mpegvideo.c 20.4 KB
Newer Older
Nick Kurshev's avatar
Nick Kurshev committed
1
/*
2 3 4
 * Optimized for ia32 CPUs by Nick Kurshev <nickols_k@mail.ru>
 * h263, mpeg1, mpeg2 dequantizer & draw_edges by Michael Niedermayer <michaelni@gmx.at>
 *
5 6 7
 * This file is part of FFmpeg.
 *
 * FFmpeg is free software; you can redistribute it and/or
8 9
 * modify it under the terms of the GNU Lesser General Public
 * License as published by the Free Software Foundation; either
10
 * version 2.1 of the License, or (at your option) any later version.
Nick Kurshev's avatar
Nick Kurshev committed
11
 *
12
 * FFmpeg is distributed in the hope that it will be useful,
Nick Kurshev's avatar
Nick Kurshev committed
13
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 15
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * Lesser General Public License for more details.
Nick Kurshev's avatar
Nick Kurshev committed
16
 *
17
 * You should have received a copy of the GNU Lesser General Public
18
 * License along with FFmpeg; if not, write to the Free Software
19
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
Nick Kurshev's avatar
Nick Kurshev committed
20 21
 */

22
#include "libavutil/attributes.h"
23
#include "libavutil/cpu.h"
24
#include "libavutil/x86/asm.h"
25
#include "libavutil/x86/cpu.h"
26 27
#include "libavcodec/avcodec.h"
#include "libavcodec/mpegvideo.h"
28

29
#if HAVE_MMX_INLINE
30

31
static void dct_unquantize_h263_intra_mmx(MpegEncContext *s,
Diego Biurrun's avatar
Diego Biurrun committed
32
                                  int16_t *block, int n, int qscale)
33
{
34
    x86_reg level, qmul, qadd, nCoeffs;
35 36

    qmul = qscale << 1;
37

38
    av_assert2(s->block_last_index[n]>=0 || s->h263_aic);
39

40 41 42 43 44 45 46 47 48
    if (!s->h263_aic) {
        if (n < 4)
            level = block[0] * s->y_dc_scale;
        else
            level = block[0] * s->c_dc_scale;
        qadd = (qscale - 1) | 1;
    }else{
        qadd = 0;
        level= block[0];
49
    }
50 51 52 53
    if(s->ac_pred)
        nCoeffs=63;
    else
        nCoeffs= s->inter_scantable.raster_end[ s->block_last_index[n] ];
54

55
__asm__ volatile(
56 57 58 59 60 61 62 63 64
                "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"
65
                ".p2align 4                     \n\t"
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
                "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"

                "add $16, %3                    \n\t"
                "jng 1b                         \n\t"
99
                ::"r" (block+nCoeffs), "rm"(qmul), "rm" (qadd), "r" (2*(-nCoeffs))
100 101
                : "memory"
        );
102 103 104 105 106
        block[0]= level;
}


static void dct_unquantize_h263_inter_mmx(MpegEncContext *s,
Diego Biurrun's avatar
Diego Biurrun committed
107
                                  int16_t *block, int n, int qscale)
108
{
109
    x86_reg qmul, qadd, nCoeffs;
110 111 112 113

    qmul = qscale << 1;
    qadd = (qscale - 1) | 1;

114
    av_assert2(s->block_last_index[n]>=0 || s->h263_aic);
115

116
    nCoeffs= s->inter_scantable.raster_end[ s->block_last_index[n] ];
117

118
__asm__ volatile(
119 120 121 122 123 124 125 126 127
                "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"
128
                ".p2align 4                     \n\t"
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
                "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"

                "add $16, %3                    \n\t"
                "jng 1b                         \n\t"
162
                ::"r" (block+nCoeffs), "rm"(qmul), "rm" (qadd), "r" (2*(-nCoeffs))
163 164
                : "memory"
        );
165 166
}

167
static void dct_unquantize_mpeg1_intra_mmx(MpegEncContext *s,
Diego Biurrun's avatar
Diego Biurrun committed
168
                                     int16_t *block, int n, int qscale)
Nick Kurshev's avatar
Nick Kurshev committed
169
{
170
    x86_reg nCoeffs;
171
    const uint16_t *quant_matrix;
172
    int block0;
173

174
    av_assert2(s->block_last_index[n]>=0);
175 176

    nCoeffs= s->intra_scantable.raster_end[ s->block_last_index[n] ]+1;
177

178
    if (n < 4)
179 180 181 182 183
        block0 = block[0] * s->y_dc_scale;
    else
        block0 = block[0] * s->c_dc_scale;
    /* XXX: only mpeg1 */
    quant_matrix = s->intra_matrix;
184
__asm__ volatile(
185 186 187 188 189 190
                "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"
                "mov %3, %%"REG_a"              \n\t"
191
                ".p2align 4                     \n\t"
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
                "1:                             \n\t"
                "movq (%0, %%"REG_a"), %%mm0    \n\t"
                "movq 8(%0, %%"REG_a"), %%mm1   \n\t"
                "movq (%1, %%"REG_a"), %%mm4    \n\t"
                "movq 8(%1, %%"REG_a"), %%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])
                "pmullw %%mm4, %%mm0            \n\t" // abs(block[i])*q
                "pmullw %%mm5, %%mm1            \n\t" // abs(block[i])*q
                "pxor %%mm4, %%mm4              \n\t"
                "pxor %%mm5, %%mm5              \n\t" // FIXME slow
                "pcmpeqw (%0, %%"REG_a"), %%mm4 \n\t" // block[i] == 0 ? -1 : 0
                "pcmpeqw 8(%0, %%"REG_a"), %%mm5\n\t" // block[i] == 0 ? -1 : 0
                "psraw $3, %%mm0                \n\t"
                "psraw $3, %%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, %%"REG_a")    \n\t"
                "movq %%mm5, 8(%0, %%"REG_a")   \n\t"

                "add $16, %%"REG_a"             \n\t"
                "js 1b                          \n\t"
230
                ::"r" (block+nCoeffs), "r"(quant_matrix+nCoeffs), "rm" (qscale), "g" (-2*nCoeffs)
231 232
                : "%"REG_a, "memory"
        );
233 234 235 236
    block[0]= block0;
}

static void dct_unquantize_mpeg1_inter_mmx(MpegEncContext *s,
Diego Biurrun's avatar
Diego Biurrun committed
237
                                     int16_t *block, int n, int qscale)
238
{
239
    x86_reg nCoeffs;
240 241
    const uint16_t *quant_matrix;

242
    av_assert2(s->block_last_index[n]>=0);
243 244

    nCoeffs= s->intra_scantable.raster_end[ s->block_last_index[n] ]+1;
245

246
        quant_matrix = s->inter_matrix;
247
__asm__ volatile(
248 249 250 251 252 253
                "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"
                "mov %3, %%"REG_a"              \n\t"
254
                ".p2align 4                     \n\t"
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
                "1:                             \n\t"
                "movq (%0, %%"REG_a"), %%mm0    \n\t"
                "movq 8(%0, %%"REG_a"), %%mm1   \n\t"
                "movq (%1, %%"REG_a"), %%mm4    \n\t"
                "movq 8(%1, %%"REG_a"), %%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, %%"REG_a"), %%mm4 \n\t" // block[i] == 0 ? -1 : 0
                "pcmpeqw 8(%0, %%"REG_a"), %%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, %%"REG_a")    \n\t"
                "movq %%mm5, 8(%0, %%"REG_a")   \n\t"

                "add $16, %%"REG_a"             \n\t"
                "js 1b                          \n\t"
297
                ::"r" (block+nCoeffs), "r"(quant_matrix+nCoeffs), "rm" (qscale), "g" (-2*nCoeffs)
298 299
                : "%"REG_a, "memory"
        );
300 301
}

302
static void dct_unquantize_mpeg2_intra_mmx(MpegEncContext *s,
Diego Biurrun's avatar
Diego Biurrun committed
303
                                     int16_t *block, int n, int qscale)
304
{
305
    x86_reg nCoeffs;
306
    const uint16_t *quant_matrix;
307
    int block0;
308

309
    av_assert2(s->block_last_index[n]>=0);
310 311 312

    if(s->alternate_scan) nCoeffs= 63; //FIXME
    else nCoeffs= s->intra_scantable.raster_end[ s->block_last_index[n] ];
313

314
    if (n < 4)
315 316 317 318
        block0 = block[0] * s->y_dc_scale;
    else
        block0 = block[0] * s->c_dc_scale;
    quant_matrix = s->intra_matrix;
319
__asm__ volatile(
320 321 322 323 324 325
                "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"
                "mov %3, %%"REG_a"              \n\t"
326
                ".p2align 4                     \n\t"
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
                "1:                             \n\t"
                "movq (%0, %%"REG_a"), %%mm0    \n\t"
                "movq 8(%0, %%"REG_a"), %%mm1   \n\t"
                "movq (%1, %%"REG_a"), %%mm4    \n\t"
                "movq 8(%1, %%"REG_a"), %%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])
                "pmullw %%mm4, %%mm0            \n\t" // abs(block[i])*q
                "pmullw %%mm5, %%mm1            \n\t" // abs(block[i])*q
                "pxor %%mm4, %%mm4              \n\t"
                "pxor %%mm5, %%mm5              \n\t" // FIXME slow
                "pcmpeqw (%0, %%"REG_a"), %%mm4 \n\t" // block[i] == 0 ? -1 : 0
                "pcmpeqw 8(%0, %%"REG_a"), %%mm5\n\t" // block[i] == 0 ? -1 : 0
                "psraw $3, %%mm0                \n\t"
                "psraw $3, %%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, %%"REG_a")    \n\t"
                "movq %%mm5, 8(%0, %%"REG_a")   \n\t"

                "add $16, %%"REG_a"             \n\t"
                "jng 1b                         \n\t"
361
                ::"r" (block+nCoeffs), "r"(quant_matrix+nCoeffs), "rm" (qscale), "g" (-2*nCoeffs)
362 363
                : "%"REG_a, "memory"
        );
364
    block[0]= block0;
Diego Biurrun's avatar
Diego Biurrun committed
365
        //Note, we do not do mismatch control for intra as errors cannot accumulate
366 367 368
}

static void dct_unquantize_mpeg2_inter_mmx(MpegEncContext *s,
Diego Biurrun's avatar
Diego Biurrun committed
369
                                     int16_t *block, int n, int qscale)
370
{
371
    x86_reg nCoeffs;
372
    const uint16_t *quant_matrix;
373

374
    av_assert2(s->block_last_index[n]>=0);
375 376 377

    if(s->alternate_scan) nCoeffs= 63; //FIXME
    else nCoeffs= s->intra_scantable.raster_end[ s->block_last_index[n] ];
378

379
        quant_matrix = s->inter_matrix;
380
__asm__ volatile(
381 382 383 384 385 386
                "pcmpeqw %%mm7, %%mm7           \n\t"
                "psrlq $48, %%mm7               \n\t"
                "movd %2, %%mm6                 \n\t"
                "packssdw %%mm6, %%mm6          \n\t"
                "packssdw %%mm6, %%mm6          \n\t"
                "mov %3, %%"REG_a"              \n\t"
387
                ".p2align 4                     \n\t"
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
                "1:                             \n\t"
                "movq (%0, %%"REG_a"), %%mm0    \n\t"
                "movq 8(%0, %%"REG_a"), %%mm1   \n\t"
                "movq (%1, %%"REG_a"), %%mm4    \n\t"
                "movq 8(%1, %%"REG_a"), %%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
                "pmullw %%mm4, %%mm0            \n\t" // abs(block[i])*2*q
                "pmullw %%mm5, %%mm1            \n\t" // abs(block[i])*2*q
                "paddw %%mm4, %%mm0             \n\t" // (abs(block[i])*2 + 1)*q
                "paddw %%mm5, %%mm1             \n\t" // (abs(block[i])*2 + 1)*q
                "pxor %%mm4, %%mm4              \n\t"
                "pxor %%mm5, %%mm5              \n\t" // FIXME slow
                "pcmpeqw (%0, %%"REG_a"), %%mm4 \n\t" // block[i] == 0 ? -1 : 0
                "pcmpeqw 8(%0, %%"REG_a"), %%mm5\n\t" // block[i] == 0 ? -1 : 0
                "psrlw $4, %%mm0                \n\t"
                "psrlw $4, %%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"
                "pxor %%mm4, %%mm7              \n\t"
                "pxor %%mm5, %%mm7              \n\t"
                "movq %%mm4, (%0, %%"REG_a")    \n\t"
                "movq %%mm5, 8(%0, %%"REG_a")   \n\t"

                "add $16, %%"REG_a"             \n\t"
                "jng 1b                         \n\t"
                "movd 124(%0, %3), %%mm0        \n\t"
                "movq %%mm7, %%mm6              \n\t"
                "psrlq $32, %%mm7               \n\t"
                "pxor %%mm6, %%mm7              \n\t"
                "movq %%mm7, %%mm6              \n\t"
                "psrlq $16, %%mm7               \n\t"
                "pxor %%mm6, %%mm7              \n\t"
                "pslld $31, %%mm7               \n\t"
                "psrlq $15, %%mm7               \n\t"
                "pxor %%mm7, %%mm0              \n\t"
                "movd %%mm0, 124(%0, %3)        \n\t"

440
                ::"r" (block+nCoeffs), "r"(quant_matrix+nCoeffs), "rm" (qscale), "r" (-2*nCoeffs)
441 442
                : "%"REG_a, "memory"
        );
Nick Kurshev's avatar
Nick Kurshev committed
443 444
}

445
#endif /* HAVE_MMX_INLINE */
446

447
av_cold void ff_mpv_common_init_x86(MpegEncContext *s)
Nick Kurshev's avatar
Nick Kurshev committed
448
{
449
#if HAVE_MMX_INLINE
450
    int cpu_flags = av_get_cpu_flags();
451

452
    if (INLINE_MMX(cpu_flags)) {
453 454 455 456
        s->dct_unquantize_h263_intra = dct_unquantize_h263_intra_mmx;
        s->dct_unquantize_h263_inter = dct_unquantize_h263_inter_mmx;
        s->dct_unquantize_mpeg1_intra = dct_unquantize_mpeg1_intra_mmx;
        s->dct_unquantize_mpeg1_inter = dct_unquantize_mpeg1_inter_mmx;
457 458
        if(!(s->flags & CODEC_FLAG_BITEXACT))
            s->dct_unquantize_mpeg2_intra = dct_unquantize_mpeg2_intra_mmx;
459
        s->dct_unquantize_mpeg2_inter = dct_unquantize_mpeg2_inter_mmx;
460
    }
461
#endif /* HAVE_MMX_INLINE */
Nick Kurshev's avatar
Nick Kurshev committed
462
}