summaryrefslogtreecommitdiff
path: root/muse_qt4_evolution/al/dspSSE.cpp
blob: 0f3d84b9a80712bd266ae6450c11ce8b602f37e5 (plain)
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
/*
    Copyright (C) 2005 Paul Davis

    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 program is distributed in the hope that it will be useful,
    but WITHOUT ANY WARRANTY; without even the implied warranty of
    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
    GNU General Public License for more details.

    You should have received a copy of the GNU General Public License
    along with this program; if not, write to the Free Software
    Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.

	Author: Sampo Savolainen

    $Id: sse_functions.s 988 2006-10-17 20:40:39Z paul $
*/


#; void x86_sse_mix_buffers_with_gain (float *dst, float *src, long nframes, float gain);

.globl x86_sse_mix_buffers_with_gain
	.type	x86_sse_mix_buffers_with_gain,@function

x86_sse_mix_buffers_with_gain:
#; 8(%ebp)	= float	*dst 	= %edi
#; 12(%ebp) = float *src	= %esi
#; 16(%ebp) = long	nframes = %ecx
#; 20(%ebp) = float	gain    = st(0)

	pushl %ebp
	movl %esp, %ebp

	#; save the registers
#;	pushl %eax
	pushl %ebx
#;	pushl %ecx
	pushl %edi
	pushl %esi
	
	#; if nframes == 0, go to end
	movl 16(%ebp), %ecx #; nframes
	cmp	$0, %ecx
	je	.MBWG_END

	#; Check for alignment

	movl 8(%ebp), %edi  #; dst 
	movl 12(%ebp), %esi #; src

	movl %edi, %eax
	andl $12, %eax #; mask alignemnt offset

	movl %esi, %ebx
	andl $12, %ebx #; mask alignment offset

	cmp %eax, %ebx
	jne .MBWG_NONALIGN #; if not aligned, calculate manually

	#; if we are aligned
	cmp $0, %ebx
	jz .MBWG_SSE
	
	#; Pre-loop, we need to run 1-3 frames "manually" without
	#; SSE instructions

	movss 20(%ebp), %xmm1 #; xmm1

.MBWG_PRELOOP:
	
	movss (%esi), %xmm0
	mulss %xmm1, %xmm0
	addss (%edi), %xmm0
	movss %xmm0, (%edi)

	addl $4, %edi #; dst++
	addl $4, %esi #; src++
	decl %ecx 	  #; nframes--
	jz .MBWG_END

#;	cmp $0, %ecx
#;	je .MBWG_END #; if we run out of frames, go to end
	
	addl $4, %ebx
	
	cmp $16, %ebx #; test if we've reached 16 byte alignment
	jne .MBWG_PRELOOP


.MBWG_SSE:

	cmp $4, %ecx #; we know it's not zero, but if it's not >=4, then
	jnge .MBWG_NONALIGN #; we jump straight to the "normal" code

	#; copy gain to fill %xmm1
	movss   20(%ebp), %xmm1
    shufps  $0x00, %xmm1, %xmm1


.MBWG_SSELOOP:

	movaps	(%esi), %xmm0 #; source => xmm0
	mulps	%xmm1,  %xmm0 #; apply gain to source
	addps	(%edi), %xmm0 #; mix with destination
	movaps  %xmm0, (%edi) #; copy result to destination
	
	addl $16, %edi #; dst+=4
	addl $16, %esi #; src+=4

	subl $4, %ecx #; nframes-=4
	cmp $4, %ecx
	jge .MBWG_SSELOOP

	cmp $0, %ecx
	je .MBWG_END

	#; if there are remaining frames, the nonalign code will do nicely
	#; for the rest 1-3 frames.
	
.MBWG_NONALIGN:
	#; not aligned!

	movss 20(%ebp), %xmm1 #; gain => xmm1

.MBWG_NONALIGNLOOP:

	movss (%esi), %xmm0
	mulss %xmm1, %xmm0
	addss (%edi), %xmm0
	movss %xmm0, (%edi)
	
	addl $4, %edi
	addl $4, %esi
	
	decl %ecx
	jnz .MBWG_NONALIGNLOOP

.MBWG_END:

	popl %esi
	popl %edi
#;	popl %ecx
	popl %ebx
#;	popl %eax
	
	#; return
	leave
	ret

.size	x86_sse_mix_buffers_with_gain, .-x86_sse_mix_buffers_with_gain




#; void x86_sse_mix_buffers_no_gain (float *dst, float *src, long nframes);

.globl x86_sse_mix_buffers_no_gain
	.type	x86_sse_mix_buffers_no_gain,@function

x86_sse_mix_buffers_no_gain:
#; 8(%ebp)	= float	*dst 	= %edi
#; 12(%ebp) = float *src	= %esi
#; 16(%ebp) = long	nframes = %ecx

	pushl %ebp
	movl %esp, %ebp

	#; save the registers
#;	pushl %eax
	pushl %ebx
#;	pushl %ecx
	pushl %edi
	pushl %esi
	
	#; the real function

	#; if nframes == 0, go to end
	movl 16(%ebp), %ecx #; nframes
	cmp	$0, %ecx
	je	.MBNG_END

	#; Check for alignment

	movl 8(%ebp), %edi  #; dst 
	movl 12(%ebp), %esi #; src

	movl %edi, %eax
	andl $12, %eax #; mask alignemnt offset

	movl %esi, %ebx
	andl $12, %ebx #; mask alignment offset

	cmp %eax, %ebx
	jne .MBNG_NONALIGN #; if not aligned, calculate manually

	cmp $0, %ebx
	je .MBNG_SSE

	#; Pre-loop, we need to run 1-3 frames "manually" without
	#; SSE instructions

.MBNG_PRELOOP:
		
	movss (%esi), %xmm0
	addss (%edi), %xmm0
	movss %xmm0, (%edi)

	addl $4, %edi #; dst++
	addl $4, %esi #; src++
	decl %ecx 	  #; nframes--
	jz	.MBNG_END
	addl $4, %ebx
	
	cmp $16, %ebx #; test if we've reached 16 byte alignment
	jne .MBNG_PRELOOP

.MBNG_SSE:

	cmp $4, %ecx #; if there are frames left, but less than 4
	jnge .MBNG_NONALIGN #; we can't run SSE

.MBNG_SSELOOP:

	movaps	(%esi), %xmm0 #; source => xmm0
	addps	(%edi), %xmm0 #; mix with destination
	movaps  %xmm0, (%edi) #; copy result to destination
	
	addl $16, %edi #; dst+=4
	addl $16, %esi #; src+=4

	subl $4, %ecx #; nframes-=4
	cmp $4, %ecx
	jge .MBNG_SSELOOP

	cmp $0, %ecx
	je .MBNG_END

	#; if there are remaining frames, the nonalign code will do nicely
	#; for the rest 1-3 frames.
	
.MBNG_NONALIGN:
	#; not aligned!

	movss (%esi), %xmm0 #; src => xmm0
	addss (%edi), %xmm0 #; xmm0 += dst
	movss %xmm0, (%edi) #; xmm0 => dst
	
	addl $4, %edi
	addl $4, %esi
	
	decl %ecx
	jnz .MBNG_NONALIGN

.MBNG_END:

	popl %esi
	popl %edi
#;	popl %ecx
	popl %ebx
#;	popl %eax
	
	#; return
	leave
	ret

.size	x86_sse_mix_buffers_no_gain, .-x86_sse_mix_buffers_no_gain




#; void x86_sse_apply_gain_to_buffer (float *buf, long nframes, float gain);

.globl x86_sse_apply_gain_to_buffer
	.type	x86_sse_apply_gain_to_buffer,@function

x86_sse_apply_gain_to_buffer:
#; 8(%ebp)	= float	*buf 	= %edi
#; 12(%ebp) = long	nframes = %ecx
#; 16(%ebp) = float	gain    = st(0)

	pushl %ebp
	movl %esp, %ebp

	#; save %edi
	pushl %edi
	
	#; the real function

	#; if nframes == 0, go to end
	movl 12(%ebp), %ecx #; nframes
	cmp	$0, %ecx
	je	.AG_END

	#; create the gain buffer in %xmm1
	movss	16(%ebp), %xmm1
	shufps	$0x00, %xmm1, %xmm1
	
	#; Check for alignment

	movl 8(%ebp), %edi #; buf 
	movl %edi, %edx #; buf => %edx
	andl $12, %edx #; mask bits 1 & 2, result = 0, 4, 8 or 12
	jz	.AG_SSE #; if buffer IS aligned

	#; PRE-LOOP
	#; we iterate 1-3 times, doing normal x87 float comparison
	#; so we reach a 16 byte aligned "buf" (=%edi) value

.AGLP_START:

	#; Load next value from the buffer
	movss (%edi), %xmm0
	mulss %xmm1, %xmm0
	movss %xmm0, (%edi)

	#; increment buffer, decrement counter
	addl $4, %edi #; buf++;
	
	decl %ecx   #; nframes--
	jz	.AG_END #; if we run out of frames, we go to the end
	
	addl $4, %edx #; one non-aligned byte less
	cmp $16, %edx
	jne .AGLP_START #; if more non-aligned frames exist, we do a do-over

.AG_SSE:

	#; We have reached the 16 byte aligned "buf" ("edi") value

	#; Figure out how many loops we should do
	movl %ecx, %eax #; copy remaining nframes to %eax for division
	movl $0, %edx   #; 0 the edx register
	
	
	pushl %edi
	movl $4, %edi
	divl %edi #; %edx = remainder == 0
	popl %edi

	#; %eax = SSE iterations
	cmp $0, %eax
	je .AGPOST_START

	
.AGLP_SSE:

	movaps (%edi), %xmm0
	mulps %xmm1, %xmm0
	movaps %xmm0, (%edi)

	addl $16, %edi
#;	subl $4, %ecx   #; nframes-=4

	decl %eax
	jnz .AGLP_SSE

	#; Next we need to post-process all remaining frames
	#; the remaining frame count is in %ecx
	
	#; if no remaining frames, jump to the end
#;	cmp $0, %ecx
	andl $3, %ecx #; nframes % 4
	je .AG_END

.AGPOST_START:

	movss (%edi), %xmm0
	mulss %xmm1, %xmm0
	movss %xmm0, (%edi)

	#; increment buffer, decrement counter
	addl $4, %edi #; buf++;
	
	decl %ecx   #; nframes--
	jnz	.AGPOST_START #; if we run out of frames, we go to the end
	
.AG_END:


	popl %edi
	
	#; return
	leave
	ret

.size	x86_sse_apply_gain_to_buffer, .-x86_sse_apply_gain_to_buffer
#; end proc



#; float x86_sse_compute_peak(float *buf, long nframes, float current);

.globl x86_sse_compute_peak
	.type	x86_sse_compute_peak,@function

x86_sse_compute_peak:
#; 8(%ebp)	= float	*buf 	= %edi
#; 12(%ebp) = long	nframes = %ecx
#; 16(%ebp) = float	current = st(0)

	pushl %ebp
	movl %esp, %ebp

	#; save %edi
	pushl %edi
	
	#; the real function

	#; Load "current" in xmm0
	movss 16(%ebp), %xmm0

	#; if nframes == 0, go to end
	movl 12(%ebp), %ecx #; nframes
	cmp	$0, %ecx
	je	.CP_END

	#; create the "abs" mask in %xmm2
	pushl	$2147483647
	movss	(%esp), %xmm2
	addl    $4, %esp
	shufps	$0x00, %xmm2, %xmm2

	#; Check for alignment

	movl 8(%ebp), %edi #; buf 
	movl %edi, %edx #; buf => %edx
	andl $12, %edx #; mask bits 1 & 2, result = 0, 4, 8 or 12
	jz	.CP_SSE #; if buffer IS aligned

	#; PRE-LOOP
	#; we iterate 1-3 times, doing normal x87 float comparison
	#; so we reach a 16 byte aligned "buf" (=%edi) value

.LP_START:

	#; Load next value from the buffer
	movss (%edi), %xmm1
	andps %xmm2, %xmm1
	maxss %xmm1, %xmm0

	#; increment buffer, decrement counter
	addl $4, %edi #; buf++;
	
	decl %ecx   #; nframes--
	jz	.CP_END #; if we run out of frames, we go to the end
	
	addl $4, %edx #; one non-aligned byte less
	cmp $16, %edx
	jne .LP_START #; if more non-aligned frames exist, we do a do-over

.CP_SSE:

	#; We have reached the 16 byte aligned "buf" ("edi") value

	#; Figure out how many loops we should do
	movl %ecx, %eax #; copy remaining nframes to %eax for division

	shr $2,%eax #; unsigned divide by 4
	jz .POST_START

	#; %eax = SSE iterations

	#; current maximum is at %xmm0, but we need to ..
	shufps $0x00, %xmm0, %xmm0 #; shuffle "current" to all 4 FP's

	#;prefetcht0 16(%edi)

.LP_SSE:

	movaps (%edi), %xmm1
	andps %xmm2, %xmm1
	maxps %xmm1, %xmm0

	addl $16, %edi

	decl %eax
	jnz .LP_SSE

	#; Calculate the maximum value contained in the 4 FP's in %xmm0
	movaps %xmm0, %xmm1
	shufps $0x4e, %xmm1, %xmm1 #; shuffle left & right pairs (1234 => 3412)
	maxps  %xmm1, %xmm0 #; maximums of the two pairs
	movaps %xmm0, %xmm1
	shufps $0xb1, %xmm1, %xmm1 #; shuffle the floats inside the two pairs (1234 => 2143)
	maxps  %xmm1, %xmm0 

	#; now every float in %xmm0 is the same value, current maximum value
	
	#; Next we need to post-process all remaining frames
	#; the remaining frame count is in %ecx
	
	#; if no remaining frames, jump to the end

	andl $3, %ecx #; nframes % 4
	jz .CP_END

.POST_START:

	movss (%edi), %xmm1
	andps %xmm2, %xmm1
	maxss %xmm1, %xmm0
	
	addl $4, %edi 	#; buf++;
	
	decl %ecx		#; nframes--;
	jnz .POST_START

.CP_END:

	#; Load the value from xmm0 to the float stack for returning
	movss %xmm0, 16(%ebp)
	flds 16(%ebp)

	popl %edi
	
	#; return
	leave
	ret

.size	x86_sse_compute_peak, .-x86_sse_compute_peak
#; end proc

#ifdef __ELF__
.section .note.GNU-stack,"",%progbits
#endif