1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
|
/***************************************************************************
* __________ __ ___.
* Open \______ \ ____ ____ | | _\_ |__ _______ ___
* Source | _// _ \_/ ___\| |/ /| __ \ / _ \ \/ /
* Jukebox | | ( <_> ) \___| < | \_\ ( <_> > < <
* Firmware |____|_ /\____/ \___ >__|_ \|___ /\____/__/\_ \
* \/ \/ \/ \/ \/
* $Id$
*
* Copyright (C) 2005 by David Bryant
*
* This program is free software; you can redistribute it and/or
* modify it under the terms of the GNU General Public License
* as published by the Free Software Foundation; either version 2
* of the License, or (at your option) any later version.
*
* This software is distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY
* KIND, either express or implied.
*
****************************************************************************/
/* This is an assembly optimized version of the following WavPack function:
*
* void decorr_stereo_pass_cont_mcf5249 (struct decorr_pass *dpp,
* long *buffer, long sample_count);
*
* It performs a single pass of stereo decorrelation on the provided buffer.
* Note that this version of the function requires that the 8 previous stereo
* samples are visible and correct. In other words, it ignores the "samples_*"
* fields in the decorr_pass structure and gets the history data directly
* from the buffer. It does, however, return the appropriate history samples
* to the decorr_pass structure before returning.
*
* This is written to work on a MCF5249 processor, or any processor based on
* the ColdFire V2 core with an EMAC unit. The EMAC is perfectly suited for
* the "apply_weight" function of WavPack decorrelation because it provides
* the requires 40-bit product. The fractional rounding mode of the EMAC is not
* configurable and uses "round to even" while WavPack uses "round to larger",
* so the rounding has to be done manually.
*/
.text
.align 2
.global decorr_stereo_pass_cont_mcf5249
decorr_stereo_pass_cont_mcf5249:
lea (-44, %sp), %sp
movem.l %d2-%d7/%a2-%a6, (%sp)
move.l 44+4(%sp), %a2 | a2 = dpp->
move.l 44+8(%sp), %a1 | a1 = bptr
move.w 2(%a2), %a3 | a3 = dpp->delta
move.w 4(%a2), %d3 | d3 = dpp->weight_A (sign extended)
ext.l %d3
move.w 6(%a2), %d4 | d4 = dpp->weight_B (sign extended)
ext.l %d4
move.l 44+12(%sp), %d0 | d0 = sample_count
jbeq return_only | if zero, nothing to do
lsl.l #3, %d0 | d5 = bptr + (sample_count * 8)
move.l %d0, %d5
add.l %a1, %d5
moveq.l #17, %d0 | left shift weights & delta 17 places
asl.l %d0, %d3
asl.l %d0, %d4
move.l %a3, %d1
asl.l %d0, %d1
move.l %d1, %a3
moveq.l #0x20, %d6
move.l %d6, %macsr | set fractional mode for MAC
move.l #0x800000, %accext01 | acc1 = 0x00 0000 80 (for rounding)
move.l #1024<<17, %d6 | d6 & d7 are weight clipping limits
move.l #-1024<<17, %d7 | (only used by negative terms)
move.w (%a2), %d0 | d0 = term
ext.l %d0
cmp.l #17, %d0
jbeq term_17 | term = 17
cmp.l #18, %d0
jbeq term_18 | term = 18
addq.l #1, %d0
jbeq term_minus_1 | term = -1
addq.l #1, %d0
jbeq term_minus_2 | term = -2
addq.l #1, %d0
jbeq term_minus_3 | term = -3
jbra term_default | default term = 1 - 8
|------------------------------------------------------------------------------
| Loop to handle term = 17 condition
|
| a0 = d0 = (2 * bptr [-1]) - bptr [-2]
| a1 = bptr d1 = initial bptr [0]
| a2 = dpp-> d2 = updated bptr [0]
| a3 = dpp->delta << 17 d3 = dpp->weight_A << 17
| a4 = d4 = dpp->weight_B << 17
| a5 = d5 = eptr
| macsr = 0x20 acc1 = 0x00 0000 80
|------------------------------------------------------------------------------
term_17:
move.l -8(%a1), %d0 | d0 = 2 * bptr [-1] - bptr [-2]
add.l %d0, %d0
sub.l -16(%a1), %d0
beq .L251 | if zero, skip calculation
move.l %acc1, %acc0
asl.l #4, %d0 | acc0 = acc1 + (d0 << 4) * weight_A
mac.l %d0, %d3, %acc0
move.l (%a1), %d1
beq .L255
eor.l %d1, %d0 | else compare signs
bge .L256 | if same, add delta to weight
sub.l %a3, %d3 | else subtract delta from weight
sub.l %a3, %d3 | subtract again instead of branch
.L256: add.l %a3, %d3 | add delta to weight
.L255: move.l %acc0, %d2 | d2 = rounded product
add.l %d1, %d2 | update bptr [0] and store
move.l %d2, (%a1)+
.L253: move.l -8(%a1), %d0 | d0 = 2 * bptr [-1] - bptr [-2]
add.l %d0, %d0
sub.l -16(%a1), %d0
beq .L257 | if zero, skip calculations
move.l %acc1, %acc0
asl.l #4, %d0 | acc0 = acc1 + (d0 << 4) * weight_B
mac.l %d0, %d4, %acc0
move.l (%a1), %d1
beq .L254
eor.l %d1, %d0 | else compare signs
bge .L259 | if same, add delta to weight
sub.l %a3, %d4 | else subtract delta from weight
sub.l %a3, %d4 | subtract again instead of branch
.L259: add.l %a3, %d4 | add delta to weight
.L254: move.l %acc0, %d2 | d2 = rounded product
add.l %d1, %d2 | update bptr [0] and store
move.l %d2, (%a1)+
.L252: cmp.l %a1, %d5 | loop if bptr < eptr
jbhi term_17
bra term_17_18_finish | exit through common path
.L251: addq.l #4, %a1 | update point and jump back into loop
bra .L253
.L257: addq.l #4, %a1 | update point and jump back into loop
bra .L252
|------------------------------------------------------------------------------
| Loop to handle term = 18 condition
|
| a0 = d0 = ((3 * bptr [-1]) - bptr [-2]) >> 1
| a1 = bptr d1 = initial bptr [0]
| a2 = dpp-> d2 = updated bptr [0]
| a3 = dpp->delta << 17 d3 = dpp->weight_A << 17
| a4 = d4 = dpp->weight_B << 17
| a5 = d5 = eptr
| macsr = 0x20 acc1 = 0x00 0000 80
|------------------------------------------------------------------------------
term_18:
move.l -8(%a1), %a0 | d0 = (3 * bptr [-1] - bptr [-2]) >> 1
lea (%a0,%a0.l*2), %a0
move.l %a0, %d0
sub.l -16(%a1), %d0
asr.l #1, %d0
beq .L260
move.l %acc1, %acc0
asl.l #4, %d0 | acc0 = acc1 + (d0 << 4) * weight_A
mac.l %d0, %d3, %acc0
move.l (%a1), %d1
beq .L266
eor.l %d1, %d0 | else compare signs
bge .L267 | if same, add delta to weight
sub.l %a3, %d3 | else subtract delta from weight
sub.l %a3, %d3 | subtract again instead of branch
.L267: add.l %a3, %d3 | add delta to weight
.L266: move.l %acc0, %d2 | d2 = rounded product
add.l %d1, %d2 | add applied weight to bptr [0], store
move.l %d2, (%a1)+
.L268: move.l -8(%a1), %a0 | d0 = (3 * bptr [-1] - bptr [-2]) >> 1
lea (%a0,%a0.l*2), %a0
move.l %a0, %d0
sub.l -16(%a1), %d0
asr.l #1, %d0
beq .L261
move.l %acc1, %acc0
asl.l #4, %d0 | acc0 = acc1 + (d0 << 4) * weight_B
mac.l %d0, %d4, %acc0
move.l (%a1), %d1
beq .L265
eor.l %d1, %d0 | else compare signs
bge .L270 | if same, add delta to weight
sub.l %a3, %d4 | else subtract delta from weight
sub.l %a3, %d4 | subtract again instead of branch
.L270: add.l %a3, %d4 | add delta to weight
.L265: move.l %acc0, %d2 | d2 = rounded product
add.l %d1, %d2 | add applied weight to bptr [0], store
move.l %d2, (%a1)+
.L269: cmp.l %a1, %d5 | loop if bptr < eptr
jbhi term_18
bra term_17_18_finish | exit through common path
.L260: addq.l #4, %a1 | bump pointer and jump back into loop
bra .L268
.L261: addq.l #4, %a1 | bump pointer and jump back into loop
bra .L269
term_17_18_finish:
move.l -4(%a1), 40(%a2) | restore dpp->samples_A [0-1], B [0-1]
move.l -8(%a1), 8(%a2)
move.l -12(%a1), 44(%a2)
move.l -16(%a1), 12(%a2)
jbra finish_up
|------------------------------------------------------------------------------
| Loop to handle default terms (i.e. 1 - 8)
|
| a0 = tptr d0 = tptr [0]
| a1 = bptr d1 = initial bptr [0]
| a2 = dpp-> d2 = updated bptr [0]
| a3 = dpp->delta << 17 d3 = dpp->weight_A << 17
| a4 = d4 = dpp->weight_B << 17
| a5 = d5 = eptr
| macsr = 0x20 acc1 = 0x00 0000 80
|------------------------------------------------------------------------------
term_default:
move.w (%a2), %d0 | a0 = a1 - (dpp->term * 8)
ext.l %d0
lsl.l #3, %d0
move.l %a1, %a0
sub.l %d0, %a0
term_default_loop:
move.l (%a0)+, %d0 | d0 = tptr [0], skip ahead if zero
beq .L271
move.l %acc1, %acc0
asl.l #4, %d0 | acc0 = acc1 + (d0 << 4) * weight_A
mac.l %d0, %d3, %acc0
move.l (%a1), %d1
beq .L277
eor.l %d1, %d0 | else compare signs
bge .L278 | if same, add delta to weight
sub.l %a3, %d3 | else subtract delta from weight
sub.l %a3, %d3 | subtract again instead of branch
.L278: add.l %a3, %d3 | add delta to weight
.L277: move.l %acc0, %d2 | d2 = rounded product
add.l %d1, %d2 | add applied weight to bptr [0], store
move.l %d2, (%a1)+
.L275: move.l (%a0)+, %d0 | d0 = tptr [0], skip ahead if zero
beq .L272
move.l %acc1, %acc0
asl.l #4, %d0 | acc0 = acc1 + (d0 << 4) * weight_B
mac.l %d0, %d4, %acc0
move.l (%a1), %d1
beq .L276
eor.l %d1, %d0 | else compare signs
bge .L281 | if same, add delta to weight
sub.l %a3, %d4 | else subtract delta from weight
sub.l %a3, %d4 | subtract again instead of branch
.L281: add.l %a3, %d4 | add delta to weight
.L276: move.l %acc0, %d2 | d2 = rounded product
add.l %d1, %d2 | add applied weight to bptr [0], store
move.l %d2, (%a1)+
.L274: cmp.l %a1, %d5 | loop back if bptr < eptr
jbhi term_default_loop
move.w (%a2), %d0 | d0 = term - 1
moveq.l #8, %d1 | d1 = loop counter
.L323: subq.l #1, %d0 | back up & mask index
and.l #7, %d0
move.l -(%a1), 40(%a2,%d0.l*4) | store dpp->samples_B [d0]
move.l -(%a1), 8(%a2,%d0.l*4) | store dpp->samples_A [d0]
subq.l #1, %d1 | loop on count
jbne .L323
jbra finish_up
.L271: addq.l #4, %a1 | bump pointer and jump back into loop
bra .L275
.L272: addq.l #4, %a1 | bump pointer and jump back into loop
bra .L274
|------------------------------------------------------------------------------
| Loop to handle term = -1 condition
|
| a0 = d0 = decorrelation sample
| a1 = bptr d1 = initial bptr [0]
| a2 = dpp-> d2 = updated bptr [0]
| a3 = dpp->delta << 17 d3 = dpp->weight_A << 17
| a4 = d4 = dpp->weight_B << 17
| a5 = d5 = eptr
| a6 = d6 = 1024 << 17
| a7 = d7 = -1024 << 17
| macsr = 0x20 acc1 = 0x00 0000 80
|------------------------------------------------------------------------------
term_minus_1:
move.l -4(%a1), %d0 | d0 = bptr [-1]
beq .L402
move.l %acc1, %acc0
asl.l #4, %d0 | acc0 = acc1 + ((d0 << 4) * weight_A)
mac.l %d0, %d3, %acc0
move.l (%a1), %d1
beq .L405
eor.l %d1, %d0 | else compare signs
bge .L404 | if same, add delta to weight
sub.l %a3, %d3 | else subtract delta from weight
cmp.l %d7, %d3 | check for negative clip limit
bge .L405
move.l %d7, %d3
bra .L405
.L404: add.l %a3, %d3 | add delta to weight
cmp.l %d6, %d3 | check for positive clip limit
ble .L405
move.l %d6, %d3
.L405: move.l %acc0, %d0 | d2 = rounded product
add.l %d1, %d0 | add applied weight to bptr [0], store
move.l %d0, (%a1)+
beq .L401
.L410: move.l %acc1, %acc0
asl.l #4, %d0 | acc0 = acc1 + ((d0 << 4) * weight_B)
mac.l %d0, %d4, %acc0
move.l (%a1), %d1
beq .L403
eor.l %d1, %d0 | else compare signs
bge .L407 | if same, add delta to weight
sub.l %a3, %d4 | else subtract delta from weight
cmp.l %d7, %d4 | check for negative clip limit
bge .L403
move.l %d7, %d4
bra .L403
.L407: add.l %a3, %d4 | add delta to weight
cmp.l %d6, %d4 | check for positive clip limit
ble .L403
move.l %d6, %d4
.L403: move.l %acc0, %d2 | d2 = rounded product
add.l %d1, %d2 | add applied weight to bptr [1], store
move.l %d2, (%a1)+
.L411: cmp.l %a1, %d5 | loop back if bptr < eptr
jbhi term_minus_1
move.l -4(%a1), 8(%a2) | dpp->samples_A [0] = bptr [-1]
jbra finish_up
.L402: move.l (%a1)+, %d0
bne .L410
.L401: addq.l #4, %a1
bra .L411
|------------------------------------------------------------------------------
| Loop to handle term = -2 condition
|
| a0 = d0 = decorrelation sample
| a1 = bptr d1 = initial bptr [0]
| a2 = dpp-> d2 = updated bptr [0]
| a3 = dpp->delta << 17 d3 = dpp->weight_A << 17
| a4 = d4 = dpp->weight_B << 17
| a5 = d5 = eptr
| a6 = d6 = 1024 << 17
| a7 = d7 = -1024 << 17
| macsr = 0x20 acc1 = 0x00 0000 80
|------------------------------------------------------------------------------
term_minus_2:
move.l -8(%a1), %d0 | d0 = bptr [-2]
beq .L511
move.l %acc1, %acc0
asl.l #4, %d0 | acc0 = acc1 + ((d0 << 4) * weight_B)
mac.l %d0, %d4, %acc0
move.l 4(%a1), %d1
beq .L505
eor.l %d1, %d0 | else compare signs
bge .L504 | if same, add delta to weight
sub.l %a3, %d4 | else subtract delta from weight
cmp.l %d7, %d4 | ckeck for negative clip limit
bge .L505
move.l %d7, %d4
bra .L505
.L504: add.l %a3, %d4 | add delta to weight
cmp.l %d6, %d4 | check for positive clip limit
ble .L505
move.l %d6, %d4
.L505: move.l %acc0, %d0 | d2 = rounded product
add.l %d1, %d0 | add applied weight to bptr [0], store
move.l %d0, 4(%a1)
beq .L512
.L510: move.l %acc1, %acc0
asl.l #4, %d0 | acc0 = acc1 + ((d0 << 4) * weight_A)
mac.l %d0, %d3, %acc0
move.l (%a1), %d1
beq .L503
eor.l %d1, %d0 | else compare signs
bge .L507 | if same, add delta to weight
sub.l %a3, %d3 | else subtract delta from weight
cmp.l %d7, %d3 | check for negative clip limit
bge .L503
move.l %d7, %d3
bra .L503
.L507: add.l %a3, %d3 | add delta to weight
cmp.l %d6, %d3 | check for negative clip limit
ble .L503
move.l %d6, %d3
.L503: move.l %acc0, %d2 | d2 = rounded product
add.l %d1, %d2 | add applied weight to bptr [1], store
move.l %d2, (%a1)
.L512: addq.l #8, %a1
cmp.l %a1, %d5 | loop if bptr < eptr
jbhi term_minus_2
move.l -8(%a1), 40(%a2) | dpp->samples_B [0] = bptr [-4]
jbra finish_up
.L511: move.l 4(%a1), %d0
beq .L512
bra .L510
|------------------------------------------------------------------------------
| Loop to handle term = -3 condition
|
| a0 = d0 = decorrelation sample
| a1 = bptr d1 = initial bptr [0]
| a2 = dpp-> d2 = updated bptr [0]
| a3 = dpp->delta << 17 d3 = dpp->weight_A << 17
| a4 = d4 = dpp->weight_B << 17
| a5 = d5 = eptr
| a6 = d6 = 1024 << 17
| a7 = d7 = -1024 << 17
| macsr = 0x20 acc1 = 0x00 0000 80
|------------------------------------------------------------------------------
term_minus_3:
move.l -4(%a1), %d0 | d0 = bptr [-1]
beq .L301
move.l %acc1, %acc0
asl.l #4, %d0 | acc0 = acc1 + ((d0 << 4) * weight_A)
mac.l %d0, %d3, %acc0
move.l (%a1), %d1
beq .L320
eor.l %d1, %d0 | else compare signs
bge .L319 | if same, add delta to weight
sub.l %a3, %d3 | else subtract delta from weight
cmp.l %d7, %d3 | check for negative clip limit
bge .L320
move.l %d7, %d3
bra .L320
.L319: add.l %a3, %d3 | add delta to weight
cmp.l %d6, %d3 | check for positive clip limit
ble .L320
move.l %d6, %d3
.L320: move.l %acc0, %d2 | d2 = rounded product
add.l %d1, %d2 | add applied weight to bptr [0], store
move.l %d2, (%a1)+
.L330: move.l -12(%a1), %d0 | d0 = bptr [-2]
beq .L302
move.l %acc1, %acc0
asl.l #4, %d0 | acc0 = acc1 + ((d0 << 4) * weight_B)
mac.l %d0, %d4, %acc0
move.l (%a1), %d1
beq .L318
eor.l %d1, %d0 | else compare signs
bge .L322 | if same, add delta to weight
sub.l %a3, %d4 | else subtract delta from weight
cmp.l %d7, %d4 | check for negative clip limit
bge .L318
move.l %d7, %d4
bra .L318
.L322: add.l %a3, %d4 | add delta to weight
cmp.l %d6, %d4 | check for positive clip limit
ble .L318
move.l %d6, %d4
.L318: move.l %acc0, %d2 | d2 = rounded product
add.l %d1, %d2 | add applied weight to bptr [1], store
move.l %d2, (%a1)+
.L331: cmp.l %a1, %d5 | bptr, eptr
jbhi term_minus_3
move.l -4(%a1), 8(%a2) | dpp->samples_A [0] = bptr [-1]
move.l -8(%a1), 40(%a2) | dpp->samples_B [0] = bptr [-2]
jbra finish_up
.L301: addq.l #4, %a1
bra .L330
.L302: addq.l #4, %a1
bra .L331
| finish and return
finish_up:
moveq.l #17, %d0
asr.l %d0, %d3
asr.l %d0, %d4
move.w %d3, 4(%a2) | weight_A, dpp->weight_A
move.w %d4, 6(%a2) | weight_B, dpp->weight_B
clr.l %d0 | clear up EMAC
move.l %d0, %acc0
move.l %d0, %acc1
return_only:
movem.l (%sp), %d2-%d7/%a2-%a6
lea (44,%sp), %sp
rts
|
