Re: [Intel-gfx] [PATCH v28 4/6] drm/i915: Add TGL+ SAGV support

["Lisovskiy, Stanislav" <stanislav.lisovskiy@intel.com>]

On Tue, May 12, 2020 at 03:03:26PM +0300, Ville Syrjälä wrote:
> On Thu, May 07, 2020 at 05:45:01PM +0300, Stanislav Lisovskiy wrote:
> > Starting from TGL we need to have a separate wm0
> > values for SAGV and non-SAGV which affects
> > how calculations are done.
> > 
> > v2: Remove long lines
> > v3: Removed COLOR_PLANE enum references
> > v4, v5, v6: Fixed rebase conflict
> > v7: - Removed skl_plane_wm_level accessor from skl_allocate_pipe_ddb(Ville)
> >     - Removed sagv_uv_wm0(Ville)
> >     - can_sagv->use_sagv_wm(Ville)
> > 
> > Signed-off-by: Stanislav Lisovskiy <stanislav.lisovskiy@intel.com>
> > ---
> >  drivers/gpu/drm/i915/display/intel_display.c  |   8 +-
> >  .../drm/i915/display/intel_display_types.h    |   2 +
> >  drivers/gpu/drm/i915/intel_pm.c               | 118 +++++++++++++++++-
> >  3 files changed, 121 insertions(+), 7 deletions(-)
> > 
> > diff --git a/drivers/gpu/drm/i915/display/intel_display.c b/drivers/gpu/drm/i915/display/intel_display.c
> > index fd6d63b03489..be5741cb7595 100644
> > --- a/drivers/gpu/drm/i915/display/intel_display.c
> > +++ b/drivers/gpu/drm/i915/display/intel_display.c
> > @@ -13961,7 +13961,9 @@ static void verify_wm_state(struct intel_crtc *crtc,
> >  		/* Watermarks */
> >  		for (level = 0; level <= max_level; level++) {
> >  			if (skl_wm_level_equals(&hw_plane_wm->wm[level],
> > -						&sw_plane_wm->wm[level]))
> > +						&sw_plane_wm->wm[level]) ||
> > +			    (level == 0 && skl_wm_level_equals(&hw_plane_wm->wm[level],
> > +							       &sw_plane_wm->sagv_wm0)))
> >  				continue;
> >  
> >  			drm_err(&dev_priv->drm,
> > @@ -14016,7 +14018,9 @@ static void verify_wm_state(struct intel_crtc *crtc,
> >  		/* Watermarks */
> >  		for (level = 0; level <= max_level; level++) {
> >  			if (skl_wm_level_equals(&hw_plane_wm->wm[level],
> > -						&sw_plane_wm->wm[level]))
> > +						&sw_plane_wm->wm[level]) ||
> > +			    (level == 0 && skl_wm_level_equals(&hw_plane_wm->wm[level],
> > +							       &sw_plane_wm->sagv_wm0)))
> >  				continue;
> >  
> >  			drm_err(&dev_priv->drm,
> > diff --git a/drivers/gpu/drm/i915/display/intel_display_types.h b/drivers/gpu/drm/i915/display/intel_display_types.h
> > index 9488449e4b94..8cede29c9562 100644
> > --- a/drivers/gpu/drm/i915/display/intel_display_types.h
> > +++ b/drivers/gpu/drm/i915/display/intel_display_types.h
> > @@ -688,11 +688,13 @@ struct skl_plane_wm {
> >  	struct skl_wm_level wm[8];
> >  	struct skl_wm_level uv_wm[8];
> >  	struct skl_wm_level trans_wm;
> > +	struct skl_wm_level sagv_wm0;
> >  	bool is_planar;
> >  };
> >  
> >  struct skl_pipe_wm {
> >  	struct skl_plane_wm planes[I915_MAX_PLANES];
> > +	bool use_sagv_wm;
> >  };
> >  
> >  enum vlv_wm_level {
> > diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c
> > index db188efee21e..934a686342ad 100644
> > --- a/drivers/gpu/drm/i915/intel_pm.c
> > +++ b/drivers/gpu/drm/i915/intel_pm.c
> > @@ -3863,25 +3863,35 @@ bool intel_can_enable_sagv(struct drm_i915_private *dev_priv,
> >  	return bw_state->pipe_sagv_reject == 0;
> >  }
> >  
> > +static bool
> > +tgl_crtc_can_enable_sagv(const struct intel_crtc_state *crtc_state);
> 
> Just put the function here instead of adding fwd decalrations.
> 
> > +
> >  static int intel_compute_sagv_mask(struct intel_atomic_state *state)
> >  {
> >  	struct drm_i915_private *dev_priv = to_i915(state->base.dev);
> >  	int ret;
> >  	struct intel_crtc *crtc;
> > -	const struct intel_crtc_state *new_crtc_state;
> > +	struct intel_crtc_state *new_crtc_state;
> >  	struct intel_bw_state *new_bw_state = NULL;
> >  	const struct intel_bw_state *old_bw_state = NULL;
> >  	int i;
> >  
> >  	for_each_new_intel_crtc_in_state(state, crtc,
> >  					 new_crtc_state, i) {
> > +		bool can_sagv;
> > +
> >  		new_bw_state = intel_atomic_get_bw_state(state);
> >  		if (IS_ERR(new_bw_state))
> >  			return PTR_ERR(new_bw_state);
> >  
> >  		old_bw_state = intel_atomic_get_old_bw_state(state);
> >  
> > -		if (skl_crtc_can_enable_sagv(new_crtc_state))
> > +		if (INTEL_GEN(dev_priv) >= 12)
> > +			can_sagv = tgl_crtc_can_enable_sagv(new_crtc_state);
> > +		else
> > +			can_sagv = skl_crtc_can_enable_sagv(new_crtc_state);
> > +
> > +		if (can_sagv)
> >  			new_bw_state->pipe_sagv_reject &= ~BIT(crtc->pipe);
> >  		else
> >  			new_bw_state->pipe_sagv_reject |= BIT(crtc->pipe);
> > @@ -3899,6 +3909,24 @@ static int intel_compute_sagv_mask(struct intel_atomic_state *state)
> >  			return ret;
> >  	}
> >  
> > +	for_each_new_intel_crtc_in_state(state, crtc,
> > +					 new_crtc_state, i) {
> > +		struct skl_pipe_wm *pipe_wm = &new_crtc_state->wm.skl.optimal;
> > +
> > +		/*
> > +		 * Due to drm limitation at commit state, when
> > +		 * changes are written the whole atomic state is
> > +		 * zeroed away => which prevents from using it,
> > +		 * so just sticking it into pipe wm state for
> > +		 * keeping it simple - anyway this is related to wm.
> > +		 * Proper way in ideal universe would be of course not
> > +		 * to lose parent atomic state object from child crtc_state,
> > +		 * and stick to OOP programming principles, which had been
> > +		 * scientifically proven to work.
> > +		 */
> 
> More ramblings. Just drop this comment too imo.

As I understand what is happening here is rather weird, so I thought 
commenting is good idea.

> 
> > +		pipe_wm->use_sagv_wm = intel_can_enable_sagv(dev_priv, new_bw_state);
> 
> I think this should be 
> gen >= 12 && intel_can_enable_sagv();
> 
> > +	}
> > +
> >  	if (intel_can_enable_sagv(dev_priv, new_bw_state) != intel_can_enable_sagv(dev_priv, old_bw_state)) {
> >  		ret = intel_atomic_serialize_global_state(&new_bw_state->base);
> >  		if (ret)
> > @@ -4642,12 +4670,39 @@ skl_plane_wm_level(const struct intel_crtc_state *crtc_state,
> >  		   int level,
> >  		   int color_plane)
> >  {
> > -	const struct skl_plane_wm *wm =
> > -		&crtc_state->wm.skl.optimal.planes[plane_id];
> > +	const struct skl_pipe_wm *pipe_wm = &crtc_state->wm.skl.optimal;
> > +	const struct skl_plane_wm *wm = &pipe_wm->planes[plane_id];
> > +
> > +	if (!level) {
> > +		if (pipe_wm->use_sagv_wm)
> > +			return &wm->sagv_wm0;
> > +	}
> 
> if (level == 0 && use_sagv_wm)
> 	return sagv_wm0;
> 
> >  
> >  	return color_plane == 0 ? &wm->wm[level] : &wm->uv_wm[level];
> >  }
> >  
> > +static bool
> > +tgl_crtc_can_enable_sagv(const struct intel_crtc_state *crtc_state)
> > +{
> > +	struct intel_crtc *crtc = to_intel_crtc(crtc_state->uapi.crtc);
> > +	enum plane_id plane_id;
> > +
> > +	if (!crtc_state->hw.active)
> > +		return true;
> > +
> > +	for_each_plane_id_on_crtc(crtc, plane_id) {
> > +		const struct skl_ddb_entry *plane_alloc =
> > +			&crtc_state->wm.skl.plane_ddb_y[plane_id];
> > +		const struct skl_plane_wm *wm =
> > +			&crtc_state->wm.skl.optimal.planes[plane_id];
> > +
> > +		if (skl_ddb_entry_size(plane_alloc) < wm->sagv_wm0.min_ddb_alloc)
> > +			return false;
> > +	}
> > +
> > +	return true;
> > +}
> > +
> >  static int
> >  skl_allocate_pipe_ddb(struct intel_crtc_state *crtc_state)
> >  {
> > @@ -4684,7 +4739,6 @@ skl_allocate_pipe_ddb(struct intel_crtc_state *crtc_state)
> >  							 plane_data_rate,
> >  							 uv_plane_data_rate);
> >  
> > -
> >  	skl_ddb_get_pipe_allocation_limits(dev_priv, crtc_state, total_data_rate,
> >  					   alloc, &num_active);
> >  	alloc_size = skl_ddb_entry_size(alloc);
> > @@ -5219,6 +5273,37 @@ skl_compute_wm_levels(const struct intel_crtc_state *crtc_state,
> >  	}
> >  }
> >  
> > +static void tgl_compute_sagv_wm(const struct intel_crtc_state *crtc_state,
> > +				const struct skl_wm_params *wm_params,
> > +				struct skl_plane_wm *plane_wm)
> > +{
> > +	struct drm_i915_private *dev_priv = to_i915(crtc_state->uapi.crtc->dev);
> > +	struct skl_wm_level *sagv_wm = &plane_wm->sagv_wm0;
> > +	struct skl_wm_level *levels = plane_wm->wm;
> > +
> > +	/*
> > +	 * For Gen12 if it is an L0 we need to also
> > +	 * consider sagv_block_time when calculating
> > +	 * L0 watermark - we will need that when making
> > +	 * a decision whether enable SAGV or not.
> > +	 * For older gens we agreed to copy L0 value for
> > +	 * compatibility.
> > +	 */
> > +	if ((INTEL_GEN(dev_priv) >= 12)) {
> 
> Drop this if-else and only call tgl_compute_sagv_wm() on tgl+.
> The comment can then go as well I think.

We actually had _agreed_ with you personally that I do copy
wm0 values for other platforms for compatibility reasons.
Now after one month you say that we are going to call this
only for gen12. 

> 
> > +		u32 latency = dev_priv->wm.skl_latency[0];
> > +
> > +		latency += dev_priv->sagv_block_time_us;
> > +		skl_compute_plane_wm(crtc_state, 0, latency,
> > +				     wm_params, &levels[0],
> > +				     sagv_wm);
> > +		DRM_DEBUG_KMS("%d L0 blocks required for SAGV vs %d for non-SAGV\n",
> > +			      sagv_wm->min_ddb_alloc, levels[0].min_ddb_alloc);
> 
> Leftover debugs. Pls remove.

I thought this might be useful, we should have some understanding what
is happening, i.e what was the reasoning and why SAGV was switched on/off.
Not crucial, but really don't understand why this is _that_ harmful.
We have plenty of wm/ddb debugs anyway, but no way to figure out how SAGV logic is working.

> 
> > +	} else {
> > +		/* Since all members are POD */
> > +		*sagv_wm = levels[0];
> > +	}
> > +}
> > +
> >  static void skl_compute_transition_wm(const struct intel_crtc_state *crtc_state,
> >  				      const struct skl_wm_params *wp,
> >  				      struct skl_plane_wm *wm)
> > @@ -5296,6 +5381,8 @@ static int skl_build_plane_wm_single(struct intel_crtc_state *crtc_state,
> >  		return ret;
> >  
> >  	skl_compute_wm_levels(crtc_state, &wm_params, wm->wm);
> > +	if (color_plane == 0)
> 
> We do want this for both the Y plane and UV plane.
> 
> > +		tgl_compute_sagv_wm(crtc_state, &wm_params, wm);
> >  	skl_compute_transition_wm(crtc_state, &wm_params, wm);
> >  
> >  	return 0;
> > @@ -5702,6 +5789,12 @@ skl_print_wm_changes(struct intel_atomic_state *state)
> >  				    enast(new_wm->wm[7].ignore_lines), new_wm->wm[7].plane_res_l,
> >  				    enast(new_wm->trans_wm.ignore_lines), new_wm->trans_wm.plane_res_l);
> >  
> > +			drm_dbg_kms(&dev_priv->drm,
> > +				    "[PLANE:%d:%s] sagv wm0 lines %4d -> %4d\n",
> > +				    plane->base.base.id, plane->base.name,
> > +				    old_wm->sagv_wm0.plane_res_l,
> > +				    new_wm->sagv_wm0.plane_res_l);
> > +
> 
> IIRC I already suggested just slapping these onto the existing debug
> prints instead of bloating the debug output with extra lines.
> 
> Ie instead of 
> wms ... -> ...
> sagv wm ... -> ...
> 
> we just get
> wms ...,old_sagv_wm0 -> ...,new_sagv_wm0
> 
> just like we already do for eg. the transition wm.

Not sure about this one. I will check, I thought you suggested the opposite,
i.e do those debugs separately.

Stan
> 
> >  			drm_dbg_kms(&dev_priv->drm,
> >  				    "[PLANE:%d:%s]  blocks %4d,%4d,%4d,%4d,%4d,%4d,%4d,%4d,%4d"
> >  				    " -> %4d,%4d,%4d,%4d,%4d,%4d,%4d,%4d,%4d\n",
> > @@ -5717,6 +5810,12 @@ skl_print_wm_changes(struct intel_atomic_state *state)
> >  				    new_wm->wm[6].plane_res_b, new_wm->wm[7].plane_res_b,
> >  				    new_wm->trans_wm.plane_res_b);
> >  
> > +			drm_dbg_kms(&dev_priv->drm,
> > +				    "[PLANE:%d:%s] sagv wm0 blocks %4d -> %4d\n",
> > +				    plane->base.base.id, plane->base.name,
> > +				    old_wm->sagv_wm0.plane_res_b,
> > +				    new_wm->sagv_wm0.plane_res_b);
> > +
> >  			drm_dbg_kms(&dev_priv->drm,
> >  				    "[PLANE:%d:%s] min_ddb %4d,%4d,%4d,%4d,%4d,%4d,%4d,%4d,%4d"
> >  				    " -> %4d,%4d,%4d,%4d,%4d,%4d,%4d,%4d,%4d\n",
> > @@ -5731,6 +5830,12 @@ skl_print_wm_changes(struct intel_atomic_state *state)
> >  				    new_wm->wm[4].min_ddb_alloc, new_wm->wm[5].min_ddb_alloc,
> >  				    new_wm->wm[6].min_ddb_alloc, new_wm->wm[7].min_ddb_alloc,
> >  				    new_wm->trans_wm.min_ddb_alloc);
> > +
> > +			drm_dbg_kms(&dev_priv->drm,
> > +				    "[PLANE:%d:%s] sagv wm0 min ddb %4d -> %4d\n",
> > +				    plane->base.base.id, plane->base.name,
> > +				    old_wm->sagv_wm0.min_ddb_alloc,
> > +				    new_wm->sagv_wm0.min_ddb_alloc);
> >  		}
> >  	}
> >  }
> > @@ -6023,6 +6128,9 @@ void skl_pipe_wm_get_hw_state(struct intel_crtc *crtc,
> >  			skl_wm_level_from_reg_val(val, &wm->wm[level]);
> >  		}
> >  
> > +		memcpy(&wm->sagv_wm0, &wm->wm[0],
> > +		       sizeof(struct skl_wm_level));
> 
> A simple assignment should suffice?

Agree.

> 
> > +
> >  		if (plane_id != PLANE_CURSOR)
> >  			val = I915_READ(PLANE_WM_TRANS(pipe, plane_id));
> >  		else
> > -- 
> > 2.24.1.485.gad05a3d8e5
> 
> -- 
> Ville Syrjälä
> Intel
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx