🚀 go-pugleaf

In article <1992Dec30.222721.29697@samba.oit.unc.edu> Brandon.Vanevery@launchpad.unc.edu (Brandon Vanevery) writes:
>
>Which compiler is "better" for OS/2 programming?  Why?  Is there even any
>contest?
------
To toss my $0.02 in, I have both compilers and prefer Watcom C 386
9.0. Why? The executables it produces are consistently tighter. For
example, I compiled my fBMP program (available as fbmp.zip in
directory /pub/os2/2.0/graphics on Hobbes) using the following
commands:

	wcl386 /l=os2v2_pm /k65536 /omaxet /zp4 /4r fbmp.c

for Watcom, and

	icc /C+ /O+ /Sp4 fbmp.c
	link386 /a:4 /base:65536 /exepack fbmp,,nul,,fbmp

for C Set/2 (*not* to be confused with C/2; does anyone else find the
distinction confusing?). The results? Watcom produced an executable of
18944 bytes; C Set/2 produced one that was 34609 bytes long. That's a
difference of 15665 bytes, or about 45%! While most program don't show
such a dramatic difference, Watcom executables tend to be 10% - 35%
tighter than their C Set/2 equivalents. And I have NEVER found a
program where Watcom made a bigger EXE than C Set/2.

Why do I keep both compilers around? Some of the huge programs I've
ported from other platforms crash and burn under Watcom, but run under
C Set/2. For instance UMB Scheme. This large interpreter was written
for Unix, but was well-engineered so it's fairly easy to port. I
compiled and linked it FOR DOS using Watcom. The resulting .EXE ran
well on 386s, and in OS/2 DOS boxes (though you have to increase the
amount of DPMI). I changed the flags to compile and link for OS/2. The
resulting .EXE blows up shortly after it starts. I then took the
original source and tried porting to C Set/2. The executable is bigger
than the one that came out of Watcom, but it actually works!

(On the other hand, there's at least *one* program that "ported" under
Watcom and failed under C Set/2. That's POV-Ray Tracer. Of course,
this is cheating, since the version I "ported" was written for Watcom
C 386 for DOS!)

Anyway, this is just my two cents worth.

Raja Thiagarajan
(sthiagar@bronze.ucs.indiana.edu)