[Intel-gfx] [PATCH 3/3] drm/i915/display: Add wrapper to Compute SAD
Golani, Mitulkumar Ajitkumar
mitulkumar.ajitkumar.golani at intel.com
Mon Aug 21 16:10:16 UTC 2023
Hi @Jani Nikula
Thanks for reviewing patch series.
> -----Original Message-----
> From: Jani Nikula <jani.nikula at linux.intel.com>
> Sent: 18 August 2023 14:08
> To: Golani, Mitulkumar Ajitkumar
> <mitulkumar.ajitkumar.golani at intel.com>; intel-gfx at lists.freedesktop.org
> Cc: jyri.sarha at linux.intel.com
> Subject: Re: [Intel-gfx] [PATCH 3/3] drm/i915/display: Add wrapper to
> Compute SAD
>
> On Thu, 17 Aug 2023, Mitul Golani
> <mitulkumar.ajitkumar.golani at intel.com> wrote:
> > Compute SADs that takes into account the supported rate and channel
> > based on the capabilities of the audio source.
> > This wrapper function should encapsulate the logic for determining the
> > supported rate and channel and should return a set of SADs that are
> > compatible with the source.
> >
> > --v1:
> > - call intel_audio_compute_eld in this commit as it is defined here
> >
> > --v2:
> > - Handle case when max frequency is less than 32k.
> > - remove drm prefix.
> > - name change for parse_sad to eld_to_sad.
> >
> > --v3:
> > - Use signed int wherever required.
> > - add debug trace when channel is limited.
> >
> > --v4:
> > - remove inline from eld_to_sad.
> > - declare index outside of for loop with int type.
> > - Correct mask value calculation.
> > - remove drm_err, instead just return if eld parsing failed.
> > - remove unncessary typecast
> > - reduce indentation while parsing sad
> > - use intel_audio_compute_eld as static and call bandwidth calculation
> > just before that.
> >
> > --v9:
> > - Handling the case when, sink supported channel is less than max
> > supported. In that case, rate needs to be calibrate in accordance with
> > available bandwidth.
> >
> > Signed-off-by: Mitul Golani <mitulkumar.ajitkumar.golani at intel.com>
> > ---
> > drivers/gpu/drm/i915/display/intel_audio.c | 107
> > ++++++++++++++++++++-
> > 1 file changed, 106 insertions(+), 1 deletion(-)
> >
> > diff --git a/drivers/gpu/drm/i915/display/intel_audio.c
> > b/drivers/gpu/drm/i915/display/intel_audio.c
> > index 79377e33a59b..c90ac2608eef 100644
> > --- a/drivers/gpu/drm/i915/display/intel_audio.c
> > +++ b/drivers/gpu/drm/i915/display/intel_audio.c
> > @@ -806,6 +806,111 @@ static void calc_audio_config_params(struct
> intel_crtc_state *pipe_config)
> > pipe_config->audio.max_channel_count = 0; }
> >
>
> This was added in previous patch:
>
> +static void calc_audio_config_params(struct intel_crtc_state *pipe_config)
> +{
> + struct drm_display_mode *adjusted_mode = &pipe_config-
> >hw.adjusted_mode;
> + int channel_count;
> + int index, rate[] = { 192000, 176000, 96000, 88000, 48000, 44100,
> 32000 };
> + int audio_req_bandwidth, available_blank_bandwidth, vblank,
> hblank;
> +
> + hblank = adjusted_mode->htotal - adjusted_mode->hdisplay;
> + vblank = adjusted_mode->vtotal - adjusted_mode->vdisplay;
> + available_blank_bandwidth = hblank * vblank *
> + drm_mode_vrefresh(adjusted_mode) *
> pipe_config->pipe_bpp;
> + for (channel_count = MAX_CHANNEL_COUNT; channel_count > 0;
> channel_count--) {
> + for (index = 0; index < ARRAY_SIZE(rate); index++) {
> + audio_req_bandwidth =
> calc_audio_bw(channel_count,
> + rate[index]);
> + if (audio_req_bandwidth <
> available_blank_bandwidth) {
> + pipe_config->audio.max_rate = rate[index];
> + pipe_config->audio.max_channel_count =
> channel_count;
> + return;
> + }
> + }
> + }
> +
> + pipe_config->audio.max_rate = 0;
> + pipe_config->audio.max_channel_count = 0;
> +}
>
> And this here:
Thanks I have removed duplicated code for calibration and calculation,
and merged them to one to avoid redundancy.
>
> > +static void calibrate_audio_config_params(struct intel_crtc_state
> > +*pipe_config, int channel) {
> > + struct drm_display_mode *adjusted_mode = &pipe_config-
> >hw.adjusted_mode;
> > + int channel_count;
> > + int index, rate[] = { 192000, 176000, 96000, 88000, 48000, 44100,
> 32000 };
> > + int audio_req_bandwidth, available_blank_bandwidth, vblank,
> hblank;
> > +
> > + hblank = adjusted_mode->htotal - adjusted_mode->hdisplay;
> > + vblank = adjusted_mode->vtotal - adjusted_mode->vdisplay;
> > + available_blank_bandwidth = hblank * vblank *
> > + drm_mode_vrefresh(adjusted_mode) *
> pipe_config->pipe_bpp;
> > +
> > + for (index = 0; index < ARRAY_SIZE(rate); index++) {
> > + audio_req_bandwidth = calc_audio_bw(channel_count,
> > + rate[index]);
> > + if (audio_req_bandwidth < available_blank_bandwidth) {
> > + pipe_config->audio.max_rate = rate[index];
> > + pipe_config->audio.max_channel_count =
> channel_count;
> > + return;
> > + }
> > + }
> > +
> > + pipe_config->audio.max_rate = 0;
> > + pipe_config->audio.max_channel_count = 0; }
>
> This kind of stuff needs to be deduplicated.
>
> > +
> > +static int sad_to_channels(const u8 *sad) {
> > + return 1 + (sad[0] & 0x7);
> > +}
> > +
> > +static u8 *eld_to_sad(u8 *eld)
> > +{
> > + int ver, mnl;
> > +
> > + ver = (eld[DRM_ELD_VER] & DRM_ELD_VER_MASK) >>
> DRM_ELD_VER_SHIFT;
> > + if (ver != 2 && ver != 31)
> > + return NULL;
> > +
> > + mnl = drm_eld_mnl(eld);
> > + if (mnl > 16)
> > + return NULL;
> > +
> > + return eld + DRM_ELD_CEA_SAD(mnl, 0); }
>
> I'm still not happy with the copy-paste here. The parsing details should be
> hidden in drm_edid.c and that should be the Single Point of Truth how to
> deal with this.
>
Thanks, I have added wrapper function adhering to
const correctness while accessing SAD information. Please suggest
if any other approach we can try.
> > +
> > +static int get_supported_freq_mask(struct intel_crtc_state
> > +*crtc_state) {
> > + int rate[] = { 32000, 44100, 48000, 88000, 96000, 176000, 192000 };
>
> This is like the third copy of the same array already... and the values are
> also present in the existing dp/hdmi aud arrays in intel_audio.c.
>
> And those have 88200 instead of 88000, so which one is correct?
This was mistake while validating previous patches. I have corrected
the same with reference to HDMI spec.
>
> > + int mask = 0, index;
> > +
> > + for (index = 0; index < ARRAY_SIZE(rate); index++) {
> > + if (rate[index] > crtc_state->audio.max_rate)
> > + break;
> > +
> > + mask |= 1 << index;
> > +
> > + if (crtc_state->audio.max_rate != rate[index])
> > + continue;
> > + }
> > +
> > + return mask;
> > +}
> > +
> > +static void intel_audio_compute_eld(struct intel_crtc_state
> > +*crtc_state) {
> > + struct drm_i915_private *i915 = to_i915(crtc_state->uapi.crtc->dev);
> > + u8 *eld, *sad;
> > + int index, mask = 0;
> > +
> > + eld = crtc_state->eld;
> > + if (!eld)
> > + return;
> > +
> > + sad = eld_to_sad(eld);
> > + if (!sad)
> > + return;
> > +
> > + calc_audio_config_params(crtc_state);
> > +
> > + mask = get_supported_freq_mask(crtc_state);
> > + for (index = 0; index < drm_eld_sad_count(eld); index++, sad += 3) {
> > + /*
> > + * Respect source restricitions. Limit capabilities to a subset
> that is
> > + * supported both by the source and the sink.
> > + */
> > + if (sad_to_channels(sad) >= crtc_state-
> >audio.max_channel_count) {
> > + sad[0] &= ~0x7;
>
> Also not happy with magic values here.
Thanks, defined the magic value.
Regards,
Mitul
>
> > + sad[0] |= crtc_state->audio.max_channel_count - 1;
> > + drm_dbg_kms(&i915->drm, "Channel count is
> limited to %d\n",
> > + crtc_state->audio.max_channel_count - 1);
> > + } else {
> > + /*
> > + * calibrate rate when, sink supported channel
> > + * count is slight less than max supported
> > + * channel count.
> > + */
> > + calibrate_audio_config_params(crtc_state,
> sad_to_channels(sad));
> > + mask = get_supported_freq_mask(crtc_state);
> > + }
> > +
> > + sad[1] &= mask;
> > + }
> > +}
> > +
> > bool intel_audio_compute_config(struct intel_encoder *encoder,
> > struct intel_crtc_state *crtc_state,
> > struct drm_connector_state *conn_state)
> @@ -827,7 +932,7 @@ bool
> > intel_audio_compute_config(struct intel_encoder *encoder,
> >
> > crtc_state->eld[6] = drm_av_sync_delay(connector, adjusted_mode)
> /
> > 2;
> >
> > - calc_audio_config_params(crtc_state);
> > + intel_audio_compute_eld(crtc_state);
> >
> > return true;
> > }
>
> --
> Jani Nikula, Intel Open Source Graphics Center
More information about the Intel-gfx
mailing list