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

Complete replacement of the PipeWire backend with a new implementation with full synchrnoisation. Works well, but still preliminary.

Browse Source
This commit is contained in:
Mike Brady
2023-10-01 09:30:29 +01:00
parent afdf0d8632
commit bde2bcc8b4
1 changed files with 114 additions and 146 deletions
Download Patch File Download Diff File Expand all files Collapse all files
audio_pw.c
+114 -146
View File
@@ -1,6 +1,6 @@
/*
* Asynchronous PipeWire Backend. This file is part of Shairport Sync.
* Copyright (c) Mike Brady 2017-2023
* Copyright (c) Mike Brady 2023
* All rights reserved.
*
* Permission is hereby granted, free of charge, to any person
@@ -24,6 +24,10 @@
* OTHER DEALINGS IN THE SOFTWARE.
*/
// This uses ideas from the tone generator sample code at:
// https://github.com/PipeWire/pipewire/blob/master/src/examples/audio-src.c
// Thanks to the Wim Taymans.
#include "audio.h"
#include "common.h"
#include <errno.h>
@@ -39,7 +43,6 @@
#define DEFAULT_FORMAT SPA_AUDIO_FORMAT_S16_LE
#define DEFAULT_RATE 44100
#define DEFAULT_CHANNELS 2
#define DEFAULT_VOLUME 0.7
// Four seconds buffer -- should be plenty
#define buffer_allocation 44100 * 4 * 2 * 2
@@ -53,9 +56,9 @@ static size_t audio_occupancy;
uint64_t starting_time;
struct timing_data {
int pw_time_is_valid; //set when the pw_time has been set
int pw_time_is_valid; // set when the pw_time has been set
struct pw_time time_info; // information about the last time a process callback occurred
size_t frames; // the number of frames sent at that time
size_t frames; // the number of frames sent at that time
};
// to avoid using a mutex, write the same data twice and check they are the same
@@ -66,8 +69,6 @@ struct timing_data timing_data_1, timing_data_2;
struct data {
struct pw_thread_loop *loop;
struct pw_stream *stream;
double accumulator;
};
// the pipewire global data structure
@@ -75,14 +76,46 @@ struct data data = {
0,
};
static int fill(void *dest, int max_frames, int stride) {
size_t bytes_we_can_transfer = max_frames * stride;
static void on_state_changed(__attribute__((unused)) void *userdata, enum pw_stream_state old,
enum pw_stream_state state,
__attribute__((unused)) const char *error) {
// struct pw_data *pw = userdata;
debug(3, "pw: stream state changed %s -> %s", pw_stream_state_as_string(old),
pw_stream_state_as_string(state));
}
static void on_process(void *userdata) {
struct data *data = userdata;
int n_frames = 0;
pthread_mutex_lock(&buffer_mutex);
if (bytes_we_can_transfer > audio_occupancy)
bytes_we_can_transfer = audio_occupancy;
pthread_mutex_unlock(&buffer_mutex);
if (bytes_we_can_transfer > 0) {
if (audio_occupancy > 0) {
// get a buffer to see how big it can be
struct pw_buffer *b = pw_stream_dequeue_buffer(data->stream);
if (b == NULL) {
pw_log_warn("out of buffers: %m");
return;
}
struct spa_buffer *buf = b->buffer;
uint8_t *dest = buf->datas[0].data;
if (dest == NULL) // the first data block does not contain a data pointer
return;
int stride = sizeof(int16_t) * DEFAULT_CHANNELS;
int max_possible_frames = SPA_MIN(b->requested, buf->datas[0].maxsize / stride);
size_t bytes_we_can_transfer = max_possible_frames * stride;
if (bytes_we_can_transfer > audio_occupancy)
bytes_we_can_transfer = audio_occupancy;
n_frames = bytes_we_can_transfer / stride;
size_t bytes_to_end_of_buffer = (size_t)(audio_umb - audio_toq); // must be zero or positive
if (bytes_we_can_transfer <= bytes_to_end_of_buffer) {
// the bytes are all in a row in the audio buffer
memcpy(dest, audio_toq, bytes_we_can_transfer);
@@ -96,81 +129,42 @@ static int fill(void *dest, int max_frames, int stride) {
memcpy(new_dest, audio_lmb, bytes_we_can_transfer - first_portion_to_write);
audio_toq = audio_lmb + bytes_we_can_transfer - first_portion_to_write;
}
// lock
pthread_mutex_lock(&buffer_mutex);
buf->datas[0].chunk->offset = 0;
buf->datas[0].chunk->stride = stride;
buf->datas[0].chunk->size = n_frames * stride;
pw_stream_queue_buffer(data->stream, b);
debug(3, "Queueing %d frames for output.", n_frames);
audio_occupancy -= bytes_we_can_transfer;
pthread_mutex_unlock(&buffer_mutex);
// unlock
}
return bytes_we_can_transfer / stride; // back to numbers of frames
}
pthread_mutex_unlock(&buffer_mutex);
/* our data processing function is in general:
*
* struct pw_buffer *b;
* b = pw_stream_dequeue_buffer(stream);
*
* .. generate stuff in the buffer ...
*
* pw_stream_queue_buffer(stream, b);
*/
static void on_process(void *userdata) {
struct data *data = userdata;
struct pw_time time_info;
memset(&time_info, 0, sizeof(time_info));
struct pw_buffer *b;
struct spa_buffer *buf;
int max_possible_frames, n_frames, stride;
uint8_t *p;
if ((b = pw_stream_dequeue_buffer(data->stream)) == NULL) {
pw_log_warn("out of buffers: %m");
return;
}
buf = b->buffer;
if ((p = buf->datas[0].data) == NULL) // the first data block does not contain a data pointer
return;
stride = sizeof(int16_t) * DEFAULT_CHANNELS;
max_possible_frames = SPA_MIN(b->requested, buf->datas[0].maxsize / stride);
do {
n_frames = fill(p, max_possible_frames, stride);
if (n_frames == 0) {
usleep(1000);
}
} while(n_frames == 0);
buf->datas[0].chunk->offset = 0;
buf->datas[0].chunk->stride = stride;
buf->datas[0].chunk->size = n_frames * stride;
debug(3, "Queueing %d frames for output.", n_frames);
if (pw_stream_get_time_n(data->stream, &timing_data_1.time_info, sizeof(time_info)) == 0)
timing_data_1.frames = n_frames;
if (pw_stream_get_time_n(data->stream, &timing_data_1.time_info, sizeof(struct timing_data)) == 0)
timing_data_1.pw_time_is_valid = 1;
else
timing_data_1.pw_time_is_valid = 0;
__sync_synchronize();
memcpy((char *)&timing_data_2, (char *)&timing_data_1, sizeof(struct timing_data));
__sync_synchronize();
pw_stream_queue_buffer(data->stream, b);
}
static const struct pw_stream_events stream_events = {
PW_VERSION_STREAM_EVENTS,
.process = on_process,
};
PW_VERSION_STREAM_EVENTS, .process = on_process, .state_changed = on_state_changed};
static void deinit(void) {
pw_thread_loop_stop(data.loop);
pw_stream_destroy(data.stream);
pw_thread_loop_destroy(data.loop);
pw_deinit();
free(audio_lmb); // deallocate that buffer
}
static int init(__attribute__((unused)) int argc, __attribute__((unused)) char **argv) {
debug(1, "pw_init");
// set up default values first
memset(&timing_data_1,0,sizeof(struct timing_data));
memset(&timing_data_2,0,sizeof(struct timing_data));
memset(&timing_data_1, 0, sizeof(struct timing_data));
memset(&timing_data_2, 0, sizeof(struct timing_data));
config.audio_backend_buffer_desired_length = 0.35;
config.audio_backend_buffer_interpolation_threshold_in_seconds =
0.02; // below this, soxr interpolation will not occur -- it'll be basic interpolation
@@ -207,30 +201,17 @@ static int init(__attribute__((unused)) int argc, __attribute__((unused)) char *
int largc = 0;
pw_init(&largc, NULL);
/* make a main loop. If you already have another main loop, you can add
* the fd of this pipewire mainloop to it. */
data.loop = pw_thread_loop_new("tone-generator", NULL);
/* make a threaded loop. */
data.loop = pw_thread_loop_new("shairport-sync", NULL);
pw_thread_loop_lock(data.loop);
pw_thread_loop_start(data.loop);
/* Create a simple stream, the simple stream manages the core and remote
* objects for you if you don't need to deal with them.
*
* If you plan to autoconnect your stream, you need to provide at least
* media, category and role properties.
*
* Pass your events and a user_data pointer as the last arguments. This
* will inform you about the stream state. The most important event
* you need to listen to is the process event where you need to produce
* the data.
*/
props = pw_properties_new(PW_KEY_MEDIA_TYPE, "Audio", PW_KEY_MEDIA_CATEGORY, "Playback",
PW_KEY_MEDIA_ROLE, "Music", PW_KEY_APP_NAME, "Shairport Sync", NULL);
data.stream = pw_stream_new_simple(pw_thread_loop_get_loop(data.loop), "audio-src-tg", props,
data.stream = pw_stream_new_simple(pw_thread_loop_get_loop(data.loop), "shairport-sync", props,
&stream_events, &data);
/* Make one parameter with the supported formats. The SPA_PARAM_EnumFormat
@@ -248,58 +229,42 @@ static int init(__attribute__((unused)) int argc, __attribute__((unused)) char *
params, 1);
pw_thread_loop_unlock(data.loop);
debug(1, "pa_init done");
return 0;
}
static void deinit(void) {
pw_thread_loop_unlock(data.loop);
debug(1, "pipewire: deinit deactivated the stream");
pw_thread_loop_signal(data.loop, false);
pw_thread_loop_wait(data.loop);
debug(1, "pipewire: deinit loop wait done");
pw_thread_loop_stop(data.loop);
debug(1, "pipewire: deinit loop stopped");
pw_stream_destroy(data.stream);
debug(1, "pipewire: deinit stream destroyed");
pw_thread_loop_destroy(data.loop);
debug(1, "pipewire: deinit loop destroyed");
pw_deinit();
debug(1, "pipewire: pw_deinit terminated");
free(audio_lmb); // deallocate that buffer
debug(1, "pipewire: deinit done");
}
static void start(__attribute__((unused)) int sample_rate,
__attribute__((unused)) int sample_format) {
}
__attribute__((unused)) int sample_format) {}
static int play(__attribute__((unused)) void *buf, int samples,
__attribute__((unused)) int sample_type, __attribute__((unused)) uint32_t timestamp,
__attribute__((unused)) uint64_t playtime) {
__attribute__((unused)) uint64_t playtime) {
// copy the samples into the queue
debug(3, "play %u samples; %u bytes already in the buffer.", samples, audio_occupancy);
size_t bytes_to_transfer = samples * 2 * 2;
size_t space_to_end_of_buffer = audio_umb - audio_eoq;
if (space_to_end_of_buffer >= bytes_to_transfer) {
memcpy(audio_eoq, buf, bytes_to_transfer);
pthread_mutex_lock(&buffer_mutex);
pthread_mutex_lock(&buffer_mutex);
size_t bytes_available = audio_size - audio_occupancy;
if (bytes_available < bytes_to_transfer)
bytes_to_transfer = bytes_available;
if (bytes_to_transfer > 0) {
size_t space_to_end_of_buffer = audio_umb - audio_eoq;
if (space_to_end_of_buffer >= bytes_to_transfer) {
memcpy(audio_eoq, buf, bytes_to_transfer);
audio_eoq += bytes_to_transfer;
} else {
memcpy(audio_eoq, buf, space_to_end_of_buffer);
buf += space_to_end_of_buffer;
memcpy(audio_lmb, buf, bytes_to_transfer - space_to_end_of_buffer);
audio_eoq = audio_lmb + bytes_to_transfer - space_to_end_of_buffer;
}
audio_occupancy += bytes_to_transfer;
pthread_mutex_unlock(&buffer_mutex);
audio_eoq += bytes_to_transfer;
} else {
memcpy(audio_eoq, buf, space_to_end_of_buffer);
buf += space_to_end_of_buffer;
memcpy(audio_lmb, buf, bytes_to_transfer - space_to_end_of_buffer);
pthread_mutex_lock(&buffer_mutex);
audio_occupancy += bytes_to_transfer;
pthread_mutex_unlock(&buffer_mutex);
audio_eoq = audio_lmb + bytes_to_transfer - space_to_end_of_buffer;
}
debug(3, "play %d samples; %d bytes in buffer.", samples, audio_occupancy);
pthread_mutex_unlock(&buffer_mutex);
return 0;
}
int delay(long *the_delay) {
long result = 0;
int reply = 0;
// find out what's already in the PipeWire system and when
struct timing_data timing_data;
int loop_count = 1;
@@ -307,40 +272,44 @@ int delay(long *the_delay) {
memcpy(&timing_data, (char *)&timing_data_1, sizeof(struct timing_data));
__sync_synchronize();
if (memcmp(&timing_data, (char *)&timing_data_2, sizeof(struct timing_data)) != 0) {
usleep(2); // microseconds
loop_count++;
__sync_synchronize();
usleep(2); // microseconds
loop_count++;
__sync_synchronize();
}
} while ((memcmp(&timing_data, (char *)&timing_data_2, sizeof(struct timing_data)) != 0) && (loop_count < 10));
} while ((memcmp(&timing_data, (char *)&timing_data_2, sizeof(struct timing_data)) != 0) &&
(loop_count < 10));
long total_delay_now_frames_long = 0;
if ((loop_count < 10) && (timing_data.pw_time_is_valid != 0)) {
struct timespec time_now;
struct timespec time_now;
clock_gettime(CLOCK_MONOTONIC, &time_now);
int64_t interval_from_process_time_to_now = SPA_TIMESPEC_TO_NSEC(&time_now) - timing_data.time_info.now;
int64_t interval_from_process_time_to_now =
SPA_TIMESPEC_TO_NSEC(&time_now) - timing_data.time_info.now;
int64_t delay_in_ns = timing_data.time_info.delay + timing_data.time_info.buffered;
delay_in_ns = delay_in_ns * 1000000000;
delay_in_ns = delay_in_ns * timing_data.time_info.rate.num;
delay_in_ns = delay_in_ns / timing_data.time_info.rate.denom;
int64_t total_delay_now_ns = delay_in_ns - interval_from_process_time_to_now;
int64_t total_delay_now_frames = (total_delay_now_ns * 44100)/1000000000 + timing_data.frames;
int64_t total_delay_now_frames = (total_delay_now_ns * 44100) / 1000000000 + timing_data.frames;
total_delay_now_frames_long = total_delay_now_frames;
debug(3, "total delay in frames: % " PRId64 ", %ld.", total_delay_now_frames, total_delay_now_frames_long);
debug(3,
"interval_from_process_time_to_now: %" PRId64 " ns, "
"delay_in_ns: %" PRId64 ", queued: %" PRId64 ", buffered: %" PRId64 ".",
// delay_timing_data.time_info.rate.num, delay_timing_data.time_info.rate.denom,
interval_from_process_time_to_now, delay_in_ns,
timing_data.time_info.queued, timing_data.time_info.buffered);
debug(3, "total delay in frames: %ld.", total_delay_now_frames_long);
if (timing_data.time_info.queued != 0) {
debug(1, "buffers queued: %d", timing_data.time_info.queued);
}
/*
debug(3,
"interval_from_process_time_to_now: %" PRId64 " ns, "
"delay_in_ns: %" PRId64 ", queued: %" PRId64 ", buffered: %" PRId64 ".",
// delay_timing_data.time_info.rate.num, delay_timing_data.time_info.rate.denom,
interval_from_process_time_to_now, delay_in_ns,
timing_data.time_info.queued, timing_data.time_info.buffered);
*/
} else {
debug(1, "can't get time info.");
}
long result = 0;
int reply = 0;
pthread_mutex_lock(&buffer_mutex);
result = total_delay_now_frames_long + audio_occupancy / (2 * 2);
pthread_mutex_unlock(&buffer_mutex);
@@ -348,19 +317,18 @@ int delay(long *the_delay) {
return reply;
}
static void flush(void) {
pthread_mutex_lock(&buffer_mutex);
audio_toq = audio_eoq = audio_lmb;
audio_umb = audio_lmb + audio_size;
pthread_mutex_lock(&buffer_mutex);
audio_occupancy = 0;
pthread_mutex_unlock(&buffer_mutex);
}
static void stop(void) {
pthread_mutex_lock(&buffer_mutex);
audio_toq = audio_eoq = audio_lmb;
audio_umb = audio_lmb + audio_size;
pthread_mutex_lock(&buffer_mutex);
audio_occupancy = 0;
pthread_mutex_unlock(&buffer_mutex);
}