Re: [PATCH 2/3] drm/i915: Generalize drain latency computation

["Ville Syrjälä" <ville.syrjala@linux.intel.com>]

On Wed, Jul 16, 2014 at 06:24:04PM +0530, Gajanan Bhat wrote:
> Modify drain latency computation to use it for any plane. Same function can be
> used for primary, cursor and sprite planes.
> 
> Signed-off-by: Gajanan Bhat <gajanan.bhat@intel.com>
> ---
>  drivers/gpu/drm/i915/i915_reg.h |    1 +
>  drivers/gpu/drm/i915/intel_pm.c |   82 ++++++++++++++++++++++-----------------
>  2 files changed, 47 insertions(+), 36 deletions(-)
> 
> diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h
> index d2a220b..a1260a2 100644
> --- a/drivers/gpu/drm/i915/i915_reg.h
> +++ b/drivers/gpu/drm/i915/i915_reg.h
> @@ -3877,6 +3877,7 @@ enum punit_power_well {
>  #define DDL_PLANE_PRECISION_64		(1<<7)
>  #define DDL_PLANE_PRECISION_32		(0<<7)
>  #define DDL_PLANE_SHIFT			0
> +#define DRAIN_LATENCY_MAX		0x7f
>  
>  /* FIFO watermark sizes etc */
>  #define G4X_FIFO_LINE_SIZE	64
> diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c
> index 90df1e8..f3a3e90 100644
> --- a/drivers/gpu/drm/i915/intel_pm.c
> +++ b/drivers/gpu/drm/i915/intel_pm.c
> @@ -1268,33 +1268,21 @@ static bool g4x_compute_srwm(struct drm_device *dev,
>  			      display, cursor);
>  }
>  
> -static bool vlv_compute_drain_latency(struct drm_device *dev,
> -				     int plane,
> -				     int *plane_prec_mult,
> -				     int *plane_dl,
> -				     int *cursor_prec_mult,
> -				     int *cursor_dl)
> +static bool vlv_compute_drain_latency(struct drm_crtc *crtc,
> +				      int pixel_size,
> +				      int *prec_mult,
> +				      int *drain_latency)
>  {
> -	struct drm_crtc *crtc;
> -	int clock, pixel_size;
>  	int entries;
> +	int clock = to_intel_crtc(crtc)->config.adjusted_mode.crtc_clock;
>  
> -	crtc = intel_get_crtc_for_plane(dev, plane);
> -	if (!intel_crtc_active(crtc))
> +	if (clock == 0 || pixel_size == 0)
>  		return false;
>  
> -	clock = to_intel_crtc(crtc)->config.adjusted_mode.crtc_clock;
> -	pixel_size = crtc->primary->fb->bits_per_pixel / 8;	/* BPP */
> -
> -	entries = (clock / 1000) * pixel_size;
> -	*plane_prec_mult = (entries > 128) ?
> -		DRAIN_LATENCY_PRECISION_64 : DRAIN_LATENCY_PRECISION_32;
> -	*plane_dl = (64 * (*plane_prec_mult) * 4) / entries;
> -
> -	entries = (clock / 1000) * 4;	/* BPP is always 4 for cursor */
> -	*cursor_prec_mult = (entries > 128) ?
> -		DRAIN_LATENCY_PRECISION_64 : DRAIN_LATENCY_PRECISION_32;
> -	*cursor_dl = (64 * (*cursor_prec_mult) * 4) / entries;
> +	entries = DIV_ROUND_UP(clock, 1000) * pixel_size;

Yeah rounding the clock up should be safe, rounding down not so much.
But this change should really be a separate patch.

> +	*prec_mult = (entries > 128) ? DRAIN_LATENCY_PRECISION_64 :
> +				       DRAIN_LATENCY_PRECISION_32;
> +	*drain_latency = (64 * (*prec_mult) * 4) / entries;

Another thing you could fix is making sure the result is <=0x7f. It's
quite possible we'd exceed that (all that's needed is a 16bpp fb and
<32MHz pixel clock, which isn't all that far fetched). This could
be part of the rounding fix patch since it's somewhat related.

>  
>  	return true;
>  }
> @@ -1309,24 +1297,46 @@ static bool vlv_compute_drain_latency(struct drm_device *dev,
>  
>  static void vlv_update_drain_latency(struct drm_crtc *crtc)
>  {
> -	struct drm_device *dev = crtc->dev;
> -	struct drm_i915_private *dev_priv = dev->dev_private;
> +	struct drm_i915_private *dev_priv = crtc->dev->dev_private;
> +	int pixel_size;
> +	int drain_latency;
>  	enum pipe pipe = to_intel_crtc(crtc)->pipe;
> -	int plane_prec, plane_dl;
> -	int cursor_prec, cursor_dl;
> -	int plane_prec_mult, cursor_prec_mult;
> +	int plane_prec, prec_mult, plane_dl;
>  
> -	if (vlv_compute_drain_latency(dev, pipe, &plane_prec_mult, &plane_dl,
> -				      &cursor_prec_mult, &cursor_dl)) {
> -		cursor_prec = (cursor_prec_mult == DRAIN_LATENCY_PRECISION_64) ?
> -			DDL_CURSOR_PRECISION_64 : DDL_CURSOR_PRECISION_32;
> -		plane_prec = (plane_prec_mult == DRAIN_LATENCY_PRECISION_64) ?
> -			DDL_PLANE_PRECISION_64 : DDL_PLANE_PRECISION_32;
> +	plane_dl = I915_READ(VLV_DDL(pipe)) & ~DDL_PLANE_PRECISION_64 &
> +		   ~DRAIN_LATENCY_MAX & ~DDL_CURSOR_PRECISION_64 &
> +		   ~(DRAIN_LATENCY_MAX << DDL_CURSOR_SHIFT);
>  
> -		I915_WRITE(VLV_DDL(pipe), cursor_prec |
> -			   (cursor_dl << DDL_CURSOR_SHIFT) |
> -			   plane_prec | (plane_dl << DDL_PLANE_SHIFT));
> +	if (!intel_crtc_active(crtc)) {
> +		I915_WRITE(VLV_DDL(pipe), plane_dl);
> +		return;
>  	}
> +
> +	/* Primary plane Drain Latency */
> +	pixel_size = crtc->primary->fb->bits_per_pixel / 8;	/* BPP */
> +	if (vlv_compute_drain_latency(crtc, pixel_size, &prec_mult, &drain_latency)) {
> +		plane_prec = (prec_mult == DRAIN_LATENCY_PRECISION_64) ?
> +					   DDL_PLANE_PRECISION_64 :
> +					   DDL_PLANE_PRECISION_32;
> +		plane_dl = plane_dl | plane_prec | drain_latency;
> +	}
> +
> +	/* Cursor Drain Latency
> +	 * BPP is always 4 for cursor
> +	 */
> +	pixel_size = 4;
> +
> +	/* Program cursor DL only if it is enabled */
> +	if ((I915_READ(CURCNTR(pipe)) & CURSOR_MODE) &&
> +	    vlv_compute_drain_latency(crtc, pixel_size, &prec_mult, &drain_latency)) {
> +		plane_prec = (prec_mult == DRAIN_LATENCY_PRECISION_64) ?
> +					   DDL_CURSOR_PRECISION_64 :
> +					   DDL_CURSOR_PRECISION_32;
> +		plane_dl = plane_dl | plane_prec |
> +			   drain_latency << DDL_CURSOR_SHIFT;
> +	}
> +
> +	I915_WRITE(VLV_DDL(pipe), plane_dl);
>  }
>  
>  #define single_plane_enabled(mask) is_power_of_2(mask)
> -- 
> 1.7.9.5
> 
> _______________________________________________
> Intel-gfx mailing list
> Intel-gfx@lists.freedesktop.org
> http://lists.freedesktop.org/mailman/listinfo/intel-gfx

-- 
Ville Syrjälä
Intel OTC