Re: [PATCH v5 26/28] drm/i915/dsc: Enable and disable appropriate power wells for VDSC

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On Fri, Oct 05, 2018 at 04:23:04PM -0700, Manasi Navare wrote:
> A separate power well 2 (PG2) is required for VDSC on eDP transcoder
> whereas all other transcoders use the power wells associated with the
> transcoders for VDSC.
> This patch adds a helper to obtain correct power domain depending on
> transcoder being used and enables/disables the power wells during
> VDSC enabling/disabling.
> 
> Suggested-by: Ville Syrjala <ville.syrjala@xxxxxxxxxxxxxxx>
> Cc: Ville Syrjala <ville.syrjala@xxxxxxxxxxxxxxx>
> Cc: Imre Deak <imre.deak@xxxxxxxxx>
> Cc: Rodrigo Vivi <rodrigo.vivi@xxxxxxxxx>
> Signed-off-by: Manasi Navare <manasi.d.navare@xxxxxxxxx>
> ---
>  drivers/gpu/drm/i915/intel_vdsc.c | 25 +++++++++++++++++++++++++
>  1 file changed, 25 insertions(+)
> 
> diff --git a/drivers/gpu/drm/i915/intel_vdsc.c b/drivers/gpu/drm/i915/intel_vdsc.c
> index 4963e80a87f0..d2b4601459c3 100644
> --- a/drivers/gpu/drm/i915/intel_vdsc.c
> +++ b/drivers/gpu/drm/i915/intel_vdsc.c
> @@ -581,6 +581,23 @@ int intel_dp_compute_dsc_params(struct intel_dp *intel_dp,
>  	return 0;
>  }
>  
> +static enum intel_display_power_domain
> +intel_dsc_get_power_domains(struct intel_crtc_state *crtc_state)

crtc_state can be const.

> +{
> +	enum transcoder cpu_transcoder = crtc_state->cpu_transcoder;
> +
> +	/*
> +         * On ICL+ PW2/ POWER_DOMAIN_VDSC_PIPE_A is required for

Tabs fail in this comment.

> +         * VDSC/joining for eDP transcoder.
> +         * For any other transcoder, VDSC/joining uses the power well associated
> +         * with the pipe/transcoder in use.

The one thing I'd like the comment to explain is why the power
domain is called PIPE_A_something but we pick it based on the
transcoder.

Maybe?
/*
 * Only the eDP/DSI VDSC (ICL) and the pipe A VDSC (*something*)
 * have their own power well. For the other VDSCs another
 * reference for the transcoder (ICL) or pipe (*something*) will
 * suffice.
 *
 * To avoid wasting power domain bits we reuse the PIPE_A_VDSC
 * power domain for both the eDP/DSI VDSC (ICL) and the
 * pipe A VDSC (*something*).
 */

Not sure if we can fill the *something* yet :)

Or maybe with the aliasing TRANSCODER_EDP_VDSC idea we
wouldn't really need to explain so much So maybe just
something simpler like
/*
 * Only eDP/DSI VDSC has its own power well. For the other
 * VDSCs another reference on the transcoder power domain
 * will suffice.
 */
and then we'd add the same comment with s/transcoder/pipe/ for
the next platform.

Anyways, just thinking out loud. If you want you can go with
what you have with the const/tabs fixed.
Reviewed-by: Ville Syrjälä <ville.syrjala@xxxxxxxxxxxxxxx>

> +         */
> +	if (cpu_transcoder == TRANSCODER_EDP)
> +		return POWER_DOMAIN_VDSC_PIPE_A;
> +	else
> +		return POWER_DOMAIN_TRANSCODER(cpu_transcoder);
> +}
> +
>  static void intel_configure_pps_for_dsc_encoder(struct intel_encoder *encoder,
>  						struct intel_crtc_state *crtc_state)
>  {
> @@ -1019,6 +1036,10 @@ void intel_dsc_enable(struct intel_encoder *encoder,
>  	if (!crtc_state->dsc_params.compression_enable)
>  		return;
>  
> +	/* Enable Power wells for VDSC/joining */
> +	intel_display_power_get(dev_priv,
> +				intel_dsc_get_power_domains(crtc_state));
> +
>  	intel_configure_pps_for_dsc_encoder(encoder, crtc_state);
>  
>  	intel_dp_send_dsc_pps_sdp(encoder, crtc_state);
> @@ -1073,4 +1094,8 @@ void intel_dsc_disable(struct intel_encoder *encoder,
>  				  RIGHT_BRANCH_VDSC_ENABLE);
>  	I915_WRITE(dss_ctl2_reg, dss_ctl2_val);
>  
> +	/* Disable Power wells for VDSC/joining */
> +	intel_display_power_put(dev_priv,
> +				intel_dsc_get_power_domains(old_crtc_state));
> +
>  }
> -- 
> 2.18.0

-- 
Ville Syrjälä
Intel
_______________________________________________
Intel-gfx mailing list
Intel-gfx@xxxxxxxxxxxxxxxxxxxxx
https://lists.freedesktop.org/mailman/listinfo/intel-gfx




[Index of Archives]     [AMD Graphics]     [Linux USB Devel]     [Linux Audio Users]     [Yosemite News]     [Linux Kernel]     [Linux SCSI]

  Powered by Linux