On Monday, 1994-11-07  Magnus Eriksson wrote to All about "Assembler
in OS/2" as follows:

ME> Is there something special you must think of when you code 
ME> in ASM for OS/2 (for multitasking purposes & PM) ?

Hi Magnus,

When coding for command-line applications, there is very little
difference in program design from DOS, unless you want to exploit
multiple threads - which are advantageous in many situations. However,
the code design is a lot more straightforward if you use the 0:32 flat
memory model instead of the 16:16 segmented model. No more fiddling
around with segment selectors every 64K when handling large chunks of
data; just access whatever field you need. But when you need to call a
module that uses 16:16 addressing, any pointers you pass must be
"thunked" down from 0:32 to 16:16 and any you receive back must be
"thunked" up to 0:32. The DosFlatToSel() and DosSelToFlat() API calls
can be used for pointer thunking or you can use inline code.

For PM programming, the cardinal rule is: DON'T HOG THE MESSAGE
QUEUE!! Whenever your program has any large volume of work to do (more
than 100 milliseconds) then a background thread should do that work
and the foreground thread should release the message queue. The
background thread can post a message to the foreground thread when the
chore is complete.

The most significant difference that applies exclusively to assembler
coding is no more INTs. The system services are supplied by API calls,
not by INT instructions branching into the operating system. You will
also find addressing bugs easier to find because of the protected mode
execution being used for all OS/2 applications.

Regards

Dave

 * KWQ/2 1.2g * "Well, Mr. Gates, there is a reason suppositories
taste bad"
--- Maximus/2 2.02