[Xitami] SMP Problem (WAS: Xitami 2.4d9/Xitami 2.5b5)

[Dr. Martin R. Hadam]

On Sun, 24 Jun 2001 06:14:01 +1200, Ewen McNeill wrote:

>I was going to say that Xitami didn't use threads, so I didn't see what
>the problem could be, but IIRC there are a few minor uses of threads for
>various auxillary things (and EMX uses threads for a few things).
>(Xitami mostly works as a single process, single thread, and uses
>select() and SMT to figure out what to do next.)

 Well, it's listed as running on two threads here. I'm quite familiar
with the situation that I'm seeing problems that no-one else is
encountering... <g> Please note that it must have been something that
changed after v2.4d6.

>Our development version makes even less use of thread-related things, so
>it's possible the problem might go away with that.  Unfortunately we
>don't curerntly have an OS/2 development system, let alone a SMP-capable
>OS/2 development system available.  However I do have a SMP Linux
>development system, and I might try playing there and see if anything
>breaks.

 whenever you have something to test (assuming that Bernd Hohman had
volunteered to do the compilation for an OS/2 version) I'll be happy to
run it here.

>And it's useful to know that single-threaded Xitami is very stable.

 It's not making it single-threaded. Execmode only restricts the
software from accessing both processors and puts all the workload for
that package on processor #1; effectively making a given software as if
running on a uniprocessor system. From what I learned by following
other SMP-related bugs I came across it could be more of a timing
problem (or the like).

 Martin


 Martin R. Hadam
 Kinderklinik - Medizinische Hochschule
 D-30623 Hannover
 Germany
 Email: Hadam.Martin@MH-Hannover.de