Re: [PATCH 1/2] ALSA: core: add hooks for audio timestamps

[Clemens Ladisch <clemens@ladisch.de>]

Pierre-Louis Bossart wrote:
>>> +                audio_frames = runtime->status->hw_ptr
>>> +                    + runtime->delay;
>>> +            audio_nsecs = audio_frames * 1000000000LL /
>>> +                runtime->rate;
>>
>> This looks like a 64-bit division.

... which needs to be handled with a function from <linux/math64.h>.

>> And what happens if audio_frames becomes negative?
>
> It's my understanding that hw_ptr represents the cumulative frames
> played since the beginning, not sure why it'd become negative, ever.

2^32 / 192 kHz = 6.2 h

> I don't understand the notion of 'boundary'

This is where hw_ptr wraps around.  It's a multiple of the buffer size
to make some calculations easier.

> and the use of the hw_ptr_base

It's the hw_ptr corresponding to the start of the buffer.

I guess audio_tstamp isn't supposed to wrap around?


Regards,
Clemens