On Mon, Apr 27, 2015 at 01:41:24PM +0100, Chris Wilson wrote:
> If the client stalls on a congested request, chosen to be 20ms old to
> match throttling, allow the client a free RPS boost.
>
> Signed-off-by: Chris Wilson <ch...@chris-wilson.co.uk>
> ---
> drivers/gpu/drm/i915/i915_gem.c | 2 +-
> drivers/gpu/drm/i915/intel_drv.h | 3 ++-
> drivers/gpu/drm/i915/intel_pm.c | 19 +++++++++++++++----
> 3 files changed, 18 insertions(+), 6 deletions(-)
>
> diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
> index 43ac75834e61..83bccb9f62d6 100644
> --- a/drivers/gpu/drm/i915/i915_gem.c
> +++ b/drivers/gpu/drm/i915/i915_gem.c
> @@ -1247,7 +1247,7 @@ int __i915_wait_request(struct drm_i915_gem_request
> *req,
> jiffies + nsecs_to_jiffies_timeout((u64)*timeout) : 0;
>
> if (INTEL_INFO(dev_priv)->gen >= 6)
> - gen6_rps_boost(dev_priv, rps);
> + gen6_rps_boost(dev_priv, rps, req->emitted_jiffies);
>
> /* Record current time in case interrupted by signal, or wedged */
> trace_i915_gem_request_wait_begin(req);
> diff --git a/drivers/gpu/drm/i915/intel_drv.h
> b/drivers/gpu/drm/i915/intel_drv.h
> index 9eb0a911343a..34cfa61f3321 100644
> --- a/drivers/gpu/drm/i915/intel_drv.h
> +++ b/drivers/gpu/drm/i915/intel_drv.h
> @@ -1361,7 +1361,8 @@ void gen6_rps_busy(struct drm_i915_private *dev_priv);
> void gen6_rps_reset_ei(struct drm_i915_private *dev_priv);
> void gen6_rps_idle(struct drm_i915_private *dev_priv);
> void gen6_rps_boost(struct drm_i915_private *dev_priv,
> - struct intel_rps_client *rps);
> + struct intel_rps_client *rps,
> + unsigned long submitted);
> void intel_queue_rps_boost_for_request(struct drm_device *dev,
> struct drm_i915_gem_request *rq);
> void ilk_wm_get_hw_state(struct drm_device *dev);
> diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c
> index dcc52f928650..850e02e1c7eb 100644
> --- a/drivers/gpu/drm/i915/intel_pm.c
> +++ b/drivers/gpu/drm/i915/intel_pm.c
> @@ -4122,10 +4122,17 @@ void gen6_rps_idle(struct drm_i915_private *dev_priv)
> }
>
> void gen6_rps_boost(struct drm_i915_private *dev_priv,
> - struct intel_rps_client *rps)
> + struct intel_rps_client *rps,
> + unsigned long submitted)
> {
> u32 val;
>
> + /* Force a RPS boost (and don't count it against the client) if
> + * the GPU is severely congested.
> + */
> + if (rps && time_after(jiffies, submitted + msecs_to_jiffies(20)))
A #define for these 20 ms used both here and in the throttle ioctl would
be neat, just to reinforce the connection. Can you please follow-up with
that patch?
-Daniel
> + rps = NULL;
> +
> mutex_lock(&dev_priv->rps.hw_lock);
> val = dev_priv->rps.max_freq_softlimit;
> if (dev_priv->rps.enabled &&
> @@ -6825,11 +6832,12 @@ struct request_boost {
> static void __intel_rps_boost_work(struct work_struct *work)
> {
> struct request_boost *boost = container_of(work, struct request_boost,
> work);
> + struct drm_i915_gem_request *rq = boost->rq;
>
> - if (!i915_gem_request_completed(boost->rq, true))
> - gen6_rps_boost(to_i915(boost->rq->ring->dev), NULL);
> + if (!i915_gem_request_completed(rq, true))
> + gen6_rps_boost(to_i915(rq->ring->dev), 0, rq->emitted_jiffies);
>
> - i915_gem_request_unreference__unlocked(boost->rq);
> + i915_gem_request_unreference__unlocked(rq);
> kfree(boost);
> }
>
> @@ -6841,6 +6849,9 @@ void intel_queue_rps_boost_for_request(struct
> drm_device *dev,
> if (rq == NULL || INTEL_INFO(dev)->gen < 6)
> return;
>
> + if (i915_gem_request_completed(rq, true))
> + return;
> +
> boost = kmalloc(sizeof(*boost), GFP_ATOMIC);
> if (boost == NULL)
> return;
> --
> 2.1.4
>
> _______________________________________________
> Intel-gfx mailing list
> Intel-gfx@lists.freedesktop.org
> http://lists.freedesktop.org/mailman/listinfo/intel-gfx
--
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
