* [PATCH] ASoC: Intel: Atom: read timestamp moved to period_elapsed
@ 2019-07-09 4:01 Alex Levin
2019-07-09 11:45 ` Andy Shevchenko
0 siblings, 1 reply; 4+ messages in thread
From: Alex Levin @ 2019-07-09 4:01 UTC (permalink / raw)
To: alsa-devel
Cc: Pierre-Louis Bossart, Liam Girdwood, Jie Yang, Mark Brown,
Jaroslav Kysela, Takashi Iwai, Andy Shevchenko, Alex Levin,
linux-kernel, benzh, cujomalainey
sst_platform_pcm_pointer is called from both snd_pcm_period_elapsed and
from snd_pcm_ioctl. Calling read timestamp results in recalculating
pcm_delay and buffer_ptr (sst_calc_tstamp) which consumes buffers in a
faster rate than intended.
In a tested BSW system with chtrt5650, for a rate of 48000, the
measured rate was sometimes 10 times more than that.
After moving the timestamp read to period elapsed, buffer consumption is
as expected.
Signed-off-by: Alex Levin <levinale@chromium.org>
---
sound/soc/intel/atom/sst-mfld-platform-pcm.c | 23 +++++++++++++-------
1 file changed, 15 insertions(+), 8 deletions(-)
diff --git a/sound/soc/intel/atom/sst-mfld-platform-pcm.c b/sound/soc/intel/atom/sst-mfld-platform-pcm.c
index 0e8b1c5eec88..196af0b30b41 100644
--- a/sound/soc/intel/atom/sst-mfld-platform-pcm.c
+++ b/sound/soc/intel/atom/sst-mfld-platform-pcm.c
@@ -265,16 +265,28 @@ static void sst_period_elapsed(void *arg)
{
struct snd_pcm_substream *substream = arg;
struct sst_runtime_stream *stream;
- int status;
+ struct snd_soc_pcm_runtime *rtd;
+ int status, ret_val;
if (!substream || !substream->runtime)
return;
stream = substream->runtime->private_data;
if (!stream)
return;
+
+ rtd = substream->private_data;
+ if (!rtd)
+ return;
+
status = sst_get_stream_status(stream);
if (status != SST_PLATFORM_RUNNING)
return;
+
+ ret_val = stream->ops->stream_read_tstamp(sst->dev, &stream->stream_info);
+ if (ret_val) {
+ dev_err(rtd->dev, "stream_read_tstamp err code = %d\n", ret_val);
+ return;
+ }
snd_pcm_period_elapsed(substream);
}
@@ -658,20 +670,15 @@ static snd_pcm_uframes_t sst_platform_pcm_pointer
(struct snd_pcm_substream *substream)
{
struct sst_runtime_stream *stream;
- int ret_val, status;
+ int status;
struct pcm_stream_info *str_info;
- struct snd_soc_pcm_runtime *rtd = substream->private_data;
stream = substream->runtime->private_data;
status = sst_get_stream_status(stream);
if (status == SST_PLATFORM_INIT)
return 0;
+
str_info = &stream->stream_info;
- ret_val = stream->ops->stream_read_tstamp(sst->dev, str_info);
- if (ret_val) {
- dev_err(rtd->dev, "sst: error code = %d\n", ret_val);
- return ret_val;
- }
substream->runtime->delay = str_info->pcm_delay;
return str_info->buffer_ptr;
}
--
2.22.0.410.gd8fdbe21b5-goog
^ permalink raw reply related [flat|nested] 4+ messages in thread
* Re: [PATCH] ASoC: Intel: Atom: read timestamp moved to period_elapsed
2019-07-09 4:01 [PATCH] ASoC: Intel: Atom: read timestamp moved to period_elapsed Alex Levin
@ 2019-07-09 11:45 ` Andy Shevchenko
2019-07-10 23:15 ` Curtis Malainey
0 siblings, 1 reply; 4+ messages in thread
From: Andy Shevchenko @ 2019-07-09 11:45 UTC (permalink / raw)
To: Alex Levin
Cc: alsa-devel, Pierre-Louis Bossart, Liam Girdwood, Jie Yang,
Mark Brown, Jaroslav Kysela, Takashi Iwai, linux-kernel, benzh,
cujomalainey
On Mon, Jul 08, 2019 at 09:01:47PM -0700, Alex Levin wrote:
> sst_platform_pcm_pointer is called from both snd_pcm_period_elapsed and
> from snd_pcm_ioctl. Calling read timestamp results in recalculating
> pcm_delay and buffer_ptr (sst_calc_tstamp) which consumes buffers in a
> faster rate than intended.
> In a tested BSW system with chtrt5650, for a rate of 48000, the
> measured rate was sometimes 10 times more than that.
> After moving the timestamp read to period elapsed, buffer consumption is
> as expected.
From code prospective it looks good. You may take mine
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Though I'm not an expert in the area, Pierre and / or Liam should give
their blessing.
>
> Signed-off-by: Alex Levin <levinale@chromium.org>
> ---
> sound/soc/intel/atom/sst-mfld-platform-pcm.c | 23 +++++++++++++-------
> 1 file changed, 15 insertions(+), 8 deletions(-)
>
> diff --git a/sound/soc/intel/atom/sst-mfld-platform-pcm.c b/sound/soc/intel/atom/sst-mfld-platform-pcm.c
> index 0e8b1c5eec88..196af0b30b41 100644
> --- a/sound/soc/intel/atom/sst-mfld-platform-pcm.c
> +++ b/sound/soc/intel/atom/sst-mfld-platform-pcm.c
> @@ -265,16 +265,28 @@ static void sst_period_elapsed(void *arg)
> {
> struct snd_pcm_substream *substream = arg;
> struct sst_runtime_stream *stream;
> - int status;
> + struct snd_soc_pcm_runtime *rtd;
> + int status, ret_val;
>
> if (!substream || !substream->runtime)
> return;
> stream = substream->runtime->private_data;
> if (!stream)
> return;
> +
> + rtd = substream->private_data;
> + if (!rtd)
> + return;
> +
> status = sst_get_stream_status(stream);
> if (status != SST_PLATFORM_RUNNING)
> return;
> +
> + ret_val = stream->ops->stream_read_tstamp(sst->dev, &stream->stream_info);
> + if (ret_val) {
> + dev_err(rtd->dev, "stream_read_tstamp err code = %d\n", ret_val);
> + return;
> + }
> snd_pcm_period_elapsed(substream);
> }
>
> @@ -658,20 +670,15 @@ static snd_pcm_uframes_t sst_platform_pcm_pointer
> (struct snd_pcm_substream *substream)
> {
> struct sst_runtime_stream *stream;
> - int ret_val, status;
> + int status;
> struct pcm_stream_info *str_info;
> - struct snd_soc_pcm_runtime *rtd = substream->private_data;
>
> stream = substream->runtime->private_data;
> status = sst_get_stream_status(stream);
> if (status == SST_PLATFORM_INIT)
> return 0;
> +
> str_info = &stream->stream_info;
> - ret_val = stream->ops->stream_read_tstamp(sst->dev, str_info);
> - if (ret_val) {
> - dev_err(rtd->dev, "sst: error code = %d\n", ret_val);
> - return ret_val;
> - }
> substream->runtime->delay = str_info->pcm_delay;
> return str_info->buffer_ptr;
> }
> --
> 2.22.0.410.gd8fdbe21b5-goog
>
--
With Best Regards,
Andy Shevchenko
^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: [PATCH] ASoC: Intel: Atom: read timestamp moved to period_elapsed
2019-07-09 11:45 ` Andy Shevchenko
@ 2019-07-10 23:15 ` Curtis Malainey
2019-07-17 15:30 ` [alsa-devel] " Pierre-Louis Bossart
0 siblings, 1 reply; 4+ messages in thread
From: Curtis Malainey @ 2019-07-10 23:15 UTC (permalink / raw)
To: Andy Shevchenko
Cc: Alex Levin, ALSA development, Pierre-Louis Bossart,
Liam Girdwood, Jie Yang, Mark Brown, Jaroslav Kysela,
Takashi Iwai, Linux Kernel Mailing List, Ben Zhang,
Curtis Malainey
On Tue, Jul 9, 2019 at 4:45 AM Andy Shevchenko
<andriy.shevchenko@linux.intel.com> wrote:
>
> On Mon, Jul 08, 2019 at 09:01:47PM -0700, Alex Levin wrote:
> > sst_platform_pcm_pointer is called from both snd_pcm_period_elapsed and
> > from snd_pcm_ioctl. Calling read timestamp results in recalculating
> > pcm_delay and buffer_ptr (sst_calc_tstamp) which consumes buffers in a
> > faster rate than intended.
> > In a tested BSW system with chtrt5650, for a rate of 48000, the
> > measured rate was sometimes 10 times more than that.
> > After moving the timestamp read to period elapsed, buffer consumption is
> > as expected.
>
> From code prospective it looks good. You may take mine
> Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
>
> Though I'm not an expert in the area, Pierre and / or Liam should give
> their blessing.
>
Agreed, Liam or Pierre should also give their ok since this is one of
the closed source firmware drivers.
Reviewed-by: Curtis Malainey <cujomalainey@chromium.org>
> >
> > Signed-off-by: Alex Levin <levinale@chromium.org>
> > ---
> > sound/soc/intel/atom/sst-mfld-platform-pcm.c | 23 +++++++++++++-------
> > 1 file changed, 15 insertions(+), 8 deletions(-)
> >
> > diff --git a/sound/soc/intel/atom/sst-mfld-platform-pcm.c b/sound/soc/intel/atom/sst-mfld-platform-pcm.c
> > index 0e8b1c5eec88..196af0b30b41 100644
> > --- a/sound/soc/intel/atom/sst-mfld-platform-pcm.c
> > +++ b/sound/soc/intel/atom/sst-mfld-platform-pcm.c
> > @@ -265,16 +265,28 @@ static void sst_period_elapsed(void *arg)
> > {
> > struct snd_pcm_substream *substream = arg;
> > struct sst_runtime_stream *stream;
> > - int status;
> > + struct snd_soc_pcm_runtime *rtd;
> > + int status, ret_val;
> >
> > if (!substream || !substream->runtime)
> > return;
> > stream = substream->runtime->private_data;
> > if (!stream)
> > return;
> > +
> > + rtd = substream->private_data;
> > + if (!rtd)
> > + return;
> > +
> > status = sst_get_stream_status(stream);
> > if (status != SST_PLATFORM_RUNNING)
> > return;
> > +
> > + ret_val = stream->ops->stream_read_tstamp(sst->dev, &stream->stream_info);
> > + if (ret_val) {
> > + dev_err(rtd->dev, "stream_read_tstamp err code = %d\n", ret_val);
> > + return;
> > + }
> > snd_pcm_period_elapsed(substream);
> > }
> >
> > @@ -658,20 +670,15 @@ static snd_pcm_uframes_t sst_platform_pcm_pointer
> > (struct snd_pcm_substream *substream)
> > {
> > struct sst_runtime_stream *stream;
> > - int ret_val, status;
> > + int status;
> > struct pcm_stream_info *str_info;
> > - struct snd_soc_pcm_runtime *rtd = substream->private_data;
> >
> > stream = substream->runtime->private_data;
> > status = sst_get_stream_status(stream);
> > if (status == SST_PLATFORM_INIT)
> > return 0;
> > +
> > str_info = &stream->stream_info;
> > - ret_val = stream->ops->stream_read_tstamp(sst->dev, str_info);
> > - if (ret_val) {
> > - dev_err(rtd->dev, "sst: error code = %d\n", ret_val);
> > - return ret_val;
> > - }
> > substream->runtime->delay = str_info->pcm_delay;
> > return str_info->buffer_ptr;
> > }
> > --
> > 2.22.0.410.gd8fdbe21b5-goog
> >
>
> --
> With Best Regards,
> Andy Shevchenko
>
>
^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: [alsa-devel] [PATCH] ASoC: Intel: Atom: read timestamp moved to period_elapsed
2019-07-10 23:15 ` Curtis Malainey
@ 2019-07-17 15:30 ` Pierre-Louis Bossart
0 siblings, 0 replies; 4+ messages in thread
From: Pierre-Louis Bossart @ 2019-07-17 15:30 UTC (permalink / raw)
To: Curtis Malainey, Andy Shevchenko
Cc: ALSA development, Linux Kernel Mailing List, Takashi Iwai,
Jie Yang, Liam Girdwood, Ben Zhang, Mark Brown, Curtis Malainey,
Alex Levin
On 7/10/19 6:15 PM, Curtis Malainey wrote:
> On Tue, Jul 9, 2019 at 4:45 AM Andy Shevchenko
> <andriy.shevchenko@linux.intel.com> wrote:
>>
>> On Mon, Jul 08, 2019 at 09:01:47PM -0700, Alex Levin wrote:
>>> sst_platform_pcm_pointer is called from both snd_pcm_period_elapsed and
>>> from snd_pcm_ioctl. Calling read timestamp results in recalculating
>>> pcm_delay and buffer_ptr (sst_calc_tstamp) which consumes buffers in a
>>> faster rate than intended.
>>> In a tested BSW system with chtrt5650, for a rate of 48000, the
>>> measured rate was sometimes 10 times more than that.
>>> After moving the timestamp read to period elapsed, buffer consumption is
>>> as expected.
>>
>> From code prospective it looks good. You may take mine
>> Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
>>
>> Though I'm not an expert in the area, Pierre and / or Liam should give
>> their blessing.
>>
> Agreed, Liam or Pierre should also give their ok since this is one of
> the closed source firmware drivers.
>
> Reviewed-by: Curtis Malainey <cujomalainey@chromium.org>
Humm, this first review after my Summer break isn't straightforward.
By moving the timestamp update to the period elapsed event, don't you
prevent the use of this driver in no-interrupt mode - which I understood
as the baseline for Chrome?
And I also don't get how this timestamp might lead to 10x speed issues,
this driver has been around for a number of years and that specific
error was never seen. What is different in this platform and can this be
seen e.g. on a Cyan device?
>>>
>>> Signed-off-by: Alex Levin <levinale@chromium.org>
>>> ---
>>> sound/soc/intel/atom/sst-mfld-platform-pcm.c | 23 +++++++++++++-------
>>> 1 file changed, 15 insertions(+), 8 deletions(-)
>>>
>>> diff --git a/sound/soc/intel/atom/sst-mfld-platform-pcm.c b/sound/soc/intel/atom/sst-mfld-platform-pcm.c
>>> index 0e8b1c5eec88..196af0b30b41 100644
>>> --- a/sound/soc/intel/atom/sst-mfld-platform-pcm.c
>>> +++ b/sound/soc/intel/atom/sst-mfld-platform-pcm.c
>>> @@ -265,16 +265,28 @@ static void sst_period_elapsed(void *arg)
>>> {
>>> struct snd_pcm_substream *substream = arg;
>>> struct sst_runtime_stream *stream;
>>> - int status;
>>> + struct snd_soc_pcm_runtime *rtd;
>>> + int status, ret_val;
>>>
>>> if (!substream || !substream->runtime)
>>> return;
>>> stream = substream->runtime->private_data;
>>> if (!stream)
>>> return;
>>> +
>>> + rtd = substream->private_data;
>>> + if (!rtd)
>>> + return;
>>> +
>>> status = sst_get_stream_status(stream);
>>> if (status != SST_PLATFORM_RUNNING)
>>> return;
>>> +
>>> + ret_val = stream->ops->stream_read_tstamp(sst->dev, &stream->stream_info);
>>> + if (ret_val) {
>>> + dev_err(rtd->dev, "stream_read_tstamp err code = %d\n", ret_val);
>>> + return;
>>> + }
>>> snd_pcm_period_elapsed(substream);
>>> }
>>>
>>> @@ -658,20 +670,15 @@ static snd_pcm_uframes_t sst_platform_pcm_pointer
>>> (struct snd_pcm_substream *substream)
>>> {
>>> struct sst_runtime_stream *stream;
>>> - int ret_val, status;
>>> + int status;
>>> struct pcm_stream_info *str_info;
>>> - struct snd_soc_pcm_runtime *rtd = substream->private_data;
>>>
>>> stream = substream->runtime->private_data;
>>> status = sst_get_stream_status(stream);
>>> if (status == SST_PLATFORM_INIT)
>>> return 0;
>>> +
>>> str_info = &stream->stream_info;
>>> - ret_val = stream->ops->stream_read_tstamp(sst->dev, str_info);
>>> - if (ret_val) {
>>> - dev_err(rtd->dev, "sst: error code = %d\n", ret_val);
>>> - return ret_val;
>>> - }
>>> substream->runtime->delay = str_info->pcm_delay;
>>> return str_info->buffer_ptr;
>>> }
>>> --
>>> 2.22.0.410.gd8fdbe21b5-goog
>>>
>>
>> --
>> With Best Regards,
>> Andy Shevchenko
>>
>>
> _______________________________________________
> Alsa-devel mailing list
> Alsa-devel@alsa-project.org
> https://mailman.alsa-project.org/mailman/listinfo/alsa-devel
>
^ permalink raw reply [flat|nested] 4+ messages in thread
end of thread, other threads:[~2019-07-17 15:30 UTC | newest]
Thread overview: 4+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2019-07-09 4:01 [PATCH] ASoC: Intel: Atom: read timestamp moved to period_elapsed Alex Levin
2019-07-09 11:45 ` Andy Shevchenko
2019-07-10 23:15 ` Curtis Malainey
2019-07-17 15:30 ` [alsa-devel] " Pierre-Louis Bossart
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).