eSpeak NG is an open source speech synthesizer that supports more than hundred languages and accents.
You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

wave_sada.cpp 15KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594
  1. /***************************************************************************
  2. * Copyright (C) 2008, Sun Microsystems, Inc. *
  3. * eSpeak driver for Solaris Audio Device Architecture (SADA) *
  4. * Written by Willie Walker, based on the eSpeak PulseAudio driver *
  5. * from Gilles Casse *
  6. * *
  7. * This program is free software; you can redistribute it and/or modify *
  8. * it under the terms of the GNU General Public License as published by *
  9. * the Free Software Foundation; either version 3 of the License, or *
  10. * (at your option) any later version. *
  11. * *
  12. * This program is distributed in the hope that it will be useful, *
  13. * but WITHOUT ANY WARRANTY; without even the implied warranty of *
  14. * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the *
  15. * GNU General Public License for more details. *
  16. * *
  17. * You should have received a copy of the GNU General Public License *
  18. * along with this program; if not, write to the *
  19. * Free Software Foundation, Inc., *
  20. * 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA. *
  21. ***************************************************************************/
  22. #include "speech.h"
  23. #ifdef USE_ASYNC
  24. // This source file is only used for asynchronious modes
  25. #include <errno.h>
  26. #include <string.h>
  27. #include <stropts.h>
  28. #include <assert.h>
  29. #include <stdlib.h>
  30. #include <unistd.h>
  31. #include <fcntl.h>
  32. #include <sys/audioio.h>
  33. #include "wave.h"
  34. #include "debug.h"
  35. enum {ONE_BILLION=1000000000};
  36. #define SAMPLE_RATE 22050
  37. #define SAMPLE_SIZE 16
  38. #ifdef USE_SADA
  39. static t_wave_callback* my_callback_is_output_enabled=NULL;
  40. static const char *sun_audio_device = "/dev/audio";
  41. static int sun_audio_fd = -1;
  42. // The total number of 16-bit samples sent to be played via the
  43. // wave_write method.
  44. //
  45. static uint32_t total_samples_sent;
  46. // The total number of samples sent to be played via the wave_write
  47. // method, but which were never played because of a call to
  48. // wave_close.
  49. //
  50. static uint32_t total_samples_skipped;
  51. // The last known playing index after a call to wave_close.
  52. //
  53. static uint32_t last_play_position=0;
  54. //>
  55. // wave_init
  56. //
  57. // DESCRIPTION:
  58. //
  59. // initializes the audio subsytem.
  60. //
  61. // GLOBALS USED/MODIFIED:
  62. //
  63. // sun_audio_fd: modified to hold the file descriptor of the opened
  64. // audio device.
  65. //
  66. //<wave_init
  67. void wave_init(int srate) {
  68. ENTER("wave_init");
  69. audio_info_t ainfo;
  70. char *audio_device = NULL;
  71. wave_samplerate = srate;
  72. audio_device = getenv("AUDIODEV");
  73. if (audio_device != NULL) {
  74. if ((sun_audio_fd = open(audio_device, O_WRONLY)) < 0) {
  75. SHOW("wave_init() could not open: %s (%d)\n",
  76. audio_device, sun_audio_fd);
  77. }
  78. }
  79. if (sun_audio_fd < 0) {
  80. if ((sun_audio_fd = open(sun_audio_device, O_WRONLY)) < 0) {
  81. SHOW("wave_init() could not open: %s (%d)\n",
  82. sun_audio_device, sun_audio_fd);
  83. }
  84. }
  85. SHOW("wave_init() sun_audio_fd: %d\n", sun_audio_fd);
  86. if (sun_audio_fd < 0) {
  87. return;
  88. }
  89. ioctl(sun_audio_fd, AUDIO_GETINFO, &ainfo);
  90. SHOW("wave_init() play buffer size: %d\n", ainfo.play.buffer_size);
  91. ainfo.play.encoding = AUDIO_ENCODING_LINEAR;
  92. ainfo.play.channels = 1;
  93. ainfo.play.sample_rate = wave_samplerate;
  94. ainfo.play.precision = SAMPLE_SIZE;
  95. if (ioctl(sun_audio_fd, AUDIO_SETINFO, &ainfo) == -1) {
  96. SHOW("wave_init() failed to set audio params: %s\n", strerror(errno));
  97. close(sun_audio_fd);
  98. return;
  99. }
  100. }
  101. //>
  102. // wave_open
  103. //
  104. // DESCRIPTION:
  105. //
  106. // opens the audio subsystem given a specific API (e.g., "alsa",
  107. // "oss", ...). We ignore the_api and just return the sun_audio_fd we
  108. // opened in wave_init. This return value will be passed in as the
  109. // theHandler parameter in all other methods.
  110. //
  111. // PARAMETERS:
  112. //
  113. // the_api: "alsa", "oss" (ignored)
  114. //
  115. // GLOBALS USED/MODIFIED:
  116. //
  117. // sun_audio_fd: used as return value
  118. //
  119. // RETURNS:
  120. //
  121. // sun_audio_fd opened in wave_init, which is passed in as theHandler
  122. // parameter in all other methods
  123. //
  124. //<wave_open
  125. void* wave_open(const char* the_api)
  126. {
  127. ENTER("wave_open");
  128. return((void*) sun_audio_fd);
  129. }
  130. //>
  131. // wave_write
  132. //
  133. // DESCRIPTION:
  134. //
  135. // Meant to be asynchronous, it supplies the wave sample to the lower
  136. // audio layer and returns. The sample is played later on. [[[WDW -
  137. // we purposely do not open the audio device as non-blocking because
  138. // managing that would be a pain. So, we rely a lot upon fifo.cpp and
  139. // event.cpp to not overload us, allowing us to get away with a
  140. // blocking write. event.cpp:polling_thread in particular appears to
  141. // use get_remaining_time to prevent flooding.]]]
  142. //
  143. // PARAMETERS:
  144. //
  145. // theHandler: the audio device file descriptor
  146. // theMono16BitsWaveBuffer: the audio data
  147. // theSize: the number of bytes (not 16-bit samples)
  148. //
  149. // GLOBALS USED/MODIFIED:
  150. //
  151. // total_samples_sent: modified based upon 16-bit samples sent
  152. //
  153. // RETURNS:
  154. //
  155. // the number of bytes (not 16-bit samples) sent
  156. //
  157. //<wave_write
  158. size_t wave_write(void* theHandler,
  159. char* theMono16BitsWaveBuffer,
  160. size_t theSize)
  161. {
  162. size_t num;
  163. ENTER("wave_write");
  164. if (my_callback_is_output_enabled && (0==my_callback_is_output_enabled())) {
  165. SHOW_TIME("wave_write > my_callback_is_output_enabled: no!");
  166. return 0;
  167. }
  168. #if defined(BYTE_ORDER) && BYTE_ORDER == BIG_ENDIAN
  169. {
  170. // BIG-ENDIAN, swap the order of bytes in each sound sample
  171. int c;
  172. char *out_ptr;
  173. char *out_end;
  174. out_ptr = (char *)theMono16BitsWaveBuffer;
  175. out_end = out_ptr + theSize;
  176. while(out_ptr < out_end)
  177. {
  178. c = out_ptr[0];
  179. out_ptr[0] = out_ptr[1];
  180. out_ptr[1] = c;
  181. out_ptr += 2;
  182. }
  183. }
  184. #endif
  185. num = write((int) theHandler, theMono16BitsWaveBuffer, theSize);
  186. // Keep track of the total number of samples sent -- we use this in
  187. // wave_get_read_position and also use it to help calculate the
  188. // total_samples_skipped in wave_close.
  189. //
  190. total_samples_sent += num / 2;
  191. if (num < theSize) {
  192. SHOW("ERROR: wave_write only wrote %d of %d bytes\n", num, theSize);
  193. } else {
  194. SHOW("wave_write wrote %d bytes\n", theSize);
  195. }
  196. SHOW_TIME("wave_write > LEAVE");
  197. return num;
  198. }
  199. //>
  200. // wave_close
  201. //
  202. // DESCRIPTION:
  203. //
  204. // Does what SADA normally would call a flush, which means to cease
  205. // all audio production in progress and throw any remaining audio
  206. // away. [[[WDW - see comment in wave_flush.]]]
  207. //
  208. // PARAMETERS:
  209. //
  210. // theHandler: the audio device file descriptor
  211. //
  212. // GLOBALS USED/MODIFIED:
  213. //
  214. // last_play_position: modified to reflect play position the last time
  215. // this method was called
  216. // total_samples_sent: used to help calculate total_samples_skipped
  217. // total_samples_skipped: modified to hold the total number of 16-bit
  218. // samples sent to wave_write, but which were
  219. // never played
  220. // sun_audio_fd: used because some calls to wave_close seem to
  221. // pass a NULL for theHandler for some odd reason
  222. //
  223. // RETURNS:
  224. //
  225. // The result of the ioctl call (non-0 means failure)
  226. //
  227. //<wave_close
  228. int wave_close(void* theHandler)
  229. {
  230. int ret;
  231. audio_info_t ainfo;
  232. int audio_fd = (int) theHandler;
  233. if (!audio_fd) {
  234. audio_fd = sun_audio_fd;
  235. }
  236. ENTER("wave_close");
  237. // [[[WDW: maybe do a pause/resume ioctl???]]]
  238. ret = ioctl(audio_fd, I_FLUSH, FLUSHRW);
  239. ioctl(audio_fd, AUDIO_GETINFO, &ainfo);
  240. // Calculate the number of samples that won't get
  241. // played. We also keep track of the last_play_position
  242. // because wave_close can be called multiple times
  243. // before another call to wave_write.
  244. //
  245. if (last_play_position != ainfo.play.samples) {
  246. last_play_position = ainfo.play.samples;
  247. total_samples_skipped = total_samples_sent - last_play_position;
  248. }
  249. SHOW_TIME("wave_close > LEAVE");
  250. return ret;
  251. }
  252. //>
  253. // wave_is_busy
  254. //
  255. // DESCRIPTION:
  256. //
  257. // Returns a non-0 value if audio is being played.
  258. //
  259. // PARAMETERS:
  260. //
  261. // theHandler: the audio device file descriptor
  262. //
  263. // GLOBALS USED/MODIFIED:
  264. //
  265. // sun_audio_fd: used because some calls to wave_is_busy seem to
  266. // pass a NULL for theHandler for some odd reason
  267. //
  268. // RETURNS:
  269. //
  270. // A non-0 value if audio is being played
  271. //
  272. //<wave_is_busy
  273. int wave_is_busy(void* theHandler)
  274. {
  275. uint32_t time;
  276. if (total_samples_sent >= 1) {
  277. wave_get_remaining_time(total_samples_sent - 1, &time);
  278. } else {
  279. time = 0;
  280. }
  281. return time != 0;
  282. }
  283. //>
  284. // wave_terminate
  285. //
  286. // DESCRIPTION:
  287. //
  288. // Used to end our session with eSpeak.
  289. //
  290. // GLOBALS USED/MODIFIED:
  291. //
  292. // sun_audio_fd: modified - closed and set to -1
  293. //
  294. //<wave_terminate
  295. void wave_terminate()
  296. {
  297. ENTER("wave_terminate");
  298. close(sun_audio_fd);
  299. sun_audio_fd = -1;
  300. SHOW_TIME("wave_terminate > LEAVE");
  301. }
  302. //>
  303. // wave_flush
  304. //
  305. // DESCRIPTION:
  306. //
  307. // Appears to want to tell the audio subsystem to make sure it plays
  308. // the audio. In our case, the system is already doing this, so this
  309. // is basically a no-op. [[[WDW - if you do a drain, you block, so
  310. // don't do that. In addition the typical SADA notion of flush is
  311. // currently handled by wave_close. I think this is most likely just
  312. // terminology conflict between eSpeak and SADA.]]]
  313. //
  314. // PARAMETERS:
  315. //
  316. // theHandler: the audio device file descriptor
  317. //
  318. //<wave_flush
  319. void wave_flush(void* theHandler)
  320. {
  321. ENTER("wave_flush");
  322. //ioctl((int) theHandler, AUDIO_DRAIN, 0);
  323. SHOW_TIME("wave_flush > LEAVE");
  324. }
  325. //>
  326. // wave_set_callback_is_output_enabled
  327. //
  328. // DESCRIPTION:
  329. //
  330. // Sets the callback to call from wave_write before it sends data to
  331. // be played. It helps wave_write determine if the data should be
  332. // thrown away or not.
  333. //
  334. // PARAMETERS:
  335. //
  336. // cb: the callback to call from wave_write
  337. //
  338. //<wave_set_callback_is_output_enabled
  339. void wave_set_callback_is_output_enabled(t_wave_callback* cb)
  340. {
  341. my_callback_is_output_enabled = cb;
  342. }
  343. //>
  344. // wave_test_get_write_buffer
  345. //
  346. // DESCRIPTION:
  347. //
  348. // Unnecessary and is used for debug output from
  349. // speak_lib.cpp:dispatch_audio.
  350. //
  351. // RETURNS:
  352. //
  353. // NULL
  354. //
  355. //<wave_test_get_write_buffer
  356. void *wave_test_get_write_buffer()
  357. {
  358. return NULL;
  359. }
  360. //>
  361. // wave_get_read_position
  362. //
  363. // DESCRIPTION:
  364. //
  365. // Concerns the sample which is currently played by the audio layer,
  366. // where 'sample' is a small buffer of synthesized wave data,
  367. // identified so that the user callback could be called when the
  368. // 'sample' is really played. The identifier is returned by
  369. // wave_get_write_position. This method is unused.
  370. //
  371. // PARAMETERS:
  372. //
  373. // theHandler: the audio device file descriptor
  374. //
  375. // RETURNS:
  376. //
  377. // The total number of 16-bit samples played by the audio system
  378. // so far.
  379. //
  380. //<wave_get_read_position
  381. uint32_t wave_get_read_position(void* theHandler)
  382. {
  383. audio_info_t ainfo;
  384. ENTER("wave_get_read_position");
  385. ioctl((int) theHandler, AUDIO_GETINFO, &ainfo);
  386. SHOW("wave_get_read_position: %d\n", ainfo.play.samples);
  387. SHOW_TIME("wave_get_read_position > LEAVE");
  388. return ainfo.play.samples;
  389. }
  390. //>
  391. // wave_get_write_position
  392. //
  393. // DESCRIPTION:
  394. //
  395. // Returns an identifier for a new sample, where 'sample' is a small
  396. // buffer of synthesized wave data, identified so that the user
  397. // callback could be called when the 'sample' is really played. This
  398. // implementation views the audio as one long continuous stream of
  399. // 16-bit samples.
  400. //
  401. // PARAMETERS:
  402. //
  403. // theHandler: the audio device file descriptor
  404. //
  405. // GLOBALS USED/MODIFIED:
  406. //
  407. // total_samples_sent: used as the return value
  408. //
  409. // RETURNS:
  410. //
  411. // total_samples_sent, which is the index for the end of this long
  412. // continuous stream. [[[WDW: with a unit32_t managing 16-bit
  413. // samples at 22050Hz, we have about 54 hours of play time before
  414. // the index wraps back to 0. We don't handle that wrapping, so
  415. // the behavior after 54 hours of play time is undefined.]]]
  416. //
  417. //<wave_get_write_position
  418. uint32_t wave_get_write_position(void* theHandler)
  419. {
  420. ENTER("wave_get_write_position");
  421. SHOW("wave_get_write_position: %d\n", total_samples_sent);
  422. SHOW_TIME("wave_get_write_position > LEAVE");
  423. return total_samples_sent;
  424. }
  425. //>
  426. // wave_get_remaining_time
  427. //
  428. // DESCRIPTION:
  429. //
  430. // Returns the remaining time (in ms) before the sample is played.
  431. // The sample in this case is a return value from a previous call to
  432. // wave_get_write_position.
  433. //
  434. // PARAMETERS:
  435. //
  436. // sample: an index returned from wave_get_write_position representing
  437. // an index into the long continuous stream of 16-bit samples
  438. // time: a return value representing the delay in milliseconds until
  439. // sample is played. A value of 0 means the sample is either
  440. // currently being played or it has already been played.
  441. //
  442. // GLOBALS USED/MODIFIED:
  443. //
  444. // sun_audio_fd: used to determine total number of samples played by
  445. // the audio system
  446. // total_samples_skipped: used in remaining time calculation
  447. //
  448. // RETURNS:
  449. //
  450. // Time in milliseconds before the sample is played or 0 if the sample
  451. // is currently playing or has already been played.
  452. //
  453. //<wave_get_remaining_time
  454. int wave_get_remaining_time(uint32_t sample, uint32_t* time)
  455. {
  456. uint32_t a_time=0;
  457. uint32_t actual_index;
  458. audio_info_t ainfo;
  459. ENTER("wave_get_remaining_time");
  460. if (!time) {
  461. return(-1);
  462. SHOW_TIME("wave_get_remaining_time > LEAVE");
  463. }
  464. ioctl(sun_audio_fd, AUDIO_GETINFO, &ainfo);
  465. // See if this sample has already been played or is currently
  466. // playing.
  467. //
  468. actual_index = sample - total_samples_skipped;
  469. if ((sample < total_samples_skipped) ||
  470. (actual_index <= ainfo.play.samples)) {
  471. *time = 0;
  472. } else {
  473. a_time = ((actual_index - ainfo.play.samples) * 1000) / wave_samplerate;
  474. *time = (uint32_t) a_time;
  475. }
  476. SHOW("wave_get_remaining_time for %d: %d\n", sample, *time);
  477. SHOW_TIME("wave_get_remaining_time > LEAVE");
  478. return 0;
  479. }
  480. #else
  481. // notdef USE_SADA
  482. void wave_init() {}
  483. void* wave_open(const char* the_api) {return (void *)1;}
  484. size_t wave_write(void* theHandler, char* theMono16BitsWaveBuffer, size_t theSize) {return theSize;}
  485. int wave_close(void* theHandler) {return 0;}
  486. int wave_is_busy(void* theHandler) {return 0;}
  487. void wave_terminate() {}
  488. uint32_t wave_get_read_position(void* theHandler) {return 0;}
  489. uint32_t wave_get_write_position(void* theHandler) {return 0;}
  490. void wave_flush(void* theHandler) {}
  491. typedef int (t_wave_callback)(void);
  492. void wave_set_callback_is_output_enabled(t_wave_callback* cb) {}
  493. extern void* wave_test_get_write_buffer() {return NULL;}
  494. int wave_get_remaining_time(uint32_t sample, uint32_t* time)
  495. {
  496. if (!time) return(-1);
  497. *time = (uint32_t)0;
  498. return 0;
  499. }
  500. #endif // of USE_PORTAUDIO
  501. //>
  502. //<clock_gettime2, add_time_in_ms
  503. void clock_gettime2(struct timespec *ts)
  504. {
  505. struct timeval tv;
  506. if (!ts)
  507. {
  508. return;
  509. }
  510. assert (gettimeofday(&tv, NULL) != -1);
  511. ts->tv_sec = tv.tv_sec;
  512. ts->tv_nsec = tv.tv_usec*1000;
  513. }
  514. void add_time_in_ms(struct timespec *ts, int time_in_ms)
  515. {
  516. if (!ts)
  517. {
  518. return;
  519. }
  520. uint64_t t_ns = (uint64_t)ts->tv_nsec + 1000000 * (uint64_t)time_in_ms;
  521. while(t_ns >= ONE_BILLION)
  522. {
  523. SHOW("event > add_time_in_ms ns: %d sec %Lu nsec \n", ts->tv_sec, t_ns);
  524. ts->tv_sec += 1;
  525. t_ns -= ONE_BILLION;
  526. }
  527. ts->tv_nsec = (long int)t_ns;
  528. }
  529. #endif // USE_ASYNC
  530. //>