On 04/02/2019 12:19, Chris Wilson wrote:
Quoting Tvrtko Ursulin (2019-02-04 12:08:50)
On 04/02/2019 10:18, Chris Wilson wrote:
+static int effective_prio(const struct i915_request *rq)
+{
+ /* Restrict mere WAIT boosts from triggering preemption */
+ return rq_prio(rq) | __NO_PREEMPTION;
+}
I suggest adding i915_request_effective_prio to i915_request.h - it is
verbose but avoids two implementation.
Too verbose... And it may differ depending on backend details...
We don't even need to or in no-preemption until later...
Hmm.. I would hope it wouldn't depend on the backend. We should at least
I think try to make things decoupled at this level.
I'm speculating about what the long term interface will be. If they can
only handle static priorities on a context level and take all
dependencies as semaphores, guc submission is a mere conduit and very
hands off.
Point taken, the force of GuC influencing the i915 design will be too
strong.
Regards,
Tvrtko
_______________________________________________
Intel-gfx mailing list
Intel-gfx@xxxxxxxxxxxxxxxxxxxxx
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
