Quality & cost of Ada compilers

[larry@VLSI.JPL.NASA.GOV]

--
Mr.  Neff makes some good points, though he seems to be a bit too
pessimistic.  For instance, compiler errors we'll have with us for decades
to come.  I still occasionally see problems with mature C and ForTran
compilers from respected companies like IBM and DEC.  The validation
process weeds out many such errors.

Efficiency of compilers obviously need improving, but already there are
several out there that are pretty good.  Some of them may produce better
code than compilers for more languages  (There's more information available
in an Ada program, so this is potentially the case.)  Tartan makes some
extravagant assertions about their compilers in comparison with (for
instance) Sun C.  (Has anyone on this list looked into their claims?)

As for cheap compilers, you can now get validated Ada compilers from
Meridian and R&R Software for about a $100.  These are "Starter" or
"Introductory" versions that have been crippled somewhat (10 compilation
units and 200 executable statements per unit for Meridian), but still
adequate for beginners.  Then if you want the full version you lay out
several hundred dollars more.  You still pay 3-4 times the cost of
equivalent C development system, optimization is limited, and the generic,
tasking, and some other facilities are primitive, so obviously the
situation can be improved more. 

Possibly the entry of Borland into the market will fuel some major
improvements.  (Or maybe not, if their Ada is as non-standard as their
ProLog system.)  Incidentally, the report I heard is that Borland has
copyrighed TurboADA, not TurboAda.  Anyone know if that's true?

(Makes me wonder if by spelling a product ADA some turkey could get around
AJPO's copyright!)
                                    Larry @ jpl-vlsi