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
|
/*
* The simplest mpeg encoder (well, it was the simplest!)
* Copyright (c) 2000,2001 Gerard Lantau.
*
* 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.
*
* Optimized for ia32 cpus by Nick Kurshev <nickols_k@mail.ru>
* h263 dequantizer by Michael Niedermayer <michaelni@gmx.at>
*/
#include "../dsputil.h"
#include "../mpegvideo.h"
extern UINT8 zigzag_end[64];
extern void (*draw_edges)(UINT8 *buf, int wrap, int width, int height, int w);
#if 0
/* XXX: GL: I don't understand why this function needs optimization
(it is called only once per frame!), so I disabled it */
void MPV_frame_start(MpegEncContext *s)
{
if (s->pict_type == B_TYPE) {
__asm __volatile(
"movl (%1), %%eax\n\t"
"movl 4(%1), %%edx\n\t"
"movl 8(%1), %%ecx\n\t"
"movl %%eax, (%0)\n\t"
"movl %%edx, 4(%0)\n\t"
"movl %%ecx, 8(%0)\n\t"
:
:"r"(s->current_picture), "r"(s->aux_picture)
:"eax","edx","ecx","memory");
} else {
/* swap next and last */
__asm __volatile(
"movl (%1), %%eax\n\t"
"movl 4(%1), %%edx\n\t"
"movl 8(%1), %%ecx\n\t"
"xchgl (%0), %%eax\n\t"
"xchgl 4(%0), %%edx\n\t"
"xchgl 8(%0), %%ecx\n\t"
"movl %%eax, (%1)\n\t"
"movl %%edx, 4(%1)\n\t"
"movl %%ecx, 8(%1)\n\t"
"movl %%eax, (%2)\n\t"
"movl %%edx, 4(%2)\n\t"
"movl %%ecx, 8(%2)\n\t"
:
:"r"(s->last_picture), "r"(s->next_picture), "r"(s->current_picture)
:"eax","edx","ecx","memory");
}
}
#endif
static const unsigned long long int mm_wabs __attribute__ ((aligned(8))) = 0xffffffffffffffffULL;
static const unsigned long long int mm_wone __attribute__ ((aligned(8))) = 0x0001000100010001ULL;
static void dct_unquantize_h263_mmx(MpegEncContext *s,
DCTELEM *block, int n, int qscale)
{
int i, level, qmul, qadd, nCoeffs;
qmul = s->qscale << 1;
qadd = (s->qscale - 1) | 1;
if (s->mb_intra) {
if (n < 4)
block[0] = block[0] * s->y_dc_scale;
else
block[0] = block[0] * s->c_dc_scale;
for(i=1; i<8; i++) {
level = block[i];
if (level) {
if (level < 0) {
level = level * qmul - qadd;
} else {
level = level * qmul + qadd;
}
block[i] = level;
}
}
nCoeffs=64;
} else {
i = 0;
nCoeffs= zigzag_end[ s->block_last_index[n] ];
}
//printf("%d %d ", qmul, qadd);
asm volatile(
"movd %1, %%mm6 \n\t" //qmul
"packssdw %%mm6, %%mm6 \n\t"
"packssdw %%mm6, %%mm6 \n\t"
"movd %2, %%mm5 \n\t" //qadd
"pxor %%mm7, %%mm7 \n\t"
"packssdw %%mm5, %%mm5 \n\t"
"packssdw %%mm5, %%mm5 \n\t"
"psubw %%mm5, %%mm7 \n\t"
"pxor %%mm4, %%mm4 \n\t"
".balign 16\n\t"
"1: \n\t"
"movq (%0, %3), %%mm0 \n\t"
"movq 8(%0, %3), %%mm1 \n\t"
"pmullw %%mm6, %%mm0 \n\t"
"pmullw %%mm6, %%mm1 \n\t"
"movq (%0, %3), %%mm2 \n\t"
"movq 8(%0, %3), %%mm3 \n\t"
"pcmpgtw %%mm4, %%mm2 \n\t" // block[i] < 0 ? -1 : 0
"pcmpgtw %%mm4, %%mm3 \n\t" // block[i] < 0 ? -1 : 0
"pxor %%mm2, %%mm0 \n\t"
"pxor %%mm3, %%mm1 \n\t"
"paddw %%mm7, %%mm0 \n\t"
"paddw %%mm7, %%mm1 \n\t"
"pxor %%mm0, %%mm2 \n\t"
"pxor %%mm1, %%mm3 \n\t"
"pcmpeqw %%mm7, %%mm0 \n\t" // block[i] == 0 ? -1 : 0
"pcmpeqw %%mm7, %%mm1 \n\t" // block[i] == 0 ? -1 : 0
"pandn %%mm2, %%mm0 \n\t"
"pandn %%mm3, %%mm1 \n\t"
"movq %%mm0, (%0, %3) \n\t"
"movq %%mm1, 8(%0, %3) \n\t"
"addl $16, %3 \n\t"
"js 1b \n\t"
::"r" (block+nCoeffs), "g"(qmul), "g" (qadd), "r" (2*(i-nCoeffs))
: "memory"
);
}
/*
NK:
Note: looking at PARANOID:
"enable all paranoid tests for rounding, overflows, etc..."
#ifdef PARANOID
if (level < -2048 || level > 2047)
fprintf(stderr, "unquant error %d %d\n", i, level);
#endif
We can suppose that result of two multiplications can't be greate of 0xFFFF
i.e. is 16-bit, so we use here only PMULLW instruction and can avoid
a complex multiplication.
=====================================================
Full formula for multiplication of 2 integer numbers
which are represent as high:low words:
input: value1 = high1:low1
value2 = high2:low2
output: value3 = value1*value2
value3=high3:low3 (on overflow: modulus 2^32 wrap-around)
this mean that for 0x123456 * 0x123456 correct result is 0x766cb0ce4
but this algorithm will compute only 0x66cb0ce4
this limited by 16-bit size of operands
---------------------------------
tlow1 = high1*low2
tlow2 = high2*low1
tlow1 = tlow1 + tlow2
high3:low3 = low1*low2
high3 += tlow1
*/
static void dct_unquantize_mpeg1_mmx(MpegEncContext *s,
DCTELEM *block, int n, int qscale)
{
int i, level, nCoeffs;
const UINT16 *quant_matrix;
if(s->alternate_scan) nCoeffs= 64;
else nCoeffs= nCoeffs= zigzag_end[ s->block_last_index[n] ];
if (s->mb_intra) {
if (n < 4)
block[0] = block[0] * s->y_dc_scale;
else
block[0] = block[0] * s->c_dc_scale;
/* isnt used anymore (we have a h263 unquantizer since some time)
if (s->out_format == FMT_H263) {
i = 1;
goto unquant_even;
}*/
/* XXX: only mpeg1 */
quant_matrix = s->intra_matrix;
i=1;
/* Align on 4 elements boundary */
while(i&3)
{
level = block[i];
if (level) {
if (level < 0) level = -level;
level = (int)(level * qscale * quant_matrix[i]) >> 3;
level = (level - 1) | 1;
if (block[i] < 0) level = -level;
block[i] = level;
}
i++;
}
__asm __volatile(
"movd %0, %%mm6\n\t" /* mm6 = qscale | 0 */
"punpckldq %%mm6, %%mm6\n\t" /* mm6 = qscale | qscale */
"movq %2, %%mm4\n\t"
"movq %%mm6, %%mm7\n\t"
"movq %1, %%mm5\n\t"
"packssdw %%mm6, %%mm7\n\t" /* mm7 = qscale | qscale | qscale | qscale */
"pxor %%mm6, %%mm6\n\t"
::"g"(qscale),"m"(mm_wone),"m"(mm_wabs):"memory");
for(;i<nCoeffs;i+=4) {
__asm __volatile(
"movq %1, %%mm0\n\t"
"movq %%mm7, %%mm1\n\t"
"movq %%mm0, %%mm2\n\t"
"movq %%mm0, %%mm3\n\t"
"pcmpgtw %%mm6, %%mm2\n\t"
"pmullw %2, %%mm1\n\t"
"pandn %%mm4, %%mm2\n\t"
"por %%mm5, %%mm2\n\t"
"pmullw %%mm2, %%mm0\n\t" /* mm0 = abs(block[i]). */
"pcmpeqw %%mm6, %%mm3\n\t"
"pmullw %%mm0, %%mm1\n\t"
"psraw $3, %%mm1\n\t"
"psubw %%mm5, %%mm1\n\t" /* block[i] --; */
"pandn %%mm4, %%mm3\n\t" /* fake of pcmpneqw : mm0 != 0 then mm1 = -1 */
"por %%mm5, %%mm1\n\t" /* block[i] |= 1 */
"pmullw %%mm2, %%mm1\n\t" /* change signs again */
"pand %%mm3, %%mm1\n\t" /* nullify if was zero */
"movq %%mm1, %0"
:"=m"(block[i])
:"m"(block[i]), "m"(quant_matrix[i])
:"memory");
}
} else {
i = 0;
unquant_even:
quant_matrix = s->non_intra_matrix;
/* Align on 4 elements boundary */
while(i&7)
{
level = block[i];
if (level) {
if (level < 0) level = -level;
level = (((level << 1) + 1) * qscale *
((int) quant_matrix[i])) >> 4;
level = (level - 1) | 1;
if(block[i] < 0) level = -level;
block[i] = level;
}
i++;
}
asm volatile(
"pcmpeqw %%mm7, %%mm7 \n\t"
"psrlw $15, %%mm7 \n\t"
"movd %2, %%mm6 \n\t"
"packssdw %%mm6, %%mm6 \n\t"
"packssdw %%mm6, %%mm6 \n\t"
".balign 16\n\t"
"1: \n\t"
"movq (%0, %3), %%mm0 \n\t"
"movq 8(%0, %3), %%mm1 \n\t"
"movq (%1, %3), %%mm4 \n\t"
"movq 8(%1, %3), %%mm5 \n\t"
"pmullw %%mm6, %%mm4 \n\t" // q=qscale*quant_matrix[i]
"pmullw %%mm6, %%mm5 \n\t" // q=qscale*quant_matrix[i]
"pxor %%mm2, %%mm2 \n\t"
"pxor %%mm3, %%mm3 \n\t"
"pcmpgtw %%mm0, %%mm2 \n\t" // block[i] < 0 ? -1 : 0
"pcmpgtw %%mm1, %%mm3 \n\t" // block[i] < 0 ? -1 : 0
"pxor %%mm2, %%mm0 \n\t"
"pxor %%mm3, %%mm1 \n\t"
"psubw %%mm2, %%mm0 \n\t" // abs(block[i])
"psubw %%mm3, %%mm1 \n\t" // abs(block[i])
"paddw %%mm0, %%mm0 \n\t" // abs(block[i])*2
"paddw %%mm1, %%mm1 \n\t" // abs(block[i])*2
"paddw %%mm7, %%mm0 \n\t" // abs(block[i])*2 + 1
"paddw %%mm7, %%mm1 \n\t" // abs(block[i])*2 + 1
"pmullw %%mm4, %%mm0 \n\t" // (abs(block[i])*2 + 1)*q
"pmullw %%mm5, %%mm1 \n\t" // (abs(block[i])*2 + 1)*q
"pxor %%mm4, %%mm4 \n\t"
"pxor %%mm5, %%mm5 \n\t" // FIXME slow
"pcmpeqw (%0, %3), %%mm4 \n\t" // block[i] == 0 ? -1 : 0
"pcmpeqw 8(%0, %3), %%mm5 \n\t" // block[i] == 0 ? -1 : 0
"psraw $4, %%mm0 \n\t"
"psraw $4, %%mm1 \n\t"
"psubw %%mm7, %%mm0 \n\t"
"psubw %%mm7, %%mm1 \n\t"
"por %%mm7, %%mm0 \n\t"
"por %%mm7, %%mm1 \n\t"
"pxor %%mm2, %%mm0 \n\t"
"pxor %%mm3, %%mm1 \n\t"
"psubw %%mm2, %%mm0 \n\t"
"psubw %%mm3, %%mm1 \n\t"
"pandn %%mm0, %%mm4 \n\t"
"pandn %%mm1, %%mm5 \n\t"
"movq %%mm4, (%0, %3) \n\t"
"movq %%mm5, 8(%0, %3) \n\t"
"addl $16, %3 \n\t"
"js 1b \n\t"
::"r" (block+nCoeffs), "r"(quant_matrix+nCoeffs), "g" (qscale), "r" (2*(i-nCoeffs))
: "memory"
);
}
}
/* draw the edges of width 'w' of an image of size width, height
this mmx version can only handle w==8 || w==16 */
static void draw_edges_mmx(UINT8 *buf, int wrap, int width, int height, int w)
{
UINT8 *ptr, *last_line;
int i;
last_line = buf + (height - 1) * wrap;
/* left and right */
ptr = buf;
if(w==8)
{
asm volatile(
"1: \n\t"
"movd (%0), %%mm0 \n\t"
"punpcklbw %%mm0, %%mm0 \n\t"
"punpcklwd %%mm0, %%mm0 \n\t"
"punpckldq %%mm0, %%mm0 \n\t"
"movq %%mm0, -8(%0) \n\t"
"movq -8(%0, %2), %%mm1 \n\t"
"punpckhbw %%mm1, %%mm1 \n\t"
"punpckhwd %%mm1, %%mm1 \n\t"
"punpckhdq %%mm1, %%mm1 \n\t"
"movq %%mm1, (%0, %2) \n\t"
"addl %1, %0 \n\t"
"cmpl %3, %0 \n\t"
" jb 1b \n\t"
: "+r" (ptr)
: "r" (wrap), "r" (width), "r" (ptr + wrap*height)
);
}
else
{
asm volatile(
"1: \n\t"
"movd (%0), %%mm0 \n\t"
"punpcklbw %%mm0, %%mm0 \n\t"
"punpcklwd %%mm0, %%mm0 \n\t"
"punpckldq %%mm0, %%mm0 \n\t"
"movq %%mm0, -8(%0) \n\t"
"movq %%mm0, -16(%0) \n\t"
"movq -8(%0, %2), %%mm1 \n\t"
"punpckhbw %%mm1, %%mm1 \n\t"
"punpckhwd %%mm1, %%mm1 \n\t"
"punpckhdq %%mm1, %%mm1 \n\t"
"movq %%mm1, (%0, %2) \n\t"
"movq %%mm1, 8(%0, %2) \n\t"
"addl %1, %0 \n\t"
"cmpl %3, %0 \n\t"
" jb 1b \n\t"
: "+r" (ptr)
: "r" (wrap), "r" (width), "r" (ptr + wrap*height)
);
}
for(i=0;i<w;i+=4) {
/* top and bottom (and hopefully also the corners) */
ptr= buf - (i + 1) * wrap - w;
asm volatile(
"1: \n\t"
"movq (%1, %0), %%mm0 \n\t"
"movq %%mm0, (%0) \n\t"
"movq %%mm0, (%0, %2) \n\t"
"movq %%mm0, (%0, %2, 2) \n\t"
"movq %%mm0, (%0, %3) \n\t"
"addl $8, %0 \n\t"
"cmpl %4, %0 \n\t"
" jb 1b \n\t"
: "+r" (ptr)
: "r" ((int)buf - (int)ptr - w), "r" (-wrap), "r" (-wrap*3), "r" (ptr+width+2*w)
);
ptr= last_line + (i + 1) * wrap - w;
asm volatile(
"1: \n\t"
"movq (%1, %0), %%mm0 \n\t"
"movq %%mm0, (%0) \n\t"
"movq %%mm0, (%0, %2) \n\t"
"movq %%mm0, (%0, %2, 2) \n\t"
"movq %%mm0, (%0, %3) \n\t"
"addl $8, %0 \n\t"
"cmpl %4, %0 \n\t"
" jb 1b \n\t"
: "+r" (ptr)
: "r" ((int)last_line - (int)ptr - w), "r" (wrap), "r" (wrap*3), "r" (ptr+width+2*w)
);
}
}
void MPV_common_init_mmx(MpegEncContext *s)
{
if (mm_flags & MM_MMX) {
if (s->out_format == FMT_H263)
s->dct_unquantize = dct_unquantize_h263_mmx;
else
s->dct_unquantize = dct_unquantize_mpeg1_mmx;
draw_edges = draw_edges_mmx;
}
}
|
