Re: A question for my personal knowledge.

[<bglbv@my-dejanews.com>]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain, Size: 2506 bytes --]

kilgallen@eisner.decus.org (Larry Kilgallen) writes:

> I have a subscription to the VMS source listings, and Charlie is
> correct, modulo some VAX-Alpha skew.  I am talking about actual
> use for code that ships in the operating system product, not whatever
> DEC may use internally for testing the compiler.

Well, source listings are of course the kind of authoritative
references I was missing. Thank you. So you are saying that there
is no trace of Ada code in VMS sources prior to release 6.2? For
completeness' sake, how far back did you look?

> > I suppose someone with access to a VMS system could go through system
> > executables and shared libraries and watch for the Ada compiler's
> > idiosyncracies in code generation. (Things like working out the size
> > of everything in bits rather than bytes, for example. Certainly if I
> > wanted to search for GNAT-generated code on an x86 I would look for
> > sections with a high density of arithmetic shifts left by 5. Similar
> > heuristics should work with other compilers and on other platforms.)
> 
> I presume that any reasonable Ada compiler would handle that before
> emitting object code, and it would never show up.

Why not? It does show up in GNAT-generated code in at least some
circumstances, and as long as the impact on performance is negligible,
who cares? It used to be possible to tell apart Fortran and Pascal
code on a VAX by noticing that Fortran used CALLG and Pascal CALLS
instructions. It would be na�ve to expect compilers for different
languages to generate totally indistinguishable object codes (as
opposed to equally efficient ones) for equivalent source programs.

> >> Not major parts mind you.  Some sub-sections where the programmers involved
> >> wanted to use Ada.
> > 
> > My understanding is that the use of Ada resulted in measurable
> > productivity gains. Pity I don't have the original reference anymore.
> 
> I doubt that DEC has done enough Ada to make such measurements.
> It takes a _lot_ of code to filter out "other factors".  See
> Marin David Condic's cautious optimism at his company after 10
> years of comparisons.

Oh, sure. As I said, I was citing from (probably not wholly accurate)
memory a source I can no longer even identify, let alone check for
trustworthiness. I don't recall that source going into any detail on
the method of measurement; the statistic may well have been meaningless.
This is at the anecdotal ("a good story doesn't need to be true")
level, no doubt about that.