Re: [POC][RFC][PATCH v2] sched: Extended Scheduler Time Slice

[Steven Rostedt]

On Thu, 26 Oct 2023 12:59:44 +0200
Peter Zijlstra <peterz@xxxxxxxxxxxxx> wrote:

> On Wed, Oct 25, 2023 at 11:54:13PM -0400, Steven Rostedt wrote:
> 
> >  static void extend(void)
> >  {
> > 	rseq_map.cr_flags = 1;
> >  }
> > 
> >  static void unextend(void)
> >  {
> > 	unsigned long prev;
> > 
> > 	prev = xchg(&rseq_map.cr_flags, 0);  
> 
> So you complain about overhead and then you add one of the most
> expensive ops possible here? xchg has an implicit LOCK prefix and you
> really don't need LOCK prefix here.

Peter, this is the user space side, where I cut and pasted the code from
the file I attached.

That has:

static inline unsigned long
xchg(volatile unsigned *ptr, unsigned new)
{
        unsigned ret = new;

	asm volatile("xchg %b0,%1"
		     : "+r"(ret), "+m"(*(ptr))
		     : : "memory");
        return ret;
}

-- Steve


> 
> > 	if (prev & 2) {
> > 		tracefs_printf(NULL, "Yield!\n");
> > 		sched_yield();
> > 	}
> >  }