[Intel-gfx] [PATCH] drm/i915: Skip execlists_dequeue() early if the list is empty

[Tvrtko Ursulin]

On 16/03/2017 14:53, Chris Wilson wrote:
> Do an early read of the execlists' queue before we take the spinlock and
> start checking. This is safe as the first writer to the execlists queue
> will cause the tasklet to be run again after a memory barrier.

Which one is the memory barrier, tasklet_hi_schedule? Wouldn't we need 
an explicit one after setting engine->execlist_first in 
execlists_submit_request now?

Regards,

Tvrtko

> Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk>
> Cc: Tvrtko Ursulin <tvrtko.ursulin at linux.intel.com>
> Cc: Joonas Lahtinen <joonas.lahtinen at linux.intel.com>
> Cc: Michał Winiarski <michal.winiarski at intel.com>
> ---
>
> With the plan to kick the tasklet before it is ready, eliminating a few
> of the heavier checks seems sensible.
> -Chris
>
> ---
>  drivers/gpu/drm/i915/intel_lrc.c | 3 +++
>  1 file changed, 3 insertions(+)
>
> diff --git a/drivers/gpu/drm/i915/intel_lrc.c b/drivers/gpu/drm/i915/intel_lrc.c
> index 89f38e7def9f..17342bc939ad 100644
> --- a/drivers/gpu/drm/i915/intel_lrc.c
> +++ b/drivers/gpu/drm/i915/intel_lrc.c
> @@ -402,6 +402,9 @@ static void execlists_dequeue(struct intel_engine_cs *engine)
>  	struct rb_node *rb;
>  	bool submit = false;
>
> +	if (!READ_ONCE(engine->execlist_first))
> +		return;
> +
>  	last = port->request;
>  	if (last)
>  		/* WaIdleLiteRestore:bdw,skl
>