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

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595
  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. static uint32_t wave_samplerate;
  55. //>
  56. // wave_init
  57. //
  58. // DESCRIPTION:
  59. //
  60. // initializes the audio subsytem.
  61. //
  62. // GLOBALS USED/MODIFIED:
  63. //
  64. // sun_audio_fd: modified to hold the file descriptor of the opened
  65. // audio device.
  66. //
  67. //<wave_init
  68. void wave_init(int srate) {
  69. ENTER("wave_init");
  70. audio_info_t ainfo;
  71. char *audio_device = NULL;
  72. wave_samplerate = srate;
  73. audio_device = getenv("AUDIODEV");
  74. if (audio_device != NULL) {
  75. if ((sun_audio_fd = open(audio_device, O_WRONLY)) < 0) {
  76. SHOW("wave_init() could not open: %s (%d)\n",
  77. audio_device, sun_audio_fd);
  78. }
  79. }
  80. if (sun_audio_fd < 0) {
  81. if ((sun_audio_fd = open(sun_audio_device, O_WRONLY)) < 0) {
  82. SHOW("wave_init() could not open: %s (%d)\n",
  83. sun_audio_device, sun_audio_fd);
  84. }
  85. }
  86. SHOW("wave_init() sun_audio_fd: %d\n", sun_audio_fd);
  87. if (sun_audio_fd < 0) {
  88. return;
  89. }
  90. ioctl(sun_audio_fd, AUDIO_GETINFO, &ainfo);
  91. SHOW("wave_init() play buffer size: %d\n", ainfo.play.buffer_size);
  92. ainfo.play.encoding = AUDIO_ENCODING_LINEAR;
  93. ainfo.play.channels = 1;
  94. ainfo.play.sample_rate = wave_samplerate;
  95. ainfo.play.precision = SAMPLE_SIZE;
  96. if (ioctl(sun_audio_fd, AUDIO_SETINFO, &ainfo) == -1) {
  97. SHOW("wave_init() failed to set audio params: %s\n", strerror(errno));
  98. close(sun_audio_fd);
  99. return;
  100. }
  101. }
  102. //>
  103. // wave_open
  104. //
  105. // DESCRIPTION:
  106. //
  107. // opens the audio subsystem given a specific API (e.g., "alsa",
  108. // "oss", ...). We ignore the_api and just return the sun_audio_fd we
  109. // opened in wave_init. This return value will be passed in as the
  110. // theHandler parameter in all other methods.
  111. //
  112. // PARAMETERS:
  113. //
  114. // the_api: "alsa", "oss" (ignored)
  115. //
  116. // GLOBALS USED/MODIFIED:
  117. //
  118. // sun_audio_fd: used as return value
  119. //
  120. // RETURNS:
  121. //
  122. // sun_audio_fd opened in wave_init, which is passed in as theHandler
  123. // parameter in all other methods
  124. //
  125. //<wave_open
  126. void* wave_open(const char* the_api)
  127. {
  128. ENTER("wave_open");
  129. return((void*) sun_audio_fd);
  130. }
  131. //>
  132. // wave_write
  133. //
  134. // DESCRIPTION:
  135. //
  136. // Meant to be asynchronous, it supplies the wave sample to the lower
  137. // audio layer and returns. The sample is played later on. [[[WDW -
  138. // we purposely do not open the audio device as non-blocking because
  139. // managing that would be a pain. So, we rely a lot upon fifo.cpp and
  140. // event.cpp to not overload us, allowing us to get away with a
  141. // blocking write. event.cpp:polling_thread in particular appears to
  142. // use get_remaining_time to prevent flooding.]]]
  143. //
  144. // PARAMETERS:
  145. //
  146. // theHandler: the audio device file descriptor
  147. // theMono16BitsWaveBuffer: the audio data
  148. // theSize: the number of bytes (not 16-bit samples)
  149. //
  150. // GLOBALS USED/MODIFIED:
  151. //
  152. // total_samples_sent: modified based upon 16-bit samples sent
  153. //
  154. // RETURNS:
  155. //
  156. // the number of bytes (not 16-bit samples) sent
  157. //
  158. //<wave_write
  159. size_t wave_write(void* theHandler,
  160. char* theMono16BitsWaveBuffer,
  161. size_t theSize)
  162. {
  163. size_t num;
  164. ENTER("wave_write");
  165. if (my_callback_is_output_enabled && (0==my_callback_is_output_enabled())) {
  166. SHOW_TIME("wave_write > my_callback_is_output_enabled: no!");
  167. return 0;
  168. }
  169. #if defined(BYTE_ORDER) && BYTE_ORDER == BIG_ENDIAN
  170. {
  171. // BIG-ENDIAN, swap the order of bytes in each sound sample
  172. int c;
  173. char *out_ptr;
  174. char *out_end;
  175. out_ptr = (char *)theMono16BitsWaveBuffer;
  176. out_end = out_ptr + theSize;
  177. while(out_ptr < out_end)
  178. {
  179. c = out_ptr[0];
  180. out_ptr[0] = out_ptr[1];
  181. out_ptr[1] = c;
  182. out_ptr += 2;
  183. }
  184. }
  185. #endif
  186. num = write((int) theHandler, theMono16BitsWaveBuffer, theSize);
  187. // Keep track of the total number of samples sent -- we use this in
  188. // wave_get_read_position and also use it to help calculate the
  189. // total_samples_skipped in wave_close.
  190. //
  191. total_samples_sent += num / 2;
  192. if (num < theSize) {
  193. SHOW("ERROR: wave_write only wrote %d of %d bytes\n", num, theSize);
  194. } else {
  195. SHOW("wave_write wrote %d bytes\n", theSize);
  196. }
  197. SHOW_TIME("wave_write > LEAVE");
  198. return num;
  199. }
  200. //>
  201. // wave_close
  202. //
  203. // DESCRIPTION:
  204. //
  205. // Does what SADA normally would call a flush, which means to cease
  206. // all audio production in progress and throw any remaining audio
  207. // away. [[[WDW - see comment in wave_flush.]]]
  208. //
  209. // PARAMETERS:
  210. //
  211. // theHandler: the audio device file descriptor
  212. //
  213. // GLOBALS USED/MODIFIED:
  214. //
  215. // last_play_position: modified to reflect play position the last time
  216. // this method was called
  217. // total_samples_sent: used to help calculate total_samples_skipped
  218. // total_samples_skipped: modified to hold the total number of 16-bit
  219. // samples sent to wave_write, but which were
  220. // never played
  221. // sun_audio_fd: used because some calls to wave_close seem to
  222. // pass a NULL for theHandler for some odd reason
  223. //
  224. // RETURNS:
  225. //
  226. // The result of the ioctl call (non-0 means failure)
  227. //
  228. //<wave_close
  229. int wave_close(void* theHandler)
  230. {
  231. int ret;
  232. audio_info_t ainfo;
  233. int audio_fd = (int) theHandler;
  234. if (!audio_fd) {
  235. audio_fd = sun_audio_fd;
  236. }
  237. ENTER("wave_close");
  238. // [[[WDW: maybe do a pause/resume ioctl???]]]
  239. ret = ioctl(audio_fd, I_FLUSH, FLUSHRW);
  240. ioctl(audio_fd, AUDIO_GETINFO, &ainfo);
  241. // Calculate the number of samples that won't get
  242. // played. We also keep track of the last_play_position
  243. // because wave_close can be called multiple times
  244. // before another call to wave_write.
  245. //
  246. if (last_play_position != ainfo.play.samples) {
  247. last_play_position = ainfo.play.samples;
  248. total_samples_skipped = total_samples_sent - last_play_position;
  249. }
  250. SHOW_TIME("wave_close > LEAVE");
  251. return ret;
  252. }
  253. //>
  254. // wave_is_busy
  255. //
  256. // DESCRIPTION:
  257. //
  258. // Returns a non-0 value if audio is being played.
  259. //
  260. // PARAMETERS:
  261. //
  262. // theHandler: the audio device file descriptor
  263. //
  264. // GLOBALS USED/MODIFIED:
  265. //
  266. // sun_audio_fd: used because some calls to wave_is_busy seem to
  267. // pass a NULL for theHandler for some odd reason
  268. //
  269. // RETURNS:
  270. //
  271. // A non-0 value if audio is being played
  272. //
  273. //<wave_is_busy
  274. int wave_is_busy(void* theHandler)
  275. {
  276. uint32_t time;
  277. if (total_samples_sent >= 1) {
  278. wave_get_remaining_time(total_samples_sent - 1, &time);
  279. } else {
  280. time = 0;
  281. }
  282. return time != 0;
  283. }
  284. //>
  285. // wave_terminate
  286. //
  287. // DESCRIPTION:
  288. //
  289. // Used to end our session with eSpeak.
  290. //
  291. // GLOBALS USED/MODIFIED:
  292. //
  293. // sun_audio_fd: modified - closed and set to -1
  294. //
  295. //<wave_terminate
  296. void wave_terminate()
  297. {
  298. ENTER("wave_terminate");
  299. close(sun_audio_fd);
  300. sun_audio_fd = -1;
  301. SHOW_TIME("wave_terminate > LEAVE");
  302. }
  303. //>
  304. // wave_flush
  305. //
  306. // DESCRIPTION:
  307. //
  308. // Appears to want to tell the audio subsystem to make sure it plays
  309. // the audio. In our case, the system is already doing this, so this
  310. // is basically a no-op. [[[WDW - if you do a drain, you block, so
  311. // don't do that. In addition the typical SADA notion of flush is
  312. // currently handled by wave_close. I think this is most likely just
  313. // terminology conflict between eSpeak and SADA.]]]
  314. //
  315. // PARAMETERS:
  316. //
  317. // theHandler: the audio device file descriptor
  318. //
  319. //<wave_flush
  320. void wave_flush(void* theHandler)
  321. {
  322. ENTER("wave_flush");
  323. //ioctl((int) theHandler, AUDIO_DRAIN, 0);
  324. SHOW_TIME("wave_flush > LEAVE");
  325. }
  326. //>
  327. // wave_set_callback_is_output_enabled
  328. //
  329. // DESCRIPTION:
  330. //
  331. // Sets the callback to call from wave_write before it sends data to
  332. // be played. It helps wave_write determine if the data should be
  333. // thrown away or not.
  334. //
  335. // PARAMETERS:
  336. //
  337. // cb: the callback to call from wave_write
  338. //
  339. //<wave_set_callback_is_output_enabled
  340. void wave_set_callback_is_output_enabled(t_wave_callback* cb)
  341. {
  342. my_callback_is_output_enabled = cb;
  343. }
  344. //>
  345. // wave_test_get_write_buffer
  346. //
  347. // DESCRIPTION:
  348. //
  349. // Unnecessary and is used for debug output from
  350. // speak_lib.cpp:dispatch_audio.
  351. //
  352. // RETURNS:
  353. //
  354. // NULL
  355. //
  356. //<wave_test_get_write_buffer
  357. void *wave_test_get_write_buffer()
  358. {
  359. return NULL;
  360. }
  361. //>
  362. // wave_get_read_position
  363. //
  364. // DESCRIPTION:
  365. //
  366. // Concerns the sample which is currently played by the audio layer,
  367. // where 'sample' is a small buffer of synthesized wave data,
  368. // identified so that the user callback could be called when the
  369. // 'sample' is really played. The identifier is returned by
  370. // wave_get_write_position. This method is unused.
  371. //
  372. // PARAMETERS:
  373. //
  374. // theHandler: the audio device file descriptor
  375. //
  376. // RETURNS:
  377. //
  378. // The total number of 16-bit samples played by the audio system
  379. // so far.
  380. //
  381. //<wave_get_read_position
  382. uint32_t wave_get_read_position(void* theHandler)
  383. {
  384. audio_info_t ainfo;
  385. ENTER("wave_get_read_position");
  386. ioctl((int) theHandler, AUDIO_GETINFO, &ainfo);
  387. SHOW("wave_get_read_position: %d\n", ainfo.play.samples);
  388. SHOW_TIME("wave_get_read_position > LEAVE");
  389. return ainfo.play.samples;
  390. }
  391. //>
  392. // wave_get_write_position
  393. //
  394. // DESCRIPTION:
  395. //
  396. // Returns an identifier for a new sample, where 'sample' is a small
  397. // buffer of synthesized wave data, identified so that the user
  398. // callback could be called when the 'sample' is really played. This
  399. // implementation views the audio as one long continuous stream of
  400. // 16-bit samples.
  401. //
  402. // PARAMETERS:
  403. //
  404. // theHandler: the audio device file descriptor
  405. //
  406. // GLOBALS USED/MODIFIED:
  407. //
  408. // total_samples_sent: used as the return value
  409. //
  410. // RETURNS:
  411. //
  412. // total_samples_sent, which is the index for the end of this long
  413. // continuous stream. [[[WDW: with a unit32_t managing 16-bit
  414. // samples at 22050Hz, we have about 54 hours of play time before
  415. // the index wraps back to 0. We don't handle that wrapping, so
  416. // the behavior after 54 hours of play time is undefined.]]]
  417. //
  418. //<wave_get_write_position
  419. uint32_t wave_get_write_position(void* theHandler)
  420. {
  421. ENTER("wave_get_write_position");
  422. SHOW("wave_get_write_position: %d\n", total_samples_sent);
  423. SHOW_TIME("wave_get_write_position > LEAVE");
  424. return total_samples_sent;
  425. }
  426. //>
  427. // wave_get_remaining_time
  428. //
  429. // DESCRIPTION:
  430. //
  431. // Returns the remaining time (in ms) before the sample is played.
  432. // The sample in this case is a return value from a previous call to
  433. // wave_get_write_position.
  434. //
  435. // PARAMETERS:
  436. //
  437. // sample: an index returned from wave_get_write_position representing
  438. // an index into the long continuous stream of 16-bit samples
  439. // time: a return value representing the delay in milliseconds until
  440. // sample is played. A value of 0 means the sample is either
  441. // currently being played or it has already been played.
  442. //
  443. // GLOBALS USED/MODIFIED:
  444. //
  445. // sun_audio_fd: used to determine total number of samples played by
  446. // the audio system
  447. // total_samples_skipped: used in remaining time calculation
  448. //
  449. // RETURNS:
  450. //
  451. // Time in milliseconds before the sample is played or 0 if the sample
  452. // is currently playing or has already been played.
  453. //
  454. //<wave_get_remaining_time
  455. int wave_get_remaining_time(uint32_t sample, uint32_t* time)
  456. {
  457. uint32_t a_time=0;
  458. uint32_t actual_index;
  459. audio_info_t ainfo;
  460. ENTER("wave_get_remaining_time");
  461. if (!time) {
  462. return(-1);
  463. SHOW_TIME("wave_get_remaining_time > LEAVE");
  464. }
  465. ioctl(sun_audio_fd, AUDIO_GETINFO, &ainfo);
  466. // See if this sample has already been played or is currently
  467. // playing.
  468. //
  469. actual_index = sample - total_samples_skipped;
  470. if ((sample < total_samples_skipped) ||
  471. (actual_index <= ainfo.play.samples)) {
  472. *time = 0;
  473. } else {
  474. a_time = ((actual_index - ainfo.play.samples) * 1000) / wave_samplerate;
  475. *time = (uint32_t) a_time;
  476. }
  477. SHOW("wave_get_remaining_time for %d: %d\n", sample, *time);
  478. SHOW_TIME("wave_get_remaining_time > LEAVE");
  479. return 0;
  480. }
  481. #else
  482. // notdef USE_SADA
  483. void wave_init() {}
  484. void* wave_open(const char* the_api) {return (void *)1;}
  485. size_t wave_write(void* theHandler, char* theMono16BitsWaveBuffer, size_t theSize) {return theSize;}
  486. int wave_close(void* theHandler) {return 0;}
  487. int wave_is_busy(void* theHandler) {return 0;}
  488. void wave_terminate() {}
  489. uint32_t wave_get_read_position(void* theHandler) {return 0;}
  490. uint32_t wave_get_write_position(void* theHandler) {return 0;}
  491. void wave_flush(void* theHandler) {}
  492. typedef int (t_wave_callback)(void);
  493. void wave_set_callback_is_output_enabled(t_wave_callback* cb) {}
  494. extern void* wave_test_get_write_buffer() {return NULL;}
  495. int wave_get_remaining_time(uint32_t sample, uint32_t* time)
  496. {
  497. if (!time) return(-1);
  498. *time = (uint32_t)0;
  499. return 0;
  500. }
  501. #endif // of USE_PORTAUDIO
  502. //>
  503. //<clock_gettime2, add_time_in_ms
  504. void clock_gettime2(struct timespec *ts)
  505. {
  506. struct timeval tv;
  507. if (!ts)
  508. {
  509. return;
  510. }
  511. assert (gettimeofday(&tv, NULL) != -1);
  512. ts->tv_sec = tv.tv_sec;
  513. ts->tv_nsec = tv.tv_usec*1000;
  514. }
  515. void add_time_in_ms(struct timespec *ts, int time_in_ms)
  516. {
  517. if (!ts)
  518. {
  519. return;
  520. }
  521. uint64_t t_ns = (uint64_t)ts->tv_nsec + 1000000 * (uint64_t)time_in_ms;
  522. while(t_ns >= ONE_BILLION)
  523. {
  524. SHOW("event > add_time_in_ms ns: %d sec %Lu nsec \n", ts->tv_sec, t_ns);
  525. ts->tv_sec += 1;
  526. t_ns -= ONE_BILLION;
  527. }
  528. ts->tv_nsec = (long int)t_ns;
  529. }
  530. #endif // USE_ASYNC
  531. //>