Re: minimal hardware for ada (newbie)

[Tucker Taft <stt@averstar.com>]

Niklas Holsti wrote:
> 
> Tucker Taft wrote:
> >
> > We have developed (and recently validated) a version of our
> > AdaMagic technology which uses optimized ANSI C as its intermediate
> > representation.  One of the reasons was to allow a programming team to
> > write their source code in Ada, while relying only on the availability
> > of an ANSI C compiler for their chosen target.  We also have
> > a version of the Ada run-time that at the lowest level depends
> > only on routines included in the standard ANSI C run-time.
> 
> I'm working on an I8051 program to control a space-based instrument,
> and have again experienced the pain of working in C. Using
> AdaMagic-to-C would have been nice, however I think it would
> only be practical if one can make the C code use the typical
> I8051 C extensions to ANSI C, such as controlling the memory space
> for each variable. In Keil C51, for example, one of the new
> keywords 'data', 'idata', 'pdata' or 'xdata' is placed in the
> variable declaration between the type-id and the variable-id
> to determine the memory space.
> 
> While one could implement one's own postprocessor to mutilate the
> C code in this way, it would be simpler and safer to make the
> necessary annotations in the Ada code, perhaps as pragmas on
> the variables or types. Does AdaMagic provide anything like this?

Not currently, though using pragma "Convention" for this might
make sense.  We have looked into something like this for a digital
signal processor which has two memory spaces.  Because we are
generating C source as the intermediate, it is a relatively small
effort to add this kind of thing.  By constrast, if we were generating 
machine code directly, it would obviously be a big deal to add support
for multiple memory spaces.

> Moreover, given the limited architecture of the I8051, it seems
> unlikely that a run-time using only ANSI C routines could perform
> within an order of magnitude of a kernel designed for the
> I8051. Any data on this? For example, how does your ANSI-C-based
> run-time handle the stack when switching tasks?

The run-time is in two layers, the upper layer is written
in Ada, and is fully portable between targets.  The lower layer
is also mostly written in Ada, with a bit of assembler or
C thrown in, and multiple versions exist, depending on the underlying
kernel.  The "pure C" version makes use of setjmp/longjmp for
task stack switching.  Other versions use the thread primitives provided
by the underlying kernel.

> TIA
> 
> Niklas Holsti
> Working at but not speaking for Space Systems Finland Ltd.

-- 
-Tucker Taft   stt@averstar.com   http://www.averstar.com/~stt/
Technical Director, Distributed IT Solutions  (www.averstar.com/tools)
AverStar (formerly Intermetrics, Inc.)   Burlington, MA  USA