Re: Another SBIR, and Ada is still missing in action

Re: Another SBIR, and Ada is still missing in action

[Matt Kennel]

Gregory Aharonian (srctran@world.std.com) wrote:
: So for anyone going to the Ada Porkit, ask AJPO why they can't get Ada into the
: DoD SBIR solicitation?  You win the big battles by winning the small battles.

: Greg Aharonian

Do you want Ada to be used for scientific contracts?

Baseline:

Scientists are poor.  We don't have that much money to spend on software.
The money we do have we spend on essentials MATLAB and Maple and Fortran.

Alleviation:

GNU Ada.  But that isn't good enough yet.  We need, at minimum, GNU Ada that
comes with

	***easy to use BLAS and LAPACK bindings ready to go***

as well as ODE integrators and minimization packages.

	
--
-Matt Kennel  		mbk@inls1.ucsd.edu
-Institute for Nonlinear Science, University of California, San Diego
-*** AD: Archive for nonlinear dynamics papers & programs: FTP to
-***     lyapunov.ucsd.edu, username "anonymous".

Re: Another SBIR, and Ada is still missing in action

[David Weller]

In article <3h6565$2gr@network.ucsd.edu>,
Matt Kennel <mbk@inls1.ucsd.edu> wrote:
>Scientists are poor.  We don't have that much money to spend on software.
>The money we do have we spend on essentials MATLAB and Maple and Fortran.
>
Hehe -- all junior scientists are taught to say that mantra!  That's
cuz they know if they ever said, "Boy, I"m happy with all the
newfangled equipment I've got", their funds would dry up in a minute!
(In reality, some labs are dirt poor, others are rich, but the truth
of the matter is that learning a new language/software tool is a
steep learning curve many scientists are unwilling to take --
especially if it takes away from their research time.)

>Alleviation:
>GNU Ada.  But that isn't good enough yet.  We need, at minimum, GNU Ada that
>comes with
>	***easy to use BLAS and LAPACK bindings ready to go***
>as well as ODE integrators and minimization packages.
>

Amen to that, brother!  Matt, I'm so glad you volunteered to do the
legwork to get this done... [sound of slamming door]...Hey, Matt,
wait, come back!

:-)


-- 
      Frustrated with C, C++, Pascal, Fortran?  Ada95 _might_ be for you!
	  For all sorts of interesting Ada95 tidbits, run the command:
"finger dweller@starbase.neosoft.com | more" (or e-mail with "finger" as subj.)
	

Re: Another SBIR, and Ada is still missing in action

[Matt Kennel]

David Weller (dweller@Starbase.NeoSoft.COM) wrote:
: In article <3h6565$2gr@network.ucsd.edu>,
: Matt Kennel <mbk@inls1.ucsd.edu> wrote:
: >Scientists are poor.  We don't have that much money to spend on software.
: >The money we do have we spend on essentials MATLAB and Maple and Fortran.
: >
: Hehe -- all junior scientists are taught to say that mantra!  That's
: cuz they know if they ever said, "Boy, I"m happy with all the
: newfangled equipment I've got", their funds would dry up in a minute!
: (In reality, some labs are dirt poor, others are rich, but the truth
: of the matter is that learning a new language/software tool is a
: steep learning curve many scientists are unwilling to take --
: especially if it takes away from their research time.)

Alright the real translation.

Those PI's who don't write the hard core executable codes but who
do manage the pursestrings don't care to spend money on compilers.

They think that "if they've bought Fortran, and C and (an inferior) C++ are
free, what is the big deal?"  They seem the same thing.

By contrast, there is a big obvious difference between MATLAB and
no-MATLAB: even a Powerbook-wielding email-tapping grant-mongering 
bigshot PI can notice.

Until those former grad students who have been programming all their
careers make it to tenured professor grade.....{and that is getting
harder and harder}

: Amen to that, brother!  Matt, I'm so glad you volunteered to do the
: legwork to get this done... [sound of slamming door]...Hey, Matt,
: wait, come back!

: :-)

	:-)

Geez I've already put my science at risk by volunteering
for one language already....

--
-Dr. Matt Kennel  		mbk@inls1.ucsd.edu
-Institute for Nonlinear Science, University of California, San Diego
-
- Archive for nonlinear dynamics papers & programs (***SITE CHANGED!!**)
-		-->  ftp://inls.ucsd.edu <--

Re: Another SBIR, and Ada is still missing in action

[Robert Dewar]

Matt says:

GNU Ada.  But that isn't good enough yet.  We need, at minimum, GNU Ada that
comes with

        ***easy to use BLAS and LAPACK bindings ready to go***

as well as ODE integrators and minimization packages.


Matt, why not just call Fortran routines directly. GNAT implements both
pragma Import Fortran, to call Fortran routines, and pragma COnvention
FOrtran (to switch the order of subsripts on arrays). Why do you need
special bindings?

Yes, I know Dave Emery will pop up and say that we need wonderful thick
super-ada-style high level bindings, and no doubt that might be nice, but
what's wrong with just using the Fortran routines directly, can't be any
better or worse than calling such routines in Fortran after all.

Re: Another SBIR, and Ada is still missing in action

[John Woodruff]

Matt and Robert are on about performing scientific computations in Ada,
possibly by using GNAT's fine pragma Import Fortran and COnvention
FOrtran (to switch the order of subsripts).

This solution is fine for a large number of computations, but Matt
specifically mentions ODE integrators; here the situation gets a little
sticky in my experience.

The ODE integrators that I experienced a few years ago required that
the integrand be a function defined on a range between T and (T+h) that
could be called non-determininistically by the integrator. This poses a
problem unless the linker can be pursuaded to invoke an Ada subprogram
from within a Fortran subprogram.  

I think that if GNAT's object-code management allows this kind of
interlanguage operation, then Robert's suggestion is sound.
--
John Woodruff
Lawrence Livermore National Lab
510 422 4661

Re: Another SBIR, and Ada is still missing in action

[Richard Kenner]

In article <WOODRUFF.95Feb7163258@neux2.addvax.llnl.gov> woodruff1@llnl.gov writes:
>This poses a problem unless the linker can be pursuaded to invoke an Ada
>subprogram from within a Fortran subprogram.  
>
>I think that if GNAT's object-code management allows this kind of
>interlanguage operation, then Robert's suggestion is sound.

Yes, it certainly does.  GNAT itself is a C main program (part of GCC)
that calls the Ada-written front end.  When it comes time for it to
generate code, it calls C code (Gigi) to interface to GCC.  Gigi, in
turn, often makes calls to Ada functions in the front end to access
the tree.

Re: Another SBIR, and Ada is still missing in action

[Robert Dewar]

Regarding Fortran calling Ada, the one missing ingrediant is pragma Export,
which should be in GNAT very soon, I don't think we will get it in 2.03,
which is coming out very soon, but it should be in version 2.04 which should
be out by the end of the month.

Re: Another SBIR, and Ada is still missing in action

[paus]

In article <WOODRUFF.95Feb7163258@neux2.addvax.llnl.gov> woodruff@neux2.addvax.llnl.gov  
(John Woodruff) writes:
> The ODE integrators that I experienced a few years ago required that
> the integrand be a function defined on a range between T and (T+h) that
> could be called non-determininistically by the integrator. This poses a
> problem unless the linker can be pursuaded to invoke an Ada subprogram
> from within a Fortran subprogram.  

We do that all the time and never had a problem with that. The only
difficulty with GNAT at the moment is that pragma Export is not yet
supported, but with a little glue code it is no problem to interface
Ada and Fortran in such a way that Fortran is calling Ada routines.
We have a large trajectory optimization program written in Fortran where
the user has to provide some low level routines which define the equations
of motion (which are then integrated internally). I have developed an Ada
binding for this program, so that the user can now specify these routines
in Ada without even noticing that the bulk of the program is not in Ada.
This runs perfectly with GNAT. So there is absolutely no excuse for not
using Ada and GNAT.

Michael

--
|----------------------------------------------------------------------|
|Dipl.-Ing. Michael Paus   (Member: Team Ada)                          |
|University of Stuttgart, Inst. of Flight Mechanics and Flight Control |
|Forststrasse 86, 70176 Stuttgart, Germany                             |
|Phone: (+49) 711-121-1434  FAX: (+49) 711-634856                      |
|Email: Michael.Paus@ifr.luftfahrt.uni-stuttgart.de (NeXT-Mail welcome)|

Re: Another SBIR, and Ada is still missing in action

[David Weller]

In article <3h8k3r$2n1@gnat.cs.nyu.edu>, Robert Dewar <dewar@cs.nyu.edu> wrote:
>GNU Ada.  But that isn't good enough yet.  We need, at minimum, GNU Ada that
>comes with
>        ***easy to use BLAS and LAPACK bindings ready to go***
>as well as ODE integrators and minimization packages.
>
>Matt, why not just call Fortran routines directly. GNAT implements both
>pragma Import Fortran, to call Fortran routines, and pragma COnvention
>FOrtran (to switch the order of subsripts on arrays). Why do you need
>special bindings?
>

How silly, Robert -- that was _exactly_ what Matt was suggesting!
Remember, _somebody_ has to take the time to write the binding to
BLAS or LAPACK (and the last time I checked, that would take a little
bit of somebody's time).  All we need is a hardy volunteer -- I'm
already doing my share, so don't ask me :-)


-- 
      Frustrated with C, C++, Pascal, Fortran?  Ada95 _might_ be for you!
	  For all sorts of interesting Ada95 tidbits, run the command:
"finger dweller@starbase.neosoft.com | more" (or e-mail with "finger" as subj.)
	

Re: Another SBIR, and Ada is still missing in action

[Robert Dewar]

Regarding Dave Wller's note "how silly robert .."
THe thin bindings needed for this kind of interfacing are trivial and
can be easily generated by a binding tool.