Quoting Tvrtko Ursulin (2018-01-09 16:16:20)
> From: Tvrtko Ursulin <tvrtko.ursu...@intel.com>
>
> Make sure busyness is correctly reported when PMU is enabled after the
> engine is already busy with a single long batch.
>
> Signed-off-by: Tvrtko Ursulin <tvrtko.ursu...@intel.com>
> ---
> tests/perf_pmu.c | 42 ++++++++++++++++++++++++++++++++++++++++++
> 1 file changed, 42 insertions(+)
>
> diff --git a/tests/perf_pmu.c b/tests/perf_pmu.c
> index 45e2f6148453..e1f449d48808 100644
> --- a/tests/perf_pmu.c
> +++ b/tests/perf_pmu.c
> @@ -157,6 +157,41 @@ single(int gem_fd, const struct intel_execution_engine2
> *e, bool busy)
> gem_quiescent_gpu(gem_fd);
> }
>
> +static void
> +busy_start(int gem_fd, const struct intel_execution_engine2 *e)
> +{
> + unsigned long slept;
> + igt_spin_t *spin;
> + uint64_t val;
> + int fd;
> +
> + /*
> + * Defeat the busy stats delayed disable, we need to guarantee we are
> + * the first user.
> + */
> + if (gem_has_execlists(gem_fd))
> + sleep(2);
I don't have a better idea than sleep, but I don't like tying this to
execlists. Make the sleep unconditional for now. Is there anyway we can
export the knowledge of the implementation through the perf api?
Different counters, or now different attrs for different busy-stats?
> +
> + spin = __igt_spin_batch_new(gem_fd, 0, e2ring(gem_fd, e), 0);
> +
> + /*
> + * Sleep for a bit after making the engine busy to make sure the PMU
> + * gets enabled when the batch is already running.
> + */
> + usleep(500000);
Just a request for 500e3.
Reviewed-by: Chris Wilson <ch...@chris-wilson.co.uk>
-Chris
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
