[Intel-gfx] [PATCH] drm/i915: Avoid waitboosting on the active request

[Chris Wilson]

Watching a light workload on Baytrail (running glxgears and a 1080p
decode), instead of the system remaining at low frequency, the glxgears
would regularly trigger waitboosting after which it would have to spend
a few seconds throttling back down. In this case, the waitboosting is
counter productive as the minimal wait for glxgears doesn't prevent it
from functioning correctly and delivering frames on time. In this case,
glxgears happens to almost always be waiting on the current request,
which we already expect to complete quickly (see i915_spin_request) and
so avoiding the waitboost on the active request and spinning instead
provides the best latency without overcommitting to upclocking.
However, if the system falls behind we still force the waitboost.
Similarly, we will also trigger upclocking if we detect the system is
not delivering frames on time - again using a mechanism that tries to
detect a miss and not preemptively upclock.

Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen at linux.intel.com>
Cc: Tvrtko Ursulin <tvrtko.ursulin at intel.com>
Cc: Radoslaw Szwichtenberg <radoslaw.szwichtenberg at intel.com>
---
 drivers/gpu/drm/i915/i915_gem.c         |  7 +++----
 drivers/gpu/drm/i915/i915_gem_request.h | 13 +++++++++++++
 2 files changed, 16 insertions(+), 4 deletions(-)

diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index 1135a77b383a..b3517676b417 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -369,7 +369,8 @@ i915_gem_object_wait_fence(struct dma_fence *fence,
 	if (i915_gem_request_completed(rq))
 		goto out;
 
-	/* This client is about to stall waiting for the GPU. In many cases
+	/*
+	 * This client is about to stall waiting for the GPU. In many cases
 	 * this is undesirable and limits the throughput of the system, as
 	 * many clients cannot continue processing user input/output whilst
 	 * blocked. RPS autotuning may take tens of milliseconds to respond
@@ -384,11 +385,9 @@ i915_gem_object_wait_fence(struct dma_fence *fence,
 	 * forcing the clocks too high for the whole system, we only allow
 	 * each client to waitboost once in a busy period.
 	 */
-	if (rps_client) {
+	if (rps_client && !i915_gem_request_started(rq)) {
 		if (INTEL_GEN(rq->i915) >= 6)
 			gen6_rps_boost(rq, rps_client);
-		else
-			rps_client = NULL;
 	}
 
 	timeout = i915_wait_request(rq, flags, timeout);
diff --git a/drivers/gpu/drm/i915/i915_gem_request.h b/drivers/gpu/drm/i915/i915_gem_request.h
index 6c607f8dbf92..2236e9188c5c 100644
--- a/drivers/gpu/drm/i915/i915_gem_request.h
+++ b/drivers/gpu/drm/i915/i915_gem_request.h
@@ -329,6 +329,19 @@ i915_gem_request_completed(const struct drm_i915_gem_request *req)
 	return __i915_gem_request_completed(req, seqno);
 }
 
+static inline bool
+i915_gem_request_started(const struct drm_i915_gem_request *req)
+{
+	u32 seqno;
+
+	seqno = i915_gem_request_global_seqno(req);
+	if (!seqno)
+		return false;
+
+	return i915_seqno_passed(intel_engine_get_seqno(req->engine),
+				 seqno - 1);
+}
+
 static inline bool i915_priotree_signaled(const struct i915_priotree *pt)
 {
 	const struct drm_i915_gem_request *rq =
-- 
2.15.1