On July 26, 2021 8:18:19 PM PDT, Chengyi Zhao <zhaochengyi@xxxxxxxxxxxxx> wrote:
>Hi guys,
>
>I found the following logic when I change the volume while playing
>audio.I guess there is some audio expertise here, would you please tell
>
>me the principle, thanks a lot!
>
>In the file sink-input.c(./src/pulsecore/sink-input.c), why does pa
>need
>to invoke the function pa_sink_input_request_rewind to rewrite and
>rewind the data?
>
>void pa_sink_input_set_volume(...) {
> ...
>
> pa_assert_se(pa_asyncmsgq_send(i->sink->asyncmsgq, PA_MSGOBJECT(i),
>PA_SINK_INPUT_MESSAGE_SET_SOFT_VOLUME, NULL, 0, NULL) == 0);
> }
>
> ...
>}
>
>int pa_sink_input_process_msg(...) {
> ...
>
> switch (code) {
>
> case PA_SINK_INPUT_MESSAGE_SET_SOFT_VOLUME:
> if (!pa_cvolume_equal(&i->thread_info.soft_volume,
>&i->soft_volume)) {
> i->thread_info.soft_volume = i->soft_volume;
> pa_sink_input_request_rewind(i, 0, true, false, false);
> }
>
> return 0;
>
> ...
>}
There's some helpful docs about what rewinding is in the context of pulseaudio here:
https://www.freedesktop.org/wiki/Software/PulseAudio/Documentation/Developer/Rewinding/
But in summary, pulse will try to get ahead of things by buffering up processed audio data. When the requested processing changes, for example a new volume level is requested, rewinding clears out the (now out of date) buffered audio samples and rewrites them with the new processing.
--Sean
