FPC compiles pusha/popa as a pushaw/popaw. Is it doing right?

[Avinash]

I would expect PUSHAD/POPAD from 32-bit compiler.

[marcov]

They have the same opcode, so the meaning is probably set by the mode ?  That might be the disassembler at fault. (or whatever the backend assembler accepts, in the case of -a)

[Avinash]

Code: Pascal  [Select][+][-]

1. {$mode objfpc}
2. {$asmmode intel}
3. var L: LongInt;
4. begin
5.   asm
6.     xor  eax,eax
7.     //pusha
8.     dec  eax
9.     //popa
10.     mov  L, eax
11.   end;
12.  
13.   WriteLn(HexStr(L, 8));
14. end.

FFFFFFFF

Code: Pascal  [Select][+][-]

1. {$mode objfpc}
2. {$asmmode intel}
3. var L: LongInt;
4. begin
5.   asm
6.     xor  eax,eax
7.     pusha
8.     dec  eax
9.     popa
10.     mov  L, eax
11.   end;
12.  
13.   WriteLn(HexStr(L, 8));
14. end.

FFFF0000

[PascalDragon]

Yes, this is by design. See this comment from the compiler's source:

Code: Pascal  [Select][+][-]

1.           { pushf/popf/pusha/popa have to default to 16 bit in Intel mode
2.           (Intel manual and Delphi-compatbile) -- setting the opsize for
3.           these instructions doesn't change anything in the internal assember,
4.           so change the opcode }