--- /dev/null
+#include "../PicoInt.h"
+#include "pcm.h"
+
+static unsigned int g_rate = 0; // 18.14 fixed point
+
+void pcm_write(unsigned int a, unsigned int d)
+{
+//printf("pcm_write(%i, %02x)\n", a, d);
+
+ if (a < 7)
+ {
+ Pico_mcd->pcm.ch[Pico_mcd->pcm.cur_ch].regs[a] = d;
+ }
+ else if (a == 7) // control register
+ {
+ if (d & 0x40) Pico_mcd->pcm.cur_ch = d & 7;
+ else Pico_mcd->pcm.bank = d & 0xf;
+ Pico_mcd->pcm.control = d;
+ // dprintf("pcm control=%02x", Pico_mcd->pcm.control);
+ }
+ else if (a == 8) // sound on/off
+ {
+ if (!(Pico_mcd->pcm.enabled & 0x01)) Pico_mcd->pcm.ch[0].addr =
+ Pico_mcd->pcm.ch[0].regs[6] << (PCM_STEP_SHIFT + 8);
+ if (!(Pico_mcd->pcm.enabled & 0x02)) Pico_mcd->pcm.ch[1].addr =
+ Pico_mcd->pcm.ch[1].regs[6] << (PCM_STEP_SHIFT + 8);
+ if (!(Pico_mcd->pcm.enabled & 0x04)) Pico_mcd->pcm.ch[2].addr =
+ Pico_mcd->pcm.ch[2].regs[6] << (PCM_STEP_SHIFT + 8);
+ if (!(Pico_mcd->pcm.enabled & 0x08)) Pico_mcd->pcm.ch[3].addr =
+ Pico_mcd->pcm.ch[3].regs[6] << (PCM_STEP_SHIFT + 8);
+ if (!(Pico_mcd->pcm.enabled & 0x10)) Pico_mcd->pcm.ch[4].addr =
+ Pico_mcd->pcm.ch[4].regs[6] << (PCM_STEP_SHIFT + 8);
+ if (!(Pico_mcd->pcm.enabled & 0x20)) Pico_mcd->pcm.ch[5].addr =
+ Pico_mcd->pcm.ch[5].regs[6] << (PCM_STEP_SHIFT + 8);
+ if (!(Pico_mcd->pcm.enabled & 0x40)) Pico_mcd->pcm.ch[6].addr =
+ Pico_mcd->pcm.ch[6].regs[6] << (PCM_STEP_SHIFT + 8);
+ if (!(Pico_mcd->pcm.enabled & 0x80)) Pico_mcd->pcm.ch[7].addr =
+ Pico_mcd->pcm.ch[7].regs[6] << (PCM_STEP_SHIFT + 8);
+// printf("addr %x %x %x %x %x %x %x %x\n", Pico_mcd->pcm.ch[0].addr, Pico_mcd->pcm.ch[1].addr
+// , Pico_mcd->pcm.ch[2].addr, Pico_mcd->pcm.ch[3].addr, Pico_mcd->pcm.ch[4].addr, Pico_mcd->pcm.ch[5].addr
+// , Pico_mcd->pcm.ch[6].addr, Pico_mcd->pcm.ch[7].addr);
+
+ Pico_mcd->pcm.enabled = ~d;
+//printf("enabled=%02x\n", Pico_mcd->pcm.enabled);
+ }
+}
+
+
+void pcm_set_rate(int rate)
+{
+ double step = 31.8 * 1024.0 / (double) rate; // max <4 @ 8000Hz
+ step *= 256*256/4;
+ g_rate = (unsigned int) step;
+ printf("g_rate: %08x\n", g_rate);
+}
+
+
+// TODO: make use of the fact that max_length == 3
+
+void pcm_update(int *buffer, int length, int stereo)
+{
+ struct pcm_chan *ch;
+ unsigned int step, addr;
+ int mul_l, mul_r, smp;
+ int i, j, k;
+ int *out;
+
+
+ // PCM disabled or all channels off (to be checked by caller)
+ //if (!(Pico_mcd->pcm.control & 0x80) || !Pico_mcd->pcm.enabled) return;
+
+//printf("-- upd %i\n", length);
+
+ for (i = 0; i < 8; i++)
+ {
+ if (!(Pico_mcd->pcm.enabled & (1 << i))) continue; // channel disabled
+
+ out = buffer;
+ ch = &Pico_mcd->pcm.ch[i];
+
+ addr = ch->addr; // >> PCM_STEP_SHIFT;
+ mul_l = ((int)ch->regs[0] * (ch->regs[1] & 0xf)) >> (5+1); // (env * pan) >> 5
+ mul_r = ((int)ch->regs[0] * (ch->regs[1] >> 4)) >> (5+1);
+ step = ((unsigned int)(*(unsigned short *)&ch->regs[2]) * g_rate) >> 14; // freq step
+// printf("step=%i, cstep=%i, mul_l=%i, mul_r=%i, ch=%i, addr=%x, en=%02x\n",
+// *(unsigned short *)&ch->regs[2], step, mul_l, mul_r, i, addr, Pico_mcd->pcm.enabled);
+
+ if (!stereo && mul_l < mul_r) mul_l = mul_r;
+
+ for (j = 0; j < length; j++)
+ {
+// printf("addr=%08x\n", addr);
+ smp = Pico_mcd->pcm_ram[addr >> PCM_STEP_SHIFT];
+
+ // test for loop signal
+ if (smp == 0xff)
+ {
+ addr = *(unsigned short *)&ch->regs[4]; // loop_addr
+ smp = Pico_mcd->pcm_ram[addr];
+ addr <<= PCM_STEP_SHIFT;
+ if (smp == 0xff) break;
+ }
+
+ if (smp & 0x80) smp = -(smp & 0x7f);
+
+ *out++ += smp * mul_l; // max 128 * 119 = 15232
+ if(stereo)
+ *out++ += smp * mul_r;
+
+ // update address register
+ k = (addr >> PCM_STEP_SHIFT) + 1;
+ addr = (addr + step) & 0x7FFFFFF;
+
+ for(; k < (addr >> PCM_STEP_SHIFT); k++)
+ {
+ if (Pico_mcd->pcm_ram[k] == 0xff)
+ {
+ addr = (unsigned int)(*(unsigned short *)&ch->regs[4]) << PCM_STEP_SHIFT; // loop_addr
+ break;
+ }
+ }
+ }
+
+ if (Pico_mcd->pcm_ram[addr >> PCM_STEP_SHIFT] == 0xff)
+ addr = (unsigned int)(*(unsigned short *)&ch->regs[4]) << PCM_STEP_SHIFT; // loop_addr
+
+ ch->addr = addr;
+ }
+}
+
--- /dev/null
+
+#define PCM_STEP_SHIFT 11
+
+void pcm_write(unsigned int a, unsigned int d);
+void pcm_set_rate(int rate);
+void pcm_update(int *buffer, int length, int stereo);
+
pcm_update(buf32, length, stereo);\r
\r
// CD: CDDA audio\r
-// if ((PicoMCD & 1) && (PicoOpt & 0x800))\r
-// mp3_update(PsndBuffer+offset, length, stereo);\r
+ if ((PicoMCD & 1) && (PicoOpt & 0x800))\r
+ mp3_update(buf32, length, stereo);\r
\r
// convert + limit to normal 16bit output\r
if (stereo)\r
#include "emu.h"\r
#include "menu.h"\r
#include "asmutils.h"\r
+#include "mp3.h"\r
#include "../../Pico/PicoInt.h"\r
-\r
-// tmp\r
#include "../../Pico/sound/mix.h"\r
\r
/* we will need some gp2x internals here */\r
\r
static unsigned char *shared_mem = 0;\r
static _940_data_t *shared_data = 0;\r
-static _940_ctl_t *shared_ctl = 0;\r
-static unsigned char *mp3_mem = 0;\r
+_940_ctl_t *shared_ctl = 0;\r
+unsigned char *mp3_mem = 0;\r
\r
#define MP3_SIZE_MAX (0x1000000 - 4*640*480)\r
\r
return 0;\r
}\r
\r
-static int g_busy = 10;\r
\r
-static void wait_busy_940(void)\r
+#define CHECK_BUSY(job) \\r
+ (gp2x_memregs[0x3b46>>1] & (1<<(job-1)))\r
+\r
+static void wait_busy_940(int job)\r
{\r
int i;\r
-#if 0\r
- printf("940 busy, entering wait loop.. (cnt: %i, wc: %i, ve: ", shared_ctl->loopc, shared_ctl->waitc);\r
- for (i = 0; i < 8; i++)\r
- printf("%i ", shared_ctl->vstarts[i]);\r
- printf(")\n");\r
\r
- for (i = 0; shared_ctl->busy; i++)\r
- {\r
- spend_cycles(1024); /* needs tuning */\r
- }\r
- printf("wait iterations: %i\n", i);\r
-#else\r
- for (i = 0; /*shared_ctl->busy*/gp2x_memregs[0x3B3E>>1] && i < 0x10000; i++)\r
+ job--;\r
+ for (i = 0; (gp2x_memregs[0x3b46>>1] & (1<<job)) && i < 0x10000; i++)\r
spend_cycles(8*1024); // tested to be best for mp3 dec\r
- //printf("i = 0x%x\n", i);\r
if (i < 0x10000) return;\r
\r
/* 940 crashed */\r
for (i = 0; i < 8; i++)\r
printf("%i ", shared_ctl->vstarts[i]);\r
printf(")\n");\r
- printf("irq pending flags: DUALCPU %04x (see 15?), SRCPND %08lx (see 26), INTPND %08lx\n",\r
+ printf("irq pending flags: DUALCPU %04x, SRCPND %08lx (see 26), INTPND %08lx\n",\r
gp2x_memregs[0x3b46>>1], gp2x_memregl[0x4500>>2], gp2x_memregl[0x4510>>2]);\r
- printf("last irq PC: %08x, lastjob: 0x%03x, busy: %x, lastbusy: %x, g_busy: %x\n", shared_ctl->last_irq_pc,\r
- shared_ctl->lastjob, gp2x_memregs[0x3B3E>>1]/*shared_ctl->busy*/, shared_ctl->lastbusy, g_busy);\r
+ printf("last lr: %08x, lastjob: %i\n", shared_ctl->last_lr, shared_ctl->lastjob);\r
printf("trying to interrupt..\n");\r
gp2x_memregs[0x3B3E>>1] = 0xffff;\r
- for (i = 0; /*shared_ctl->busy*/gp2x_memregs[0x3B3E>>1] && i < 0x10000; i++)\r
+ for (i = 0; gp2x_memregs[0x3b46>>1] && i < 0x10000; i++)\r
spend_cycles(8*1024);\r
printf("i = 0x%x\n", i);\r
- printf("irq pending flags: DUALCPU %04x (see 15?), SRCPND %08lx (see 26), INTPND %08lx\n",\r
+ printf("irq pending flags: DUALCPU %04x, SRCPND %08lx (see 26), INTPND %08lx\n",\r
gp2x_memregs[0x3b46>>1], gp2x_memregl[0x4500>>2], gp2x_memregl[0x4510>>2]);\r
- printf("last irq PC: %08x, lastjob: 0x%03x, busy: %x\n", shared_ctl->last_irq_pc, shared_ctl->lastjob, gp2x_memregs[0x3B3E>>1]/*shared_ctl->busy*/);\r
+ printf("last lr: %08x, lastjob: %i\n", shared_ctl->last_lr, shared_ctl->lastjob);\r
\r
strcpy(menuErrorMsg, "940 crashed.");\r
engineState = PGS_Menu;\r
crashed_940 = 1;\r
-#endif\r
}\r
\r
\r
-static void add_job_940(int job0, int job1)\r
+static void add_job_940(int job)\r
{\r
-/* if (gp2x_memregs[0x3b46>>1] || shared_ctl->busy)\r
- {\r
- printf("!!add_job_940: irq pending flags: DUALCPU %04x (see 15?), SRCPND %08lx (see 26), INTPND %08lx, busy: %x\n",\r
- gp2x_memregs[0x3b46>>1], gp2x_memregl[0x4500>>2], gp2x_memregl[0x4510>>2], shared_ctl->busy);\r
+ if (job <= 0 || job > 16) {\r
+ printf("add_job_940: bad job: %i\n", job);\r
+ return;\r
}\r
-*/\r
- shared_ctl->jobs[0] = job0;\r
- shared_ctl->jobs[1] = job1;\r
- //shared_ctl->busy = ++g_busy; // 1;\r
- gp2x_memregs[0x3B3E>>1] = ++g_busy; // set busy flag\r
-// gp2x_memregs[0x3B3E>>1] = 0xffff; // cause interrupt\r
+\r
+ // generate interrupt for this job\r
+ job--;\r
+ gp2x_memregs[(0x3B20+job*2)>>1] = 1;\r
+\r
+// printf("added %i, pending %04x\n", job+1, gp2x_memregs[0x3b46>>1]);\r
}\r
\r
\r
{\r
int i, old_A1 = addr_A1;\r
\r
- if (/*shared_ctl->busy*/gp2x_memregs[0x3B3E>>1]) wait_busy_940();\r
-\r
// feed all the registers and update internal state\r
for(i = 0; i < 0x100; i++) {\r
YM2612Write_940(0, i);\r
\r
addr_A1 = old_A1;\r
\r
- add_job_940(JOB940_PICOSTATELOAD, 0);\r
+ add_job_940(JOB940_PICOSTATELOAD);\r
}\r
\r
\r
gp2x_memregs[0x3B40>>1] = 0; // disable DUALCPU interrupts for 920\r
gp2x_memregs[0x3B42>>1] = 1; // enable DUALCPU interrupts for 940\r
\r
+ gp2x_memregl[0x4504>>2] = 0; // make sure no FIQs will be generated\r
+ gp2x_memregl[0x4508>>2] = ~(1<<26); // unmask DUALCPU ints in the undocumented 940's interrupt controller\r
+\r
+\r
+\r
if (shared_mem == NULL)\r
{\r
shared_mem = (unsigned char *) mmap(0, 0x210000, PROT_READ|PROT_WRITE, MAP_SHARED, memdev, 0x2000000);\r
\r
loaded_mp3 = 0;\r
\r
- /* now cause 940 to init it's ym2612 stuff */\r
- shared_ctl->baseclock = baseclock;\r
- shared_ctl->rate = rate;\r
- shared_ctl->jobs[0] = JOB940_INITALL;\r
- shared_ctl->jobs[1] = 0;\r
- //shared_ctl->busy = 1;\r
- gp2x_memregs[0x3B3E>>1] = 1; // set busy flag\r
-\r
gp2x_memregs[0x3B46>>1] = 0xffff; // clear pending DUALCPU interrupts for 940\r
- gp2x_memregl[0x4500>>2] = 0; // clear pending IRQs in SRCPND\r
- gp2x_memregl[0x4510>>2] = 0; // clear pending IRQs in INTPND\r
- gp2x_memregl[0x4508>>2] = ~(1<<26); // unmask DUALCPU ints in the undocumented 940's interrupt controller\r
+ gp2x_memregl[0x4500>>2] = 0xffffffff; // clear pending IRQs in SRCPND\r
+ gp2x_memregl[0x4510>>2] = 0xffffffff; // clear pending IRQs in INTPND\r
\r
/* start the 940 */\r
Reset940(0, 2);\r
Pause940(0);\r
\r
// YM2612ResetChip_940(); // will be done on JOB940_YM2612INIT\r
+\r
+ /* now cause 940 to init it's ym2612 stuff */\r
+ shared_ctl->baseclock = baseclock;\r
+ shared_ctl->rate = rate;\r
+ add_job_940(JOB940_INITALL);\r
}\r
\r
\r
return;\r
}\r
\r
- if (/*shared_ctl->busy*/gp2x_memregs[0x3B3E>>1]) wait_busy_940();\r
-\r
internal_reset();\r
\r
- add_job_940(JOB940_YM2612RESETCHIP, 0);\r
+ add_job_940(JOB940_YM2612RESETCHIP);\r
}\r
\r
\r
-//extern int pcm_buffer[2*44100/50];\r
-/*\r
-static void mix_samples(int *dest_buf, short *mp3_buf, int len, int stereo)\r
+int YM2612UpdateOne_940(int *buffer, int length, int stereo, int is_buf_empty)\r
{\r
-// int *pcm = pcm_buffer + offset * 2;\r
+ int *ym_buf = shared_data->ym_buffer;\r
\r
- if (stereo)\r
- {\r
- for (; len > 0; len--)\r
- {\r
- int lm, rm;\r
- lm = *mp3_buf++; rm = *mp3_buf++;\r
- *dest_buf++ += lm - lm/2; *dest_buf++ += rm - rm/2;\r
- }\r
+ //printf("YM2612UpdateOne_940()\n");\r
+\r
+ if (CHECK_BUSY(JOB940_YM2612UPDATEONE)) wait_busy_940(JOB940_YM2612UPDATEONE);\r
+\r
+ // mix in ym buffer\r
+ if (is_buf_empty) memcpy32(buffer, ym_buf, length<<stereo);\r
+ // else TODO\r
+\r
+// for (len = length << stereo; len > 0; len--)\r
+// {\r
+// *dest_buf++ += *ym_buf++;\r
+// }\r
+\r
+ if (shared_ctl->writebuffsel == 1) {\r
+ shared_ctl->writebuff0[writebuff_ptr] = 0xffff;\r
} else {\r
- for (; len > 0; len--)\r
- {\r
- int l = *mp3_buf++;\r
- *dest_buf++ = l - l/2;\r
- }\r
+ shared_ctl->writebuff1[writebuff_ptr] = 0xffff;\r
}\r
+ writebuff_ptr = 0;\r
+\r
+ /* predict sample counter for next frame */\r
+ if (PsndLen_exc_add) {\r
+ if (PsndLen_exc_cnt + PsndLen_exc_add >= 0x10000) length = PsndLen + 1;\r
+ else length = PsndLen;\r
+ }\r
+\r
+ /* give 940 ym job */\r
+ shared_ctl->writebuffsel ^= 1;\r
+ shared_ctl->length = length;\r
+ shared_ctl->stereo = stereo;\r
+\r
+ add_job_940(JOB940_YM2612UPDATEONE);\r
+\r
+ return 1;\r
}\r
-*/\r
\r
-// here we assume that length is different between games, but constant in one game\r
\r
static int mp3_samples_ready = 0, mp3_buffer_offs = 0;\r
-static int mp3_play_bufsel = 0;\r
+static int mp3_play_bufsel = 0, mp3_job_started = 0;\r
\r
-int YM2612UpdateOne_940(int *buffer, int length, int stereo, int is_buf_empty)\r
+void mp3_update(int *buffer, int length, int stereo)\r
{\r
- int length_mp3 = Pico.m.pal ? 44100/50 : 44100/60; // mp3s are locked to 44100Hz stereo\r
- int *ym_buf = shared_data->mix_buffer;\r
-// int *dest_buf = buffer;\r
- int cdda_on, mp3_job = 0;\r
-// int len;\r
+ int length_mp3;\r
+ int cdda_on;\r
\r
- // emulating CD, enabled in opts, not data track, CDC is reading, playback was started, track not ended\r
- cdda_on = (PicoMCD & 1) && (PicoOpt & 0x800) && !(Pico_mcd->s68k_regs[0x36] & 1) &&\r
- (Pico_mcd->scd.Status_CDC & 1) && loaded_mp3;\r
+ // not data track, CDC is reading, playback was started, track not ended\r
+ cdda_on = !(Pico_mcd->s68k_regs[0x36] & 1) && (Pico_mcd->scd.Status_CDC & 1) &&\r
+ loaded_mp3 && shared_ctl->mp3_offs < shared_ctl->mp3_len;\r
\r
- //printf("YM2612UpdateOne_940()\n");\r
- if (/*shared_ctl->busy*/gp2x_memregs[0x3B3E>>1]) wait_busy_940();\r
+ if (!cdda_on) return;\r
\r
- // track ended?\r
- cdda_on = cdda_on && shared_ctl->mp3_offs < shared_ctl->mp3_len;\r
+ if (!(PicoOpt&0x200)) {\r
+ mp3_update_local(buffer, length, stereo);\r
+ return;\r
+ }\r
\r
- // mix in ym buffer\r
- if (is_buf_empty) memcpy32(buffer, ym_buf, length<<stereo);\r
- // else TODO\r
+ length_mp3 = length;\r
+ if (PsndRate == 22050) length_mp3 <<= 1; // mp3s are locked to 44100Hz stereo\r
+ else if (PsndRate == 11025) length_mp3 <<= 2; // so make length 44100ish\r
\r
-// for (len = length << stereo; len > 0; len--)\r
-// {\r
-// *dest_buf++ += *ym_buf++;\r
-// }\r
+ /* do we have to wait? */\r
+ if (mp3_job_started && mp3_samples_ready < length_mp3) {\r
+ if (CHECK_BUSY(JOB940_MP3DECODE)) wait_busy_940(JOB940_MP3DECODE);\r
+ mp3_job_started = 0;\r
+ mp3_samples_ready += 1152;\r
+ }\r
\r
/* mix mp3 data, only stereo */\r
- if (cdda_on && mp3_samples_ready >= length_mp3)\r
+ if (mp3_samples_ready >= length_mp3)\r
{\r
int shr = 0;\r
void (*mix_samples)(int *dest_buf, short *mp3_buf, int count) = mix_16h_to_32;\r
else if (PsndRate == 11025) { mix_samples = mix_16h_to_32_s2; shr = 2; }\r
\r
if (1152 - mp3_buffer_offs >= length_mp3) {\r
- mix_samples(buffer, shared_data->mp3_buffer[mp3_play_bufsel] + mp3_buffer_offs*2, (length_mp3>>shr)<<1);\r
+ mix_samples(buffer, shared_data->mp3_buffer[mp3_play_bufsel] + mp3_buffer_offs*2, length<<1);\r
\r
mp3_buffer_offs += length_mp3;\r
} else {\r
- // collect from both buffers..\r
+ // collect samples from both buffers..\r
int left = 1152 - mp3_buffer_offs;\r
if (mp3_play_bufsel == 0)\r
{\r
- mix_samples(buffer, shared_data->mp3_buffer[0] + mp3_buffer_offs*2, (length_mp3>>shr)<<1);\r
+ mix_samples(buffer, shared_data->mp3_buffer[0] + mp3_buffer_offs*2, length<<1);\r
mp3_buffer_offs = length_mp3 - left;\r
mp3_play_bufsel = 1;\r
} else {\r
mp3_samples_ready -= length_mp3;\r
}\r
\r
- if (shared_ctl->writebuffsel == 1) {\r
- shared_ctl->writebuff0[writebuff_ptr] = 0xffff;\r
- } else {\r
- shared_ctl->writebuff1[writebuff_ptr] = 0xffff;\r
- }\r
- writebuff_ptr = 0;\r
-\r
- /* predict sample counter for next frame */\r
- if (PsndLen_exc_add) {\r
- if (PsndLen_exc_cnt + PsndLen_exc_add >= 0x10000) length = PsndLen + 1;\r
- else length = PsndLen;\r
- }\r
-\r
- /* give 940 ym job */\r
- shared_ctl->writebuffsel ^= 1;\r
- shared_ctl->length = length;\r
- shared_ctl->stereo = stereo;\r
-\r
- // make sure we will have enough mp3 samples next frame\r
- if (cdda_on && mp3_samples_ready < length_mp3)\r
+ // ask to decode more if we already can\r
+ if (!mp3_job_started)\r
{\r
+ mp3_job_started = 1;\r
shared_ctl->mp3_buffsel ^= 1;\r
- mp3_job = JOB940_MP3DECODE;\r
- mp3_samples_ready += 1152;\r
+ add_job_940(JOB940_MP3DECODE);\r
}\r
-\r
- add_job_940(JOB940_YM2612UPDATEONE, mp3_job);\r
- //spend_cycles(512);\r
- //printf("SRCPND: %08lx, INTMODE: %08lx, INTMASK: %08lx, INTPEND: %08lx\n",\r
- // gp2x_memregl[0x4500>>2], gp2x_memregl[0x4504>>2], gp2x_memregl[0x4508>>2], gp2x_memregl[0x4510>>2]);\r
-\r
- return 1;\r
}\r
\r
\r
// else printf("done. mp3 too large, not all data loaded.\n");\r
shared_ctl->mp3_len = ftell(f);\r
loaded_mp3 = f;\r
+\r
+ if (PicoOpt&0x200) {\r
+ // as we are going to change 940's cacheable area, we must invalidate it's cache..\r
+ if (CHECK_BUSY(JOB940_MP3DECODE)) wait_busy_940(JOB940_MP3DECODE);\r
+ add_job_940(JOB940_INVALIDATE_DCACHE);\r
+ }\r
}\r
\r
// seek..\r
\r
shared_ctl->mp3_offs = byte_offs;\r
\r
- // reset buffer pointers..\r
+ // reset buffer pointers and stuff..\r
mp3_samples_ready = mp3_buffer_offs = mp3_play_bufsel = 0;\r
+ mp3_job_started = 0;\r
shared_ctl->mp3_buffsel = 1; // will change to 0 on first decode\r
+\r
+ if (!(PicoOpt&0x200)) mp3_start_local();\r
}\r
\r
\r
int cdda_on;\r
\r
cdda_on = (PicoMCD & 1) && (PicoOpt&0x800) && !(Pico_mcd->s68k_regs[0x36] & 1) &&\r
- (Pico_mcd->scd.Status_CDC & 1) && loaded_mp3 && shared_ctl->mp3_offs < shared_ctl->mp3_len;\r
+ (Pico_mcd->scd.Status_CDC & 1) && loaded_mp3;\r
\r
if (cdda_on) {\r
offs1024 = shared_ctl->mp3_offs << 7;\r
YM2612 *ym2612_940;\r
\r
// from init.s\r
-void wait_irq(void);\r
+int wait_get_job(int oldjob);\r
void spend_cycles(int c);\r
-void cache_clean(void);\r
-void cache_clean_flush(void);\r
+void dcache_clean(void);\r
+void dcache_clean_flush(void);\r
+void drain_wb(void);\r
// this should help to resolve race confition where shared var\r
// is changed by other core just before we update it\r
void set_if_not_changed(int *val, int oldval, int newval);\r
shared_data->mp3_buffer[shared_ctl->mp3_buffsel], 0);\r
if (err) {\r
if (err == ERR_MP3_INDATA_UNDERFLOW) {\r
- shared_ctl->mp3_offs = shared_ctl->mp3_len; // EOF\r
set_if_not_changed(&shared_ctl->mp3_offs, mp3_offs, shared_ctl->mp3_len);\r
return;\r
} else if (err <= -6 && err >= -12) {\r
}\r
\r
\r
-void Main940(int startvector, int pc_at_irq)\r
+void Main940(void)\r
{\r
- int mix_buffer = shared_data->mix_buffer;\r
+ int *ym_buffer = shared_data->ym_buffer;\r
+ int job = 0;\r
ym2612_940 = &shared_data->ym2612;\r
\r
- // debug\r
- shared_ctl->vstarts[startvector]++;\r
- shared_ctl->last_irq_pc = pc_at_irq;\r
- // asm volatile ("mcr p15, 0, r0, c7, c10, 4" ::: "r0");\r
\r
-\r
-// for (;;)\r
+ for (;;)\r
{\r
- int job_num = 0;\r
-/*\r
- while (!shared_ctl->busy)\r
- {\r
- //shared_ctl->waitc++;\r
- spend_cycles(8*1024);\r
- }\r
-*/\r
-/*\r
- if (!shared_ctl->busy)\r
- {\r
- wait_irq();\r
- }\r
- shared_ctl->lastbusy = shared_ctl->busy;\r
-*/\r
+ job = wait_get_job(job);\r
+\r
+ shared_ctl->lastjob = job;\r
\r
- for (job_num = 0; job_num < MAX_940JOBS; job_num++)\r
+ switch (job)\r
{\r
- shared_ctl->lastjob = (job_num << 8) | shared_ctl->jobs[job_num];\r
-\r
- switch (shared_ctl->jobs[job_num])\r
- {\r
- case JOB940_INITALL:\r
- /* ym2612 */\r
- shared_ctl->writebuff0[0] = shared_ctl->writebuff1[0] = 0xffff;\r
- YM2612Init_(shared_ctl->baseclock, shared_ctl->rate);\r
- /* Helix mp3 decoder */\r
- shared_data->mp3dec = MP3InitDecoder();\r
- break;\r
-\r
- case JOB940_YM2612RESETCHIP:\r
- YM2612ResetChip_();\r
- break;\r
-\r
- case JOB940_PICOSTATELOAD:\r
- YM2612PicoStateLoad_();\r
- break;\r
-\r
- case JOB940_YM2612UPDATEONE: {\r
- int i, dw, *wbuff;\r
- if (shared_ctl->writebuffsel == 1) {\r
- wbuff = (int *) shared_ctl->writebuff1;\r
- } else {\r
- wbuff = (int *) shared_ctl->writebuff0;\r
- }\r
-\r
- /* playback all writes */\r
- for (i = 2048/2; i > 0; i--) {\r
- UINT16 d;\r
- dw = *wbuff++;\r
- d = dw;\r
- if (d == 0xffff) break;\r
- YM2612Write_(d >> 8, d);\r
- d = (dw>>16);\r
- if (d == 0xffff) break;\r
- YM2612Write_(d >> 8, d);\r
- }\r
-\r
- YM2612UpdateOne_(mix_buffer, shared_ctl->length, shared_ctl->stereo, 1);\r
- break;\r
+ case JOB940_INITALL:\r
+ /* ym2612 */\r
+ shared_ctl->writebuff0[0] = shared_ctl->writebuff1[0] = 0xffff;\r
+ YM2612Init_(shared_ctl->baseclock, shared_ctl->rate);\r
+ /* Helix mp3 decoder */\r
+ shared_data->mp3dec = MP3InitDecoder();\r
+ break;\r
+\r
+ case JOB940_INVALIDATE_DCACHE:\r
+ drain_wb();\r
+ dcache_clean_flush();\r
+ break;\r
+\r
+ case JOB940_YM2612RESETCHIP:\r
+ YM2612ResetChip_();\r
+ break;\r
+\r
+ case JOB940_PICOSTATELOAD:\r
+ YM2612PicoStateLoad_();\r
+ break;\r
+\r
+ case JOB940_YM2612UPDATEONE: {\r
+ int i, dw, *wbuff;\r
+ if (shared_ctl->writebuffsel == 1) {\r
+ wbuff = (int *) shared_ctl->writebuff1;\r
+ } else {\r
+ wbuff = (int *) shared_ctl->writebuff0;\r
}\r
\r
- case JOB940_MP3DECODE:\r
- mp3_decode();\r
- break;\r
+ /* playback all writes */\r
+ for (i = 2048/2; i > 0; i--) {\r
+ UINT16 d;\r
+ dw = *wbuff++;\r
+ d = dw;\r
+ if (d == 0xffff) break;\r
+ YM2612Write_(d >> 8, d);\r
+ d = (dw>>16);\r
+ if (d == 0xffff) break;\r
+ YM2612Write_(d >> 8, d);\r
+ }\r
+\r
+ YM2612UpdateOne_(ym_buffer, shared_ctl->length, shared_ctl->stereo, 1);\r
+ break;\r
}\r
- }\r
\r
- cache_clean();\r
-// asm volatile ("mov r0, #0" ::: "r0");\r
-// asm volatile ("mcr p15, 0, r0, c7, c10, 4" ::: "r0"); /* drain write buffer, should be done on nonbuffered write */\r
-// cache_clean_flush();\r
+ case JOB940_MP3DECODE:\r
+ mp3_decode();\r
+ break;\r
+ }\r
\r
shared_ctl->loopc++;\r
-\r
-// // shared_ctl->busy = 0; // shared mem is not reliable?\r
-\r
- wait_irq();\r
+ dcache_clean();\r
}\r
}\r
\r
-.global code940\r
+@ vim:filetype=armasm:\r
+\r
\r
.equ mmsp2_regs, (0xc0000000-0x02000000) @ assume we live @ 0x2000000 bank\r
+.equ shared_ctl, 0x00200000 @ this is where shared_ctl struncture is located\r
+\r
\r
-code940: @ interrupt table:\r
+@ exception table:\r
+.global code940\r
+code940:\r
b .b_reset @ reset\r
b .b_undef @ undefined instructions\r
b .b_swi @ software interrupt\r
mov r12, #5\r
b .Begin\r
.b_irq:\r
- mov r12, #6\r
mov sp, #0x100000 @ reset stack\r
sub sp, sp, #4\r
- mov r1, #mmsp2_regs\r
- orr r2, r1, #0x3B00\r
- orr r2, r2, #0x0046\r
- mvn r3, #0\r
- strh r3, [r2] @ clear any pending interrupts from the DUALCPU unit\r
- orr r2, r1, #0x4500\r
- str r3, [r2] @ clear all pending interrupts in irq controller's SRCPND register\r
- orr r2, r2, #0x0010\r
- str r3, [r2] @ clear all pending interrupts in irq controller's INTPND register\r
- b .Enter\r
+ mov r0, #shared_ctl @ remember where we were when interrupt happened\r
+ add r0, r0, #0x20\r
+ str lr, [r0]\r
+ mov r0, #shared_ctl @ increment exception counter (for debug)\r
+ add r0, r0, #(6*4)\r
+ ldr r1, [r0]\r
+ add r1, r1, #1\r
+ str r1, [r0]\r
+\r
+ bl Main940\r
+\r
+ @ we should never get here\r
+ b .b_reserved\r
+\r
+\r
.b_fiq:\r
mov r12, #7\r
b .Begin\r
mcr p15, 0, r0, c6, c4, 0\r
mcr p15, 0, r0, c6, c4, 1\r
\r
- @ set regions 1 and 4 to be cacheable (so the first 2M and mp3 area will be cacheable)\r
- mov r0, #(1<<1)|(1<<4)\r
+ @ region 5: 4K 0x00000000-0x00001000 (boot code protection region)\r
+ mov r0, #(0x0b<<1)|1\r
+ mcr p15, 0, r0, c6, c5, 0\r
+ mcr p15, 0, r0, c6, c5, 1\r
+\r
+ @ set regions 1, 4 and 5 to be cacheable (so the first 2M and mp3 area will be cacheable)\r
+ mov r0, #(1<<1)|(1<<4)|(1<<5)\r
mcr p15, 0, r0, c2, c0, 0\r
mcr p15, 0, r0, c2, c0, 1\r
\r
mov r0, #(1<<1)\r
mcr p15, 0, r0, c3, c0, 0\r
\r
- @ set protection, allow access only to regions 1 and 2\r
- mov r0, #(3<<8)|(3<<6)|(3<<4)|(3<<2)|(0) @ data: [full, full, full, full, no access] for regions [4 3 2 1 0]\r
+ @ set access protection\r
+ @ data: [no, full, full, full, full, no access] for regions [5 4 3 2 1 0]\r
+ mov r0, #(0<<10)|(3<<8)|(3<<6)|(3<<4)|(3<<2)|(0)\r
mcr p15, 0, r0, c5, c0, 0\r
- mov r0, #(0<<8)|(0<<6)|(0<<4)|(3<<2)|(0) @ instructions: [no access, no, no, full, no]\r
+ @ instructions: [full, no access, no, no, full, no]\r
+ mov r0, #(0<< 6)|(0<<4)|(3<<2)|(0)\r
+ orr r0, r0, #(3<<10)|(0<<8)\r
mcr p15, 0, r0, c5, c0, 1\r
\r
mrc p15, 0, r0, c1, c0, 0 @ fetch current control reg\r
mov r0, #0\r
mcr p15, 0, r0, c7, c6, 0\r
\r
-.Enter:\r
- mov r0, r12\r
- mov r1, lr\r
- bl Main940\r
+ @ remember which exception vector we came from (increment counter for debug)\r
+ mov r0, #shared_ctl\r
+ add r0, r0, r12, lsl #2\r
+ ldr r1, [r0]\r
+ add r1, r1, #1\r
+ str r1, [r0]\r
+ \r
+ @ remember last lr (for debug)\r
+ mov r0, #shared_ctl\r
+ add r0, r0, #0x20\r
+ str lr, [r0]\r
+\r
+ @ ready to take first job-interrupt\r
+wait_for_irq:\r
+ mrs r0, cpsr\r
+ bic r0, r0, #0x80\r
+ msr cpsr_c, r0 @ enable interrupts\r
+\r
+ mov r0, #0\r
+ mcr p15, 0, r0, c7, c0, 4 @ wait for IRQ\r
+@ mcr p15, 0, r0, c15, c8, 2\r
+ nop\r
+ nop\r
+ b .b_reserved\r
+\r
\r
- @ we should never get here\r
-@.b_deadloop:\r
-@ b .b_deadloop\r
- b .b_reserved\r
\r
+@ next job getter\r
+.global wait_get_job @ int oldjob\r
\r
+wait_get_job:\r
+ mov r3, #mmsp2_regs\r
+ orr r2, r3, #0x3B00\r
+ orr r2, r2, #0x0046 @ DUALPEND940 register\r
+ ldrh r12,[r2]\r
\r
-@ so asm utils are also defined here:\r
+ tst r0, r0\r
+ beq wgj_no_old\r
+ sub r0, r0, #1\r
+ mov r1, #1\r
+ mov r1, r1, lsl r0\r
+ strh r1, [r2] @ clear finished job's pending bit\r
+ bic r12,r12,r1\r
+\r
+wgj_no_old:\r
+ tst r12,r12\r
+ beq wgj_no_jobs\r
+ mov r0, #0\r
+wgj_loop:\r
+ add r0, r0, #1\r
+ movs r12,r12,lsr #1\r
+ bxcs lr\r
+ b wgj_loop\r
+\r
+wgj_no_jobs:\r
+ mvn r0, #0\r
+ orr r2, r3, #0x4500\r
+ str r0, [r2] @ clear all pending interrupts in irq controller's SRCPND register\r
+ orr r2, r2, #0x0010\r
+ str r0, [r2] @ clear all pending interrupts in irq controller's INTPND register\r
+ b wait_for_irq\r
+\r
+.pool\r
+\r
+\r
+\r
+\r
+@ some asm utils are also defined here:\r
.global spend_cycles @ c\r
\r
spend_cycles:\r
\r
\r
@ clean-flush function from ARM940T technical reference manual\r
-.global cache_clean_flush\r
+.global dcache_clean_flush\r
\r
-cache_clean_flush:\r
+dcache_clean_flush:\r
mov r1, #0 @ init line counter\r
ccf_outer_loop:\r
mov r0, #0 @ segment counter\r
ccf_inner_loop:\r
orr r2, r1, r0 @ make segment and line address\r
mcr p15, 0, r2, c7, c14, 2 @ clean and flush that line\r
- add r0, r0, #0x10 @ incremet secment counter\r
+ add r0, r0, #0x10 @ incremet segment counter\r
cmp r0, #0x40 @ complete all 4 segments?\r
bne ccf_inner_loop\r
add r1, r1, #0x04000000 @ increment line counter\r
bx lr\r
\r
\r
+\r
@ clean-only version\r
-.global cache_clean\r
+.global dcache_clean\r
\r
-cache_clean:\r
+dcache_clean:\r
mov r1, #0 @ init line counter\r
cf_outer_loop:\r
mov r0, #0 @ segment counter\r
cf_inner_loop:\r
orr r2, r1, r0 @ make segment and line address\r
mcr p15, 0, r2, c7, c10, 2 @ clean that line\r
- add r0, r0, #0x10 @ incremet secment counter\r
+ add r0, r0, #0x10 @ incremet segment counter\r
cmp r0, #0x40 @ complete all 4 segments?\r
bne cf_inner_loop\r
add r1, r1, #0x04000000 @ increment line counter\r
bx lr\r
\r
\r
-.global wait_irq\r
-\r
-wait_irq:\r
- mov r0, #mmsp2_regs\r
- orr r0, r0, #0x3B00\r
- orr r1, r0, #0x0042\r
- mov r3, #0\r
- strh r3, [r1] @ disable interrupts\r
- orr r2, r0, #0x003E\r
- strh r3, [r2] @ remove busy flag\r
- mov r3, #1\r
- strh r3, [r1] @ enable interrupts\r
-\r
- mrs r0, cpsr\r
- bic r0, r0, #0x80\r
- msr cpsr_c, r0 @ enable interrupts\r
+@ drain write buffer\r
+.global drain_wb\r
\r
+drain_wb:\r
mov r0, #0\r
- mcr p15, 0, r0, c7, c0, 4 @ wait for IRQ\r
-@ mcr p15, 0, r0, c15, c8, 2\r
- nop\r
- nop\r
- b .b_reserved\r
-\r
-.pool\r
+ mcr p15, 0, r0, c7, c10, 4\r
+ bx lr\r
\r
\r
.global set_if_not_changed @ int *val, int oldval, int newval\r
strne r3, [r0] @ restore value which was changed there by other core\r
bx lr\r
\r
-@ vim:filetype=armasm:\r
+\r
+\r
+@ pad the protected region.\r
+.rept 1024\r
+.long 0\r
+.endr\r
+\r
#include "../../../Pico/sound/ym2612.h"\r
#include "../helix/pub/mp3dec.h"\r
\r
+// max 16 jobs\r
enum _940_job_t {\r
JOB940_INITALL = 1,\r
+ JOB940_INVALIDATE_DCACHE,\r
JOB940_YM2612RESETCHIP,\r
JOB940_YM2612UPDATEONE,\r
- JOB940_PICOSTATELOAD,\r
JOB940_MP3DECODE,\r
- JOB940_NUMJOBS\r
+ JOB940_PICOSTATELOAD,\r
};\r
\r
-#define MAX_940JOBS 2\r
+//#define MAX_940JOBS 2\r
\r
typedef struct\r
{\r
- YM2612 ym2612; /* current state of the emulated YM2612 */\r
+ YM2612 ym2612; /* current state of the emulated YM2612 */\r
HMP3Decoder mp3dec; /* mp3 decoder's handle */\r
- int mix_buffer[44100/50*2]; /* this is where the YM2612 samples will be mixed to */\r
+ int ym_buffer[44100/50*2]; /* this is where the YM2612 samples will be mixed to */\r
short mp3_buffer[2][1152*2]; /* buffers for mp3 decoder's output */\r
} _940_data_t;\r
\r
\r
typedef struct\r
{\r
- int jobs[MAX_940JOBS]; /* jobs for second core */\r
- int busy_; /* unused */\r
+ int vstarts[8]; /* debug: 00: number of starts from each of 8 vectors */\r
+ int last_lr; /* debug: 20: last exception's lr */\r
+// int jobs[MAX_940JOBS]; /* jobs for second core */\r
+// int busy_; /* unused */\r
int length; /* number of samples to mix (882 max) */\r
int stereo; /* mix samples as stereo, doubles sample count automatically */\r
int baseclock; /* ym2612 settings */\r
int mp3_len; /* data len of loaded mp3 */\r
int mp3_offs; /* current playback offset (just after last decoded frame) */\r
int mp3_buffsel; /* which output buffer to decode to */\r
- int vstarts[8]; /* debug: number of starts from each of 8 vectors */\r
- int loopc; /* debug: main loop counter */\r
+ int loopc; /* debug: main loop counter */\r
int mp3_errors; /* debug: mp3 decoder's error counter */\r
int mp3_lasterr; /* debug: mp3 decoder's last error */\r
- int last_irq_pc; /* debug: PC value when IRQ happened */\r
int lastjob; /* debug: last job id */\r
- int lastbusy; /* debug: */\r
} _940_ctl_t;\r
--- /dev/null
+#include <stdio.h>
+#include <string.h>
+
+#include "../../Pico/sound/mix.h"
+#include "code940/940shared.h"
+#include "helix/pub/mp3dec.h"
+
+static short mp3_out_buffer[2*1152];
+static HMP3Decoder mp3dec = 0;
+static int mp3_buffer_offs = 0;
+
+extern _940_ctl_t *shared_ctl;
+extern unsigned char *mp3_mem;
+extern int PsndRate;
+
+
+static int try_get_header(unsigned char *buff, MP3FrameInfo *fi)
+{
+ int ret, offs1, offs = 0;
+
+ while (1)
+ {
+ offs1 = MP3FindSyncWord(buff + offs, 2048 - offs);
+ if (offs1 < 0) return -2;
+ offs += offs1;
+ if (2048 - offs < 4) return -3;
+
+ // printf("trying header %08x\n", *(int *)(buff + offs));
+
+ ret = MP3GetNextFrameInfo(mp3dec, fi, buff + offs);
+ if (ret == 0 && fi->bitrate != 0) break;
+ offs++;
+ }
+
+ return ret;
+}
+
+int mp3_get_bitrate(FILE *f, int len)
+{
+ unsigned char buff[2048];
+ MP3FrameInfo fi;
+ int ret;
+
+ memset(buff, 0, 2048);
+
+ if (!mp3dec) mp3dec = MP3InitDecoder();
+
+ fseek(f, 0, SEEK_SET);
+ ret = fread(buff, 1, 2048, f);
+ fseek(f, 0, SEEK_SET);
+ if (ret <= 0) return -1;
+
+ ret = try_get_header(buff, &fi);
+ if (ret != 0 || fi.bitrate == 0) {
+ // try to read somewhere around the middle
+ fseek(f, len>>1, SEEK_SET);
+ fread(buff, 1, 2048, f);
+ fseek(f, 0, SEEK_SET);
+ ret = try_get_header(buff, &fi);
+ }
+ if (ret != 0) return ret;
+
+ // printf("bitrate: %i\n", fi.bitrate / 1000);
+
+ return fi.bitrate / 1000;
+}
+
+
+static void mp3_decode(void)
+{
+ // tried copying this to cached mem, no improvement noticed
+ int mp3_offs = shared_ctl->mp3_offs;
+ unsigned char *readPtr = mp3_mem + mp3_offs;
+ int bytesLeft = shared_ctl->mp3_len - mp3_offs;
+ int offset; // frame offset from readPtr
+ int err;
+
+ if (bytesLeft <= 0) return; // EOF, nothing to do
+
+ offset = MP3FindSyncWord(readPtr, bytesLeft);
+ if (offset < 0) {
+ shared_ctl->mp3_offs = shared_ctl->mp3_len;
+ return; // EOF
+ }
+ readPtr += offset;
+ bytesLeft -= offset;
+
+ err = MP3Decode(mp3dec, &readPtr, &bytesLeft, mp3_out_buffer, 0);
+ if (err) {
+ if (err == ERR_MP3_INDATA_UNDERFLOW) {
+ shared_ctl->mp3_offs = shared_ctl->mp3_len; // EOF
+ return;
+ } else if (err <= -6 && err >= -12) {
+ // ERR_MP3_INVALID_FRAMEHEADER, ERR_MP3_INVALID_*
+ // just try to skip the offending frame..
+ readPtr++;
+ }
+ shared_ctl->mp3_errors++;
+ shared_ctl->mp3_lasterr = err;
+ }
+ shared_ctl->mp3_offs = readPtr - mp3_mem;
+}
+
+
+void mp3_update_local(int *buffer, int length, int stereo)
+{
+ int length_mp3, shr = 0;
+ void (*mix_samples)(int *dest_buf, short *mp3_buf, int count) = mix_16h_to_32;
+
+ length_mp3 = length;
+ if (PsndRate == 22050) { mix_samples = mix_16h_to_32_s1; length_mp3 <<= 1; shr = 1; }
+ else if (PsndRate == 11025) { mix_samples = mix_16h_to_32_s2; length_mp3 <<= 2; shr = 2; }
+
+ if (1152 - mp3_buffer_offs >= length_mp3) {
+ mix_samples(buffer, mp3_out_buffer + mp3_buffer_offs*2, length<<1);
+
+ mp3_buffer_offs += length_mp3;
+ } else {
+ int left = 1152 - mp3_buffer_offs;
+
+ mix_samples(buffer, mp3_out_buffer + mp3_buffer_offs*2, (left>>shr)<<1);
+ mp3_decode();
+ mp3_buffer_offs = length_mp3 - left;
+ mix_samples(buffer + ((left>>shr)<<1), mp3_out_buffer, (mp3_buffer_offs>>shr)<<1);
+ }
+}
+
+
+void mp3_start_local(void)
+{
+ mp3_buffer_offs = 0;
+ mp3_decode();
+}
+
--- /dev/null
+
+void mp3_update_local(int *buffer, int length, int stereo);
+void mp3_start_local(void);
+
notaz.gp2x.de / picodrive.git / commitdiff