Newer
Older
* Copyright (c) 2010 The WebM project authors. All Rights Reserved.
* Use of this source code is governed by a BSD-style license
* that can be found in the LICENSE file in the root of the source
* tree. An additional intellectual property rights grant can be found
* in the file PATENTS. All contributing project authors may
* be found in the AUTHORS file in the root of the source tree.
#include "vp9/decoder/vp9_onyxd_int.h"
#include "vp9/common/vp9_header.h"
#include "vp9/common/vp9_reconintra.h"
#include "vp9/common/vp9_reconintra4x4.h"
#include "vp9/common/vp9_reconinter.h"
#include "vp9/decoder/vp9_detokenize.h"
#include "vp9/common/vp9_invtrans.h"
#include "vp9/common/vp9_alloccommon.h"
#include "vp9/common/vp9_entropymode.h"
#include "vp9/common/vp9_quant_common.h"
#include "vp9/common/vp9_setupintrarecon.h"
#include "vp9/decoder/vp9_decodemv.h"
#include "vp9/common/vp9_extend.h"
#include "vp9/common/vp9_modecont.h"
#include "vp9/decoder/vp9_dboolhuff.h"
#define COEFCOUNT_TESTING
#ifdef DEC_DEBUG
int dec_debug = 0;
#endif
static int merge_index(int v, int n, int modulus) {
int max1 = (n - 1 - modulus / 2) / modulus + 1;
if (v < max1) v = v * modulus + modulus / 2;
else {
int w;
v -= max1;
w = v;
v += (v + modulus - modulus / 2) / modulus;
while (v % modulus == modulus / 2 ||
w != v - (v + modulus - modulus / 2) / modulus) v++;
}
return v;
static int inv_remap_prob(int v, int m) {
const int n = 256;
const int modulus = MODULUS_PARAM;
i = vp9_inv_recenter_nonneg(v + 1, m);
i = n - 1 - vp9_inv_recenter_nonneg(v + 1, n - 1 - m);
static vp9_prob read_prob_diff_update(vp9_reader *const bc, int oldp) {
int delp = vp9_decode_term_subexp(bc, SUBEXP_PARAM, 255);
return (vp9_prob)inv_remap_prob(delp, oldp);
void vp9_init_de_quantizer(VP9D_COMP *pbi) {
VP9_COMMON *const pc = &pbi->common;
pc->Y1dequant[Q][0] = (int16_t)vp9_dc_quant(Q, pc->y1dc_delta_q);
pc->Y2dequant[Q][0] = (int16_t)vp9_dc2quant(Q, pc->y2dc_delta_q);
pc->UVdequant[Q][0] = (int16_t)vp9_dc_uv_quant(Q, pc->uvdc_delta_q);
int rc = vp9_default_zig_zag1d_4x4[i];
pc->Y1dequant[Q][rc] = (int16_t)vp9_ac_yquant(Q);
pc->Y2dequant[Q][rc] = (int16_t)vp9_ac2quant(Q, pc->y2ac_delta_q);
pc->UVdequant[Q][rc] = (int16_t)vp9_ac_uv_quant(Q, pc->uvac_delta_q);
static void mb_init_dequantizer(VP9D_COMP *pbi, MACROBLOCKD *xd) {
VP9_COMMON *const pc = &pbi->common;
int segment_id = xd->mode_info_context->mbmi.segment_id;
// Set the Q baseline allowing for any segment level adjustment
if (vp9_segfeature_active(xd, segment_id, SEG_LVL_ALT_Q)) {
/* Abs Value */
if (xd->mb_segment_abs_delta == SEGMENT_ABSDATA)
QIndex = vp9_get_segdata(xd, segment_id, SEG_LVL_ALT_Q);
/* Delta Value */
else {
QIndex = pc->base_qindex +
vp9_get_segdata(xd, segment_id, SEG_LVL_ALT_Q);
QIndex = (QIndex >= 0) ? ((QIndex <= MAXQ) ? QIndex : MAXQ) : 0; /* Clamp to valid range */
/* Set up the block level dequant pointers */
for (i = 0; i < 16; i++) {
xd->block[i].dequant = pc->Y1dequant[QIndex];
}
pbi->mb.inv_xform4x4_1_x8 = vp9_short_inv_walsh4x4_1_x8;
pbi->mb.inv_xform4x4_x8 = vp9_short_inv_walsh4x4_x8;
pbi->mb.inv_walsh4x4_1 = vp9_short_inv_walsh4x4_1_lossless;
pbi->mb.inv_walsh4x4_lossless = vp9_short_inv_walsh4x4_lossless;
pbi->idct_add = vp9_dequant_idct_add_lossless_c;
pbi->dc_idct_add = vp9_dequant_dc_idct_add_lossless_c;
pbi->dc_idct_add_y_block = vp9_dequant_dc_idct_add_y_block_lossless_c;
pbi->idct_add_y_block = vp9_dequant_idct_add_y_block_lossless_c;
pbi->idct_add_uv_block = vp9_dequant_idct_add_uv_block_lossless_c;
pbi->mb.inv_xform4x4_1_x8 = vp9_short_idct4x4llm_1;
pbi->mb.inv_xform4x4_x8 = vp9_short_idct4x4llm;
pbi->mb.inv_walsh4x4_1 = vp9_short_inv_walsh4x4_1;
pbi->mb.inv_walsh4x4_lossless = vp9_short_inv_walsh4x4;
pbi->idct_add = vp9_dequant_idct_add;
pbi->dc_idct_add = vp9_dequant_dc_idct_add;
pbi->dc_idct_add_y_block = vp9_dequant_dc_idct_add_y_block;
pbi->idct_add_y_block = vp9_dequant_idct_add_y_block;
pbi->idct_add_uv_block = vp9_dequant_idct_add_uv_block;
pbi->mb.inv_xform4x4_1_x8 = vp9_short_idct4x4llm_1;
pbi->mb.inv_xform4x4_x8 = vp9_short_idct4x4llm;
pbi->mb.inv_walsh4x4_1 = vp9_short_inv_walsh4x4_1;
pbi->mb.inv_walsh4x4_lossless = vp9_short_inv_walsh4x4;
pbi->idct_add = vp9_dequant_idct_add;
pbi->dc_idct_add = vp9_dequant_dc_idct_add;
pbi->dc_idct_add_y_block = vp9_dequant_dc_idct_add_y_block;
pbi->idct_add_y_block = vp9_dequant_idct_add_y_block;
pbi->idct_add_uv_block = vp9_dequant_idct_add_uv_block;
for (i = 16; i < 24; i++) {
xd->block[i].dequant = pc->UVdequant[QIndex];
}
/* skip_recon_mb() is Modified: Instead of writing the result to predictor buffer and then copying it
* to dst buffer, we can write the result directly to dst buffer. This eliminates unnecessary copy.
*/
static void skip_recon_mb(VP9D_COMP *pbi, MACROBLOCKD *xd) {
if (xd->mode_info_context->mbmi.ref_frame == INTRA_FRAME) {
if (xd->mode_info_context->mbmi.sb_type == BLOCK_SIZE_SB64X64) {
vp9_build_intra_predictors_sb64uv_s(xd);
vp9_build_intra_predictors_sb64y_s(xd);
} else if (xd->mode_info_context->mbmi.sb_type == BLOCK_SIZE_SB32X32) {
vp9_build_intra_predictors_sbuv_s(xd);
vp9_build_intra_predictors_sby_s(xd);
vp9_build_intra_predictors_mbuv_s(xd);
vp9_build_intra_predictors_mby_s(xd);
if (xd->mode_info_context->mbmi.sb_type == BLOCK_SIZE_SB64X64) {
vp9_build_inter64x64_predictors_sb(xd,
xd->dst.y_buffer,
xd->dst.u_buffer,
xd->dst.v_buffer,
xd->dst.y_stride,
xd->dst.uv_stride);
} else if (xd->mode_info_context->mbmi.sb_type == BLOCK_SIZE_SB32X32) {
vp9_build_inter32x32_predictors_sb(xd,
xd->dst.y_buffer,
xd->dst.u_buffer,
xd->dst.v_buffer,
xd->dst.y_stride,
xd->dst.uv_stride);
xd->dst.y_buffer,
xd->dst.u_buffer,
xd->dst.v_buffer,
xd->dst.y_stride,
xd->dst.uv_stride);
if (xd->mode_info_context->mbmi.second_ref_frame > 0) {
vp9_build_2nd_inter16x16_predictors_mb(xd,
xd->dst.y_buffer,
xd->dst.u_buffer,
xd->dst.v_buffer,
xd->dst.y_stride,
xd->dst.uv_stride);
}
#if CONFIG_COMP_INTERINTRA_PRED
else if (xd->mode_info_context->mbmi.second_ref_frame == INTRA_FRAME) {
vp9_build_interintra_16x16_predictors_mb(xd,
xd->dst.y_buffer,
xd->dst.u_buffer,
xd->dst.v_buffer,
xd->dst.y_stride,
xd->dst.uv_stride);
}
static void decode_16x16(VP9D_COMP *pbi, MACROBLOCKD *xd,
BOOL_DECODER* const bc) {
BLOCKD *bd = &xd->block[0];
TX_TYPE tx_type = get_tx_type_16x16(xd, bd);
assert(get_2nd_order_usage(xd) == 0);
#ifdef DEC_DEBUG
if (dec_debug) {
int i;
printf("\n");
printf("qcoeff 16x16\n");
for (i = 0; i < 400; i++) {
printf("%3d ", xd->qcoeff[i]);
if (i % 16 == 15) printf("\n");
}
printf("\n");
printf("predictor\n");
for (i = 0; i < 400; i++) {
printf("%3d ", xd->predictor[i]);
if (i % 16 == 15) printf("\n");
}
}
#endif
if (tx_type != DCT_DCT) {
vp9_ht_dequant_idct_add_16x16_c(tx_type, xd->qcoeff,
xd->block[0].dequant, xd->predictor,
xd->dst.y_buffer, 16, xd->dst.y_stride,
xd->eobs[0]);
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
} else {
vp9_dequant_idct_add_16x16(xd->qcoeff, xd->block[0].dequant,
xd->predictor, xd->dst.y_buffer,
16, xd->dst.y_stride, xd->eobs[0]);
}
vp9_dequant_idct_add_uv_block_8x8(
xd->qcoeff + 16 * 16, xd->block[16].dequant,
xd->predictor + 16 * 16, xd->dst.u_buffer, xd->dst.v_buffer,
xd->dst.uv_stride, xd->eobs + 16, xd);
}
static void decode_8x8(VP9D_COMP *pbi, MACROBLOCKD *xd,
BOOL_DECODER* const bc) {
// First do Y
// if the first one is DCT_DCT assume all the rest are as well
TX_TYPE tx_type = get_tx_type_8x8(xd, &xd->block[0]);
#ifdef DEC_DEBUG
if (dec_debug) {
int i;
printf("\n");
printf("qcoeff 8x8\n");
for (i = 0; i < 400; i++) {
printf("%3d ", xd->qcoeff[i]);
if (i % 16 == 15) printf("\n");
}
}
#endif
if (tx_type != DCT_DCT || xd->mode_info_context->mbmi.mode == I8X8_PRED) {
int i;
assert(get_2nd_order_usage(xd) == 0);
for (i = 0; i < 4; i++) {
int ib = vp9_i8x8_block[i];
int idx = (ib & 0x02) ? (ib + 2) : ib;
int16_t *q = xd->block[idx].qcoeff;
int16_t *dq = xd->block[0].dequant;
uint8_t *pre = xd->block[ib].predictor;
uint8_t *dst = *(xd->block[ib].base_dst) + xd->block[ib].dst;
int stride = xd->dst.y_stride;
BLOCKD *b = &xd->block[ib];
if (xd->mode_info_context->mbmi.mode == I8X8_PRED) {
int i8x8mode = b->bmi.as_mode.first;
vp9_intra8x8_predict(b, i8x8mode, b->predictor);
}
tx_type = get_tx_type_8x8(xd, &xd->block[ib]);
if (tx_type != DCT_DCT) {
vp9_ht_dequant_idct_add_8x8_c(tx_type, q, dq, pre, dst, 16, stride,
xd->eobs[idx]);
} else {
vp9_dequant_idct_add_8x8_c(q, dq, pre, dst, 16, stride,
0, xd->eobs[idx]);
}
}
} else if (xd->mode_info_context->mbmi.mode == SPLITMV) {
assert(get_2nd_order_usage(xd) == 0);
vp9_dequant_idct_add_y_block_8x8(xd->qcoeff,
xd->block[0].dequant,
xd->predictor,
xd->dst.y_buffer,
xd->dst.y_stride,
xd->eobs, xd);
} else {
BLOCKD *b = &xd->block[24];
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
vp9_dequantize_b_2x2(b);
vp9_short_ihaar2x2(&b->dqcoeff[0], b->diff, 8);
((int *)b->qcoeff)[0] = 0; // 2nd order block are set to 0 after idct
((int *)b->qcoeff)[1] = 0;
((int *)b->qcoeff)[2] = 0;
((int *)b->qcoeff)[3] = 0;
((int *)b->qcoeff)[4] = 0;
((int *)b->qcoeff)[5] = 0;
((int *)b->qcoeff)[6] = 0;
((int *)b->qcoeff)[7] = 0;
vp9_dequant_dc_idct_add_y_block_8x8(xd->qcoeff,
xd->block[0].dequant,
xd->predictor,
xd->dst.y_buffer,
xd->dst.y_stride,
xd->eobs,
xd->block[24].diff,
xd);
}
// Now do UV
if (xd->mode_info_context->mbmi.mode == I8X8_PRED) {
int i;
for (i = 0; i < 4; i++) {
int ib = vp9_i8x8_block[i];
BLOCKD *b = &xd->block[ib];
int i8x8mode = b->bmi.as_mode.first;
b = &xd->block[16 + i];
vp9_intra_uv4x4_predict(&xd->block[16 + i], i8x8mode, b->predictor);
pbi->idct_add(b->qcoeff, b->dequant, b->predictor,
*(b->base_dst) + b->dst, 8, b->dst_stride);
b = &xd->block[20 + i];
vp9_intra_uv4x4_predict(&xd->block[20 + i], i8x8mode, b->predictor);
pbi->idct_add(b->qcoeff, b->dequant, b->predictor,
*(b->base_dst) + b->dst, 8, b->dst_stride);
}
} else if (xd->mode_info_context->mbmi.mode == SPLITMV) {
pbi->idct_add_uv_block(xd->qcoeff + 16 * 16, xd->block[16].dequant,
xd->predictor + 16 * 16, xd->dst.u_buffer, xd->dst.v_buffer,
xd->dst.uv_stride, xd->eobs + 16);
} else {
vp9_dequant_idct_add_uv_block_8x8
(xd->qcoeff + 16 * 16, xd->block[16].dequant,
xd->predictor + 16 * 16, xd->dst.u_buffer, xd->dst.v_buffer,
xd->dst.uv_stride, xd->eobs + 16, xd);
}
#ifdef DEC_DEBUG
if (dec_debug) {
int i;
printf("\n");
printf("predictor\n");
for (i = 0; i < 384; i++) {
printf("%3d ", xd->predictor[i]);
if (i % 16 == 15) printf("\n");
}
}
#endif
}
static void decode_4x4(VP9D_COMP *pbi, MACROBLOCKD *xd,
BOOL_DECODER* const bc) {
TX_TYPE tx_type;
int i, eobtotal = 0;
MB_PREDICTION_MODE mode = xd->mode_info_context->mbmi.mode;
if (mode == I8X8_PRED) {
assert(get_2nd_order_usage(xd) == 0);
for (i = 0; i < 4; i++) {
int ib = vp9_i8x8_block[i];
const int iblock[4] = {0, 1, 4, 5};
int j;
int i8x8mode;
BLOCKD *b;
b = &xd->block[ib];
i8x8mode = b->bmi.as_mode.first;
vp9_intra8x8_predict(b, i8x8mode, b->predictor);
for (j = 0; j < 4; j++) {
b = &xd->block[ib + iblock[j]];
tx_type = get_tx_type_4x4(xd, b);
if (tx_type != DCT_DCT) {
vp9_ht_dequant_idct_add_c(tx_type, b->qcoeff,
b->dequant, b->predictor,
*(b->base_dst) + b->dst, 16,
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
} else {
vp9_dequant_idct_add(b->qcoeff, b->dequant, b->predictor,
*(b->base_dst) + b->dst, 16, b->dst_stride);
}
}
b = &xd->block[16 + i];
vp9_intra_uv4x4_predict(b, i8x8mode, b->predictor);
pbi->idct_add(b->qcoeff, b->dequant, b->predictor,
*(b->base_dst) + b->dst, 8, b->dst_stride);
b = &xd->block[20 + i];
vp9_intra_uv4x4_predict(b, i8x8mode, b->predictor);
pbi->idct_add(b->qcoeff, b->dequant, b->predictor,
*(b->base_dst) + b->dst, 8, b->dst_stride);
}
} else if (mode == B_PRED) {
assert(get_2nd_order_usage(xd) == 0);
for (i = 0; i < 16; i++) {
int b_mode;
BLOCKD *b = &xd->block[i];
b_mode = xd->mode_info_context->bmi[i].as_mode.first;
#if CONFIG_NEWBINTRAMODES
xd->mode_info_context->bmi[i].as_mode.context = b->bmi.as_mode.context =
vp9_find_bpred_context(b);
#endif
if (!xd->mode_info_context->mbmi.mb_skip_coeff)
eobtotal += vp9_decode_coefs_4x4(pbi, xd, bc, PLANE_TYPE_Y_WITH_DC, i);
vp9_intra4x4_predict(b, b_mode, b->predictor);
tx_type = get_tx_type_4x4(xd, b);
if (tx_type != DCT_DCT) {
vp9_ht_dequant_idct_add_c(tx_type, b->qcoeff,
b->dequant, b->predictor,
*(b->base_dst) + b->dst, 16, b->dst_stride,
b->eob);
} else {
vp9_dequant_idct_add(b->qcoeff, b->dequant, b->predictor,
*(b->base_dst) + b->dst, 16, b->dst_stride);
}
}
if (!xd->mode_info_context->mbmi.mb_skip_coeff) {
vp9_decode_mb_tokens_4x4_uv(pbi, xd, bc);
}
xd->above_context->y2 = 0;
xd->left_context->y2 = 0;
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
vp9_build_intra_predictors_mbuv(xd);
pbi->idct_add_uv_block(xd->qcoeff + 16 * 16,
xd->block[16].dequant,
xd->predictor + 16 * 16,
xd->dst.u_buffer,
xd->dst.v_buffer,
xd->dst.uv_stride,
xd->eobs + 16);
} else if (mode == SPLITMV) {
assert(get_2nd_order_usage(xd) == 0);
pbi->idct_add_y_block(xd->qcoeff,
xd->block[0].dequant,
xd->predictor,
xd->dst.y_buffer,
xd->dst.y_stride,
xd->eobs);
pbi->idct_add_uv_block(xd->qcoeff + 16 * 16,
xd->block[16].dequant,
xd->predictor + 16 * 16,
xd->dst.u_buffer,
xd->dst.v_buffer,
xd->dst.uv_stride,
xd->eobs + 16);
} else {
#ifdef DEC_DEBUG
if (dec_debug) {
int i;
printf("\n");
printf("qcoeff 4x4\n");
for (i = 0; i < 400; i++) {
printf("%3d ", xd->qcoeff[i]);
if (i % 16 == 15) printf("\n");
}
printf("\n");
printf("predictor\n");
for (i = 0; i < 400; i++) {
printf("%3d ", xd->predictor[i]);
if (i % 16 == 15) printf("\n");
}
}
#endif
tx_type = get_tx_type_4x4(xd, &xd->block[0]);
if (tx_type != DCT_DCT) {
assert(get_2nd_order_usage(xd) == 0);
for (i = 0; i < 16; i++) {
BLOCKD *b = &xd->block[i];
tx_type = get_tx_type_4x4(xd, b);
if (tx_type != DCT_DCT) {
vp9_ht_dequant_idct_add_c(tx_type, b->qcoeff,
b->dequant, b->predictor,
*(b->base_dst) + b->dst, 16,
} else {
vp9_dequant_idct_add(b->qcoeff, b->dequant, b->predictor,
*(b->base_dst) + b->dst, 16, b->dst_stride);
}
}
} else {
BLOCKD *b = &xd->block[24];
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
538
539
vp9_dequantize_b(b);
if (xd->eobs[24] > 1) {
vp9_short_inv_walsh4x4(&b->dqcoeff[0], b->diff);
((int *)b->qcoeff)[0] = 0;
((int *)b->qcoeff)[1] = 0;
((int *)b->qcoeff)[2] = 0;
((int *)b->qcoeff)[3] = 0;
((int *)b->qcoeff)[4] = 0;
((int *)b->qcoeff)[5] = 0;
((int *)b->qcoeff)[6] = 0;
((int *)b->qcoeff)[7] = 0;
} else {
xd->inv_walsh4x4_1(&b->dqcoeff[0], b->diff);
((int *)b->qcoeff)[0] = 0;
}
vp9_dequantize_b(b);
pbi->dc_idct_add_y_block(xd->qcoeff,
xd->block[0].dequant,
xd->predictor,
xd->dst.y_buffer,
xd->dst.y_stride,
xd->eobs,
xd->block[24].diff);
}
pbi->idct_add_uv_block(xd->qcoeff + 16 * 16,
xd->block[16].dequant,
xd->predictor + 16 * 16,
xd->dst.u_buffer,
xd->dst.v_buffer,
xd->dst.uv_stride,
xd->eobs + 16);
}
}
static void decode_16x16_sb(VP9D_COMP *pbi, MACROBLOCKD *xd,
BOOL_DECODER* const bc, int n,
int maska, int shiftb) {
int x_idx = n & maska, y_idx = n >> shiftb;
TX_TYPE tx_type = get_tx_type_16x16(xd, &xd->block[0]);
if (tx_type != DCT_DCT) {
vp9_ht_dequant_idct_add_16x16_c(
tx_type, xd->qcoeff, xd->block[0].dequant,
xd->dst.y_buffer + y_idx * 16 * xd->dst.y_stride + x_idx * 16,
xd->dst.y_buffer + y_idx * 16 * xd->dst.y_stride + x_idx * 16,
xd->dst.y_stride, xd->dst.y_stride, xd->block[0].eob);
} else {
vp9_dequant_idct_add_16x16(
xd->qcoeff, xd->block[0].dequant,
xd->dst.y_buffer + y_idx * 16 * xd->dst.y_stride + x_idx * 16,
xd->dst.y_buffer + y_idx * 16 * xd->dst.y_stride + x_idx * 16,
xd->dst.y_stride, xd->dst.y_stride, xd->eobs[0]);
}
vp9_dequant_idct_add_uv_block_8x8_inplace_c(
xd->qcoeff + 16 * 16,
xd->block[16].dequant,
xd->dst.u_buffer + y_idx * 8 * xd->dst.uv_stride + x_idx * 8,
xd->dst.v_buffer + y_idx * 8 * xd->dst.uv_stride + x_idx * 8,
xd->dst.uv_stride, xd->eobs + 16, xd);
};
static void decode_8x8_sb(VP9D_COMP *pbi, MACROBLOCKD *xd,
BOOL_DECODER* const bc, int n,
int maska, int shiftb) {
int x_idx = n & maska, y_idx = n >> shiftb;
BLOCKD *b = &xd->block[24];
TX_TYPE tx_type = get_tx_type_8x8(xd, &xd->block[0]);
if (tx_type != DCT_DCT) {
int i;
for (i = 0; i < 4; i++) {
int ib = vp9_i8x8_block[i];
int idx = (ib & 0x02) ? (ib + 2) : ib;
int16_t *q = xd->block[idx].qcoeff;
int16_t *dq = xd->block[0].dequant;
int stride = xd->dst.y_stride;
BLOCKD *b = &xd->block[ib];
tx_type = get_tx_type_8x8(xd, &xd->block[ib]);
if (tx_type != DCT_DCT) {
vp9_ht_dequant_idct_add_8x8_c(
tx_type, q, dq,
xd->dst.y_buffer + (y_idx * 16 + (i / 2) * 8) * xd->dst.y_stride
+ x_idx * 16 + (i & 1) * 8,
xd->dst.y_buffer + (y_idx * 16 + (i / 2) * 8) * xd->dst.y_stride
+ x_idx * 16 + (i & 1) * 8,
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
} else {
vp9_dequant_idct_add_8x8_c(
q, dq,
xd->dst.y_buffer + (y_idx * 16 + (i / 2) * 8) * xd->dst.y_stride
+ x_idx * 16 + (i & 1) * 8,
xd->dst.y_buffer + (y_idx * 16 + (i / 2) * 8) * xd->dst.y_stride
+ x_idx * 16 + (i & 1) * 8,
stride, stride, 0, b->eob);
}
vp9_dequant_idct_add_uv_block_8x8_inplace_c(
xd->qcoeff + 16 * 16, xd->block[16].dequant,
xd->dst.u_buffer + y_idx * 8 * xd->dst.uv_stride + x_idx * 8,
xd->dst.v_buffer + y_idx * 8 * xd->dst.uv_stride + x_idx * 8,
xd->dst.uv_stride, xd->eobs + 16, xd);
}
} else {
vp9_dequantize_b_2x2(b);
vp9_short_ihaar2x2(&b->dqcoeff[0], b->diff, 8);
((int *)b->qcoeff)[0] = 0; // 2nd order block are set to 0 after idct
((int *)b->qcoeff)[1] = 0;
((int *)b->qcoeff)[2] = 0;
((int *)b->qcoeff)[3] = 0;
((int *)b->qcoeff)[4] = 0;
((int *)b->qcoeff)[5] = 0;
((int *)b->qcoeff)[6] = 0;
((int *)b->qcoeff)[7] = 0;
vp9_dequant_dc_idct_add_y_block_8x8_inplace_c(
xd->qcoeff, xd->block[0].dequant,
xd->dst.y_buffer + y_idx * 16 * xd->dst.y_stride + x_idx * 16,
xd->dst.y_stride, xd->eobs, xd->block[24].diff, xd);
vp9_dequant_idct_add_uv_block_8x8_inplace_c(
xd->qcoeff + 16 * 16, xd->block[16].dequant,
xd->dst.u_buffer + y_idx * 8 * xd->dst.uv_stride + x_idx * 8,
xd->dst.v_buffer + y_idx * 8 * xd->dst.uv_stride + x_idx * 8,
xd->dst.uv_stride, xd->eobs + 16, xd);
}
};
static void decode_4x4_sb(VP9D_COMP *pbi, MACROBLOCKD *xd,
BOOL_DECODER* const bc, int n,
int maska, int shiftb) {
int x_idx = n & maska, y_idx = n >> shiftb;
BLOCKD *b = &xd->block[24];
TX_TYPE tx_type = get_tx_type_4x4(xd, &xd->block[0]);
if (tx_type != DCT_DCT) {
int i;
for (i = 0; i < 16; i++) {
BLOCKD *b = &xd->block[i];
tx_type = get_tx_type_4x4(xd, b);
if (tx_type != DCT_DCT) {
vp9_ht_dequant_idct_add_c(
tx_type, b->qcoeff, b->dequant,
xd->dst.y_buffer + (y_idx * 16 + (i / 4) * 4) * xd->dst.y_stride
+ x_idx * 16 + (i & 3) * 4,
xd->dst.y_buffer + (y_idx * 16 + (i / 4) * 4) * xd->dst.y_stride
+ x_idx * 16 + (i & 3) * 4,
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
} else {
vp9_dequant_idct_add_c(
b->qcoeff, b->dequant,
xd->dst.y_buffer + (y_idx * 16 + (i / 4) * 4) * xd->dst.y_stride
+ x_idx * 16 + (i & 3) * 4,
xd->dst.y_buffer + (y_idx * 16 + (i / 4) * 4) * xd->dst.y_stride
+ x_idx * 16 + (i & 3) * 4,
xd->dst.y_stride, xd->dst.y_stride);
}
}
} else {
vp9_dequantize_b(b);
if (xd->eobs[24] > 1) {
vp9_short_inv_walsh4x4(&b->dqcoeff[0], b->diff);
((int *)b->qcoeff)[0] = 0;
((int *)b->qcoeff)[1] = 0;
((int *)b->qcoeff)[2] = 0;
((int *)b->qcoeff)[3] = 0;
((int *)b->qcoeff)[4] = 0;
((int *)b->qcoeff)[5] = 0;
((int *)b->qcoeff)[6] = 0;
((int *)b->qcoeff)[7] = 0;
} else {
xd->inv_walsh4x4_1(&b->dqcoeff[0], b->diff);
((int *)b->qcoeff)[0] = 0;
}
vp9_dequant_dc_idct_add_y_block_4x4_inplace_c(
xd->qcoeff, xd->block[0].dequant,
xd->dst.y_buffer + y_idx * 16 * xd->dst.y_stride + x_idx * 16,
xd->dst.y_stride, xd->eobs, xd->block[24].diff, xd);
}
vp9_dequant_idct_add_uv_block_4x4_inplace_c(
xd->qcoeff + 16 * 16, xd->block[16].dequant,
xd->dst.u_buffer + y_idx * 8 * xd->dst.uv_stride + x_idx * 8,
xd->dst.v_buffer + y_idx * 8 * xd->dst.uv_stride + x_idx * 8,
xd->dst.uv_stride, xd->eobs + 16, xd);
};
static void decode_superblock64(VP9D_COMP *pbi, MACROBLOCKD *xd,
int mb_row, unsigned int mb_col,
BOOL_DECODER* const bc) {
int i, n, eobtotal;
TX_SIZE tx_size = xd->mode_info_context->mbmi.txfm_size;
VP9_COMMON *const pc = &pbi->common;
MODE_INFO *orig_mi = xd->mode_info_context;
const int mis = pc->mode_info_stride;
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
assert(xd->mode_info_context->mbmi.sb_type == BLOCK_SIZE_SB64X64);
if (pbi->common.frame_type != KEY_FRAME)
vp9_setup_interp_filters(xd, xd->mode_info_context->mbmi.interp_filter, pc);
// re-initialize macroblock dequantizer before detokenization
if (xd->segmentation_enabled)
mb_init_dequantizer(pbi, xd);
if (xd->mode_info_context->mbmi.mb_skip_coeff) {
int n;
vp9_reset_mb_tokens_context(xd);
for (n = 1; n <= 3; n++) {
if (mb_col < pc->mb_cols - n)
xd->above_context += n;
if (mb_row < pc->mb_rows - n)
xd->left_context += n;
vp9_reset_mb_tokens_context(xd);
if (mb_col < pc->mb_cols - n)
xd->above_context -= n;
if (mb_row < pc->mb_rows - n)
xd->left_context -= n;
}
/* Special case: Force the loopfilter to skip when eobtotal and
* mb_skip_coeff are zero.
*/
skip_recon_mb(pbi, xd);
return;
}
/* do prediction */
if (xd->mode_info_context->mbmi.ref_frame == INTRA_FRAME) {
vp9_build_intra_predictors_sb64y_s(xd);
vp9_build_intra_predictors_sb64uv_s(xd);
} else {
vp9_build_inter64x64_predictors_sb(xd, xd->dst.y_buffer,
xd->dst.u_buffer, xd->dst.v_buffer,
xd->dst.y_stride, xd->dst.uv_stride);
}
/* dequantization and idct */
if (xd->mode_info_context->mbmi.txfm_size == TX_32X32) {
for (n = 0; n < 4; n++) {
const int x_idx = n & 1, y_idx = n >> 1;
if (mb_col + x_idx * 2 >= pc->mb_cols ||
mb_row + y_idx * 2 >= pc->mb_rows)
continue;
xd->left_context = pc->left_context + (y_idx << 1);
xd->above_context = pc->above_context + mb_col + (x_idx << 1);
xd->mode_info_context = orig_mi + x_idx * 2 + y_idx * 2 * mis;
eobtotal = vp9_decode_sb_tokens(pbi, xd, bc);
if (eobtotal == 0) { // skip loopfilter
xd->mode_info_context->mbmi.mb_skip_coeff = 1;
if (mb_col + 1 < pc->mb_cols)
xd->mode_info_context[1].mbmi.mb_skip_coeff = 1;
if (mb_row + 1 < pc->mb_rows) {
xd->mode_info_context[mis].mbmi.mb_skip_coeff = 1;
if (mb_col + 1 < pc->mb_cols)
xd->mode_info_context[mis + 1].mbmi.mb_skip_coeff = 1;
}
} else {
vp9_dequant_idct_add_32x32(xd->sb_coeff_data.qcoeff, xd->block[0].dequant,
xd->dst.y_buffer + x_idx * 32 +
xd->dst.y_stride * y_idx * 32,
xd->dst.y_buffer + x_idx * 32 +
xd->dst.y_stride * y_idx * 32,
xd->dst.y_stride, xd->dst.y_stride,
xd->eobs[0]);
vp9_dequant_idct_add_uv_block_16x16_c(xd->sb_coeff_data.qcoeff + 1024,
xd->block[16].dequant,
xd->dst.u_buffer + x_idx * 16 +
xd->dst.uv_stride * y_idx * 16,
xd->dst.v_buffer + x_idx * 16 +
xd->dst.uv_stride * y_idx * 16,
xd->dst.uv_stride, xd->eobs + 16);
}
}
} else {
for (n = 0; n < 16; n++) {
int x_idx = n & 3, y_idx = n >> 2;
if (mb_col + x_idx >= pc->mb_cols || mb_row + y_idx >= pc->mb_rows)
continue;
xd->above_context = pc->above_context + mb_col + x_idx;
xd->left_context = pc->left_context + y_idx;
xd->mode_info_context = orig_mi + x_idx + y_idx * mis;
for (i = 0; i < 25; i++) {
xd->block[i].eob = 0;
xd->eobs[i] = 0;
}
eobtotal = vp9_decode_mb_tokens(pbi, xd, bc);
if (eobtotal == 0) { // skip loopfilter
xd->mode_info_context->mbmi.mb_skip_coeff = 1;
continue;
}
if (tx_size == TX_16X16) {
decode_16x16_sb(pbi, xd, bc, n, 3, 2);
} else if (tx_size == TX_8X8) {
decode_8x8_sb(pbi, xd, bc, n, 3, 2);
} else {
decode_4x4_sb(pbi, xd, bc, n, 3, 2);
}
}
}
xd->above_context = pc->above_context + mb_col;
xd->left_context = pc->left_context;
xd->mode_info_context = orig_mi;
}
static void decode_superblock32(VP9D_COMP *pbi, MACROBLOCKD *xd,
int mb_row, unsigned int mb_col,
BOOL_DECODER* const bc) {
int i, n, eobtotal;
TX_SIZE tx_size = xd->mode_info_context->mbmi.txfm_size;
VP9_COMMON *const pc = &pbi->common;
MODE_INFO *orig_mi = xd->mode_info_context;
const int mis = pc->mode_info_stride;
assert(xd->mode_info_context->mbmi.sb_type == BLOCK_SIZE_SB32X32);
if (pbi->common.frame_type != KEY_FRAME)
vp9_setup_interp_filters(xd, xd->mode_info_context->mbmi.interp_filter, pc);
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
// re-initialize macroblock dequantizer before detokenization
if (xd->segmentation_enabled)
mb_init_dequantizer(pbi, xd);
if (xd->mode_info_context->mbmi.mb_skip_coeff) {
vp9_reset_mb_tokens_context(xd);
if (mb_col < pc->mb_cols - 1)
xd->above_context++;
if (mb_row < pc->mb_rows - 1)
xd->left_context++;
vp9_reset_mb_tokens_context(xd);
if (mb_col < pc->mb_cols - 1)
xd->above_context--;
if (mb_row < pc->mb_rows - 1)
xd->left_context--;
/* Special case: Force the loopfilter to skip when eobtotal and
* mb_skip_coeff are zero.
*/
skip_recon_mb(pbi, xd);
return;
}
/* do prediction */
if (xd->mode_info_context->mbmi.ref_frame == INTRA_FRAME) {
vp9_build_intra_predictors_sby_s(xd);
vp9_build_intra_predictors_sbuv_s(xd);
} else {
vp9_build_inter32x32_predictors_sb(xd, xd->dst.y_buffer,
xd->dst.u_buffer, xd->dst.v_buffer,
xd->dst.y_stride, xd->dst.uv_stride);
}
/* dequantization and idct */
if (xd->mode_info_context->mbmi.txfm_size == TX_32X32) {
eobtotal = vp9_decode_sb_tokens(pbi, xd, bc);
if (eobtotal == 0) { // skip loopfilter
xd->mode_info_context->mbmi.mb_skip_coeff = 1;
if (mb_col + 1 < pc->mb_cols)
xd->mode_info_context[1].mbmi.mb_skip_coeff = 1;
if (mb_row + 1 < pc->mb_rows) {
xd->mode_info_context[mis].mbmi.mb_skip_coeff = 1;
if (mb_col + 1 < pc->mb_cols)
xd->mode_info_context[mis + 1].mbmi.mb_skip_coeff = 1;
}
} else {
vp9_dequant_idct_add_32x32(xd->sb_coeff_data.qcoeff, xd->block[0].dequant,
xd->dst.y_buffer, xd->dst.y_buffer,
xd->dst.y_stride, xd->dst.y_stride,
xd->eobs[0]);
vp9_dequant_idct_add_uv_block_16x16_c(xd->sb_coeff_data.qcoeff + 1024,
xd->block[16].dequant,
xd->dst.u_buffer, xd->dst.v_buffer,
xd->dst.uv_stride, xd->eobs + 16);
}
for (n = 0; n < 4; n++) {
int x_idx = n & 1, y_idx = n >> 1;
if (mb_col + x_idx >= pc->mb_cols || mb_row + y_idx >= pc->mb_rows)
continue;
xd->above_context = pc->above_context + mb_col + x_idx;
xd->left_context = pc->left_context + y_idx + (mb_row & 2);
xd->mode_info_context = orig_mi + x_idx + y_idx * mis;
for (i = 0; i < 25; i++) {
xd->block[i].eob = 0;
xd->eobs[i] = 0;
}
eobtotal = vp9_decode_mb_tokens(pbi, xd, bc);
if (eobtotal == 0) { // skip loopfilter
xd->mode_info_context->mbmi.mb_skip_coeff = 1;
continue;
}
if (tx_size == TX_16X16) {
decode_16x16_sb(pbi, xd, bc, n, 1, 1);
} else if (tx_size == TX_8X8) {
decode_8x8_sb(pbi, xd, bc, n, 1, 1);
} else {
decode_4x4_sb(pbi, xd, bc, n, 1, 1);
}
xd->above_context = pc->above_context + mb_col;
xd->left_context = pc->left_context + (mb_row & 2);
xd->mode_info_context = orig_mi;
static void decode_macroblock(VP9D_COMP *pbi, MACROBLOCKD *xd,
assert(!xd->mode_info_context->mbmi.sb_type);
// re-initialize macroblock dequantizer before detokenization
if (xd->segmentation_enabled)
mb_init_dequantizer(pbi, xd);
tx_size = xd->mode_info_context->mbmi.txfm_size;
mode = xd->mode_info_context->mbmi.mode;
vp9_reset_mb_tokens_context(xd);
for (i = 0; i < 25; i++) {
xd->block[i].eob = 0;
xd->eobs[i] = 0;
if (mode != B_PRED) {
eobtotal = vp9_decode_mb_tokens(pbi, xd, bc);
//mode = xd->mode_info_context->mbmi.mode;
if (pbi->common.frame_type != KEY_FRAME)
vp9_setup_interp_filters(xd, xd->mode_info_context->mbmi.interp_filter,
if (eobtotal == 0 && mode != B_PRED && mode != SPLITMV
&& mode != I8X8_PRED
/* Special case: Force the loopfilter to skip when eobtotal and
* mb_skip_coeff are zero.
* */
xd->mode_info_context->mbmi.mb_skip_coeff = 1;
skip_recon_mb(pbi, xd);
return;
#ifdef DEC_DEBUG
if (dec_debug)
printf("Decoding mb: %d %d\n", xd->mode_info_context->mbmi.mode, tx_size);
#endif
// moved to be performed before detokenization
// if (xd->segmentation_enabled)
// mb_init_dequantizer(pbi, xd);
/* do prediction */
if (xd->mode_info_context->mbmi.ref_frame == INTRA_FRAME) {
if (mode != I8X8_PRED) {
vp9_build_intra_predictors_mbuv(xd);
vp9_build_intra_predictors_mby(xd);
#ifdef DEC_DEBUG
if (dec_debug)
printf("Decoding mb: %d %d interp %d\n",
xd->mode_info_context->mbmi.mode, tx_size,
xd->mode_info_context->mbmi.interp_filter);
#endif
vp9_build_inter_predictors_mb(xd);
if (tx_size == TX_16X16) {
decode_16x16(pbi, xd, bc);
} else if (tx_size == TX_8X8) {
decode_8x8(pbi, xd, bc);
} else {
decode_4x4(pbi, xd, bc);
}
#ifdef DEC_DEBUG
if (dec_debug) {
int i, j;
printf("\n");
printf("final y\n");
for (j = 0; j < 16; j++)