Turning a 16-bit Protected Segment into a 32-bit Segment

• Microsoft Macro Assembler for MS-DOS, versions 6.0, 6.1, and 6.11

SUMMARY

By default, 32-bit code created in MASM exists in 16-bit segments. This article shows by example how to use DPMI calls to change a 16-bit segment into a 32-bit segment.

NOTE: On a 386, almost no improvement in performance results from using 32-bit code in a 32-bit segment over using 32-bit code in a 16-bit segment. There may be slight performance improvement on a 486 due to the effects of the 66h prefix byte on instruction pipelining.

MORE INFORMATION

The following code fragment uses DPMI calls to turn a 16-bit protected mode segment into a 32-bit protected mode segment. A number of the instructions are encoded using the DB directive because you need to generate USE16 instructions inside of a USE32 segment. The sample includes an interface function so that you can call the 32-bit segment from a 16-bit C, C++, or FORTRAN module.

NOTE: Your debugger may still report these segments to be 16-bit segments. It all depends on what the debugger expects.

Sample Code

   .MODEL large
   .386


   .CODE
   SwitchInterface PROC FAR16
      call FAR32 PTR Switch16bitTo32bit
      ret
   SwitchInterface ENDP


   CODE32 SEGMENT PARA PUBLIC USE32 'CODE'
   Switch16bitTo32bit PROC FAR32
      xor eax, eax                       ;zero set
      xor eax, DWORD PTR 0C08B0000h      ;zero clear for 32-bit
      jz Still16bit                      ;zero set for 16-bit
      ret

   Still16bit:
      DB 083h, 0ECh, 008h        ;sub sp, 8
      DB 08Ch, 0D0h              ;mov ax, ss
      DB 08Eh, 0C0h              ;mov es, ax
      DB 08Bh, 0FCh              ;mov di, sp

      DB 08Ch, 0CBh              ;mov bx, cs
      DB 0B8h, 00Bh, 0000h       ;mov ax, 000Bh
      int 31h

      DB 026h, 080h, 04Dh, 006h, 040h    ;or BYTE PTR es:[di+6], 40h
      DB 0B8h, 00Ch, 0000h       ;mov ax, 000Ch
      int 31h

   Now32bit:
      add sp, 8
      ret
   Switch16bitTo32bit ENDP
   CODE32 ENDS

   END