Re: IEEE float I/O in DEC Ada

Re: IEEE float I/O in DEC Ada

[Robert Dewar]

In article <374EB9A9.F226AEE7@Maths.UniNe.CH>,
  Gautier <Gautier.deMontmollin@Maths.UniNe.CH> wrote:
> Hi all.
>
> Since DEC Ada wants either
>
>   pragma float_representation(vax_float);
>
> or
>
>   pragma float_representation(ieee_float);
>
> for a whole "project" and a part of mine requires
> the VAX representation, so I must use it.
>
> _But_ I wish to save results in IEEE format -
> I can live with text format but the IEEE is
> a bit more compact...
>
> DEC Ada doesn't seem to have special types or
> conversions functions for both floats sorts.
>
> Anyway, are there conversion routines available ?
>
> NB: the version is :
>  "DEC Ada V3.2-8 with the Professional Development Option"
>
> TIA


You can definitely handle both types within a single program
and convert between them (we know this only too well since it
was one of the relatively difficult requirements that the
GNAT implementation had to meet in copying DEC Ada 83).


Sent via Deja.com http://www.deja.com/
Share what you know. Learn what you don't.

IEEE float I/O in DEC Ada

[Gautier]

Hi all.

Since DEC Ada wants either

  pragma float_representation(vax_float);

or

  pragma float_representation(ieee_float);

for a whole "project" and a part of mine requires
the VAX representation, so I must use it.

_But_ I wish to save results in IEEE format -
I can live with text format but the IEEE is
a bit more compact...

DEC Ada doesn't seem to have special types or
conversions functions for both floats sorts.

Anyway, are there conversion routines available ?

NB: the version is :
 "DEC Ada V3.2-8 with the Professional Development Option"

TIA

-- 
Gautier

--------
http://members.xoom.com/gdemont/

Re: IEEE float I/O in DEC Ada

[GitaSrinivasan]

It is not clear whether you are talking about a VAX or an Alpha AXP
platform.

OpenVMS on AXP contains routines to perform such conversions -- I would look
for :

HELP RTL_Routines LIB$

In addition, if memory serves me correctly System.t_float is a valid data
type -- so simply declaring a variable of that type and assigning your
"float" variable should do the trick :

       declare
              myfloat : float ;
              ieee_float : system.t_float ;
       begin
             ieee_float := system.t_float( myfloat ) ;
       end ;

 I would look ada ada$predefined:system_.adc
On VAX systems, we had to develop a routine to perform these conversions. I
am not sure if the recent Open VMS releases on VAX'en would have these
routines are not.


Gautier wrote in message <374EB9A9.F226AEE7@Maths.UniNe.CH>...
>Hi all.
>
>Since DEC Ada wants either
>
>  pragma float_representation(vax_float);
>
>or
>
>  pragma float_representation(ieee_float);
>
>for a whole "project" and a part of mine requires
>the VAX representation, so I must use it.
>
>_But_ I wish to save results in IEEE format -
>I can live with text format but the IEEE is
>a bit more compact...
>
>DEC Ada doesn't seem to have special types or
>conversions functions for both floats sorts.
>
>Anyway, are there conversion routines available ?
>
>NB: the version is :
> "DEC Ada V3.2-8 with the Professional Development Option"
>
>TIA
>
>--
>Gautier
>
>--------
>http://members.xoom.com/gdemont/

Re: IEEE float I/O in DEC Ada

[Thierry Lelegard]

The pragma float_representation(vax_float) simply specifies
the representation format for the "standard" types named
Float.

However, the package System contains explicit (and non-standard)
type definitions for all VAX and IEEE floats. If you have
specific (and explicit) requirements for float representation,
I would suggest to use the explicit and non-standard types
and _not_ use the pragma which is confusing and error-prone.

-Thierry
________________________________________________________
Thierry Lelegard, Paris, France
E-mail: lelegard@club-internet.fr

Re: IEEE float I/O in DEC Ada

[Gautier]

Thank you all for your help!

The link to file ada$predefined:system_.adc
was decisive. The conversions work - I precise
that the machine is an Alphaserver 8420 under OpenVMS V6.2.

Testing works - apart a carriage return that
DEC Ada adds after each binary record !

It seems even usual; dir /full outputs:
...
Record format:      Fixed length 4 byte records
Record attributes:  Carriage return carriage control

But now the situation is under control...

Thank you again!

-- 
Gautier

--------
http://members.xoom.com/gdemont/

PS: test programs.

-- Test IEEE float output on AXP with DEC Ada

with system;
with sequential_io;
with text_io;                use text_io;

procedure wr_ieee is

  subtype ieee_real is system.ieee_single_float;
  subtype real is long_float;

  package rio  is new float_io(real);      use rio;
  package irio is new float_io(ieee_real); use irio;

  package sio is new sequential_io(ieee_real); use sio;

  f: sio.file_type;

  d: constant array(1..6) of real:=
    (  1.2345678901234567890,  1.0e40,  1.0e-39,
      -1.2345678901234567890, -1.0e40, -1.0e-39 );

  function To_IEEE(r: real) return ieee_real is
    begin
      if r > real(ieee_real'last) then      -- overflow
        return ieee_real'last;
      elsif r < real(ieee_real'first) then  -- overflow
        return ieee_real'first;
--      elsif abs(r) < real(ieee_real'small) then -- underflow (seems
useless)
--        return 0.0;
      else
        return ieee_real(r);
      end if;
    end To_IEEE;

  begin
    for i in d'range loop
      Put("   real  ");    Put(d(i));
      Put("   ieee  ");    Put(to_ieee(d(i)));
      New_Line;
    end loop;
    Create(f, name=> "IEEE.DAT");
    for i in d'range loop Write(f, to_ieee(d(i)) ); end loop;
    Close(f);
  end wr_ieee;

$ ru wr_ieee
   real   1.23456789012346E+00   ieee   1.23457E+00
   real   1.00000000000000E+40   ieee   3.40282E+38
   real   1.00000000000000E-39   ieee   0.00000E+00
   real  -1.23456789012346E+00   ieee  -1.23457E+00
   real  -1.00000000000000E+40   ieee  -3.40282E+38
   real  -1.00000000000000E-39   ieee   0.00000E+00

-- Test IEEE float input on PC with GNAT

with sequential_io;
with text_io;                use text_io;

procedure rd_ieee is

  subtype ieee_real is float;
  package irio is new float_io(ieee_real); use irio;
  package sio is new sequential_io(ieee_real); use sio;

  f: sio.file_type;
  d: array(1..6) of ieee_real;

  begin
    Open(f, in_file, "IEEE.DAT");
    for i in d'range loop
      Read(f,d(i));
      Put("   ieee  ");    Put(d(i));
      New_Line;
    end loop;
    Close(f);
  end;

C:\USERS\GAUTIER\ADA\TESTS>rd_ieee.exe
   ieee   1.23457E+00
   ieee   3.40282E+38
   ieee   0.00000E+00
   ieee  -1.23457E+00
   ieee  -3.40282E+38
   ieee   0.00000E+00