Intel-GFX Archive on lore.kernel.org
 help / color / mirror / Atom feed
From: "Hogander, Jouni" <jouni.hogander@intel.com>
To: "intel-xe@lists.freedesktop.org" <intel-xe@lists.freedesktop.org>,
	"Nautiyal, Ankit K" <ankit.k.nautiyal@intel.com>,
	"intel-gfx@lists.freedesktop.org"
	<intel-gfx@lists.freedesktop.org>
Cc: "ville.syrjala@linux.intel.com" <ville.syrjala@linux.intel.com>,
	"Manna, Animesh" <animesh.manna@intel.com>
Subject: Re: [PATCH 08/10] drm/i915/psr: Check if final vblank is sufficient for PSR features
Date: Wed, 15 Oct 2025 08:23:48 +0000	[thread overview]
Message-ID: <6aa087759e3b2467ebe47d25309afdc56a4e986b.camel@intel.com> (raw)
In-Reply-To: <20251015072217.1710717-9-ankit.k.nautiyal@intel.com>

On Wed, 2025-10-15 at 12:52 +0530, Ankit Nautiyal wrote:
> Currently, wake line latency checks rely on the vblank length,
> which does not account for either the extra vblank delay for icl/tgl
> or for
> the optimized guardband which will come into picture later at some
> point.
> 
> Validate whether the final vblank (with extra vblank delay) or
> guardband
> is sufficient to support wake line latencies required by Panel Replay
> and
> PSR2 selective update. Disable the PSR features if their wake
> requirements
> cannot be accomodated.
> 
> v2: Add comments clarifying wake line checks and rationale for not
>     resetting SCL. (Jouni)
> v3: Reset other psr flags based on features that are dropped. (Jouni)
> v4: Update commit message.
> 
> Signed-off-by: Ankit Nautiyal <ankit.k.nautiyal@intel.com>
> Cc: Animesh Manna <animesh.manna@intel.com>
> Cc: Jouni Högander <jouni.hogander@intel.com>
> ---
>  drivers/gpu/drm/i915/display/intel_psr.c | 78 +++++++++++++++++++++-
> --
>  1 file changed, 70 insertions(+), 8 deletions(-)
> 
> diff --git a/drivers/gpu/drm/i915/display/intel_psr.c
> b/drivers/gpu/drm/i915/display/intel_psr.c
> index 383e6dc1ed63..816ad67e33cd 100644
> --- a/drivers/gpu/drm/i915/display/intel_psr.c
> +++ b/drivers/gpu/drm/i915/display/intel_psr.c
> @@ -1405,6 +1405,20 @@ int _intel_psr_min_set_context_latency(const
> struct intel_crtc_state *crtc_state
>  		return 1;
>  }
>  
> +static bool _wake_lines_fit_into_vblank(const struct
> intel_crtc_state *crtc_state,
> +					int vblank,
> +					int wake_lines)
> +{
> +	if (crtc_state->req_psr2_sdp_prior_scanline)
> +		vblank -= 1;
> +
> +	/* Vblank >= PSR2_CTL Block Count Number maximum line count
> */
> +	if (vblank < wake_lines)
> +		return false;
> +
> +	return true;
> +}
> +
>  static bool wake_lines_fit_into_vblank(struct intel_dp *intel_dp,
>  				       const struct intel_crtc_state
> *crtc_state,
>  				       bool aux_less,
> @@ -1428,14 +1442,16 @@ static bool wake_lines_fit_into_vblank(struct
> intel_dp *intel_dp,
>  					       crtc_state-
> >alpm_state.fast_wake_lines) :
>  			crtc_state->alpm_state.io_wake_lines;
>  
> -	if (crtc_state->req_psr2_sdp_prior_scanline)
> -		vblank -= 1;
> -
> -	/* Vblank >= PSR2_CTL Block Count Number maximum line count
> */
> -	if (vblank < wake_lines)
> -		return false;
> -
> -	return true;
> +	/*
> +	 * Guardband has not been computed yet, so we conservatively
> check if the
> +	 * full vblank duration is sufficient to accommodate wake
> line requirements
> +	 * for PSR features like Panel Replay and Selective Update.
> +	 *
> +	 * Once the actual guardband is available, a more accurate
> validation is
> +	 * performed in intel_psr_compute_config_late(), and PSR
> features are
> +	 * disabled if wake lines exceed the available guardband.
> +	 */
> +	return _wake_lines_fit_into_vblank(crtc_state, vblank,
> wake_lines);
>  }
>  
>  static bool alpm_config_valid(struct intel_dp *intel_dp,
> @@ -4351,6 +4367,52 @@ void intel_psr_compute_config_late(struct
> intel_dp *intel_dp,
>  				   struct intel_crtc_state
> *crtc_state)
>  {
>  	struct intel_display *display = to_intel_display(intel_dp);
> +	int vblank = intel_crtc_vblank_length(crtc_state);
> +	int aux_less_wake_lines;
> +	int wake_lines;
> +
> +	if (!intel_psr_needs_alpm(intel_dp, crtc_state))

Now as psr_compute_config_late is not only about vblank you can't have
this early return. You could have (or do it as you see is best):

	if (intel_psr_needs_alpm_aux_less(intel_dp, crtc_state))
		wake_lines = DISPLAY_VER(display) < 20 ?
			psr2_block_count_lines(crtc_state->alpm_state.io_wake_lines,
					       crtc_state->alpm_state.fast_wake_lines) :
			crtc_state->alpm_state.io_wake_lines;
	else if (intel_psr_needs_alpm_aux_less(intel_dp, crtc_state))
		wake_lines = crtc_state->alpm_state.aux_less_wake_lines;
	else
		wake_lines = 0;

	if (!wake_lines || !_wake_lines_fit_into_vblank(crtc_state, vblank, wake_lines)) {
		drm_dbg_kms(display->drm,
			    "Adjusting PSR/PR mode: vblank too short for wake lines = %d\n",
			    wake_lines);

		if (crtc_state->has_panel_replay) {
			crtc_state->has_panel_replay = false;
			/*
			 * #TODO : Add fall back to PSR/PSR2
			 * Since panel replay cannot be supported, we can fall back to PSR/PSR2.
			 * This will require calling compute_config for psr and psr2 with check for
			 * actual guardband instead of vblank_length.
			 */
			crtc_state->has_psr = false;
		}
		crtc_state->has_sel_update = false;
		crtc_state->enable_psr2_su_region_et = false;
		crtc_state->enable_psr2_sel_fetch = false;
	}

BR,

Jouni Högander

> +		return;
> +
> +	aux_less_wake_lines = crtc_state-
> >alpm_state.aux_less_wake_lines;
> +	wake_lines = DISPLAY_VER(display) < 20 ?
> +		     psr2_block_count_lines(crtc_state-
> >alpm_state.io_wake_lines,
> +					    crtc_state-
> >alpm_state.fast_wake_lines) :
> +		     crtc_state->alpm_state.io_wake_lines;
> +
> +	/*
> +	 * Disable the PSR features if wake lines exceed the
> available vblank.
> +	 * Though SCL is computed based on these PSR features, it is
> not reset
> +	 * even if the PSR features are disabled to avoid changing
> vblank start
> +	 * at this stage.
> +	 */
> +	if (!crtc_state->has_panel_replay && crtc_state-
> >has_sel_update &&
> +	    !_wake_lines_fit_into_vblank(crtc_state, vblank,
> wake_lines)) {
> +		drm_dbg_kms(display->drm,
> +			    "Disabling Selective Update: vblank too
> short for wake lines = %d\n",
> +			    wake_lines);
> +		crtc_state->has_sel_update = false;
> +		crtc_state->enable_psr2_su_region_et = false;
> +		crtc_state->enable_psr2_sel_fetch = false;
> +	}
> +
> +	if (intel_psr_needs_alpm_aux_less(intel_dp, crtc_state) &&
> +	    !_wake_lines_fit_into_vblank(crtc_state, vblank,
> aux_less_wake_lines)) {
> +		drm_dbg_kms(display->drm,
> +			    "Disabling Panel replay: vblank too
> short for wake lines = %d\n",
> +			    aux_less_wake_lines);
> +		crtc_state->has_panel_replay = false;
> +		/*
> +		 * #TODO : Add fall back to PSR/PSR2
> +		 * Since panel replay cannot be supported, we can
> fall back to PSR/PSR2.
> +		 * This will require calling compute_config for psr
> and psr2 with check for
> +		 * actual guardband instead of vblank_length.
> +		 */
> +		crtc_state->has_psr = false;
> +		crtc_state->has_sel_update = false;
> +		crtc_state->enable_psr2_su_region_et = false;
> +		crtc_state->enable_psr2_sel_fetch = false;
> +	}
>  
>  	/* Wa_18037818876 */
>  	if (intel_psr_needs_wa_18037818876(intel_dp, crtc_state)) {


  reply	other threads:[~2025-10-15  8:23 UTC|newest]

Thread overview: 23+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2025-10-15  7:22 [PATCH 00/10] Preparatory patches for guardband optimization Ankit Nautiyal
2025-10-15  7:22 ` [PATCH 01/10] drm/i915/vrr: Use crtc_vsync_start/end for computing vrr.vsync_start/end Ankit Nautiyal
2025-10-15  7:22 ` [PATCH 02/10] drm/i915/display: Move intel_dpll_crtc_compute_clock early Ankit Nautiyal
2025-10-15  7:22 ` [PATCH 03/10] drm/i915/vrr: s/intel_vrr_compute_config_late/intel_vrr_compute_guardband Ankit Nautiyal
2025-10-15  7:22 ` [PATCH 04/10] drm/i915/vblank: Add helper to get correct vblank length Ankit Nautiyal
2025-10-15  7:22 ` [PATCH 05/10] drm/i915/psr: Consider SCL lines when validating vblank for wake latency Ankit Nautiyal
2025-10-15  7:22 ` [PATCH 06/10] drm/i915/psr: Introduce helper intel_psr_set_non_psr_pipes() Ankit Nautiyal
2025-10-15  7:57   ` Hogander, Jouni
2025-10-15  7:22 ` [PATCH 07/10] drm/i915/display: Introduce dp/psr_compute_config_late() Ankit Nautiyal
2025-10-15  7:59   ` Hogander, Jouni
2025-10-15  7:22 ` [PATCH 08/10] drm/i915/psr: Check if final vblank is sufficient for PSR features Ankit Nautiyal
2025-10-15  8:23   ` Hogander, Jouni [this message]
2025-10-15  8:33     ` Hogander, Jouni
2025-10-15  9:14   ` Ankit Nautiyal
2025-10-15  9:32     ` Hogander, Jouni
2025-10-15  7:22 ` [PATCH 09/10] drm/i915/display: Add vblank_start adjustment logic for always-on VRR TG Ankit Nautiyal
2025-10-15 13:54   ` Ville Syrjälä
2025-10-15  7:22 ` [PATCH 10/10] drm/i915/display: Prepare for vblank_delay for LRR Ankit Nautiyal
2025-10-15 15:00   ` Ville Syrjälä
2025-10-15 11:07 ` ✓ i915.CI.BAT: success for Preparatory patches for guardband optimization (rev7) Patchwork
2025-10-15 17:12 ` ✗ i915.CI.Full: failure " Patchwork
2025-10-16  5:24 ` ✓ i915.CI.Full: success " Patchwork
  -- strict thread matches above, loose matches on Subject: below --
2025-10-16  5:54 [PATCH 00/10] Preparatory patches for guardband optimization Ankit Nautiyal
2025-10-16  5:54 ` [PATCH 08/10] drm/i915/psr: Check if final vblank is sufficient for PSR features Ankit Nautiyal

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=6aa087759e3b2467ebe47d25309afdc56a4e986b.camel@intel.com \
    --to=jouni.hogander@intel.com \
    --cc=animesh.manna@intel.com \
    --cc=ankit.k.nautiyal@intel.com \
    --cc=intel-gfx@lists.freedesktop.org \
    --cc=intel-xe@lists.freedesktop.org \
    --cc=ville.syrjala@linux.intel.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox