[Intel-gfx] [PATCH 14/19] drm/i915: Reconstruct active state on starting busy-stats

Tvrtko Ursulin tvrtko.ursulin at linux.intel.com
Mon Jan 8 14:18:53 UTC 2018


On 02/01/2018 15:12, Chris Wilson wrote:
> We have a hole in our busy-stat accounting if the pmu is enabled during
> a long running batch, the pmu will not start accumulating busy-time
> until the next context switch. This then fails tests that are only
> sampling a single batch.

Oops, something in recent perf_pmu rework uncovered this?

> Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk>
> Cc: Tvrtko Ursulin <tvrtko.ursulin at intel.com>
> ---
>   drivers/gpu/drm/i915/intel_engine_cs.c | 9 ++++++++-
>   1 file changed, 8 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/gpu/drm/i915/intel_engine_cs.c b/drivers/gpu/drm/i915/intel_engine_cs.c
> index ebdcbcbacb3c..c2f01ff96ce1 100644
> --- a/drivers/gpu/drm/i915/intel_engine_cs.c
> +++ b/drivers/gpu/drm/i915/intel_engine_cs.c
> @@ -1946,8 +1946,15 @@ int intel_enable_engine_stats(struct intel_engine_cs *engine)
>   	spin_lock_irqsave(&engine->stats.lock, flags);
>   	if (engine->stats.enabled == ~0)
>   		goto busy;
> -	if (engine->stats.enabled++ == 0)
> +	if (engine->stats.enabled++ == 0) {
>   		engine->stats.enabled_at = ktime_get();
> +
> +		/* XXX submission method oblivious */

You mean once busy stats support for the GuC gets added?

> +		engine->stats.active = port_count(&engine->execlists.port[1]);
> +		engine->stats.active += port_count(&engine->execlists.port[0]);

Hm, wouldn't this have the potential to over-account the active counter 
resulting in permanent 100%? Port count is [0, 2], where 2 is 
re-submission of port 0, while engine->stats.active is [0, 2], where 
this 2 is number of ports. In other words I think the correct method 
would be:

if (port_count(0))
	active++;
if (port_count(1))
	active++;

Have to yet think how to handle GuC for this hole.

Regards,

Tvrtko

> +		if (engine->stats.active)
> +			engine->stats.start = engine->stats.enabled_at;
> +	}
>   	spin_unlock_irqrestore(&engine->stats.lock, flags);
>   
>   	return 0;
> 


More information about the Intel-gfx mailing list