[haiku-development] Re: Res: Scheduler algorithm improvements

  • From: "Axel Dörfler" <axeld@xxxxxxxxxxxxxxxx>
  • To: haiku-development@xxxxxxxxxxxxx
  • Date: Sat, 07 Nov 2009 17:55:37 +0100

Fredrik Holmqvist <fredrik.holmqvist@xxxxxxxxx> wrote:
> There is a timing-test app someone wrote that showed quite awful
> timing issues with threads during some tests. I ran it on my system
> yesterday and it still shows really bad results. Can't find the thread
> where it was discussed though. Btw is there a way to profile the
> system now?

Use the "scheduling_recorder <output> <command>", and the 
"DebugAnalyzer" on its output.
You have to build both manually, though.

> I'm not really sure what to do else except hunting sections
> with interrupts disabled and proof-reading mediakit code.

For testing I had hacked a panic into 
enable_interrupts()/restore_interrupts() when interrupts were turned 
off for more than x usecs. That only worked for single CPU machines, 
that's why I didn't commit it; I may rework it to support SMP as well, 
or clean it up at least.

Bye,
   Axel.


Other related posts: