Re: [PATCH v4 3/3] qspinlock, x86: Add x86 specific optimization for 2 contending tasks

[Waiman Long]

On 02/21/2014 12:28 PM, Peter Zijlstra wrote:
> On Mon, Feb 17, 2014 at 03:41:24PM -0500, Waiman Long wrote:
> > +	struct {
> > +		u8  lock;	/* Lock bit	*/
> > +		u8  wait;	/* Waiting bit	*/
> > +		u16 qcode;	/* Queue code	*/
> > +	};
> 16 bit code would result in 14 bits for the CPU number, that's only 16k,
> I think SGI actually had a machine with that many CPUs in.

If NR_CPUS >= 16k, the 2-task optimized code path will be disabled. 
However, with that many CPUs, it is high spinlock contention behavior 
that is more important than a bit better performance at low contention 
level.

-Longman
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/