We're only requesting preemption for requests that have high enough
priority (above threshold). Currently we're also ignoring requests that
have dependencies on different engines.
Signed-off-by: Michał Winiarski <michal.winiarski@xxxxxxxxx>
---
drivers/gpu/drm/i915/i915_guc_submission.c | 3 ++
drivers/gpu/drm/i915/intel_lrc.c | 81 +++++++++++++++++++++++++++++-
2 files changed, 83 insertions(+), 1 deletion(-)
diff --git a/drivers/gpu/drm/i915/i915_guc_submission.c b/drivers/gpu/drm/i915/i915_guc_submission.c
index feccd65..6d9431d 100644
--- a/drivers/gpu/drm/i915/i915_guc_submission.c
+++ b/drivers/gpu/drm/i915/i915_guc_submission.c
@@ -491,6 +491,9 @@ static int guc_ring_doorbell(struct i915_guc_client *client)
static int i915_guc_preempt_noop(struct intel_engine_cs *engine)
{
+ engine->preempt_requested = false;
+ intel_write_status_page(engine, I915_GEM_HWS_PREEMPT_INDEX, 0);
+
return 0;
}
diff --git a/drivers/gpu/drm/i915/intel_lrc.c b/drivers/gpu/drm/i915/intel_lrc.c
index d4cfaa1..869b96e 100644
--- a/drivers/gpu/drm/i915/intel_lrc.c
+++ b/drivers/gpu/drm/i915/intel_lrc.c
@@ -674,16 +674,93 @@ pt_lock_engine(struct i915_priotree *pt, struct intel_engine_cs *locked)
return engine;
}
+#define EXECLISTS_PREEMPT_THRESHOLD 512
+
+static void __execlists_try_preempt(struct intel_engine_cs *engine,
+ int prio)
+{
+ struct drm_i915_gem_request *rq;
+ int highest_prio = INT_MIN;
+ int ret;
+
+ spin_lock_irq(&engine->timeline->lock);
+
+ /* Engine is idle */
+ if (execlists_elsp_idle(engine))
+ goto out_unlock;
+
+ if (engine->preempt_requested)
+ goto out_unlock;
+
+ list_for_each_entry_reverse(rq, &engine->timeline->requests, link) {
+ if (!i915_gem_request_completed(rq)) {
+ highest_prio = (rq->priotree.priority > highest_prio) ?
+ rq->priotree.priority : highest_prio;
+ } else
+ break;
+ }
+
+ /* Bail out if our priority is lower than any of the inflight requests
+ * (also if there are none requests) */
+ if (highest_prio == INT_MIN || prio <= highest_prio)
+ goto out_unlock;
+
+ engine->preempt_requested = true;
+
+ spin_unlock_irq(&engine->timeline->lock);
+
+ ret = engine->preempt(engine);
+ if (ret) {
+ spin_lock_irq(&engine->timeline->lock);
+ engine->preempt_requested = false;
+ spin_unlock_irq(&engine->timeline->lock);
+ }
+
+ return;
+
+out_unlock:
+ spin_unlock_irq(&engine->timeline->lock);
+}
+
+static void execlists_try_preempt(struct intel_engine_cs *engine,
+ int prio,
+ unsigned long *engines_bumped)
+{
+ int num_engines_bumped = bitmap_weight(engines_bumped,
+ I915_NUM_ENGINES);
+
+ /* Preemption is disabled */
+ if (!engine->preempt)
+ return;
+
+ /* We're not a high priority request */
+ if (prio < EXECLISTS_PREEMPT_THRESHOLD)
+ return;
+
+ /* We have dependencies on many engines */
+ if (num_engines_bumped > 1)
+ return;
+
+ /* We have dependency on a single engine - but it's not our engine */
+ if (num_engines_bumped == 1 && !test_bit(engine->id, engines_bumped))
+ return;
+
+ __execlists_try_preempt(engine, prio);
+}
+
static void execlists_schedule(struct drm_i915_gem_request *request, int prio)
{
struct intel_engine_cs *engine = NULL;
struct i915_dependency *dep, *p;
struct i915_dependency stack;
LIST_HEAD(dfs);
+ DECLARE_BITMAP(engine_bumped, I915_NUM_ENGINES);
if (prio <= READ_ONCE(request->priotree.priority))
return;
+ bitmap_zero(engine_bumped, I915_NUM_ENGINES);
+
/* Need BKL in order to use the temporary link inside i915_dependency */
lockdep_assert_held(&request->i915->drm.struct_mutex);
@@ -719,6 +796,7 @@ static void execlists_schedule(struct drm_i915_gem_request *request, int prio)
continue;
engine = pt_lock_engine(pt, engine);
+ __set_bit(engine->id, engine_bumped);
/* If it is not already in the rbtree, we can update the
* priority inplace and skip over it (and its dependencies)
@@ -737,6 +815,7 @@ static void execlists_schedule(struct drm_i915_gem_request *request, int prio)
INIT_LIST_HEAD(&dep->dfs_link);
engine = pt_lock_engine(pt, engine);
+ __set_bit(engine->id, engine_bumped);
if (prio <= pt->priority)
continue;
@@ -752,7 +831,7 @@ static void execlists_schedule(struct drm_i915_gem_request *request, int prio)
if (engine)
spin_unlock_irq(&engine->timeline->lock);
- /* XXX Do we need to preempt to make room for us and our deps? */
+ execlists_try_preempt(engine, prio, engine_bumped);
}
static int execlists_context_pin(struct intel_engine_cs *engine,
--
2.9.3
_______________________________________________
Intel-gfx mailing list
Intel-gfx@xxxxxxxxxxxxxxxxxxxxx
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
