Omission in the Ada Standard

Omission in the Ada Standard

[Robert C. Leif]

While I was working on my latest project, the CellFuge, I was shocked
to learn that Jerry van Dijk's Io_Ports package that I have did not
compile with GNAT. I do need an up to date version. Since the Intel
Pentium class of processors and their clones are a very large part of
the software market, the possibility of port based IO should be
included in Ada.  I would suggest a child library of Package System.
Something like the source text below. The with System and use System
were included to permit compilation.
with System;
use System;
package System.Ports is
   type Port is new Address; 
   Null_Port : constant Port := Port (Null_Address);  
   Port_Unit : constant      := Storage_Unit;  

   -- Address Comparison: 
   function "<" (
         Left,        
         Right : Port ) 
     return Boolean; 
   function "<=" (
         Left,        
         Right : Port ) 
     return Boolean; 
   function ">" (
         Left,        
         Right : Port ) 
     return Boolean; 
   function ">=" (
         Left,        
         Right : Port ) 
     return Boolean; 
   function "=" (
         Left,        
         Right : Port ) 
     return Boolean; 
   -- function "/=" (Left, Right : Port) return Boolean; 
   -- "/=" is implicitly defined 
end System.Ports;
This would be followed by variants of the other System child
libraries. One way to do this is to create a generic package that
could be instantiated into both System and Ports.
Bob
Robert C. Leif, Ph.D.
Email rleif@rleif.com

Re: Omission in the Ada Standard

[Larry Kilgallen]

In article <657ea3e3.0309250705.3537b0b0@posting.google.com>, rleif@rleif.com (Robert C. Leif) writes:
> While I was working on my latest project, the CellFuge, I was shocked
> to learn that Jerry van Dijk's Io_Ports package that I have did not
> compile with GNAT. I do need an up to date version. Since the Intel
> Pentium class of processors and their clones are a very large part of
> the software market, the possibility of port based IO should be
> included in Ada.

And what would the meaning be on some other processor for which a
standard-compliant Ada compiler is needed ?

Omission in the Ada Standard

[Robert C. Leif]

This should be Re: Omission in the Ada Standard. However, neither the
original nor Mr. Kilgallen's reply have appeared on the Google news
reader, which will not transmit with the Re:

Larry Kilgallen stated, "And what would the meaning be on some other
processor for which a standard-compliant Ada compiler is needed?"

If a CPU does NOT have ports, the part(s) of the compiler specific for
ports do not have to be implemented.

Bob Leif
Robert C. Leif, Ph.D.
rleif@rleif.com
www.newportinstruments.com

RE: Omission in the Ada Standard

[Beard, Frank Randolph CIV]

-----Original Message-----
From: Larry Kilgallen

>> Robert C. Leif writes:
>> While I was working on my latest project, the CellFuge, I was shocked
>> to learn that Jerry van Dijk's Io_Ports package that I have did not
>> compile with GNAT. I do need an up to date version. Since the Intel
>> Pentium class of processors and their clones are a very large part of
>> the software market, the possibility of port based IO should be
>> included in Ada.
>
> And what would the meaning be on some other processor for which a
> standard-compliant Ada compiler is needed ?

I guess the same as GNAT.  Over half of the GNAT source I see, that
is of interest to me, won't compile on my Aonix compiler because of
some GNAT.Some_Package reference.  I have to go in and surgically
modify the code. Portability ends up moving from the language
to the tool.  So, "my GNAT code is portable so long as GNAT has been
ported to the new platform".  This has been an issue for a while and
no one seems to be too concerned.

Granted, in some cases I can pick up that particular GNAT source easily
enough and add it to my project.  But, other times, the GNAT.Some_Package
is so entangled in the rest of GNAT that it becomes a monumental task.  

So, why is adding it to the standard that big of a concern?
Just handle it the same way as Ada.Command_Line.  If it's supported,
great.  If not, don't use it.  And if someone does try to use it,
raise an exception.  

Frank

Re: Omission in the Ada Standard

[Simon Wright]

Kilgallen@SpamCop.net (Larry Kilgallen) writes:

> In article <657ea3e3.0309250705.3537b0b0@posting.google.com>, rleif@rleif.com (Robert C. Leif) writes:
> > While I was working on my latest project, the CellFuge, I was shocked
> > to learn that Jerry van Dijk's Io_Ports package that I have did not
> > compile with GNAT. I do need an up to date version. Since the Intel
> > Pentium class of processors and their clones are a very large part of
> > the software market, the possibility of port based IO should be
> > included in Ada.
> 
> And what would the meaning be on some other processor for which a
> standard-compliant Ada compiler is needed ?

About as little as GNAT's Interfaces.VxWorks package, I guess! (I
think they have a System.VxWorks too).

RE: Omission in the Ada Standard

[Larry Kilgallen]

In article <mailman.3.1064517922.25614.comp.lang.ada@ada-france.org>, "Beard, Frank Randolph CIV" <frank.beard@navy.mil> writes:
> -----Original Message-----
> From: Larry Kilgallen
> 
>>> Robert C. Leif writes:
>>> While I was working on my latest project, the CellFuge, I was shocked
>>> to learn that Jerry van Dijk's Io_Ports package that I have did not
>>> compile with GNAT. I do need an up to date version. Since the Intel
>>> Pentium class of processors and their clones are a very large part of
>>> the software market, the possibility of port based IO should be
>>> included in Ada.
>>
>> And what would the meaning be on some other processor for which a
>> standard-compliant Ada compiler is needed ?
> 
> I guess the same as GNAT.  Over half of the GNAT source I see, that
> is of interest to me, won't compile on my Aonix compiler because of
> some GNAT.Some_Package reference.  I have to go in and surgically
> modify the code. Portability ends up moving from the language
> to the tool.  So, "my GNAT code is portable so long as GNAT has been
> ported to the new platform".  This has been an issue for a while and
> no one seems to be too concerned.

But the suggestion here is that the Ada standard would be platform-specific.
That is quite a different thing than an individual compiler.

Re: Omission in the Ada Standard

[Jeff C,]

"Robert C. Leif" <rleif@rleif.com> wrote in message
news:657ea3e3.0309250705.3537b0b0@posting.google.com...
> While I was working on my latest project, the CellFuge, I was shocked
> to learn that Jerry van Dijk's Io_Ports package that I have did not
> compile with GNAT. I do need an up to date version. Since the Intel
> Pentium class of processors and their clones are a very large part of
> the software market, the possibility of port based IO should be


A few notes...

First....this is not in any language standard...So Ada would be the only
language to support it.

Second...Code that accesses the ports directly is now obsolete (mostly)
since it
is not legal under NT, 2000, XP (although there are some way to simulate it
this is
another area where no compiler (including the VC++) allows you to do this
out of the box.

Third...What this really points out is not so much a problem with the Ada
standard as much as
a problem with providing standard platform specific libraries. It has always
seemed to me to be
a shame that there is no good standardization for platform specific
libraries in Ada.
(i.e. a standard way of getting access to Altivec extensions for PowerPC or
a standard way of
doing what you are asking for)..

So, while I don't think it is appropriate to say these features are missing
from the language standard I do think
that a better job could be done with standard libraries. I think what
prevents it is not having one dominant Monopoly like
microsoft for Ada "setting the standard" such that other vendors feel they
are required to be compatible.

Re: Omission in the Ada Standard

[Larry Kilgallen]

In article <kYMcb.582614$o%2.263727@sccrnsc02>, "Jeff C," <nolongersafeto@userealemailsniff.com> writes:

> Third...What this really points out is not so much a problem with the Ada
> standard as much as
> a problem with providing standard platform specific libraries. It has always
> seemed to me to be
> a shame that there is no good standardization for platform specific
> libraries in Ada.

On VMS there is Starlet, which is used by both DEC Ada and GNAT.

But more important, it is also used by Compaq Basic, HP C, Kednos PL/I,
etc.  So accessing OS features is well documented for all languages.

Re: Omission in the Ada Standard

[Vinzent 'Gadget' Hoefler]

Jeff C, wrote:

>Second...Code that accesses the ports directly is now obsolete (mostly)
>since it is not legal under NT, 2000, XP [...]

I beg to differ. Especially because Ada is well suited for embedded
systems, where port accesses to address external devices are quite
usual.


Vinzent.

Re: Omission in the Ada Standard

[Robert C. Leif]

I believe that other chips besides Intel's have ports. My suggestion
would also work for systems with multiple memory banks. If Ada is to
be used for real-time embedded systems, Ada compilers MUST work out of
the box with major processors. Otherwise, normal engineers will go to
some other language.

Incidentally in order to program special boards, one needs to read
from and write to ports under Windows. I might note after posting my
message and several follow-ups, I have yet to receive any suggestions
as to how to read and write to a port with GNAT.
Bob Leif

Kilgallen@SpamCop.net (Larry Kilgallen) wrote in message news:<OOiLZwwgQEAY@eisner.encompasserve.org>...
> In article <mailman.3.1064517922.25614.comp.lang.ada@ada-france.org>, "Beard, Frank Randolph CIV" <frank.beard@navy.mil> writes:
> > -----Original Message-----
> > From: Larry Kilgallen
> > 
> >>> Robert C. Leif writes:
> >>> While I was working on my latest project, the CellFuge, I was shocked
> >>> to learn that Jerry van Dijk's Io_Ports package that I have did not
> >>> compile with GNAT. I do need an up to date version. Since the Intel
> >>> Pentium class of processors and their clones are a very large part of
> >>> the software market, the possibility of port based IO should be
> >>> included in Ada.
> >>
> >> And what would the meaning be on some other processor for which a
> >> standard-compliant Ada compiler is needed ?
> > 
> > I guess the same as GNAT.  Over half of the GNAT source I see, that
> > is of interest to me, won't compile on my Aonix compiler because of
> > some GNAT.Some_Package reference.  I have to go in and surgically
> > modify the code. Portability ends up moving from the language
> > to the tool.  So, "my GNAT code is portable so long as GNAT has been
> > ported to the new platform".  This has been an issue for a while and
> > no one seems to be too concerned.
> 
> But the suggestion here is that the Ada standard would be platform-specific.
> That is quite a different thing than an individual compiler.

Re: Omission in the Ada Standard

[Jeff C,]

"Robert C. Leif" <rleif@rleif.com> wrote in message
news:657ea3e3.0309260810.4b6682d2@posting.google.com...
> I believe that other chips besides Intel's have ports. My suggestion
> would also work for systems with multiple memory banks. If Ada is to
> be used for real-time embedded systems, Ada compilers MUST work out of
> the box with major processors. Otherwise, normal engineers will go to
> some other language.
>

I don't think we disagree as much as you might think...I think that
any compiler for x86 designed for an OS that supports it should come
with predefined libraries to do port IO. I even wish it were relatively
standard..

Note again that no language I know of has the predefined ability to talk to
ports as
part of the language definition..

It is pretty easy to get GNAT to do it. I'll have to dust off my old wind 95
laptop since once I did some port
IO stuff on there but I never took it off before I put it away.





> Incidentally in order to program special boards, one needs to read
> from and write to ports under Windows. I might note after posting my
> message and several follow-ups, I have yet to receive any suggestions
> as to how to read and write to a port with GNAT.
> Bob Leif
>

Depends what version of Windows you mean. In general one can not execute
the various x86 port instructions on NT, 2000, XP series windows from user
space
code. Even if you get a compiler to generate the instruction, the OS will
trap it and raise an error.

So, the example code I hope to find will absolutely NOT work under NT, 2000,
XP.

Re: Omission in the Ada Standard

[Jeff C,]

"Robert C. Leif" <rleif@rleif.com> wrote in message
news:657ea3e3.0309260810.4b6682d2@posting.google.com...
> I believe that other chips besides Intel's have ports. My suggestion
> would also work for systems with multiple memory banks. If Ada is to
> be used for real-time embedded systems, Ada compilers MUST work out of
> the box with major processors. Otherwise, normal engineers will go to
> some other language.
>

I don't think we disagree as much as you might think...I think that
any compiler for x86 designed for an OS that supports it should come
with predefined libraries to do port IO. I even wish it were relatively
standard..

Note again that no language I know of has the predefined ability to talk to
ports as
part of the language definition..

It is pretty easy to get GNAT to do it. I'll have to dust off my old wind 95
laptop since once I did some port
IO stuff on there but I never took it off before I put it away.





> Incidentally in order to program special boards, one needs to read
> from and write to ports under Windows. I might note after posting my
> message and several follow-ups, I have yet to receive any suggestions
> as to how to read and write to a port with GNAT.
> Bob Leif
>

Depends what version of Windows you mean. In general one can not execute
the various x86 port instructions on NT, 2000, XP series windows from user
space
code. Even if you get a compiler to generate the instruction, the OS will
trap it and raise an error.

So, the example code I hope to find will absolutely NOT work under NT, 2000,
XP.

Re: Omission in the Ada Standard

[Keith Thompson]

"Jeff C," <nolongersafeto@userealemailsniff.com> writes:
[...]
> Note again that no language I know of has the predefined ability to
> talk to ports as part of the language definition..

Ada 83 had:

package LOW_LEVEL_IO is
   --  declarations of the possible types for DEVICE and DATA;
   --  declarations of overloaded procedures for these types:
   procedure SEND_CONTROL    (DEVICE : device_type; DATA : in out data_type);
   procedure RECEIVE_CONTROL (DEVICE : device_type; DATA : in out data_type);
end;

I don't know whether anybody actually implemented it, and it was
dropped in Ada 95 (it's not just obsolescent, there's no reference to
it at all).

-- 
Keith Thompson (The_Other_Keith) kst@cts.com  <http://www.ghoti.net/~kst>
San Diego Supercomputer Center           <*>  <http://www.sdsc.edu/~kst>
Schroedinger does Shakespeare: "To be *and* not to be"

Re: Omission in the Ada Standard

[Dave Thompson]

On Fri, 26 Sep 2003 01:51:16 GMT, "Jeff C,"
<nolongersafeto@userealemailsniff.com> wrote:

> 
> "Robert C. Leif" <rleif@rleif.com> wrote in message
> news:657ea3e3.0309250705.3537b0b0@posting.google.com...
> > While I was working on my latest project, the CellFuge, I was shocked
> > to learn that Jerry van Dijk's Io_Ports package that I have did not
> > compile with GNAT. I do need an up to date version. Since the Intel
> > Pentium class of processors and their clones are a very large part of
> > the software market, the possibility of port based IO should be
> 
> 
> A few notes...
> 
> First....this is not in any language standard...So Ada would be the only
> language to support it.
> 
However the C committee (JTC1 SC22 WG14) appears close to adopting a
Technical Report, TR 18037, that includes this as one of several
extensions for embedded systems.  Such a TR has roughly the same
official status as an optional Annex.

> Second...Code that accesses the ports directly is now obsolete (mostly)
> since it
> is not legal under NT, 2000, XP (although there are some way to simulate it
> this is
> another area where no compiler (including the VC++) allows you to do this
> out of the box.
> 
Although one of the advantages (and goals) of Ada is suitability for
programming on bare hardware as well as bloated OSes.

But in this particular case, why not just use machine-code insertions?
(As I see someone else has already detailed.) It's platform dependent,
but the desired functionality is anyway.

- David.Thompson1 at worldnet.att.net

Re: Omission in the Ada Standard

[Jerry Petrey]

Keith Thompson wrote:

> "Jeff C," <nolongersafeto@userealemailsniff.com> writes:
> [...]
> > Note again that no language I know of has the predefined ability to
> > talk to ports as part of the language definition..
>
> Ada 83 had:
>
> package LOW_LEVEL_IO is
>    --  declarations of the possible types for DEVICE and DATA;
>    --  declarations of overloaded procedures for these types:
>    procedure SEND_CONTROL    (DEVICE : device_type; DATA : in out data_type);
>    procedure RECEIVE_CONTROL (DEVICE : device_type; DATA : in out data_type);
> end;
>
> I don't know whether anybody actually implemented it, and it was
> dropped in Ada 95 (it's not just obsolescent, there's no reference to
> it at all).
>
> --
> Keith Thompson (The_Other_Keith) kst@cts.com  <http://www.ghoti.net/~kst>
>

DDCI implemented Low_Level_IO for their 80x86 cross compiler.  We used it on a
project I just finished working on.

Jerry

--
---------------------------------------------------------------------------------

-- Jerry Petrey
-- Senior Principal Systems Engineer - Navigation (GPS/INS), Guidance, & Control
-- Raytheon Missile Systems          - Member Team Ada & Team Forth
-- NOTE: please remove <NOSPAM> in email address to reply
---------------------------------------------------------------------------------

Re: Omission in the Ada Standard

[Jeff C,]

"Jerry Petrey @raytheon.com>" <"jdpetrey<NOSPAM> wrote in message
news:3F859331.AD61B137@raytheon.com...
>
>
> Keith Thompson wrote:
>
> > "Jeff C," <nolongersafeto@userealemailsniff.com> writes:
> > [...]
> > > Note again that no language I know of has the predefined ability to
> > > talk to ports as part of the language definition..
> >
> > Ada 83 had:
> >
> > package LOW_LEVEL_IO is
> >    --  declarations of the possible types for DEVICE and DATA;
> >    --  declarations of overloaded procedures for these types:
> >    procedure SEND_CONTROL    (DEVICE : device_type; DATA : in out
data_type);
> >    procedure RECEIVE_CONTROL (DEVICE : device_type; DATA : in out
data_type);
> > end;
> >
> > I don't know whether anybody actually implemented it, and it was
> > dropped in Ada 95 (it's not just obsolescent, there's no reference to
> > it at all).
> >
> > --
> > Keith Thompson (The_Other_Keith) kst@cts.com
<http://www.ghoti.net/~kst>
> >
>
> DDCI implemented Low_Level_IO for their 80x86 cross compiler.  We used it
on a
> project I just finished working on.
>
> Jerry
>

Of course the standard never said what the heck low level IO was even
supposed to do (although it
did have a flavor of x86 IO ports to some extent).. In any case, the fact
that some vendor chose to implement
port access via low_level_io is a far cry from claiming that this behaviour
was defined by the standard.