[igt-dev] [PATCH i-g-t] igt/perf_pmu: Flush to idle after hang

[Tvrtko Ursulin]

On 25/05/2018 16:14, Chris Wilson wrote:
> We may not idle immediately after a hang, and indeed may send a pulse
> down the pipeline periodically to become idle. Rather than make a flimsy
> assumption about how long we need to sleep before the system idles,
> wait for the system to declare itself idle; flushing it to idle in the
> process!
> 
> Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk>
> Cc: Tvrtko Ursulin <tvrtko.ursulin at linux.intel.com>
> ---
>   tests/perf_pmu.c | 11 ++---------
>   1 file changed, 2 insertions(+), 9 deletions(-)
> 
> diff --git a/tests/perf_pmu.c b/tests/perf_pmu.c
> index 590e6526b..9af192dd8 100644
> --- a/tests/perf_pmu.c
> +++ b/tests/perf_pmu.c
> @@ -281,16 +281,9 @@ single(int gem_fd, const struct intel_execution_engine2 *e, unsigned int flags)
>   
>   	/* Check for idle after hang. */
>   	if (flags & FLAG_HANG) {
> -		/* Sleep for a bit for reset unwind to settle. */
> -		usleep(500e3);
> -		/*
> -		 * Ensure batch was executing before reset, meaning it must be
> -		 * idle by now. Unless it did not even manage to start before we
> -		 * triggered the reset, in which case the idleness check below
> -		 * might fail. The latter is very unlikely since there are two
> -		 * sleeps during which it had an opportunity to start.
> -		 */
> +		gem_quiescent_gpu(gem_fd);
>   		igt_assert(!gem_bo_busy(gem_fd, spin->handle));
> +
>   		val = pmu_read_single(fd);
>   		slept = measured_usleep(batch_duration_ns / 1000);
>   		val = pmu_read_single(fd) - val;
> 

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

Regards,

Tvrtko