🚀 go-pugleaf

nevin@CS.Arizona.EDU (Nevin ":-]" Liber) wrote:
>In article <adtDpntEF.Epw@netcom.com>, adt@netcom.com (Anthony D.
>Tribelli) wrote:
>
>> Which is why I believe CMT applications require extra care when written,
>> i.e. they are created (written) differently than a PMT application (the
>> contradiction to your earlier "no difference" statement). A threaded CMT
>> application needs a little more care to ensure that some interaction
>> between threads does not cause the app to hog the CPU. 
>
>And PMT threads have to do sychronization (semaphores, etc.) to make sure
>that shared data remains consistent, while CMT threads do not.  It's just
>a tradeoff.

   Worth noting, in your assertion there is no advantage to multi-threading a
CMT app because everything still becomes single-threaded.  However, if you
compare CMT to PMT threads in a multiprocessor environment, then both types of
threads will need a synchronization method if they expect to update common
data.  Thus, your assertion makes no sense.  Consider re-entrant coding with
multiple threads of execution but each thread with its own data(as in more
advanced OSs, threaded apps and such) and you should be able to see what
threads buys you.  In an application, threaded execution on multiple processors
of common or thread specific data will buy your app faster throughput, but, it
must be managed and that's true whether we use a PMT or a CMT scheme.  I do
have one question, have you ever programmed re-entrant, multi-threaded apps or
OS level code?  This stuff has been going on since the late 60s/early 70s so
the methodologies are well understood.  WRT using semaphores to regulate the
access/update of common data sets, most such facilities get rapidly embedded in
to OO class methods or into a series of macros in more primative environments,
pretty much the same as Apple app main event loops.  For my money, given all
the really cool, sophisticated tools available for the MacOS, I find
programming on the Mac more painful than any other environment I've played in.
To be fair, that may be somewhat of a function of familiarity but even C++
presents a litany of little gotchas in the MacOS environment.  I still intend
to stick it out because I like some of the other features and I like the Apple
HI pretty well.  I'd prefer a more robust OS but hopefully that will come with
Copland. 
>-- 
>        Nevin ":-)" Liber       nevin@CS.Arizona.EDU    (520) 293-2799

-- 
Joseph Crowe
jcrowe@isd.tandem.com