| 1 | /*************************************************************************** |
| 2 | spu.c - description |
| 3 | ------------------- |
| 4 | begin : Wed May 15 2002 |
| 5 | copyright : (C) 2002 by Pete Bernert |
| 6 | email : BlackDove@addcom.de |
| 7 | |
| 8 | Portions (C) GraÅžvydas "notaz" Ignotas, 2010-2012,2014,2015 |
| 9 | |
| 10 | ***************************************************************************/ |
| 11 | /*************************************************************************** |
| 12 | * * |
| 13 | * This program is free software; you can redistribute it and/or modify * |
| 14 | * it under the terms of the GNU General Public License as published by * |
| 15 | * the Free Software Foundation; either version 2 of the License, or * |
| 16 | * (at your option) any later version. See also the license.txt file for * |
| 17 | * additional informations. * |
| 18 | * * |
| 19 | ***************************************************************************/ |
| 20 | |
| 21 | #if !defined(_WIN32) && !defined(NO_OS) |
| 22 | #include <sys/time.h> // gettimeofday in xa.c |
| 23 | #define THREAD_ENABLED 1 |
| 24 | #endif |
| 25 | #include "stdafx.h" |
| 26 | |
| 27 | #define _IN_SPU |
| 28 | |
| 29 | #include "externals.h" |
| 30 | #include "registers.h" |
| 31 | #include "out.h" |
| 32 | #include "spu_config.h" |
| 33 | |
| 34 | #ifdef __arm__ |
| 35 | #include "arm_features.h" |
| 36 | #endif |
| 37 | |
| 38 | #ifdef HAVE_ARMV7 |
| 39 | #define ssat32_to_16(v) \ |
| 40 | asm("ssat %0,#16,%1" : "=r" (v) : "r" (v)) |
| 41 | #else |
| 42 | #define ssat32_to_16(v) do { \ |
| 43 | if (v < -32768) v = -32768; \ |
| 44 | else if (v > 32767) v = 32767; \ |
| 45 | } while (0) |
| 46 | #endif |
| 47 | |
| 48 | #define PSXCLK 33868800 /* 33.8688 MHz */ |
| 49 | |
| 50 | // intended to be ~1 frame |
| 51 | #define IRQ_NEAR_BLOCKS 32 |
| 52 | |
| 53 | /* |
| 54 | #if defined (USEMACOSX) |
| 55 | static char * libraryName = N_("Mac OS X Sound"); |
| 56 | #elif defined (USEALSA) |
| 57 | static char * libraryName = N_("ALSA Sound"); |
| 58 | #elif defined (USEOSS) |
| 59 | static char * libraryName = N_("OSS Sound"); |
| 60 | #elif defined (USESDL) |
| 61 | static char * libraryName = N_("SDL Sound"); |
| 62 | #elif defined (USEPULSEAUDIO) |
| 63 | static char * libraryName = N_("PulseAudio Sound"); |
| 64 | #else |
| 65 | static char * libraryName = N_("NULL Sound"); |
| 66 | #endif |
| 67 | |
| 68 | static char * libraryInfo = N_("P.E.Op.S. Sound Driver V1.7\nCoded by Pete Bernert and the P.E.Op.S. team\n"); |
| 69 | */ |
| 70 | |
| 71 | // globals |
| 72 | |
| 73 | SPUInfo spu; |
| 74 | SPUConfig spu_config; |
| 75 | |
| 76 | static int iFMod[NSSIZE]; |
| 77 | static int RVB[NSSIZE * 2]; |
| 78 | int ChanBuf[NSSIZE]; |
| 79 | |
| 80 | #define CDDA_BUFFER_SIZE (16384 * sizeof(uint32_t)) // must be power of 2 |
| 81 | |
| 82 | //////////////////////////////////////////////////////////////////////// |
| 83 | // CODE AREA |
| 84 | //////////////////////////////////////////////////////////////////////// |
| 85 | |
| 86 | // dirty inline func includes |
| 87 | |
| 88 | #include "reverb.c" |
| 89 | #include "adsr.c" |
| 90 | |
| 91 | //////////////////////////////////////////////////////////////////////// |
| 92 | // helpers for simple interpolation |
| 93 | |
| 94 | // |
| 95 | // easy interpolation on upsampling, no special filter, just "Pete's common sense" tm |
| 96 | // |
| 97 | // instead of having n equal sample values in a row like: |
| 98 | // ____ |
| 99 | // |____ |
| 100 | // |
| 101 | // we compare the current delta change with the next delta change. |
| 102 | // |
| 103 | // if curr_delta is positive, |
| 104 | // |
| 105 | // - and next delta is smaller (or changing direction): |
| 106 | // \. |
| 107 | // -__ |
| 108 | // |
| 109 | // - and next delta significant (at least twice) bigger: |
| 110 | // --_ |
| 111 | // \. |
| 112 | // |
| 113 | // - and next delta is nearly same: |
| 114 | // \. |
| 115 | // \. |
| 116 | // |
| 117 | // |
| 118 | // if curr_delta is negative, |
| 119 | // |
| 120 | // - and next delta is smaller (or changing direction): |
| 121 | // _-- |
| 122 | // / |
| 123 | // |
| 124 | // - and next delta significant (at least twice) bigger: |
| 125 | // / |
| 126 | // __- |
| 127 | // |
| 128 | // - and next delta is nearly same: |
| 129 | // / |
| 130 | // / |
| 131 | // |
| 132 | |
| 133 | static void InterpolateUp(int *SB, int sinc) |
| 134 | { |
| 135 | if(SB[32]==1) // flag == 1? calc step and set flag... and don't change the value in this pass |
| 136 | { |
| 137 | const int id1=SB[30]-SB[29]; // curr delta to next val |
| 138 | const int id2=SB[31]-SB[30]; // and next delta to next-next val :) |
| 139 | |
| 140 | SB[32]=0; |
| 141 | |
| 142 | if(id1>0) // curr delta positive |
| 143 | { |
| 144 | if(id2<id1) |
| 145 | {SB[28]=id1;SB[32]=2;} |
| 146 | else |
| 147 | if(id2<(id1<<1)) |
| 148 | SB[28]=(id1*sinc)>>16; |
| 149 | else |
| 150 | SB[28]=(id1*sinc)>>17; |
| 151 | } |
| 152 | else // curr delta negative |
| 153 | { |
| 154 | if(id2>id1) |
| 155 | {SB[28]=id1;SB[32]=2;} |
| 156 | else |
| 157 | if(id2>(id1<<1)) |
| 158 | SB[28]=(id1*sinc)>>16; |
| 159 | else |
| 160 | SB[28]=(id1*sinc)>>17; |
| 161 | } |
| 162 | } |
| 163 | else |
| 164 | if(SB[32]==2) // flag 1: calc step and set flag... and don't change the value in this pass |
| 165 | { |
| 166 | SB[32]=0; |
| 167 | |
| 168 | SB[28]=(SB[28]*sinc)>>17; |
| 169 | //if(sinc<=0x8000) |
| 170 | // SB[29]=SB[30]-(SB[28]*((0x10000/sinc)-1)); |
| 171 | //else |
| 172 | SB[29]+=SB[28]; |
| 173 | } |
| 174 | else // no flags? add bigger val (if possible), calc smaller step, set flag1 |
| 175 | SB[29]+=SB[28]; |
| 176 | } |
| 177 | |
| 178 | // |
| 179 | // even easier interpolation on downsampling, also no special filter, again just "Pete's common sense" tm |
| 180 | // |
| 181 | |
| 182 | static void InterpolateDown(int *SB, int sinc) |
| 183 | { |
| 184 | if(sinc>=0x20000L) // we would skip at least one val? |
| 185 | { |
| 186 | SB[29]+=(SB[30]-SB[29])/2; // add easy weight |
| 187 | if(sinc>=0x30000L) // we would skip even more vals? |
| 188 | SB[29]+=(SB[31]-SB[30])/2; // add additional next weight |
| 189 | } |
| 190 | } |
| 191 | |
| 192 | //////////////////////////////////////////////////////////////////////// |
| 193 | // helpers for gauss interpolation |
| 194 | |
| 195 | #define gval0 (((short*)(&SB[29]))[gpos&3]) |
| 196 | #define gval(x) ((int)((short*)(&SB[29]))[(gpos+x)&3]) |
| 197 | |
| 198 | #include "gauss_i.h" |
| 199 | |
| 200 | //////////////////////////////////////////////////////////////////////// |
| 201 | |
| 202 | #include "xa.c" |
| 203 | |
| 204 | static void do_irq(void) |
| 205 | { |
| 206 | //if(!(spu.spuStat & STAT_IRQ)) |
| 207 | { |
| 208 | spu.spuStat |= STAT_IRQ; // asserted status? |
| 209 | if(spu.irqCallback) spu.irqCallback(); |
| 210 | } |
| 211 | } |
| 212 | |
| 213 | static int check_irq(int ch, unsigned char *pos) |
| 214 | { |
| 215 | if((spu.spuCtrl & CTRL_IRQ) && pos == spu.pSpuIrq) |
| 216 | { |
| 217 | //printf("ch%d irq %04x\n", ch, pos - spu.spuMemC); |
| 218 | do_irq(); |
| 219 | return 1; |
| 220 | } |
| 221 | return 0; |
| 222 | } |
| 223 | |
| 224 | //////////////////////////////////////////////////////////////////////// |
| 225 | // START SOUND... called by main thread to setup a new sound on a channel |
| 226 | //////////////////////////////////////////////////////////////////////// |
| 227 | |
| 228 | static void StartSoundSB(int *SB) |
| 229 | { |
| 230 | SB[26]=0; // init mixing vars |
| 231 | SB[27]=0; |
| 232 | |
| 233 | SB[28]=0; |
| 234 | SB[29]=0; // init our interpolation helpers |
| 235 | SB[30]=0; |
| 236 | SB[31]=0; |
| 237 | } |
| 238 | |
| 239 | static void StartSoundMain(int ch) |
| 240 | { |
| 241 | SPUCHAN *s_chan = &spu.s_chan[ch]; |
| 242 | |
| 243 | StartADSR(ch); |
| 244 | StartREVERB(ch); |
| 245 | |
| 246 | s_chan->prevflags=2; |
| 247 | s_chan->iSBPos=27; |
| 248 | s_chan->spos=0; |
| 249 | |
| 250 | spu.dwNewChannel&=~(1<<ch); // clear new channel bit |
| 251 | spu.dwChannelOn|=1<<ch; |
| 252 | spu.dwChannelDead&=~(1<<ch); |
| 253 | } |
| 254 | |
| 255 | static void StartSound(int ch) |
| 256 | { |
| 257 | StartSoundMain(ch); |
| 258 | StartSoundSB(spu.SB + ch * SB_SIZE); |
| 259 | } |
| 260 | |
| 261 | //////////////////////////////////////////////////////////////////////// |
| 262 | // ALL KIND OF HELPERS |
| 263 | //////////////////////////////////////////////////////////////////////// |
| 264 | |
| 265 | INLINE int FModChangeFrequency(int *SB, int pitch, int ns) |
| 266 | { |
| 267 | unsigned int NP=pitch; |
| 268 | int sinc; |
| 269 | |
| 270 | NP=((32768L+iFMod[ns])*NP)>>15; |
| 271 | |
| 272 | if(NP>0x3fff) NP=0x3fff; |
| 273 | if(NP<0x1) NP=0x1; |
| 274 | |
| 275 | sinc=NP<<4; // calc frequency |
| 276 | if(spu_config.iUseInterpolation==1) // freq change in simple interpolation mode |
| 277 | SB[32]=1; |
| 278 | iFMod[ns]=0; |
| 279 | |
| 280 | return sinc; |
| 281 | } |
| 282 | |
| 283 | //////////////////////////////////////////////////////////////////////// |
| 284 | |
| 285 | INLINE void StoreInterpolationVal(int *SB, int sinc, int fa, int fmod_freq) |
| 286 | { |
| 287 | if(fmod_freq) // fmod freq channel |
| 288 | SB[29]=fa; |
| 289 | else |
| 290 | { |
| 291 | ssat32_to_16(fa); |
| 292 | |
| 293 | if(spu_config.iUseInterpolation>=2) // gauss/cubic interpolation |
| 294 | { |
| 295 | int gpos = SB[28]; |
| 296 | gval0 = fa; |
| 297 | gpos = (gpos+1) & 3; |
| 298 | SB[28] = gpos; |
| 299 | } |
| 300 | else |
| 301 | if(spu_config.iUseInterpolation==1) // simple interpolation |
| 302 | { |
| 303 | SB[28] = 0; |
| 304 | SB[29] = SB[30]; // -> helpers for simple linear interpolation: delay real val for two slots, and calc the two deltas, for a 'look at the future behaviour' |
| 305 | SB[30] = SB[31]; |
| 306 | SB[31] = fa; |
| 307 | SB[32] = 1; // -> flag: calc new interolation |
| 308 | } |
| 309 | else SB[29]=fa; // no interpolation |
| 310 | } |
| 311 | } |
| 312 | |
| 313 | //////////////////////////////////////////////////////////////////////// |
| 314 | |
| 315 | INLINE int iGetInterpolationVal(int *SB, int sinc, int spos, int fmod_freq) |
| 316 | { |
| 317 | int fa; |
| 318 | |
| 319 | if(fmod_freq) return SB[29]; |
| 320 | |
| 321 | switch(spu_config.iUseInterpolation) |
| 322 | { |
| 323 | //--------------------------------------------------// |
| 324 | case 3: // cubic interpolation |
| 325 | { |
| 326 | long xd;int gpos; |
| 327 | xd = (spos >> 1)+1; |
| 328 | gpos = SB[28]; |
| 329 | |
| 330 | fa = gval(3) - 3*gval(2) + 3*gval(1) - gval0; |
| 331 | fa *= (xd - (2<<15)) / 6; |
| 332 | fa >>= 15; |
| 333 | fa += gval(2) - gval(1) - gval(1) + gval0; |
| 334 | fa *= (xd - (1<<15)) >> 1; |
| 335 | fa >>= 15; |
| 336 | fa += gval(1) - gval0; |
| 337 | fa *= xd; |
| 338 | fa >>= 15; |
| 339 | fa = fa + gval0; |
| 340 | |
| 341 | } break; |
| 342 | //--------------------------------------------------// |
| 343 | case 2: // gauss interpolation |
| 344 | { |
| 345 | int vl, vr;int gpos; |
| 346 | vl = (spos >> 6) & ~3; |
| 347 | gpos = SB[28]; |
| 348 | vr=(gauss[vl]*(int)gval0) >> 15; |
| 349 | vr+=(gauss[vl+1]*gval(1)) >> 15; |
| 350 | vr+=(gauss[vl+2]*gval(2)) >> 15; |
| 351 | vr+=(gauss[vl+3]*gval(3)) >> 15; |
| 352 | fa = vr; |
| 353 | } break; |
| 354 | //--------------------------------------------------// |
| 355 | case 1: // simple interpolation |
| 356 | { |
| 357 | if(sinc<0x10000L) // -> upsampling? |
| 358 | InterpolateUp(SB, sinc); // --> interpolate up |
| 359 | else InterpolateDown(SB, sinc); // --> else down |
| 360 | fa=SB[29]; |
| 361 | } break; |
| 362 | //--------------------------------------------------// |
| 363 | default: // no interpolation |
| 364 | { |
| 365 | fa=SB[29]; |
| 366 | } break; |
| 367 | //--------------------------------------------------// |
| 368 | } |
| 369 | |
| 370 | return fa; |
| 371 | } |
| 372 | |
| 373 | static void decode_block_data(int *dest, const unsigned char *src, int predict_nr, int shift_factor) |
| 374 | { |
| 375 | static const int f[16][2] = { |
| 376 | { 0, 0 }, |
| 377 | { 60, 0 }, |
| 378 | { 115, -52 }, |
| 379 | { 98, -55 }, |
| 380 | { 122, -60 } |
| 381 | }; |
| 382 | int nSample; |
| 383 | int fa, s_1, s_2, d, s; |
| 384 | |
| 385 | s_1 = dest[27]; |
| 386 | s_2 = dest[26]; |
| 387 | |
| 388 | for (nSample = 0; nSample < 28; src++) |
| 389 | { |
| 390 | d = (int)*src; |
| 391 | s = (int)(signed short)((d & 0x0f) << 12); |
| 392 | |
| 393 | fa = s >> shift_factor; |
| 394 | fa += ((s_1 * f[predict_nr][0])>>6) + ((s_2 * f[predict_nr][1])>>6); |
| 395 | s_2=s_1;s_1=fa; |
| 396 | |
| 397 | dest[nSample++] = fa; |
| 398 | |
| 399 | s = (int)(signed short)((d & 0xf0) << 8); |
| 400 | fa = s >> shift_factor; |
| 401 | fa += ((s_1 * f[predict_nr][0])>>6) + ((s_2 * f[predict_nr][1])>>6); |
| 402 | s_2=s_1;s_1=fa; |
| 403 | |
| 404 | dest[nSample++] = fa; |
| 405 | } |
| 406 | } |
| 407 | |
| 408 | static int decode_block(void *unused, int ch, int *SB) |
| 409 | { |
| 410 | SPUCHAN *s_chan = &spu.s_chan[ch]; |
| 411 | unsigned char *start; |
| 412 | int predict_nr, shift_factor, flags; |
| 413 | int ret = 0; |
| 414 | |
| 415 | start = s_chan->pCurr; // set up the current pos |
| 416 | if (start == spu.spuMemC) // ? |
| 417 | ret = 1; |
| 418 | |
| 419 | if (s_chan->prevflags & 1) // 1: stop/loop |
| 420 | { |
| 421 | if (!(s_chan->prevflags & 2)) |
| 422 | ret = 1; |
| 423 | |
| 424 | start = s_chan->pLoop; |
| 425 | } |
| 426 | else |
| 427 | check_irq(ch, start); // hack, see check_irq below.. |
| 428 | |
| 429 | predict_nr = start[0]; |
| 430 | shift_factor = predict_nr & 0xf; |
| 431 | predict_nr >>= 4; |
| 432 | |
| 433 | decode_block_data(SB, start + 2, predict_nr, shift_factor); |
| 434 | |
| 435 | flags = start[1]; |
| 436 | if (flags & 4) |
| 437 | s_chan->pLoop = start; // loop adress |
| 438 | |
| 439 | start += 16; |
| 440 | |
| 441 | if (flags & 1) { // 1: stop/loop |
| 442 | start = s_chan->pLoop; |
| 443 | check_irq(ch, start); // hack.. :( |
| 444 | } |
| 445 | |
| 446 | if (start - spu.spuMemC >= 0x80000) |
| 447 | start = spu.spuMemC; |
| 448 | |
| 449 | s_chan->pCurr = start; // store values for next cycle |
| 450 | s_chan->prevflags = flags; |
| 451 | |
| 452 | return ret; |
| 453 | } |
| 454 | |
| 455 | // do block, but ignore sample data |
| 456 | static int skip_block(int ch) |
| 457 | { |
| 458 | SPUCHAN *s_chan = &spu.s_chan[ch]; |
| 459 | unsigned char *start = s_chan->pCurr; |
| 460 | int flags; |
| 461 | int ret = 0; |
| 462 | |
| 463 | if (s_chan->prevflags & 1) { |
| 464 | if (!(s_chan->prevflags & 2)) |
| 465 | ret = 1; |
| 466 | |
| 467 | start = s_chan->pLoop; |
| 468 | } |
| 469 | else |
| 470 | check_irq(ch, start); |
| 471 | |
| 472 | flags = start[1]; |
| 473 | if (flags & 4) |
| 474 | s_chan->pLoop = start; |
| 475 | |
| 476 | start += 16; |
| 477 | |
| 478 | if (flags & 1) { |
| 479 | start = s_chan->pLoop; |
| 480 | check_irq(ch, start); |
| 481 | } |
| 482 | |
| 483 | s_chan->pCurr = start; |
| 484 | s_chan->prevflags = flags; |
| 485 | |
| 486 | return ret; |
| 487 | } |
| 488 | |
| 489 | // if irq is going to trigger sooner than in upd_samples, set upd_samples |
| 490 | static void scan_for_irq(int ch, unsigned int *upd_samples) |
| 491 | { |
| 492 | SPUCHAN *s_chan = &spu.s_chan[ch]; |
| 493 | int pos, sinc, sinc_inv, end; |
| 494 | unsigned char *block; |
| 495 | int flags; |
| 496 | |
| 497 | block = s_chan->pCurr; |
| 498 | pos = s_chan->spos; |
| 499 | sinc = s_chan->sinc; |
| 500 | end = pos + *upd_samples * sinc; |
| 501 | |
| 502 | pos += (28 - s_chan->iSBPos) << 16; |
| 503 | while (pos < end) |
| 504 | { |
| 505 | if (block == spu.pSpuIrq) |
| 506 | break; |
| 507 | flags = block[1]; |
| 508 | block += 16; |
| 509 | if (flags & 1) { // 1: stop/loop |
| 510 | block = s_chan->pLoop; |
| 511 | if (block == spu.pSpuIrq) // hack.. (see decode_block) |
| 512 | break; |
| 513 | } |
| 514 | pos += 28 << 16; |
| 515 | } |
| 516 | |
| 517 | if (pos < end) |
| 518 | { |
| 519 | sinc_inv = s_chan->sinc_inv; |
| 520 | if (sinc_inv == 0) |
| 521 | sinc_inv = s_chan->sinc_inv = (0x80000000u / (uint32_t)sinc) << 1; |
| 522 | |
| 523 | pos -= s_chan->spos; |
| 524 | *upd_samples = (((uint64_t)pos * sinc_inv) >> 32) + 1; |
| 525 | //xprintf("ch%02d: irq sched: %3d %03d\n", |
| 526 | // ch, *upd_samples, *upd_samples * 60 * 263 / 44100); |
| 527 | } |
| 528 | } |
| 529 | |
| 530 | #define make_do_samples(name, fmod_code, interp_start, interp1_code, interp2_code, interp_end) \ |
| 531 | static noinline int do_samples_##name( \ |
| 532 | int (*decode_f)(void *context, int ch, int *SB), void *ctx, \ |
| 533 | int ch, int ns_to, int *SB, int sinc, int *spos, int *sbpos) \ |
| 534 | { \ |
| 535 | int ns, d, fa; \ |
| 536 | int ret = ns_to; \ |
| 537 | interp_start; \ |
| 538 | \ |
| 539 | for (ns = 0; ns < ns_to; ns++) \ |
| 540 | { \ |
| 541 | fmod_code; \ |
| 542 | \ |
| 543 | *spos += sinc; \ |
| 544 | while (*spos >= 0x10000) \ |
| 545 | { \ |
| 546 | fa = SB[(*sbpos)++]; \ |
| 547 | if (*sbpos >= 28) \ |
| 548 | { \ |
| 549 | *sbpos = 0; \ |
| 550 | d = decode_f(ctx, ch, SB); \ |
| 551 | if (d && ns < ret) \ |
| 552 | ret = ns; \ |
| 553 | } \ |
| 554 | \ |
| 555 | interp1_code; \ |
| 556 | *spos -= 0x10000; \ |
| 557 | } \ |
| 558 | \ |
| 559 | interp2_code; \ |
| 560 | } \ |
| 561 | \ |
| 562 | interp_end; \ |
| 563 | \ |
| 564 | return ret; \ |
| 565 | } |
| 566 | |
| 567 | #define fmod_recv_check \ |
| 568 | if(spu.s_chan[ch].bFMod==1 && iFMod[ns]) \ |
| 569 | sinc = FModChangeFrequency(SB, spu.s_chan[ch].iRawPitch, ns) |
| 570 | |
| 571 | make_do_samples(default, fmod_recv_check, , |
| 572 | StoreInterpolationVal(SB, sinc, fa, spu.s_chan[ch].bFMod==2), |
| 573 | ChanBuf[ns] = iGetInterpolationVal(SB, sinc, *spos, spu.s_chan[ch].bFMod==2), ) |
| 574 | make_do_samples(noint, , fa = SB[29], , ChanBuf[ns] = fa, SB[29] = fa) |
| 575 | |
| 576 | #define simple_interp_store \ |
| 577 | SB[28] = 0; \ |
| 578 | SB[29] = SB[30]; \ |
| 579 | SB[30] = SB[31]; \ |
| 580 | SB[31] = fa; \ |
| 581 | SB[32] = 1 |
| 582 | |
| 583 | #define simple_interp_get \ |
| 584 | if(sinc<0x10000) /* -> upsampling? */ \ |
| 585 | InterpolateUp(SB, sinc); /* --> interpolate up */ \ |
| 586 | else InterpolateDown(SB, sinc); /* --> else down */ \ |
| 587 | ChanBuf[ns] = SB[29] |
| 588 | |
| 589 | make_do_samples(simple, , , |
| 590 | simple_interp_store, simple_interp_get, ) |
| 591 | |
| 592 | static int do_samples_skip(int ch, int ns_to) |
| 593 | { |
| 594 | SPUCHAN *s_chan = &spu.s_chan[ch]; |
| 595 | int spos = s_chan->spos; |
| 596 | int sinc = s_chan->sinc; |
| 597 | int ret = ns_to, ns, d; |
| 598 | |
| 599 | spos += s_chan->iSBPos << 16; |
| 600 | |
| 601 | for (ns = 0; ns < ns_to; ns++) |
| 602 | { |
| 603 | spos += sinc; |
| 604 | while (spos >= 28*0x10000) |
| 605 | { |
| 606 | d = skip_block(ch); |
| 607 | if (d && ns < ret) |
| 608 | ret = ns; |
| 609 | spos -= 28*0x10000; |
| 610 | } |
| 611 | } |
| 612 | |
| 613 | s_chan->iSBPos = spos >> 16; |
| 614 | s_chan->spos = spos & 0xffff; |
| 615 | |
| 616 | return ret; |
| 617 | } |
| 618 | |
| 619 | static void do_lsfr_samples(int ns_to, int ctrl, |
| 620 | unsigned int *dwNoiseCount, unsigned int *dwNoiseVal) |
| 621 | { |
| 622 | unsigned int counter = *dwNoiseCount; |
| 623 | unsigned int val = *dwNoiseVal; |
| 624 | unsigned int level, shift, bit; |
| 625 | int ns; |
| 626 | |
| 627 | // modified from DrHell/shalma, no fraction |
| 628 | level = (ctrl >> 10) & 0x0f; |
| 629 | level = 0x8000 >> level; |
| 630 | |
| 631 | for (ns = 0; ns < ns_to; ns++) |
| 632 | { |
| 633 | counter += 2; |
| 634 | if (counter >= level) |
| 635 | { |
| 636 | counter -= level; |
| 637 | shift = (val >> 10) & 0x1f; |
| 638 | bit = (0x69696969 >> shift) & 1; |
| 639 | bit ^= (val >> 15) & 1; |
| 640 | val = (val << 1) | bit; |
| 641 | } |
| 642 | |
| 643 | ChanBuf[ns] = (signed short)val; |
| 644 | } |
| 645 | |
| 646 | *dwNoiseCount = counter; |
| 647 | *dwNoiseVal = val; |
| 648 | } |
| 649 | |
| 650 | static int do_samples_noise(int ch, int ns_to) |
| 651 | { |
| 652 | int ret; |
| 653 | |
| 654 | ret = do_samples_skip(ch, ns_to); |
| 655 | |
| 656 | do_lsfr_samples(ns_to, spu.spuCtrl, &spu.dwNoiseCount, &spu.dwNoiseVal); |
| 657 | |
| 658 | return ret; |
| 659 | } |
| 660 | |
| 661 | #ifdef HAVE_ARMV5 |
| 662 | // asm code; lv and rv must be 0-3fff |
| 663 | extern void mix_chan(int *SSumLR, int count, int lv, int rv); |
| 664 | extern void mix_chan_rvb(int *SSumLR, int count, int lv, int rv, int *rvb); |
| 665 | #else |
| 666 | static void mix_chan(int *SSumLR, int count, int lv, int rv) |
| 667 | { |
| 668 | const int *src = ChanBuf; |
| 669 | int l, r; |
| 670 | |
| 671 | while (count--) |
| 672 | { |
| 673 | int sval = *src++; |
| 674 | |
| 675 | l = (sval * lv) >> 14; |
| 676 | r = (sval * rv) >> 14; |
| 677 | *SSumLR++ += l; |
| 678 | *SSumLR++ += r; |
| 679 | } |
| 680 | } |
| 681 | |
| 682 | static void mix_chan_rvb(int *SSumLR, int count, int lv, int rv, int *rvb) |
| 683 | { |
| 684 | const int *src = ChanBuf; |
| 685 | int *dst = SSumLR; |
| 686 | int *drvb = rvb; |
| 687 | int l, r; |
| 688 | |
| 689 | while (count--) |
| 690 | { |
| 691 | int sval = *src++; |
| 692 | |
| 693 | l = (sval * lv) >> 14; |
| 694 | r = (sval * rv) >> 14; |
| 695 | *dst++ += l; |
| 696 | *dst++ += r; |
| 697 | *drvb++ += l; |
| 698 | *drvb++ += r; |
| 699 | } |
| 700 | } |
| 701 | #endif |
| 702 | |
| 703 | // 0x0800-0x0bff Voice 1 |
| 704 | // 0x0c00-0x0fff Voice 3 |
| 705 | static noinline void do_decode_bufs(unsigned short *mem, int which, |
| 706 | int count, int decode_pos) |
| 707 | { |
| 708 | unsigned short *dst = &mem[0x800/2 + which*0x400/2]; |
| 709 | const int *src = ChanBuf; |
| 710 | int cursor = decode_pos; |
| 711 | |
| 712 | while (count-- > 0) |
| 713 | { |
| 714 | cursor &= 0x1ff; |
| 715 | dst[cursor] = *src++; |
| 716 | cursor++; |
| 717 | } |
| 718 | |
| 719 | // decode_pos is updated and irqs are checked later, after voice loop |
| 720 | } |
| 721 | |
| 722 | static void do_silent_chans(int ns_to, int silentch) |
| 723 | { |
| 724 | unsigned int mask; |
| 725 | SPUCHAN *s_chan; |
| 726 | int ch; |
| 727 | |
| 728 | mask = silentch & 0xffffff; |
| 729 | for (ch = 0; mask != 0; ch++, mask >>= 1) |
| 730 | { |
| 731 | if (!(mask & 1)) continue; |
| 732 | if (spu.dwChannelDead & (1<<ch)) continue; |
| 733 | |
| 734 | s_chan = &spu.s_chan[ch]; |
| 735 | if (s_chan->pCurr > spu.pSpuIrq && s_chan->pLoop > spu.pSpuIrq) |
| 736 | continue; |
| 737 | |
| 738 | s_chan->spos += s_chan->iSBPos << 16; |
| 739 | s_chan->iSBPos = 0; |
| 740 | |
| 741 | s_chan->spos += s_chan->sinc * ns_to; |
| 742 | while (s_chan->spos >= 28 * 0x10000) |
| 743 | { |
| 744 | unsigned char *start = s_chan->pCurr; |
| 745 | |
| 746 | skip_block(ch); |
| 747 | if (start == s_chan->pCurr || start - spu.spuMemC < 0x1000) |
| 748 | { |
| 749 | // looping on self or stopped(?) |
| 750 | spu.dwChannelDead |= 1<<ch; |
| 751 | s_chan->spos = 0; |
| 752 | break; |
| 753 | } |
| 754 | |
| 755 | s_chan->spos -= 28 * 0x10000; |
| 756 | } |
| 757 | } |
| 758 | } |
| 759 | |
| 760 | static void do_channels(int ns_to) |
| 761 | { |
| 762 | unsigned int mask; |
| 763 | int do_rvb, ch, d; |
| 764 | SPUCHAN *s_chan; |
| 765 | int *SB, sinc; |
| 766 | |
| 767 | do_rvb = spu.rvb->StartAddr && spu_config.iUseReverb; |
| 768 | if (do_rvb) |
| 769 | memset(RVB, 0, ns_to * sizeof(RVB[0]) * 2); |
| 770 | |
| 771 | mask = spu.dwNewChannel & 0xffffff; |
| 772 | for (ch = 0; mask != 0; ch++, mask >>= 1) { |
| 773 | if (mask & 1) |
| 774 | StartSound(ch); |
| 775 | } |
| 776 | |
| 777 | mask = spu.dwChannelOn & 0xffffff; |
| 778 | for (ch = 0; mask != 0; ch++, mask >>= 1) // loop em all... |
| 779 | { |
| 780 | if (!(mask & 1)) continue; // channel not playing? next |
| 781 | |
| 782 | s_chan = &spu.s_chan[ch]; |
| 783 | SB = spu.SB + ch * SB_SIZE; |
| 784 | sinc = s_chan->sinc; |
| 785 | |
| 786 | if (s_chan->bNoise) |
| 787 | d = do_samples_noise(ch, ns_to); |
| 788 | else if (s_chan->bFMod == 2 |
| 789 | || (s_chan->bFMod == 0 && spu_config.iUseInterpolation == 0)) |
| 790 | d = do_samples_noint(decode_block, NULL, ch, ns_to, |
| 791 | SB, sinc, &s_chan->spos, &s_chan->iSBPos); |
| 792 | else if (s_chan->bFMod == 0 && spu_config.iUseInterpolation == 1) |
| 793 | d = do_samples_simple(decode_block, NULL, ch, ns_to, |
| 794 | SB, sinc, &s_chan->spos, &s_chan->iSBPos); |
| 795 | else |
| 796 | d = do_samples_default(decode_block, NULL, ch, ns_to, |
| 797 | SB, sinc, &s_chan->spos, &s_chan->iSBPos); |
| 798 | |
| 799 | d = MixADSR(&s_chan->ADSRX, d); |
| 800 | if (d < ns_to) { |
| 801 | spu.dwChannelOn &= ~(1 << ch); |
| 802 | s_chan->ADSRX.EnvelopeVol = 0; |
| 803 | memset(&ChanBuf[d], 0, (ns_to - d) * sizeof(ChanBuf[0])); |
| 804 | } |
| 805 | |
| 806 | if (ch == 1 || ch == 3) |
| 807 | { |
| 808 | do_decode_bufs(spu.spuMem, ch/2, ns_to, spu.decode_pos); |
| 809 | spu.decode_dirty_ch |= 1 << ch; |
| 810 | } |
| 811 | |
| 812 | if (s_chan->bFMod == 2) // fmod freq channel |
| 813 | memcpy(iFMod, &ChanBuf, ns_to * sizeof(iFMod[0])); |
| 814 | if (s_chan->bRVBActive && do_rvb) |
| 815 | mix_chan_rvb(spu.SSumLR, ns_to, s_chan->iLeftVolume, s_chan->iRightVolume, RVB); |
| 816 | else |
| 817 | mix_chan(spu.SSumLR, ns_to, s_chan->iLeftVolume, s_chan->iRightVolume); |
| 818 | } |
| 819 | |
| 820 | if (spu.rvb->StartAddr) { |
| 821 | if (do_rvb) |
| 822 | REVERBDo(spu.SSumLR, RVB, ns_to, spu.rvb->CurrAddr); |
| 823 | |
| 824 | spu.rvb->CurrAddr += ns_to / 2; |
| 825 | while (spu.rvb->CurrAddr >= 0x40000) |
| 826 | spu.rvb->CurrAddr -= 0x40000 - spu.rvb->StartAddr; |
| 827 | } |
| 828 | } |
| 829 | |
| 830 | static void do_samples_finish(int *SSumLR, int ns_to, |
| 831 | int silentch, int decode_pos); |
| 832 | |
| 833 | // optional worker thread handling |
| 834 | |
| 835 | #if defined(THREAD_ENABLED) || defined(WANT_THREAD_CODE) |
| 836 | |
| 837 | // worker thread state |
| 838 | static struct spu_worker { |
| 839 | union { |
| 840 | struct { |
| 841 | unsigned int exit_thread; |
| 842 | unsigned int i_ready; |
| 843 | unsigned int i_reaped; |
| 844 | unsigned int last_boot_cnt; // dsp |
| 845 | unsigned int ram_dirty; |
| 846 | }; |
| 847 | // aligning for C64X_DSP |
| 848 | unsigned int _pad0[128/4]; |
| 849 | }; |
| 850 | union { |
| 851 | struct { |
| 852 | unsigned int i_done; |
| 853 | unsigned int active; // dsp |
| 854 | unsigned int boot_cnt; |
| 855 | }; |
| 856 | unsigned int _pad1[128/4]; |
| 857 | }; |
| 858 | struct work_item { |
| 859 | int ns_to; |
| 860 | int ctrl; |
| 861 | int decode_pos; |
| 862 | int rvb_addr; |
| 863 | unsigned int channels_new; |
| 864 | unsigned int channels_on; |
| 865 | unsigned int channels_silent; |
| 866 | struct { |
| 867 | int spos; |
| 868 | int sbpos; |
| 869 | int sinc; |
| 870 | int start; |
| 871 | int loop; |
| 872 | int ns_to; |
| 873 | short vol_l; |
| 874 | short vol_r; |
| 875 | ADSRInfoEx adsr; |
| 876 | // might also want to add fmod flags.. |
| 877 | } ch[24]; |
| 878 | int SSumLR[NSSIZE * 2]; |
| 879 | } i[4]; |
| 880 | } *worker; |
| 881 | |
| 882 | #define WORK_MAXCNT (sizeof(worker->i) / sizeof(worker->i[0])) |
| 883 | #define WORK_I_MASK (WORK_MAXCNT - 1) |
| 884 | |
| 885 | static void thread_work_start(void); |
| 886 | static void thread_work_wait_sync(struct work_item *work, int force); |
| 887 | static void thread_sync_caches(void); |
| 888 | static int thread_get_i_done(void); |
| 889 | |
| 890 | static int decode_block_work(void *context, int ch, int *SB) |
| 891 | { |
| 892 | const unsigned char *ram = spu.spuMemC; |
| 893 | int predict_nr, shift_factor, flags; |
| 894 | struct work_item *work = context; |
| 895 | int start = work->ch[ch].start; |
| 896 | int loop = work->ch[ch].loop; |
| 897 | |
| 898 | predict_nr = ram[start]; |
| 899 | shift_factor = predict_nr & 0xf; |
| 900 | predict_nr >>= 4; |
| 901 | |
| 902 | decode_block_data(SB, ram + start + 2, predict_nr, shift_factor); |
| 903 | |
| 904 | flags = ram[start + 1]; |
| 905 | if (flags & 4) |
| 906 | loop = start; // loop adress |
| 907 | |
| 908 | start += 16; |
| 909 | |
| 910 | if (flags & 1) // 1: stop/loop |
| 911 | start = loop; |
| 912 | |
| 913 | work->ch[ch].start = start & 0x7ffff; |
| 914 | work->ch[ch].loop = loop; |
| 915 | |
| 916 | return 0; |
| 917 | } |
| 918 | |
| 919 | static void queue_channel_work(int ns_to, unsigned int silentch) |
| 920 | { |
| 921 | struct work_item *work; |
| 922 | SPUCHAN *s_chan; |
| 923 | unsigned int mask; |
| 924 | int ch, d; |
| 925 | |
| 926 | work = &worker->i[worker->i_ready & WORK_I_MASK]; |
| 927 | work->ns_to = ns_to; |
| 928 | work->ctrl = spu.spuCtrl; |
| 929 | work->decode_pos = spu.decode_pos; |
| 930 | work->channels_silent = silentch; |
| 931 | |
| 932 | mask = work->channels_new = spu.dwNewChannel & 0xffffff; |
| 933 | for (ch = 0; mask != 0; ch++, mask >>= 1) { |
| 934 | if (mask & 1) |
| 935 | StartSoundMain(ch); |
| 936 | } |
| 937 | |
| 938 | mask = work->channels_on = spu.dwChannelOn & 0xffffff; |
| 939 | spu.decode_dirty_ch |= mask & 0x0a; |
| 940 | |
| 941 | for (ch = 0; mask != 0; ch++, mask >>= 1) |
| 942 | { |
| 943 | if (!(mask & 1)) continue; |
| 944 | |
| 945 | s_chan = &spu.s_chan[ch]; |
| 946 | work->ch[ch].spos = s_chan->spos; |
| 947 | work->ch[ch].sbpos = s_chan->iSBPos; |
| 948 | work->ch[ch].sinc = s_chan->sinc; |
| 949 | work->ch[ch].adsr = s_chan->ADSRX; |
| 950 | work->ch[ch].vol_l = s_chan->iLeftVolume; |
| 951 | work->ch[ch].vol_r = s_chan->iRightVolume; |
| 952 | work->ch[ch].start = s_chan->pCurr - spu.spuMemC; |
| 953 | work->ch[ch].loop = s_chan->pLoop - spu.spuMemC; |
| 954 | if (s_chan->prevflags & 1) |
| 955 | work->ch[ch].start = work->ch[ch].loop; |
| 956 | |
| 957 | d = do_samples_skip(ch, ns_to); |
| 958 | work->ch[ch].ns_to = d; |
| 959 | |
| 960 | // note: d is not accurate on skip |
| 961 | d = SkipADSR(&s_chan->ADSRX, d); |
| 962 | if (d < ns_to) { |
| 963 | spu.dwChannelOn &= ~(1 << ch); |
| 964 | s_chan->ADSRX.EnvelopeVol = 0; |
| 965 | } |
| 966 | } |
| 967 | |
| 968 | work->rvb_addr = 0; |
| 969 | if (spu.rvb->StartAddr) { |
| 970 | if (spu_config.iUseReverb) |
| 971 | work->rvb_addr = spu.rvb->CurrAddr; |
| 972 | |
| 973 | spu.rvb->CurrAddr += ns_to / 2; |
| 974 | while (spu.rvb->CurrAddr >= 0x40000) |
| 975 | spu.rvb->CurrAddr -= 0x40000 - spu.rvb->StartAddr; |
| 976 | } |
| 977 | |
| 978 | worker->i_ready++; |
| 979 | thread_work_start(); |
| 980 | } |
| 981 | |
| 982 | static void do_channel_work(struct work_item *work) |
| 983 | { |
| 984 | unsigned int mask; |
| 985 | unsigned int decode_dirty_ch = 0; |
| 986 | const SPUCHAN *s_chan; |
| 987 | int *SB, sinc, spos, sbpos; |
| 988 | int d, ch, ns_to; |
| 989 | |
| 990 | ns_to = work->ns_to; |
| 991 | |
| 992 | if (work->rvb_addr) |
| 993 | memset(RVB, 0, ns_to * sizeof(RVB[0]) * 2); |
| 994 | |
| 995 | mask = work->channels_new; |
| 996 | for (ch = 0; mask != 0; ch++, mask >>= 1) { |
| 997 | if (mask & 1) |
| 998 | StartSoundSB(spu.SB + ch * SB_SIZE); |
| 999 | } |
| 1000 | |
| 1001 | mask = work->channels_on; |
| 1002 | for (ch = 0; mask != 0; ch++, mask >>= 1) |
| 1003 | { |
| 1004 | if (!(mask & 1)) continue; |
| 1005 | |
| 1006 | d = work->ch[ch].ns_to; |
| 1007 | spos = work->ch[ch].spos; |
| 1008 | sbpos = work->ch[ch].sbpos; |
| 1009 | sinc = work->ch[ch].sinc; |
| 1010 | |
| 1011 | s_chan = &spu.s_chan[ch]; |
| 1012 | SB = spu.SB + ch * SB_SIZE; |
| 1013 | |
| 1014 | if (s_chan->bNoise) |
| 1015 | do_lsfr_samples(d, work->ctrl, &spu.dwNoiseCount, &spu.dwNoiseVal); |
| 1016 | else if (s_chan->bFMod == 2 |
| 1017 | || (s_chan->bFMod == 0 && spu_config.iUseInterpolation == 0)) |
| 1018 | do_samples_noint(decode_block_work, work, ch, d, SB, sinc, &spos, &sbpos); |
| 1019 | else if (s_chan->bFMod == 0 && spu_config.iUseInterpolation == 1) |
| 1020 | do_samples_simple(decode_block_work, work, ch, d, SB, sinc, &spos, &sbpos); |
| 1021 | else |
| 1022 | do_samples_default(decode_block_work, work, ch, d, SB, sinc, &spos, &sbpos); |
| 1023 | |
| 1024 | d = MixADSR(&work->ch[ch].adsr, d); |
| 1025 | if (d < ns_to) { |
| 1026 | work->ch[ch].adsr.EnvelopeVol = 0; |
| 1027 | memset(&ChanBuf[d], 0, (ns_to - d) * sizeof(ChanBuf[0])); |
| 1028 | } |
| 1029 | |
| 1030 | if (ch == 1 || ch == 3) |
| 1031 | { |
| 1032 | do_decode_bufs(spu.spuMem, ch/2, ns_to, work->decode_pos); |
| 1033 | decode_dirty_ch |= 1 << ch; |
| 1034 | } |
| 1035 | |
| 1036 | if (s_chan->bFMod == 2) // fmod freq channel |
| 1037 | memcpy(iFMod, &ChanBuf, ns_to * sizeof(iFMod[0])); |
| 1038 | if (s_chan->bRVBActive && work->rvb_addr) |
| 1039 | mix_chan_rvb(work->SSumLR, ns_to, |
| 1040 | work->ch[ch].vol_l, work->ch[ch].vol_r, RVB); |
| 1041 | else |
| 1042 | mix_chan(work->SSumLR, ns_to, work->ch[ch].vol_l, work->ch[ch].vol_r); |
| 1043 | } |
| 1044 | |
| 1045 | if (work->rvb_addr) |
| 1046 | REVERBDo(work->SSumLR, RVB, ns_to, work->rvb_addr); |
| 1047 | } |
| 1048 | |
| 1049 | static void sync_worker_thread(int force) |
| 1050 | { |
| 1051 | struct work_item *work; |
| 1052 | int done, used_space; |
| 1053 | |
| 1054 | // rvb offsets will change, thread may be using them |
| 1055 | force |= spu.rvb->dirty && spu.rvb->StartAddr; |
| 1056 | |
| 1057 | done = thread_get_i_done() - worker->i_reaped; |
| 1058 | used_space = worker->i_ready - worker->i_reaped; |
| 1059 | |
| 1060 | //printf("done: %d use: %d dsp: %u/%u\n", done, used_space, |
| 1061 | // worker->boot_cnt, worker->i_done); |
| 1062 | |
| 1063 | while ((force && used_space > 0) || used_space >= WORK_MAXCNT || done > 0) { |
| 1064 | work = &worker->i[worker->i_reaped & WORK_I_MASK]; |
| 1065 | thread_work_wait_sync(work, force); |
| 1066 | |
| 1067 | do_samples_finish(work->SSumLR, work->ns_to, |
| 1068 | work->channels_silent, work->decode_pos); |
| 1069 | |
| 1070 | worker->i_reaped++; |
| 1071 | done = thread_get_i_done() - worker->i_reaped; |
| 1072 | used_space = worker->i_ready - worker->i_reaped; |
| 1073 | } |
| 1074 | if (force) |
| 1075 | thread_sync_caches(); |
| 1076 | } |
| 1077 | |
| 1078 | #else |
| 1079 | |
| 1080 | static void queue_channel_work(int ns_to, int silentch) {} |
| 1081 | static void sync_worker_thread(int force) {} |
| 1082 | |
| 1083 | static const void * const worker = NULL; |
| 1084 | |
| 1085 | #endif // THREAD_ENABLED |
| 1086 | |
| 1087 | //////////////////////////////////////////////////////////////////////// |
| 1088 | // MAIN SPU FUNCTION |
| 1089 | // here is the main job handler... |
| 1090 | //////////////////////////////////////////////////////////////////////// |
| 1091 | |
| 1092 | void do_samples(unsigned int cycles_to, int do_direct) |
| 1093 | { |
| 1094 | unsigned int silentch; |
| 1095 | int cycle_diff; |
| 1096 | int ns_to; |
| 1097 | |
| 1098 | cycle_diff = cycles_to - spu.cycles_played; |
| 1099 | if (cycle_diff < -2*1048576 || cycle_diff > 2*1048576) |
| 1100 | { |
| 1101 | //xprintf("desync %u %d\n", cycles_to, cycle_diff); |
| 1102 | spu.cycles_played = cycles_to; |
| 1103 | return; |
| 1104 | } |
| 1105 | |
| 1106 | silentch = ~(spu.dwChannelOn | spu.dwNewChannel) & 0xffffff; |
| 1107 | |
| 1108 | do_direct |= (silentch == 0xffffff); |
| 1109 | if (worker != NULL) |
| 1110 | sync_worker_thread(do_direct); |
| 1111 | |
| 1112 | if (cycle_diff < 2 * 768) |
| 1113 | return; |
| 1114 | |
| 1115 | ns_to = (cycle_diff / 768 + 1) & ~1; |
| 1116 | if (ns_to > NSSIZE) { |
| 1117 | // should never happen |
| 1118 | //xprintf("ns_to oflow %d %d\n", ns_to, NSSIZE); |
| 1119 | ns_to = NSSIZE; |
| 1120 | } |
| 1121 | |
| 1122 | ////////////////////////////////////////////////////// |
| 1123 | // special irq handling in the decode buffers (0x0000-0x1000) |
| 1124 | // we know: |
| 1125 | // the decode buffers are located in spu memory in the following way: |
| 1126 | // 0x0000-0x03ff CD audio left |
| 1127 | // 0x0400-0x07ff CD audio right |
| 1128 | // 0x0800-0x0bff Voice 1 |
| 1129 | // 0x0c00-0x0fff Voice 3 |
| 1130 | // and decoded data is 16 bit for one sample |
| 1131 | // we assume: |
| 1132 | // even if voices 1/3 are off or no cd audio is playing, the internal |
| 1133 | // play positions will move on and wrap after 0x400 bytes. |
| 1134 | // Therefore: we just need a pointer from spumem+0 to spumem+3ff, and |
| 1135 | // increase this pointer on each sample by 2 bytes. If this pointer |
| 1136 | // (or 0x400 offsets of this pointer) hits the spuirq address, we generate |
| 1137 | // an IRQ. |
| 1138 | |
| 1139 | if (unlikely((spu.spuCtrl & CTRL_IRQ) |
| 1140 | && spu.pSpuIrq < spu.spuMemC+0x1000)) |
| 1141 | { |
| 1142 | int irq_pos = (spu.pSpuIrq - spu.spuMemC) / 2 & 0x1ff; |
| 1143 | int left = (irq_pos - spu.decode_pos) & 0x1ff; |
| 1144 | if (0 < left && left <= ns_to) |
| 1145 | { |
| 1146 | //xprintf("decoder irq %x\n", spu.decode_pos); |
| 1147 | do_irq(); |
| 1148 | } |
| 1149 | } |
| 1150 | |
| 1151 | if (unlikely(spu.rvb->dirty)) |
| 1152 | REVERBPrep(); |
| 1153 | |
| 1154 | if (do_direct || worker == NULL || !spu_config.iUseThread) { |
| 1155 | do_channels(ns_to); |
| 1156 | do_samples_finish(spu.SSumLR, ns_to, silentch, spu.decode_pos); |
| 1157 | } |
| 1158 | else { |
| 1159 | queue_channel_work(ns_to, silentch); |
| 1160 | } |
| 1161 | |
| 1162 | // advance "stopped" channels that can cause irqs |
| 1163 | // (all chans are always playing on the real thing..) |
| 1164 | if (spu.spuCtrl & CTRL_IRQ) |
| 1165 | do_silent_chans(ns_to, silentch); |
| 1166 | |
| 1167 | spu.cycles_played += ns_to * 768; |
| 1168 | spu.decode_pos = (spu.decode_pos + ns_to) & 0x1ff; |
| 1169 | } |
| 1170 | |
| 1171 | static void do_samples_finish(int *SSumLR, int ns_to, |
| 1172 | int silentch, int decode_pos) |
| 1173 | { |
| 1174 | int volmult = spu_config.iVolume; |
| 1175 | int ns; |
| 1176 | int d; |
| 1177 | |
| 1178 | // must clear silent channel decode buffers |
| 1179 | if(unlikely(silentch & spu.decode_dirty_ch & (1<<1))) |
| 1180 | { |
| 1181 | memset(&spu.spuMem[0x800/2], 0, 0x400); |
| 1182 | spu.decode_dirty_ch &= ~(1<<1); |
| 1183 | } |
| 1184 | if(unlikely(silentch & spu.decode_dirty_ch & (1<<3))) |
| 1185 | { |
| 1186 | memset(&spu.spuMem[0xc00/2], 0, 0x400); |
| 1187 | spu.decode_dirty_ch &= ~(1<<3); |
| 1188 | } |
| 1189 | |
| 1190 | MixXA(SSumLR, ns_to, decode_pos); |
| 1191 | |
| 1192 | if((spu.spuCtrl&0x4000)==0) // muted? (rare, don't optimize for this) |
| 1193 | { |
| 1194 | memset(spu.pS, 0, ns_to * 2 * sizeof(spu.pS[0])); |
| 1195 | spu.pS += ns_to * 2; |
| 1196 | } |
| 1197 | else |
| 1198 | for (ns = 0; ns < ns_to * 2; ) |
| 1199 | { |
| 1200 | d = SSumLR[ns]; SSumLR[ns] = 0; |
| 1201 | d = d * volmult >> 10; |
| 1202 | ssat32_to_16(d); |
| 1203 | *spu.pS++ = d; |
| 1204 | ns++; |
| 1205 | |
| 1206 | d = SSumLR[ns]; SSumLR[ns] = 0; |
| 1207 | d = d * volmult >> 10; |
| 1208 | ssat32_to_16(d); |
| 1209 | *spu.pS++ = d; |
| 1210 | ns++; |
| 1211 | } |
| 1212 | } |
| 1213 | |
| 1214 | void schedule_next_irq(void) |
| 1215 | { |
| 1216 | unsigned int upd_samples; |
| 1217 | int ch; |
| 1218 | |
| 1219 | if (spu.scheduleCallback == NULL) |
| 1220 | return; |
| 1221 | |
| 1222 | upd_samples = 44100 / 50; |
| 1223 | |
| 1224 | for (ch = 0; ch < MAXCHAN; ch++) |
| 1225 | { |
| 1226 | if (spu.dwChannelDead & (1 << ch)) |
| 1227 | continue; |
| 1228 | if ((unsigned long)(spu.pSpuIrq - spu.s_chan[ch].pCurr) > IRQ_NEAR_BLOCKS * 16 |
| 1229 | && (unsigned long)(spu.pSpuIrq - spu.s_chan[ch].pLoop) > IRQ_NEAR_BLOCKS * 16) |
| 1230 | continue; |
| 1231 | |
| 1232 | scan_for_irq(ch, &upd_samples); |
| 1233 | } |
| 1234 | |
| 1235 | if (unlikely(spu.pSpuIrq < spu.spuMemC + 0x1000)) |
| 1236 | { |
| 1237 | int irq_pos = (spu.pSpuIrq - spu.spuMemC) / 2 & 0x1ff; |
| 1238 | int left = (irq_pos - spu.decode_pos) & 0x1ff; |
| 1239 | if (0 < left && left < upd_samples) { |
| 1240 | //xprintf("decode: %3d (%3d/%3d)\n", left, spu.decode_pos, irq_pos); |
| 1241 | upd_samples = left; |
| 1242 | } |
| 1243 | } |
| 1244 | |
| 1245 | if (upd_samples < 44100 / 50) |
| 1246 | spu.scheduleCallback(upd_samples * 768); |
| 1247 | } |
| 1248 | |
| 1249 | // SPU ASYNC... even newer epsxe func |
| 1250 | // 1 time every 'cycle' cycles... harhar |
| 1251 | |
| 1252 | // rearmed: called dynamically now |
| 1253 | |
| 1254 | void CALLBACK SPUasync(unsigned int cycle, unsigned int flags) |
| 1255 | { |
| 1256 | do_samples(cycle, spu_config.iUseFixedUpdates); |
| 1257 | |
| 1258 | if (spu.spuCtrl & CTRL_IRQ) |
| 1259 | schedule_next_irq(); |
| 1260 | |
| 1261 | if (flags & 1) { |
| 1262 | out_current->feed(spu.pSpuBuffer, (unsigned char *)spu.pS - spu.pSpuBuffer); |
| 1263 | spu.pS = (short *)spu.pSpuBuffer; |
| 1264 | |
| 1265 | if (spu_config.iTempo) { |
| 1266 | if (!out_current->busy()) |
| 1267 | // cause more samples to be generated |
| 1268 | // (and break some games because of bad sync) |
| 1269 | spu.cycles_played -= 44100 / 60 / 2 * 768; |
| 1270 | } |
| 1271 | } |
| 1272 | } |
| 1273 | |
| 1274 | // SPU UPDATE... new epsxe func |
| 1275 | // 1 time every 32 hsync lines |
| 1276 | // (312/32)x50 in pal |
| 1277 | // (262/32)x60 in ntsc |
| 1278 | |
| 1279 | // since epsxe 1.5.2 (linux) uses SPUupdate, not SPUasync, I will |
| 1280 | // leave that func in the linux port, until epsxe linux is using |
| 1281 | // the async function as well |
| 1282 | |
| 1283 | void CALLBACK SPUupdate(void) |
| 1284 | { |
| 1285 | } |
| 1286 | |
| 1287 | // XA AUDIO |
| 1288 | |
| 1289 | void CALLBACK SPUplayADPCMchannel(xa_decode_t *xap) |
| 1290 | { |
| 1291 | if(!xap) return; |
| 1292 | if(!xap->freq) return; // no xa freq ? bye |
| 1293 | |
| 1294 | FeedXA(xap); // call main XA feeder |
| 1295 | } |
| 1296 | |
| 1297 | // CDDA AUDIO |
| 1298 | int CALLBACK SPUplayCDDAchannel(short *pcm, int nbytes) |
| 1299 | { |
| 1300 | if (!pcm) return -1; |
| 1301 | if (nbytes<=0) return -1; |
| 1302 | |
| 1303 | return FeedCDDA((unsigned char *)pcm, nbytes); |
| 1304 | } |
| 1305 | |
| 1306 | // to be called after state load |
| 1307 | void ClearWorkingState(void) |
| 1308 | { |
| 1309 | memset(iFMod, 0, sizeof(iFMod)); |
| 1310 | spu.pS=(short *)spu.pSpuBuffer; // setup soundbuffer pointer |
| 1311 | } |
| 1312 | |
| 1313 | // SETUPSTREAMS: init most of the spu buffers |
| 1314 | static void SetupStreams(void) |
| 1315 | { |
| 1316 | spu.pSpuBuffer = (unsigned char *)malloc(32768); // alloc mixing buffer |
| 1317 | spu.SSumLR = calloc(NSSIZE * 2, sizeof(spu.SSumLR[0])); |
| 1318 | |
| 1319 | spu.XAStart = // alloc xa buffer |
| 1320 | (uint32_t *)malloc(44100 * sizeof(uint32_t)); |
| 1321 | spu.XAEnd = spu.XAStart + 44100; |
| 1322 | spu.XAPlay = spu.XAStart; |
| 1323 | spu.XAFeed = spu.XAStart; |
| 1324 | |
| 1325 | spu.CDDAStart = // alloc cdda buffer |
| 1326 | (uint32_t *)malloc(CDDA_BUFFER_SIZE); |
| 1327 | spu.CDDAEnd = spu.CDDAStart + 16384; |
| 1328 | spu.CDDAPlay = spu.CDDAStart; |
| 1329 | spu.CDDAFeed = spu.CDDAStart; |
| 1330 | |
| 1331 | ClearWorkingState(); |
| 1332 | } |
| 1333 | |
| 1334 | // REMOVESTREAMS: free most buffer |
| 1335 | static void RemoveStreams(void) |
| 1336 | { |
| 1337 | free(spu.pSpuBuffer); // free mixing buffer |
| 1338 | spu.pSpuBuffer = NULL; |
| 1339 | free(spu.SSumLR); |
| 1340 | spu.SSumLR = NULL; |
| 1341 | free(spu.XAStart); // free XA buffer |
| 1342 | spu.XAStart = NULL; |
| 1343 | free(spu.CDDAStart); // free CDDA buffer |
| 1344 | spu.CDDAStart = NULL; |
| 1345 | } |
| 1346 | |
| 1347 | #if defined(C64X_DSP) |
| 1348 | |
| 1349 | /* special code for TI C64x DSP */ |
| 1350 | #include "spu_c64x.c" |
| 1351 | |
| 1352 | #elif defined(THREAD_ENABLED) |
| 1353 | |
| 1354 | #include <pthread.h> |
| 1355 | #include <semaphore.h> |
| 1356 | #include <unistd.h> |
| 1357 | |
| 1358 | static struct { |
| 1359 | pthread_t thread; |
| 1360 | sem_t sem_avail; |
| 1361 | sem_t sem_done; |
| 1362 | } t; |
| 1363 | |
| 1364 | /* generic pthread implementation */ |
| 1365 | |
| 1366 | static void thread_work_start(void) |
| 1367 | { |
| 1368 | sem_post(&t.sem_avail); |
| 1369 | } |
| 1370 | |
| 1371 | static void thread_work_wait_sync(struct work_item *work, int force) |
| 1372 | { |
| 1373 | sem_wait(&t.sem_done); |
| 1374 | } |
| 1375 | |
| 1376 | static int thread_get_i_done(void) |
| 1377 | { |
| 1378 | return worker->i_done; |
| 1379 | } |
| 1380 | |
| 1381 | static void thread_sync_caches(void) |
| 1382 | { |
| 1383 | } |
| 1384 | |
| 1385 | static void *spu_worker_thread(void *unused) |
| 1386 | { |
| 1387 | struct work_item *work; |
| 1388 | |
| 1389 | while (1) { |
| 1390 | sem_wait(&t.sem_avail); |
| 1391 | if (worker->exit_thread) |
| 1392 | break; |
| 1393 | |
| 1394 | work = &worker->i[worker->i_done & WORK_I_MASK]; |
| 1395 | do_channel_work(work); |
| 1396 | worker->i_done++; |
| 1397 | |
| 1398 | sem_post(&t.sem_done); |
| 1399 | } |
| 1400 | |
| 1401 | return NULL; |
| 1402 | } |
| 1403 | |
| 1404 | static void init_spu_thread(void) |
| 1405 | { |
| 1406 | int ret; |
| 1407 | |
| 1408 | if (sysconf(_SC_NPROCESSORS_ONLN) <= 1) |
| 1409 | return; |
| 1410 | |
| 1411 | worker = calloc(1, sizeof(*worker)); |
| 1412 | if (worker == NULL) |
| 1413 | return; |
| 1414 | ret = sem_init(&t.sem_avail, 0, 0); |
| 1415 | if (ret != 0) |
| 1416 | goto fail_sem_avail; |
| 1417 | ret = sem_init(&t.sem_done, 0, 0); |
| 1418 | if (ret != 0) |
| 1419 | goto fail_sem_done; |
| 1420 | |
| 1421 | ret = pthread_create(&t.thread, NULL, spu_worker_thread, NULL); |
| 1422 | if (ret != 0) |
| 1423 | goto fail_thread; |
| 1424 | |
| 1425 | spu_config.iThreadAvail = 1; |
| 1426 | return; |
| 1427 | |
| 1428 | fail_thread: |
| 1429 | sem_destroy(&t.sem_done); |
| 1430 | fail_sem_done: |
| 1431 | sem_destroy(&t.sem_avail); |
| 1432 | fail_sem_avail: |
| 1433 | free(worker); |
| 1434 | worker = NULL; |
| 1435 | spu_config.iThreadAvail = 0; |
| 1436 | } |
| 1437 | |
| 1438 | static void exit_spu_thread(void) |
| 1439 | { |
| 1440 | if (worker == NULL) |
| 1441 | return; |
| 1442 | worker->exit_thread = 1; |
| 1443 | sem_post(&t.sem_avail); |
| 1444 | pthread_join(t.thread, NULL); |
| 1445 | sem_destroy(&t.sem_done); |
| 1446 | sem_destroy(&t.sem_avail); |
| 1447 | free(worker); |
| 1448 | worker = NULL; |
| 1449 | } |
| 1450 | |
| 1451 | #else // if !THREAD_ENABLED |
| 1452 | |
| 1453 | static void init_spu_thread(void) |
| 1454 | { |
| 1455 | } |
| 1456 | |
| 1457 | static void exit_spu_thread(void) |
| 1458 | { |
| 1459 | } |
| 1460 | |
| 1461 | #endif |
| 1462 | |
| 1463 | // SPUINIT: this func will be called first by the main emu |
| 1464 | long CALLBACK SPUinit(void) |
| 1465 | { |
| 1466 | int i; |
| 1467 | |
| 1468 | spu.spuMemC = calloc(1, 512 * 1024); |
| 1469 | InitADSR(); |
| 1470 | |
| 1471 | spu.s_chan = calloc(MAXCHAN+1, sizeof(spu.s_chan[0])); // channel + 1 infos (1 is security for fmod handling) |
| 1472 | spu.rvb = calloc(1, sizeof(REVERBInfo)); |
| 1473 | spu.SB = calloc(MAXCHAN, sizeof(spu.SB[0]) * SB_SIZE); |
| 1474 | |
| 1475 | spu.spuAddr = 0; |
| 1476 | spu.decode_pos = 0; |
| 1477 | spu.pSpuIrq = spu.spuMemC; |
| 1478 | |
| 1479 | SetupStreams(); // prepare streaming |
| 1480 | |
| 1481 | if (spu_config.iVolume == 0) |
| 1482 | spu_config.iVolume = 768; // 1024 is 1.0 |
| 1483 | |
| 1484 | init_spu_thread(); |
| 1485 | |
| 1486 | for (i = 0; i < MAXCHAN; i++) // loop sound channels |
| 1487 | { |
| 1488 | spu.s_chan[i].ADSRX.SustainLevel = 0xf; // -> init sustain |
| 1489 | spu.s_chan[i].ADSRX.SustainIncrease = 1; |
| 1490 | spu.s_chan[i].pLoop = spu.spuMemC; |
| 1491 | spu.s_chan[i].pCurr = spu.spuMemC; |
| 1492 | } |
| 1493 | |
| 1494 | spu.bSpuInit=1; // flag: we are inited |
| 1495 | |
| 1496 | return 0; |
| 1497 | } |
| 1498 | |
| 1499 | // SPUOPEN: called by main emu after init |
| 1500 | long CALLBACK SPUopen(void) |
| 1501 | { |
| 1502 | if (spu.bSPUIsOpen) return 0; // security for some stupid main emus |
| 1503 | |
| 1504 | SetupSound(); // setup sound (before init!) |
| 1505 | |
| 1506 | spu.bSPUIsOpen = 1; |
| 1507 | |
| 1508 | return PSE_SPU_ERR_SUCCESS; |
| 1509 | } |
| 1510 | |
| 1511 | // SPUCLOSE: called before shutdown |
| 1512 | long CALLBACK SPUclose(void) |
| 1513 | { |
| 1514 | if (!spu.bSPUIsOpen) return 0; // some security |
| 1515 | |
| 1516 | spu.bSPUIsOpen = 0; // no more open |
| 1517 | |
| 1518 | out_current->finish(); // no more sound handling |
| 1519 | |
| 1520 | return 0; |
| 1521 | } |
| 1522 | |
| 1523 | // SPUSHUTDOWN: called by main emu on final exit |
| 1524 | long CALLBACK SPUshutdown(void) |
| 1525 | { |
| 1526 | SPUclose(); |
| 1527 | |
| 1528 | exit_spu_thread(); |
| 1529 | |
| 1530 | free(spu.spuMemC); |
| 1531 | spu.spuMemC = NULL; |
| 1532 | free(spu.SB); |
| 1533 | spu.SB = NULL; |
| 1534 | free(spu.s_chan); |
| 1535 | spu.s_chan = NULL; |
| 1536 | free(spu.rvb); |
| 1537 | spu.rvb = NULL; |
| 1538 | |
| 1539 | RemoveStreams(); // no more streaming |
| 1540 | spu.bSpuInit=0; |
| 1541 | |
| 1542 | return 0; |
| 1543 | } |
| 1544 | |
| 1545 | // SPUTEST: we don't test, we are always fine ;) |
| 1546 | long CALLBACK SPUtest(void) |
| 1547 | { |
| 1548 | return 0; |
| 1549 | } |
| 1550 | |
| 1551 | // SPUCONFIGURE: call config dialog |
| 1552 | long CALLBACK SPUconfigure(void) |
| 1553 | { |
| 1554 | #ifdef _MACOSX |
| 1555 | DoConfiguration(); |
| 1556 | #else |
| 1557 | // StartCfgTool("CFG"); |
| 1558 | #endif |
| 1559 | return 0; |
| 1560 | } |
| 1561 | |
| 1562 | // SPUABOUT: show about window |
| 1563 | void CALLBACK SPUabout(void) |
| 1564 | { |
| 1565 | #ifdef _MACOSX |
| 1566 | DoAbout(); |
| 1567 | #else |
| 1568 | // StartCfgTool("ABOUT"); |
| 1569 | #endif |
| 1570 | } |
| 1571 | |
| 1572 | // SETUP CALLBACKS |
| 1573 | // this functions will be called once, |
| 1574 | // passes a callback that should be called on SPU-IRQ/cdda volume change |
| 1575 | void CALLBACK SPUregisterCallback(void (CALLBACK *callback)(void)) |
| 1576 | { |
| 1577 | spu.irqCallback = callback; |
| 1578 | } |
| 1579 | |
| 1580 | void CALLBACK SPUregisterCDDAVolume(void (CALLBACK *CDDAVcallback)(unsigned short,unsigned short)) |
| 1581 | { |
| 1582 | spu.cddavCallback = CDDAVcallback; |
| 1583 | } |
| 1584 | |
| 1585 | void CALLBACK SPUregisterScheduleCb(void (CALLBACK *callback)(unsigned int)) |
| 1586 | { |
| 1587 | spu.scheduleCallback = callback; |
| 1588 | } |
| 1589 | |
| 1590 | // COMMON PLUGIN INFO FUNCS |
| 1591 | /* |
| 1592 | char * CALLBACK PSEgetLibName(void) |
| 1593 | { |
| 1594 | return _(libraryName); |
| 1595 | } |
| 1596 | |
| 1597 | unsigned long CALLBACK PSEgetLibType(void) |
| 1598 | { |
| 1599 | return PSE_LT_SPU; |
| 1600 | } |
| 1601 | |
| 1602 | unsigned long CALLBACK PSEgetLibVersion(void) |
| 1603 | { |
| 1604 | return (1 << 16) | (6 << 8); |
| 1605 | } |
| 1606 | |
| 1607 | char * SPUgetLibInfos(void) |
| 1608 | { |
| 1609 | return _(libraryInfo); |
| 1610 | } |
| 1611 | */ |
| 1612 | |
| 1613 | // debug |
| 1614 | void spu_get_debug_info(int *chans_out, int *run_chans, int *fmod_chans_out, int *noise_chans_out) |
| 1615 | { |
| 1616 | int ch = 0, fmod_chans = 0, noise_chans = 0, irq_chans = 0; |
| 1617 | |
| 1618 | if (spu.s_chan == NULL) |
| 1619 | return; |
| 1620 | |
| 1621 | for(;ch<MAXCHAN;ch++) |
| 1622 | { |
| 1623 | if (!(spu.dwChannelOn & (1<<ch))) |
| 1624 | continue; |
| 1625 | if (spu.s_chan[ch].bFMod == 2) |
| 1626 | fmod_chans |= 1 << ch; |
| 1627 | if (spu.s_chan[ch].bNoise) |
| 1628 | noise_chans |= 1 << ch; |
| 1629 | if((spu.spuCtrl&CTRL_IRQ) && spu.s_chan[ch].pCurr <= spu.pSpuIrq && spu.s_chan[ch].pLoop <= spu.pSpuIrq) |
| 1630 | irq_chans |= 1 << ch; |
| 1631 | } |
| 1632 | |
| 1633 | *chans_out = spu.dwChannelOn; |
| 1634 | *run_chans = ~spu.dwChannelOn & ~spu.dwChannelDead & irq_chans; |
| 1635 | *fmod_chans_out = fmod_chans; |
| 1636 | *noise_chans_out = noise_chans; |
| 1637 | } |
| 1638 | |
| 1639 | // vim:shiftwidth=1:expandtab |