Ada
> > was the choice for everything, embittered the defense
> > companies and their Engineers.
>
> I don't think so, some contractors and engineers were annoyed,
> others ignored the mandate in any case, others embraced it and
> became enthusiastic supporters.

At least NVL viewed it as an all or nothing deal.  We had a
signal processing intensive application that we wanted to do on
2 320c25 DSP's which had no Ada support.  At the time (1985-88) there
was not an Ada supporting processor that could do the Signal processing
at a reasonable rate.  The signal processing was pretty much a
fixed process once designed, and all the SW that was likely to
change was on one processor.  We offered to use a 680x0 processor with
Ada for the second processor, but that wasn't a pure solution.

In addition, the government raised our SW costs by 2x by insisting that
it either be Ada or assembly and thus no C.  Since the 320c25 or the
Motorala 5600 DSP had C compliers, but no Ada we wrote everything in
Assembly instead of just the inner loops.




>
> It sounds like you were not around at the time, perhaps that is
> unfair, but your tendency to paint things black and white, when
> in fact they were not nearly so clear cut, suggests it.

I didn't think I painted anything as black and white, I think you
should read my post again.  I have worked in either Processor design,
or software in the defense industry for 15 years. I was there, and I
believe I paint a more Ada Friendly picture than most.


>
> > 6. All of the above has created a situlation, where staffing
> > Ada programmers is very, very difficult.
>
> No, not so difficult.
If you can find Ada programmers so easily please contact Raytheon
personnel or call be.  I can get 1500 a piece finder's fee.  The big
staffing problem we have is that we have to fight for SW staff with
the 12's of communications companies in the DFW area.  They use C or
their own languages.  That means it is much, much easier to find C
programmers, and Engineers want to know C to keep their value up.
Ada just makes things harder.


> Let me get this right. On the one hand it is impossible to find
> Ada programmers, on the other hand knowing Ada is not a valuable
> job skill. SOmething doesn't compute here :-)
>
As I stated above, the engineers are smart enough to realize that
they better have mobile skills.  In this market, any SW person can
get a job, but it last for ever.


> Actually it is very rarely the case that defense programs use
> state of the art processors. For one thing, it is often the
> case that special or hardened versions of processors are
> required.
Say what?  I used a prototypes C25, c30 and c40 and others used
proto-C80.  We are using the Power PC G4 processor in this design.
These may not be the ultimate Bleeding edge, but they are not
80386's either.  By the time we get to production, are products are
dated, but we have to start at near the leading edge or face supply
problems.



>
> In practice, multiple Ada compilers have been available very
> rapidly for nearly every processor being seriously considered
> for defence (and commercial) applications.
>
Our tools are dated, and I haven't used Ada 95.  BTW you have
Ada compliers for ARM, PIC and other Micro-controllers.  I will admit
there is a complier for the Power PC.


> > BTW I know Ada and it is better than C, though the tools
> > aren't too hot.
>
>
> But the important thing is to remember that just because your
> company moved from Ada to C++ does not mean that it was a
> sensible decision, and more importantly does not mean that all
> other companies are making the same mistake (if indeed it was
> a mistake in your case, as you imply).

I not sure it is not a mistake myself, but the reasons seem to be
good.  Another problem is defense contractors don't pay as much
money, and thus it is very hard for them to add additional requirements
on recruiting.  The major fears are fear of future support of new
processors and staffing.

BTW several people have said Ada 95 is like Java.  Does it not allow
pointers, and does it have automatic garbage collection?  Java is
dog slow, but it apparently has yielded a large productivy increase
for some companies.

Muddy

Muddy


>
> Sent via Deja.com http://www.deja.com/
> Share what you know. Learn what you don't.
>


Sent via Deja.com http://www.deja.com/
Share what you know. Learn what you don't.