mad pascal optimizes indexing wrong

[aydindemircioglu]

i think mad pascal again over-optimizes incorrectly, i do not have a minimal example at hand,
but my code is like this:

    weaponNames:   array[0..8] Of String[16] =   ('H', 'M', 'X', 'Y', 'Z'~);
    fp_N : array[0..2] of byte absolute $e400;
    fp_weapon: array[0..16*2-1] of byte absolute $e462;
    fp_Gang: array[0..2] of String[15] absolute $e482;
    fp_name: array[0..16*2-1] of String[16];

    fp_N[0] := 4;
    fp_N[1] := 2;
    fp_Gang[0] := 'Tunkar'~;
    fp_Gang[1] := 'Mari Bani'~;
    fp_name[0] := 'Peter 1'~;
<...>
    fp_name[16] := 'Borke 1'~;
<...>
    fp_name[17] := 'Borke 2'~;
    fp_weapon[0] := 0;
    fp_weapon[1] := 1;
    fp_weapon[2] := 2;
    fp_weapon[3] := 3;
    fp_weapon[16] := 3;
    fp_weapon[17] := 5;

        For s := 0 To 1 Do
            For i := 0 To fp_N[s]-1 Do
                Begin;
                    If fp_energy[s SHL 4 +i] = 0 Then continue;

                    // player stats
                    // fill old stuff first 
                    FillChar (Pointer (MAP_SCR_ADDRESS+19*40), 5*40, ' '~);
                    ofs := 20*s;
                    CRT_GotoxY(ofs,20);
                    CRT_Write(fp_Name[s SHL 4+i]);
< more stuff involving s SHL 4 +i >
                    CRT_Write(fp_weapon[s SHL 4 + i]);
                    CRT_Write(weaponNames[fp_weapon[s SHL 4 +i]]);
...

this results that the weapon names from player from gang 1 (=0,1,2,3) are OK, but weapon names from players from the other side (with index 16, 17, 18) show the wrong string, possibly because mad-pascal tries to optimize something incorrectly?
this is resolved by just introducing a new byte z like this.

        For s := 0 To 1 Do
            For i := 0 To fp_N[s]-1 Do
                Begin;
                    z := s SHL 4 + i;
                   CRT_Write(weaponNames[fp_weapon[z]]);
...

(i know introducing this byte z is better than writing s SHL 4 + i all the times, but this is debug stadium, optimization will be done later).

am using mad pascal 1.7.0 still, not sure if this was fixed already.

[tebe6502]

in the latest version of MP, there is an error message when trying to use the modifier ABSOLUTE for a string array

fp_Gang: array[0..2] of String[15] absolute $e482;

compiler has no way of instantiating such an array with valid pointers to strings
such an array will contain pointers with a value of 0

[aydindemircioglu]

ok, but how to do it else? what i want would be at $e482 a space of 15 chars (+1 for the string length) that the user can fill (gang name is typed in at start of the game).

[tebe6502]

fp_Gang: array[0..2] of pointer absolute $e482;
gang0, gang1, gang2 : string[15];

begin
fp_Gang[0] := @gang0;
fp_Gang[1] := @gang1;
fp_Gang[2] := @gang2;

fp_Gang[2]:='string2';
fp_Gang[1]:='string1';
fp_Gang[0]:='string0';

writeln(Pstring(fp_Gang[2]));

end.

[zbyti]

{$define basicoff}
{$define romoff}

type TGang = record
   name: string[15];
   logo: string[15];
end;

var
   fp_Gang: array[0..2] of ^TGang absolute $e482;
   gang0, gang1, gang2 : TGang;

begin

gang0.name := 'string1';
gang1.name := 'string1';
gang2.name := 'string2';
gang0.logo := 'red';
gang1.logo := 'white';
gang2.logo := 'black';

fp_Gang[2] := @gang2;
fp_Gang[1] := @gang1;
fp_Gang[0] := @gang0;

writeln(Pstring(fp_Gang[2].name));
writeln(Pstring(fp_Gang[1]));
writeln(Pstring(fp_Gang[2].logo));
writeln(Pstring(fp_Gang[1].logo));

repeat until false;

end.

[aydindemircioglu]

thanks to both of you!
i will try to see if the problem is there with the latest version and if i use correct code.
it may take a couple of days (or weeks), need to finish other work first.

[zbyti]

var
   fp_Gang: array[0..2] of String[15] absolute $e482;

begin
   repeat until false;
end.

Mad Pascal Compiler version 1.7.1 [2024/02/03] for 6502
Compiling test.pas
test.pas (5,52) Error: ABSOLUTE modifier is not available for this type of array

[aydindemircioglu]

but that was not the initial problem. in the code above weaponnames are a usual array of strings, not fixed to an absolute address. the fp_weapon was absolute, but its an array of byte. yet, it seemed to me that mad pascal confused the indexing. as said i will check this later and reopen if need be.

[zbyti]

I'll check it

I remember that I had a similar problem in previous versions (2-3 years ago) and I wrote a workaround then.

I also had a problem when I was trying to calculate an index inside array brackets (the problematic array was nested within another array), and the solution was to make this calculation one step earlier

[zbyti]

@aydindemircioglu I wrote some simple examples considering your case and got the correct result, please provide full example of your code.

[aydindemircioglu]

thanks, as said, might be due to my older version. the code is embedded in a long procedure, i have to take my time to get a minimal example. for now i will then close this issue, and will reopen it if i am able to reproduce with the minimal example.