[alsa-devel] [PATCH v4 2/3] ASoC: Add multiple CPU DAI support for PCM ops
Charles Keepax
ckeepax at opensource.cirrus.com
Wed May 16 14:55:39 CEST 2018
On Tue, May 08, 2018 at 04:15:34PM +0530, Shreyas NC wrote:
> Add support in PCM operations to invoke multiple cpu dais as we do
> for multiple codec dais. Also the symmetry calculations are updated to
> reflect multiple cpu dais.
>
> Signed-off-by: Vinod Koul <vkoul at kernel.org>
> Signed-off-by: Shreyas NC <shreyas.nc at intel.com>
> ---
> sound/soc/soc-pcm.c | 487 +++++++++++++++++++++++++++++++++++-----------------
> 1 file changed, 326 insertions(+), 161 deletions(-)
Getting down to the last few comments from me :-)
> @@ -313,13 +333,18 @@ static int soc_pcm_params_symmetry(struct snd_pcm_substream *substream,
> static bool soc_pcm_has_symmetry(struct snd_pcm_substream *substream)
> {
> struct snd_soc_pcm_runtime *rtd = substream->private_data;
> - struct snd_soc_dai_driver *cpu_driver = rtd->cpu_dai->driver;
> struct snd_soc_dai_link *link = rtd->dai_link;
> - unsigned int symmetry, i;
> + unsigned int symmetry = 0, i;
No need to init this to zero you are explicitly setting it
straight after.
>
> - symmetry = cpu_driver->symmetric_rates || link->symmetric_rates ||
> - cpu_driver->symmetric_channels || link->symmetric_channels ||
> - cpu_driver->symmetric_samplebits || link->symmetric_samplebits;
> + symmetry = link->symmetric_rates || link->symmetric_channels ||
> + link->symmetric_samplebits;
> @@ -427,30 +466,55 @@ static void soc_pcm_init_runtime_hw(struct snd_pcm_substream *substream)
> rates = snd_pcm_rate_mask_intersect(codec_stream->rates, rates);
> }
>
> + for (i = 0; i < rtd->num_cpu_dai; i++) {
> + cpu_dai_drv = rtd->cpu_dais[i]->driver;
> +
> + if (substream->stream == SNDRV_PCM_STREAM_PLAYBACK)
> + cpu_stream = &cpu_dai_drv->playback;
> + else
> + cpu_stream = &cpu_dai_drv->capture;
> +
> + cpu_chan_min = max(hw->channels_min,
> + cpu_stream->channels_min);
> + cpu_chan_max = min(hw->channels_max,
> + cpu_stream->channels_max);
At the end of the loop cpu_chan_min and cpu_chan_max will only
have considered the channels_min/max from the last cpu DAI, is
that the behaviour you wanted?
> +
> + if (hw->formats)
> + hw->formats &= cpu_stream->formats;
> + else
> + hw->formats = cpu_stream->formats;
> +
> + cpu_rates = snd_pcm_rate_mask_intersect(cpu_rates,
> + cpu_stream->rates);
> +
> + cpu_rate_min = max(hw->rate_min, cpu_stream->rate_min);
> + cpu_rate_max = min_not_zero(hw->rate_max, cpu_stream->rate_max);
Same here with cpu_rate_min and cpu_rate_max all the first n-1 CPU
DAIs are ignored and only the last DAI is considered.
> @@ -1162,7 +1292,7 @@ static snd_pcm_uframes_t soc_pcm_pointer(struct snd_pcm_substream *substream)
> struct snd_soc_pcm_runtime *rtd = substream->private_data;
> struct snd_soc_component *component;
> struct snd_soc_rtdcom_list *rtdcom;
> - struct snd_soc_dai *cpu_dai = rtd->cpu_dai;
> + struct snd_soc_dai *cpu_dai;
> struct snd_soc_dai *codec_dai;
> struct snd_pcm_runtime *runtime = substream->runtime;
> snd_pcm_uframes_t offset = 0;
> @@ -1182,8 +1312,12 @@ static snd_pcm_uframes_t soc_pcm_pointer(struct snd_pcm_substream *substream)
> break;
> }
>
> - if (cpu_dai->driver->ops->delay)
> - delay += cpu_dai->driver->ops->delay(substream, cpu_dai);
> + for (i = 0; i < rtd->num_cpu_dai; i++) {
> + cpu_dai = rtd->cpu_dais[i];
> + if (cpu_dai->driver->ops->delay)
> + delay += cpu_dai->driver->ops->delay(substream,
> + cpu_dai);
> + }
Should we be adding these delays? Wouldn't it be better to use
the max CPU DAI delay as we do for the CODECs, or is there a
reason these are additive?
>
> for (i = 0; i < rtd->num_codecs; i++) {
> codec_dai = rtd->codec_dais[i];
> @@ -1306,12 +1440,16 @@ static struct snd_soc_pcm_runtime *dpcm_get_be(struct snd_soc_card *card,
> if (!be->dai_link->no_pcm)
> continue;
>
> - dev_dbg(card->dev, "ASoC: try BE : %s\n",
> - be->cpu_dai->playback_widget ?
> - be->cpu_dai->playback_widget->name : "(not set)");
> + for (i = 0; i < be->num_cpu_dai; i++) {
> + struct snd_soc_dai *cpu_dai = be->cpu_dais[i];
> +
> + dev_dbg(card->dev, "ASoC: try BE : %s\n",
> + cpu_dai->playback_widget ?
> + cpu_dai->playback_widget->name : "(not set)");
>
> - if (be->cpu_dai->playback_widget == widget)
> - return be;
> + if (cpu_dai->playback_widget == widget)
> + return be;
> + }
>
> for (i = 0; i < be->num_codecs; i++) {
> struct snd_soc_dai *dai = be->codec_dais[i];
> @@ -1326,12 +1464,16 @@ static struct snd_soc_pcm_runtime *dpcm_get_be(struct snd_soc_card *card,
> if (!be->dai_link->no_pcm)
> continue;
>
> - dev_dbg(card->dev, "ASoC: try BE %s\n",
> - be->cpu_dai->capture_widget ?
> - be->cpu_dai->capture_widget->name : "(not set)");
> + for (i = 0; i < be->num_cpu_dai; i++) {
> + struct snd_soc_dai *cpu_dai = be->cpu_dais[i];
>
> - if (be->cpu_dai->capture_widget == widget)
> - return be;
> + dev_dbg(card->dev, "ASoC: try BE %s\n",
> + cpu_dai->capture_widget ?
> + cpu_dai->capture_widget->name : "(not set)");
> +
> + if (cpu_dai->capture_widget == widget)
> + return be;
> + }
>
> for (i = 0; i < be->num_codecs; i++) {
> struct snd_soc_dai *dai = be->codec_dais[i];
I still think you need to get Liam or someone to review this DPCM
stuff. Multi-CPU DAIs feels like it could be a tricky thing to
merge with DPCM and I am not sure I am confident enough in me
reviewing it right.
Thanks,
Charles
More information about the Alsa-devel
mailing list