Hi,

Looking at the FPC Language Reference for Variable Declaration syntax:

http://www.freepascal.org/docs-html/ref/refse21.html

Specifically, ABSOLUTE declarations. The documentation gives this example:

var
  curterm1 : integer;
  curterm6 : integer absolute curterm1;

Now according to the syntax diagram the syntax after ABSOLUTE can be an Identifier or an Integer Expression. So the above is an example where an Identifier is used. So what is (code example) of valid syntax when the Integer Expression syntax is used?

Regards,
- Graeme -

In our previous episode, Graeme Geldenhuys said:

IIRC on Dos you can use addresses in [segment:offset] notation.

Graeme Geldenhuys wrote

So what is (code example) of valid syntax when the Integer Expression syntax is used?

var
  NilPointer: Pointer absolute 0;

Bart

;-)

There's a lot of use in embedded targets:
const
   ADC1_BASE              = $40012000;
var
   ADC1                   : TADC_Registers absolute ADC1_BASE;

Hello Michael,

Michael Ring wrote

There's a lot of use in embedded targets:

Ah, thank you. Finally an answer I can understand. :)

Regards,
- Graeme -

fpGUI Toolkit - a cross-platform GUI toolkit using Free Pascal http://fpgui.sourceforge.net/

My public PGP key: http://tinyurl.com/graeme-pgp

Graeme Geldenhuys wrote

On 2015-10-13 at 22:13, Michael Ring wrote: > There's a lot of use in embedded targets: Ah, thank you. Finally an answer I can understand. :)

The case of MS-DOS using it e.g. to provide direct access to the video adapter memory ($B800:0000), low-level information maintained by BIOS ($40:xxx), etc., as already mentioned by someone else (Marco?) is in principle the same as use cases relevant for embedded targets.

Tomas

In our previous episode, Tomas Hajny said:

Yes, but I thought it was square brackets, not parentheses. Address is linearized using seg*16+ofs and addressed via %fs (for go32v2, for i8086 it probably is used as is)

[Checks] No bracket or parentheses in the case of Turbo Pascal, but I'd point out that the syntax was $B800:$0000 i.e. the absolute address was two concatenated 16-bit words rather than a colon being acceptable in the middle of a 32-bit word. Modula-2 IIRC was about the same.

The C crowd would probably like to use (square) brackets on account of their array semantics, and in the general case it's arguable whether defining an absolute variable is better or worse than defining a pointer with an absolute initial value. I suppose that a compiler/linker could potentially interpret an absolute declaration as indicating that that location should not be swapped/overlaid/relocated.

There are no parentheses nor brackets when used with "absolute":

---------------
var
Shifts1: byte absolute $40:$17; (* shift status as set by BIOS *) Shifts2: byte absolute $40:$18; (* (extended and normal) *) ===============

Brackets are used when using the other option for direct memory access - accessing the memory as array via Mem[] / MemW[].

Tomas