nils/shairport-sync
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Make this threadsafe and modify the clock drift diagnostics

Browse Source
This commit is contained in:
Mike Brady
2017-04-07 18:37:44 +01:00
parent 8ab4f3050e
commit 9a8a0f0764
1 changed files with 25 additions and 24 deletions
Download Patch File Download Diff File Expand all files Collapse all files
rtp.c
+25 -24
View File
@@ -391,10 +391,10 @@ void *rtp_timing_receiver(void *arg) {
uint64_t local_time_by_remote_clock = distant_transmit_time + return_time / 2;
unsigned int cc;
unsigned int cc,chosen;
for (cc = time_ping_history - 1; cc > 0; cc--) {
conn->time_pings[cc] = conn->time_pings[cc - 1];
conn->time_pings[cc].dispersion = (conn->time_pings[cc].dispersion * 133) /
conn->time_pings[cc].dispersion = (conn->time_pings[cc].dispersion * 110) /
100; // make the dispersions 'age' by this rational factor
}
// these are for diagnostics only -- not used
@@ -412,9 +412,11 @@ void *rtp_timing_receiver(void *arg) {
// now pick the timestamp with the lowest dispersion
uint64_t l2rtd = conn->time_pings[0].local_to_remote_difference;
uint64_t tld = conn->time_pings[0].dispersion;
chosen = 0;
for (cc = 1; cc < conn->time_ping_count; cc++)
if (conn->time_pings[cc].dispersion < tld) {
l2rtd = conn->time_pings[cc].local_to_remote_difference;
chosen = cc;
tld = conn->time_pings[cc].dispersion;
local_time_chosen = conn->time_pings[cc].local_time;
remote_time_chosen = conn->time_pings[cc].remote_time;
@@ -444,6 +446,7 @@ void *rtp_timing_receiver(void *arg) {
}
int64_t clock_drift, clock_drift_in_usec;
double clock_drift_ppm = 0.0;
if (first_local_time == 0) {
first_local_time = local_time_chosen;
first_remote_time = remote_time_chosen;
@@ -456,28 +459,30 @@ void *rtp_timing_receiver(void *arg) {
clock_drift = remote_time_change - local_time_change;
else
clock_drift = -(local_time_change - remote_time_change);
}
if (clock_drift >= 0)
clock_drift_in_usec = (clock_drift * 1000000) >> 32;
else
clock_drift_in_usec = -(((-clock_drift) * 1000000) >> 32);
if (clock_drift >= 0)
clock_drift_in_usec = (clock_drift * 1000000) >> 32;
else
clock_drift_in_usec = -(((-clock_drift) * 1000000) >> 32);
clock_drift_ppm = (1.0*clock_drift_in_usec)/(local_time_change>>32);
}
int64_t source_drift_usec;
if (play_segment_reference_frame != 0) {
if (conn->play_segment_reference_frame != 0) {
int64_t reference_timestamp;
uint64_t reference_timestamp_time, remote_reference_timestamp_time;
get_reference_timestamp_stuff(&reference_timestamp, &reference_timestamp_time,
&remote_reference_timestamp_time, conn);
uint64_t frame_difference = 0;
if (reference_timestamp >= play_segment_reference_frame)
frame_difference = (uint64_t)reference_timestamp - (uint64_t)play_segment_reference_frame;
if (reference_timestamp >= conn->play_segment_reference_frame)
frame_difference = (uint64_t)reference_timestamp - (uint64_t)conn->play_segment_reference_frame;
else // rollover
frame_difference =
(uint64_t)reference_timestamp + 0x100000000 - (uint64_t)play_segment_reference_frame;
(uint64_t)reference_timestamp + 0x100000000 - (uint64_t)conn->play_segment_reference_frame;
uint64_t frame_time_difference_calculated = (((uint64_t)frame_difference << 32) / 44100);
uint64_t frame_time_difference_actual =
remote_reference_timestamp_time -
play_segment_reference_frame_remote_time; // this is all done by reference to the
conn->play_segment_reference_frame_remote_time; // this is all done by reference to the
// sources' system clock
// debug(1,"%llu frames since play started, %llu usec calculated, %llu usec
// actual",frame_difference, (frame_time_difference_calculated*1000000)>>32,
@@ -500,18 +505,14 @@ void *rtp_timing_receiver(void *arg) {
// config.output->delay(&current_delay);
//}
// Useful for troubleshooting:
// clock_drift between source and local clock -- +ve means source is faster
// session_corrections -- the amount of correction done, in microseconds. +ve means frames
// added
// current_delay = delay in DAC buffer in frames
// source_drift_usec = how much faster (+ve) or slower the source DAC is running relative
// to the source clock
// buffer_occupancy = the number of buffers occupied. Crude, but should show no long term
// trend if source and device are in sync.
// return_time = the time from soliciting a timing packet to getting it back. It should be
// short ( < 5 ms) and pretty consistent.
// debug(1, "%lld\t%lld\t%ld\t%lld\t%u\t%llu",
// clock_drift_in_usec,(session_corrections*1000000)/44100,current_delay,source_drift_usec,buffer_occupancy,(return_time*1000000)>>32);
//debug(1, "clock_drift_ppm %f\tchosen %5d\tsource_drift_usec %10.1lld\treturn_time_in_usec %10.1llu",
//clock_drift_ppm,
//chosen,
//(session_corrections*1000000)/44100,
//current_delay,
//source_drift_usec,
//buffer_occupancy,
//(return_time*1000000)>>32);
} else {
debug(1, "Timing port -- Unknown RTP packet of type 0x%02X length %d.", packet[1], nread);