[pulseaudio-discuss] [PATCH 2/2] loopback: Track and use average adjust time

[Georg Chini]

On 04.08.2017 15:37, Tanu Kaskinen wrote:
> On Mon, 2017-04-17 at 21:37 +0200, Georg Chini wrote:
>> @@ -327,11 +335,20 @@ static void adjust_rates(struct userdata *u) {
>>       }
>>   
>>       /* Allow one underrun per hour */
>> -    if (u->iteration_counter * u->adjust_time / PA_USEC_PER_SEC / 3600 > run_hours) {
>> +    if (u->iteration_counter * u->real_adjust_time / PA_USEC_PER_SEC / 3600 > run_hours) {
>>           u->underrun_counter = PA_CLIP_SUB(u->underrun_counter, 1u);
>>           pa_log_info("Underrun counter: %u", u->underrun_counter);
>>       }
>>   
>> +    /* Calculate real adjust time */
>> +    now = pa_rtclock_now();
>> +    if (!u->source_sink_changed) {
>> +        u->adjust_counter++;
>> +        u->real_adjust_time_sum += now - u->adjust_time_stamp;
>> +        u->real_adjust_time = u->real_adjust_time_sum / u->adjust_counter;
>> +    }
>> +    u->adjust_time_stamp = now;
>> +
>>       /* Rates and latencies*/
>>       old_rate = u->sink_input->sample_spec.rate;
>>       base_rate = u->source_output->sample_spec.rate;
>> @@ -364,7 +381,9 @@ static void adjust_rates(struct userdata *u) {
>>       pa_log_debug("Loopback latency at base rate is %0.2f ms", (double)latency_at_optimum_rate / PA_USEC_PER_MSEC);
>>   
>>       /* Calculate new rate */
>> -    new_rate = rate_controller(base_rate, u->adjust_time, latency_difference);
>> +    new_rate = rate_controller(base_rate, u->real_adjust_time, latency_difference);
>> +
>> +    u->source_sink_changed = false;
> This "source_sink_changed = false" assignment would be better right
> after the code that checks its value, at least based on this patch
> alone. But maybe later patches use this variable too in a way that
> makes this suggestion less good?
>
> Otherwise looks fine.
>
Yes, the variable is later user within rate_controller(). Also there will be
other variables that indicate state changes between two calls of
adjust_rate() (for example if the latency offset changed). I'll push the 
patch
to next as is if you have no objection.