Re: [AD] Test patch

[Peter Wang <tjaden@xxxxxxxxxx>]

Chris wrote:

> Peter Wang wrote:
>
> > I think automatically switching to another thread would really hurt 
> > performance if all mutexes worked that way (and this is surely the 
> > domain of the thread system and not Xlib).
>
>
> I wouldn't be so sure. The thread that's blocked waiting would pretty 
> much be asleep so the main program is getting a boost while it's 
> holding the lock. Once it unlocks, it should pass control to the 
> waiting thread since it left prematurely before.


Context switches are not free, though.  The more switching, the lower 
the system throughput.  It increases responsiveness, of course.

> > Maybe we could try something using condition variables?
>
>
> That looks ugly... I'd be content just warning to yield or rest every 
> loop iteration if you're doing a lot of operations on the screen/video 
> bitmaps.


I did a mini test using the condition variables idea and it didn't seem 
to work.  I'll search around for something else.

I don't like the idea of telling the user to yield.  It ought to Just 
Work, even if yielding is a good idea in many cases.

BTW, don't you find that you get rather different numbers in the test 
program with your rest(0) patch?

Peter