[Intel-gfx] [PATCH RFC] drm/i915/vlv: Ramp up gpu freq gradually
Mika Kuoppala
mika.kuoppala at linux.intel.com
Tue Jan 16 15:21:16 UTC 2018
There is a suspicion that with aggressive upclocking, power rail
voltage fluctuations can disrupt c state transition, leading
to system hang.
When upclocking with 4 cpu Baytrails, bring up cpus to c1 and then
go through bins gradually towards target frequency to give leeway
for hw.
We go towards requested frequency on 1 millisecond intervals. For
each 1 millisecond, we increase the frequency by half of bins
that are in between current frequency and target.
This will have an adverse effect on client boosting, delaying
reaching full frequency by about 8ms.
References: https://bugzilla.kernel.org/show_bug.cgi?id=109051
References: https://bugs.freedesktop.org/show_bug.cgi?id=102657
Cc: Chris Wilson <chris at chris-wilson.co.uk>
Cc: Hans de Goede <hdegoede at redhat.com>
Signed-off-by: Mika Kuoppala <mika.kuoppala at linux.intel.com>
---
drivers/gpu/drm/i915/i915_debugfs.c | 7 ++++
drivers/gpu/drm/i915/i915_drv.c | 5 +++
drivers/gpu/drm/i915/i915_drv.h | 4 +++
drivers/gpu/drm/i915/intel_pm.c | 72 +++++++++++++++++++++++++++++++++++--
4 files changed, 86 insertions(+), 2 deletions(-)
diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c
index cc659b4b2a45..9c718c2811b8 100644
--- a/drivers/gpu/drm/i915/i915_debugfs.c
+++ b/drivers/gpu/drm/i915/i915_debugfs.c
@@ -1043,6 +1043,13 @@ static int i915_frequency_info(struct seq_file *m, void *unused)
seq_printf(m, "current GPU freq: %d MHz\n",
intel_gpu_freq(dev_priv, rps->cur_freq));
+ if (IS_VALLEYVIEW(dev_priv)) {
+ seq_printf(m, "target GPU freq: %d MHz\n",
+ intel_gpu_freq(dev_priv, rps->target_freq));
+ seq_printf(m, "QOS CPU DMA Latency %d usecs\n",
+ pm_qos_request(PM_QOS_CPU_DMA_LATENCY));
+ }
+
seq_printf(m, "max GPU freq: %d MHz\n",
intel_gpu_freq(dev_priv, rps->max_freq));
diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c
index 6c8da9d20c33..9718afacfa30 100644
--- a/drivers/gpu/drm/i915/i915_drv.c
+++ b/drivers/gpu/drm/i915/i915_drv.c
@@ -896,6 +896,10 @@ static int i915_driver_init_early(struct drm_i915_private *dev_priv,
BUG_ON(device_info->gen > sizeof(device_info->gen_mask) * BITS_PER_BYTE);
device_info->gen_mask = BIT(device_info->gen - 1);
+ pm_qos_add_request(&dev_priv->gt_pm.rps.qos,
+ PM_QOS_CPU_DMA_LATENCY,
+ PM_QOS_DEFAULT_VALUE);
+
spin_lock_init(&dev_priv->irq_lock);
spin_lock_init(&dev_priv->gpu_error.lock);
mutex_init(&dev_priv->backlight_lock);
@@ -953,6 +957,7 @@ static void i915_driver_cleanup_early(struct drm_i915_private *dev_priv)
intel_irq_fini(dev_priv);
i915_workqueues_cleanup(dev_priv);
i915_engines_cleanup(dev_priv);
+ pm_qos_remove_request(&dev_priv->gt_pm.rps.qos);
}
static int i915_mmio_setup(struct drm_i915_private *dev_priv)
diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
index c42015b05b47..90cc2788fbbb 100644
--- a/drivers/gpu/drm/i915/i915_drv.h
+++ b/drivers/gpu/drm/i915/i915_drv.h
@@ -936,6 +936,10 @@ struct intel_rps {
/* manual wa residency calculations */
struct intel_rps_ei ei;
+
+ u8 target_freq; /* vlv target to reach */
+ struct pm_qos_request qos;
+ struct delayed_work vlv_rps_work;
};
struct intel_rc6 {
diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c
index 1db79a860b96..00ef4cb7bd8f 100644
--- a/drivers/gpu/drm/i915/intel_pm.c
+++ b/drivers/gpu/drm/i915/intel_pm.c
@@ -6200,7 +6200,7 @@ static int gen6_set_rps(struct drm_i915_private *dev_priv, u8 val)
return 0;
}
-static int valleyview_set_rps(struct drm_i915_private *dev_priv, u8 val)
+static int valleyview_set_hw_rps(struct drm_i915_private *dev_priv, u8 val)
{
int err;
@@ -6224,6 +6224,69 @@ static int valleyview_set_rps(struct drm_i915_private *dev_priv, u8 val)
return 0;
}
+static void valleyview_queue_rps_change(struct drm_i915_private *dev_priv)
+{
+ struct intel_rps * const rps = &dev_priv->gt_pm.rps;
+ unsigned long delay;
+
+ delay = 1 + msecs_to_jiffies(1);
+ queue_delayed_work(system_wq, &rps->vlv_rps_work, delay);
+}
+
+static int valleyview_set_rps(struct drm_i915_private *dev_priv, u8 val)
+{
+ struct intel_rps * const rps = &dev_priv->gt_pm.rps;
+
+ /* 2 cpu Baytrails seems to be safe */
+ if (IS_CHERRYVIEW(dev_priv) || num_online_cpus() <= 2)
+ return valleyview_set_hw_rps(dev_priv, val);
+
+ /*
+ * For >2 cpu Baytrails we need to get cpus to C1
+ * and limit the freq rampup.
+ */
+ rps->target_freq = val;
+ pm_qos_update_request(&dev_priv->gt_pm.rps.qos, 0);
+
+ valleyview_queue_rps_change(dev_priv);
+
+ return 0;
+}
+
+static void vlv_rps_write_work(struct work_struct *work)
+{
+ struct drm_i915_private *dev_priv =
+ container_of(work, typeof(*dev_priv),
+ gt_pm.rps.vlv_rps_work.work);
+ struct intel_rps * const rps = &dev_priv->gt_pm.rps;
+ u8 next;
+
+ mutex_lock(&dev_priv->pcu_lock);
+
+ if (rps->target_freq == rps->cur_freq) {
+ pm_qos_update_request(&dev_priv->gt_pm.rps.qos,
+ PM_QOS_DEFAULT_VALUE);
+ mutex_unlock(&dev_priv->pcu_lock);
+ return;
+ }
+
+ if (rps->target_freq > rps->cur_freq) {
+ u8 diff = rps->target_freq - rps->cur_freq;
+
+ if (diff > 1)
+ diff = diff >> 1;
+
+ next = rps->cur_freq + diff;
+ } else {
+ next = rps->target_freq;
+ }
+
+ valleyview_set_hw_rps(dev_priv, next);
+ mutex_unlock(&dev_priv->pcu_lock);
+
+ valleyview_queue_rps_change(dev_priv);
+}
+
/* vlv_set_rps_idle: Set the frequency to idle, if Gfx clocks are down
*
* * If Gfx is Idle, then
@@ -6237,9 +6300,11 @@ static void vlv_set_rps_idle(struct drm_i915_private *dev_priv)
u32 val = rps->idle_freq;
int err;
- if (rps->cur_freq <= val)
+ if (IS_CHERRYVIEW(dev_priv) && rps->cur_freq <= val)
return;
+ rps->target_freq = val;
+
/* The punit delays the write of the frequency and voltage until it
* determines the GPU is awake. During normal usage we don't want to
* waste power changing the frequency if the GPU is sleeping (rc6).
@@ -9336,6 +9401,9 @@ int intel_freq_opcode(struct drm_i915_private *dev_priv, int val)
void intel_pm_setup(struct drm_i915_private *dev_priv)
{
+ INIT_DELAYED_WORK(&dev_priv->gt_pm.rps.vlv_rps_work,
+ vlv_rps_write_work);
+
mutex_init(&dev_priv->pcu_lock);
atomic_set(&dev_priv->gt_pm.rps.num_waiters, 0);
--
2.14.1
More information about the Intel-gfx
mailing list