Excerpted from message dated 03-18-97, Paul Edwards to Henk den Adel:

HdA> I would appreciate if anyone could explain me why the stack is not 
HdA> automatically chosen to be sufficiently large, or better how to

PE>Much easier to allow the user to specify the stack size!

HdA> order the compiler to choose the right stack size. I'm using ibm's C 
HdA> compiler, version 2.01.

Hi Henk--

    PMFJI at this late date, but I apparently missed the beginning of
this thread.  If you are using a good compiler designed for OS/2 2.x (or
later), it doesn't cost anything to over-specify stack size.  Stack will
be loaded into "real" memory only as needed, one 4K page at a time, as
long as there is still any committed stack address space.  So it is a
false economy (under OS/2) to try to save memory by minimizing committed
(specified) stack space.

    I don't program in C any more, but I have ILINK (the new IBM 32-bit
linker) set up for 2MB stack size for any PL/I program I write.  (I
suppose I could run out of memory-address space for that process, but
I've never gotten anywhere near it!)  An example: one of my little PL/I
multitasking utilities has a 2 MB stack specified (committed address
space) for the main thread, and a 1 MB stack for the secondary.
According to OS20MEMU, the program only occupies 216 KB of real memory
after the secondary thread kicks in.  The rest of the committed stack
space is just sitting around in address space, just in case.

    However, you have to be sure your compiler takes advantage of the
OS/2 memory-management system this way.  You can test your compiler:
Check out what happens to "in-memory" usage as you increase the
committed stack address space on a working program.  My PL/I for OS/2
compiler does so, and I am fairly certain the IBM C/C++ does, too (they
have the same back end).  I know that Borland's OS/2 compilers don't.  I
can't speak for any of the others.

    Regards,

          --Murray

___
 * MR/2 2.25 #120 * Watching for speed bumps on the Information Highwy

--- Maximus/2 2.02