Intel-GFX Archive on lore.kernel.org
 help / color / Atom feed
From: Chris Wilson <chris@chris-wilson.co.uk>
To: intel-gfx@lists.freedesktop.org
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Subject: [Intel-gfx] [PATCH 41/57] drm/i915: Move saturated workload detection back to the context
Date: Mon,  1 Feb 2021 08:56:59 +0000
Message-ID: <20210201085715.27435-41-chris@chris-wilson.co.uk> (raw)
In-Reply-To: <20210201085715.27435-1-chris@chris-wilson.co.uk>

When we introduced the saturated workload detection to tell us to back
off from semaphore usage [semaphores have a noticeable impact on
contended bus cycles with the CPU for some heavy workloads], we first
introduced it as a per-context tracker. This allows individual contexts
to try and optimise their own usage, but we found that with the local
tracking and the no-semaphore boosting, the first context to disable
semaphores got a massive priority boost and so would starve the rest and
all new contexts (as they started with semaphores enabled and lower
priority). Hence we moved the saturated workload detection to the
engine, and a consequence had to disable semaphores on virtual engines.

Now that we do not have semaphore priority boosting, and try to fairly
schedule irrespective of semaphore usage, we can move the tracking back
to the context and virtual engines can now utilise the faster inter-engine
synchronisation. If we see that any context fairs to use the semaphore,
because the system is oversubscribed and was busy doing something else
instead of spinning on the semaphore, we disable further usage of
semaphores with that context until it idles again. This should restrict
the semaphores to lightly utilised system where the latency between
requests is more noticeable, and curtail the bus-contention from checking
for signaled semaphores.

References: 44d89409a12e ("drm/i915: Make the semaphore saturation mask global")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
---
 drivers/gpu/drm/i915/gt/intel_context.c           |  3 +++
 drivers/gpu/drm/i915/gt/intel_context_types.h     |  2 ++
 drivers/gpu/drm/i915/gt/intel_engine_pm.c         |  2 --
 drivers/gpu/drm/i915/gt/intel_engine_types.h      |  2 --
 .../gpu/drm/i915/gt/intel_execlists_submission.c  | 15 ---------------
 drivers/gpu/drm/i915/i915_request.c               |  6 +++---
 6 files changed, 8 insertions(+), 22 deletions(-)

diff --git a/drivers/gpu/drm/i915/gt/intel_context.c b/drivers/gpu/drm/i915/gt/intel_context.c
index daf537d1e415..57b6bde2b736 100644
--- a/drivers/gpu/drm/i915/gt/intel_context.c
+++ b/drivers/gpu/drm/i915/gt/intel_context.c
@@ -344,6 +344,9 @@ static int __intel_context_active(struct i915_active *active)
 {
 	struct intel_context *ce = container_of(active, typeof(*ce), active);
 
+	CE_TRACE(ce, "active\n");
+	ce->saturated = 0;
+
 	intel_context_get(ce);
 
 	/* everything should already be activated by intel_context_pre_pin() */
diff --git a/drivers/gpu/drm/i915/gt/intel_context_types.h b/drivers/gpu/drm/i915/gt/intel_context_types.h
index 0ea18c9e2aca..d1a35c3055a7 100644
--- a/drivers/gpu/drm/i915/gt/intel_context_types.h
+++ b/drivers/gpu/drm/i915/gt/intel_context_types.h
@@ -109,6 +109,8 @@ struct intel_context {
 	} lrc;
 	u32 tag; /* cookie passed to HW to track this context on submission */
 
+	intel_engine_mask_t saturated; /* submitting semaphores too late? */
+
 	/** stats: Context GPU engine busyness tracking. */
 	struct intel_context_stats {
 		u64 active;
diff --git a/drivers/gpu/drm/i915/gt/intel_engine_pm.c b/drivers/gpu/drm/i915/gt/intel_engine_pm.c
index ef5064ea54e5..44948abe4bf8 100644
--- a/drivers/gpu/drm/i915/gt/intel_engine_pm.c
+++ b/drivers/gpu/drm/i915/gt/intel_engine_pm.c
@@ -253,8 +253,6 @@ static int __engine_park(struct intel_wakeref *wf)
 	struct intel_engine_cs *engine =
 		container_of(wf, typeof(*engine), wakeref);
 
-	engine->saturated = 0;
-
 	/*
 	 * If one and only one request is completed between pm events,
 	 * we know that we are inside the kernel context and it is
diff --git a/drivers/gpu/drm/i915/gt/intel_engine_types.h b/drivers/gpu/drm/i915/gt/intel_engine_types.h
index dc12cbdfda46..e94c99dee5cb 100644
--- a/drivers/gpu/drm/i915/gt/intel_engine_types.h
+++ b/drivers/gpu/drm/i915/gt/intel_engine_types.h
@@ -303,8 +303,6 @@ struct intel_engine_cs {
 
 	struct intel_context *kernel_context; /* pinned */
 
-	intel_engine_mask_t saturated; /* submitting semaphores too late? */
-
 	struct {
 		struct delayed_work work;
 		struct i915_request *systole;
diff --git a/drivers/gpu/drm/i915/gt/intel_execlists_submission.c b/drivers/gpu/drm/i915/gt/intel_execlists_submission.c
index 9c929688a955..e8f192984e88 100644
--- a/drivers/gpu/drm/i915/gt/intel_execlists_submission.c
+++ b/drivers/gpu/drm/i915/gt/intel_execlists_submission.c
@@ -3413,21 +3413,6 @@ intel_execlists_create_virtual(struct intel_engine_cs **siblings,
 	ve->base.instance = I915_ENGINE_CLASS_INVALID_VIRTUAL;
 	ve->base.uabi_instance = I915_ENGINE_CLASS_INVALID_VIRTUAL;
 
-	/*
-	 * The decision on whether to submit a request using semaphores
-	 * depends on the saturated state of the engine. We only compute
-	 * this during HW submission of the request, and we need for this
-	 * state to be globally applied to all requests being submitted
-	 * to this engine. Virtual engines encompass more than one physical
-	 * engine and so we cannot accurately tell in advance if one of those
-	 * engines is already saturated and so cannot afford to use a semaphore
-	 * and be pessimized in priority for doing so -- if we are the only
-	 * context using semaphores after all other clients have stopped, we
-	 * will be starved on the saturated system. Such a global switch for
-	 * semaphores is less than ideal, but alas is the current compromise.
-	 */
-	ve->base.saturated = ALL_ENGINES;
-
 	snprintf(ve->base.name, sizeof(ve->base.name), "virtual");
 
 	intel_engine_init_execlists(&ve->base);
diff --git a/drivers/gpu/drm/i915/i915_request.c b/drivers/gpu/drm/i915/i915_request.c
index aa12289ea14b..352083889b97 100644
--- a/drivers/gpu/drm/i915/i915_request.c
+++ b/drivers/gpu/drm/i915/i915_request.c
@@ -516,7 +516,7 @@ bool __i915_request_submit(struct i915_request *request)
 	 */
 	if (request->sched.semaphores &&
 	    i915_sw_fence_signaled(&request->semaphore))
-		engine->saturated |= request->sched.semaphores;
+		request->context->saturated |= request->sched.semaphores;
 
 	engine->emit_fini_breadcrumb(request,
 				     request->ring->vaddr + request->postfix);
@@ -977,7 +977,7 @@ already_busywaiting(struct i915_request *rq)
 	 *
 	 * See the are-we-too-late? check in __i915_request_submit().
 	 */
-	return rq->sched.semaphores | READ_ONCE(rq->engine->saturated);
+	return rq->sched.semaphores | READ_ONCE(rq->context->saturated);
 }
 
 static int
@@ -1071,7 +1071,7 @@ emit_semaphore_wait(struct i915_request *to,
 	if (__emit_semaphore_wait(to, from, from->fence.seqno))
 		goto await_fence;
 
-	to->sched.semaphores |= mask;
+	to->sched.semaphores |= mask & ~to->engine->mask;
 	wait = &to->semaphore;
 
 await_fence:
-- 
2.20.1

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

  parent reply index

Thread overview: 103+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-02-01  8:56 [Intel-gfx] [PATCH 01/57] drm/i915/gt: Restrict the GT clock override to just Icelake Chris Wilson
2021-02-01  8:56 ` [Intel-gfx] [PATCH 02/57] drm/i915/selftests: Exercise relative mmio paths to non-privileged registers Chris Wilson
2021-02-01 14:34   ` Mika Kuoppala
2021-02-01  8:56 ` [Intel-gfx] [PATCH 03/57] drm/i915/selftests: Exercise cross-process context isolation Chris Wilson
2021-02-01 16:37   ` Mika Kuoppala
2021-02-01  8:56 ` [Intel-gfx] [PATCH 04/57] drm/i915: Protect against request freeing during cancellation on wedging Chris Wilson
2021-02-02  9:55   ` Mika Kuoppala
2021-02-01  8:56 ` [Intel-gfx] [PATCH 05/57] drm/i915: Take rcu_read_lock for querying fence's driver/timeline names Chris Wilson
2021-02-02 18:33   ` Mika Kuoppala
2021-02-01  8:56 ` [Intel-gfx] [PATCH 06/57] drm/i915/gt: Always flush the submission queue on checking for idle Chris Wilson
2021-02-01  8:56 ` [Intel-gfx] [PATCH 07/57] drm/i915/gt: Move engine setup out of set_default_submission Chris Wilson
2021-02-02 11:57   ` Tvrtko Ursulin
2021-02-01  8:56 ` [Intel-gfx] [PATCH 08/57] drm/i915/gt: Move submission_method into intel_gt Chris Wilson
2021-02-02 12:03   ` Tvrtko Ursulin
2021-02-02 12:18     ` Chris Wilson
2021-02-01  8:56 ` [Intel-gfx] [PATCH 09/57] drm/i915: Replace engine->schedule() with a known request operation Chris Wilson
2021-02-01  8:56 ` [Intel-gfx] [PATCH 10/57] drm/i915: Restructure priority inheritance Chris Wilson
2021-02-01  8:56 ` [Intel-gfx] [PATCH 11/57] drm/i915/selftests: Measure set-priority duration Chris Wilson
2021-02-01  8:56 ` [Intel-gfx] [PATCH 12/57] drm/i915/selftests: Exercise priority inheritance around an engine loop Chris Wilson
2021-02-01  8:56 ` [Intel-gfx] [PATCH 13/57] drm/i915/selftests: Force a rewind if at first we don't succeed Chris Wilson
2021-02-01  8:56 ` [Intel-gfx] [PATCH 14/57] drm/i915: Improve DFS for priority inheritance Chris Wilson
2021-02-01  8:56 ` [Intel-gfx] [PATCH 15/57] drm/i915: Extract request submission from execlists Chris Wilson
2021-02-01  8:56 ` [Intel-gfx] [PATCH 16/57] drm/i915: Extract request rewinding " Chris Wilson
2021-02-02 13:08   ` Tvrtko Ursulin
2021-02-01  8:56 ` [Intel-gfx] [PATCH 17/57] drm/i915: Extract request suspension from the execlists Chris Wilson
2021-02-02 13:15   ` Tvrtko Ursulin
2021-02-02 13:26     ` Chris Wilson
2021-02-02 13:32       ` Tvrtko Ursulin
2021-02-02 13:27     ` Chris Wilson
2021-02-01  8:56 ` [Intel-gfx] [PATCH 18/57] drm/i915: Extract the ability to defer and rerun a request later Chris Wilson
2021-02-02 13:18   ` Tvrtko Ursulin
2021-02-01  8:56 ` [Intel-gfx] [PATCH 19/57] drm/i915: Fix the iterative dfs for defering requests Chris Wilson
2021-02-02 14:10   ` Tvrtko Ursulin
2021-02-01  8:56 ` [Intel-gfx] [PATCH 20/57] drm/i915: Wrap access to intel_engine.active Chris Wilson
2021-02-04 11:07   ` Tvrtko Ursulin
2021-02-04 11:18     ` Chris Wilson
2021-02-04 11:56       ` Chris Wilson
2021-02-04 12:08         ` Tvrtko Ursulin
2021-02-01  8:56 ` [Intel-gfx] [PATCH 21/57] drm/i915: Move common active lists from engine to i915_scheduler Chris Wilson
2021-02-04 11:12   ` Tvrtko Ursulin
2021-02-01  8:56 ` [Intel-gfx] [PATCH 22/57] drm/i915: Move scheduler queue Chris Wilson
2021-02-04 11:19   ` Tvrtko Ursulin
2021-02-04 11:32     ` Chris Wilson
2021-02-04 11:40     ` Chris Wilson
2021-02-01  8:56 ` [Intel-gfx] [PATCH 23/57] drm/i915: Move tasklet from execlists to sched Chris Wilson
2021-02-04 14:06   ` Tvrtko Ursulin
2021-02-01  8:56 ` [Intel-gfx] [PATCH 24/57] drm/i915/gt: Only kick the scheduler on timeslice/preemption change Chris Wilson
2021-02-04 14:09   ` Tvrtko Ursulin
2021-02-04 14:43     ` Chris Wilson
2021-02-01  8:56 ` [Intel-gfx] [PATCH 25/57] drm/i915: Move submit_request to i915_sched_engine Chris Wilson
2021-02-04 14:13   ` Tvrtko Ursulin
2021-02-04 14:45     ` Chris Wilson
2021-02-01  8:56 ` [Intel-gfx] [PATCH 26/57] drm/i915: Move finding the current active request to the scheduler Chris Wilson
2021-02-04 14:30   ` Tvrtko Ursulin
2021-02-04 14:59     ` Chris Wilson
2021-02-01  8:56 ` [Intel-gfx] [PATCH 27/57] drm/i915: Show execlists queues when dumping state Chris Wilson
2021-02-04 15:04   ` Tvrtko Ursulin
2021-02-01  8:56 ` [Intel-gfx] [PATCH 28/57] drm/i915: Wrap i915_request_use_semaphores() Chris Wilson
2021-02-04 15:05   ` Tvrtko Ursulin
2021-02-01  8:56 ` [Intel-gfx] [PATCH 29/57] drm/i915: Move scheduler flags Chris Wilson
2021-02-04 15:14   ` Tvrtko Ursulin
2021-02-04 16:05     ` Chris Wilson
2021-02-01  8:56 ` [Intel-gfx] [PATCH 30/57] drm/i915: Move timeslicing flag to scheduler Chris Wilson
2021-02-04 15:18   ` Tvrtko Ursulin
2021-02-04 16:11     ` Chris Wilson
2021-02-05  9:48       ` Tvrtko Ursulin
2021-02-01  8:56 ` [Intel-gfx] [PATCH 31/57] drm/i915/gt: Declare when we enabled timeslicing Chris Wilson
2021-02-04 15:26   ` Tvrtko Ursulin
2021-02-01  8:56 ` [Intel-gfx] [PATCH 32/57] drm/i915: Move needs-breadcrumb flags to scheduler Chris Wilson
2021-02-04 15:28   ` Tvrtko Ursulin
2021-02-04 16:12     ` Chris Wilson
2021-02-01  8:56 ` [Intel-gfx] [PATCH 33/57] drm/i915: Move busywaiting control to the scheduler Chris Wilson
2021-02-04 15:32   ` Tvrtko Ursulin
2021-02-01  8:56 ` [Intel-gfx] [PATCH 34/57] drm/i915: Move preempt-reset flag " Chris Wilson
2021-02-04 15:34   ` Tvrtko Ursulin
2021-02-01  8:56 ` [Intel-gfx] [PATCH 35/57] drm/i915: Replace priolist rbtree with a skiplist Chris Wilson
2021-02-01  8:56 ` [Intel-gfx] [PATCH 36/57] drm/i915: Wrap cmpxchg64 with try_cmpxchg64() helper Chris Wilson
2021-02-01  8:56 ` [Intel-gfx] [PATCH 37/57] drm/i915: Fair low-latency scheduling Chris Wilson
2021-02-01  8:56 ` [Intel-gfx] [PATCH 38/57] drm/i915/gt: Specify a deadline for the heartbeat Chris Wilson
2021-02-01  8:56 ` [Intel-gfx] [PATCH 39/57] drm/i915: Extend the priority boosting for the display with a deadline Chris Wilson
2021-02-01  8:56 ` [Intel-gfx] [PATCH 40/57] drm/i915/gt: Support virtual engine queues Chris Wilson
2021-02-01  8:56 ` Chris Wilson [this message]
2021-02-01  8:57 ` [Intel-gfx] [PATCH 42/57] drm/i915: Bump default timeslicing quantum to 5ms Chris Wilson
2021-02-01  8:57 ` [Intel-gfx] [PATCH 43/57] drm/i915/gt: Delay taking irqoff for execlists submission Chris Wilson
2021-02-01  8:57 ` [Intel-gfx] [PATCH 44/57] drm/i915/gt: Wrap intel_timeline.has_initial_breadcrumb Chris Wilson
2021-02-01  8:57 ` [Intel-gfx] [PATCH 45/57] drm/i915/gt: Track timeline GGTT offset separately from subpage offset Chris Wilson
2021-02-01  8:57 ` [Intel-gfx] [PATCH 46/57] drm/i915/gt: Add timeline "mode" Chris Wilson
2021-02-01  8:57 ` [Intel-gfx] [PATCH 47/57] drm/i915/gt: Use indices for writing into relative timelines Chris Wilson
2021-02-01  8:57 ` [Intel-gfx] [PATCH 48/57] drm/i915/selftests: Exercise relative timeline modes Chris Wilson
2021-02-01  8:57 ` [Intel-gfx] [PATCH 49/57] drm/i915/gt: Use ppHWSP for unshared non-semaphore related timelines Chris Wilson
2021-02-01  8:57 ` [Intel-gfx] [PATCH 50/57] Restore "drm/i915: drop engine_pin/unpin_breadcrumbs_irq" Chris Wilson
2021-02-01  8:57 ` [Intel-gfx] [PATCH 51/57] drm/i915/gt: Couple tasklet scheduling for all CS interrupts Chris Wilson
2021-02-01  8:57 ` [Intel-gfx] [PATCH 52/57] drm/i915/gt: Support creation of 'internal' rings Chris Wilson
2021-02-01  8:57 ` [Intel-gfx] [PATCH 53/57] drm/i915/gt: Use client timeline address for seqno writes Chris Wilson
2021-02-01  8:57 ` [Intel-gfx] [PATCH 54/57] drm/i915/gt: Infrastructure for ring scheduling Chris Wilson
2021-02-01  8:57 ` [Intel-gfx] [PATCH 55/57] drm/i915/gt: Implement ring scheduler for gen4-7 Chris Wilson
2021-02-01  8:57 ` [Intel-gfx] [PATCH 56/57] drm/i915/gt: Enable ring scheduling for gen5-7 Chris Wilson
2021-02-01  8:57 ` [Intel-gfx] [PATCH 57/57] drm/i915: Support secure dispatch on gen6/gen7 Chris Wilson
2021-02-01 14:13 ` [Intel-gfx] ✗ Fi.CI.CHECKPATCH: warning for series starting with [01/57] drm/i915/gt: Restrict the GT clock override to just Icelake Patchwork
2021-02-01 14:15 ` [Intel-gfx] ✗ Fi.CI.SPARSE: " Patchwork
2021-02-01 14:15 ` [Intel-gfx] [PATCH 01/57] " Mika Kuoppala
2021-02-01 14:41 ` [Intel-gfx] ✓ Fi.CI.BAT: success for series starting with [01/57] " Patchwork
2021-02-01 19:33 ` [Intel-gfx] ✗ Fi.CI.IGT: failure " Patchwork

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20210201085715.27435-41-chris@chris-wilson.co.uk \
    --to=chris@chris-wilson.co.uk \
    --cc=intel-gfx@lists.freedesktop.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link

Intel-GFX Archive on lore.kernel.org

Archives are clonable:
	git clone --mirror https://lore.kernel.org/intel-gfx/0 intel-gfx/git/0.git
	git clone --mirror https://lore.kernel.org/intel-gfx/1 intel-gfx/git/1.git

	# If you have public-inbox 1.1+ installed, you may
	# initialize and index your mirror using the following commands:
	public-inbox-init -V2 intel-gfx intel-gfx/ https://lore.kernel.org/intel-gfx \
		intel-gfx@lists.freedesktop.org
	public-inbox-index intel-gfx

Example config snippet for mirrors

Newsgroup available over NNTP:
	nntp://nntp.lore.kernel.org/org.freedesktop.lists.intel-gfx


AGPL code for this site: git clone https://public-inbox.org/public-inbox.git