www.delorie.com/djgpp/doc/ug/compiling/portother.html   search  
Porting code from other 32-bit compilers to DJGPP

This chapter discusses porting 32-bit code written for other compilers to DJGPP. Mainly for console (TTY or text mode) applications, and I am bias towards Watcom V11.0 (and 10.5). In this chapter I will cover:

  1. Watcom
    1. Runtime environment differences
    2. Using DJGPP tools
    3. Using the right optimizer switches
  2. LCC-WIN32
    1. Runtime environment differences
    2. Using DJGPP tools
  3. Tips on writting code that ports easily.


You may be wondering, if you have Watcom, why use DJGPP in the first place. Well I will be the first to admit that Watcom generates some of the tightest code I have ever seen from a compiler. However DJGPP is fairly respectiable too. DJGPP and WATCOM both share alot of the same optimizations. However the area where WATCOM really shines is in register calling convention and floating point math. In this section I will cover the runtime differences, which are few. The tools and how to use them (compared to Watcom) and what optimizations switches you would like to use (based on what you used for WATCOM).

Runtime Difference

The nice thing about standards, is that they differ very slightly. Now you make think that a standard doesn't change, but you would be thinking wrong. First off WATCOM's differences are to make programming PC's (i386) a whole lot easier. Where DJGPP strives to be compatible with GCC (which it is based on).

Hardware I/O

Hardware IO, is provided via the following functions: inportb, inportw, inportl, outportb, outportw, outportl. All of which are defined in PC.H. You use the inportX functions to read from the ports and outportX to write to them. The X denotes one of the following. b for a 8-bit byte sized operand. w for a 16-bit word sized operand. l for a 32-bit long sized operand.

/* Code example to write and read from ports */
#include <pc.io>

int main(void)
    char a;

    /* read a byte from port 0x100 */
    a = inportb(0x100);

    /* write it to port 0x200 */
    outportb(0x200, a);

It's that simple

Inline Assembly

We all know that sometimes C doesn't suffice and you must use inline assembly for performace. Well in DJGPP that too is no exception. Therefore DJGPP provides a method for doing so. I will write a simple yet complete inline function in watcom, then covert it to DJGPP and comment on it.

DJ help me, I couldn't find info on inline asm in the gcc.inf!!! [it's under "Externsions...C..Asm" - DJ]

Inline Functions

Sometimes a function is small and written in C, and since you want it to become portable you decided to keep it in C. However you would like to inline to save time. Well that's ok, DJGPP supports this too. Making an inlined function is really quite simple watch.

/* TEST.H */

__INLINE__ int sum(int a, int b)
    return a + b;

That simple. Note the __INLINE__ is used in ANSI C headers only. In C code you would use inline, but this only effects the one module only. In C and objective C, the inline has no effect on how the object is linked in externally. Put simply, if you use the inline keyword, the function will be inlined for all the other functions in that one compilation, and leads me to suggest not to define inline functions in headers, since that is bad design.

/* Test.C */

inline int sum(int a, int b)
    return a + b;

int test(int a, b, c)
    return sum(a,b) + sum(b,c) + sum(c,a);

In this case, the function sum is inlined in test. Had you made a header to call sum from outside of this compilation, you would not have it inlined, even if you add __INLINE__.

Zero based segments

In WATCOM you can write to the video memory, using the address 0xA0000 however this is not true in DJGPP. I suggest you read 'porting from 16-bit compilers' also written by me. It's here.

Compiler Tools

In every compiler suite you must master (or atleast be able to use) the compiler tools contained within. However for most ms-dos users, who are used to Sybase or Borland style compilers, may find DJGPP slightly harder to get started with. However this is not the case.

WCL386 vs. GCC

WCL386 is Sybases' frontend tool, that will compile, assemble and link together source/object code to make a final image (read EXE file). DJGPP includes a front end too, and in my opinion is easier to use. It's called GCC and you will find it in \DJGPP\BIN\. One difference between them two that bothered me at first, is that GCC doesn't provide the help screens. However, determination got me thru it. It's quite simple.

Compile Only

With WCL386 you can type in: 'WCL386 /c file1.c file2.c file3.c' to just compile and assemble the source code. This will get you object files that you may wish to distribute. With DJGPP it's quite the same. Just type 'GCC -c file1.c file2.c file3.c' to get the resulting object files. You should note that with DJGPP objects files are in COFF format and have the .O file name extension.

Linking libraries

With WCL386 you can type in 'WCL386 file1.c file2.lib' to compile, assemble and link file1.c with file2.lib. In GCC it gets a little more complicated. In DJGPP library files have a .A file name extension. Now the way you link in libraries is dependant on one thing. If the library is in your \DJGPP\LIB directory, the library must have LIB as the first three letters of the filename. I.e test2.lib would be called libtest2.a. Now to compile/link it, type 'GCC test1.c -ltest2 -o test1.exe' I just introduced something else, the -o switch. (Note: make sure you use a small letter o, otherwise it will toggle the optimizer). This will make it produce test1.exe instead of the default A.EXE. Now if the library is in your current directory. You don't need to append the letters LIB to the filename. And to compile/link type 'GCC test1.c test2.a -o test1.exe'

Making libraries

To make a library with WATCOM you would type 'WLIB mylib.lib +test.obj +test2.obj' and in DJGPP it's just as simple. Just type 'AR RS mylib.a test.o test2.o' and you have library called mylib.a for your use. If you want to put this library in \DJGPP\LIB thne you should append LIB to the beginning of the filename. For example: 'AR RS LIBMYLIB.A test.o test2.o'.


With WATCOM you have WMAKE, to process makefiles. In DJGPP the program is the same, only it's called MAKE. However sometimes WATCOM makefiles will not port to DJGPP because they use WATCOM extenstions.

Optimizer Switches


This concludes my section on Watcom. If you have any other ideas, questions, comments, queries, or problems, just email me.


Runtime Differences

Well LCC-WIN32 is not a dos compiler so there are a few differences. :)

  1. No DPMI
  2. No INT (can't call realmode ones, and can't set one up)

Basically when you program for LCC-WIN32 in console mode, or want to program portably, you must stick to the standard. That is what LCC (LCC- WIN32 is based on) was designed for. It follows the ANSI C standard closer in more repsects to GCC, but has no addons. LCC-WIN32 is a port of LCC to make Win32 code (which you must already know :) ).

Although there is no INT support (IRQ handler, or call real mode int) there is always a work around. Say you wanted to hook a INT to tell when to play music, well in LCC-WIN32 you just use Windows Media Functions.

Other then that, porting console mode (text mode) applications to LCC-WIN32 is really quite simple. In fact, most GNU software will port to LCC- WIN32 quite quickly.

Compiler Tools


In LCC-WIN32 you would type 'LCC -c test.c' to compile test.c to an object. In DJGPP it's just as easy. 'GCC -c test.c' is all you need.

Linking Libraries

In LCC-WIN32 to link in a library you would type 'LCCLNK test.obj test2.lib' And that would like test.obj and test2.lib together (along with the standard c lib) to make a program. In DJGPP it's quite similar. However there is slight difference. (Read the watcom section to understand).

Making Libs

Please read the watcom section.


in LCC-WIN32 you would type 'LCC -O test.c' to compile and optimize test.c. Probably the closest match is 'GCC -O2 test.c' however DJGPP optimizes a lot better then LCC-WIN32, the -O2 generally doesn't make the code any bigger, and in fact makes it smaller most of the time.

This sums up this chapter. I wish to make ammenmants to this section, such as better code snipets, and more detailed explanations. If you have questions about any area presented here please email me.

This section was provided by Tom St Denis

Email questions or comments to: tomstdenis@goplay.com

  webmaster     delorie software   privacy  
  Copyright 1998   by DJ Delorie     Updated Sep 1998