We dropped calling process_csb prior to handling direct submission in
order to avoid the nesting of spinlocks and lift process_csb() and the
majority of the tasklet out of irq-off. However, we do want to avoid
ksoftirqd latency in the fast path, so try and pull the interrupt-bh
local to direct submission if we can acquire the tasklet's lock.
v2: Tweak the balance to avoid over submitting lite-restores
Signed-off-by: Chris Wilson <chris@xxxxxxxxxxxxxxxxxx>
Cc: Francisco Jerez <currojerez@xxxxxxxxxx>
Cc: Tvrtko Ursulin <tvrtko.ursulin@xxxxxxxxxxxxxxx>
---
drivers/gpu/drm/i915/gt/intel_lrc.c | 40 +++++++++++++++++++++--------
1 file changed, 29 insertions(+), 11 deletions(-)
diff --git a/drivers/gpu/drm/i915/gt/intel_lrc.c b/drivers/gpu/drm/i915/gt/intel_lrc.c
index f09dd87324b9..0d9c6ea4adaa 100644
--- a/drivers/gpu/drm/i915/gt/intel_lrc.c
+++ b/drivers/gpu/drm/i915/gt/intel_lrc.c
@@ -2884,29 +2884,47 @@ static void queue_request(struct intel_engine_cs *engine,
set_bit(I915_FENCE_FLAG_PQUEUE, &rq->fence.flags);
}
-static void __submit_queue_imm(struct intel_engine_cs *engine)
+static bool pending_csb(const struct intel_engine_execlists *el)
{
- struct intel_engine_execlists * const execlists = &engine->execlists;
-
- if (reset_in_progress(execlists))
- return; /* defer until we restart the engine following reset */
-
- if (execlists->tasklet.func == execlists_submission_tasklet)
- __execlists_submission_tasklet(engine);
- else
- tasklet_hi_schedule(&execlists->tasklet);
+ return READ_ONCE(*el->csb_write) != READ_ONCE(el->csb_head);
}
static void submit_queue(struct intel_engine_cs *engine,
const struct i915_request *rq)
{
struct intel_engine_execlists *execlists = &engine->execlists;
+ struct i915_request *inflight;
if (rq_prio(rq) <= execlists->queue_priority_hint)
return;
+ if (reset_in_progress(execlists))
+ return; /* defer until we restart the engine following reset */
+
+ /* Hopefully we clear execlists->pending[] to let us through */
+ if (execlists->pending[0] && tasklet_trylock(&execlists->tasklet)) {
+ process_csb(engine);
+ tasklet_unlock(&execlists->tasklet);
+ }
+
+ /*
+ * Suppress immediate lite-restores, leave that to the tasklet.
+ *
+ * However, we leave the queue_priority_hint unset so that if we do
+ * submit a second context, we push that into ELSP[1] immediately.
+ */
+ inflight = execlists_active(&engine->execlists);
+ if (inflight && inflight->context == rq->context)
+ return;
+
execlists->queue_priority_hint = rq_prio(rq);
- __submit_queue_imm(engine);
+ __execlists_submission_tasklet(engine);
+
+ /* Try and pull an interrupt-bh queued on another CPU to here */
+ if (pending_csb(execlists) && tasklet_trylock(&execlists->tasklet)) {
+ process_csb(engine);
+ tasklet_unlock(&execlists->tasklet);
+ }
}
static bool ancestor_on_hold(const struct intel_engine_cs *engine,
--
2.20.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@xxxxxxxxxxxxxxxxxxxxx
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
