🚀 go-pugleaf

Hi Rob,
the problem with Sleep( 1) on Microsoft Windows systems is, that the minimum
sleep time is a time slice ( and that is 10/15 ms, depending on the
processor type). On my current tests ( which I think are different from my
test years before) I recognized that a sleep bet 1 and 15 causes in real
sleeptimes of 15 - 18 ms, sleeps between 16 and 20 causes in real times of
around 32 ms.
That is the reason for looking another type of a sleep function.
Detlef

"Rob" <nospam@nonexistant.com> schrieb im Newsbeitrag
news:402d421b@newsgroups.borland.com...
> "Malcolm Smith" <mjfreelancing@optusnet.com.au> wrote in message
> news:402b6ed2$1@newsgroups.borland.com...
> > I made an observation a long time ago that Sleep(0) causes real problems
> in
> > tight loops.  Change it to Sleep(1) - all problems go away.
> >
>
> The reason that Sleep(0) can cause problems in tight loops is
> that all it does is yield the processor to other threads in priority
> order.   Higher priority threads will often get CPU resources,
> with or without the Sleep(0) as your thread will be preempted.
> Threads at equal priority will get some access to CPU
> resources, but will often be starved of CPU resources
> (eg other user apps will be starved by one that hogs resources).
> Lower priority threads are very likely to be starved of
> CPU resources (although the scheduler sometimes does
> things to address that, it often elects not to).
>
> Changing to Sleep(1) simply means that you slow down
> your loops, so that other apps have a larger timeslice in
> which to run and in which your tight loop is guaranteed
> not to.  The flip side of that is that your "tight loop" will
> take longer to complete.
>
>
>