Skip to contentSkip to conversation

drm/i915: Make own struct for execlist items

PATCH In review Mika Kuoppala 2 participants 7 messages Sep 22, 2017 – Sep 25, 2017

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH 1/5 drm/i915: Make own struct for execlist items

Engine's execlist related items have been increasing to a point where a separate struct is warranted. Carve execlist specific items to a dedicated struct to add clarity.

v2: add kerneldoc and fix whitespace (Joonas, Chris)
v3: csb_mmio changes, rebase
v4: s/\b(el|execlist)\b/execlists/ (Joonas)

---
 drivers/gpu/drm/i915/i915_debugfs.c        |   8 +--
 drivers/gpu/drm/i915/i915_gem.c            |   6 +-
 drivers/gpu/drm/i915/i915_gpu_error.c      |   4 +-
 drivers/gpu/drm/i915/i915_guc_submission.c |  31 +++++----
 drivers/gpu/drm/i915/i915_irq.c            |   5 +-
 drivers/gpu/drm/i915/intel_engine_cs.c     |  12 ++--
 drivers/gpu/drm/i915/intel_lrc.c           | 100 +++++++++++++++--------------
 drivers/gpu/drm/i915/intel_ringbuffer.h    | 100 +++++++++++++++++++++++------
 8 files changed, 167 insertions(+), 99 deletions(-)
drivers/gpu/drm/i915/i915_debugfs.c +4 -4
index 13fc25997d65..1470754e53d2 100644
@@ -3324,7 +3324,7 @@ static int i915_engine_info(struct seq_file *m, void *unused)
 			read = GEN8_CSB_READ_PTR(ptr);
 			write = GEN8_CSB_WRITE_PTR(ptr);
 			seq_printf(m, "\tExeclist CSB read %d [%d cached], write %d [%d from hws], interrupt posted? %s\n",
-				   read, engine->csb_head,
+				   read, engine->execlists.csb_head,
 				   write,
 				   intel_read_status_page(engine, intel_hws_csb_write_index(engine->i915)),
 				   yesno(test_bit(ENGINE_IRQ_EXECLIST,
@@ -3346,10 +3346,10 @@ static int i915_engine_info(struct seq_file *m, void *unused)
 			}
 
 			rcu_read_lock();
-			for (idx = 0; idx < ARRAY_SIZE(engine->execlist_port); idx++) {
+			for (idx = 0; idx < ARRAY_SIZE(engine->execlists.port); idx++) {
 				unsigned int count;
 
-				rq = port_unpack(&engine->execlist_port[idx],
+				rq = port_unpack(&engine->execlists.port[idx],
 						 &count);
 				if (rq) {
 					seq_printf(m, "\t\tELSP[%d] count=%d, ",
@@ -3363,7 +3363,7 @@ static int i915_engine_info(struct seq_file *m, void *unused)
 			rcu_read_unlock();
 
 			spin_lock_irq(&engine->timeline->lock);
-			for (rb = engine->execlist_first; rb; rb = rb_next(rb)){
+			for (rb = engine->execlists.first; rb; rb = rb_next(rb)) {
 				struct i915_priolist *p =
 					rb_entry(rb, typeof(*p), node);
 
drivers/gpu/drm/i915/i915_gem.c +3 -3
index 2a650f92aa74..73eeb6b1f1cd 100644
@@ -2815,8 +2815,8 @@ i915_gem_reset_prepare_engine(struct intel_engine_cs *engine)
 	 * Turning off the engine->irq_tasklet until the reset is over
 	 * prevents the race.
 	 */
-	tasklet_kill(&engine->irq_tasklet);
-	tasklet_disable(&engine->irq_tasklet);
+	tasklet_kill(&engine->execlists.irq_tasklet);
+	tasklet_disable(&engine->execlists.irq_tasklet);
 
 	if (engine->irq_seqno_barrier)
 		engine->irq_seqno_barrier(engine);
@@ -2995,7 +2995,7 @@ void i915_gem_reset(struct drm_i915_private *dev_priv)
 
 void i915_gem_reset_finish_engine(struct intel_engine_cs *engine)
 {
-	tasklet_enable(&engine->irq_tasklet);
+	tasklet_enable(&engine->execlists.irq_tasklet);
 	kthread_unpark(engine->breadcrumbs.signaler);
 }
 
drivers/gpu/drm/i915/i915_gpu_error.c +2 -2
index c7aaf628e7e0..45189795be35 100644
@@ -1327,10 +1327,10 @@ static void engine_record_requests(struct intel_engine_cs *engine,
 static void error_record_engine_execlists(struct intel_engine_cs *engine,
 					  struct drm_i915_error_engine *ee)
 {
-	const struct execlist_port *port = engine->execlist_port;
+	const struct execlist_port *port = engine->execlists.port;
 	unsigned int n;
 
-	for (n = 0; n < ARRAY_SIZE(engine->execlist_port); n++) {
+	for (n = 0; n < ARRAY_SIZE(engine->execlists.port); n++) {
 		struct drm_i915_gem_request *rq = port_request(&port[n]);
 
 		if (!rq)
drivers/gpu/drm/i915/i915_guc_submission.c +17 -14
index 06a26c610806..bce3f1b5892b 100644
@@ -494,11 +494,12 @@ static void i915_guc_submit(struct intel_engine_cs *engine)
 	struct drm_i915_private *dev_priv = engine->i915;
 	struct intel_guc *guc = &dev_priv->guc;
 	struct i915_guc_client *client = guc->execbuf_client;
-	struct execlist_port *port = engine->execlist_port;
-	unsigned int engine_id = engine->id;
+	struct intel_engine_execlists * const execlists = &engine->execlists;
+	struct execlist_port *port = execlists->port;
+	const unsigned int engine_id = engine->id;
 	unsigned int n;
 
-	for (n = 0; n < ARRAY_SIZE(engine->execlist_port); n++) {
+	for (n = 0; n < ARRAY_SIZE(execlists->port); n++) {
 		struct drm_i915_gem_request *rq;
 		unsigned int count;
 
@@ -558,7 +559,8 @@ static void port_assign(struct execlist_port *port,
 
 static void i915_guc_dequeue(struct intel_engine_cs *engine)
 {
-	struct execlist_port *port = engine->execlist_port;
+	struct intel_engine_execlists * const execlists = &engine->execlists;
+	struct execlist_port *port = execlists->port;
 	struct drm_i915_gem_request *last = NULL;
 	bool submit = false;
 	struct rb_node *rb;
@@ -567,15 +569,15 @@ static void i915_guc_dequeue(struct intel_engine_cs *engine)
 		port++;
 
 	spin_lock_irq(&engine->timeline->lock);
-	rb = engine->execlist_first;
-	GEM_BUG_ON(rb_first(&engine->execlist_queue) != rb);
+	rb = execlists->first;
+	GEM_BUG_ON(rb_first(&execlists->queue) != rb);
 	while (rb) {
 		struct i915_priolist *p = rb_entry(rb, typeof(*p), node);
 		struct drm_i915_gem_request *rq, *rn;
 
 		list_for_each_entry_safe(rq, rn, &p->requests, priotree.link) {
 			if (last && rq->ctx != last->ctx) {
-				if (port != engine->execlist_port) {
+				if (port != execlists->port) {
 					__list_del_many(&p->requests,
 							&rq->priotree.link);
 					goto done;
@@ -596,13 +598,13 @@ static void i915_guc_dequeue(struct intel_engine_cs *engine)
 		}
 
 		rb = rb_next(rb);
-		rb_erase(&p->node, &engine->execlist_queue);
+		rb_erase(&p->node, &execlists->queue);
 		INIT_LIST_HEAD(&p->requests);
 		if (p->priority != I915_PRIORITY_NORMAL)
 			kmem_cache_free(engine->i915->priorities, p);
 	}
 done:
-	engine->execlist_first = rb;
+	execlists->first = rb;
 	if (submit) {
 		port_assign(port, last);
 		i915_guc_submit(engine);
@@ -612,8 +614,8 @@ static void i915_guc_dequeue(struct intel_engine_cs *engine)
 
 static void i915_guc_irq_handler(unsigned long data)
 {
-	struct intel_engine_cs *engine = (struct intel_engine_cs *)data;
-	struct execlist_port *port = engine->execlist_port;
+	struct intel_engine_cs * const engine = (struct intel_engine_cs *)data;
+	struct execlist_port *port = engine->execlists.port;
 	struct drm_i915_gem_request *rq;
 
 	rq = port_request(&port[0]);
@@ -1144,7 +1146,7 @@ int i915_guc_submission_enable(struct drm_i915_private *dev_priv)
 	 * and it is guaranteed that it will remove the work item from the
 	 * queue before our request is completed.
 	 */
-	BUILD_BUG_ON(ARRAY_SIZE(engine->execlist_port) *
+	BUILD_BUG_ON(ARRAY_SIZE(engine->execlists.port) *
 		     sizeof(struct guc_wq_item) *
 		     I915_NUM_ENGINES > GUC_WQ_SIZE);
 
@@ -1175,14 +1177,15 @@ int i915_guc_submission_enable(struct drm_i915_private *dev_priv)
 	guc_interrupts_capture(dev_priv);
 
 	for_each_engine(engine, dev_priv, id) {
+		struct intel_engine_execlists * const execlists = &engine->execlists;
 		/* The tasklet was initialised by execlists, and may be in
 		 * a state of flux (across a reset) and so we just want to
 		 * take over the callback without changing any other state
 		 * in the tasklet.
 		 */
-		engine->irq_tasklet.func = i915_guc_irq_handler;
+		execlists->irq_tasklet.func = i915_guc_irq_handler;
 		clear_bit(ENGINE_IRQ_EXECLIST, &engine->irq_posted);
-		tasklet_schedule(&engine->irq_tasklet);
+		tasklet_schedule(&execlists->irq_tasklet);
 	}
 
 	return 0;
drivers/gpu/drm/i915/i915_irq.c +3 -2
index b1bab7605db9..af82bd721dbc 100644
@@ -1346,10 +1346,11 @@ static void snb_gt_irq_handler(struct drm_i915_private *dev_priv,
 static void
 gen8_cs_irq_handler(struct intel_engine_cs *engine, u32 iir, int test_shift)
 {
+	struct intel_engine_execlists * const execlists = &engine->execlists;
 	bool tasklet = false;
 
 	if (iir & (GT_CONTEXT_SWITCH_INTERRUPT << test_shift)) {
-		if (port_count(&engine->execlist_port[0])) {
+		if (port_count(&execlists->port[0])) {
 			__set_bit(ENGINE_IRQ_EXECLIST, &engine->irq_posted);
 			tasklet = true;
 		}
@@ -1361,7 +1362,7 @@ gen8_cs_irq_handler(struct intel_engine_cs *engine, u32 iir, int test_shift)
 	}
 
 	if (tasklet)
-		tasklet_hi_schedule(&engine->irq_tasklet);
+		tasklet_hi_schedule(&execlists->irq_tasklet);
 }
 
 static irqreturn_t gen8_gt_irq_ack(struct drm_i915_private *dev_priv,
drivers/gpu/drm/i915/intel_engine_cs.c +6 -6
index d755a2ae4223..957b0d47f635 100644
@@ -391,8 +391,8 @@ static void intel_engine_init_timeline(struct intel_engine_cs *engine)
  */
 void intel_engine_setup_common(struct intel_engine_cs *engine)
 {
-	engine->execlist_queue = RB_ROOT;
-	engine->execlist_first = NULL;
+	engine->execlists.queue = RB_ROOT;
+	engine->execlists.first = NULL;
 
 	intel_engine_init_timeline(engine);
 	intel_engine_init_hangcheck(engine);
@@ -1473,11 +1473,11 @@ bool intel_engine_is_idle(struct intel_engine_cs *engine)
 		return false;
 
 	/* Both ports drained, no more ELSP submission? */
-	if (port_request(&engine->execlist_port[0]))
+	if (port_request(&engine->execlists.port[0]))
 		return false;
 
 	/* ELSP is empty, but there are ready requests? */
-	if (READ_ONCE(engine->execlist_first))
+	if (READ_ONCE(engine->execlists.first))
 		return false;
 
 	/* Ring stopped? */
@@ -1526,8 +1526,8 @@ void intel_engines_mark_idle(struct drm_i915_private *i915)
 	for_each_engine(engine, i915, id) {
 		intel_engine_disarm_breadcrumbs(engine);
 		i915_gem_batch_pool_fini(&engine->batch_pool);
-		tasklet_kill(&engine->irq_tasklet);
-		engine->no_priolist = false;
+		tasklet_kill(&engine->execlists.irq_tasklet);
+		engine->execlists.no_priolist = false;
 	}
 }
 
drivers/gpu/drm/i915/intel_lrc.c +53 -47

323 lines not on this page · raw message

drivers/gpu/drm/i915/intel_ringbuffer.h +79 -21

112 lines not on this page · raw message

Signature · 6 lines
2.11.0

_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
#1 lore raw

PATCH 2/5 drm/i915: Move execlist initialization into intel_engine_cs.c

Move execlist init into a common engine setup. As it is common to both guc and hw execlists.

v2: rebase with csb changes
v3: rebase

---
 drivers/gpu/drm/i915/intel_engine_cs.c | 30 ++++++++++++++++++++++++++++--
 drivers/gpu/drm/i915/intel_lrc.c       | 19 -------------------
 2 files changed, 28 insertions(+), 21 deletions(-)
drivers/gpu/drm/i915/intel_engine_cs.c +28 -2
index 957b0d47f635..9e4edd170bed 100644
@@ -380,6 +380,33 @@ static void intel_engine_init_timeline(struct intel_engine_cs *engine)
 	engine->timeline = &engine->i915->gt.global_timeline.engine[engine->id];
 }
 
+static bool csb_force_mmio(struct drm_i915_private *i915)
+{
+	/* GVT emulation depends upon intercepting CSB mmio */
+	if (intel_vgpu_active(i915))
+		return true;
+
+	/*
+	 * IOMMU adds unpredictable latency causing the CSB write (from the
+	 * GPU into the HWSP) to only be visible some time after the interrupt
+	 * (missed breadcrumb syndrome).
+	 */
+	if (intel_vtd_active())
+		return true;
+
+	return false;
+}
+
+static void intel_engine_init_execlist(struct intel_engine_cs *engine)
+{
+	struct intel_engine_execlists * const execlists = &engine->execlists;
+
+	execlists->csb_use_mmio = csb_force_mmio(engine->i915);
+
+	execlists->queue = RB_ROOT;
+	execlists->first = NULL;
+}
+
 /**
  * intel_engines_setup_common - setup engine state not requiring hw access
  * @engine: Engine to setup.
@@ -391,8 +418,7 @@ static void intel_engine_init_timeline(struct intel_engine_cs *engine)
  */
 void intel_engine_setup_common(struct intel_engine_cs *engine)
 {
-	engine->execlists.queue = RB_ROOT;
-	engine->execlists.first = NULL;
+	intel_engine_init_execlist(engine);
 
 	intel_engine_init_timeline(engine);
 	intel_engine_init_hangcheck(engine);
drivers/gpu/drm/i915/intel_lrc.c +0 -19
index 4f202b840e3d..3186be54bbd8 100644
@@ -1784,23 +1784,6 @@ logical_ring_default_irqs(struct intel_engine_cs *engine)
 	engine->irq_keep_mask = GT_CONTEXT_SWITCH_INTERRUPT << shift;
 }
 
-static bool irq_handler_force_mmio(struct drm_i915_private *i915)
-{
-	/* GVT emulation depends upon intercepting CSB mmio */
-	if (intel_vgpu_active(i915))
-		return true;
-
-	/*
-	 * IOMMU adds unpredictable latency causing the CSB write (from the
-	 * GPU into the HWSP) to only be visible some time after the interrupt
-	 * (missed breadcrumb syndrome).
-	 */
-	if (intel_vtd_active())
-		return true;
-
-	return false;
-}
-
 static void
 logical_ring_setup(struct intel_engine_cs *engine)
 {
@@ -1812,8 +1795,6 @@ logical_ring_setup(struct intel_engine_cs *engine)
 	/* Intentionally left blank. */
 	engine->buffer = NULL;
 
-	engine->execlists.csb_use_mmio = irq_handler_force_mmio(dev_priv);
-
 	fw_domains = intel_uncore_forcewake_for_reg(dev_priv,
 						    RING_ELSP(engine),
 						    FW_REG_WRITE);
Signature · 6 lines
2.11.0

_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
#2 lore raw

PATCH 3/5 drm/i915: Wrap port cancellation into a function

In reply to Mika Kuoppala

On reset and wedged path, we want to release the requests that are tied to ports and then mark the ports to be unset. Introduce a function for this.

v2: rebase
v3: drop local, keep GEM_BUG_ON (Michał, Chris)
v4: rebase

---
 drivers/gpu/drm/i915/intel_lrc.c | 25 ++++++++++++++-----------
 1 file changed, 14 insertions(+), 11 deletions(-)
drivers/gpu/drm/i915/intel_lrc.c +14 -11
index 3186be54bbd8..4f625371b5fe 100644
@@ -568,21 +568,27 @@ static void execlists_dequeue(struct intel_engine_cs *engine)
 		execlists_submit_ports(engine);
 }
 
+static void execlist_cancel_port_requests(struct intel_engine_execlists *execlists)
+{
+	unsigned int i;
+
+	for (i = 0; i < ARRAY_SIZE(execlists->port); i++)
+		i915_gem_request_put(port_request(&execlists->port[i]));
+
+	memset(execlists->port, 0, sizeof(execlists->port));
+}
+
 static void execlists_cancel_requests(struct intel_engine_cs *engine)
 {
 	struct intel_engine_execlists * const execlists = &engine->execlists;
-	struct execlist_port *port = execlists->port;
 	struct drm_i915_gem_request *rq, *rn;
 	struct rb_node *rb;
 	unsigned long flags;
-	unsigned long n;
 
 	spin_lock_irqsave(&engine->timeline->lock, flags);
 
 	/* Cancel the requests on the HW and clear the ELSP tracker. */
-	for (n = 0; n < ARRAY_SIZE(execlists->port); n++)
-		i915_gem_request_put(port_request(&port[n]));
-	memset(execlists->port, 0, sizeof(execlists->port));
+	execlist_cancel_port_requests(execlists);
 
 	/* Mark all executing requests as skipped. */
 	list_for_each_entry(rq, &engine->timeline->requests, link) {
@@ -613,9 +619,10 @@ static void execlists_cancel_requests(struct intel_engine_cs *engine)
 
 	/* Remaining _unready_ requests will be nop'ed when submitted */
 
+
 	execlists->queue = RB_ROOT;
 	execlists->first = NULL;
-	GEM_BUG_ON(port_isset(&port[0]));
+	GEM_BUG_ON(port_isset(&execlists->port[0]));
 
 	/*
 	 * The port is checked prior to scheduling a tasklet, but
@@ -1372,11 +1379,9 @@ static void reset_common_ring(struct intel_engine_cs *engine,
 			      struct drm_i915_gem_request *request)
 {
 	struct intel_engine_execlists * const execlists = &engine->execlists;
-	struct execlist_port *port = execlists->port;
 	struct drm_i915_gem_request *rq, *rn;
 	struct intel_context *ce;
 	unsigned long flags;
-	unsigned int n;
 
 	spin_lock_irqsave(&engine->timeline->lock, flags);
 
@@ -1389,9 +1394,7 @@ static void reset_common_ring(struct intel_engine_cs *engine,
 	 * guessing the missed context-switch events by looking at what
 	 * requests were completed.
 	 */
-	for (n = 0; n < ARRAY_SIZE(execlists->port); n++)
-		i915_gem_request_put(port_request(&port[n]));
-	memset(execlists->port, 0, sizeof(execlists->port));
+	execlist_cancel_port_requests(execlists);
 
 	/* Push back any incomplete requests for replay after the reset. */
 	list_for_each_entry_safe_reverse(rq, rn,
Signature · 6 lines
2.11.0

_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
#3 lore raw

PATCH 4/5 drm/i915: Add execlist_port_complete

In reply to Mika Kuoppala

When first execlist entry is processed, we move the port (contents). Introduce function for this as execlist and guc use this common operation.

v2: rebase. s/GEM_DEBUG_BUG/GEM_BUG (Chris)
v3: rebase

---
 drivers/gpu/drm/i915/i915_guc_submission.c |  8 ++++----
 drivers/gpu/drm/i915/intel_lrc.c           | 22 +++++++++++-----------
 drivers/gpu/drm/i915/intel_ringbuffer.h    | 14 +++++++++++++-
 3 files changed, 28 insertions(+), 16 deletions(-)
drivers/gpu/drm/i915/i915_guc_submission.c +4 -4
index bce3f1b5892b..55e15a57c3d9 100644
@@ -592,7 +592,7 @@ static void i915_guc_dequeue(struct intel_engine_cs *engine)
 			rq->priotree.priority = INT_MAX;
 
 			__i915_gem_request_submit(rq);
-			trace_i915_gem_request_in(rq, port_index(port, engine));
+			trace_i915_gem_request_in(rq, port_index(port, execlists));
 			last = rq;
 			submit = true;
 		}
@@ -615,7 +615,8 @@ static void i915_guc_dequeue(struct intel_engine_cs *engine)
 static void i915_guc_irq_handler(unsigned long data)
 {
 	struct intel_engine_cs * const engine = (struct intel_engine_cs *)data;
-	struct execlist_port *port = engine->execlists.port;
+	struct intel_engine_execlists * const execlists = &engine->execlists;
+	struct execlist_port *port = execlists->port;
 	struct drm_i915_gem_request *rq;
 
 	rq = port_request(&port[0]);
@@ -623,8 +624,7 @@ static void i915_guc_irq_handler(unsigned long data)
 		trace_i915_gem_request_out(rq);
 		i915_gem_request_put(rq);
 
-		port[0] = port[1];
-		memset(&port[1], 0, sizeof(port[1]));
+		execlists_port_complete(execlists, port);
 
 		rq = port_request(&port[0]);
 	}
drivers/gpu/drm/i915/intel_lrc.c +11 -11
index 4f625371b5fe..3b03f19f1395 100644
@@ -454,7 +454,8 @@ static void port_assign(struct execlist_port *port,
 static void execlists_dequeue(struct intel_engine_cs *engine)
 {
 	struct drm_i915_gem_request *last;
-	struct execlist_port *port = engine->execlists.port;
+	struct intel_engine_execlists * const execlists = &engine->execlists;
+	struct execlist_port *port = execlists->port;
 	struct rb_node *rb;
 	bool submit = false;
 
@@ -468,8 +469,6 @@ static void execlists_dequeue(struct intel_engine_cs *engine)
 		 */
 		last->tail = last->wa_tail;
 
-	GEM_BUG_ON(port_isset(&port[1]));
-
 	/* Hardware submission is through 2 ports. Conceptually each port
 	 * has a (RING_START, RING_HEAD, RING_TAIL) tuple. RING_START is
 	 * static for a context, and unique to each, so we only execute
@@ -492,8 +491,8 @@ static void execlists_dequeue(struct intel_engine_cs *engine)
 	 */
 
 	spin_lock_irq(&engine->timeline->lock);
-	rb = engine->execlists.first;
-	GEM_BUG_ON(rb_first(&engine->execlists.queue) != rb);
+	rb = execlists->first;
+	GEM_BUG_ON(rb_first(&execlists->queue) != rb);
 	while (rb) {
 		struct i915_priolist *p = rb_entry(rb, typeof(*p), node);
 		struct drm_i915_gem_request *rq, *rn;
@@ -516,7 +515,7 @@ static void execlists_dequeue(struct intel_engine_cs *engine)
 				 * combine this request with the last, then we
 				 * are done.
 				 */
-				if (port != engine->execlists.port) {
+				if (port != execlists->port) {
 					__list_del_many(&p->requests,
 							&rq->priotree.link);
 					goto done;
@@ -541,25 +540,27 @@ static void execlists_dequeue(struct intel_engine_cs *engine)
 				if (submit)
 					port_assign(port, last);
 				port++;
+
+				GEM_BUG_ON(port_isset(port));
 			}
 
 			INIT_LIST_HEAD(&rq->priotree.link);
 			rq->priotree.priority = INT_MAX;
 
 			__i915_gem_request_submit(rq);
-			trace_i915_gem_request_in(rq, port_index(port, engine));
+			trace_i915_gem_request_in(rq, port_index(port, execlists));
 			last = rq;
 			submit = true;
 		}
 
 		rb = rb_next(rb);
-		rb_erase(&p->node, &engine->execlists.queue);
+		rb_erase(&p->node, &execlists->queue);
 		INIT_LIST_HEAD(&p->requests);
 		if (p->priority != I915_PRIORITY_NORMAL)
 			kmem_cache_free(engine->i915->priorities, p);
 	}
 done:
-	engine->execlists.first = rb;
+	execlists->first = rb;
 	if (submit)
 		port_assign(port, last);
 	spin_unlock_irq(&engine->timeline->lock);
@@ -748,8 +749,7 @@ static void intel_lrc_irq_handler(unsigned long data)
 				trace_i915_gem_request_out(rq);
 				i915_gem_request_put(rq);
 
-				port[0] = port[1];
-				memset(&port[1], 0, sizeof(port[1]));
+				execlists_port_complete(execlists, port);
 			} else {
 				port_set(port, port_pack(rq, count));
 			}
drivers/gpu/drm/i915/intel_ringbuffer.h +13 -1
index 421e769adb79..0eae5936bc3c 100644
@@ -228,7 +228,7 @@ struct intel_engine_execlists {
 #define port_unpack(p, count) ptr_unpack_bits((p)->request_count, count, EXECLIST_COUNT_BITS)
 #define port_set(p, packed) ((p)->request_count = (packed))
 #define port_isset(p) ((p)->request_count)
-#define port_index(p, e) ((p) - (e)->execlists.port)
+#define port_index(p, execlists) ((p) - (execlists)->port)
 
 		/**
 		 * @context_id: context ID for port
@@ -511,6 +511,18 @@ struct intel_engine_cs {
 	u32 (*get_cmd_length_mask)(u32 cmd_header);
 };
 
+static inline void
+execlists_port_complete(struct intel_engine_execlists * const execlists,
+			struct execlist_port * const port)
+{
+	struct execlist_port * const port1 = &execlists->port[1];
+
+	GEM_BUG_ON(port_index(port, execlists) != 0);
+
+	*port = *port1;
+	memset(port1, 0, sizeof(struct execlist_port));
+}
+
 static inline unsigned int
 intel_engine_flag(const struct intel_engine_cs *engine)
 {
Signature · 6 lines
2.11.0

_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
#4 lore raw

PATCH 5/5 drm/i915: Make execlist port count variable

In reply to Mika Kuoppala

As we emulate execlists on top of the GuC workqueue, it is not restricted to just 2 ports and we can increase that number arbitrarily to trade-off queue depth (i.e. scheduling latency) against pipeline bubbles.

v2: rebase. better commit msg (Chris)
v3: rebase

---
 drivers/gpu/drm/i915/i915_debugfs.c        | 10 +++++-----
 drivers/gpu/drm/i915/i915_drv.h            |  3 ++-
 drivers/gpu/drm/i915/i915_gpu_error.c      | 17 ++++++++++++-----
 drivers/gpu/drm/i915/i915_guc_submission.c |  8 ++++++--
 drivers/gpu/drm/i915/intel_engine_cs.c     |  4 ++++
 drivers/gpu/drm/i915/intel_lrc.c           |  6 ++++--
 drivers/gpu/drm/i915/intel_ringbuffer.h    | 21 +++++++++++++++++----
 7 files changed, 50 insertions(+), 19 deletions(-)
drivers/gpu/drm/i915/i915_debugfs.c +5 -5
index 1470754e53d2..847f8e8d6b58 100644
@@ -3313,6 +3313,7 @@ static int i915_engine_info(struct seq_file *m, void *unused)
 
 		if (i915_modparams.enable_execlists) {
 			const u32 *hws = &engine->status_page.page_addr[I915_HWS_CSB_BUF0_INDEX];
+			struct intel_engine_execlists * const execlists = &engine->execlists;
 			u32 ptr, read, write;
 			unsigned int idx;
 
@@ -3324,7 +3325,7 @@ static int i915_engine_info(struct seq_file *m, void *unused)
 			read = GEN8_CSB_READ_PTR(ptr);
 			write = GEN8_CSB_WRITE_PTR(ptr);
 			seq_printf(m, "\tExeclist CSB read %d [%d cached], write %d [%d from hws], interrupt posted? %s\n",
-				   read, engine->execlists.csb_head,
+				   read, execlists->csb_head,
 				   write,
 				   intel_read_status_page(engine, intel_hws_csb_write_index(engine->i915)),
 				   yesno(test_bit(ENGINE_IRQ_EXECLIST,
@@ -3346,11 +3347,10 @@ static int i915_engine_info(struct seq_file *m, void *unused)
 			}
 
 			rcu_read_lock();
-			for (idx = 0; idx < ARRAY_SIZE(engine->execlists.port); idx++) {
+			for (idx = 0; idx < execlists_num_ports(execlists); idx++) {
 				unsigned int count;
 
-				rq = port_unpack(&engine->execlists.port[idx],
-						 &count);
+				rq = port_unpack(&execlists->port[idx], &count);
 				if (rq) {
 					seq_printf(m, "\t\tELSP[%d] count=%d, ",
 						   idx, count);
@@ -3363,7 +3363,7 @@ static int i915_engine_info(struct seq_file *m, void *unused)
 			rcu_read_unlock();
 
 			spin_lock_irq(&engine->timeline->lock);
-			for (rb = engine->execlists.first; rb; rb = rb_next(rb)) {
+			for (rb = execlists->first; rb; rb = rb_next(rb)) {
 				struct i915_priolist *p =
 					rb_entry(rb, typeof(*p), node);
 
drivers/gpu/drm/i915/i915_drv.h +2 -1
index f5d0e816008d..c1e93a61d81b 100644
@@ -1000,7 +1000,8 @@ struct i915_gpu_state {
 			u32 seqno;
 			u32 head;
 			u32 tail;
-		} *requests, execlist[2];
+		} *requests, execlist[EXECLIST_MAX_PORTS];
+		unsigned int num_ports;
 
 		struct drm_i915_error_waiter {
 			char comm[TASK_COMM_LEN];
drivers/gpu/drm/i915/i915_gpu_error.c +12 -5
index 45189795be35..12146d70a62e 100644
@@ -396,6 +396,8 @@ static void error_print_context(struct drm_i915_error_state_buf *m,
 static void error_print_engine(struct drm_i915_error_state_buf *m,
 			       const struct drm_i915_error_engine *ee)
 {
+	int n;
+
 	err_printf(m, "%s command stream:\n", engine_str(ee->engine_id));
 	err_printf(m, "  START: 0x%08x\n", ee->start);
 	err_printf(m, "  HEAD:  0x%08x [0x%08x]\n", ee->head, ee->rq_head);
@@ -465,8 +467,11 @@ static void error_print_engine(struct drm_i915_error_state_buf *m,
 		   jiffies_to_msecs(jiffies - ee->hangcheck_timestamp));
 	err_printf(m, "  engine reset count: %u\n", ee->reset_count);
 
-	error_print_request(m, "  ELSP[0]: ", &ee->execlist[0]);
-	error_print_request(m, "  ELSP[1]: ", &ee->execlist[1]);
+	for (n = 0; n < ee->num_ports; n++) {
+		err_printf(m, "  ELSP[%d]:", n);
+		error_print_request(m, " ", &ee->execlist[n]);
+	}
+
 	error_print_context(m, "  Active context: ", &ee->context);
 }
 
@@ -1327,17 +1332,19 @@ static void engine_record_requests(struct intel_engine_cs *engine,
 static void error_record_engine_execlists(struct intel_engine_cs *engine,
 					  struct drm_i915_error_engine *ee)
 {
-	const struct execlist_port *port = engine->execlists.port;
+	const struct intel_engine_execlists * const execlists = &engine->execlists;
 	unsigned int n;
 
-	for (n = 0; n < ARRAY_SIZE(engine->execlists.port); n++) {
-		struct drm_i915_gem_request *rq = port_request(&port[n]);
+	for (n = 0; n < execlists_num_ports(execlists); n++) {
+		struct drm_i915_gem_request *rq = port_request(&execlists->port[n]);
 
 		if (!rq)
 			break;
 
 		record_request(rq, &ee->execlist[n]);
 	}
+
+	ee->num_ports = n;
 }
 
 static void record_context(struct drm_i915_error_context *e,
drivers/gpu/drm/i915/i915_guc_submission.c +6 -2
index 55e15a57c3d9..04f1281d81a5 100644
@@ -562,6 +562,8 @@ static void i915_guc_dequeue(struct intel_engine_cs *engine)
 	struct intel_engine_execlists * const execlists = &engine->execlists;
 	struct execlist_port *port = execlists->port;
 	struct drm_i915_gem_request *last = NULL;
+	const struct execlist_port * const last_port =
+		&execlists->port[execlists->port_mask];
 	bool submit = false;
 	struct rb_node *rb;
 
@@ -577,7 +579,7 @@ static void i915_guc_dequeue(struct intel_engine_cs *engine)
 
 		list_for_each_entry_safe(rq, rn, &p->requests, priotree.link) {
 			if (last && rq->ctx != last->ctx) {
-				if (port != execlists->port) {
+				if (port == last_port) {
 					__list_del_many(&p->requests,
 							&rq->priotree.link);
 					goto done;
@@ -617,6 +619,8 @@ static void i915_guc_irq_handler(unsigned long data)
 	struct intel_engine_cs * const engine = (struct intel_engine_cs *)data;
 	struct intel_engine_execlists * const execlists = &engine->execlists;
 	struct execlist_port *port = execlists->port;
+	const struct execlist_port * const last_port =
+		&execlists->port[execlists->port_mask];
 	struct drm_i915_gem_request *rq;
 
 	rq = port_request(&port[0]);
@@ -629,7 +633,7 @@ static void i915_guc_irq_handler(unsigned long data)
 		rq = port_request(&port[0]);
 	}
 
-	if (!port_isset(&port[1]))
+	if (!port_isset(last_port))
 		i915_guc_dequeue(engine);
 }
 
drivers/gpu/drm/i915/intel_engine_cs.c +4 -0
index 9e4edd170bed..9d42d4895235 100644
@@ -403,6 +403,10 @@ static void intel_engine_init_execlist(struct intel_engine_cs *engine)
 
 	execlists->csb_use_mmio = csb_force_mmio(engine->i915);
 
+	execlists->port_mask = 1;
+	BUILD_BUG_ON_NOT_POWER_OF_2(execlists_num_ports(execlists));
+	GEM_BUG_ON(execlists_num_ports(execlists) > EXECLIST_MAX_PORTS);
+
 	execlists->queue = RB_ROOT;
 	execlists->first = NULL;
 }
drivers/gpu/drm/i915/intel_lrc.c +4 -2
index 3b03f19f1395..f91e126a7f97 100644
@@ -399,7 +399,7 @@ static void execlists_submit_ports(struct intel_engine_cs *engine)
 		engine->i915->regs + i915_mmio_reg_offset(RING_ELSP(engine));
 	unsigned int n;
 
-	for (n = ARRAY_SIZE(engine->execlists.port); n--; ) {
+	for (n = execlists_num_ports(&engine->execlists); n--; ) {
 		struct drm_i915_gem_request *rq;
 		unsigned int count;
 		u64 desc;
@@ -456,6 +456,8 @@ static void execlists_dequeue(struct intel_engine_cs *engine)
 	struct drm_i915_gem_request *last;
 	struct intel_engine_execlists * const execlists = &engine->execlists;
 	struct execlist_port *port = execlists->port;
+	const struct execlist_port * const last_port =
+		&execlists->port[execlists->port_mask];
 	struct rb_node *rb;
 	bool submit = false;
 
@@ -515,7 +517,7 @@ static void execlists_dequeue(struct intel_engine_cs *engine)
 				 * combine this request with the last, then we
 				 * are done.
 				 */
-				if (port != execlists->port) {
+				if (port == last_port) {
 					__list_del_many(&p->requests,
 							&rq->priotree.link);
 					goto done;
drivers/gpu/drm/i915/intel_ringbuffer.h +17 -4
index 0eae5936bc3c..56d7ae9f298b 100644
@@ -234,7 +234,14 @@ struct intel_engine_execlists {
 		 * @context_id: context ID for port
 		 */
 		GEM_DEBUG_DECL(u32 context_id);
-	} port[2];
+
+#define EXECLIST_MAX_PORTS 2
+	} port[EXECLIST_MAX_PORTS];
+
+	/**
+	 * @port_mask: number of execlist ports - 1
+	 */
+	unsigned int port_mask;
 
 	/**
 	 * @queue: queue of requests, in priority lists
@@ -511,16 +518,22 @@ struct intel_engine_cs {
 	u32 (*get_cmd_length_mask)(u32 cmd_header);
 };
 
+static inline unsigned int
+execlists_num_ports(const struct intel_engine_execlists * const execlists)
+{
+	return execlists->port_mask + 1;
+}
+
 static inline void
 execlists_port_complete(struct intel_engine_execlists * const execlists,
 			struct execlist_port * const port)
 {
-	struct execlist_port * const port1 = &execlists->port[1];
+	const unsigned int m = execlists->port_mask;
 
 	GEM_BUG_ON(port_index(port, execlists) != 0);
 
-	*port = *port1;
-	memset(port1, 0, sizeof(struct execlist_port));
+	memmove(port, port + 1, m * sizeof(struct execlist_port));
+	memset(port + m, 0, sizeof(struct execlist_port));
 }
 
 static inline unsigned int
Signature · 6 lines
2.11.0

_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
#5 lore raw

✗ Fi.CI.BAT: warning for series starting with [CI,1/5] drm/i915: Make own struct for execlist items

In reply to Mika Kuoppala
== Series Details ==

== Series Details ==

Series: series starting with [CI,1/5] drm/i915: Make own struct for execlist items
URL : https://patchwork.freedesktop.org/series/30761/
State : warning

== Summary ==

Series 30761v1 series starting with [CI,1/5] drm/i915: Make own struct for execlist items
https://patchwork.freedesktop.org/api/1.0/series/30761/revisions/1/mbox/

Test gem_mmap_gtt:

        Subgroup basic-write-cpu-read-gtt:
                pass       -> DMESG-WARN (fi-kbl-7500u)
Test gem_pread:
        Subgroup basic:
                pass       -> DMESG-WARN (fi-kbl-7500u)
Test gem_sync:
        Subgroup basic-store-all:
                pass       -> DMESG-WARN (fi-kbl-7500u)
Test kms_addfb_basic:
        Subgroup bad-pitch-32:
                pass       -> DMESG-WARN (fi-kbl-7500u)
Test kms_busy:
        Subgroup basic-flip-a:
                pass       -> DMESG-WARN (fi-kbl-7500u)
Test kms_cursor_legacy:
        Subgroup basic-busy-flip-before-cursor-legacy:
                pass       -> FAIL       (fi-snb-2600) fdo#100215
Test kms_pipe_crc_basic:
        Subgroup suspend-read-crc-pipe-a:
                pass       -> INCOMPLETE (fi-kbl-7500u) fdo#102850
Test drv_module_reload:
        Subgroup basic-no-display:
                dmesg-warn -> PASS       (fi-glk-1) fdo#102777 +1

fdo#100215 https://bugs.freedesktop.org/show_bug.cgi?id=100215
fdo#102850 https://bugs.freedesktop.org/show_bug.cgi?id=102850
fdo#102777 https://bugs.freedesktop.org/show_bug.cgi?id=102777

fi-bdw-5557u     total:289  pass:268  dwarn:0   dfail:0   fail:0   skip:21  time:437s
fi-bdw-gvtdvm    total:289  pass:265  dwarn:0   dfail:0   fail:0   skip:24  time:469s
fi-blb-e6850     total:289  pass:224  dwarn:1   dfail:0   fail:0   skip:64  time:415s
fi-bsw-n3050     total:289  pass:243  dwarn:0   dfail:0   fail:0   skip:46  time:522s
fi-bwr-2160      total:289  pass:184  dwarn:0   dfail:0   fail:0   skip:105 time:277s
fi-bxt-j4205     total:289  pass:260  dwarn:0   dfail:0   fail:0   skip:29  time:500s
fi-byt-j1900     total:289  pass:254  dwarn:1   dfail:0   fail:0   skip:34  time:490s
fi-byt-n2820     total:289  pass:250  dwarn:1   dfail:0   fail:0   skip:38  time:492s
fi-cfl-s         total:289  pass:223  dwarn:34  dfail:0   fail:0   skip:32  time:541s
fi-elk-e7500     total:289  pass:230  dwarn:0   dfail:0   fail:0   skip:59  time:414s
fi-glk-1         total:289  pass:259  dwarn:1   dfail:0   fail:0   skip:29  time:568s
fi-hsw-4770      total:289  pass:263  dwarn:0   dfail:0   fail:0   skip:26  time:424s
fi-hsw-4770r     total:289  pass:263  dwarn:0   dfail:0   fail:0   skip:26  time:406s
fi-ilk-650       total:289  pass:229  dwarn:0   dfail:0   fail:0   skip:60  time:430s
fi-ivb-3520m     total:289  pass:261  dwarn:0   dfail:0   fail:0   skip:28  time:488s
fi-ivb-3770      total:289  pass:261  dwarn:0   dfail:0   fail:0   skip:28  time:462s
fi-kbl-7500u     total:245  pass:218  dwarn:6   dfail:0   fail:0   skip:20 
fi-kbl-7560u     total:289  pass:270  dwarn:0   dfail:0   fail:0   skip:19  time:574s
fi-kbl-r         total:289  pass:262  dwarn:0   dfail:0   fail:0   skip:27  time:588s
fi-pnv-d510      total:289  pass:223  dwarn:1   dfail:0   fail:0   skip:65  time:539s
fi-skl-6260u     total:289  pass:269  dwarn:0   dfail:0   fail:0   skip:20  time:455s
fi-skl-6700k     total:289  pass:265  dwarn:0   dfail:0   fail:0   skip:24  time:747s
fi-skl-6770hq    total:289  pass:269  dwarn:0   dfail:0   fail:0   skip:20  time:492s
fi-skl-gvtdvm    total:289  pass:266  dwarn:0   dfail:0   fail:0   skip:23  time:469s
fi-snb-2520m     total:289  pass:251  dwarn:0   dfail:0   fail:0   skip:38  time:566s
fi-snb-2600      total:289  pass:248  dwarn:0   dfail:0   fail:2   skip:39  time:420s

e0e308721fd283e1c5777657a5941f178f0d49e6 drm-tip: 2017y-09m-22d-13h-31m-38s UTC integration manifest 2fd2949c33c1 drm/i915: Make execlist port count variable
869653ef354d drm/i915: Add execlist_port_complete
0362499b6132 drm/i915: Wrap port cancellation into a function
43f2486ca3d6 drm/i915: Move execlist initialization into intel_engine_cs.c f0df0ae37cca drm/i915: Make own struct for execlist items

== Logs ==

For more details see: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_5792/
_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

#6 lore raw

Re: ✗ Fi.CI.BAT: warning for series starting with [CI,1/5] drm/i915: Make own struct for execlist items

Quoted 26 lines · Patchwork

== Series Details ==

Series: series starting with [CI,1/5] drm/i915: Make own struct for execlist items
URL : https://patchwork.freedesktop.org/series/30761/
State : warning

== Summary ==

Series 30761v1 series starting with [CI,1/5] drm/i915: Make own struct for execlist items
https://patchwork.freedesktop.org/api/1.0/series/30761/revisions/1/mbox/

Test gem_mmap_gtt:

        Subgroup basic-write-cpu-read-gtt:
                pass       -> DMESG-WARN (fi-kbl-7500u)
Test gem_pread:
        Subgroup basic:
                pass       -> DMESG-WARN (fi-kbl-7500u)
Test gem_sync:
        Subgroup basic-store-all:
                pass       -> DMESG-WARN (fi-kbl-7500u)
Test kms_addfb_basic:
        Subgroup bad-pitch-32:
                pass       -> DMESG-WARN (fi-kbl-7500u)
Test kms_busy:
        Subgroup basic-flip-a:
                pass       -> DMESG-WARN (fi-kbl-7500u)
Quoted 1 line

Test kms_cursor_legacy:

All above:

https://bugs.freedesktop.org/show_bug.cgi?id=102849

Quoted 47 lines
        Subgroup basic-busy-flip-before-cursor-legacy:
                pass       -> FAIL       (fi-snb-2600) fdo#100215
Test kms_pipe_crc_basic:
        Subgroup suspend-read-crc-pipe-a:
                pass       -> INCOMPLETE (fi-kbl-7500u) fdo#102850
Test drv_module_reload:
        Subgroup basic-no-display:
                dmesg-warn -> PASS       (fi-glk-1) fdo#102777 +1

fdo#100215 https://bugs.freedesktop.org/show_bug.cgi?id=100215
fdo#102850 https://bugs.freedesktop.org/show_bug.cgi?id=102850
fdo#102777 https://bugs.freedesktop.org/show_bug.cgi?id=102777

fi-bdw-5557u     total:289  pass:268  dwarn:0   dfail:0   fail:0   skip:21  time:437s
fi-bdw-gvtdvm    total:289  pass:265  dwarn:0   dfail:0   fail:0   skip:24  time:469s
fi-blb-e6850     total:289  pass:224  dwarn:1   dfail:0   fail:0   skip:64  time:415s
fi-bsw-n3050     total:289  pass:243  dwarn:0   dfail:0   fail:0   skip:46  time:522s
fi-bwr-2160      total:289  pass:184  dwarn:0   dfail:0   fail:0   skip:105 time:277s
fi-bxt-j4205     total:289  pass:260  dwarn:0   dfail:0   fail:0   skip:29  time:500s
fi-byt-j1900     total:289  pass:254  dwarn:1   dfail:0   fail:0   skip:34  time:490s
fi-byt-n2820     total:289  pass:250  dwarn:1   dfail:0   fail:0   skip:38  time:492s
fi-cfl-s         total:289  pass:223  dwarn:34  dfail:0   fail:0   skip:32  time:541s
fi-elk-e7500     total:289  pass:230  dwarn:0   dfail:0   fail:0   skip:59  time:414s
fi-glk-1         total:289  pass:259  dwarn:1   dfail:0   fail:0   skip:29  time:568s
fi-hsw-4770      total:289  pass:263  dwarn:0   dfail:0   fail:0   skip:26  time:424s
fi-hsw-4770r     total:289  pass:263  dwarn:0   dfail:0   fail:0   skip:26  time:406s
fi-ilk-650       total:289  pass:229  dwarn:0   dfail:0   fail:0   skip:60  time:430s
fi-ivb-3520m     total:289  pass:261  dwarn:0   dfail:0   fail:0   skip:28  time:488s
fi-ivb-3770      total:289  pass:261  dwarn:0   dfail:0   fail:0   skip:28  time:462s
fi-kbl-7500u     total:245  pass:218  dwarn:6   dfail:0   fail:0   skip:20 
fi-kbl-7560u     total:289  pass:270  dwarn:0   dfail:0   fail:0   skip:19  time:574s
fi-kbl-r         total:289  pass:262  dwarn:0   dfail:0   fail:0   skip:27  time:588s
fi-pnv-d510      total:289  pass:223  dwarn:1   dfail:0   fail:0   skip:65  time:539s
fi-skl-6260u     total:289  pass:269  dwarn:0   dfail:0   fail:0   skip:20  time:455s
fi-skl-6700k     total:289  pass:265  dwarn:0   dfail:0   fail:0   skip:24  time:747s
fi-skl-6770hq    total:289  pass:269  dwarn:0   dfail:0   fail:0   skip:20  time:492s
fi-skl-gvtdvm    total:289  pass:266  dwarn:0   dfail:0   fail:0   skip:23  time:469s
fi-snb-2520m     total:289  pass:251  dwarn:0   dfail:0   fail:0   skip:38  time:566s
fi-snb-2600      total:289  pass:248  dwarn:0   dfail:0   fail:2   skip:39  time:420s

e0e308721fd283e1c5777657a5941f178f0d49e6 drm-tip: 2017y-09m-22d-13h-31m-38s UTC integration manifest 2fd2949c33c1 drm/i915: Make execlist port count variable
869653ef354d drm/i915: Add execlist_port_complete
0362499b6132 drm/i915: Wrap port cancellation into a function
43f2486ca3d6 drm/i915: Move execlist initialization into intel_engine_cs.c f0df0ae37cca drm/i915: Make own struct for execlist items

Quoted 3 lines

== Logs ==

For more details see: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_5792/

_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx