[Intel-gfx] [RFC v3 4/9] drm/i915/tgl: Gen-12 render decompression

Matt Roper matthew.d.roper at intel.com
Wed Oct 2 22:32:41 UTC 2019


On Mon, Sep 23, 2019 at 03:29:30AM -0700, Dhinakaran Pandiyan wrote:
> Gen-12 display decompression operates on Y-tiled compressed main surface.
> The CCS is linear

I'd mention in the commit message that we opt to to treat the
CCS as if it were a bunch of 64x1 tiles.

There are a handful of cosmetic changes suggested below, but otherwise,

Reviewed-by: Matt Roper <matthew.d.roper at intel.com>

> and has 4 bits of metadata for each main surface cache
> line pair, a size ratio of 1:256. Gen-12 display decompression is
> incompatible with buffers compressed by earlier GPUs, so make use of a new
> modifier to identify gen-12 compression. Another notable change is that
> render decompression is supported on all planes except cursor and on all
> pipes. Start by adding render decompression support for [A,X]BGR888 pixel
> formats.
> 
> v2: Fix checkpatch warnings (Lucas)
> v3:
> Rebase, disable color clear, styling changes and modify
> intel_tile_width_bytes and intel_tile_height to handle linear CCS
> 
> Cc: Ville Syrjälä <ville.syrjala at linux.intel.com>
> Cc: Matt Roper <matthew.d.roper at intel.com>
> Cc: Nanley G Chery <nanley.g.chery at intel.com>
> Cc: Jason Ekstrand <jason at jlekstrand.net>
> Signed-off-by: Dhinakaran Pandiyan <dhinakaran.pandiyan at intel.com>
> Signed-off-by: Lucas De Marchi <lucas.demarchi at intel.com>
> ---
>  drivers/gpu/drm/i915/display/intel_display.c | 85 ++++++++++++++++----
>  drivers/gpu/drm/i915/display/intel_sprite.c  | 23 ++++--
>  drivers/gpu/drm/i915/i915_reg.h              |  1 +
>  3 files changed, 84 insertions(+), 25 deletions(-)
> 
> diff --git a/drivers/gpu/drm/i915/display/intel_display.c b/drivers/gpu/drm/i915/display/intel_display.c
> index c437f00c2072..6fec43cdddf4 100644
> --- a/drivers/gpu/drm/i915/display/intel_display.c
> +++ b/drivers/gpu/drm/i915/display/intel_display.c
> @@ -1911,6 +1911,10 @@ intel_tile_width_bytes(const struct drm_framebuffer *fb, int color_plane)
>  		if (color_plane == 1)
>  			return 128;
>  		/* fall through */
> +	case I915_FORMAT_MOD_Y_TILED_GEN12_RC_CCS:
> +		if (color_plane == 1)
> +			return 64;
> +		/* fall through */
>  	case I915_FORMAT_MOD_Y_TILED:
>  		if (IS_GEN(dev_priv, 2) || HAS_128_BYTE_Y_TILING(dev_priv))
>  			return 128;
> @@ -1944,8 +1948,15 @@ intel_tile_width_bytes(const struct drm_framebuffer *fb, int color_plane)
>  static unsigned int
>  intel_tile_height(const struct drm_framebuffer *fb, int color_plane)
>  {
> -	return intel_tile_size(to_i915(fb->dev)) /
> -		intel_tile_width_bytes(fb, color_plane);
> +	switch (fb->modifier) {
> +	case I915_FORMAT_MOD_Y_TILED_GEN12_RC_CCS:
> +		if (color_plane == 1)
> +			return 1;
> +		/* fall through */
> +	default:
> +		return intel_tile_size(to_i915(fb->dev)) /
> +			intel_tile_width_bytes(fb, color_plane);
> +	}

I'd just use an 'if' instead of a 'switch' here since I can't see us
ending up with a bunch of different cases in the future.  Either we have
real tiles that follow normal rules or else we have pretend tiles for
stuff like CCS that have a height of 1.

>  }
>  
>  /* Return the tile dimensions in pixel units */
> @@ -2044,6 +2055,8 @@ static unsigned int intel_surf_alignment(const struct drm_framebuffer *fb,
>  		if (INTEL_GEN(dev_priv) >= 9)
>  			return 256 * 1024;
>  		return 0;
> +	case I915_FORMAT_MOD_Y_TILED_GEN12_RC_CCS:
> +		return 16 * 1024;
>  	case I915_FORMAT_MOD_Y_TILED_CCS:
>  	case I915_FORMAT_MOD_Yf_TILED_CCS:
>  	case I915_FORMAT_MOD_Y_TILED:
> @@ -2243,7 +2256,8 @@ static u32 intel_adjust_tile_offset(int *x, int *y,
>  
>  static bool is_surface_linear(u64 modifier, int color_plane)
>  {
> -	return modifier == DRM_FORMAT_MOD_LINEAR;
> +	return modifier == DRM_FORMAT_MOD_LINEAR ||
> +	       (modifier == I915_FORMAT_MOD_Y_TILED_GEN12_RC_CCS && color_plane == 1);

Long line; needs a split to stay under 80 chars.

>  }
>  
>  static u32 intel_adjust_aligned_offset(int *x, int *y,
> @@ -2430,6 +2444,7 @@ static unsigned int intel_fb_modifier_to_tiling(u64 fb_modifier)
>  		return I915_TILING_X;
>  	case I915_FORMAT_MOD_Y_TILED:
>  	case I915_FORMAT_MOD_Y_TILED_CCS:
> +	case I915_FORMAT_MOD_Y_TILED_GEN12_RC_CCS:
>  		return I915_TILING_Y;
>  	default:
>  		return I915_TILING_NONE;
> @@ -2450,7 +2465,7 @@ static unsigned int intel_fb_modifier_to_tiling(u64 fb_modifier)
>   * us a ratio of one byte in the CCS for each 8x16 pixels in the
>   * main surface.
>   */
> -static const struct drm_format_info ccs_formats[] = {
> +static const struct drm_format_info skl_ccs_formats[] = {
>  	{ .format = DRM_FORMAT_XRGB8888, .depth = 24, .num_planes = 2,
>  	  .cpp = { 4, 1, }, .hsub = 8, .vsub = 16, },
>  	{ .format = DRM_FORMAT_XBGR8888, .depth = 24, .num_planes = 2,
> @@ -2461,6 +2476,24 @@ static const struct drm_format_info ccs_formats[] = {
>  	  .cpp = { 4, 1, }, .hsub = 8, .vsub = 16, .has_alpha = true, },
>  };
>  
> +/*
> + * Gen-12 compression uses 4 bits of CCS data for each cache line pair in the
> + * main surface. And each 64B CCS cache line represents an area of 4x1 Y-tiles
> + * in the main surface. With 4 byte pixels and each Y-tile having dimensions of
> + * 32x32 pixels, the ratio turns out to 1B in the CCS for every 2x32 pixels in
> + * the main surface.
> + */
> +static const struct drm_format_info gen12_ccs_formats[] = {
> +	{ .format = DRM_FORMAT_XRGB8888, .depth = 24, .num_planes = 2,
> +	  .cpp = { 4, 1, }, .hsub = 2, .vsub = 32, },
> +	{ .format = DRM_FORMAT_XBGR8888, .depth = 24, .num_planes = 2,
> +	  .cpp = { 4, 1, }, .hsub = 2, .vsub = 32, },
> +	{ .format = DRM_FORMAT_ARGB8888, .depth = 32, .num_planes = 2,
> +	  .cpp = { 4, 1, }, .hsub = 2, .vsub = 32, .has_alpha = true },
> +	{ .format = DRM_FORMAT_ABGR8888, .depth = 32, .num_planes = 2,
> +	  .cpp = { 4, 1, }, .hsub = 2, .vsub = 32, .has_alpha = true },
> +};
> +
>  static const struct drm_format_info *
>  lookup_format_info(const struct drm_format_info formats[],
>  		   int num_formats, u32 format)
> @@ -2481,8 +2514,12 @@ intel_get_format_info(const struct drm_mode_fb_cmd2 *cmd)
>  	switch (cmd->modifier[0]) {
>  	case I915_FORMAT_MOD_Y_TILED_CCS:
>  	case I915_FORMAT_MOD_Yf_TILED_CCS:
> -		return lookup_format_info(ccs_formats,
> -					  ARRAY_SIZE(ccs_formats),
> +		return lookup_format_info(skl_ccs_formats,
> +					  ARRAY_SIZE(skl_ccs_formats),
> +					  cmd->pixel_format);
> +	case I915_FORMAT_MOD_Y_TILED_GEN12_RC_CCS:
> +		return lookup_format_info(gen12_ccs_formats,
> +					  ARRAY_SIZE(gen12_ccs_formats),
>  					  cmd->pixel_format);
>  	default:
>  		return NULL;
> @@ -2491,7 +2528,8 @@ intel_get_format_info(const struct drm_mode_fb_cmd2 *cmd)
>  
>  bool is_ccs_modifier(u64 modifier)
>  {
> -	return modifier == I915_FORMAT_MOD_Y_TILED_CCS ||
> +	return modifier == I915_FORMAT_MOD_Y_TILED_GEN12_RC_CCS ||
> +	       modifier == I915_FORMAT_MOD_Y_TILED_CCS ||
>  	       modifier == I915_FORMAT_MOD_Yf_TILED_CCS;
>  }
>  
> @@ -2536,8 +2574,9 @@ static u32
>  intel_fb_stride_alignment(const struct drm_framebuffer *fb, int color_plane)
>  {
>  	struct drm_i915_private *dev_priv = to_i915(fb->dev);
> +	u32 tile_width;
>  
> -	if (fb->modifier == DRM_FORMAT_MOD_LINEAR) {
> +	if (is_surface_linear(fb->modifier, color_plane)) {
>  		u32 max_stride = intel_plane_fb_max_stride(dev_priv,
>  							   fb->format->format,
>  							   fb->modifier);
> @@ -2546,13 +2585,14 @@ intel_fb_stride_alignment(const struct drm_framebuffer *fb, int color_plane)
>  		 * To make remapping with linear generally feasible
>  		 * we need the stride to be page aligned.
>  		 */
> -		if (fb->pitches[color_plane] > max_stride)
> +		if (fb->pitches[color_plane] > max_stride && !is_ccs_modifier(fb->modifier))

Need to split this line to keep it under 80 chars.

>  			return intel_tile_size(dev_priv);
>  		else
>  			return 64;
> -	} else {
> -		u32 tile_width = intel_tile_width_bytes(fb, color_plane);
> +	}
>  
> +	tile_width = intel_tile_width_bytes(fb, color_plane);
> +	if (is_ccs_modifier(fb->modifier) && color_plane == 0) {
>  		/*
>  		 * Display WA #0531: skl,bxt,kbl,glk
>  		 *
> @@ -2562,12 +2602,16 @@ intel_fb_stride_alignment(const struct drm_framebuffer *fb, int color_plane)
>  		 * require the entire fb to accommodate that to avoid
>  		 * potential runtime errors at plane configuration time.
>  		 */
> -		if (IS_GEN(dev_priv, 9) && is_ccs_modifier(fb->modifier) &&
> -		    color_plane == 0 && fb->width > 3840)
> +		if (IS_GEN(dev_priv, 9) && fb->width > 3840)
> +			tile_width *= 4;
> +		/*
> +		 * The main surface pitch must be padded to a multiple of four
> +		 * tile widths.
> +		 */
> +		else if (INTEL_GEN(dev_priv) >= 12)
>  			tile_width *= 4;
> -
> -		return tile_width;
>  	}
> +	return tile_width;
>  }
>  
>  bool intel_plane_can_remap(const struct intel_plane_state *plane_state)
> @@ -2676,6 +2720,7 @@ intel_fill_fb_info(struct drm_i915_private *dev_priv,
>  			int ccs_x, ccs_y;
>  
>  			intel_tile_dims(fb, i, &tile_width, &tile_height);
> +

Unwanted change?

>  			tile_width *= hsub;
>  			tile_height *= vsub;
>  
> @@ -3972,7 +4017,7 @@ static unsigned int skl_plane_stride_mult(const struct drm_framebuffer *fb,
>  	 * The stride is either expressed as a multiple of 64 bytes chunks for
>  	 * linear buffers or in number of tiles for tiled buffers.
>  	 */
> -	if (fb->modifier == DRM_FORMAT_MOD_LINEAR)
> +	if (is_surface_linear(fb->modifier, color_plane))
>  		return 64;
>  	else if (drm_rotation_90_or_270(rotation))
>  		return intel_tile_height(fb, color_plane);
> @@ -4098,6 +4143,10 @@ static u32 skl_plane_ctl_tiling(u64 fb_modifier)
>  		return PLANE_CTL_TILED_Y;
>  	case I915_FORMAT_MOD_Y_TILED_CCS:
>  		return PLANE_CTL_TILED_Y | PLANE_CTL_RENDER_DECOMPRESSION_ENABLE;
> +	case I915_FORMAT_MOD_Y_TILED_GEN12_RC_CCS:
> +		return PLANE_CTL_TILED_Y |
> +		       PLANE_CTL_RENDER_DECOMPRESSION_ENABLE |
> +		       PLANE_CTL_CLEAR_COLOR_DISABLE;
>  	case I915_FORMAT_MOD_Yf_TILED:
>  		return PLANE_CTL_TILED_YF;
>  	case I915_FORMAT_MOD_Yf_TILED_CCS:
> @@ -9899,7 +9948,9 @@ skylake_get_initial_plane_config(struct intel_crtc *crtc,
>  	case PLANE_CTL_TILED_Y:
>  		plane_config->tiling = I915_TILING_Y;
>  		if (val & PLANE_CTL_RENDER_DECOMPRESSION_ENABLE)
> -			fb->modifier = I915_FORMAT_MOD_Y_TILED_CCS;
> +			fb->modifier = INTEL_GEN(dev_priv) >= 12 ?
> +				I915_FORMAT_MOD_Y_TILED_GEN12_RC_CCS :
> +				I915_FORMAT_MOD_Y_TILED_CCS;
>  		else
>  			fb->modifier = I915_FORMAT_MOD_Y_TILED;
>  		break;
> diff --git a/drivers/gpu/drm/i915/display/intel_sprite.c b/drivers/gpu/drm/i915/display/intel_sprite.c
> index 7a7078d0ba23..866d25d38d04 100644
> --- a/drivers/gpu/drm/i915/display/intel_sprite.c
> +++ b/drivers/gpu/drm/i915/display/intel_sprite.c
> @@ -535,6 +535,7 @@ skl_program_plane(struct intel_plane *plane,
>  	const struct drm_intel_sprite_colorkey *key = &plane_state->ckey;
>  	u32 surf_addr = plane_state->color_plane[color_plane].offset;
>  	u32 stride = skl_plane_stride(plane_state, color_plane);
> +	u32 aux_dist = plane_state->color_plane[1].offset - surf_addr;
>  	u32 aux_stride = skl_plane_stride(plane_state, 1);
>  	int crtc_x = plane_state->base.dst.x1;
>  	int crtc_y = plane_state->base.dst.y1;
> @@ -576,8 +577,10 @@ skl_program_plane(struct intel_plane *plane,
>  	I915_WRITE_FW(PLANE_STRIDE(pipe, plane_id), stride);
>  	I915_WRITE_FW(PLANE_POS(pipe, plane_id), (crtc_y << 16) | crtc_x);
>  	I915_WRITE_FW(PLANE_SIZE(pipe, plane_id), (src_h << 16) | src_w);
> -	I915_WRITE_FW(PLANE_AUX_DIST(pipe, plane_id),
> -		      (plane_state->color_plane[1].offset - surf_addr) | aux_stride);
> +
> +	if (INTEL_GEN(dev_priv) < 12)
> +		aux_dist |= aux_stride;
> +	I915_WRITE_FW(PLANE_AUX_DIST(pipe, plane_id), aux_dist);
>  
>  	if (icl_is_hdr_plane(dev_priv, plane_id)) {
>  		u32 cus_ctl = 0;
> @@ -1733,7 +1736,8 @@ static int skl_plane_check_fb(const struct intel_crtc_state *crtc_state,
>  	    (fb->modifier == I915_FORMAT_MOD_Y_TILED ||
>  	     fb->modifier == I915_FORMAT_MOD_Yf_TILED ||
>  	     fb->modifier == I915_FORMAT_MOD_Y_TILED_CCS ||
> -	     fb->modifier == I915_FORMAT_MOD_Yf_TILED_CCS)) {
> +	     fb->modifier == I915_FORMAT_MOD_Yf_TILED_CCS ||
> +	     fb->modifier == I915_FORMAT_MOD_Y_TILED_GEN12_RC_CCS)) {
>  		DRM_DEBUG_KMS("Y/Yf tiling not supported in IF-ID mode\n");
>  		return -EINVAL;
>  	}
> @@ -2145,7 +2149,8 @@ static const u64 skl_plane_format_modifiers_ccs[] = {
>  	DRM_FORMAT_MOD_INVALID
>  };
>  
> -static const u64 gen12_plane_format_modifiers_noccs[] = {
> +static const u64 gen12_plane_format_modifiers_ccs[] = {
> +	I915_FORMAT_MOD_Y_TILED_GEN12_RC_CCS,
>  	I915_FORMAT_MOD_Y_TILED,
>  	I915_FORMAT_MOD_X_TILED,
>  	DRM_FORMAT_MOD_LINEAR,
> @@ -2307,6 +2312,7 @@ static bool gen12_plane_format_mod_supported(struct drm_plane *_plane,
>  	case DRM_FORMAT_MOD_LINEAR:
>  	case I915_FORMAT_MOD_X_TILED:
>  	case I915_FORMAT_MOD_Y_TILED:
> +	case I915_FORMAT_MOD_Y_TILED_GEN12_RC_CCS:
>  		break;
>  	default:
>  		return false;
> @@ -2317,6 +2323,9 @@ static bool gen12_plane_format_mod_supported(struct drm_plane *_plane,
>  	case DRM_FORMAT_XBGR8888:
>  	case DRM_FORMAT_ARGB8888:
>  	case DRM_FORMAT_ABGR8888:
> +		if (is_ccs_modifier(modifier))
> +			return true;
> +		/* fall through */
>  	case DRM_FORMAT_RGB565:
>  	case DRM_FORMAT_XRGB2101010:
>  	case DRM_FORMAT_XBGR2101010:
> @@ -2525,13 +2534,11 @@ skl_universal_plane_create(struct drm_i915_private *dev_priv,
>  		formats = skl_get_plane_formats(dev_priv, pipe,
>  						plane_id, &num_formats);
>  
> +	plane->has_ccs = skl_plane_has_ccs(dev_priv, pipe, plane_id);
>  	if (INTEL_GEN(dev_priv) >= 12) {
> -		/* TODO: Implement support for gen-12 CCS modifiers */
> -		plane->has_ccs = false;
> -		modifiers = gen12_plane_format_modifiers_noccs;
> +		modifiers = gen12_plane_format_modifiers_ccs;
>  		plane_funcs = &gen12_plane_funcs;
>  	} else {
> -		plane->has_ccs = skl_plane_has_ccs(dev_priv, pipe, plane_id);
>  		if (plane->has_ccs)
>  			modifiers = skl_plane_format_modifiers_ccs;
>  		else
> diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h
> index 6ecb64c042ef..b465eae11763 100644
> --- a/drivers/gpu/drm/i915/i915_reg.h
> +++ b/drivers/gpu/drm/i915/i915_reg.h
> @@ -6688,6 +6688,7 @@ enum {
>  #define   PLANE_CTL_YUV422_VYUY			(3 << 16)
>  #define   PLANE_CTL_RENDER_DECOMPRESSION_ENABLE	(1 << 15)
>  #define   PLANE_CTL_TRICKLE_FEED_DISABLE	(1 << 14)
> +#define   PLANE_CTL_CLEAR_COLOR_DISABLE		(1 << 13) /* TGL+ */
>  #define   PLANE_CTL_PLANE_GAMMA_DISABLE		(1 << 13) /* Pre-GLK */
>  #define   PLANE_CTL_TILED_MASK			(0x7 << 10)
>  #define   PLANE_CTL_TILED_LINEAR		(0 << 10)
> -- 
> 2.17.1
> 

-- 
Matt Roper
Graphics Software Engineer
VTT-OSGC Platform Enablement
Intel Corporation
(916) 356-2795


More information about the Intel-gfx mailing list