diff options
Diffstat (limited to 'lib/rbcodec/codecs/libspc')
-rw-r--r-- | lib/rbcodec/codecs/libspc/SOURCES | 4 | ||||
-rw-r--r-- | lib/rbcodec/codecs/libspc/libspc.make | 18 | ||||
-rw-r--r-- | lib/rbcodec/codecs/libspc/spc_codec.h | 491 | ||||
-rw-r--r-- | lib/rbcodec/codecs/libspc/spc_cpu.c | 1049 | ||||
-rw-r--r-- | lib/rbcodec/codecs/libspc/spc_dsp.c | 1594 | ||||
-rw-r--r-- | lib/rbcodec/codecs/libspc/spc_emu.c | 397 | ||||
-rw-r--r-- | lib/rbcodec/codecs/libspc/spc_profiler.c | 66 | ||||
-rw-r--r-- | lib/rbcodec/codecs/libspc/spc_profiler.h | 72 |
8 files changed, 3691 insertions, 0 deletions
diff --git a/lib/rbcodec/codecs/libspc/SOURCES b/lib/rbcodec/codecs/libspc/SOURCES new file mode 100644 index 0000000000..901232a6eb --- /dev/null +++ b/lib/rbcodec/codecs/libspc/SOURCES | |||
@@ -0,0 +1,4 @@ | |||
1 | spc_cpu.c | ||
2 | spc_dsp.c | ||
3 | spc_emu.c | ||
4 | spc_profiler.c | ||
diff --git a/lib/rbcodec/codecs/libspc/libspc.make b/lib/rbcodec/codecs/libspc/libspc.make new file mode 100644 index 0000000000..a005f7914a --- /dev/null +++ b/lib/rbcodec/codecs/libspc/libspc.make | |||
@@ -0,0 +1,18 @@ | |||
1 | # __________ __ ___. | ||
2 | # Open \______ \ ____ ____ | | _\_ |__ _______ ___ | ||
3 | # Source | _// _ \_/ ___\| |/ /| __ \ / _ \ \/ / | ||
4 | # Jukebox | | ( <_> ) \___| < | \_\ ( <_> > < < | ||
5 | # Firmware |____|_ /\____/ \___ >__|_ \|___ /\____/__/\_ \ | ||
6 | # \/ \/ \/ \/ \/ | ||
7 | # $Id$ | ||
8 | # | ||
9 | |||
10 | # libspc | ||
11 | SPCLIB := $(CODECDIR)/libspc.a | ||
12 | SPCLIB_SRC := $(call preprocess, $(RBCODECLIB_DIR)/codecs/libspc/SOURCES) | ||
13 | SPCLIB_OBJ := $(call c2obj, $(SPCLIB_SRC)) | ||
14 | OTHER_SRC += $(SPCLIB_SRC) | ||
15 | |||
16 | $(SPCLIB): $(SPCLIB_OBJ) | ||
17 | $(SILENT)$(shell rm -f $@) | ||
18 | $(call PRINTS,AR $(@F))$(AR) rcs $@ $^ >/dev/null | ||
diff --git a/lib/rbcodec/codecs/libspc/spc_codec.h b/lib/rbcodec/codecs/libspc/spc_codec.h new file mode 100644 index 0000000000..7f6b6e2e9f --- /dev/null +++ b/lib/rbcodec/codecs/libspc/spc_codec.h | |||
@@ -0,0 +1,491 @@ | |||
1 | /*************************************************************************** | ||
2 | * __________ __ ___. | ||
3 | * Open \______ \ ____ ____ | | _\_ |__ _______ ___ | ||
4 | * Source | _// _ \_/ ___\| |/ /| __ \ / _ \ \/ / | ||
5 | * Jukebox | | ( <_> ) \___| < | \_\ ( <_> > < < | ||
6 | * Firmware |____|_ /\____/ \___ >__|_ \|___ /\____/__/\_ \ | ||
7 | * \/ \/ \/ \/ \/ | ||
8 | * $Id$ | ||
9 | * | ||
10 | * Copyright (C) 2007-2008 Michael Sevakis (jhMikeS) | ||
11 | * Copyright (C) 2006-2007 Adam Gashlin (hcs) | ||
12 | * Copyright (C) 2004-2007 Shay Green (blargg) | ||
13 | * Copyright (C) 2002 Brad Martin | ||
14 | * | ||
15 | * This program is free software; you can redistribute it and/or | ||
16 | * modify it under the terms of the GNU General Public License | ||
17 | * as published by the Free Software Foundation; either version 2 | ||
18 | * of the License, or (at your option) any later version. | ||
19 | * | ||
20 | * This software is distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY | ||
21 | * KIND, either express or implied. | ||
22 | * | ||
23 | ****************************************************************************/ | ||
24 | |||
25 | /* lovingly ripped off from Game_Music_Emu 0.5.2. http://www.slack.net/~ant/ */ | ||
26 | /* DSP Based on Brad Martin's OpenSPC DSP emulator */ | ||
27 | /* tag reading from sexyspc by John Brawn (John_Brawn@yahoo.com) and others */ | ||
28 | |||
29 | #ifndef _SPC_CODEC_H_ | ||
30 | #define _SPC_CODEC_H_ | ||
31 | |||
32 | /* rather than comment out asserts, just define NDEBUG */ | ||
33 | #ifndef NDEBUG | ||
34 | #define NDEBUG | ||
35 | #endif | ||
36 | #include <assert.h> | ||
37 | |||
38 | /** Basic configuration options **/ | ||
39 | |||
40 | #ifndef ARM_ARCH | ||
41 | #define ARM_ARCH 0 | ||
42 | #endif | ||
43 | |||
44 | #define SPC_DUAL_CORE 1 | ||
45 | |||
46 | #if !defined(SPC_DUAL_CORE) || NUM_CORES == 1 | ||
47 | #undef SPC_DUAL_CORE | ||
48 | #define SPC_DUAL_CORE 0 | ||
49 | #endif | ||
50 | |||
51 | /* Only some targets are fast enough for gaussian and realtime BRR decode */ | ||
52 | #if CONFIG_CPU == S3C2440 || CONFIG_CPU == IMX31L || \ | ||
53 | CONFIG_CPU == AS3525 || CONFIG_CPU == AS3525v2 || \ | ||
54 | defined(CPU_S5L870X) || \ | ||
55 | (CONFIG_PLATFORM & PLATFORM_HOSTED) || MEMORYSIZE <= 2 | ||
56 | /* Don't cache BRR waves */ | ||
57 | #define SPC_BRRCACHE 0 | ||
58 | |||
59 | /* Allow gaussian interpolation */ | ||
60 | #define SPC_NOINTERP 0 | ||
61 | |||
62 | /* Allow echo processing */ | ||
63 | #define SPC_NOECHO 0 | ||
64 | #elif defined(CPU_COLDFIRE) | ||
65 | /* Cache BRR waves */ | ||
66 | #define SPC_BRRCACHE 1 | ||
67 | |||
68 | /* Disable gaussian interpolation */ | ||
69 | #define SPC_NOINTERP 1 | ||
70 | |||
71 | /* Allow echo processing */ | ||
72 | #define SPC_NOECHO 0 | ||
73 | #elif defined (CPU_PP) && SPC_DUAL_CORE | ||
74 | /* Cache BRR waves */ | ||
75 | #define SPC_BRRCACHE 1 | ||
76 | |||
77 | /* Disable gaussian interpolation */ | ||
78 | #define SPC_NOINTERP 1 | ||
79 | |||
80 | /* Allow echo processing */ | ||
81 | #define SPC_NOECHO 0 | ||
82 | #else | ||
83 | /* Cache BRR waves */ | ||
84 | #define SPC_BRRCACHE 1 | ||
85 | |||
86 | /* Disable gaussian interpolation */ | ||
87 | #define SPC_NOINTERP 1 | ||
88 | |||
89 | /* Disable echo processing */ | ||
90 | #define SPC_NOECHO 1 | ||
91 | #endif | ||
92 | |||
93 | #if (CONFIG_CPU == MCF5250) | ||
94 | #define IBSS_ATTR_SPC IBSS_ATTR | ||
95 | #define ICODE_ATTR_SPC ICODE_ATTR | ||
96 | #define ICONST_ATTR_SPC ICONST_ATTR | ||
97 | /* Not enough IRAM available to move further data to it. */ | ||
98 | #define IBSS_ATTR_SPC_LARGE_IRAM | ||
99 | |||
100 | #elif (CONFIG_CPU == PP5020) | ||
101 | /* spc is slower on PP5020 when moving data to IRAM. */ | ||
102 | #define IBSS_ATTR_SPC | ||
103 | #define ICODE_ATTR_SPC | ||
104 | #define ICONST_ATTR_SPC | ||
105 | /* Not enough IRAM available to move further data to it. */ | ||
106 | #define IBSS_ATTR_SPC_LARGE_IRAM | ||
107 | |||
108 | #elif (CONFIG_CPU == PP5022) || (CONFIG_CPU == PP5024) | ||
109 | #define IBSS_ATTR_SPC IBSS_ATTR | ||
110 | #define ICODE_ATTR_SPC ICODE_ATTR | ||
111 | #define ICONST_ATTR_SPC ICONST_ATTR | ||
112 | /* Not enough IRAM available to move further data to it. */ | ||
113 | #define IBSS_ATTR_SPC_LARGE_IRAM | ||
114 | |||
115 | #elif defined(CPU_S5L870X) | ||
116 | #define IBSS_ATTR_SPC IBSS_ATTR | ||
117 | #define ICODE_ATTR_SPC ICODE_ATTR | ||
118 | #define ICONST_ATTR_SPC ICONST_ATTR | ||
119 | /* Very large IRAM. Move even more data to it. */ | ||
120 | #define IBSS_ATTR_SPC_LARGE_IRAM IBSS_ATTR | ||
121 | |||
122 | #else | ||
123 | #define IBSS_ATTR_SPC IBSS_ATTR | ||
124 | #define ICODE_ATTR_SPC ICODE_ATTR | ||
125 | #define ICONST_ATTR_SPC ICONST_ATTR | ||
126 | /* Not enough IRAM available to move further data to it. */ | ||
127 | #define IBSS_ATTR_SPC_LARGE_IRAM | ||
128 | #endif | ||
129 | |||
130 | #if SPC_DUAL_CORE | ||
131 | #undef SHAREDBSS_ATTR | ||
132 | #define SHAREDBSS_ATTR __attribute__ ((section(".ibss"))) | ||
133 | #undef SHAREDDATA_ATTR | ||
134 | #define SHAREDDATA_ATTR __attribute__((section(".idata"))) | ||
135 | #endif | ||
136 | |||
137 | /* Samples per channel per iteration */ | ||
138 | #if defined(CPU_PP) && NUM_CORES == 1 | ||
139 | #define WAV_CHUNK_SIZE 2048 | ||
140 | #else | ||
141 | #define WAV_CHUNK_SIZE 1024 | ||
142 | #endif | ||
143 | |||
144 | /**************** Little-endian handling ****************/ | ||
145 | |||
146 | static inline unsigned get_le16( void const* p ) | ||
147 | { | ||
148 | return ((unsigned char const*) p) [1] * 0x100u + | ||
149 | ((unsigned char const*) p) [0]; | ||
150 | } | ||
151 | |||
152 | static inline int get_le16s( void const* p ) | ||
153 | { | ||
154 | return ((signed char const*) p) [1] * 0x100 + | ||
155 | ((unsigned char const*) p) [0]; | ||
156 | } | ||
157 | |||
158 | static inline void set_le16( void* p, unsigned n ) | ||
159 | { | ||
160 | ((unsigned char*) p) [1] = (unsigned char) (n >> 8); | ||
161 | ((unsigned char*) p) [0] = (unsigned char) n; | ||
162 | } | ||
163 | |||
164 | #define GET_LE16( addr ) get_le16( addr ) | ||
165 | #define GET_LE16A( addr ) get_le16( addr ) | ||
166 | #define SET_LE16( addr, data ) set_le16( addr, data ) | ||
167 | #define INT16A( addr ) (*(uint16_t*) (addr)) | ||
168 | #define INT16SA( addr ) (*(int16_t*) (addr)) | ||
169 | |||
170 | #ifdef ROCKBOX_LITTLE_ENDIAN | ||
171 | #define GET_LE16SA( addr ) (*( int16_t*) (addr)) | ||
172 | #define SET_LE16A( addr, data ) (void) (*(uint16_t*) (addr) = (data)) | ||
173 | #else | ||
174 | #define GET_LE16SA( addr ) get_le16s( addr ) | ||
175 | #define SET_LE16A( addr, data ) set_le16 ( addr, data ) | ||
176 | #endif | ||
177 | |||
178 | struct Spc_Emu; | ||
179 | #define THIS struct Spc_Emu* const this | ||
180 | |||
181 | /* The CPU portion (shock!) */ | ||
182 | |||
183 | struct cpu_regs_t | ||
184 | { | ||
185 | long pc; /* more than 16 bits to allow overflow detection */ | ||
186 | uint8_t a; | ||
187 | uint8_t x; | ||
188 | uint8_t y; | ||
189 | uint8_t status; | ||
190 | uint8_t sp; | ||
191 | }; | ||
192 | |||
193 | struct src_dir | ||
194 | { | ||
195 | uint16_t start; | ||
196 | uint16_t loop; | ||
197 | }; | ||
198 | |||
199 | struct cpu_ram_t | ||
200 | { | ||
201 | union { | ||
202 | uint8_t padding1 [0x100]; | ||
203 | uint16_t align; | ||
204 | } padding1 [1]; | ||
205 | union { | ||
206 | uint8_t ram [0x10000]; | ||
207 | struct src_dir sd [0x10000/sizeof(struct src_dir)]; | ||
208 | }; | ||
209 | uint8_t padding2 [0x100]; | ||
210 | }; | ||
211 | |||
212 | #undef RAM | ||
213 | #define RAM ram.ram | ||
214 | extern struct cpu_ram_t ram; | ||
215 | |||
216 | long CPU_run( THIS, long start_time ) ICODE_ATTR_SPC; | ||
217 | void CPU_Init( THIS ); | ||
218 | |||
219 | /* The DSP portion (awe!) */ | ||
220 | enum { VOICE_COUNT = 8 }; | ||
221 | enum { REGISTER_COUNT = 128 }; | ||
222 | |||
223 | struct raw_voice_t | ||
224 | { | ||
225 | int8_t volume [2]; | ||
226 | uint8_t rate [2]; | ||
227 | uint8_t waveform; | ||
228 | uint8_t adsr [2]; /* envelope rates for attack, decay, and sustain */ | ||
229 | uint8_t gain; /* envelope gain (if not using ADSR) */ | ||
230 | int8_t envx; /* current envelope level */ | ||
231 | int8_t outx; /* current sample */ | ||
232 | int8_t unused [6]; | ||
233 | }; | ||
234 | |||
235 | struct globals_t | ||
236 | { | ||
237 | int8_t unused1 [12]; | ||
238 | int8_t volume_0; /* 0C Main Volume Left (-.7) */ | ||
239 | int8_t echo_feedback; /* 0D Echo Feedback (-.7) */ | ||
240 | int8_t unused2 [14]; | ||
241 | int8_t volume_1; /* 1C Main Volume Right (-.7) */ | ||
242 | int8_t unused3 [15]; | ||
243 | int8_t echo_volume_0; /* 2C Echo Volume Left (-.7) */ | ||
244 | uint8_t pitch_mods; /* 2D Pitch Modulation on/off for each voice */ | ||
245 | int8_t unused4 [14]; | ||
246 | int8_t echo_volume_1; /* 3C Echo Volume Right (-.7) */ | ||
247 | uint8_t noise_enables; /* 3D Noise output on/off for each voice */ | ||
248 | int8_t unused5 [14]; | ||
249 | uint8_t key_ons; /* 4C Key On for each voice */ | ||
250 | uint8_t echo_ons; /* 4D Echo on/off for each voice */ | ||
251 | int8_t unused6 [14]; | ||
252 | uint8_t key_offs; /* 5C key off for each voice | ||
253 | (instantiates release mode) */ | ||
254 | uint8_t wave_page; /* 5D source directory (wave table offsets) */ | ||
255 | int8_t unused7 [14]; | ||
256 | uint8_t flags; /* 6C flags and noise freq */ | ||
257 | uint8_t echo_page; /* 6D */ | ||
258 | int8_t unused8 [14]; | ||
259 | uint8_t wave_ended; /* 7C */ | ||
260 | uint8_t echo_delay; /* 7D ms >> 4 */ | ||
261 | char unused9 [2]; | ||
262 | }; | ||
263 | |||
264 | enum state_t | ||
265 | { /* -1, 0, +1 allows more efficient if statements */ | ||
266 | state_decay = -1, | ||
267 | state_sustain = 0, | ||
268 | state_attack = +1, | ||
269 | state_release = 2 | ||
270 | }; | ||
271 | |||
272 | struct cache_entry_t | ||
273 | { | ||
274 | int16_t const* samples; | ||
275 | unsigned end; /* past-the-end position */ | ||
276 | unsigned loop; /* number of samples in loop */ | ||
277 | unsigned start_addr; | ||
278 | }; | ||
279 | |||
280 | enum { BRR_BLOCK_SIZE = 16 }; | ||
281 | enum { BRR_CACHE_SIZE = 0x20000 + 32} ; | ||
282 | |||
283 | struct voice_t | ||
284 | { | ||
285 | #if SPC_BRRCACHE | ||
286 | int16_t const* samples; | ||
287 | long wave_end; | ||
288 | int wave_loop; | ||
289 | #else | ||
290 | int16_t samples [3 + BRR_BLOCK_SIZE + 1]; | ||
291 | int block_header; /* header byte from current block */ | ||
292 | #endif | ||
293 | uint8_t const* addr; | ||
294 | short volume [2]; | ||
295 | long position;/* position in samples buffer, with 12-bit fraction */ | ||
296 | short envx; | ||
297 | short env_mode; | ||
298 | short env_timer; | ||
299 | short key_on_delay; | ||
300 | }; | ||
301 | |||
302 | #if SPC_BRRCACHE | ||
303 | /* a little extra for samples that go past end */ | ||
304 | extern int16_t BRRcache [BRR_CACHE_SIZE]; | ||
305 | #endif | ||
306 | |||
307 | enum { FIR_BUF_HALF = 8 }; | ||
308 | |||
309 | #if defined(CPU_COLDFIRE) | ||
310 | /* global because of the large aligment requirement for hardware masking - | ||
311 | * L-R interleaved 16-bit samples for easy loading and mac.w use. | ||
312 | */ | ||
313 | enum | ||
314 | { | ||
315 | FIR_BUF_CNT = FIR_BUF_HALF, | ||
316 | FIR_BUF_SIZE = FIR_BUF_CNT * sizeof ( int32_t ), | ||
317 | FIR_BUF_ALIGN = FIR_BUF_SIZE * 2, | ||
318 | FIR_BUF_MASK = ~((FIR_BUF_ALIGN / 2) | (sizeof ( int32_t ) - 1)) | ||
319 | }; | ||
320 | #elif defined (CPU_ARM) | ||
321 | #if ARM_ARCH >= 6 | ||
322 | enum | ||
323 | { | ||
324 | FIR_BUF_CNT = FIR_BUF_HALF * 2, | ||
325 | FIR_BUF_SIZE = FIR_BUF_CNT * sizeof ( int32_t ), | ||
326 | FIR_BUF_ALIGN = FIR_BUF_SIZE, | ||
327 | FIR_BUF_MASK = ~((FIR_BUF_ALIGN / 2) | (sizeof ( int32_t ) - 1)) | ||
328 | }; | ||
329 | #else | ||
330 | enum | ||
331 | { | ||
332 | FIR_BUF_CNT = FIR_BUF_HALF * 2 * 2, | ||
333 | FIR_BUF_SIZE = FIR_BUF_CNT * sizeof ( int32_t ), | ||
334 | FIR_BUF_ALIGN = FIR_BUF_SIZE, | ||
335 | FIR_BUF_MASK = ~((FIR_BUF_ALIGN / 2) | (sizeof ( int32_t ) * 2 - 1)) | ||
336 | }; | ||
337 | #endif /* ARM_ARCH */ | ||
338 | #endif /* CPU_* */ | ||
339 | |||
340 | struct Spc_Dsp | ||
341 | { | ||
342 | union | ||
343 | { | ||
344 | struct raw_voice_t voice [VOICE_COUNT]; | ||
345 | uint8_t reg [REGISTER_COUNT]; | ||
346 | struct globals_t g; | ||
347 | int16_t align; | ||
348 | } r; | ||
349 | |||
350 | unsigned echo_pos; | ||
351 | int keys_down; | ||
352 | int noise_count; | ||
353 | uint16_t noise; /* also read as int16_t */ | ||
354 | |||
355 | #if defined(CPU_COLDFIRE) | ||
356 | /* FIR history is interleaved. Hardware handles wrapping by mask. | ||
357 | * |LR|LR|LR|LR|LR|LR|LR|LR| */ | ||
358 | int32_t *fir_ptr; | ||
359 | /* wrapped address just behind current position - | ||
360 | allows mac.w to increment and mask fir_ptr */ | ||
361 | int32_t *last_fir_ptr; | ||
362 | /* copy of echo FIR constants as int16_t for use with mac.w */ | ||
363 | int16_t fir_coeff [VOICE_COUNT]; | ||
364 | #elif defined (CPU_ARM) | ||
365 | /* fir_buf [i + 8] == fir_buf [i], to avoid wrap checking in FIR code */ | ||
366 | int32_t *fir_ptr; | ||
367 | #if ARM_ARCH >= 6 | ||
368 | /* FIR history is interleaved with guard to eliminate wrap checking | ||
369 | * when convolving. | ||
370 | * |LR|LR|LR|LR|LR|LR|LR|LR|--|--|--|--|--|--|--|--| */ | ||
371 | /* copy of echo FIR constants as int16_t, loaded as int32 for | ||
372 | * halfword, packed multiples */ | ||
373 | int16_t fir_coeff [VOICE_COUNT]; | ||
374 | #else | ||
375 | /* FIR history is interleaved with guard to eliminate wrap checking | ||
376 | * when convolving. | ||
377 | * |LL|RR|LL|RR|LL|RR|LL|RR|LL|RR|LL|RR|LL|RR|LL|RR|... | ||
378 | * |--|--|--|--|--|--|--|--|--|--|--|--|--|--|--|--| */ | ||
379 | /* copy of echo FIR constants as int32_t, for faster access */ | ||
380 | int32_t fir_coeff [VOICE_COUNT]; | ||
381 | #endif /* ARM_ARCH */ | ||
382 | #else /* Unoptimized CPU */ | ||
383 | /* fir_buf [i + 8] == fir_buf [i], to avoid wrap checking in FIR code */ | ||
384 | int fir_pos; /* (0 to 7) */ | ||
385 | int fir_buf [FIR_BUF_HALF * 2] [2]; | ||
386 | /* copy of echo FIR constants as int, for faster access */ | ||
387 | int fir_coeff [VOICE_COUNT]; | ||
388 | #endif | ||
389 | |||
390 | struct voice_t voice_state [VOICE_COUNT]; | ||
391 | |||
392 | #if SPC_BRRCACHE | ||
393 | uint8_t oldsize; | ||
394 | struct cache_entry_t wave_entry [256]; | ||
395 | struct cache_entry_t wave_entry_old [256]; | ||
396 | #endif | ||
397 | }; | ||
398 | |||
399 | void DSP_run_( struct Spc_Dsp* this, long count, int32_t* out_buf ) ICODE_ATTR_SPC; | ||
400 | void DSP_reset( struct Spc_Dsp* this ); | ||
401 | |||
402 | static inline void DSP_run( struct Spc_Dsp* this, long count, int32_t* out ) | ||
403 | { | ||
404 | /* Should we just fill the buffer with silence? Flags won't be cleared */ | ||
405 | /* during this run so it seems it should keep resetting every sample. */ | ||
406 | if ( this->r.g.flags & 0x80 ) | ||
407 | DSP_reset( this ); | ||
408 | |||
409 | DSP_run_( this, count, out ); | ||
410 | } | ||
411 | |||
412 | /**************** SPC emulator ****************/ | ||
413 | /* 1.024 MHz clock / 32000 samples per second */ | ||
414 | enum { CLOCKS_PER_SAMPLE = 32 }; | ||
415 | |||
416 | enum { EXTRA_CLOCKS = CLOCKS_PER_SAMPLE / 2 }; | ||
417 | |||
418 | /* using this disables timer (since this will always be in the future) */ | ||
419 | enum { TIMER_DISABLED_TIME = 127 }; | ||
420 | |||
421 | enum { ROM_SIZE = 64 }; | ||
422 | enum { ROM_ADDR = 0xFFC0 }; | ||
423 | |||
424 | enum { TIMER_COUNT = 3 }; | ||
425 | |||
426 | struct Timer | ||
427 | { | ||
428 | long next_tick; | ||
429 | int period; | ||
430 | int count; | ||
431 | int shift; | ||
432 | int enabled; | ||
433 | int counter; | ||
434 | }; | ||
435 | |||
436 | struct Spc_Emu | ||
437 | { | ||
438 | uint8_t cycle_table [0x100]; | ||
439 | struct cpu_regs_t r; | ||
440 | |||
441 | int32_t* sample_buf; | ||
442 | long next_dsp; | ||
443 | int rom_enabled; | ||
444 | int extra_cycles; | ||
445 | |||
446 | struct Timer timer [TIMER_COUNT]; | ||
447 | |||
448 | /* large objects at end */ | ||
449 | struct Spc_Dsp dsp; | ||
450 | uint8_t extra_ram [ROM_SIZE]; | ||
451 | uint8_t boot_rom [ROM_SIZE]; | ||
452 | }; | ||
453 | |||
454 | enum { SPC_FILE_SIZE = 0x10180 }; | ||
455 | |||
456 | struct spc_file_t | ||
457 | { | ||
458 | char signature [27]; | ||
459 | char unused [10]; | ||
460 | uint8_t pc [2]; | ||
461 | uint8_t a; | ||
462 | uint8_t x; | ||
463 | uint8_t y; | ||
464 | uint8_t status; | ||
465 | uint8_t sp; | ||
466 | char unused2 [212]; | ||
467 | uint8_t ram [0x10000]; | ||
468 | uint8_t dsp [128]; | ||
469 | uint8_t ipl_rom [128]; | ||
470 | }; | ||
471 | |||
472 | void SPC_Init( THIS ); | ||
473 | |||
474 | int SPC_load_spc( THIS, const void* data, long size ); | ||
475 | |||
476 | /**************** DSP interaction ****************/ | ||
477 | void DSP_write( struct Spc_Dsp* this, int i, int data ) ICODE_ATTR_SPC; | ||
478 | |||
479 | static inline int DSP_read( struct Spc_Dsp* this, int i ) | ||
480 | { | ||
481 | assert( (unsigned) i < REGISTER_COUNT ); | ||
482 | return this->r.reg [i]; | ||
483 | } | ||
484 | |||
485 | int SPC_read( THIS, unsigned addr, long const time ) ICODE_ATTR_SPC; | ||
486 | void SPC_write( THIS, unsigned addr, int data, long const time ) ICODE_ATTR_SPC; | ||
487 | |||
488 | /**************** Sample generation ****************/ | ||
489 | int SPC_play( THIS, long count, int32_t* out ) ICODE_ATTR_SPC; | ||
490 | |||
491 | #endif /* _SPC_CODEC_H_ */ | ||
diff --git a/lib/rbcodec/codecs/libspc/spc_cpu.c b/lib/rbcodec/codecs/libspc/spc_cpu.c new file mode 100644 index 0000000000..23dcc257de --- /dev/null +++ b/lib/rbcodec/codecs/libspc/spc_cpu.c | |||
@@ -0,0 +1,1049 @@ | |||
1 | /*************************************************************************** | ||
2 | * __________ __ ___. | ||
3 | * Open \______ \ ____ ____ | | _\_ |__ _______ ___ | ||
4 | * Source | _// _ \_/ ___\| |/ /| __ \ / _ \ \/ / | ||
5 | * Jukebox | | ( <_> ) \___| < | \_\ ( <_> > < < | ||
6 | * Firmware |____|_ /\____/ \___ >__|_ \|___ /\____/__/\_ \ | ||
7 | * \/ \/ \/ \/ \/ | ||
8 | * $Id$ | ||
9 | * | ||
10 | * Copyright (C) 2006-2007 Adam Gashlin (hcs) | ||
11 | * Copyright (C) 2004-2007 Shay Green (blargg) | ||
12 | * Copyright (C) 2002 Brad Martin | ||
13 | * | ||
14 | * This program is free software; you can redistribute it and/or | ||
15 | * modify it under the terms of the GNU General Public License | ||
16 | * as published by the Free Software Foundation; either version 2 | ||
17 | * of the License, or (at your option) any later version. | ||
18 | * | ||
19 | * This software is distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY | ||
20 | * KIND, either express or implied. | ||
21 | * | ||
22 | ****************************************************************************/ | ||
23 | |||
24 | /* The CPU portion (shock!) */ | ||
25 | #include "codeclib.h" | ||
26 | #include "spc_codec.h" | ||
27 | #include "spc_profiler.h" | ||
28 | |||
29 | #undef check | ||
30 | #define check assert | ||
31 | |||
32 | #define READ( addr ) (SPC_read( this, addr, spc_time_ )) | ||
33 | #define WRITE( addr, value ) (SPC_write( this, addr, value, spc_time_ )) | ||
34 | |||
35 | #define READ_DP( addr ) READ( (addr) + dp ) | ||
36 | #define WRITE_DP( addr, value ) WRITE( (addr) + dp, value ) | ||
37 | |||
38 | #define READ_PROG16( addr ) GET_LE16( RAM + (addr) ) | ||
39 | |||
40 | #define READ_PC( pc ) (*(pc)) | ||
41 | #define READ_PC16( pc ) GET_LE16( pc ) | ||
42 | |||
43 | #define SET_PC( n ) (pc = RAM + (n)) | ||
44 | #define GET_PC() (pc - RAM) | ||
45 | |||
46 | static unsigned char const cycle_table [0x100] = { | ||
47 | 2,8,4,5,3,4,3,6,2,6,5,4,5,4,6,8, /* 0 */ | ||
48 | 2,8,4,5,4,5,5,6,5,5,6,5,2,2,4,6, /* 1 */ | ||
49 | 2,8,4,5,3,4,3,6,2,6,5,4,5,4,5,4, /* 2 */ | ||
50 | 2,8,4,5,4,5,5,6,5,5,6,5,2,2,3,8, /* 3 */ | ||
51 | 2,8,4,5,3,4,3,6,2,6,4,4,5,4,6,6, /* 4 */ | ||
52 | 2,8,4,5,4,5,5,6,5,5,4,5,2,2,4,3, /* 5 */ | ||
53 | 2,8,4,5,3,4,3,6,2,6,4,4,5,4,5,5, /* 6 */ | ||
54 | 2,8,4,5,4,5,5,6,5,5,5,5,2,2,3,6, /* 7 */ | ||
55 | 2,8,4,5,3,4,3,6,2,6,5,4,5,2,4,5, /* 8 */ | ||
56 | 2,8,4,5,4,5,5,6,5,5,5,5,2,2,12,5,/* 9 */ | ||
57 | 3,8,4,5,3,4,3,6,2,6,4,4,5,2,4,4, /* A */ | ||
58 | 2,8,4,5,4,5,5,6,5,5,5,5,2,2,3,4, /* B */ | ||
59 | 3,8,4,5,4,5,4,7,2,5,6,4,5,2,4,9, /* C */ | ||
60 | 2,8,4,5,5,6,6,7,4,5,4,5,2,2,6,3, /* D */ | ||
61 | 2,8,4,5,3,4,3,6,2,4,5,3,4,3,4,3, /* E */ | ||
62 | 2,8,4,5,4,5,5,6,3,4,5,4,2,2,4,3 /* F */ | ||
63 | }; | ||
64 | |||
65 | #define MEM_BIT() CPU_mem_bit( this, pc, spc_time_ ) | ||
66 | |||
67 | static unsigned CPU_mem_bit( THIS, uint8_t const* pc, long const spc_time_ ) | ||
68 | ICODE_ATTR_SPC; | ||
69 | |||
70 | static unsigned CPU_mem_bit( THIS, uint8_t const* pc, long const spc_time_ ) | ||
71 | { | ||
72 | unsigned addr = READ_PC16( pc ); | ||
73 | unsigned t = READ( addr & 0x1FFF ) >> (addr >> 13); | ||
74 | return (t << 8) & 0x100; | ||
75 | } | ||
76 | |||
77 | /* status flags */ | ||
78 | enum { st_n = 0x80 }; | ||
79 | enum { st_v = 0x40 }; | ||
80 | enum { st_p = 0x20 }; | ||
81 | enum { st_b = 0x10 }; | ||
82 | enum { st_h = 0x08 }; | ||
83 | enum { st_i = 0x04 }; | ||
84 | enum { st_z = 0x02 }; | ||
85 | enum { st_c = 0x01 }; | ||
86 | |||
87 | #define IS_NEG (nz & 0x880) | ||
88 | |||
89 | #define CALC_STATUS( out )\ | ||
90 | {\ | ||
91 | out = status & ~(st_n | st_z | st_c);\ | ||
92 | out |= (c >> 8) & st_c;\ | ||
93 | out |= (dp >> 3) & st_p;\ | ||
94 | if ( IS_NEG ) out |= st_n;\ | ||
95 | if ( !(nz & 0xFF) ) out |= st_z;\ | ||
96 | } | ||
97 | |||
98 | #define SET_STATUS( in )\ | ||
99 | {\ | ||
100 | status = in & ~(st_n | st_z | st_c | st_p);\ | ||
101 | c = in << 8;\ | ||
102 | nz = ((in << 4) & 0x800) | (~in & st_z);\ | ||
103 | dp = (in << 3) & 0x100;\ | ||
104 | } | ||
105 | |||
106 | |||
107 | /* stack */ | ||
108 | #define PUSH( v ) (*--sp = (uint8_t) (v)) | ||
109 | #define PUSH16( v ) (sp -= 2, SET_LE16( sp, v )) | ||
110 | #define POP() (*sp++) | ||
111 | #define SET_SP( v ) (sp = RAM + 0x101 + (v)) | ||
112 | #define GET_SP() (sp - 0x101 - RAM) | ||
113 | |||
114 | long CPU_run( THIS, long start_time ) | ||
115 | { | ||
116 | #if 0 | ||
117 | ENTER_TIMER(cpu); | ||
118 | #endif | ||
119 | |||
120 | register long spc_time_ = start_time; | ||
121 | |||
122 | #ifdef CPU_ARM | ||
123 | uint8_t* const ram_ = ram.ram; | ||
124 | #undef RAM | ||
125 | #define RAM ram_ | ||
126 | #endif | ||
127 | |||
128 | int a = this->r.a; | ||
129 | int x = this->r.x; | ||
130 | int y = this->r.y; | ||
131 | |||
132 | uint8_t const* pc; | ||
133 | SET_PC( this->r.pc ); | ||
134 | |||
135 | uint8_t* sp; | ||
136 | SET_SP( this->r.sp ); | ||
137 | |||
138 | int status; | ||
139 | int c; | ||
140 | int nz; | ||
141 | unsigned dp; | ||
142 | { | ||
143 | int temp = this->r.status; | ||
144 | SET_STATUS( temp ); | ||
145 | } | ||
146 | |||
147 | goto loop; | ||
148 | |||
149 | /* main loop */ | ||
150 | cbranch_taken_loop: | ||
151 | pc += *(int8_t const*) pc; | ||
152 | spc_time_ += 2; | ||
153 | inc_pc_loop: | ||
154 | pc++; | ||
155 | loop: | ||
156 | check( (unsigned) GET_PC() < 0x10000 ); | ||
157 | check( (unsigned) GET_SP() < 0x100 ); | ||
158 | check( (unsigned) a < 0x100 ); | ||
159 | check( (unsigned) x < 0x100 ); | ||
160 | check( (unsigned) y < 0x100 ); | ||
161 | |||
162 | unsigned opcode = *pc; | ||
163 | int cycles = this->cycle_table [opcode]; | ||
164 | unsigned data = *++pc; | ||
165 | if ( (spc_time_ += cycles) > 0 ) | ||
166 | goto out_of_time; | ||
167 | switch ( opcode ) | ||
168 | { | ||
169 | |||
170 | /* Common instructions */ | ||
171 | |||
172 | #define BRANCH( cond )\ | ||
173 | {\ | ||
174 | pc++;\ | ||
175 | int offset = (int8_t) data;\ | ||
176 | if ( cond ) {\ | ||
177 | pc += offset;\ | ||
178 | spc_time_ += 2;\ | ||
179 | }\ | ||
180 | goto loop;\ | ||
181 | } | ||
182 | |||
183 | case 0xF0: /* BEQ (most common) */ | ||
184 | BRANCH( !(uint8_t) nz ) | ||
185 | |||
186 | case 0xD0: /* BNE */ | ||
187 | BRANCH( (uint8_t) nz ) | ||
188 | |||
189 | case 0x3F: /* CALL */ | ||
190 | PUSH16( GET_PC() + 2 ); | ||
191 | SET_PC( READ_PC16( pc ) ); | ||
192 | goto loop; | ||
193 | |||
194 | case 0x6F: /* RET */ | ||
195 | SET_PC( POP() ); | ||
196 | pc += POP() * 0x100; | ||
197 | goto loop; | ||
198 | |||
199 | #define CASE( n ) case n: | ||
200 | |||
201 | /* Define common address modes based on opcode for immediate mode. Execution | ||
202 | ends with data set to the address of the operand. */ | ||
203 | #define ADDR_MODES( op )\ | ||
204 | CASE( op - 0x02 ) /* (X) */\ | ||
205 | data = x + dp;\ | ||
206 | pc--;\ | ||
207 | goto end_##op;\ | ||
208 | CASE( op + 0x0F ) /* (dp)+Y */\ | ||
209 | data = READ_PROG16( data + dp ) + y;\ | ||
210 | goto end_##op;\ | ||
211 | CASE( op - 0x01 ) /* (dp+X) */\ | ||
212 | data = READ_PROG16( ((uint8_t) (data + x)) + dp );\ | ||
213 | goto end_##op;\ | ||
214 | CASE( op + 0x0E ) /* abs+Y */\ | ||
215 | data += y;\ | ||
216 | goto abs_##op;\ | ||
217 | CASE( op + 0x0D ) /* abs+X */\ | ||
218 | data += x;\ | ||
219 | CASE( op - 0x03 ) /* abs */\ | ||
220 | abs_##op:\ | ||
221 | data += 0x100 * READ_PC( ++pc );\ | ||
222 | goto end_##op;\ | ||
223 | CASE( op + 0x0C ) /* dp+X */\ | ||
224 | data = (uint8_t) (data + x);\ | ||
225 | CASE( op - 0x04 ) /* dp */\ | ||
226 | data += dp;\ | ||
227 | end_##op: | ||
228 | |||
229 | /* 1. 8-bit Data Transmission Commands. Group I */ | ||
230 | |||
231 | ADDR_MODES( 0xE8 ) /* MOV A,addr */ | ||
232 | /*case 0xE4:*/ /* MOV a,dp (most common) */ | ||
233 | mov_a_addr: | ||
234 | a = nz = READ( data ); | ||
235 | goto inc_pc_loop; | ||
236 | case 0xBF: /* MOV A,(X)+ */ | ||
237 | data = x + dp; | ||
238 | x = (uint8_t) (x + 1); | ||
239 | pc--; | ||
240 | goto mov_a_addr; | ||
241 | |||
242 | case 0xE8: /* MOV A,imm */ | ||
243 | a = data; | ||
244 | nz = data; | ||
245 | goto inc_pc_loop; | ||
246 | |||
247 | case 0xF9: /* MOV X,dp+Y */ | ||
248 | data = (uint8_t) (data + y); | ||
249 | case 0xF8: /* MOV X,dp */ | ||
250 | data += dp; | ||
251 | goto mov_x_addr; | ||
252 | case 0xE9: /* MOV X,abs */ | ||
253 | data = READ_PC16( pc ); | ||
254 | pc++; | ||
255 | mov_x_addr: | ||
256 | data = READ( data ); | ||
257 | case 0xCD: /* MOV X,imm */ | ||
258 | x = data; | ||
259 | nz = data; | ||
260 | goto inc_pc_loop; | ||
261 | |||
262 | case 0xFB: /* MOV Y,dp+X */ | ||
263 | data = (uint8_t) (data + x); | ||
264 | case 0xEB: /* MOV Y,dp */ | ||
265 | data += dp; | ||
266 | goto mov_y_addr; | ||
267 | case 0xEC: /* MOV Y,abs */ | ||
268 | data = READ_PC16( pc ); | ||
269 | pc++; | ||
270 | mov_y_addr: | ||
271 | data = READ( data ); | ||
272 | case 0x8D: /* MOV Y,imm */ | ||
273 | y = data; | ||
274 | nz = data; | ||
275 | goto inc_pc_loop; | ||
276 | |||
277 | /* 2. 8-BIT DATA TRANSMISSION COMMANDS, GROUP 2 */ | ||
278 | |||
279 | ADDR_MODES( 0xC8 ) /* MOV addr,A */ | ||
280 | WRITE( data, a ); | ||
281 | goto inc_pc_loop; | ||
282 | |||
283 | { | ||
284 | int temp; | ||
285 | case 0xCC: /* MOV abs,Y */ | ||
286 | temp = y; | ||
287 | goto mov_abs_temp; | ||
288 | case 0xC9: /* MOV abs,X */ | ||
289 | temp = x; | ||
290 | mov_abs_temp: | ||
291 | WRITE( READ_PC16( pc ), temp ); | ||
292 | pc += 2; | ||
293 | goto loop; | ||
294 | } | ||
295 | |||
296 | case 0xD9: /* MOV dp+Y,X */ | ||
297 | data = (uint8_t) (data + y); | ||
298 | case 0xD8: /* MOV dp,X */ | ||
299 | WRITE( data + dp, x ); | ||
300 | goto inc_pc_loop; | ||
301 | |||
302 | case 0xDB: /* MOV dp+X,Y */ | ||
303 | data = (uint8_t) (data + x); | ||
304 | case 0xCB: /* MOV dp,Y */ | ||
305 | WRITE( data + dp, y ); | ||
306 | goto inc_pc_loop; | ||
307 | |||
308 | case 0xFA: /* MOV dp,dp */ | ||
309 | data = READ( data + dp ); | ||
310 | case 0x8F: /* MOV dp,#imm */ | ||
311 | WRITE_DP( READ_PC( ++pc ), data ); | ||
312 | goto inc_pc_loop; | ||
313 | |||
314 | /* 3. 8-BIT DATA TRANSMISSIN COMMANDS, GROUP 3. */ | ||
315 | |||
316 | case 0x7D: /* MOV A,X */ | ||
317 | a = x; | ||
318 | nz = x; | ||
319 | goto loop; | ||
320 | |||
321 | case 0xDD: /* MOV A,Y */ | ||
322 | a = y; | ||
323 | nz = y; | ||
324 | goto loop; | ||
325 | |||
326 | case 0x5D: /* MOV X,A */ | ||
327 | x = a; | ||
328 | nz = a; | ||
329 | goto loop; | ||
330 | |||
331 | case 0xFD: /* MOV Y,A */ | ||
332 | y = a; | ||
333 | nz = a; | ||
334 | goto loop; | ||
335 | |||
336 | case 0x9D: /* MOV X,SP */ | ||
337 | x = nz = GET_SP(); | ||
338 | goto loop; | ||
339 | |||
340 | case 0xBD: /* MOV SP,X */ | ||
341 | SET_SP( x ); | ||
342 | goto loop; | ||
343 | |||
344 | /*case 0xC6:*/ /* MOV (X),A (handled by MOV addr,A in group 2) */ | ||
345 | |||
346 | case 0xAF: /* MOV (X)+,A */ | ||
347 | WRITE_DP( x, a ); | ||
348 | x++; | ||
349 | goto loop; | ||
350 | |||
351 | /* 5. 8-BIT LOGIC OPERATION COMMANDS */ | ||
352 | |||
353 | #define LOGICAL_OP( op, func )\ | ||
354 | ADDR_MODES( op ) /* addr */\ | ||
355 | data = READ( data );\ | ||
356 | case op: /* imm */\ | ||
357 | nz = a func##= data;\ | ||
358 | goto inc_pc_loop;\ | ||
359 | { unsigned addr;\ | ||
360 | case op + 0x11: /* X,Y */\ | ||
361 | data = READ_DP( y );\ | ||
362 | addr = x + dp;\ | ||
363 | pc--;\ | ||
364 | goto addr_##op;\ | ||
365 | case op + 0x01: /* dp,dp */\ | ||
366 | data = READ_DP( data );\ | ||
367 | case op + 0x10: /*dp,imm*/\ | ||
368 | addr = READ_PC( ++pc ) + dp;\ | ||
369 | addr_##op:\ | ||
370 | nz = data func READ( addr );\ | ||
371 | WRITE( addr, nz );\ | ||
372 | goto inc_pc_loop;\ | ||
373 | } | ||
374 | |||
375 | LOGICAL_OP( 0x28, & ); /* AND */ | ||
376 | |||
377 | LOGICAL_OP( 0x08, | ); /* OR */ | ||
378 | |||
379 | LOGICAL_OP( 0x48, ^ ); /* EOR */ | ||
380 | |||
381 | /* 4. 8-BIT ARITHMETIC OPERATION COMMANDS */ | ||
382 | |||
383 | ADDR_MODES( 0x68 ) /* CMP addr */ | ||
384 | data = READ( data ); | ||
385 | case 0x68: /* CMP imm */ | ||
386 | nz = a - data; | ||
387 | c = ~nz; | ||
388 | nz &= 0xFF; | ||
389 | goto inc_pc_loop; | ||
390 | |||
391 | case 0x79: /* CMP (X),(Y) */ | ||
392 | data = READ_DP( x ); | ||
393 | nz = data - READ_DP( y ); | ||
394 | c = ~nz; | ||
395 | nz &= 0xFF; | ||
396 | goto loop; | ||
397 | |||
398 | case 0x69: /* CMP (dp),(dp) */ | ||
399 | data = READ_DP( data ); | ||
400 | case 0x78: /* CMP dp,imm */ | ||
401 | nz = READ_DP( READ_PC( ++pc ) ) - data; | ||
402 | c = ~nz; | ||
403 | nz &= 0xFF; | ||
404 | goto inc_pc_loop; | ||
405 | |||
406 | case 0x3E: /* CMP X,dp */ | ||
407 | data += dp; | ||
408 | goto cmp_x_addr; | ||
409 | case 0x1E: /* CMP X,abs */ | ||
410 | data = READ_PC16( pc ); | ||
411 | pc++; | ||
412 | cmp_x_addr: | ||
413 | data = READ( data ); | ||
414 | case 0xC8: /* CMP X,imm */ | ||
415 | nz = x - data; | ||
416 | c = ~nz; | ||
417 | nz &= 0xFF; | ||
418 | goto inc_pc_loop; | ||
419 | |||
420 | case 0x7E: /* CMP Y,dp */ | ||
421 | data += dp; | ||
422 | goto cmp_y_addr; | ||
423 | case 0x5E: /* CMP Y,abs */ | ||
424 | data = READ_PC16( pc ); | ||
425 | pc++; | ||
426 | cmp_y_addr: | ||
427 | data = READ( data ); | ||
428 | case 0xAD: /* CMP Y,imm */ | ||
429 | nz = y - data; | ||
430 | c = ~nz; | ||
431 | nz &= 0xFF; | ||
432 | goto inc_pc_loop; | ||
433 | |||
434 | { | ||
435 | int addr; | ||
436 | case 0xB9: /* SBC (x),(y) */ | ||
437 | case 0x99: /* ADC (x),(y) */ | ||
438 | pc--; /* compensate for inc later */ | ||
439 | data = READ_DP( x ); | ||
440 | addr = y + dp; | ||
441 | goto adc_addr; | ||
442 | case 0xA9: /* SBC dp,dp */ | ||
443 | case 0x89: /* ADC dp,dp */ | ||
444 | data = READ_DP( data ); | ||
445 | case 0xB8: /* SBC dp,imm */ | ||
446 | case 0x98: /* ADC dp,imm */ | ||
447 | addr = READ_PC( ++pc ) + dp; | ||
448 | adc_addr: | ||
449 | nz = READ( addr ); | ||
450 | goto adc_data; | ||
451 | |||
452 | /* catch ADC and SBC together, then decode later based on operand */ | ||
453 | #undef CASE | ||
454 | #define CASE( n ) case n: case (n) + 0x20: | ||
455 | ADDR_MODES( 0x88 ) /* ADC/SBC addr */ | ||
456 | data = READ( data ); | ||
457 | case 0xA8: /* SBC imm */ | ||
458 | case 0x88: /* ADC imm */ | ||
459 | addr = -1; /* A */ | ||
460 | nz = a; | ||
461 | adc_data: { | ||
462 | if ( opcode & 0x20 ) | ||
463 | data ^= 0xFF; /* SBC */ | ||
464 | int carry = (c >> 8) & 1; | ||
465 | int ov = (nz ^ 0x80) + carry + (int8_t) data; /* sign-extend */ | ||
466 | int hc = (nz & 15) + carry; | ||
467 | c = nz += data + carry; | ||
468 | hc = (nz & 15) - hc; | ||
469 | status = (status & ~(st_v | st_h)) | ((ov >> 2) & st_v) | | ||
470 | ((hc >> 1) & st_h); | ||
471 | if ( addr < 0 ) { | ||
472 | a = (uint8_t) nz; | ||
473 | goto inc_pc_loop; | ||
474 | } | ||
475 | WRITE( addr, (uint8_t) nz ); | ||
476 | goto inc_pc_loop; | ||
477 | } | ||
478 | |||
479 | } | ||
480 | |||
481 | /* 6. ADDITION & SUBTRACTION COMMANDS */ | ||
482 | |||
483 | #define INC_DEC_REG( reg, n )\ | ||
484 | nz = reg + n;\ | ||
485 | reg = (uint8_t) nz;\ | ||
486 | goto loop; | ||
487 | |||
488 | case 0xBC: INC_DEC_REG( a, 1 ) /* INC A */ | ||
489 | case 0x3D: INC_DEC_REG( x, 1 ) /* INC X */ | ||
490 | case 0xFC: INC_DEC_REG( y, 1 ) /* INC Y */ | ||
491 | |||
492 | case 0x9C: INC_DEC_REG( a, -1 ) /* DEC A */ | ||
493 | case 0x1D: INC_DEC_REG( x, -1 ) /* DEC X */ | ||
494 | case 0xDC: INC_DEC_REG( y, -1 ) /* DEC Y */ | ||
495 | |||
496 | case 0x9B: /* DEC dp+X */ | ||
497 | case 0xBB: /* INC dp+X */ | ||
498 | data = (uint8_t) (data + x); | ||
499 | case 0x8B: /* DEC dp */ | ||
500 | case 0xAB: /* INC dp */ | ||
501 | data += dp; | ||
502 | goto inc_abs; | ||
503 | case 0x8C: /* DEC abs */ | ||
504 | case 0xAC: /* INC abs */ | ||
505 | data = READ_PC16( pc ); | ||
506 | pc++; | ||
507 | inc_abs: | ||
508 | nz = ((opcode >> 4) & 2) - 1; | ||
509 | nz += READ( data ); | ||
510 | WRITE( data, (uint8_t) nz ); | ||
511 | goto inc_pc_loop; | ||
512 | |||
513 | /* 7. SHIFT, ROTATION COMMANDS */ | ||
514 | |||
515 | case 0x5C: /* LSR A */ | ||
516 | c = 0; | ||
517 | case 0x7C:{/* ROR A */ | ||
518 | nz = ((c >> 1) & 0x80) | (a >> 1); | ||
519 | c = a << 8; | ||
520 | a = nz; | ||
521 | goto loop; | ||
522 | } | ||
523 | |||
524 | case 0x1C: /* ASL A */ | ||
525 | c = 0; | ||
526 | case 0x3C:{/* ROL A */ | ||
527 | int temp = (c >> 8) & 1; | ||
528 | c = a << 1; | ||
529 | nz = c | temp; | ||
530 | a = (uint8_t) nz; | ||
531 | goto loop; | ||
532 | } | ||
533 | |||
534 | case 0x0B: /* ASL dp */ | ||
535 | c = 0; | ||
536 | data += dp; | ||
537 | goto rol_mem; | ||
538 | case 0x1B: /* ASL dp+X */ | ||
539 | c = 0; | ||
540 | case 0x3B: /* ROL dp+X */ | ||
541 | data = (uint8_t) (data + x); | ||
542 | case 0x2B: /* ROL dp */ | ||
543 | data += dp; | ||
544 | goto rol_mem; | ||
545 | case 0x0C: /* ASL abs */ | ||
546 | c = 0; | ||
547 | case 0x2C: /* ROL abs */ | ||
548 | data = READ_PC16( pc ); | ||
549 | pc++; | ||
550 | rol_mem: | ||
551 | nz = (c >> 8) & 1; | ||
552 | nz |= (c = READ( data ) << 1); | ||
553 | WRITE( data, (uint8_t) nz ); | ||
554 | goto inc_pc_loop; | ||
555 | |||
556 | case 0x4B: /* LSR dp */ | ||
557 | c = 0; | ||
558 | data += dp; | ||
559 | goto ror_mem; | ||
560 | case 0x5B: /* LSR dp+X */ | ||
561 | c = 0; | ||
562 | case 0x7B: /* ROR dp+X */ | ||
563 | data = (uint8_t) (data + x); | ||
564 | case 0x6B: /* ROR dp */ | ||
565 | data += dp; | ||
566 | goto ror_mem; | ||
567 | case 0x4C: /* LSR abs */ | ||
568 | c = 0; | ||
569 | case 0x6C: /* ROR abs */ | ||
570 | data = READ_PC16( pc ); | ||
571 | pc++; | ||
572 | ror_mem: { | ||
573 | int temp = READ( data ); | ||
574 | nz = ((c >> 1) & 0x80) | (temp >> 1); | ||
575 | c = temp << 8; | ||
576 | WRITE( data, nz ); | ||
577 | goto inc_pc_loop; | ||
578 | } | ||
579 | |||
580 | case 0x9F: /* XCN */ | ||
581 | nz = a = (a >> 4) | (uint8_t) (a << 4); | ||
582 | goto loop; | ||
583 | |||
584 | /* 8. 16-BIT TRANSMISION COMMANDS */ | ||
585 | |||
586 | case 0xBA: /* MOVW YA,dp */ | ||
587 | a = READ_DP( data ); | ||
588 | nz = (a & 0x7F) | (a >> 1); | ||
589 | y = READ_DP( (uint8_t) (data + 1) ); | ||
590 | nz |= y; | ||
591 | goto inc_pc_loop; | ||
592 | |||
593 | case 0xDA: /* MOVW dp,YA */ | ||
594 | WRITE_DP( data, a ); | ||
595 | WRITE_DP( (uint8_t) (data + 1), y ); | ||
596 | goto inc_pc_loop; | ||
597 | |||
598 | /* 9. 16-BIT OPERATION COMMANDS */ | ||
599 | |||
600 | case 0x3A: /* INCW dp */ | ||
601 | case 0x1A:{/* DECW dp */ | ||
602 | data += dp; | ||
603 | |||
604 | /* low byte */ | ||
605 | int temp = READ( data ); | ||
606 | temp += ((opcode >> 4) & 2) - 1; /* +1 for INCW, -1 for DECW */ | ||
607 | nz = ((temp >> 1) | temp) & 0x7F; | ||
608 | WRITE( data, (uint8_t) temp ); | ||
609 | |||
610 | /* high byte */ | ||
611 | data = ((uint8_t) (data + 1)) + dp; | ||
612 | temp >>= 8; | ||
613 | temp = (uint8_t) (temp + READ( data )); | ||
614 | nz |= temp; | ||
615 | WRITE( data, temp ); | ||
616 | |||
617 | goto inc_pc_loop; | ||
618 | } | ||
619 | |||
620 | case 0x9A: /* SUBW YA,dp */ | ||
621 | case 0x7A: /* ADDW YA,dp */ | ||
622 | { | ||
623 | /* read 16-bit addend */ | ||
624 | int temp = READ_DP( data ); | ||
625 | int sign = READ_DP( (uint8_t) (data + 1) ); | ||
626 | temp += 0x100 * sign; | ||
627 | status &= ~(st_v | st_h); | ||
628 | |||
629 | /* to do: fix half-carry for SUBW (it's probably wrong) */ | ||
630 | |||
631 | /* for SUBW, negate and truncate to 16 bits */ | ||
632 | if ( opcode & 0x80 ) { | ||
633 | temp = (temp ^ 0xFFFF) + 1; | ||
634 | sign = temp >> 8; | ||
635 | } | ||
636 | |||
637 | /* add low byte (A) */ | ||
638 | temp += a; | ||
639 | a = (uint8_t) temp; | ||
640 | nz = (temp | (temp >> 1)) & 0x7F; | ||
641 | |||
642 | /* add high byte (Y) */ | ||
643 | temp >>= 8; | ||
644 | c = y + temp; | ||
645 | nz = (nz | c) & 0xFF; | ||
646 | |||
647 | /* half-carry (temporary avoids CodeWarrior optimizer bug) */ | ||
648 | unsigned hc = (c & 15) - (y & 15); | ||
649 | status |= (hc >> 4) & st_h; | ||
650 | |||
651 | /* overflow if sign of YA changed when previous sign | ||
652 | and addend sign were same */ | ||
653 | status |= (((c ^ y) & ~(y ^ sign)) >> 1) & st_v; | ||
654 | |||
655 | y = (uint8_t) c; | ||
656 | |||
657 | goto inc_pc_loop; | ||
658 | } | ||
659 | |||
660 | case 0x5A: { /* CMPW YA,dp */ | ||
661 | int temp = a - READ_DP( data ); | ||
662 | nz = ((temp >> 1) | temp) & 0x7F; | ||
663 | temp = y + (temp >> 8); | ||
664 | temp -= READ_DP( (uint8_t) (data + 1) ); | ||
665 | nz |= temp; | ||
666 | c = ~temp; | ||
667 | nz &= 0xFF; | ||
668 | goto inc_pc_loop; | ||
669 | } | ||
670 | |||
671 | /* 10. MULTIPLICATION & DIVISON COMMANDS */ | ||
672 | |||
673 | case 0xCF: { /* MUL YA */ | ||
674 | unsigned temp = y * a; | ||
675 | a = (uint8_t) temp; | ||
676 | nz = ((temp >> 1) | temp) & 0x7F; | ||
677 | y = temp >> 8; | ||
678 | nz |= y; | ||
679 | goto loop; | ||
680 | } | ||
681 | |||
682 | case 0x9E: /* DIV YA,X */ | ||
683 | { | ||
684 | /* behavior based on SPC CPU tests */ | ||
685 | |||
686 | status &= ~(st_h | st_v); | ||
687 | |||
688 | if ( (y & 15) >= (x & 15) ) | ||
689 | status |= st_h; | ||
690 | |||
691 | if ( y >= x ) | ||
692 | status |= st_v; | ||
693 | |||
694 | unsigned ya = y * 0x100 + a; | ||
695 | if ( y < x * 2 ) | ||
696 | { | ||
697 | a = ya / x; | ||
698 | y = ya - a * x; | ||
699 | } | ||
700 | else | ||
701 | { | ||
702 | a = 255 - (ya - x * 0x200) / (256 - x); | ||
703 | y = x + (ya - x * 0x200) % (256 - x); | ||
704 | } | ||
705 | |||
706 | nz = (uint8_t) a; | ||
707 | a = (uint8_t) a; | ||
708 | |||
709 | goto loop; | ||
710 | } | ||
711 | |||
712 | /* 11. DECIMAL COMPENSATION COMMANDS */ | ||
713 | |||
714 | /* seem unused */ | ||
715 | /* case 0xDF: */ /* DAA */ | ||
716 | /* case 0xBE: */ /* DAS */ | ||
717 | |||
718 | /* 12. BRANCHING COMMANDS */ | ||
719 | |||
720 | case 0x2F: /* BRA rel */ | ||
721 | pc += (int8_t) data; | ||
722 | goto inc_pc_loop; | ||
723 | |||
724 | case 0x30: /* BMI */ | ||
725 | BRANCH( IS_NEG ) | ||
726 | |||
727 | case 0x10: /* BPL */ | ||
728 | BRANCH( !IS_NEG ) | ||
729 | |||
730 | case 0xB0: /* BCS */ | ||
731 | BRANCH( c & 0x100 ) | ||
732 | |||
733 | case 0x90: /* BCC */ | ||
734 | BRANCH( !(c & 0x100) ) | ||
735 | |||
736 | case 0x70: /* BVS */ | ||
737 | BRANCH( status & st_v ) | ||
738 | |||
739 | case 0x50: /* BVC */ | ||
740 | BRANCH( !(status & st_v) ) | ||
741 | |||
742 | case 0x03: /* BBS dp.bit,rel */ | ||
743 | case 0x23: | ||
744 | case 0x43: | ||
745 | case 0x63: | ||
746 | case 0x83: | ||
747 | case 0xA3: | ||
748 | case 0xC3: | ||
749 | case 0xE3: | ||
750 | pc++; | ||
751 | if ( (READ_DP( data ) >> (opcode >> 5)) & 1 ) | ||
752 | goto cbranch_taken_loop; | ||
753 | goto inc_pc_loop; | ||
754 | |||
755 | case 0x13: /* BBC dp.bit,rel */ | ||
756 | case 0x33: | ||
757 | case 0x53: | ||
758 | case 0x73: | ||
759 | case 0x93: | ||
760 | case 0xB3: | ||
761 | case 0xD3: | ||
762 | case 0xF3: | ||
763 | pc++; | ||
764 | if ( !((READ_DP( data ) >> (opcode >> 5)) & 1) ) | ||
765 | goto cbranch_taken_loop; | ||
766 | goto inc_pc_loop; | ||
767 | |||
768 | case 0xDE: /* CBNE dp+X,rel */ | ||
769 | data = (uint8_t) (data + x); | ||
770 | /* fall through */ | ||
771 | case 0x2E: /* CBNE dp,rel */ | ||
772 | pc++; | ||
773 | if ( READ_DP( data ) != a ) | ||
774 | goto cbranch_taken_loop; | ||
775 | goto inc_pc_loop; | ||
776 | |||
777 | case 0xFE: /* DBNZ Y,rel */ | ||
778 | y = (uint8_t) (y - 1); | ||
779 | BRANCH( y ) | ||
780 | |||
781 | case 0x6E: { /* DBNZ dp,rel */ | ||
782 | pc++; | ||
783 | unsigned temp = READ_DP( data ) - 1; | ||
784 | WRITE_DP( (uint8_t) data, (uint8_t) temp ); | ||
785 | if ( temp ) | ||
786 | goto cbranch_taken_loop; | ||
787 | goto inc_pc_loop; | ||
788 | } | ||
789 | |||
790 | case 0x1F: /* JMP (abs+X) */ | ||
791 | SET_PC( READ_PC16( pc ) + x ); | ||
792 | /* fall through */ | ||
793 | case 0x5F: /* JMP abs */ | ||
794 | SET_PC( READ_PC16( pc ) ); | ||
795 | goto loop; | ||
796 | |||
797 | /* 13. SUB-ROUTINE CALL RETURN COMMANDS */ | ||
798 | |||
799 | case 0x0F:{/* BRK */ | ||
800 | check( 0 ); /* untested */ | ||
801 | PUSH16( GET_PC() + 1 ); | ||
802 | SET_PC( READ_PROG16( 0xFFDE ) ); /* vector address verified */ | ||
803 | int temp; | ||
804 | CALC_STATUS( temp ); | ||
805 | PUSH( temp ); | ||
806 | status = (status | st_b) & ~st_i; | ||
807 | goto loop; | ||
808 | } | ||
809 | |||
810 | case 0x4F: /* PCALL offset */ | ||
811 | PUSH16( GET_PC() + 1 ); | ||
812 | SET_PC( 0xFF00 + data ); | ||
813 | goto loop; | ||
814 | |||
815 | case 0x01: /* TCALL n */ | ||
816 | case 0x11: | ||
817 | case 0x21: | ||
818 | case 0x31: | ||
819 | case 0x41: | ||
820 | case 0x51: | ||
821 | case 0x61: | ||
822 | case 0x71: | ||
823 | case 0x81: | ||
824 | case 0x91: | ||
825 | case 0xA1: | ||
826 | case 0xB1: | ||
827 | case 0xC1: | ||
828 | case 0xD1: | ||
829 | case 0xE1: | ||
830 | case 0xF1: | ||
831 | PUSH16( GET_PC() ); | ||
832 | SET_PC( READ_PROG16( 0xFFDE - (opcode >> 3) ) ); | ||
833 | goto loop; | ||
834 | |||
835 | /* 14. STACK OPERATION COMMANDS */ | ||
836 | |||
837 | { | ||
838 | int temp; | ||
839 | case 0x7F: /* RET1 */ | ||
840 | temp = POP(); | ||
841 | SET_PC( POP() ); | ||
842 | pc += POP() << 8; | ||
843 | goto set_status; | ||
844 | case 0x8E: /* POP PSW */ | ||
845 | temp = POP(); | ||
846 | set_status: | ||
847 | SET_STATUS( temp ); | ||
848 | goto loop; | ||
849 | } | ||
850 | |||
851 | case 0x0D: { /* PUSH PSW */ | ||
852 | int temp; | ||
853 | CALC_STATUS( temp ); | ||
854 | PUSH( temp ); | ||
855 | goto loop; | ||
856 | } | ||
857 | |||
858 | case 0x2D: /* PUSH A */ | ||
859 | PUSH( a ); | ||
860 | goto loop; | ||
861 | |||
862 | case 0x4D: /* PUSH X */ | ||
863 | PUSH( x ); | ||
864 | goto loop; | ||
865 | |||
866 | case 0x6D: /* PUSH Y */ | ||
867 | PUSH( y ); | ||
868 | goto loop; | ||
869 | |||
870 | case 0xAE: /* POP A */ | ||
871 | a = POP(); | ||
872 | goto loop; | ||
873 | |||
874 | case 0xCE: /* POP X */ | ||
875 | x = POP(); | ||
876 | goto loop; | ||
877 | |||
878 | case 0xEE: /* POP Y */ | ||
879 | y = POP(); | ||
880 | goto loop; | ||
881 | |||
882 | /* 15. BIT OPERATION COMMANDS */ | ||
883 | |||
884 | case 0x02: /* SET1 */ | ||
885 | case 0x22: | ||
886 | case 0x42: | ||
887 | case 0x62: | ||
888 | case 0x82: | ||
889 | case 0xA2: | ||
890 | case 0xC2: | ||
891 | case 0xE2: | ||
892 | case 0x12: /* CLR1 */ | ||
893 | case 0x32: | ||
894 | case 0x52: | ||
895 | case 0x72: | ||
896 | case 0x92: | ||
897 | case 0xB2: | ||
898 | case 0xD2: | ||
899 | case 0xF2: { | ||
900 | data += dp; | ||
901 | int bit = 1 << (opcode >> 5); | ||
902 | int mask = ~bit; | ||
903 | if ( opcode & 0x10 ) | ||
904 | bit = 0; | ||
905 | WRITE( data, (READ( data ) & mask) | bit ); | ||
906 | goto inc_pc_loop; | ||
907 | } | ||
908 | |||
909 | case 0x0E: /* TSET1 abs */ | ||
910 | case 0x4E:{/* TCLR1 abs */ | ||
911 | data = READ_PC16( pc ); | ||
912 | pc += 2; | ||
913 | unsigned temp = READ( data ); | ||
914 | nz = temp & a; | ||
915 | temp &= ~a; | ||
916 | if ( !(opcode & 0x40) ) | ||
917 | temp |= a; | ||
918 | WRITE( data, temp ); | ||
919 | goto loop; | ||
920 | } | ||
921 | |||
922 | case 0x4A: /* AND1 C,mem.bit */ | ||
923 | c &= MEM_BIT(); | ||
924 | pc += 2; | ||
925 | goto loop; | ||
926 | |||
927 | case 0x6A: /* AND1 C,/mem.bit */ | ||
928 | check( 0 ); /* untested */ | ||
929 | c &= ~MEM_BIT(); | ||
930 | pc += 2; | ||
931 | goto loop; | ||
932 | |||
933 | case 0x0A: /* OR1 C,mem.bit */ | ||
934 | check( 0 ); /* untested */ | ||
935 | c |= MEM_BIT(); | ||
936 | pc += 2; | ||
937 | goto loop; | ||
938 | |||
939 | case 0x2A: /* OR1 C,/mem.bit */ | ||
940 | check( 0 ); /* untested */ | ||
941 | c |= ~MEM_BIT(); | ||
942 | pc += 2; | ||
943 | goto loop; | ||
944 | |||
945 | case 0x8A: /* EOR1 C,mem.bit */ | ||
946 | c ^= MEM_BIT(); | ||
947 | pc += 2; | ||
948 | goto loop; | ||
949 | |||
950 | case 0xEA: { /* NOT1 mem.bit */ | ||
951 | data = READ_PC16( pc ); | ||
952 | pc += 2; | ||
953 | unsigned temp = READ( data & 0x1FFF ); | ||
954 | temp ^= 1 << (data >> 13); | ||
955 | WRITE( data & 0x1FFF, temp ); | ||
956 | goto loop; | ||
957 | } | ||
958 | |||
959 | case 0xCA: { /* MOV1 mem.bit,C */ | ||
960 | data = READ_PC16( pc ); | ||
961 | pc += 2; | ||
962 | unsigned temp = READ( data & 0x1FFF ); | ||
963 | unsigned bit = data >> 13; | ||
964 | temp = (temp & ~(1 << bit)) | (((c >> 8) & 1) << bit); | ||
965 | WRITE( data & 0x1FFF, temp ); | ||
966 | goto loop; | ||
967 | } | ||
968 | |||
969 | case 0xAA: /* MOV1 C,mem.bit */ | ||
970 | c = MEM_BIT(); | ||
971 | pc += 2; | ||
972 | goto loop; | ||
973 | |||
974 | /* 16. PROGRAM STATUS FLAG OPERATION COMMANDS */ | ||
975 | |||
976 | case 0x60: /* CLRC */ | ||
977 | c = 0; | ||
978 | goto loop; | ||
979 | |||
980 | case 0x80: /* SETC */ | ||
981 | c = ~0; | ||
982 | goto loop; | ||
983 | |||
984 | case 0xED: /* NOTC */ | ||
985 | c ^= 0x100; | ||
986 | goto loop; | ||
987 | |||
988 | case 0xE0: /* CLRV */ | ||
989 | status &= ~(st_v | st_h); | ||
990 | goto loop; | ||
991 | |||
992 | case 0x20: /* CLRP */ | ||
993 | dp = 0; | ||
994 | goto loop; | ||
995 | |||
996 | case 0x40: /* SETP */ | ||
997 | dp = 0x100; | ||
998 | goto loop; | ||
999 | |||
1000 | case 0xA0: /* EI */ | ||
1001 | check( 0 ); /* untested */ | ||
1002 | status |= st_i; | ||
1003 | goto loop; | ||
1004 | |||
1005 | case 0xC0: /* DI */ | ||
1006 | check( 0 ); /* untested */ | ||
1007 | status &= ~st_i; | ||
1008 | goto loop; | ||
1009 | |||
1010 | /* 17. OTHER COMMANDS */ | ||
1011 | |||
1012 | case 0x00: /* NOP */ | ||
1013 | goto loop; | ||
1014 | |||
1015 | /*case 0xEF:*/ /* SLEEP */ | ||
1016 | /*case 0xFF:*/ /* STOP */ | ||
1017 | case 0xFF: | ||
1018 | c |= 1; /* force switch table to have 256 entries, | ||
1019 | hopefully helping optimizer */ | ||
1020 | } /* switch */ | ||
1021 | |||
1022 | /* unhandled instructions fall out of switch so emulator can catch them */ | ||
1023 | |||
1024 | out_of_time: | ||
1025 | /* undo partial execution of opcode */ | ||
1026 | spc_time_ -= this->cycle_table [*--pc]; | ||
1027 | { | ||
1028 | int temp; | ||
1029 | CALC_STATUS( temp ); | ||
1030 | this->r.status = (uint8_t) temp; | ||
1031 | } | ||
1032 | |||
1033 | this->r.pc = GET_PC(); | ||
1034 | this->r.sp = (uint8_t) GET_SP(); | ||
1035 | this->r.a = (uint8_t) a; | ||
1036 | this->r.x = (uint8_t) x; | ||
1037 | this->r.y = (uint8_t) y; | ||
1038 | |||
1039 | #if 0 | ||
1040 | EXIT_TIMER(cpu); | ||
1041 | #endif | ||
1042 | return spc_time_; | ||
1043 | } | ||
1044 | |||
1045 | void CPU_Init( THIS ) | ||
1046 | { | ||
1047 | ci->memcpy( this->cycle_table, cycle_table, sizeof cycle_table ); | ||
1048 | } | ||
1049 | |||
diff --git a/lib/rbcodec/codecs/libspc/spc_dsp.c b/lib/rbcodec/codecs/libspc/spc_dsp.c new file mode 100644 index 0000000000..6350c4c331 --- /dev/null +++ b/lib/rbcodec/codecs/libspc/spc_dsp.c | |||
@@ -0,0 +1,1594 @@ | |||
1 | /*************************************************************************** | ||
2 | * __________ __ ___. | ||
3 | * Open \______ \ ____ ____ | | _\_ |__ _______ ___ | ||
4 | * Source | _// _ \_/ ___\| |/ /| __ \ / _ \ \/ / | ||
5 | * Jukebox | | ( <_> ) \___| < | \_\ ( <_> > < < | ||
6 | * Firmware |____|_ /\____/ \___ >__|_ \|___ /\____/__/\_ \ | ||
7 | * \/ \/ \/ \/ \/ | ||
8 | * $Id$ | ||
9 | * | ||
10 | * Copyright (C) 2007-2008 Michael Sevakis (jhMikeS) | ||
11 | * Copyright (C) 2006-2007 Adam Gashlin (hcs) | ||
12 | * Copyright (C) 2004-2007 Shay Green (blargg) | ||
13 | * Copyright (C) 2002 Brad Martin | ||
14 | * | ||
15 | * This program is free software; you can redistribute it and/or | ||
16 | * modify it under the terms of the GNU General Public License | ||
17 | * as published by the Free Software Foundation; either version 2 | ||
18 | * of the License, or (at your option) any later version. | ||
19 | * | ||
20 | * This software is distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY | ||
21 | * KIND, either express or implied. | ||
22 | * | ||
23 | ****************************************************************************/ | ||
24 | |||
25 | /* The DSP portion (awe!) */ | ||
26 | #include "codeclib.h" | ||
27 | #include "spc_codec.h" | ||
28 | #include "spc_profiler.h" | ||
29 | |||
30 | #if defined(CPU_COLDFIRE) || defined (CPU_ARM) | ||
31 | int32_t fir_buf[FIR_BUF_CNT] IBSS_ATTR_SPC | ||
32 | __attribute__((aligned(FIR_BUF_ALIGN*1))); | ||
33 | #endif | ||
34 | #if SPC_BRRCACHE | ||
35 | /* a little extra for samples that go past end */ | ||
36 | int16_t BRRcache [BRR_CACHE_SIZE] CACHEALIGN_ATTR; | ||
37 | #endif | ||
38 | |||
39 | void DSP_write( struct Spc_Dsp* this, int i, int data ) | ||
40 | { | ||
41 | assert( (unsigned) i < REGISTER_COUNT ); | ||
42 | |||
43 | this->r.reg [i] = data; | ||
44 | int high = i >> 4; | ||
45 | int low = i & 0x0F; | ||
46 | if ( low < 2 ) /* voice volumes */ | ||
47 | { | ||
48 | int left = *(int8_t const*) &this->r.reg [i & ~1]; | ||
49 | int right = *(int8_t const*) &this->r.reg [i | 1]; | ||
50 | struct voice_t* v = this->voice_state + high; | ||
51 | v->volume [0] = left; | ||
52 | v->volume [1] = right; | ||
53 | } | ||
54 | else if ( low == 0x0F ) /* fir coefficients */ | ||
55 | { | ||
56 | this->fir_coeff [7 - high] = (int8_t) data; /* sign-extend */ | ||
57 | } | ||
58 | } | ||
59 | |||
60 | #define CLAMP16( n ) clip_sample_16( n ) | ||
61 | |||
62 | #if SPC_BRRCACHE | ||
63 | static void decode_brr( struct Spc_Dsp* this, unsigned start_addr, | ||
64 | struct voice_t* voice, | ||
65 | struct raw_voice_t const* const raw_voice ) ICODE_ATTR_SPC; | ||
66 | static void decode_brr( struct Spc_Dsp* this, unsigned start_addr, | ||
67 | struct voice_t* voice, | ||
68 | struct raw_voice_t const* const raw_voice ) | ||
69 | { | ||
70 | /* setup same variables as where decode_brr() is called from */ | ||
71 | #undef RAM | ||
72 | #define RAM ram.ram | ||
73 | |||
74 | struct src_dir const* const sd = | ||
75 | &ram.sd[this->r.g.wave_page * 0x100/sizeof(struct src_dir)]; | ||
76 | struct cache_entry_t* const wave_entry = | ||
77 | &this->wave_entry [raw_voice->waveform]; | ||
78 | |||
79 | /* the following block can be put in place of the call to | ||
80 | decode_brr() below | ||
81 | */ | ||
82 | { | ||
83 | DEBUGF( "decode at %08x (wave #%d)\n", | ||
84 | start_addr, raw_voice->waveform ); | ||
85 | |||
86 | /* see if in cache */ | ||
87 | int i; | ||
88 | for ( i = 0; i < this->oldsize; i++ ) | ||
89 | { | ||
90 | struct cache_entry_t* e = &this->wave_entry_old [i]; | ||
91 | if ( e->start_addr == start_addr ) | ||
92 | { | ||
93 | DEBUGF( "found in wave_entry_old (oldsize=%d)\n", | ||
94 | this->oldsize ); | ||
95 | *wave_entry = *e; | ||
96 | goto wave_in_cache; | ||
97 | } | ||
98 | } | ||
99 | |||
100 | wave_entry->start_addr = start_addr; | ||
101 | |||
102 | uint8_t const* const loop_ptr = | ||
103 | RAM + letoh16(sd[raw_voice->waveform].loop); | ||
104 | short* loop_start = 0; | ||
105 | |||
106 | short* out = BRRcache + start_addr * 2; | ||
107 | wave_entry->samples = out; | ||
108 | *out++ = 0; | ||
109 | int smp1 = 0; | ||
110 | int smp2 = 0; | ||
111 | |||
112 | uint8_t const* addr = RAM + start_addr; | ||
113 | int block_header; | ||
114 | do | ||
115 | { | ||
116 | if ( addr == loop_ptr ) | ||
117 | { | ||
118 | loop_start = out; | ||
119 | DEBUGF( "loop at %08lx (wave #%d)\n", | ||
120 | (unsigned long)(addr - RAM), raw_voice->waveform ); | ||
121 | } | ||
122 | |||
123 | /* header */ | ||
124 | block_header = *addr; | ||
125 | addr += 9; | ||
126 | voice->addr = addr; | ||
127 | int const filter = (block_header & 0x0C) - 0x08; | ||
128 | |||
129 | /* scaling | ||
130 | (invalid scaling gives -4096 for neg nybble, 0 for pos) */ | ||
131 | static unsigned char const right_shifts [16] = { | ||
132 | 5, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 29, 29, 29, | ||
133 | }; | ||
134 | static unsigned char const left_shifts [16] = { | ||
135 | 0, 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 11, 11, 11 | ||
136 | }; | ||
137 | int const scale = block_header >> 4; | ||
138 | int const right_shift = right_shifts [scale]; | ||
139 | int const left_shift = left_shifts [scale]; | ||
140 | |||
141 | /* output position */ | ||
142 | out += BRR_BLOCK_SIZE; | ||
143 | int offset = -BRR_BLOCK_SIZE << 2; | ||
144 | |||
145 | do /* decode and filter 16 samples */ | ||
146 | { | ||
147 | /* Get nybble, sign-extend, then scale | ||
148 | get byte, select which nybble, sign-extend, then shift based | ||
149 | on scaling. also handles invalid scaling values. */ | ||
150 | int delta = (int) (int8_t) (addr [offset >> 3] << (offset & 4)) | ||
151 | >> right_shift << left_shift; | ||
152 | |||
153 | out [offset >> 2] = smp2; | ||
154 | |||
155 | if ( filter == 0 ) /* mode 0x08 (30-90% of the time) */ | ||
156 | { | ||
157 | delta -= smp2 >> 1; | ||
158 | delta += smp2 >> 5; | ||
159 | smp2 = smp1; | ||
160 | delta += smp1; | ||
161 | delta += (-smp1 - (smp1 >> 1)) >> 5; | ||
162 | } | ||
163 | else | ||
164 | { | ||
165 | if ( filter == -4 ) /* mode 0x04 */ | ||
166 | { | ||
167 | delta += smp1 >> 1; | ||
168 | delta += (-smp1) >> 5; | ||
169 | } | ||
170 | else if ( filter > -4 ) /* mode 0x0C */ | ||
171 | { | ||
172 | delta -= smp2 >> 1; | ||
173 | delta += (smp2 + (smp2 >> 1)) >> 4; | ||
174 | delta += smp1; | ||
175 | delta += (-smp1 * 13) >> 7; | ||
176 | } | ||
177 | smp2 = smp1; | ||
178 | } | ||
179 | |||
180 | delta = CLAMP16( delta ); | ||
181 | smp1 = (int16_t) (delta * 2); /* sign-extend */ | ||
182 | } | ||
183 | while ( (offset += 4) != 0 ); | ||
184 | |||
185 | /* if next block has end flag set, this block ends early */ | ||
186 | /* (verified) */ | ||
187 | if ( (block_header & 3) != 3 && (*addr & 3) == 1 ) | ||
188 | { | ||
189 | /* skip last 9 samples */ | ||
190 | out -= 9; | ||
191 | goto early_end; | ||
192 | } | ||
193 | } | ||
194 | while ( !(block_header & 1) && addr < RAM + 0x10000 ); | ||
195 | |||
196 | out [0] = smp2; | ||
197 | out [1] = smp1; | ||
198 | |||
199 | early_end: | ||
200 | wave_entry->end = (out - 1 - wave_entry->samples) << 12; | ||
201 | |||
202 | wave_entry->loop = 0; | ||
203 | if ( (block_header & 2) ) | ||
204 | { | ||
205 | if ( loop_start ) | ||
206 | { | ||
207 | int loop = out - loop_start; | ||
208 | wave_entry->loop = loop; | ||
209 | wave_entry->end += 0x3000; | ||
210 | out [2] = loop_start [2]; | ||
211 | out [3] = loop_start [3]; | ||
212 | out [4] = loop_start [4]; | ||
213 | } | ||
214 | else | ||
215 | { | ||
216 | DEBUGF( "loop point outside initial wave\n" ); | ||
217 | } | ||
218 | } | ||
219 | |||
220 | DEBUGF( "end at %08lx (wave #%d)\n", | ||
221 | (unsigned long)(addr - RAM), raw_voice->waveform ); | ||
222 | |||
223 | /* add to cache */ | ||
224 | this->wave_entry_old [this->oldsize++] = *wave_entry; | ||
225 | wave_in_cache:; | ||
226 | } | ||
227 | } | ||
228 | #endif | ||
229 | |||
230 | static void key_on(struct Spc_Dsp* const this, struct voice_t* const voice, | ||
231 | struct src_dir const* const sd, | ||
232 | struct raw_voice_t const* const raw_voice, | ||
233 | const int key_on_delay, const int vbit) ICODE_ATTR_SPC; | ||
234 | static void key_on(struct Spc_Dsp* const this, struct voice_t* const voice, | ||
235 | struct src_dir const* const sd, | ||
236 | struct raw_voice_t const* const raw_voice, | ||
237 | const int key_on_delay, const int vbit) { | ||
238 | #undef RAM | ||
239 | #define RAM ram.ram | ||
240 | int const env_rate_init = 0x7800; | ||
241 | voice->key_on_delay = key_on_delay; | ||
242 | if ( key_on_delay == 0 ) | ||
243 | { | ||
244 | this->keys_down |= vbit; | ||
245 | voice->envx = 0; | ||
246 | voice->env_mode = state_attack; | ||
247 | voice->env_timer = env_rate_init; /* TODO: inaccurate? */ | ||
248 | unsigned start_addr = letoh16(sd[raw_voice->waveform].start); | ||
249 | #if !SPC_BRRCACHE | ||
250 | { | ||
251 | voice->addr = RAM + start_addr; | ||
252 | /* BRR filter uses previous samples */ | ||
253 | voice->samples [BRR_BLOCK_SIZE + 1] = 0; | ||
254 | voice->samples [BRR_BLOCK_SIZE + 2] = 0; | ||
255 | /* decode three samples immediately */ | ||
256 | voice->position = (BRR_BLOCK_SIZE + 3) * 0x1000 - 1; | ||
257 | voice->block_header = 0; /* "previous" BRR header */ | ||
258 | } | ||
259 | #else | ||
260 | { | ||
261 | voice->position = 3 * 0x1000 - 1; | ||
262 | struct cache_entry_t* const wave_entry = | ||
263 | &this->wave_entry [raw_voice->waveform]; | ||
264 | |||
265 | /* predecode BRR if not already */ | ||
266 | if ( wave_entry->start_addr != start_addr ) | ||
267 | { | ||
268 | /* the following line can be replaced by the indicated block | ||
269 | in decode_brr() */ | ||
270 | decode_brr( this, start_addr, voice, raw_voice ); | ||
271 | } | ||
272 | |||
273 | voice->samples = wave_entry->samples; | ||
274 | voice->wave_end = wave_entry->end; | ||
275 | voice->wave_loop = wave_entry->loop; | ||
276 | } | ||
277 | #endif | ||
278 | } | ||
279 | } | ||
280 | |||
281 | void DSP_run_( struct Spc_Dsp* this, long count, int32_t* out_buf ) | ||
282 | { | ||
283 | #undef RAM | ||
284 | #if defined(CPU_ARM) && !SPC_BRRCACHE | ||
285 | uint8_t* const ram_ = ram.ram; | ||
286 | #define RAM ram_ | ||
287 | #else | ||
288 | #define RAM ram.ram | ||
289 | #endif | ||
290 | #if 0 | ||
291 | EXIT_TIMER(cpu); | ||
292 | ENTER_TIMER(dsp); | ||
293 | #endif | ||
294 | |||
295 | /* Here we check for keys on/off. Docs say that successive writes | ||
296 | to KON/KOF must be separated by at least 2 Ts periods or risk | ||
297 | being neglected. Therefore DSP only looks at these during an | ||
298 | update, and not at the time of the write. Only need to do this | ||
299 | once however, since the regs haven't changed over the whole | ||
300 | period we need to catch up with. */ | ||
301 | |||
302 | { | ||
303 | int key_ons = this->r.g.key_ons; | ||
304 | int key_offs = this->r.g.key_offs; | ||
305 | /* keying on a voice resets that bit in ENDX */ | ||
306 | this->r.g.wave_ended &= ~key_ons; | ||
307 | /* key_off bits prevent key_on from being acknowledged */ | ||
308 | this->r.g.key_ons = key_ons & key_offs; | ||
309 | |||
310 | /* process key events outside loop, since they won't re-occur */ | ||
311 | struct voice_t* voice = this->voice_state + 8; | ||
312 | int vbit = 0x80; | ||
313 | do | ||
314 | { | ||
315 | --voice; | ||
316 | if ( key_offs & vbit ) | ||
317 | { | ||
318 | voice->env_mode = state_release; | ||
319 | voice->key_on_delay = 0; | ||
320 | } | ||
321 | else if ( key_ons & vbit ) | ||
322 | { | ||
323 | voice->key_on_delay = 8; | ||
324 | } | ||
325 | } | ||
326 | while ( (vbit >>= 1) != 0 ); | ||
327 | } | ||
328 | |||
329 | struct src_dir const* const sd = | ||
330 | &ram.sd[this->r.g.wave_page * 0x100/sizeof(struct src_dir)]; | ||
331 | |||
332 | #ifdef ROCKBOX_BIG_ENDIAN | ||
333 | /* Convert endiannesses before entering loops - these | ||
334 | get used alot */ | ||
335 | const uint32_t rates[VOICE_COUNT] = | ||
336 | { | ||
337 | GET_LE16A( this->r.voice[0].rate ) & 0x3FFF, | ||
338 | GET_LE16A( this->r.voice[1].rate ) & 0x3FFF, | ||
339 | GET_LE16A( this->r.voice[2].rate ) & 0x3FFF, | ||
340 | GET_LE16A( this->r.voice[3].rate ) & 0x3FFF, | ||
341 | GET_LE16A( this->r.voice[4].rate ) & 0x3FFF, | ||
342 | GET_LE16A( this->r.voice[5].rate ) & 0x3FFF, | ||
343 | GET_LE16A( this->r.voice[6].rate ) & 0x3FFF, | ||
344 | GET_LE16A( this->r.voice[7].rate ) & 0x3FFF, | ||
345 | }; | ||
346 | #define VOICE_RATE(x) *(x) | ||
347 | #define IF_RBE(...) __VA_ARGS__ | ||
348 | #ifdef CPU_COLDFIRE | ||
349 | /* Initialize mask register with the buffer address mask */ | ||
350 | asm volatile ("move.l %[m], %%mask" : : [m]"i"(FIR_BUF_MASK)); | ||
351 | const int echo_wrap = (this->r.g.echo_delay & 15) * 0x800; | ||
352 | const int echo_start = this->r.g.echo_page * 0x100; | ||
353 | #endif /* CPU_COLDFIRE */ | ||
354 | #else | ||
355 | #define VOICE_RATE(x) (GET_LE16(raw_voice->rate) & 0x3FFF) | ||
356 | #define IF_RBE(...) | ||
357 | #endif /* ROCKBOX_BIG_ENDIAN */ | ||
358 | |||
359 | #if !SPC_NOINTERP | ||
360 | int const slow_gaussian = (this->r.g.pitch_mods >> 1) | | ||
361 | this->r.g.noise_enables; | ||
362 | #endif | ||
363 | /* (g.flags & 0x40) ? 30 : 14 */ | ||
364 | int const global_muting = ((this->r.g.flags & 0x40) >> 2) + 14 - 8; | ||
365 | int const global_vol_0 = this->r.g.volume_0; | ||
366 | int const global_vol_1 = this->r.g.volume_1; | ||
367 | |||
368 | /* each rate divides exactly into 0x7800 without remainder */ | ||
369 | int const env_rate_init = 0x7800; | ||
370 | static unsigned short const env_rates [0x20] ICONST_ATTR_SPC = | ||
371 | { | ||
372 | 0x0000, 0x000F, 0x0014, 0x0018, 0x001E, 0x0028, 0x0030, 0x003C, | ||
373 | 0x0050, 0x0060, 0x0078, 0x00A0, 0x00C0, 0x00F0, 0x0140, 0x0180, | ||
374 | 0x01E0, 0x0280, 0x0300, 0x03C0, 0x0500, 0x0600, 0x0780, 0x0A00, | ||
375 | 0x0C00, 0x0F00, 0x1400, 0x1800, 0x1E00, 0x2800, 0x3C00, 0x7800 | ||
376 | }; | ||
377 | |||
378 | do /* one pair of output samples per iteration */ | ||
379 | { | ||
380 | /* Noise */ | ||
381 | if ( this->r.g.noise_enables ) | ||
382 | { | ||
383 | if ( (this->noise_count -= | ||
384 | env_rates [this->r.g.flags & 0x1F]) <= 0 ) | ||
385 | { | ||
386 | this->noise_count = env_rate_init; | ||
387 | int feedback = (this->noise << 13) ^ (this->noise << 14); | ||
388 | this->noise = (feedback & 0x8000) ^ (this->noise >> 1 & ~1); | ||
389 | } | ||
390 | } | ||
391 | |||
392 | #if !SPC_NOECHO | ||
393 | int echo_0 = 0; | ||
394 | int echo_1 = 0; | ||
395 | #endif | ||
396 | long prev_outx = 0; /* TODO: correct value for first channel? */ | ||
397 | int chans_0 = 0; | ||
398 | int chans_1 = 0; | ||
399 | /* TODO: put raw_voice pointer in voice_t? */ | ||
400 | struct raw_voice_t * raw_voice = this->r.voice; | ||
401 | struct voice_t* voice = this->voice_state; | ||
402 | int vbit = 1; | ||
403 | IF_RBE( const uint32_t* vr = rates; ) | ||
404 | for ( ; vbit < 0x100; vbit <<= 1, ++voice, ++raw_voice IF_RBE( , ++vr ) ) | ||
405 | { | ||
406 | /* pregen involves checking keyon, etc */ | ||
407 | #if 0 | ||
408 | ENTER_TIMER(dsp_pregen); | ||
409 | #endif | ||
410 | |||
411 | /* Key on events are delayed */ | ||
412 | int key_on_delay = voice->key_on_delay; | ||
413 | |||
414 | if ( UNLIKELY ( --key_on_delay >= 0 ) ) /* <1% of the time */ | ||
415 | { | ||
416 | key_on(this,voice,sd,raw_voice,key_on_delay,vbit); | ||
417 | } | ||
418 | |||
419 | if ( !(this->keys_down & vbit) ) /* Silent channel */ | ||
420 | { | ||
421 | silent_chan: | ||
422 | raw_voice->envx = 0; | ||
423 | raw_voice->outx = 0; | ||
424 | prev_outx = 0; | ||
425 | continue; | ||
426 | } | ||
427 | |||
428 | /* Envelope */ | ||
429 | { | ||
430 | int const ENV_RANGE = 0x800; | ||
431 | int env_mode = voice->env_mode; | ||
432 | int adsr0 = raw_voice->adsr [0]; | ||
433 | int env_timer; | ||
434 | if ( LIKELY ( env_mode != state_release ) ) /* 99% of the time */ | ||
435 | { | ||
436 | env_timer = voice->env_timer; | ||
437 | if ( LIKELY ( adsr0 & 0x80 ) ) /* 79% of the time */ | ||
438 | { | ||
439 | int adsr1 = raw_voice->adsr [1]; | ||
440 | if ( LIKELY ( env_mode == state_sustain ) ) /* 74% of the time */ | ||
441 | { | ||
442 | if ( (env_timer -= env_rates [adsr1 & 0x1F]) > 0 ) | ||
443 | goto write_env_timer; | ||
444 | |||
445 | int envx = voice->envx; | ||
446 | envx--; /* envx *= 255 / 256 */ | ||
447 | envx -= envx >> 8; | ||
448 | voice->envx = envx; | ||
449 | /* TODO: should this be 8? */ | ||
450 | raw_voice->envx = envx >> 4; | ||
451 | goto init_env_timer; | ||
452 | } | ||
453 | else if ( env_mode < 0 ) /* 25% state_decay */ | ||
454 | { | ||
455 | int envx = voice->envx; | ||
456 | if ( (env_timer -= | ||
457 | env_rates [(adsr0 >> 3 & 0x0E) + 0x10]) <= 0 ) | ||
458 | { | ||
459 | envx--; /* envx *= 255 / 256 */ | ||
460 | envx -= envx >> 8; | ||
461 | voice->envx = envx; | ||
462 | /* TODO: should this be 8? */ | ||
463 | raw_voice->envx = envx >> 4; | ||
464 | env_timer = env_rate_init; | ||
465 | } | ||
466 | |||
467 | int sustain_level = adsr1 >> 5; | ||
468 | if ( envx <= (sustain_level + 1) * 0x100 ) | ||
469 | voice->env_mode = state_sustain; | ||
470 | |||
471 | goto write_env_timer; | ||
472 | } | ||
473 | else /* state_attack */ | ||
474 | { | ||
475 | int t = adsr0 & 0x0F; | ||
476 | if ( (env_timer -= env_rates [t * 2 + 1]) > 0 ) | ||
477 | goto write_env_timer; | ||
478 | |||
479 | int envx = voice->envx; | ||
480 | |||
481 | int const step = ENV_RANGE / 64; | ||
482 | envx += step; | ||
483 | if ( t == 15 ) | ||
484 | envx += ENV_RANGE / 2 - step; | ||
485 | |||
486 | if ( envx >= ENV_RANGE ) | ||
487 | { | ||
488 | envx = ENV_RANGE - 1; | ||
489 | voice->env_mode = state_decay; | ||
490 | } | ||
491 | voice->envx = envx; | ||
492 | /* TODO: should this be 8? */ | ||
493 | raw_voice->envx = envx >> 4; | ||
494 | goto init_env_timer; | ||
495 | } | ||
496 | } | ||
497 | else /* gain mode */ | ||
498 | { | ||
499 | int t = raw_voice->gain; | ||
500 | if ( t < 0x80 ) | ||
501 | { | ||
502 | raw_voice->envx = t; | ||
503 | voice->envx = t << 4; | ||
504 | goto env_end; | ||
505 | } | ||
506 | else | ||
507 | { | ||
508 | if ( (env_timer -= env_rates [t & 0x1F]) > 0 ) | ||
509 | goto write_env_timer; | ||
510 | |||
511 | int envx = voice->envx; | ||
512 | int mode = t >> 5; | ||
513 | if ( mode <= 5 ) /* decay */ | ||
514 | { | ||
515 | int step = ENV_RANGE / 64; | ||
516 | if ( mode == 5 ) /* exponential */ | ||
517 | { | ||
518 | envx--; /* envx *= 255 / 256 */ | ||
519 | step = envx >> 8; | ||
520 | } | ||
521 | if ( (envx -= step) < 0 ) | ||
522 | { | ||
523 | envx = 0; | ||
524 | if ( voice->env_mode == state_attack ) | ||
525 | voice->env_mode = state_decay; | ||
526 | } | ||
527 | } | ||
528 | else /* attack */ | ||
529 | { | ||
530 | int const step = ENV_RANGE / 64; | ||
531 | envx += step; | ||
532 | if ( mode == 7 && | ||
533 | envx >= ENV_RANGE * 3 / 4 + step ) | ||
534 | envx += ENV_RANGE / 256 - step; | ||
535 | |||
536 | if ( envx >= ENV_RANGE ) | ||
537 | envx = ENV_RANGE - 1; | ||
538 | } | ||
539 | voice->envx = envx; | ||
540 | /* TODO: should this be 8? */ | ||
541 | raw_voice->envx = envx >> 4; | ||
542 | goto init_env_timer; | ||
543 | } | ||
544 | } | ||
545 | } | ||
546 | else /* state_release */ | ||
547 | { | ||
548 | int envx = voice->envx; | ||
549 | if ( (envx -= ENV_RANGE / 256) > 0 ) | ||
550 | { | ||
551 | voice->envx = envx; | ||
552 | raw_voice->envx = envx >> 8; | ||
553 | goto env_end; | ||
554 | } | ||
555 | else | ||
556 | { | ||
557 | /* bit was set, so this clears it */ | ||
558 | this->keys_down ^= vbit; | ||
559 | voice->envx = 0; | ||
560 | goto silent_chan; | ||
561 | } | ||
562 | } | ||
563 | init_env_timer: | ||
564 | env_timer = env_rate_init; | ||
565 | write_env_timer: | ||
566 | voice->env_timer = env_timer; | ||
567 | env_end:; | ||
568 | } | ||
569 | #if 0 | ||
570 | EXIT_TIMER(dsp_pregen); | ||
571 | |||
572 | ENTER_TIMER(dsp_gen); | ||
573 | #endif | ||
574 | #if !SPC_BRRCACHE | ||
575 | /* Decode BRR block */ | ||
576 | if ( voice->position >= BRR_BLOCK_SIZE * 0x1000 ) | ||
577 | { | ||
578 | voice->position -= BRR_BLOCK_SIZE * 0x1000; | ||
579 | |||
580 | uint8_t const* addr = voice->addr; | ||
581 | if ( addr >= RAM + 0x10000 ) | ||
582 | addr -= 0x10000; | ||
583 | |||
584 | /* action based on previous block's header */ | ||
585 | if ( voice->block_header & 1 ) | ||
586 | { | ||
587 | addr = RAM + letoh16(sd[raw_voice->waveform].loop); | ||
588 | this->r.g.wave_ended |= vbit; | ||
589 | if ( !(voice->block_header & 2) ) /* 1% of the time */ | ||
590 | { | ||
591 | /* first block was end block; | ||
592 | don't play anything (verified) */ | ||
593 | /* bit was set, so this clears it */ | ||
594 | this->keys_down ^= vbit; | ||
595 | |||
596 | /* since voice->envx is 0, | ||
597 | samples and position don't matter */ | ||
598 | raw_voice->envx = 0; | ||
599 | voice->envx = 0; | ||
600 | goto skip_decode; | ||
601 | } | ||
602 | } | ||
603 | |||
604 | /* header */ | ||
605 | int const block_header = *addr; | ||
606 | addr += 9; | ||
607 | voice->addr = addr; | ||
608 | voice->block_header = block_header; | ||
609 | |||
610 | /* previous samples */ | ||
611 | int smp2 = voice->samples [BRR_BLOCK_SIZE + 1]; | ||
612 | int smp1 = voice->samples [BRR_BLOCK_SIZE + 2]; | ||
613 | voice->samples [0] = voice->samples [BRR_BLOCK_SIZE]; | ||
614 | |||
615 | /* output position */ | ||
616 | short* out = voice->samples + (1 + BRR_BLOCK_SIZE); | ||
617 | int offset = -BRR_BLOCK_SIZE << 2; | ||
618 | |||
619 | /* if next block has end flag set, | ||
620 | this block ends early (verified) */ | ||
621 | if ( (block_header & 3) != 3 && (*addr & 3) == 1 ) | ||
622 | { | ||
623 | /* arrange for last 9 samples to be skipped */ | ||
624 | int const skip = 9; | ||
625 | out += (skip & 1); | ||
626 | voice->samples [skip] = voice->samples [BRR_BLOCK_SIZE]; | ||
627 | voice->position += skip * 0x1000; | ||
628 | offset = (-BRR_BLOCK_SIZE + (skip & ~1)) << 2; | ||
629 | addr -= skip / 2; | ||
630 | /* force sample to end on next decode */ | ||
631 | voice->block_header = 1; | ||
632 | } | ||
633 | |||
634 | int const filter = block_header & 0x0c; | ||
635 | int const scale = block_header >> 4; | ||
636 | |||
637 | if ( filter == 0x08 ) /* filter 2 (30-90% of the time) */ | ||
638 | { | ||
639 | /* y[n] = x[n] + 61/32 * y[n-1] - 15/16 * y[n-2] */ | ||
640 | do /* decode and filter 16 samples */ | ||
641 | { | ||
642 | /* Get nybble, sign-extend, then scale | ||
643 | get byte, select which nybble, sign-extend, then shift | ||
644 | based on scaling. */ | ||
645 | int delta = (int8_t)(addr [offset >> 3] << (offset & 4)) >> 4; | ||
646 | delta = (delta << scale) >> 1; | ||
647 | |||
648 | if (scale > 0xc) | ||
649 | delta = (delta >> 17) << 11; | ||
650 | |||
651 | out [offset >> 2] = smp2; | ||
652 | |||
653 | delta -= smp2 >> 1; | ||
654 | delta += smp2 >> 5; | ||
655 | delta += smp1; | ||
656 | delta += (-smp1 - (smp1 >> 1)) >> 5; | ||
657 | |||
658 | delta = CLAMP16( delta ); | ||
659 | smp2 = smp1; | ||
660 | smp1 = (int16_t) (delta * 2); /* sign-extend */ | ||
661 | } | ||
662 | while ( (offset += 4) != 0 ); | ||
663 | } | ||
664 | else if ( filter == 0x04 ) /* filter 1 */ | ||
665 | { | ||
666 | /* y[n] = x[n] + 15/16 * y[n-1] */ | ||
667 | do /* decode and filter 16 samples */ | ||
668 | { | ||
669 | /* Get nybble, sign-extend, then scale | ||
670 | get byte, select which nybble, sign-extend, then shift | ||
671 | based on scaling. */ | ||
672 | int delta = (int8_t)(addr [offset >> 3] << (offset & 4)) >> 4; | ||
673 | delta = (delta << scale) >> 1; | ||
674 | |||
675 | if (scale > 0xc) | ||
676 | delta = (delta >> 17) << 11; | ||
677 | |||
678 | out [offset >> 2] = smp2; | ||
679 | |||
680 | delta += smp1 >> 1; | ||
681 | delta += (-smp1) >> 5; | ||
682 | |||
683 | delta = CLAMP16( delta ); | ||
684 | smp2 = smp1; | ||
685 | smp1 = (int16_t) (delta * 2); /* sign-extend */ | ||
686 | } | ||
687 | while ( (offset += 4) != 0 ); | ||
688 | } | ||
689 | else if ( filter == 0x0c ) /* filter 3 */ | ||
690 | { | ||
691 | /* y[n] = x[n] + 115/64 * y[n-1] - 13/16 * y[n-2] */ | ||
692 | do /* decode and filter 16 samples */ | ||
693 | { | ||
694 | /* Get nybble, sign-extend, then scale | ||
695 | get byte, select which nybble, sign-extend, then shift | ||
696 | based on scaling. */ | ||
697 | int delta = (int8_t)(addr [offset >> 3] << (offset & 4)) >> 4; | ||
698 | delta = (delta << scale) >> 1; | ||
699 | |||
700 | if (scale > 0xc) | ||
701 | delta = (delta >> 17) << 11; | ||
702 | |||
703 | out [offset >> 2] = smp2; | ||
704 | |||
705 | delta -= smp2 >> 1; | ||
706 | delta += (smp2 + (smp2 >> 1)) >> 4; | ||
707 | delta += smp1; | ||
708 | delta += (-smp1 * 13) >> 7; | ||
709 | |||
710 | delta = CLAMP16( delta ); | ||
711 | smp2 = smp1; | ||
712 | smp1 = (int16_t) (delta * 2); /* sign-extend */ | ||
713 | } | ||
714 | while ( (offset += 4) != 0 ); | ||
715 | } | ||
716 | else /* filter 0 */ | ||
717 | { | ||
718 | /* y[n] = x[n] */ | ||
719 | do /* decode and filter 16 samples */ | ||
720 | { | ||
721 | /* Get nybble, sign-extend, then scale | ||
722 | get byte, select which nybble, sign-extend, then shift | ||
723 | based on scaling. */ | ||
724 | int delta = (int8_t)(addr [offset >> 3] << (offset & 4)) >> 4; | ||
725 | delta = (delta << scale) >> 1; | ||
726 | |||
727 | if (scale > 0xc) | ||
728 | delta = (delta >> 17) << 11; | ||
729 | |||
730 | out [offset >> 2] = smp2; | ||
731 | |||
732 | smp2 = smp1; | ||
733 | smp1 = delta * 2; | ||
734 | } | ||
735 | while ( (offset += 4) != 0 ); | ||
736 | } | ||
737 | |||
738 | out [0] = smp2; | ||
739 | out [1] = smp1; | ||
740 | |||
741 | skip_decode:; | ||
742 | } | ||
743 | #endif /* !SPC_BRRCACHE */ | ||
744 | /* Get rate (with possible modulation) */ | ||
745 | int rate = VOICE_RATE(vr); | ||
746 | if ( this->r.g.pitch_mods & vbit ) | ||
747 | rate = (rate * (prev_outx + 32768)) >> 15; | ||
748 | |||
749 | #if !SPC_NOINTERP | ||
750 | /* Interleved gauss table (to improve cache coherency). */ | ||
751 | /* gauss [i * 2 + j] = normal_gauss [(1 - j) * 256 + i] */ | ||
752 | static short const gauss [512] ICONST_ATTR_SPC MEM_ALIGN_ATTR = | ||
753 | { | ||
754 | 370,1305, 366,1305, 362,1304, 358,1304, 354,1304, 351,1304, 347,1304, 343,1303, | ||
755 | 339,1303, 336,1303, 332,1302, 328,1302, 325,1301, 321,1300, 318,1300, 314,1299, | ||
756 | 311,1298, 307,1297, 304,1297, 300,1296, 297,1295, 293,1294, 290,1293, 286,1292, | ||
757 | 283,1291, 280,1290, 276,1288, 273,1287, 270,1286, 267,1284, 263,1283, 260,1282, | ||
758 | 257,1280, 254,1279, 251,1277, 248,1275, 245,1274, 242,1272, 239,1270, 236,1269, | ||
759 | 233,1267, 230,1265, 227,1263, 224,1261, 221,1259, 218,1257, 215,1255, 212,1253, | ||
760 | 210,1251, 207,1248, 204,1246, 201,1244, 199,1241, 196,1239, 193,1237, 191,1234, | ||
761 | 188,1232, 186,1229, 183,1227, 180,1224, 178,1221, 175,1219, 173,1216, 171,1213, | ||
762 | 168,1210, 166,1207, 163,1205, 161,1202, 159,1199, 156,1196, 154,1193, 152,1190, | ||
763 | 150,1186, 147,1183, 145,1180, 143,1177, 141,1174, 139,1170, 137,1167, 134,1164, | ||
764 | 132,1160, 130,1157, 128,1153, 126,1150, 124,1146, 122,1143, 120,1139, 118,1136, | ||
765 | 117,1132, 115,1128, 113,1125, 111,1121, 109,1117, 107,1113, 106,1109, 104,1106, | ||
766 | 102,1102, 100,1098, 99,1094, 97,1090, 95,1086, 94,1082, 92,1078, 90,1074, | ||
767 | 89,1070, 87,1066, 86,1061, 84,1057, 83,1053, 81,1049, 80,1045, 78,1040, | ||
768 | 77,1036, 76,1032, 74,1027, 73,1023, 71,1019, 70,1014, 69,1010, 67,1005, | ||
769 | 66,1001, 65, 997, 64, 992, 62, 988, 61, 983, 60, 978, 59, 974, 58, 969, | ||
770 | 56, 965, 55, 960, 54, 955, 53, 951, 52, 946, 51, 941, 50, 937, 49, 932, | ||
771 | 48, 927, 47, 923, 46, 918, 45, 913, 44, 908, 43, 904, 42, 899, 41, 894, | ||
772 | 40, 889, 39, 884, 38, 880, 37, 875, 36, 870, 36, 865, 35, 860, 34, 855, | ||
773 | 33, 851, 32, 846, 32, 841, 31, 836, 30, 831, 29, 826, 29, 821, 28, 816, | ||
774 | 27, 811, 27, 806, 26, 802, 25, 797, 24, 792, 24, 787, 23, 782, 23, 777, | ||
775 | 22, 772, 21, 767, 21, 762, 20, 757, 20, 752, 19, 747, 19, 742, 18, 737, | ||
776 | 17, 732, 17, 728, 16, 723, 16, 718, 15, 713, 15, 708, 15, 703, 14, 698, | ||
777 | 14, 693, 13, 688, 13, 683, 12, 678, 12, 674, 11, 669, 11, 664, 11, 659, | ||
778 | 10, 654, 10, 649, 10, 644, 9, 640, 9, 635, 9, 630, 8, 625, 8, 620, | ||
779 | 8, 615, 7, 611, 7, 606, 7, 601, 6, 596, 6, 592, 6, 587, 6, 582, | ||
780 | 5, 577, 5, 573, 5, 568, 5, 563, 4, 559, 4, 554, 4, 550, 4, 545, | ||
781 | 4, 540, 3, 536, 3, 531, 3, 527, 3, 522, 3, 517, 2, 513, 2, 508, | ||
782 | 2, 504, 2, 499, 2, 495, 2, 491, 2, 486, 1, 482, 1, 477, 1, 473, | ||
783 | 1, 469, 1, 464, 1, 460, 1, 456, 1, 451, 1, 447, 1, 443, 1, 439, | ||
784 | 0, 434, 0, 430, 0, 426, 0, 422, 0, 418, 0, 414, 0, 410, 0, 405, | ||
785 | 0, 401, 0, 397, 0, 393, 0, 389, 0, 385, 0, 381, 0, 378, 0, 374, | ||
786 | }; | ||
787 | /* Gaussian interpolation using most recent 4 samples */ | ||
788 | long position = voice->position; | ||
789 | voice->position += rate; | ||
790 | short const* interp = voice->samples + (position >> 12); | ||
791 | int offset = position >> 4 & 0xFF; | ||
792 | |||
793 | /* Only left half of gaussian kernel is in table, so we must mirror | ||
794 | for right half */ | ||
795 | short const* fwd = gauss + offset * 2; | ||
796 | short const* rev = gauss + 510 - offset * 2; | ||
797 | |||
798 | /* Use faster gaussian interpolation when exact result isn't needed | ||
799 | by pitch modulator of next channel */ | ||
800 | int amp_0, amp_1; /* Also serve as temps _0, and _1 */ | ||
801 | if ( LIKELY ( !(slow_gaussian & vbit) ) ) /* 99% of the time */ | ||
802 | { | ||
803 | /* Main optimization is lack of clamping. Not a problem since | ||
804 | output never goes more than +/- 16 outside 16-bit range and | ||
805 | things are clamped later anyway. Other optimization is to | ||
806 | preserve fractional accuracy, eliminating several masks. */ | ||
807 | #if defined (CPU_ARM) | ||
808 | int output; | ||
809 | int _2, _3; /* All-purpose temps */ | ||
810 | /* Multiple ASM blocks keep regs free and reduce result | ||
811 | * latency issues. */ | ||
812 | #if ARM_ARCH >= 6 | ||
813 | /* Interpolate */ | ||
814 | asm volatile ( | ||
815 | "ldr %[_0], [%[interp]] \r\n" /* _0=i0i1 */ | ||
816 | "ldr %[_2], [%[fwd]] \r\n" /* _2=f0f1 */ | ||
817 | "ldr %[_1], [%[interp], #4] \r\n" /* _1=i2i3 */ | ||
818 | "ldr %[_3], [%[rev]] \r\n" /* _3=r0r1 */ | ||
819 | "smuad %[out], %[_0], %[_2] \r\n" /* out=f0*i0 + f1*i1 */ | ||
820 | "smladx %[out], %[_1], %[_3], %[out] \r\n" /* out+=r1*i2 + r0*i3 */ | ||
821 | : [out]"=r"(output), | ||
822 | [_0]"=&r"(amp_0), [_1]"=&r"(amp_1), | ||
823 | [_2]"=&r"(_2), [_3]"=r"(_3) | ||
824 | : [fwd]"r"(fwd), [rev]"r"(rev), | ||
825 | [interp]"r"(interp)); | ||
826 | /* Apply voice envelope */ | ||
827 | asm volatile ( | ||
828 | "mov %[_2], %[out], asr #(11-5) \r\n" /* To do >> 16 later */ | ||
829 | "mul %[out], %[_2], %[envx] \r\n" /* and avoid exp. shift */ | ||
830 | : [out]"+r"(output), [_2]"=&r"(_2) | ||
831 | : [envx]"r"((int)voice->envx)); | ||
832 | /* Apply left and right volume */ | ||
833 | asm volatile ( | ||
834 | "smulwb %[amp_0], %[out], %[vvol_0] \r\n" /* (32x16->48)[47:16]->[31:0] */ | ||
835 | "smulwb %[amp_1], %[out], %[vvol_1] \r\n" | ||
836 | : [out]"+r"(output), | ||
837 | [amp_0]"=&r"(amp_0), [amp_1]"=r"(amp_1) | ||
838 | : [vvol_0]"r"(voice->volume[0]), | ||
839 | [vvol_1]"r"(voice->volume[1])); | ||
840 | |||
841 | raw_voice->outx = output >> (8+5); /* 'output' still 5 bits too big */ | ||
842 | #else /* ARM_ARCH < 6 */ | ||
843 | /* Perform gaussian interpolation on four samples */ | ||
844 | asm volatile ( | ||
845 | "ldrsh %[_0], [%[interp]] \r\n" | ||
846 | "ldrsh %[_2], [%[fwd]] \r\n" | ||
847 | "ldrsh %[_1], [%[interp], #2] \r\n" | ||
848 | "ldrsh %[_3], [%[fwd], #2] \r\n" | ||
849 | "mul %[out], %[_0], %[_2] \r\n" /* out= fwd[0]*interp[0] */ | ||
850 | "ldrsh %[_0], [%[interp], #4] \r\n" | ||
851 | "ldrsh %[_2], [%[rev], #2] \r\n" | ||
852 | "mla %[out], %[_1], %[_3], %[out] \r\n" /* out+=fwd[1]*interp[1] */ | ||
853 | "ldrsh %[_1], [%[interp], #6] \r\n" | ||
854 | "ldrsh %[_3], [%[rev]] \r\n" | ||
855 | "mla %[out], %[_0], %[_2], %[out] \r\n" /* out+=rev[1]*interp[2] */ | ||
856 | "mla %[out], %[_1], %[_3], %[out] \r\n" /* out+=rev[0]*interp[3] */ | ||
857 | : [out]"=&r"(output), | ||
858 | [_0]"=&r"(amp_0), [_1]"=&r"(amp_1), | ||
859 | [_2]"=&r"(_2), [_3]"=&r"(_3) | ||
860 | : [fwd]"r"(fwd), [rev]"r"(rev), | ||
861 | [interp]"r"(interp)); | ||
862 | /* Apply voice envelope */ | ||
863 | asm volatile ( | ||
864 | "mov %[_2], %[out], asr #11 \r\n" | ||
865 | "mul %[out], %[_2], %[envx] \r\n" | ||
866 | : [out]"+r"(output), [_2]"=&r"(_2) | ||
867 | : [envx]"r"((int)voice->envx)); | ||
868 | /* Reduce and apply left and right volume */ | ||
869 | asm volatile ( | ||
870 | "mov %[out], %[out], asr #11 \r\n" | ||
871 | "mul %[amp_0], %[out], %[vvol_0] \r\n" | ||
872 | "mul %[amp_1], %[out], %[vvol_1] \r\n" | ||
873 | : [out]"+r"(output), | ||
874 | [amp_0]"=&r"(amp_0), [amp_1]"=r"(amp_1) | ||
875 | : [vvol_0]"r"((int)voice->volume[0]), | ||
876 | [vvol_1]"r"((int)voice->volume[1])); | ||
877 | |||
878 | raw_voice->outx = output >> 8; | ||
879 | #endif /* ARM_ARCH */ | ||
880 | #else /* Unoptimized CPU */ | ||
881 | int output = (((fwd [0] * interp [0] + | ||
882 | fwd [1] * interp [1] + | ||
883 | rev [1] * interp [2] + | ||
884 | rev [0] * interp [3] ) >> 11) * voice->envx) >> 11; | ||
885 | |||
886 | /* duplicated here to give compiler more to run in parallel */ | ||
887 | amp_0 = voice->volume [0] * output; | ||
888 | amp_1 = voice->volume [1] * output; | ||
889 | |||
890 | raw_voice->outx = output >> 8; | ||
891 | #endif /* CPU_* */ | ||
892 | } | ||
893 | else /* slow gaussian */ | ||
894 | { | ||
895 | #if defined(CPU_ARM) | ||
896 | #if ARM_ARCH >= 6 | ||
897 | int output = *(int16_t*) &this->noise; | ||
898 | |||
899 | if ( !(this->r.g.noise_enables & vbit) ) | ||
900 | { | ||
901 | /* Interpolate */ | ||
902 | int _2, _3; | ||
903 | asm volatile ( | ||
904 | /* NOTE: often-unaligned accesses */ | ||
905 | "ldr %[_0], [%[interp]] \r\n" /* _0=i0i1 */ | ||
906 | "ldr %[_2], [%[fwd]] \r\n" /* _2=f0f1 */ | ||
907 | "ldr %[_1], [%[interp], #4] \r\n" /* _1=i2i3 */ | ||
908 | "ldr %[_3], [%[rev]] \r\n" /* _3=f2f3 */ | ||
909 | "smulbb %[out], %[_0], %[_2] \r\n" /* out=f0*i0 */ | ||
910 | "smultt %[_0], %[_0], %[_2] \r\n" /* _0=f1*i1 */ | ||
911 | "smulbt %[_2], %[_1], %[_3] \r\n" /* _2=r1*i2 */ | ||
912 | "smultb %[_3], %[_1], %[_3] \r\n" /* _3=r0*i3 */ | ||
913 | : [out]"=r"(output), | ||
914 | [_0]"=&r"(amp_0), [_1]"=&r"(amp_1), | ||
915 | [_2]"=&r"(_2), [_3]"=r"(_3) | ||
916 | : [fwd]"r"(fwd), [rev]"r"(rev), | ||
917 | [interp]"r"(interp)); | ||
918 | asm volatile ( | ||
919 | "mov %[out], %[out], asr#12 \r\n" | ||
920 | "add %[_0], %[out], %[_0], asr #12 \r\n" | ||
921 | "add %[_2], %[_0], %[_2], asr #12 \r\n" | ||
922 | "pkhbt %[_0], %[_2], %[_3], asl #4 \r\n" /* _3[31:16], _2[15:0] */ | ||
923 | "sadd16 %[_0], %[_0], %[_0] \r\n" /* _3[31:16]*2, _2[15:0]*2 */ | ||
924 | "qsubaddx %[out], %[_0], %[_0] \r\n" /* out[15:0]= | ||
925 | * sat16(_3[31:16]+_2[15:0]) */ | ||
926 | : [out]"+r"(output), | ||
927 | [_0]"+r"(amp_0), [_2]"+r"(_2), [_3]"+r"(_3)); | ||
928 | } | ||
929 | /* Apply voice envelope */ | ||
930 | asm volatile ( | ||
931 | "smulbb %[out], %[out], %[envx] \r\n" | ||
932 | : [out]"+r"(output) | ||
933 | : [envx]"r"(voice->envx)); | ||
934 | /* Reduce and apply left and right volume */ | ||
935 | asm volatile ( | ||
936 | "mov %[out], %[out], asr #11 \r\n" | ||
937 | "bic %[out], %[out], #0x1 \r\n" | ||
938 | "mul %[amp_0], %[out], %[vvol_0] \r\n" | ||
939 | "mul %[amp_1], %[out], %[vvol_1] \r\n" | ||
940 | : [out]"+r"(output), | ||
941 | [amp_0]"=&r"(amp_0), [amp_1]"=r"(amp_1) | ||
942 | : [vvol_0]"r"((int)voice->volume[0]), | ||
943 | [vvol_1]"r"((int)voice->volume[1])); | ||
944 | |||
945 | prev_outx = output; | ||
946 | raw_voice->outx = output >> 8; | ||
947 | #else /* ARM_ARCH < 6 */ | ||
948 | int output = *(int16_t*) &this->noise; | ||
949 | |||
950 | if ( !(this->r.g.noise_enables & vbit) ) | ||
951 | { | ||
952 | /* Interpolate */ | ||
953 | int _2, _3; | ||
954 | asm volatile ( | ||
955 | "ldrsh %[_0], [%[interp]] \r\n" | ||
956 | "ldrsh %[_2], [%[fwd]] \r\n" | ||
957 | "ldrsh %[_1], [%[interp], #2] \r\n" | ||
958 | "ldrsh %[_3], [%[fwd], #2] \r\n" | ||
959 | "mul %[out], %[_2], %[_0] \r\n" /* fwd[0]*interp[0] */ | ||
960 | "ldrsh %[_2], [%[rev], #2] \r\n" | ||
961 | "mul %[_0], %[_3], %[_1] \r\n" /* fwd[1]*interp[1] */ | ||
962 | "ldrsh %[_1], [%[interp], #4] \r\n" | ||
963 | "mov %[out], %[out], asr #12 \r\n" | ||
964 | "ldrsh %[_3], [%[rev]] \r\n" | ||
965 | "mul %[_2], %[_1], %[_2] \r\n" /* rev[1]*interp[2] */ | ||
966 | "ldrsh %[_1], [%[interp], #6] \r\n" | ||
967 | "add %[_0], %[out], %[_0], asr #12 \r\n" | ||
968 | "mul %[_3], %[_1], %[_3] \r\n" /* rev[0]*interp[3] */ | ||
969 | "add %[_2], %[_0], %[_2], asr #12 \r\n" | ||
970 | "mov %[_2], %[_2], lsl #17 \r\n" | ||
971 | "mov %[_3], %[_3], asr #12 \r\n" | ||
972 | "mov %[_3], %[_3], asl #1 \r\n" | ||
973 | "add %[out], %[_3], %[_2], asr #16 \r\n" | ||
974 | : [out]"=&r"(output), | ||
975 | [_0]"=&r"(amp_0), [_1]"=&r"(amp_1), | ||
976 | [_2]"=&r"(_2), [_3]"=&r"(_3) | ||
977 | : [fwd]"r"(fwd), [rev]"r"(rev), | ||
978 | [interp]"r"(interp)); | ||
979 | |||
980 | output = CLAMP16(output); | ||
981 | } | ||
982 | /* Apply voice envelope */ | ||
983 | asm volatile ( | ||
984 | "mul %[_0], %[out], %[envx] \r\n" | ||
985 | : [_0]"=r"(amp_0) | ||
986 | : [out]"r"(output), [envx]"r"((int)voice->envx)); | ||
987 | /* Reduce and apply left and right volume */ | ||
988 | asm volatile ( | ||
989 | "mov %[out], %[amp_0], asr #11 \r\n" /* amp_0 = _0 */ | ||
990 | "bic %[out], %[out], #0x1 \r\n" | ||
991 | "mul %[amp_0], %[out], %[vvol_0] \r\n" | ||
992 | "mul %[amp_1], %[out], %[vvol_1] \r\n" | ||
993 | : [out]"+r"(output), | ||
994 | [amp_0]"+r"(amp_0), [amp_1]"=r"(amp_1) | ||
995 | : [vvol_0]"r"((int)voice->volume[0]), | ||
996 | [vvol_1]"r"((int)voice->volume[1])); | ||
997 | |||
998 | prev_outx = output; | ||
999 | raw_voice->outx = output >> 8; | ||
1000 | #endif /* ARM_ARCH >= 6 */ | ||
1001 | #else /* Unoptimized CPU */ | ||
1002 | int output = *(int16_t*) &this->noise; | ||
1003 | |||
1004 | if ( !(this->r.g.noise_enables & vbit) ) | ||
1005 | { | ||
1006 | output = (fwd [0] * interp [0]) & ~0xFFF; | ||
1007 | output = (output + fwd [1] * interp [1]) & ~0xFFF; | ||
1008 | output = (output + rev [1] * interp [2]) >> 12; | ||
1009 | output = (int16_t) (output * 2); | ||
1010 | output += ((rev [0] * interp [3]) >> 12) * 2; | ||
1011 | output = CLAMP16( output ); | ||
1012 | } | ||
1013 | output = (output * voice->envx) >> 11 & ~1; | ||
1014 | |||
1015 | /* duplicated here to give compiler more to run in parallel */ | ||
1016 | amp_0 = voice->volume [0] * output; | ||
1017 | amp_1 = voice->volume [1] * output; | ||
1018 | |||
1019 | prev_outx = output; | ||
1020 | raw_voice->outx = output >> 8; | ||
1021 | #endif /* CPU_* */ | ||
1022 | } | ||
1023 | #else /* SPCNOINTERP */ | ||
1024 | /* two-point linear interpolation */ | ||
1025 | #ifdef CPU_COLDFIRE | ||
1026 | int amp_0 = (int16_t)this->noise; | ||
1027 | int amp_1; | ||
1028 | |||
1029 | if ( (this->r.g.noise_enables & vbit) == 0 ) | ||
1030 | { | ||
1031 | uint32_t f = voice->position; | ||
1032 | int32_t y0; | ||
1033 | |||
1034 | /** | ||
1035 | * Formula (fastest found so far of MANY): | ||
1036 | * output = y0 + f*y1 - f*y0 | ||
1037 | */ | ||
1038 | asm volatile ( | ||
1039 | /* separate fractional and whole parts */ | ||
1040 | "move.l %[f], %[y1] \r\n" | ||
1041 | "and.l #0xfff, %[f] \r\n" | ||
1042 | "lsr.l %[sh], %[y1] \r\n" | ||
1043 | /* load samples y0 (upper) & y1 (lower) */ | ||
1044 | "move.l 2(%[s], %[y1].l*2), %[y1] \r\n" | ||
1045 | /* %acc0 = f*y1 */ | ||
1046 | "mac.w %[f]l, %[y1]l, %%acc0 \r\n" | ||
1047 | /* %acc0 -= f*y0 */ | ||
1048 | "msac.w %[f]l, %[y1]u, %%acc0 \r\n" | ||
1049 | /* separate out y0 and sign extend */ | ||
1050 | "swap %[y1] \r\n" | ||
1051 | "movea.w %[y1], %[y0] \r\n" | ||
1052 | /* fetch result, scale down and add y0 */ | ||
1053 | "movclr.l %%acc0, %[y1] \r\n" | ||
1054 | /* output = y0 + (result >> 12) */ | ||
1055 | "asr.l %[sh], %[y1] \r\n" | ||
1056 | "add.l %[y0], %[y1] \r\n" | ||
1057 | : [f]"+d"(f), [y0]"=&a"(y0), [y1]"=&d"(amp_0) | ||
1058 | : [s]"a"(voice->samples), [sh]"d"(12)); | ||
1059 | } | ||
1060 | |||
1061 | /* apply voice envelope to output */ | ||
1062 | asm volatile ( | ||
1063 | "mac.w %[out]l, %[envx]l, %%acc0 \r\n" | ||
1064 | : | ||
1065 | : [out]"r"(amp_0), [envx]"r"(voice->envx)); | ||
1066 | |||
1067 | /* advance voice position */ | ||
1068 | voice->position += rate; | ||
1069 | |||
1070 | /* fetch output, scale and apply left and right | ||
1071 | voice volume */ | ||
1072 | asm volatile ( | ||
1073 | "movclr.l %%acc0, %[out] \r\n" | ||
1074 | "asr.l %[sh], %[out] \r\n" | ||
1075 | "mac.l %[vvol_0], %[out], %%acc0 \r\n" | ||
1076 | "mac.l %[vvol_1], %[out], %%acc1 \r\n" | ||
1077 | : [out]"=&d"(amp_0) | ||
1078 | : [vvol_0]"r"((int)voice->volume[0]), | ||
1079 | [vvol_1]"r"((int)voice->volume[1]), | ||
1080 | [sh]"d"(11)); | ||
1081 | |||
1082 | /* save this output into previous, scale and save in | ||
1083 | output register */ | ||
1084 | prev_outx = amp_0; | ||
1085 | raw_voice->outx = amp_0 >> 8; | ||
1086 | |||
1087 | /* fetch final voice output */ | ||
1088 | asm volatile ( | ||
1089 | "movclr.l %%acc0, %[amp_0] \r\n" | ||
1090 | "movclr.l %%acc1, %[amp_1] \r\n" | ||
1091 | : [amp_0]"=r"(amp_0), [amp_1]"=r"(amp_1)); | ||
1092 | #elif defined (CPU_ARM) | ||
1093 | int amp_0, amp_1; | ||
1094 | |||
1095 | if ( (this->r.g.noise_enables & vbit) != 0 ) | ||
1096 | { | ||
1097 | amp_0 = *(int16_t *)&this->noise; | ||
1098 | } | ||
1099 | else | ||
1100 | { | ||
1101 | uint32_t f = voice->position; | ||
1102 | amp_0 = (uint32_t)voice->samples; | ||
1103 | |||
1104 | asm volatile( | ||
1105 | "mov %[y1], %[f], lsr #12 \r\n" | ||
1106 | "eor %[f], %[f], %[y1], lsl #12 \r\n" | ||
1107 | "add %[y1], %[y0], %[y1], lsl #1 \r\n" | ||
1108 | "ldrsh %[y0], [%[y1], #2] \r\n" | ||
1109 | "ldrsh %[y1], [%[y1], #4] \r\n" | ||
1110 | "sub %[y1], %[y1], %[y0] \r\n" | ||
1111 | "mul %[f], %[y1], %[f] \r\n" | ||
1112 | "add %[y0], %[y0], %[f], asr #12 \r\n" | ||
1113 | : [f]"+r"(f), [y0]"+r"(amp_0), [y1]"=&r"(amp_1)); | ||
1114 | } | ||
1115 | |||
1116 | voice->position += rate; | ||
1117 | |||
1118 | asm volatile( | ||
1119 | "mul %[amp_1], %[amp_0], %[envx] \r\n" | ||
1120 | "mov %[amp_0], %[amp_1], asr #11 \r\n" | ||
1121 | "mov %[amp_1], %[amp_0], asr #8 \r\n" | ||
1122 | : [amp_0]"+r"(amp_0), [amp_1]"=r"(amp_1) | ||
1123 | : [envx]"r"(voice->envx)); | ||
1124 | |||
1125 | prev_outx = amp_0; | ||
1126 | raw_voice->outx = (int8_t)amp_1; | ||
1127 | |||
1128 | asm volatile( | ||
1129 | "mul %[amp_1], %[amp_0], %[vol_1] \r\n" | ||
1130 | "mul %[amp_0], %[vol_0], %[amp_0] \r\n" | ||
1131 | : [amp_0]"+r"(amp_0), [amp_1]"=&r"(amp_1) | ||
1132 | : [vol_0]"r"((int)voice->volume[0]), | ||
1133 | [vol_1]"r"((int)voice->volume[1])); | ||
1134 | #else /* Unoptimized CPU */ | ||
1135 | int output; | ||
1136 | |||
1137 | if ( (this->r.g.noise_enables & vbit) == 0 ) | ||
1138 | { | ||
1139 | int const fraction = voice->position & 0xfff; | ||
1140 | short const* const pos = (voice->samples + (voice->position >> 12)) + 1; | ||
1141 | output = pos[0] + ((fraction * (pos[1] - pos[0])) >> 12); | ||
1142 | } else { | ||
1143 | output = *(int16_t *)&this->noise; | ||
1144 | } | ||
1145 | |||
1146 | voice->position += rate; | ||
1147 | |||
1148 | output = (output * voice->envx) >> 11; | ||
1149 | |||
1150 | /* duplicated here to give compiler more to run in parallel */ | ||
1151 | int amp_0 = voice->volume [0] * output; | ||
1152 | int amp_1 = voice->volume [1] * output; | ||
1153 | |||
1154 | prev_outx = output; | ||
1155 | raw_voice->outx = (int8_t) (output >> 8); | ||
1156 | #endif /* CPU_* */ | ||
1157 | #endif /* SPCNOINTERP */ | ||
1158 | |||
1159 | #if SPC_BRRCACHE | ||
1160 | if ( voice->position >= voice->wave_end ) | ||
1161 | { | ||
1162 | long loop_len = voice->wave_loop << 12; | ||
1163 | voice->position -= loop_len; | ||
1164 | this->r.g.wave_ended |= vbit; | ||
1165 | if ( !loop_len ) | ||
1166 | { | ||
1167 | this->keys_down ^= vbit; | ||
1168 | raw_voice->envx = 0; | ||
1169 | voice->envx = 0; | ||
1170 | } | ||
1171 | } | ||
1172 | #endif | ||
1173 | #if 0 | ||
1174 | EXIT_TIMER(dsp_gen); | ||
1175 | |||
1176 | ENTER_TIMER(dsp_mix); | ||
1177 | #endif | ||
1178 | chans_0 += amp_0; | ||
1179 | chans_1 += amp_1; | ||
1180 | #if !SPC_NOECHO | ||
1181 | if ( this->r.g.echo_ons & vbit ) | ||
1182 | { | ||
1183 | echo_0 += amp_0; | ||
1184 | echo_1 += amp_1; | ||
1185 | } | ||
1186 | #endif | ||
1187 | #if 0 | ||
1188 | EXIT_TIMER(dsp_mix); | ||
1189 | #endif | ||
1190 | } | ||
1191 | /* end of voice loop */ | ||
1192 | |||
1193 | #if !SPC_NOECHO | ||
1194 | #ifdef CPU_COLDFIRE | ||
1195 | /* Read feedback from echo buffer */ | ||
1196 | int echo_pos = this->echo_pos; | ||
1197 | uint8_t* const echo_ptr = RAM + ((echo_start + echo_pos) & 0xFFFF); | ||
1198 | echo_pos += 4; | ||
1199 | if ( echo_pos >= echo_wrap ) | ||
1200 | echo_pos = 0; | ||
1201 | this->echo_pos = echo_pos; | ||
1202 | int fb = swap_odd_even32(*(int32_t *)echo_ptr); | ||
1203 | int out_0, out_1; | ||
1204 | |||
1205 | /* Keep last 8 samples */ | ||
1206 | *this->last_fir_ptr = fb; | ||
1207 | this->last_fir_ptr = this->fir_ptr; | ||
1208 | |||
1209 | /* Apply echo FIR filter to output samples read from echo buffer - | ||
1210 | circular buffer is hardware incremented and masked; FIR | ||
1211 | coefficients and buffer history are loaded in parallel with | ||
1212 | multiply accumulate operations. Shift left by one here and once | ||
1213 | again when calculating feedback to have sample values justified | ||
1214 | to bit 31 in the output to ease endian swap, interleaving and | ||
1215 | clamping before placing result in the program's echo buffer. */ | ||
1216 | int _0, _1, _2; | ||
1217 | asm volatile ( | ||
1218 | "move.l (%[fir_c]) , %[_2] \r\n" | ||
1219 | "mac.w %[fb]u, %[_2]u, <<, (%[fir_p])+&, %[_0], %%acc0 \r\n" | ||
1220 | "mac.w %[fb]l, %[_2]u, <<, (%[fir_p])& , %[_1], %%acc1 \r\n" | ||
1221 | "mac.w %[_0]u, %[_2]l, << , %%acc0 \r\n" | ||
1222 | "mac.w %[_0]l, %[_2]l, <<, 4(%[fir_c]) , %[_2], %%acc1 \r\n" | ||
1223 | "mac.w %[_1]u, %[_2]u, <<, 4(%[fir_p])& , %[_0], %%acc0 \r\n" | ||
1224 | "mac.w %[_1]l, %[_2]u, <<, 8(%[fir_p])& , %[_1], %%acc1 \r\n" | ||
1225 | "mac.w %[_0]u, %[_2]l, << , %%acc0 \r\n" | ||
1226 | "mac.w %[_0]l, %[_2]l, <<, 8(%[fir_c]) , %[_2], %%acc1 \r\n" | ||
1227 | "mac.w %[_1]u, %[_2]u, <<, 12(%[fir_p])& , %[_0], %%acc0 \r\n" | ||
1228 | "mac.w %[_1]l, %[_2]u, <<, 16(%[fir_p])& , %[_1], %%acc1 \r\n" | ||
1229 | "mac.w %[_0]u, %[_2]l, << , %%acc0 \r\n" | ||
1230 | "mac.w %[_0]l, %[_2]l, <<, 12(%[fir_c]) , %[_2], %%acc1 \r\n" | ||
1231 | "mac.w %[_1]u, %[_2]u, <<, 20(%[fir_p])& , %[_0], %%acc0 \r\n" | ||
1232 | "mac.w %[_1]l, %[_2]u, << , %%acc1 \r\n" | ||
1233 | "mac.w %[_0]u, %[_2]l, << , %%acc0 \r\n" | ||
1234 | "mac.w %[_0]l, %[_2]l, << , %%acc1 \r\n" | ||
1235 | : [_0]"=&r"(_0), [_1]"=&r"(_1), [_2]"=&r"(_2), | ||
1236 | [fir_p]"+a"(this->fir_ptr) | ||
1237 | : [fir_c]"a"(this->fir_coeff), [fb]"r"(fb) | ||
1238 | ); | ||
1239 | |||
1240 | /* Generate output */ | ||
1241 | asm volatile ( | ||
1242 | /* fetch filter results _after_ gcc loads asm | ||
1243 | block parameters to eliminate emac stalls */ | ||
1244 | "movclr.l %%acc0, %[out_0] \r\n" | ||
1245 | "movclr.l %%acc1, %[out_1] \r\n" | ||
1246 | /* apply global volume */ | ||
1247 | "mac.l %[chans_0], %[gv_0] , %%acc2 \r\n" | ||
1248 | "mac.l %[chans_1], %[gv_1] , %%acc3 \r\n" | ||
1249 | /* apply echo volume and add to final output */ | ||
1250 | "mac.l %[ev_0], %[out_0], >>, %%acc2 \r\n" | ||
1251 | "mac.l %[ev_1], %[out_1], >>, %%acc3 \r\n" | ||
1252 | : [out_0]"=&r"(out_0), [out_1]"=&r"(out_1) | ||
1253 | : [chans_0]"r"(chans_0), [gv_0]"r"(global_vol_0), | ||
1254 | [ev_0]"r"((int)this->r.g.echo_volume_0), | ||
1255 | [chans_1]"r"(chans_1), [gv_1]"r"(global_vol_1), | ||
1256 | [ev_1]"r"((int)this->r.g.echo_volume_1) | ||
1257 | ); | ||
1258 | |||
1259 | /* Feedback into echo buffer */ | ||
1260 | if ( !(this->r.g.flags & 0x20) ) | ||
1261 | { | ||
1262 | int sh = 1 << 9; | ||
1263 | |||
1264 | asm volatile ( | ||
1265 | /* scale echo voices; saturate if overflow */ | ||
1266 | "mac.l %[sh], %[e1] , %%acc1 \r\n" | ||
1267 | "mac.l %[sh], %[e0] , %%acc0 \r\n" | ||
1268 | /* add scaled output from FIR filter */ | ||
1269 | "mac.l %[out_1], %[ef], <<, %%acc1 \r\n" | ||
1270 | "mac.l %[out_0], %[ef], <<, %%acc0 \r\n" | ||
1271 | /* swap and fetch feedback results - simply | ||
1272 | swap_odd_even32 mixed in between macs and | ||
1273 | movclrs to mitigate stall issues */ | ||
1274 | "move.l #0x00ff00ff, %[sh] \r\n" | ||
1275 | "movclr.l %%acc1, %[e1] \r\n" | ||
1276 | "swap %[e1] \r\n" | ||
1277 | "movclr.l %%acc0, %[e0] \r\n" | ||
1278 | "move.w %[e1], %[e0] \r\n" | ||
1279 | "and.l %[e0], %[sh] \r\n" | ||
1280 | "eor.l %[sh], %[e0] \r\n" | ||
1281 | "lsl.l #8, %[sh] \r\n" | ||
1282 | "lsr.l #8, %[e0] \r\n" | ||
1283 | "or.l %[sh], %[e0] \r\n" | ||
1284 | /* save final feedback into echo buffer */ | ||
1285 | "move.l %[e0], (%[echo_ptr]) \r\n" | ||
1286 | : [e0]"+d"(echo_0), [e1]"+d"(echo_1), [sh]"+d"(sh) | ||
1287 | : [out_0]"r"(out_0), [out_1]"r"(out_1), | ||
1288 | [ef]"r"((int)this->r.g.echo_feedback), | ||
1289 | [echo_ptr]"a"((int32_t *)echo_ptr) | ||
1290 | ); | ||
1291 | } | ||
1292 | |||
1293 | /* Output final samples */ | ||
1294 | asm volatile ( | ||
1295 | /* fetch output saved in %acc2 and %acc3 */ | ||
1296 | "movclr.l %%acc2, %[out_0] \r\n" | ||
1297 | "movclr.l %%acc3, %[out_1] \r\n" | ||
1298 | /* scale right by global_muting shift */ | ||
1299 | "asr.l %[gm], %[out_0] \r\n" | ||
1300 | "asr.l %[gm], %[out_1] \r\n" | ||
1301 | : [out_0]"=&d"(out_0), [out_1]"=&d"(out_1) | ||
1302 | : [gm]"d"(global_muting) | ||
1303 | ); | ||
1304 | |||
1305 | out_buf [ 0] = out_0; | ||
1306 | out_buf [WAV_CHUNK_SIZE] = out_1; | ||
1307 | out_buf ++; | ||
1308 | #elif defined (CPU_ARM) | ||
1309 | /* Read feedback from echo buffer */ | ||
1310 | int echo_pos = this->echo_pos; | ||
1311 | uint8_t* const echo_ptr = RAM + | ||
1312 | ((this->r.g.echo_page * 0x100 + echo_pos) & 0xFFFF); | ||
1313 | echo_pos += 4; | ||
1314 | if ( echo_pos >= (this->r.g.echo_delay & 15) * 0x800 ) | ||
1315 | echo_pos = 0; | ||
1316 | this->echo_pos = echo_pos; | ||
1317 | |||
1318 | #if ARM_ARCH >= 6 | ||
1319 | int32_t *fir_ptr, *fir_coeff; | ||
1320 | int fb_0, fb_1; | ||
1321 | |||
1322 | /* Apply FIR */ | ||
1323 | |||
1324 | /* Keep last 8 samples */ | ||
1325 | asm volatile ( | ||
1326 | "ldr %[fb_0], [%[echo_p]] \r\n" | ||
1327 | "add %[fir_p], %[t_fir_p], #4 \r\n" | ||
1328 | "bic %[t_fir_p], %[fir_p], %[mask] \r\n" | ||
1329 | "str %[fb_0], [%[fir_p], #-4] \r\n" | ||
1330 | /* duplicate at +8 eliminates wrap checking below */ | ||
1331 | "str %[fb_0], [%[fir_p], #28] \r\n" | ||
1332 | : [fir_p]"=&r"(fir_ptr), [t_fir_p]"+r"(this->fir_ptr), | ||
1333 | [fb_0]"=&r"(fb_0) | ||
1334 | : [echo_p]"r"(echo_ptr), [mask]"i"(~FIR_BUF_MASK)); | ||
1335 | |||
1336 | fir_coeff = (int32_t *)this->fir_coeff; | ||
1337 | |||
1338 | /* Fugly, but the best version found. */ | ||
1339 | int _0; | ||
1340 | asm volatile ( /* L0R0 = acc0 */ | ||
1341 | "ldmia %[fir_p]!, { r2-r5 } \r\n" /* L1R1-L4R4 = r2-r5 */ | ||
1342 | "ldmia %[fir_c]!, { r0-r1 } \r\n" /* C0C1-C2C3 = r0-r1 */ | ||
1343 | "pkhbt %[_0], %[acc0], r2, asl #16 \r\n" /* L0R0,L1R1->L0L1,R0R1 */ | ||
1344 | "pkhtb r2, r2, %[acc0], asr #16 \r\n" | ||
1345 | "smuad %[acc0], %[_0], r0 \r\n" /* acc0=L0*C0+L1*C1 */ | ||
1346 | "smuad %[acc1], r2, r0 \r\n" /* acc1=R0*C0+R1*C1 */ | ||
1347 | "pkhbt %[_0], r3, r4, asl #16 \r\n" /* L2R2,L3R3->L2L3,R2R3 */ | ||
1348 | "pkhtb r4, r4, r3, asr #16 \r\n" | ||
1349 | "smlad %[acc0], %[_0], r1, %[acc0] \r\n" /* acc0+=L2*C2+L3*C3 */ | ||
1350 | "smlad %[acc1], r4, r1, %[acc1] \r\n" /* acc1+=R2*C2+R3*C3 */ | ||
1351 | "ldmia %[fir_p], { r2-r4 } \r\n" /* L5R5-L7R7 = r2-r4 */ | ||
1352 | "ldmia %[fir_c], { r0-r1 } \r\n" /* C4C5-C6C7 = r0-r1 */ | ||
1353 | "pkhbt %[_0], r5, r2, asl #16 \r\n" /* L4R4,L5R5->L4L5,R4R5 */ | ||
1354 | "pkhtb r2, r2, r5, asr #16 \r\n" | ||
1355 | "smlad %[acc0], %[_0], r0, %[acc0] \r\n" /* acc0+=L4*C4+L5*C5 */ | ||
1356 | "smlad %[acc1], r2, r0, %[acc1] \r\n" /* acc1+=R4*C4+R5*C5 */ | ||
1357 | "pkhbt %[_0], r3, r4, asl #16 \r\n" /* L6R6,L7R7->L6L7,R6R7 */ | ||
1358 | "pkhtb r4, r4, r3, asr #16 \r\n" | ||
1359 | "smlad %[acc0], %[_0], r1, %[acc0] \r\n" /* acc0+=L6*C6+L7*C7 */ | ||
1360 | "smlad %[acc1], r4, r1, %[acc1] \r\n" /* acc1+=R6*C6+R7*C7 */ | ||
1361 | : [acc0]"+r"(fb_0), [acc1]"=&r"(fb_1), [_0]"=&r"(_0), | ||
1362 | [fir_p]"+r"(fir_ptr), [fir_c]"+r"(fir_coeff) | ||
1363 | : | ||
1364 | : "r0", "r1", "r2", "r3", "r4", "r5"); | ||
1365 | |||
1366 | /* Generate output */ | ||
1367 | int amp_0, amp_1; | ||
1368 | |||
1369 | asm volatile ( | ||
1370 | "mul %[amp_0], %[gvol_0], %[chans_0] \r\n" | ||
1371 | "mul %[amp_1], %[gvol_1], %[chans_1] \r\n" | ||
1372 | : [amp_0]"=&r"(amp_0), [amp_1]"=r"(amp_1) | ||
1373 | : [gvol_0]"r"(global_vol_0), [gvol_1]"r"(global_vol_1), | ||
1374 | [chans_0]"r"(chans_0), [chans_1]"r"(chans_1)); | ||
1375 | asm volatile ( | ||
1376 | "mla %[amp_0], %[fb_0], %[ev_0], %[amp_0] \r\n" | ||
1377 | "mla %[amp_1], %[fb_1], %[ev_1], %[amp_1] \r\n" | ||
1378 | : [amp_0]"+r"(amp_0), [amp_1]"+r"(amp_1) | ||
1379 | : [fb_0]"r"(fb_0), [fb_1]"r"(fb_1), | ||
1380 | [ev_0]"r"((int)this->r.g.echo_volume_0), | ||
1381 | [ev_1]"r"((int)this->r.g.echo_volume_1)); | ||
1382 | |||
1383 | out_buf [ 0] = amp_0 >> global_muting; | ||
1384 | out_buf [WAV_CHUNK_SIZE] = amp_1 >> global_muting; | ||
1385 | out_buf ++; | ||
1386 | |||
1387 | if ( !(this->r.g.flags & 0x20) ) | ||
1388 | { | ||
1389 | /* Feedback into echo buffer */ | ||
1390 | int e0, e1; | ||
1391 | |||
1392 | asm volatile ( | ||
1393 | "mov %[e0], %[echo_0], asl #7 \r\n" | ||
1394 | "mov %[e1], %[echo_1], asl #7 \r\n" | ||
1395 | "mla %[e0], %[fb_0], %[efb], %[e0] \r\n" | ||
1396 | "mla %[e1], %[fb_1], %[efb], %[e1] \r\n" | ||
1397 | : [e0]"=&r"(e0), [e1]"=&r"(e1) | ||
1398 | : [echo_0]"r"(echo_0), [echo_1]"r"(echo_1), | ||
1399 | [fb_0]"r"(fb_0), [fb_1]"r"(fb_1), | ||
1400 | [efb]"r"((int)this->r.g.echo_feedback)); | ||
1401 | asm volatile ( | ||
1402 | "ssat %[e0], #16, %[e0], asr #14 \r\n" | ||
1403 | "ssat %[e1], #16, %[e1], asr #14 \r\n" | ||
1404 | "pkhbt %[e0], %[e0], %[e1], lsl #16 \r\n" | ||
1405 | "str %[e0], [%[echo_p]] \r\n" | ||
1406 | : [e0]"+r"(e0), [e1]"+r"(e1) | ||
1407 | : [echo_p]"r"(echo_ptr)); | ||
1408 | } | ||
1409 | #else /* ARM_ARCH < 6 */ | ||
1410 | int fb_0 = GET_LE16SA( echo_ptr ); | ||
1411 | int fb_1 = GET_LE16SA( echo_ptr + 2 ); | ||
1412 | int32_t *fir_ptr, *fir_coeff; | ||
1413 | |||
1414 | /* Keep last 8 samples */ | ||
1415 | |||
1416 | /* Apply FIR */ | ||
1417 | asm volatile ( | ||
1418 | "add %[fir_p], %[t_fir_p], #8 \r\n" | ||
1419 | "bic %[t_fir_p], %[fir_p], %[mask] \r\n" | ||
1420 | "str %[fb_0], [%[fir_p], #-8] \r\n" | ||
1421 | "str %[fb_1], [%[fir_p], #-4] \r\n" | ||
1422 | /* duplicate at +8 eliminates wrap checking below */ | ||
1423 | "str %[fb_0], [%[fir_p], #56] \r\n" | ||
1424 | "str %[fb_1], [%[fir_p], #60] \r\n" | ||
1425 | : [fir_p]"=&r"(fir_ptr), [t_fir_p]"+r"(this->fir_ptr) | ||
1426 | : [fb_0]"r"(fb_0), [fb_1]"r"(fb_1), [mask]"i"(~FIR_BUF_MASK)); | ||
1427 | |||
1428 | fir_coeff = this->fir_coeff; | ||
1429 | |||
1430 | asm volatile ( | ||
1431 | "ldmia %[fir_c]!, { r0-r1 } \r\n" | ||
1432 | "ldmia %[fir_p]!, { r4-r5 } \r\n" | ||
1433 | "mul %[fb_0], r0, %[fb_0] \r\n" | ||
1434 | "mul %[fb_1], r0, %[fb_1] \r\n" | ||
1435 | "mla %[fb_0], r4, r1, %[fb_0] \r\n" | ||
1436 | "mla %[fb_1], r5, r1, %[fb_1] \r\n" | ||
1437 | "ldmia %[fir_c]!, { r0-r1 } \r\n" | ||
1438 | "ldmia %[fir_p]!, { r2-r5 } \r\n" | ||
1439 | "mla %[fb_0], r2, r0, %[fb_0] \r\n" | ||
1440 | "mla %[fb_1], r3, r0, %[fb_1] \r\n" | ||
1441 | "mla %[fb_0], r4, r1, %[fb_0] \r\n" | ||
1442 | "mla %[fb_1], r5, r1, %[fb_1] \r\n" | ||
1443 | "ldmia %[fir_c]!, { r0-r1 } \r\n" | ||
1444 | "ldmia %[fir_p]!, { r2-r5 } \r\n" | ||
1445 | "mla %[fb_0], r2, r0, %[fb_0] \r\n" | ||
1446 | "mla %[fb_1], r3, r0, %[fb_1] \r\n" | ||
1447 | "mla %[fb_0], r4, r1, %[fb_0] \r\n" | ||
1448 | "mla %[fb_1], r5, r1, %[fb_1] \r\n" | ||
1449 | "ldmia %[fir_c]!, { r0-r1 } \r\n" | ||
1450 | "ldmia %[fir_p]!, { r2-r5 } \r\n" | ||
1451 | "mla %[fb_0], r2, r0, %[fb_0] \r\n" | ||
1452 | "mla %[fb_1], r3, r0, %[fb_1] \r\n" | ||
1453 | "mla %[fb_0], r4, r1, %[fb_0] \r\n" | ||
1454 | "mla %[fb_1], r5, r1, %[fb_1] \r\n" | ||
1455 | : [fb_0]"+r"(fb_0), [fb_1]"+r"(fb_1), | ||
1456 | [fir_p]"+r"(fir_ptr), [fir_c]"+r"(fir_coeff) | ||
1457 | : | ||
1458 | : "r0", "r1", "r2", "r3", "r4", "r5"); | ||
1459 | |||
1460 | /* Generate output */ | ||
1461 | int amp_0 = (chans_0 * global_vol_0 + fb_0 * this->r.g.echo_volume_0) | ||
1462 | >> global_muting; | ||
1463 | int amp_1 = (chans_1 * global_vol_1 + fb_1 * this->r.g.echo_volume_1) | ||
1464 | >> global_muting; | ||
1465 | |||
1466 | out_buf [ 0] = amp_0; | ||
1467 | out_buf [WAV_CHUNK_SIZE] = amp_1; | ||
1468 | out_buf ++; | ||
1469 | |||
1470 | if ( !(this->r.g.flags & 0x20) ) | ||
1471 | { | ||
1472 | /* Feedback into echo buffer */ | ||
1473 | int e0 = (echo_0 >> 7) + ((fb_0 * this->r.g.echo_feedback) >> 14); | ||
1474 | int e1 = (echo_1 >> 7) + ((fb_1 * this->r.g.echo_feedback) >> 14); | ||
1475 | e0 = CLAMP16( e0 ); | ||
1476 | SET_LE16A( echo_ptr , e0 ); | ||
1477 | e1 = CLAMP16( e1 ); | ||
1478 | SET_LE16A( echo_ptr + 2, e1 ); | ||
1479 | } | ||
1480 | #endif /* ARM_ARCH */ | ||
1481 | #else /* Unoptimized CPU */ | ||
1482 | /* Read feedback from echo buffer */ | ||
1483 | int echo_pos = this->echo_pos; | ||
1484 | uint8_t* const echo_ptr = RAM + | ||
1485 | ((this->r.g.echo_page * 0x100 + echo_pos) & 0xFFFF); | ||
1486 | echo_pos += 4; | ||
1487 | if ( echo_pos >= (this->r.g.echo_delay & 15) * 0x800 ) | ||
1488 | echo_pos = 0; | ||
1489 | this->echo_pos = echo_pos; | ||
1490 | int fb_0 = GET_LE16SA( echo_ptr ); | ||
1491 | int fb_1 = GET_LE16SA( echo_ptr + 2 ); | ||
1492 | |||
1493 | /* Keep last 8 samples */ | ||
1494 | int (* const fir_ptr) [2] = this->fir_buf + this->fir_pos; | ||
1495 | this->fir_pos = (this->fir_pos + 1) & (FIR_BUF_HALF - 1); | ||
1496 | fir_ptr [ 0] [0] = fb_0; | ||
1497 | fir_ptr [ 0] [1] = fb_1; | ||
1498 | /* duplicate at +8 eliminates wrap checking below */ | ||
1499 | fir_ptr [FIR_BUF_HALF] [0] = fb_0; | ||
1500 | fir_ptr [FIR_BUF_HALF] [1] = fb_1; | ||
1501 | |||
1502 | /* Apply FIR */ | ||
1503 | fb_0 *= this->fir_coeff [0]; | ||
1504 | fb_1 *= this->fir_coeff [0]; | ||
1505 | |||
1506 | #define DO_PT( i )\ | ||
1507 | fb_0 += fir_ptr [i] [0] * this->fir_coeff [i];\ | ||
1508 | fb_1 += fir_ptr [i] [1] * this->fir_coeff [i]; | ||
1509 | |||
1510 | DO_PT( 1 ) | ||
1511 | DO_PT( 2 ) | ||
1512 | DO_PT( 3 ) | ||
1513 | DO_PT( 4 ) | ||
1514 | DO_PT( 5 ) | ||
1515 | DO_PT( 6 ) | ||
1516 | DO_PT( 7 ) | ||
1517 | |||
1518 | /* Generate output */ | ||
1519 | int amp_0 = (chans_0 * global_vol_0 + fb_0 * this->r.g.echo_volume_0) | ||
1520 | >> global_muting; | ||
1521 | int amp_1 = (chans_1 * global_vol_1 + fb_1 * this->r.g.echo_volume_1) | ||
1522 | >> global_muting; | ||
1523 | out_buf [ 0] = amp_0; | ||
1524 | out_buf [WAV_CHUNK_SIZE] = amp_1; | ||
1525 | out_buf ++; | ||
1526 | |||
1527 | if ( !(this->r.g.flags & 0x20) ) | ||
1528 | { | ||
1529 | /* Feedback into echo buffer */ | ||
1530 | int e0 = (echo_0 >> 7) + ((fb_0 * this->r.g.echo_feedback) >> 14); | ||
1531 | int e1 = (echo_1 >> 7) + ((fb_1 * this->r.g.echo_feedback) >> 14); | ||
1532 | e0 = CLAMP16( e0 ); | ||
1533 | SET_LE16A( echo_ptr , e0 ); | ||
1534 | e1 = CLAMP16( e1 ); | ||
1535 | SET_LE16A( echo_ptr + 2, e1 ); | ||
1536 | } | ||
1537 | #endif /* CPU_* */ | ||
1538 | #else /* SPCNOECHO == 1*/ | ||
1539 | /* Generate output */ | ||
1540 | int amp_0 = (chans_0 * global_vol_0) >> global_muting; | ||
1541 | int amp_1 = (chans_1 * global_vol_1) >> global_muting; | ||
1542 | out_buf [ 0] = amp_0; | ||
1543 | out_buf [WAV_CHUNK_SIZE] = amp_1; | ||
1544 | out_buf ++; | ||
1545 | #endif /* SPCNOECHO */ | ||
1546 | } | ||
1547 | while ( --count ); | ||
1548 | #if 0 | ||
1549 | EXIT_TIMER(dsp); | ||
1550 | ENTER_TIMER(cpu); | ||
1551 | #endif | ||
1552 | } | ||
1553 | |||
1554 | void DSP_reset( struct Spc_Dsp* this ) | ||
1555 | { | ||
1556 | this->keys_down = 0; | ||
1557 | this->echo_pos = 0; | ||
1558 | this->noise_count = 0; | ||
1559 | this->noise = 2; | ||
1560 | |||
1561 | this->r.g.flags = 0xE0; /* reset, mute, echo off */ | ||
1562 | this->r.g.key_ons = 0; | ||
1563 | |||
1564 | ci->memset( this->voice_state, 0, sizeof this->voice_state ); | ||
1565 | |||
1566 | int i; | ||
1567 | for ( i = VOICE_COUNT; --i >= 0; ) | ||
1568 | { | ||
1569 | struct voice_t* v = this->voice_state + i; | ||
1570 | v->env_mode = state_release; | ||
1571 | v->addr = ram.ram; | ||
1572 | } | ||
1573 | |||
1574 | #if SPC_BRRCACHE | ||
1575 | this->oldsize = 0; | ||
1576 | for ( i = 0; i < 256; i++ ) | ||
1577 | this->wave_entry [i].start_addr = -1; | ||
1578 | #endif | ||
1579 | |||
1580 | #if defined(CPU_COLDFIRE) | ||
1581 | this->fir_ptr = fir_buf; | ||
1582 | this->last_fir_ptr = &fir_buf [7]; | ||
1583 | ci->memset( fir_buf, 0, sizeof fir_buf ); | ||
1584 | #elif defined (CPU_ARM) | ||
1585 | this->fir_ptr = fir_buf; | ||
1586 | ci->memset( fir_buf, 0, sizeof fir_buf ); | ||
1587 | #else | ||
1588 | this->fir_pos = 0; | ||
1589 | ci->memset( this->fir_buf, 0, sizeof this->fir_buf ); | ||
1590 | #endif | ||
1591 | |||
1592 | assert( offsetof (struct globals_t,unused9 [2]) == REGISTER_COUNT ); | ||
1593 | assert( sizeof (this->r.voice) == REGISTER_COUNT ); | ||
1594 | } | ||
diff --git a/lib/rbcodec/codecs/libspc/spc_emu.c b/lib/rbcodec/codecs/libspc/spc_emu.c new file mode 100644 index 0000000000..5ea5b0cdeb --- /dev/null +++ b/lib/rbcodec/codecs/libspc/spc_emu.c | |||
@@ -0,0 +1,397 @@ | |||
1 | /*************************************************************************** | ||
2 | * __________ __ ___. | ||
3 | * Open \______ \ ____ ____ | | _\_ |__ _______ ___ | ||
4 | * Source | _// _ \_/ ___\| |/ /| __ \ / _ \ \/ / | ||
5 | * Jukebox | | ( <_> ) \___| < | \_\ ( <_> > < < | ||
6 | * Firmware |____|_ /\____/ \___ >__|_ \|___ /\____/__/\_ \ | ||
7 | * \/ \/ \/ \/ \/ | ||
8 | * $Id$ | ||
9 | * | ||
10 | * Copyright (C) 2006-2007 Adam Gashlin (hcs) | ||
11 | * Copyright (C) 2004-2007 Shay Green (blargg) | ||
12 | * Copyright (C) 2002 Brad Martin | ||
13 | * | ||
14 | * This program is free software; you can redistribute it and/or | ||
15 | * modify it under the terms of the GNU General Public License | ||
16 | * as published by the Free Software Foundation; either version 2 | ||
17 | * of the License, or (at your option) any later version. | ||
18 | * | ||
19 | * This software is distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY | ||
20 | * KIND, either express or implied. | ||
21 | * | ||
22 | ****************************************************************************/ | ||
23 | #include "codeclib.h" | ||
24 | #include "spc_codec.h" | ||
25 | #include "spc_profiler.h" | ||
26 | |||
27 | /* lovingly ripped off from Game_Music_Emu 0.5.2. http://www.slack.net/~ant/ */ | ||
28 | /* DSP Based on Brad Martin's OpenSPC DSP emulator */ | ||
29 | /* tag reading from sexyspc by John Brawn (John_Brawn@yahoo.com) and others */ | ||
30 | |||
31 | struct cpu_ram_t ram IBSS_ATTR_SPC_LARGE_IRAM CACHEALIGN_ATTR; | ||
32 | |||
33 | /**************** Timers ****************/ | ||
34 | |||
35 | static void Timer_run_( struct Timer* t, long time ) ICODE_ATTR_SPC; | ||
36 | static void Timer_run_( struct Timer* t, long time ) | ||
37 | { | ||
38 | /* when disabled, next_tick should always be in the future */ | ||
39 | assert( t->enabled ); | ||
40 | |||
41 | int elapsed = ((time - t->next_tick) >> t->shift) + 1; | ||
42 | t->next_tick += elapsed << t->shift; | ||
43 | |||
44 | elapsed += t->count; | ||
45 | if ( elapsed >= t->period ) /* avoid unnecessary division */ | ||
46 | { | ||
47 | int n = elapsed / t->period; | ||
48 | elapsed -= n * t->period; | ||
49 | t->counter = (t->counter + n) & 15; | ||
50 | } | ||
51 | t->count = elapsed; | ||
52 | } | ||
53 | |||
54 | static inline void Timer_run( struct Timer* t, long time ) | ||
55 | { | ||
56 | if ( time >= t->next_tick ) | ||
57 | Timer_run_( t, time ); | ||
58 | } | ||
59 | |||
60 | /**************** SPC emulator ****************/ | ||
61 | /* 1.024 MHz clock / 32000 samples per second */ | ||
62 | |||
63 | static void SPC_enable_rom( THIS, int enable ) | ||
64 | { | ||
65 | if ( this->rom_enabled != enable ) | ||
66 | { | ||
67 | this->rom_enabled = enable; | ||
68 | ci->memcpy( RAM + ROM_ADDR, (enable ? this->boot_rom : this->extra_ram), ROM_SIZE ); | ||
69 | /* TODO: ROM can still get overwritten when DSP writes to echo buffer */ | ||
70 | } | ||
71 | } | ||
72 | |||
73 | void SPC_Init( THIS ) | ||
74 | { | ||
75 | this->timer [0].shift = 4 + 3; /* 8 kHz */ | ||
76 | this->timer [1].shift = 4 + 3; /* 8 kHz */ | ||
77 | this->timer [2].shift = 4; /* 8 kHz */ | ||
78 | |||
79 | /* Put STOP instruction around memory to catch PC underflow/overflow. */ | ||
80 | ci->memset( ram.padding1, 0xFF, sizeof ram.padding1 ); | ||
81 | ci->memset( ram.padding2, 0xFF, sizeof ram.padding2 ); | ||
82 | |||
83 | /* A few tracks read from the last four bytes of IPL ROM */ | ||
84 | this->boot_rom [sizeof this->boot_rom - 2] = 0xC0; | ||
85 | this->boot_rom [sizeof this->boot_rom - 1] = 0xFF; | ||
86 | ci->memset( this->boot_rom, 0, sizeof this->boot_rom - 2 ); | ||
87 | |||
88 | /* Have DSP in a defined state in case EMU is run and hasn't loaded | ||
89 | * a program yet */ | ||
90 | DSP_reset(&this->dsp); | ||
91 | } | ||
92 | |||
93 | static void SPC_load_state( THIS, struct cpu_regs_t const* cpu_state, | ||
94 | const void* new_ram, const void* dsp_state ) | ||
95 | { | ||
96 | ci->memcpy(&(this->r),cpu_state,sizeof this->r); | ||
97 | |||
98 | /* ram */ | ||
99 | ci->memcpy( RAM, new_ram, sizeof RAM ); | ||
100 | ci->memcpy( this->extra_ram, RAM + ROM_ADDR, sizeof this->extra_ram ); | ||
101 | |||
102 | /* boot rom (have to force enable_rom() to update it) */ | ||
103 | this->rom_enabled = !(RAM [0xF1] & 0x80); | ||
104 | SPC_enable_rom( this, !this->rom_enabled ); | ||
105 | |||
106 | /* dsp */ | ||
107 | /* some SPCs rely on DSP immediately generating one sample */ | ||
108 | this->extra_cycles = 32; | ||
109 | DSP_reset( &this->dsp ); | ||
110 | int i; | ||
111 | for ( i = 0; i < REGISTER_COUNT; i++ ) | ||
112 | DSP_write( &this->dsp, i, ((uint8_t const*) dsp_state) [i] ); | ||
113 | |||
114 | /* timers */ | ||
115 | for ( i = 0; i < TIMER_COUNT; i++ ) | ||
116 | { | ||
117 | struct Timer* t = &this->timer [i]; | ||
118 | |||
119 | t->next_tick = -EXTRA_CLOCKS; | ||
120 | t->enabled = (RAM [0xF1] >> i) & 1; | ||
121 | if ( !t->enabled ) | ||
122 | t->next_tick = TIMER_DISABLED_TIME; | ||
123 | t->count = 0; | ||
124 | t->counter = RAM [0xFD + i] & 15; | ||
125 | |||
126 | int p = RAM [0xFA + i]; | ||
127 | if ( !p ) | ||
128 | p = 0x100; | ||
129 | t->period = p; | ||
130 | } | ||
131 | |||
132 | /* Handle registers which already give 0 when read by | ||
133 | setting RAM and not changing it. | ||
134 | Put STOP instruction in registers which can be read, | ||
135 | to catch attempted execution. */ | ||
136 | RAM [0xF0] = 0; | ||
137 | RAM [0xF1] = 0; | ||
138 | RAM [0xF3] = 0xFF; | ||
139 | RAM [0xFA] = 0; | ||
140 | RAM [0xFB] = 0; | ||
141 | RAM [0xFC] = 0; | ||
142 | RAM [0xFD] = 0xFF; | ||
143 | RAM [0xFE] = 0xFF; | ||
144 | RAM [0xFF] = 0xFF; | ||
145 | } | ||
146 | |||
147 | static void clear_echo( THIS ) | ||
148 | { | ||
149 | if ( !(DSP_read( &this->dsp, 0x6C ) & 0x20) ) | ||
150 | { | ||
151 | unsigned addr = 0x100 * DSP_read( &this->dsp, 0x6D ); | ||
152 | size_t size = 0x800 * DSP_read( &this->dsp, 0x7D ); | ||
153 | size_t max_size = sizeof RAM - addr; | ||
154 | if ( size > max_size ) | ||
155 | size = sizeof RAM - addr; | ||
156 | ci->memset( RAM + addr, 0xFF, size ); | ||
157 | } | ||
158 | } | ||
159 | |||
160 | int SPC_load_spc( THIS, const void* data, long size ) | ||
161 | { | ||
162 | struct spc_file_t const* spc = (struct spc_file_t const*) data; | ||
163 | struct cpu_regs_t regs; | ||
164 | |||
165 | if ( size < SPC_FILE_SIZE ) | ||
166 | return -1; | ||
167 | |||
168 | if ( ci->memcmp( spc->signature, "SNES-SPC700 Sound File Data", 27 ) != 0 ) | ||
169 | return -1; | ||
170 | |||
171 | regs.pc = spc->pc [1] * 0x100 + spc->pc [0]; | ||
172 | regs.a = spc->a; | ||
173 | regs.x = spc->x; | ||
174 | regs.y = spc->y; | ||
175 | regs.status = spc->status; | ||
176 | regs.sp = spc->sp; | ||
177 | |||
178 | if ( (unsigned long) size >= sizeof *spc ) | ||
179 | ci->memcpy( this->boot_rom, spc->ipl_rom, sizeof this->boot_rom ); | ||
180 | |||
181 | SPC_load_state( this, ®s, spc->ram, spc->dsp ); | ||
182 | |||
183 | clear_echo(this); | ||
184 | |||
185 | return 0; | ||
186 | } | ||
187 | |||
188 | /**************** DSP interaction ****************/ | ||
189 | static void SPC_run_dsp_( THIS, long time ) ICODE_ATTR_SPC; | ||
190 | static void SPC_run_dsp_( THIS, long time ) | ||
191 | { | ||
192 | /* divide by CLOCKS_PER_SAMPLE */ | ||
193 | int count = ((time - this->next_dsp) >> 5) + 1; | ||
194 | int32_t* buf = this->sample_buf; | ||
195 | this->sample_buf = buf + count; | ||
196 | this->next_dsp += count * CLOCKS_PER_SAMPLE; | ||
197 | DSP_run( &this->dsp, count, buf ); | ||
198 | } | ||
199 | |||
200 | static inline void SPC_run_dsp( THIS, long time ) | ||
201 | { | ||
202 | if ( time >= this->next_dsp ) | ||
203 | SPC_run_dsp_( this, time ); | ||
204 | } | ||
205 | |||
206 | int SPC_read( THIS, unsigned addr, long const time ) | ||
207 | { | ||
208 | int result = RAM [addr]; | ||
209 | |||
210 | if ( ((unsigned) (addr - 0xF0)) < 0x10 ) | ||
211 | { | ||
212 | assert( 0xF0 <= addr && addr <= 0xFF ); | ||
213 | |||
214 | /* counters */ | ||
215 | int i = addr - 0xFD; | ||
216 | if ( i >= 0 ) | ||
217 | { | ||
218 | struct Timer* t = &this->timer [i]; | ||
219 | Timer_run( t, time ); | ||
220 | result = t->counter; | ||
221 | t->counter = 0; | ||
222 | } | ||
223 | /* dsp */ | ||
224 | else if ( addr == 0xF3 ) | ||
225 | { | ||
226 | SPC_run_dsp( this, time ); | ||
227 | result = DSP_read( &this->dsp, RAM [0xF2] & 0x7F ); | ||
228 | } | ||
229 | } | ||
230 | return result; | ||
231 | } | ||
232 | |||
233 | void SPC_write( THIS, unsigned addr, int data, long const time ) | ||
234 | { | ||
235 | /* first page is very common */ | ||
236 | if ( addr < 0xF0 ) | ||
237 | { | ||
238 | RAM [addr] = (uint8_t) data; | ||
239 | } | ||
240 | else switch ( addr ) | ||
241 | { | ||
242 | /* RAM */ | ||
243 | default: | ||
244 | if ( addr < ROM_ADDR ) | ||
245 | { | ||
246 | RAM [addr] = (uint8_t) data; | ||
247 | } | ||
248 | else | ||
249 | { | ||
250 | this->extra_ram [addr - ROM_ADDR] = (uint8_t) data; | ||
251 | if ( !this->rom_enabled ) | ||
252 | RAM [addr] = (uint8_t) data; | ||
253 | } | ||
254 | break; | ||
255 | |||
256 | /* DSP */ | ||
257 | /*case 0xF2:*/ /* mapped to RAM */ | ||
258 | case 0xF3: { | ||
259 | SPC_run_dsp( this, time ); | ||
260 | int reg = RAM [0xF2]; | ||
261 | if ( reg < REGISTER_COUNT ) { | ||
262 | DSP_write( &this->dsp, reg, data ); | ||
263 | } | ||
264 | else { | ||
265 | /*dprintf( "DSP write to $%02X\n", (int) reg ); */ | ||
266 | } | ||
267 | break; | ||
268 | } | ||
269 | |||
270 | case 0xF0: /* Test register */ | ||
271 | /*dprintf( "Wrote $%02X to $F0\n", (int) data ); */ | ||
272 | break; | ||
273 | |||
274 | /* Config */ | ||
275 | case 0xF1: | ||
276 | { | ||
277 | int i; | ||
278 | /* timers */ | ||
279 | for ( i = 0; i < TIMER_COUNT; i++ ) | ||
280 | { | ||
281 | struct Timer * t = this->timer+i; | ||
282 | if ( !(data & (1 << i)) ) | ||
283 | { | ||
284 | t->enabled = 0; | ||
285 | t->next_tick = TIMER_DISABLED_TIME; | ||
286 | } | ||
287 | else if ( !t->enabled ) | ||
288 | { | ||
289 | /* just enabled */ | ||
290 | t->enabled = 1; | ||
291 | t->counter = 0; | ||
292 | t->count = 0; | ||
293 | t->next_tick = time; | ||
294 | } | ||
295 | } | ||
296 | |||
297 | /* port clears */ | ||
298 | if ( data & 0x10 ) | ||
299 | { | ||
300 | RAM [0xF4] = 0; | ||
301 | RAM [0xF5] = 0; | ||
302 | } | ||
303 | if ( data & 0x20 ) | ||
304 | { | ||
305 | RAM [0xF6] = 0; | ||
306 | RAM [0xF7] = 0; | ||
307 | } | ||
308 | |||
309 | SPC_enable_rom( this, (data & 0x80) != 0 ); | ||
310 | break; | ||
311 | } | ||
312 | |||
313 | /* Ports */ | ||
314 | case 0xF4: | ||
315 | case 0xF5: | ||
316 | case 0xF6: | ||
317 | case 0xF7: | ||
318 | /* to do: handle output ports */ | ||
319 | break; | ||
320 | |||
321 | /* verified on SNES that these are read/write (RAM) */ | ||
322 | /*case 0xF8: */ | ||
323 | /*case 0xF9: */ | ||
324 | |||
325 | /* Timers */ | ||
326 | case 0xFA: | ||
327 | case 0xFB: | ||
328 | case 0xFC: { | ||
329 | int i = addr - 0xFA; | ||
330 | struct Timer* t = &this->timer [i]; | ||
331 | if ( (t->period & 0xFF) != data ) | ||
332 | { | ||
333 | Timer_run( t, time ); | ||
334 | this->timer[i].period = data ? data : 0x100; | ||
335 | } | ||
336 | break; | ||
337 | } | ||
338 | |||
339 | /* Counters (cleared on write) */ | ||
340 | case 0xFD: | ||
341 | case 0xFE: | ||
342 | case 0xFF: | ||
343 | /*dprintf( "Wrote to counter $%02X\n", (int) addr ); */ | ||
344 | this->timer [addr - 0xFD].counter = 0; | ||
345 | break; | ||
346 | } | ||
347 | } | ||
348 | |||
349 | /**************** Sample generation ****************/ | ||
350 | int SPC_play( THIS, long count, int32_t* out ) | ||
351 | { | ||
352 | int i; | ||
353 | assert( count % 2 == 0 ); /* output is always in pairs of samples */ | ||
354 | |||
355 | long start_time = -(count >> 1) * CLOCKS_PER_SAMPLE - EXTRA_CLOCKS; | ||
356 | |||
357 | /* DSP output is made on-the-fly when DSP registers are read or written */ | ||
358 | this->sample_buf = out; | ||
359 | this->next_dsp = start_time + CLOCKS_PER_SAMPLE; | ||
360 | |||
361 | /* Localize timer next_tick times and run them to the present to prevent | ||
362 | a running but ignored timer's next_tick from getting too far behind | ||
363 | and overflowing. */ | ||
364 | for ( i = 0; i < TIMER_COUNT; i++ ) | ||
365 | { | ||
366 | struct Timer* t = &this->timer [i]; | ||
367 | if ( t->enabled ) | ||
368 | { | ||
369 | t->next_tick += start_time + EXTRA_CLOCKS; | ||
370 | Timer_run( t, start_time ); | ||
371 | } | ||
372 | } | ||
373 | |||
374 | /* Run from start_time to 0, pre-advancing by extra cycles from last run */ | ||
375 | this->extra_cycles = CPU_run( this, start_time + this->extra_cycles ) + | ||
376 | EXTRA_CLOCKS; | ||
377 | if ( this->extra_cycles < 0 ) | ||
378 | { | ||
379 | /*dprintf( "Unhandled instruction $%02X, pc = $%04X\n", | ||
380 | (int) CPU_read( r.pc ), (unsigned) r.pc ); */ | ||
381 | |||
382 | return -1; | ||
383 | } | ||
384 | |||
385 | /* Catch DSP up to present */ | ||
386 | #if 0 | ||
387 | ENTER_TIMER(cpu); | ||
388 | #endif | ||
389 | SPC_run_dsp( this, -EXTRA_CLOCKS ); | ||
390 | #if 0 | ||
391 | EXIT_TIMER(cpu); | ||
392 | #endif | ||
393 | assert( this->next_dsp == CLOCKS_PER_SAMPLE - EXTRA_CLOCKS ); | ||
394 | assert( this->sample_buf - out == count ); | ||
395 | |||
396 | return 0; | ||
397 | } | ||
diff --git a/lib/rbcodec/codecs/libspc/spc_profiler.c b/lib/rbcodec/codecs/libspc/spc_profiler.c new file mode 100644 index 0000000000..0ced8b5bd3 --- /dev/null +++ b/lib/rbcodec/codecs/libspc/spc_profiler.c | |||
@@ -0,0 +1,66 @@ | |||
1 | /*************************************************************************** | ||
2 | * __________ __ ___. | ||
3 | * Open \______ \ ____ ____ | | _\_ |__ _______ ___ | ||
4 | * Source | _// _ \_/ ___\| |/ /| __ \ / _ \ \/ / | ||
5 | * Jukebox | | ( <_> ) \___| < | \_\ ( <_> > < < | ||
6 | * Firmware |____|_ /\____/ \___ >__|_ \|___ /\____/__/\_ \ | ||
7 | * \/ \/ \/ \/ \/ | ||
8 | * $Id$ | ||
9 | * | ||
10 | * Copyright (C) 2006-2007 Adam Gashlin (hcs) | ||
11 | * | ||
12 | * This program is free software; you can redistribute it and/or | ||
13 | * modify it under the terms of the GNU General Public License | ||
14 | * as published by the Free Software Foundation; either version 2 | ||
15 | * of the License, or (at your option) any later version. | ||
16 | * | ||
17 | * This software is distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY | ||
18 | * KIND, either express or implied. | ||
19 | * | ||
20 | ****************************************************************************/ | ||
21 | |||
22 | /* lovingly ripped off from Game_Music_Emu 0.5.2. http://www.slack.net/~ant/ */ | ||
23 | /* DSP Based on Brad Martin's OpenSPC DSP emulator */ | ||
24 | /* tag reading from sexyspc by John Brawn (John_Brawn@yahoo.com) and others */ | ||
25 | |||
26 | #if defined(SPC_PROFILE) && defined(USEC_TIMER) | ||
27 | |||
28 | #include "codeclib.h" | ||
29 | #include "spc_codec.h" | ||
30 | #define SPC_DEFINE_PROFILER_TIMERS | ||
31 | #include "spc_profiler.h" | ||
32 | |||
33 | void reset_profile_timers(void) | ||
34 | { | ||
35 | RESET_TIMER(total); | ||
36 | RESET_TIMER(render); | ||
37 | #if 0 | ||
38 | RESET_TIMER(cpu); | ||
39 | RESET_TIMER(dsp); | ||
40 | RESET_TIMER(dsp_pregen); | ||
41 | RESET_TIMER(dsp_gen); | ||
42 | RESET_TIMER(dsp_mix); | ||
43 | #endif | ||
44 | } | ||
45 | |||
46 | void print_timers(char * path) | ||
47 | { | ||
48 | int logfd = ci->open("/spclog.txt",O_WRONLY|O_CREAT|O_APPEND, 0666); | ||
49 | ci->fdprintf(logfd,"%s:\t",path); | ||
50 | ci->fdprintf(logfd,"%10ld total\t",READ_TIMER(total)); | ||
51 | PRINT_TIMER_PCT(render,total,"render"); | ||
52 | #if 0 | ||
53 | PRINT_TIMER_PCT(cpu,total,"CPU"); | ||
54 | PRINT_TIMER_PCT(dsp,total,"DSP"); | ||
55 | ci->fdprintf(logfd,"("); | ||
56 | PRINT_TIMER_PCT(dsp_pregen,dsp,"pregen"); | ||
57 | PRINT_TIMER_PCT(dsp_gen,dsp,"gen"); | ||
58 | PRINT_TIMER_PCT(dsp_mix,dsp,"mix"); | ||
59 | #endif | ||
60 | ci->fdprintf(logfd,"\n"); | ||
61 | |||
62 | ci->close(logfd); | ||
63 | logfd=-1; | ||
64 | } | ||
65 | |||
66 | #endif /* #if defined(SPC_PROFILE) && defined(USEC_TIMER) */ | ||
diff --git a/lib/rbcodec/codecs/libspc/spc_profiler.h b/lib/rbcodec/codecs/libspc/spc_profiler.h new file mode 100644 index 0000000000..405ee43ef9 --- /dev/null +++ b/lib/rbcodec/codecs/libspc/spc_profiler.h | |||
@@ -0,0 +1,72 @@ | |||
1 | /*************************************************************************** | ||
2 | * __________ __ ___. | ||
3 | * Open \______ \ ____ ____ | | _\_ |__ _______ ___ | ||
4 | * Source | _// _ \_/ ___\| |/ /| __ \ / _ \ \/ / | ||
5 | * Jukebox | | ( <_> ) \___| < | \_\ ( <_> > < < | ||
6 | * Firmware |____|_ /\____/ \___ >__|_ \|___ /\____/__/\_ \ | ||
7 | * \/ \/ \/ \/ \/ | ||
8 | * $Id$ | ||
9 | * | ||
10 | * Copyright (C) 2006-2007 Adam Gashlin (hcs) | ||
11 | * | ||
12 | * This program is free software; you can redistribute it and/or | ||
13 | * modify it under the terms of the GNU General Public License | ||
14 | * as published by the Free Software Foundation; either version 2 | ||
15 | * of the License, or (at your option) any later version. | ||
16 | * | ||
17 | * This software is distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY | ||
18 | * KIND, either express or implied. | ||
19 | * | ||
20 | ****************************************************************************/ | ||
21 | |||
22 | /* a fun simple elapsed time profiler */ | ||
23 | #ifndef _SPC_PROFILER_H_ | ||
24 | #define _SPC_PROFILER_H_ | ||
25 | |||
26 | #if defined(SPC_PROFILE) && defined(USEC_TIMER) | ||
27 | |||
28 | #ifdef SPC_DEFINE_PROFILER_TIMERS | ||
29 | #define CREATE_TIMER(name) uint32_t spc_timer_##name##_start,\ | ||
30 | spc_timer_##name##_total | ||
31 | #else | ||
32 | #define CREATE_TIMER(name) extern uint32_t spc_timer_##name##_start,\ | ||
33 | spc_timer_##name##_total | ||
34 | #endif | ||
35 | |||
36 | #define ENTER_TIMER(name) spc_timer_##name##_start=USEC_TIMER | ||
37 | #define EXIT_TIMER(name) spc_timer_##name##_total+=\ | ||
38 | (USEC_TIMER-spc_timer_##name##_start) | ||
39 | #define READ_TIMER(name) (spc_timer_##name##_total) | ||
40 | #define RESET_TIMER(name) spc_timer_##name##_total=0 | ||
41 | |||
42 | #define PRINT_TIMER_PCT(bname,tname,nstr) ci->fdprintf( \ | ||
43 | logfd,"%10ld ",READ_TIMER(bname));\ | ||
44 | ci->fdprintf(logfd,"(%3d%%) " nstr "\t",\ | ||
45 | ((uint64_t)READ_TIMER(bname))*100/READ_TIMER(tname)) | ||
46 | |||
47 | CREATE_TIMER(total); | ||
48 | CREATE_TIMER(render); | ||
49 | #if 0 | ||
50 | CREATE_TIMER(cpu); | ||
51 | CREATE_TIMER(dsp); | ||
52 | CREATE_TIMER(dsp_pregen); | ||
53 | CREATE_TIMER(dsp_gen); | ||
54 | CREATE_TIMER(dsp_mix); | ||
55 | #endif | ||
56 | |||
57 | void reset_profile_timers(void); | ||
58 | void print_timers(char * path); | ||
59 | |||
60 | #else | ||
61 | |||
62 | #define CREATE_TIMER(name) | ||
63 | #define ENTER_TIMER(name) | ||
64 | #define EXIT_TIMER(name) | ||
65 | #define READ_TIMER(name) | ||
66 | #define RESET_TIMER(name) | ||
67 | #define print_timers(path) | ||
68 | #define reset_profile_timers() | ||
69 | |||
70 | #endif | ||
71 | |||
72 | #endif /* _SPC_PROFILER_H_ */ | ||