SGI Developer Magic - Ada95

SGI Developer Magic - Ada95

[Viqar Abbasi]

Has anyone out there used the "SGI Developer Magic ProDev Ada95"
compiler?

The information provided on the SGI website
(http://www.sgi.com/Products/DevMagic/products/project_ada95.html)
seems to indicate that its the tool for the task at hand.

The task at hand is to incorporate Ada into large infrastructure
based in C and Fortran, on the SGI Irix platform.  I want to
call the Ada from C and Fortran, without any limitations.  
The Verdix compiler we have now doesn't allow "in out" parameters
in procedures.. just "out" parameters.  Also, we're able to link
and run basic programs in GNAT... but have been having other
problems with newer versions of GNAT.  The last stable version 
seems to be 3.09.  Therefore, we're a bit GNAT-wary. 

We don't need all the graphical-debugger and fancy stuff like 
that provided by SGI.  We just want a stable, robust Ada compiler,
which produces object files we can link in with our existing software. 
Is SGI Ada95 the answer? 

V.

Re: SGI Developer Magic - Ada95

[Corey Ashford]

Viqar Abbasi wrote in message <35ABDCDF.7B8F@cae.ca>...
>Has anyone out there used the "SGI Developer Magic ProDev Ada95"
>compiler?
>
>The information provided on the SGI website
>(http://www.sgi.com/Products/DevMagic/products/project_ada95.html)
>seems to indicate that its the tool for the task at hand.
>
[snip]
>
>We don't need all the graphical-debugger and fancy stuff like
>that provided by SGI.  We just want a stable, robust Ada compiler,
>which produces object files we can link in with our existing software.
>Is SGI Ada95 the answer?


I'm fairly certain SGI's Ada95 compilation system is based on GNAT.
At least it was a year or so ago.

- Corey

Re: SGI Developer Magic - Ada95

[Robert Dewar]

Corey said

<<I'm fairly certain SGI's Ada95 compilation system is based on GNAT.
At least it was a year or so ago.
>>

It still is. The new SGI release will be essentially similar to the
forthcoming 3.11b release of GNAT. SGI customizes GNAT to work nicely
with workshop (ACT reacquires these patches), and also provides a library
of bindings to the SGI libraries. In addition, you can obtain from SGI
a version of Workshop customized for GNAT. The latest incarnation of
this technology is 100% Ada aware. For more information contact your
SGI representative.

Robert Dewar
Ada Core Technologies

Note that SGI is one of the only work station companies to provide a fully
integrated, fully supported Ada 95 technology. The only other such company
(as far as I am aware) is Siemens Nixdorf, whose Ada technology is also
based on GNAT.

Re: SGI Developer Magic - Ada95

[David C. Hoos, Sr.]

Viqar Abbasi wrote in message <35ABDCDF.7B8F@cae.ca>...
>Has anyone out there used the "SGI Developer Magic ProDev Ada95"
>compiler?
>
>The information provided on the SGI website
>(http://www.sgi.com/Products/DevMagic/products/project_ada95.html)
>seems to indicate that its the tool for the task at hand.
>
>The task at hand is to incorporate Ada into large infrastructure
>based in C and Fortran, on the SGI Irix platform.  I want to
>call the Ada from C and Fortran, without any limitations.
>The Verdix compiler we have now doesn't allow "in out" parameters
>in procedures.. just "out" parameters.  Also, we're able to link
>and run basic programs in GNAT... but have been having other
>problems with newer versions of GNAT.  The last stable version
>seems to be 3.09.  Therefore, we're a bit GNAT-wary.
>
>We don't need all the graphical-debugger and fancy stuff like
>that provided by SGI.  We just want a stable, robust Ada compiler,
>which produces object files we can link in with our existing software.
>Is SGI Ada95 the answer?

SGI Ada95 id gnat-based.  A year-and-a-half ago, I began a substantial
project with SGI Version 1.2 which was (if memory serves me correctly) gnat
3.07-based, and later I upgraded with a patch provided on the SGI web site
which upgraded to gnat 3.09.

SGI provides some libraies -- e.g., a Posix binding based on the
now-obsolete forest -- which are not part of the public gnat releases.

SGI Version 1.3 is now in beta, but since I have been involved in a Solaris
project for several months, I haven't looked at 1.3 yet, so don't know on
which gnat version it's based.

On another SGI project with which I am involved we had been using the latest
public release (3.10), but because we needed some problems fixed we
purchased a support contract, and have been using "wavefront" releases of
3.11.

The thing which was particularly attractive to us in January 1997 about the
SGI version 1.2 was the GUI-based debugger which was much superior to the
Ada-aware GDB available at that time.

Also, debugging of tasks was made easier by the sproc-based tasking model,
and I imagine that SGI was able to leverage sproc-based thread debugging for
other languages in implementing the Ada debugger.

With regard to your comment about the Verdix limitation on procedure
parameter types -- I believe you're talking about procedures implemented by
means of pragma Interface or exported to be called by other languages by
means of pragma Export.  I seem to remember the limitation, and I know we
worked around it, but I don't rememeber the details.

My own experience (for what it's worth) is that the gnat product is as good
as anything else out there (and better in many respects).  I don't feel
you'll find better support than what you can get from a gnat support
contract.  I still have some outstanding unfixed Verdix bugs which go back
four years.

I hope this helps.

David C. Hoos, Sr.

Re: SGI Developer Magic - Ada95

[dennison]

In article <35ABDCDF.7B8F@cae.ca>,
  viqar@cae.ca wrote:
...
> based in C and Fortran, on the SGI Irix platform.  I want to
> call the Ada from C and Fortran, without any limitations.
> The Verdix compiler we have now doesn't allow "in out" parameters
> in procedures.. just "out" parameters.  Also, we're able to link

Wow! That's quite a restriction.

I suspect what you mean to say here is that the Verdix compiler only permits
parameters of mode "in" when interfacing to C or Fortran. That is the
restriction the old VADS-based compiler I have used had.

That is an annoyance, but its certianly not fatal. That is, after all, the way
C passes parameters. It doesn't seem to have deterred many people from using C
compilers! :-)

It would be nice to have a more convienent interface, but I certianly
wouldn't change compilers over it. Of course VADS is only Ada83. That's a
good reason to change.

T.E.D.

-----== Posted via Deja News, The Leader in Internet Discussion ==-----
http://www.dejanews.com/rg_mkgrp.xp   Create Your Own Free Member Forum

Compilers (was Re: SGI Developer Magic - Ada95)

[Viqar Abbasi]

Thanks for the help on this issue.  I'm actually glad to hear that
the SGI compiler is based on GNAT, since GNAT has been the best
we've dealt with for C/Fortran/Ada interfacing.  I'm not sure
exactly what kind of problems people have had with the newer
versions.  Suffice to say, we'll need a support contract, with
whichever compiler is ultimately chosen for the task.  One of 
my colleagues is going to investigate APEX as well... do any of
you helpful people have any experience with this one?

Oh... and for the curious...

> In article <35ABDCDF.7B8F@cae.ca>,
>   viqar@cae.ca wrote:
> ...
> > based in C and Fortran, on the SGI Irix platform.  I want to
> > call the Ada from C and Fortran, without any limitations.
> > The Verdix compiler we have now doesn't allow "in out" parameters
> > in procedures.. just "out" parameters.
> 

Everyone guessed right that I was referring to exporting a 
procedure from Ada, and Verdix restricted procedure parameters
to "in", (not "out", as I wrote).  Sure we can get around it...
but all the Ada code has already been written.  We want miminal
changes, if any at all...

V.

Re: Compilers (was Re: SGI Developer Magic - Ada95)

[David C. Hoos, Sr.]

Viqar Abbasi wrote in message <35ACBD51.6777@cae.ca>...
>Thanks for the help on this issue.  I'm actually glad to hear that
>the SGI compiler is based on GNAT, since GNAT has been the best
>we've dealt with for C/Fortran/Ada interfacing.  I'm not sure
>exactly what kind of problems people have had with the newer
>versions.  Suffice to say, we'll need a support contract, with
>whichever compiler is ultimately chosen for the task.  One of
>my colleagues is going to investigate APEX as well... do any of
>you helpful people have any experience with this one?
>
Well, the the Solaris and SGI gnat projects to which I referred in my
earlier post were actually developed in the Apex environment hosted on
Solaris, using the Apex RCI.

The reason Apex was hosted on Solaris is that one of the tools (Rational
Rose, I believe it was) was not available hosted on SGI.

Apex comes with RCI customizations for gnat (version 3.07 at the time we
obtained Apex), but the customizations were unsatisfactory, so I re-did
them, and included Posix bindings (the SGI-supplied forest on SGI, and
florist on Solaris).

The reason for using gnat was that (at least at the time) Apex's Ada95
implementation was lacking so many of the features of the language as to be
unusable.

So, basically, we used the facilities of Apex for CM and source code
development, with its many powerful features, but not for code generation,
or debugging.  We used the SGI debugger on the SGIs and the Ada-aware GDB on
the Solaris targets.

Both our Solaris and SGI systems were on a common network (but with distinct
NIS domains), with the files being served to all workstations by a
four-processor Onyx, so we used the rsh_nfs RCI connection mode.

 Hope this helps,

David C. Hoos, Sr.

Re: SGI Developer Magic - Ada95

[Robert Dewar]

Viqar said

<<The task at hand is to incorporate Ada into large infrastructure
based in C and Fortran, on the SGI Irix platform.  I want to
call the Ada from C and Fortran, without any limitations.
The Verdix compiler we have now doesn't allow "in out" parameters
in procedures.. just "out" parameters.  Also, we're able to link
and run basic programs in GNAT... but have been having other
problems with newer versions of GNAT.  The last stable version
seems to be 3.09.  Therefore, we're a bit GNAT-wary.
>>

I am not quite sure what version of GNAT Viqar has tried, but we are
certainly not aware of any customer problems involving regressions
with respect to 3.09. There can of course be cases where 3.09 works
and 3.10 does not, although we don't know many of these (that is at
this stage pretty much a moot point anyway, since 3.11 is the current
version and has been for a while).

Incidentally, the great majority of cases where complaints have been made
that something worked with version xxx and not with a later version turn
out to be cases where either the program is erroneous and the behavior has
changed, or alternatively that the program was illegal, and the illegality
was not properly diagnosed by the earlier version. This is not to say that
there are no cases of regressions, but they are relatively rare compared
to these kind of issues (a lot of our support work is in sorting out
problems of this kind!)

Robert Dewar
Ada Core Technologies