Re: Current state of Ada 9X compilers...?

[agate!doc.ic.ac.uk!pipex!sunic!news.funet.fi!klaava!klaava!not-for-mail@u cbvax.Berkeley.EDU (Patric M Stickler)]

In article <27f4gs$m9k@cmcl2.NYU.EDU> kenner@lab.ultra.nyu.edu (Richard Kenner)
 writes:
>In article <27eqku$o5t@klaava.Helsinki.FI> stickler@klaava.Helsinki.FI (Patric
 M Stickler) writes:
>>What about GNAT. Is it just an academic exercise, or is it expected
>>to be an "industrial strength" compiler for "real" projects. I'm in
>>general very impressed with the quality and portability of GNU
>>applications, and for that matter sympathetic to the FSF cause, but
>>could someone in-the-know tell me what the actual involvement of the
>>FSF is in the GNAT project. Does NYU simply have permission to base
>>GNAT on gcc, or does the FSF actually plan to promote the use of Ada 9X.
>>Will GNAT become like GNU Smalltalk, which although OK for learning,
>>is too clumsy for real development? (IMHO, no flames please)
>
>Let me address a number of your questions.  First of all, GNAT is
>certainly not an "academic exercise".  I don't know what an
>"industrial strength" compiler would necessarily mean, but our goal is
>to make it at least as reliable as GNU C.  Even in its present very
>preliminary state, it is being used for a significant "real" project:
>itself!  Most of the GNAT developers no longer use any Ada compiler
>other than GNAT.

This is what I had hoped (and half expected) to hear, but it's nice to
hear it in any case.

>
>Although we have not yet done any significant Ada-specific
>optimizations, we are quite pleased both with our compilation speed
>and performance of the generated code.  Both are at least 3 times
>better than that of the commercial compiler we previously used, which
>is why we switched to using GNAT internally.  If you write an Ada
>subprogram at the same semantic level as a C function, you should get
>the identical machine code for both on most GCC targets.  Once we
>finish, you should even be getting better code for the Ada subprogram
>in some cases, since we can do a better job determining which pointers
>cannot alias to each other and other variables than can be done for C.

I find this *very* encouraging. Although I'd be willing to take a slight
performance hit for the other benefits of the language, being able to claim
equal performance and even potential increases in performance over C and C++
helps to justify the choice of Ada 9X for our projects.

>
>We don't need "permission" to base GNAT on GCC, since the whole
>purpose of the GNU Public License is precisely to encourage such
>noncommercial uses of GCC.  The FSF does not itself promote the use of
>any one of its tools, so it will not promote Ada.

Sorry, I should've thought this part of the question through better. Of
course the FSF won't be playing favorites with any of it's tools. I was more
concerned with robustness and completeness, and worried that GNAT would be
like GNU Smalltalk (i.e.  rather clumsy and in many ways incomplete and "not
yet ready for prime time" and not showing any signs of improvement). 

>
>There is significant cooperation between the FSF and the GNAT project.
>We have had a number of meetings with Stallman (RMS).  I am the member
>of the GNAT project that is most responsible for the interface between
>the front end and GCC and the person who is responsible for making the
>necessary enhancements to GCC.  I am also one of the primary GCC
>maintainers for the FSF and will shortly have primary responsibility
>for the maintainance of GCC as a whole.

Boy is *this* encouraging, the primary maintainer of GCC with a positive
attitude towards Ada 9X. Of course, we wouldn't want Ada 9X to get any extra
favors over the rest of the "children" now, would we ;-)

On a side note, are there plans to produce an OOP library for Ada 9X similar
to (or perhaps even ported from) libg++? Or is one in the works or already
finished?

Regards,


/////////////////////////////////////////////////////////////////////////
  Patrick M. Stickler  OH2LUV, KC4YYY    The comments contained herein
  WSOY - Information Systems Division    do not necessarily reflect the
  Helsinki, FINLAND    (psti@wsoy.fi)    official views of my employer.
/////////////////////////////////////////////////////////////////////////