Re: [PATCH next 4/5] locking/osq_lock: Optimise per-cpu data accesses.

[Ingo Molnar <mingo@kernel.org>]

* David Laight <David.Laight@ACULAB.COM> wrote:

>  bool osq_lock(struct optimistic_spin_queue *lock)
>  {
> -	struct optimistic_spin_node *node = this_cpu_ptr(&osq_node);
> +	struct optimistic_spin_node *node = raw_cpu_read(osq_node.self);
>  	struct optimistic_spin_node *prev, *next;
>  	int old;
>  
> -	if (unlikely(node->cpu == OSQ_UNLOCKED_VAL))
> -		node->cpu = encode_cpu(smp_processor_id());
> +	if (unlikely(!node)) {
> +		int cpu = encode_cpu(smp_processor_id());
> +		node = decode_cpu(cpu);
> +		node->self = node;
> +		node->cpu = cpu;

This whole initialization sequence is suboptimal and needs to be 
cleaned up first: the node->cpu field is constant once initialized, so 
it should be initialized from appropriate init methods, not runtime in 
osq_lock(), right?

Eliminating that initialization branch is a useful micro-optimization 
in itself for the hot path.

Thanks,

	Ingo