[Intel-gfx] [PATCH igt 1/3] igt/perf_pmu: Reduce arbitrary delays

[Tvrtko Ursulin]

On 23/11/2017 08:22, Chris Wilson wrote:
> gem_quiescent_gpu() is supposed to ensure that the HW is idle, and in
> the process kick the GPU into rc6, so we should not need a long delay
> afterwards to ensure that we are indeed in rc6. We do however need a
> small delay in order to be sure that rc6 cycle counter has started and
> stopped.
> 
> v2: Apply to rc6p as well.
> v3: The longest rc6 timeout (before the HW kicks in and enables rc6 on
> an idle GPU) is 50ms, so make sure that at least that time has passed
> since we were busy.
> 
> Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk>
> Cc: Tvrtko Ursulin <tvrtko.ursulin at intel.com>
> ---
>   tests/perf_pmu.c | 6 ++++--
>   1 file changed, 4 insertions(+), 2 deletions(-)
> 
> diff --git a/tests/perf_pmu.c b/tests/perf_pmu.c
> index ef5de658..5118c023 100644
> --- a/tests/perf_pmu.c
> +++ b/tests/perf_pmu.c
> @@ -1003,7 +1003,7 @@ test_rc6(int gem_fd)
>   	fd = open_pmu(I915_PMU_RC6_RESIDENCY);
>   
>   	gem_quiescent_gpu(gem_fd);
> -	usleep(1e6);
> +	usleep(100e3); /* wait for the rc6 cycle counter to kick in */
>   
>   	/* Go idle and check full RC6. */
>   	prev = pmu_read_single(fd);
> @@ -1015,6 +1015,7 @@ test_rc6(int gem_fd)
>   	/* Wake up device and check no RC6. */
>   	fw = igt_open_forcewake_handle(gem_fd);
>   	igt_assert(fw >= 0);
> +	usleep(1e3); /* wait for the rc6 cycle counter to stop ticking */
>   
>   	prev = pmu_read_single(fd);
>   	usleep(duration_ns / 1000);
> @@ -1047,7 +1048,7 @@ test_rc6p(int gem_fd)
>   	igt_require(num_pmu == 3);
>   
>   	gem_quiescent_gpu(gem_fd);
> -	usleep(1e6);
> +	usleep(100e3); /* wait for the rc6 cycle counter to kick in */
>   
>   	/* Go idle and check full RC6. */
>   	pmu_read_multi(fd, num_pmu, prev);
> @@ -1060,6 +1061,7 @@ test_rc6p(int gem_fd)
>   	/* Wake up device and check no RC6. */
>   	fw = igt_open_forcewake_handle(gem_fd);
>   	igt_assert(fw >= 0);
> +	usleep(1e3); /* wait for the rc6 cycle counter to stop ticking */
>   
>   	pmu_read_multi(fd, num_pmu, prev);
>   	usleep(duration_ns / 1000);
> 

As we talked yesterday I added the sleep thinking I am letting the GPU 
go to RC6, but I missed the fact gem_queiscent_gpu already does that. So 
as you explain here it is actually about hardware behaviour.

Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin at intel.com>

Regards,

Tvrtko