vpx/vp9/decoder/vp9_decodemv.c

1236 lines
42 KiB
C
Raw Normal View History

2010-05-18 17:58:33 +02:00
/*
Copyright (c) 2010 The WebM project authors. All Rights Reserved.
2010-05-18 17:58:33 +02:00
*
* 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.
2010-05-18 17:58:33 +02:00
*/
#include "vp9/decoder/vp9_treereader.h"
#include "vp9/common/vp9_entropymv.h"
#include "vp9/common/vp9_entropymode.h"
#include "vp9/decoder/vp9_onyxd_int.h"
#include "vp9/common/vp9_findnearmv.h"
#include "vp9/common/vp9_common.h"
#include "vp9/common/vp9_seg_common.h"
#include "vp9/common/vp9_pred_common.h"
#include "vp9/common/vp9_entropy.h"
#include "vp9/decoder/vp9_decodemv.h"
#include "vp9/common/vp9_mvref_common.h"
2010-05-18 17:58:33 +02:00
#if CONFIG_DEBUG
#include <assert.h>
#endif
Supporting high precision 1/8-pel motion vectors This is the initial patch for supporting 1/8th pel motion. Currently if we configure with enable-high-precision-mv, all motion vectors would default to 1/8 pel. Encode and decode syncs fine with the current code. In the next phase the code will be refactored so that we can choose the 1/8 pel mode adaptively at a frame/segment/mb level. Derf results: http://www.corp.google.com/~debargha/vp8_results/enhinterp_hpmv.html (about 0.83% better than 8-tap interpoaltion) Patch 3: Rebased. Also adding 1/16th pel interpolation for U and V Patch 4: HD results. http://www.corp.google.com/~debargha/vp8_results/enhinterp_hd_hpmv.html Seems impressive (unless I am doing something wrong). Patch 5: Added mmx/sse for bilateral filtering, as well as enforced use of c-versions of subpel filters with 8-taps and 1/16th pel; Also redesigned the 8-tap filters to reduce the cut-off in order to introduce a denoising effect. There is a new configure option sixteenth-subpel-uv which will use 1/16 th pel interpolation for uv, if the motion vectors have 1/8 pel accuracy. With the fixes the results are promising on the derf set. The enhanced interpolation option with 8-taps alone gives 3% improvement over thei derf set: http://www.corp.google.com/~debargha/vp8_results/enhinterpn.html Results on high precision mv and on the hd set are to follow. Patch 6: Adding a missing condition for CONFIG_SIXTEENTH_SUBPEL_UV in vp8/common/x86/x86_systemdependent.c Patch 7: Cleaning up various debug messages. Patch 8: Merge conflict Change-Id: I5b1d844457aefd7414a9e4e0e06c6ed38fd8cc04
2012-02-16 18:29:54 +01:00
// #define DEBUG_DEC_MV
Supporting high precision 1/8-pel motion vectors This is the initial patch for supporting 1/8th pel motion. Currently if we configure with enable-high-precision-mv, all motion vectors would default to 1/8 pel. Encode and decode syncs fine with the current code. In the next phase the code will be refactored so that we can choose the 1/8 pel mode adaptively at a frame/segment/mb level. Derf results: http://www.corp.google.com/~debargha/vp8_results/enhinterp_hpmv.html (about 0.83% better than 8-tap interpoaltion) Patch 3: Rebased. Also adding 1/16th pel interpolation for U and V Patch 4: HD results. http://www.corp.google.com/~debargha/vp8_results/enhinterp_hd_hpmv.html Seems impressive (unless I am doing something wrong). Patch 5: Added mmx/sse for bilateral filtering, as well as enforced use of c-versions of subpel filters with 8-taps and 1/16th pel; Also redesigned the 8-tap filters to reduce the cut-off in order to introduce a denoising effect. There is a new configure option sixteenth-subpel-uv which will use 1/16 th pel interpolation for uv, if the motion vectors have 1/8 pel accuracy. With the fixes the results are promising on the derf set. The enhanced interpolation option with 8-taps alone gives 3% improvement over thei derf set: http://www.corp.google.com/~debargha/vp8_results/enhinterpn.html Results on high precision mv and on the hd set are to follow. Patch 6: Adding a missing condition for CONFIG_SIXTEENTH_SUBPEL_UV in vp8/common/x86/x86_systemdependent.c Patch 7: Cleaning up various debug messages. Patch 8: Merge conflict Change-Id: I5b1d844457aefd7414a9e4e0e06c6ed38fd8cc04
2012-02-16 18:29:54 +01:00
#ifdef DEBUG_DEC_MV
int dec_mvcount = 0;
#endif
// #define DEC_DEBUG
#ifdef DEC_DEBUG
extern int dec_debug;
#endif
Supporting high precision 1/8-pel motion vectors This is the initial patch for supporting 1/8th pel motion. Currently if we configure with enable-high-precision-mv, all motion vectors would default to 1/8 pel. Encode and decode syncs fine with the current code. In the next phase the code will be refactored so that we can choose the 1/8 pel mode adaptively at a frame/segment/mb level. Derf results: http://www.corp.google.com/~debargha/vp8_results/enhinterp_hpmv.html (about 0.83% better than 8-tap interpoaltion) Patch 3: Rebased. Also adding 1/16th pel interpolation for U and V Patch 4: HD results. http://www.corp.google.com/~debargha/vp8_results/enhinterp_hd_hpmv.html Seems impressive (unless I am doing something wrong). Patch 5: Added mmx/sse for bilateral filtering, as well as enforced use of c-versions of subpel filters with 8-taps and 1/16th pel; Also redesigned the 8-tap filters to reduce the cut-off in order to introduce a denoising effect. There is a new configure option sixteenth-subpel-uv which will use 1/16 th pel interpolation for uv, if the motion vectors have 1/8 pel accuracy. With the fixes the results are promising on the derf set. The enhanced interpolation option with 8-taps alone gives 3% improvement over thei derf set: http://www.corp.google.com/~debargha/vp8_results/enhinterpn.html Results on high precision mv and on the hd set are to follow. Patch 6: Adding a missing condition for CONFIG_SIXTEENTH_SUBPEL_UV in vp8/common/x86/x86_systemdependent.c Patch 7: Cleaning up various debug messages. Patch 8: Merge conflict Change-Id: I5b1d844457aefd7414a9e4e0e06c6ed38fd8cc04
2012-02-16 18:29:54 +01:00
static int read_bmode(vp9_reader *bc, const vp9_prob *p) {
B_PREDICTION_MODE m = treed_read(bc, vp9_bmode_tree, p);
#if CONFIG_NEWBINTRAMODES
if (m == B_CONTEXT_PRED - CONTEXT_PRED_REPLACEMENTS)
m = B_CONTEXT_PRED;
assert(m < B_CONTEXT_PRED - CONTEXT_PRED_REPLACEMENTS || m == B_CONTEXT_PRED);
#endif
return m;
}
static int read_kf_bmode(vp9_reader *bc, const vp9_prob *p) {
return treed_read(bc, vp9_kf_bmode_tree, p);
}
static int read_ymode(vp9_reader *bc, const vp9_prob *p) {
return treed_read(bc, vp9_ymode_tree, p);
}
static int read_sb_ymode(vp9_reader *bc, const vp9_prob *p) {
return treed_read(bc, vp9_sb_ymode_tree, p);
}
static int read_kf_sb_ymode(vp9_reader *bc, const vp9_prob *p) {
return treed_read(bc, vp9_uv_mode_tree, p);
}
static int read_kf_mb_ymode(vp9_reader *bc, const vp9_prob *p) {
return treed_read(bc, vp9_kf_ymode_tree, p);
}
static int read_i8x8_mode(vp9_reader *bc, const vp9_prob *p) {
return treed_read(bc, vp9_i8x8_mode_tree, p);
}
static int read_uv_mode(vp9_reader *bc, const vp9_prob *p) {
return treed_read(bc, vp9_uv_mode_tree, p);
}
// This function reads the current macro block's segnent id from the bitstream
// It should only be called if a segment map update is indicated.
static void read_mb_segid(vp9_reader *r, MB_MODE_INFO *mi,
MACROBLOCKD *xd) {
/* Is segmentation enabled */
if (xd->segmentation_enabled && xd->update_mb_segmentation_map) {
/* If so then read the segment id. */
if (vp9_read(r, xd->mb_segment_tree_probs[0]))
mi->segment_id =
(unsigned char)(2 + vp9_read(r, xd->mb_segment_tree_probs[2]));
else
mi->segment_id =
(unsigned char)(vp9_read(r, xd->mb_segment_tree_probs[1]));
}
}
// This function reads the current macro block's segnent id from the bitstream
// It should only be called if a segment map update is indicated.
static void read_mb_segid_except(VP9_COMMON *cm,
vp9_reader *r, MB_MODE_INFO *mi,
MACROBLOCKD *xd, int mb_row, int mb_col) {
int pred_seg_id = vp9_get_pred_mb_segid(cm, xd,
mb_row * cm->mb_cols + mb_col);
const vp9_prob *p = xd->mb_segment_tree_probs;
vp9_prob p1 = xd->mb_segment_mispred_tree_probs[pred_seg_id];
/* Is segmentation enabled */
if (xd->segmentation_enabled && xd->update_mb_segmentation_map) {
/* If so then read the segment id. */
if (vp9_read(r, p1)) {
if (pred_seg_id < 2)
mi->segment_id = 2 + vp9_read(r, p[2]);
else
mi->segment_id = 2 + (pred_seg_id == 2);
} else {
if (pred_seg_id >= 2)
mi->segment_id = vp9_read(r, p[1]);
else
mi->segment_id = pred_seg_id == 0;
}
}
}
#if CONFIG_NEW_MVREF
int vp9_read_mv_ref_id(vp9_reader *r,
vp9_prob * ref_id_probs) {
int ref_index = 0;
if (vp9_read(r, ref_id_probs[0])) {
ref_index++;
if (vp9_read(r, ref_id_probs[1])) {
ref_index++;
if (vp9_read(r, ref_id_probs[2]))
ref_index++;
}
}
return ref_index;
}
#endif
extern const int vp9_i8x8_block[4];
static void kfread_modes(VP9D_COMP *pbi,
MODE_INFO *m,
int mb_row,
int mb_col,
BOOL_DECODER* const bc) {
VP9_COMMON *const cm = &pbi->common;
[WIP] Add column-based tiling. This patch adds column-based tiling. The idea is to make each tile independently decodable (after reading the common frame header) and also independendly encodable (minus within-frame cost adjustments in the RD loop) to speed-up hardware & software en/decoders if they used multi-threading. Column-based tiling has the added advantage (over other tiling methods) that it minimizes realtime use-case latency, since all threads can start encoding data as soon as the first SB-row worth of data is available to the encoder. There is some test code that does random tile ordering in the decoder, to confirm that each tile is indeed independently decodable from other tiles in the same frame. At tile edges, all contexts assume default values (i.e. 0, 0 motion vector, no coefficients, DC intra4x4 mode), and motion vector search and ordering do not cross tiles in the same frame. t log Tile independence is not maintained between frames ATM, i.e. tile 0 of frame 1 is free to use motion vectors that point into any tile of frame 0. We support 1 (i.e. no tiling), 2 or 4 column-tiles. The loopfilter crosses tile boundaries. I discussed this briefly with Aki and he says that's OK. An in-loop loopfilter would need to do some sync between tile threads, but that shouldn't be a big issue. Resuls: with tiling disabled, we go up slightly because of improved edge use in the intra4x4 prediction. With 2 tiles, we lose about ~1% on derf, ~0.35% on HD and ~0.55% on STD/HD. With 4 tiles, we lose another ~1.5% on derf ~0.77% on HD and ~0.85% on STD/HD. Most of this loss is concentrated in the low-bitrate end of clips, and most of it is because of the loss of edges at tile boundaries and the resulting loss of intra predictors. TODO: - more tiles (perhaps allow row-based tiling also, and max. 8 tiles)? - maybe optionally (for EC purposes), motion vectors themselves should not cross tile edges, or we should emulate such borders as if they were off-frame, to limit error propagation to within one tile only. This doesn't have to be the default behaviour but could be an optional bitstream flag. Change-Id: I5951c3a0742a767b20bc9fb5af685d9892c2c96f
2013-02-01 18:35:28 +01:00
MACROBLOCKD *const xd = &pbi->mb;
const int mis = pbi->common.mode_info_stride;
int map_index = mb_row * pbi->common.mb_cols + mb_col;
MB_PREDICTION_MODE y_mode;
// Read the Macroblock segmentation map if it is being updated explicitly
// this frame (reset to 0 by default).
m->mbmi.segment_id = 0;
if (pbi->mb.update_mb_segmentation_map) {
read_mb_segid(bc, &m->mbmi, &pbi->mb);
if (m->mbmi.sb_type) {
const int nmbs = 1 << m->mbmi.sb_type;
const int ymbs = MIN(cm->mb_rows - mb_row, nmbs);
const int xmbs = MIN(cm->mb_cols - mb_col, nmbs);
int x, y;
for (y = 0; y < ymbs; y++) {
for (x = 0; x < xmbs; x++) {
cm->last_frame_seg_map[map_index + x + y * cm->mb_cols] =
m->mbmi.segment_id;
}
}
} else {
cm->last_frame_seg_map[map_index] = m->mbmi.segment_id;
}
}
m->mbmi.mb_skip_coeff = 0;
if (pbi->common.mb_no_coeff_skip &&
(!vp9_segfeature_active(&pbi->mb,
m->mbmi.segment_id, SEG_LVL_SKIP))) {
MACROBLOCKD *const xd = &pbi->mb;
m->mbmi.mb_skip_coeff =
vp9_read(bc, vp9_get_pred_prob(cm, xd, PRED_MBSKIP));
} else {
if (vp9_segfeature_active(&pbi->mb,
m->mbmi.segment_id, SEG_LVL_SKIP)) {
m->mbmi.mb_skip_coeff = 1;
} else
m->mbmi.mb_skip_coeff = 0;
}
if (m->mbmi.sb_type) {
y_mode = (MB_PREDICTION_MODE) read_kf_sb_ymode(bc,
pbi->common.sb_kf_ymode_prob[pbi->common.kf_ymode_probs_index]);
} else {
y_mode = (MB_PREDICTION_MODE) read_kf_mb_ymode(bc,
pbi->common.kf_ymode_prob[pbi->common.kf_ymode_probs_index]);
}
m->mbmi.ref_frame = INTRA_FRAME;
if ((m->mbmi.mode = y_mode) == B_PRED) {
int i = 0;
do {
const B_PREDICTION_MODE A = above_block_mode(m, i, mis);
[WIP] Add column-based tiling. This patch adds column-based tiling. The idea is to make each tile independently decodable (after reading the common frame header) and also independendly encodable (minus within-frame cost adjustments in the RD loop) to speed-up hardware & software en/decoders if they used multi-threading. Column-based tiling has the added advantage (over other tiling methods) that it minimizes realtime use-case latency, since all threads can start encoding data as soon as the first SB-row worth of data is available to the encoder. There is some test code that does random tile ordering in the decoder, to confirm that each tile is indeed independently decodable from other tiles in the same frame. At tile edges, all contexts assume default values (i.e. 0, 0 motion vector, no coefficients, DC intra4x4 mode), and motion vector search and ordering do not cross tiles in the same frame. t log Tile independence is not maintained between frames ATM, i.e. tile 0 of frame 1 is free to use motion vectors that point into any tile of frame 0. We support 1 (i.e. no tiling), 2 or 4 column-tiles. The loopfilter crosses tile boundaries. I discussed this briefly with Aki and he says that's OK. An in-loop loopfilter would need to do some sync between tile threads, but that shouldn't be a big issue. Resuls: with tiling disabled, we go up slightly because of improved edge use in the intra4x4 prediction. With 2 tiles, we lose about ~1% on derf, ~0.35% on HD and ~0.55% on STD/HD. With 4 tiles, we lose another ~1.5% on derf ~0.77% on HD and ~0.85% on STD/HD. Most of this loss is concentrated in the low-bitrate end of clips, and most of it is because of the loss of edges at tile boundaries and the resulting loss of intra predictors. TODO: - more tiles (perhaps allow row-based tiling also, and max. 8 tiles)? - maybe optionally (for EC purposes), motion vectors themselves should not cross tile edges, or we should emulate such borders as if they were off-frame, to limit error propagation to within one tile only. This doesn't have to be the default behaviour but could be an optional bitstream flag. Change-Id: I5951c3a0742a767b20bc9fb5af685d9892c2c96f
2013-02-01 18:35:28 +01:00
const B_PREDICTION_MODE L = (xd->left_available || (i & 3)) ?
left_block_mode(m, i) : B_DC_PRED;
m->bmi[i].as_mode.first =
(B_PREDICTION_MODE) read_kf_bmode(
bc, pbi->common.kf_bmode_prob [A] [L]);
} while (++i < 16);
}
if ((m->mbmi.mode = y_mode) == I8X8_PRED) {
int i;
int mode8x8;
for (i = 0; i < 4; i++) {
int ib = vp9_i8x8_block[i];
mode8x8 = read_i8x8_mode(bc, pbi->common.fc.i8x8_mode_prob);
m->bmi[ib + 0].as_mode.first = mode8x8;
m->bmi[ib + 1].as_mode.first = mode8x8;
m->bmi[ib + 4].as_mode.first = mode8x8;
m->bmi[ib + 5].as_mode.first = mode8x8;
}
} else
m->mbmi.uv_mode = (MB_PREDICTION_MODE)read_uv_mode(bc,
pbi->common.kf_uv_mode_prob[m->mbmi.mode]);
if (cm->txfm_mode == TX_MODE_SELECT && m->mbmi.mb_skip_coeff == 0 &&
m->mbmi.mode <= I8X8_PRED) {
// FIXME(rbultje) code ternary symbol once all experiments are merged
m->mbmi.txfm_size = vp9_read(bc, cm->prob_tx[0]);
32x32 transform for superblocks. This adds Debargha's DCT/DWT hybrid and a regular 32x32 DCT, and adds code all over the place to wrap that in the bitstream/encoder/decoder/RD. Some implementation notes (these probably need careful review): - token range is extended by 1 bit, since the value range out of this transform is [-16384,16383]. - the coefficients coming out of the FDCT are manually scaled back by 1 bit, or else they won't fit in int16_t (they are 17 bits). Because of this, the RD error scoring does not right-shift the MSE score by two (unlike for 4x4/8x8/16x16). - to compensate for this loss in precision, the quantizer is halved also. This is currently a little hacky. - FDCT and IDCT is double-only right now. Needs a fixed-point impl. - There are no default probabilities for the 32x32 transform yet; I'm simply using the 16x16 luma ones. A future commit will add newly generated probabilities for all transforms. - No ADST version. I don't think we'll add one for this level; if an ADST is desired, transform-size selection can scale back to 16x16 or lower, and use an ADST at that level. Additional notes specific to Debargha's DWT/DCT hybrid: - coefficient scale is different for the top/left 16x16 (DCT-over-DWT) block than for the rest (DWT pixel differences) of the block. Therefore, RD error scoring isn't easily scalable between coefficient and pixel domain. Thus, unfortunately, we need to compute the RD distortion in the pixel domain until we figure out how to scale these appropriately. Change-Id: I00386f20f35d7fabb19aba94c8162f8aee64ef2b
2012-12-07 23:45:05 +01:00
if (m->mbmi.txfm_size != TX_4X4 && m->mbmi.mode != I8X8_PRED) {
m->mbmi.txfm_size += vp9_read(bc, cm->prob_tx[1]);
if (m->mbmi.txfm_size != TX_8X8 && m->mbmi.sb_type)
32x32 transform for superblocks. This adds Debargha's DCT/DWT hybrid and a regular 32x32 DCT, and adds code all over the place to wrap that in the bitstream/encoder/decoder/RD. Some implementation notes (these probably need careful review): - token range is extended by 1 bit, since the value range out of this transform is [-16384,16383]. - the coefficients coming out of the FDCT are manually scaled back by 1 bit, or else they won't fit in int16_t (they are 17 bits). Because of this, the RD error scoring does not right-shift the MSE score by two (unlike for 4x4/8x8/16x16). - to compensate for this loss in precision, the quantizer is halved also. This is currently a little hacky. - FDCT and IDCT is double-only right now. Needs a fixed-point impl. - There are no default probabilities for the 32x32 transform yet; I'm simply using the 16x16 luma ones. A future commit will add newly generated probabilities for all transforms. - No ADST version. I don't think we'll add one for this level; if an ADST is desired, transform-size selection can scale back to 16x16 or lower, and use an ADST at that level. Additional notes specific to Debargha's DWT/DCT hybrid: - coefficient scale is different for the top/left 16x16 (DCT-over-DWT) block than for the rest (DWT pixel differences) of the block. Therefore, RD error scoring isn't easily scalable between coefficient and pixel domain. Thus, unfortunately, we need to compute the RD distortion in the pixel domain until we figure out how to scale these appropriately. Change-Id: I00386f20f35d7fabb19aba94c8162f8aee64ef2b
2012-12-07 23:45:05 +01:00
m->mbmi.txfm_size += vp9_read(bc, cm->prob_tx[2]);
}
} else if (cm->txfm_mode >= ALLOW_32X32 && m->mbmi.sb_type) {
32x32 transform for superblocks. This adds Debargha's DCT/DWT hybrid and a regular 32x32 DCT, and adds code all over the place to wrap that in the bitstream/encoder/decoder/RD. Some implementation notes (these probably need careful review): - token range is extended by 1 bit, since the value range out of this transform is [-16384,16383]. - the coefficients coming out of the FDCT are manually scaled back by 1 bit, or else they won't fit in int16_t (they are 17 bits). Because of this, the RD error scoring does not right-shift the MSE score by two (unlike for 4x4/8x8/16x16). - to compensate for this loss in precision, the quantizer is halved also. This is currently a little hacky. - FDCT and IDCT is double-only right now. Needs a fixed-point impl. - There are no default probabilities for the 32x32 transform yet; I'm simply using the 16x16 luma ones. A future commit will add newly generated probabilities for all transforms. - No ADST version. I don't think we'll add one for this level; if an ADST is desired, transform-size selection can scale back to 16x16 or lower, and use an ADST at that level. Additional notes specific to Debargha's DWT/DCT hybrid: - coefficient scale is different for the top/left 16x16 (DCT-over-DWT) block than for the rest (DWT pixel differences) of the block. Therefore, RD error scoring isn't easily scalable between coefficient and pixel domain. Thus, unfortunately, we need to compute the RD distortion in the pixel domain until we figure out how to scale these appropriately. Change-Id: I00386f20f35d7fabb19aba94c8162f8aee64ef2b
2012-12-07 23:45:05 +01:00
m->mbmi.txfm_size = TX_32X32;
} else if (cm->txfm_mode >= ALLOW_16X16 && m->mbmi.mode <= TM_PRED) {
m->mbmi.txfm_size = TX_16X16;
} else if (cm->txfm_mode >= ALLOW_8X8 && m->mbmi.mode != B_PRED) {
m->mbmi.txfm_size = TX_8X8;
} else {
m->mbmi.txfm_size = TX_4X4;
}
}
2010-05-18 17:58:33 +02:00
static int read_nmv_component(vp9_reader *r,
int rv,
const nmv_component *mvcomp) {
int v, s, z, c, o, d;
s = vp9_read(r, mvcomp->sign);
c = treed_read(r, vp9_mv_class_tree, mvcomp->classes);
if (c == MV_CLASS_0) {
d = treed_read(r, vp9_mv_class0_tree, mvcomp->class0);
} else {
int i, b;
d = 0;
b = c + CLASS0_BITS - 1; /* number of bits */
for (i = 0; i < b; ++i)
d |= (vp9_read(r, mvcomp->bits[i]) << i);
}
o = d << 3;
z = vp9_get_mv_mag(c, o);
v = (s ? -(z + 8) : (z + 8));
return v;
}
static int read_nmv_component_fp(vp9_reader *r,
int v,
int rv,
const nmv_component *mvcomp,
int usehp) {
int s, z, c, o, d, e, f;
s = v < 0;
z = (s ? -v : v) - 1; /* magnitude - 1 */
z &= ~7;
c = vp9_get_mv_class(z, &o);
d = o >> 3;
if (c == MV_CLASS_0) {
f = treed_read(r, vp9_mv_fp_tree, mvcomp->class0_fp[d]);
} else {
f = treed_read(r, vp9_mv_fp_tree, mvcomp->fp);
}
o += (f << 1);
if (usehp) {
if (c == MV_CLASS_0) {
e = vp9_read(r, mvcomp->class0_hp);
} else {
e = vp9_read(r, mvcomp->hp);
}
o += e;
} else {
++o; /* Note if hp is not used, the default value of the hp bit is 1 */
}
z = vp9_get_mv_mag(c, o);
v = (s ? -(z + 1) : (z + 1));
return v;
}
static void read_nmv(vp9_reader *r, MV *mv, const MV *ref,
const nmv_context *mvctx) {
MV_JOINT_TYPE j = treed_read(r, vp9_mv_joint_tree, mvctx->joints);
mv->row = mv-> col = 0;
if (j == MV_JOINT_HZVNZ || j == MV_JOINT_HNZVNZ) {
mv->row = read_nmv_component(r, ref->row, &mvctx->comps[0]);
}
if (j == MV_JOINT_HNZVZ || j == MV_JOINT_HNZVNZ) {
mv->col = read_nmv_component(r, ref->col, &mvctx->comps[1]);
}
}
static void read_nmv_fp(vp9_reader *r, MV *mv, const MV *ref,
const nmv_context *mvctx, int usehp) {
MV_JOINT_TYPE j = vp9_get_mv_joint(*mv);
usehp = usehp && vp9_use_nmv_hp(ref);
if (j == MV_JOINT_HZVNZ || j == MV_JOINT_HNZVNZ) {
mv->row = read_nmv_component_fp(r, mv->row, ref->row, &mvctx->comps[0],
usehp);
}
if (j == MV_JOINT_HNZVZ || j == MV_JOINT_HNZVNZ) {
mv->col = read_nmv_component_fp(r, mv->col, ref->col, &mvctx->comps[1],
usehp);
}
/*
printf("MV: %d %d REF: %d %d\n", mv->row + ref->row, mv->col + ref->col,
ref->row, ref->col);
*/
}
static void update_nmv(vp9_reader *bc, vp9_prob *const p,
const vp9_prob upd_p) {
if (vp9_read(bc, upd_p)) {
#ifdef LOW_PRECISION_MV_UPDATE
*p = (vp9_read_literal(bc, 7) << 1) | 1;
#else
*p = (vp9_read_literal(bc, 8));
#endif
}
}
static void read_nmvprobs(vp9_reader *bc, nmv_context *mvctx,
int usehp) {
int i, j, k;
#ifdef MV_GROUP_UPDATE
if (!vp9_read_bit(bc)) return;
#endif
for (j = 0; j < MV_JOINTS - 1; ++j) {
update_nmv(bc, &mvctx->joints[j],
VP9_NMV_UPDATE_PROB);
}
for (i = 0; i < 2; ++i) {
update_nmv(bc, &mvctx->comps[i].sign,
VP9_NMV_UPDATE_PROB);
for (j = 0; j < MV_CLASSES - 1; ++j) {
update_nmv(bc, &mvctx->comps[i].classes[j],
VP9_NMV_UPDATE_PROB);
}
for (j = 0; j < CLASS0_SIZE - 1; ++j) {
update_nmv(bc, &mvctx->comps[i].class0[j],
VP9_NMV_UPDATE_PROB);
}
for (j = 0; j < MV_OFFSET_BITS; ++j) {
update_nmv(bc, &mvctx->comps[i].bits[j],
VP9_NMV_UPDATE_PROB);
}
}
for (i = 0; i < 2; ++i) {
for (j = 0; j < CLASS0_SIZE; ++j) {
for (k = 0; k < 3; ++k)
update_nmv(bc, &mvctx->comps[i].class0_fp[j][k],
VP9_NMV_UPDATE_PROB);
}
for (j = 0; j < 3; ++j) {
update_nmv(bc, &mvctx->comps[i].fp[j],
VP9_NMV_UPDATE_PROB);
}
}
if (usehp) {
for (i = 0; i < 2; ++i) {
update_nmv(bc, &mvctx->comps[i].class0_hp,
VP9_NMV_UPDATE_PROB);
update_nmv(bc, &mvctx->comps[i].hp,
VP9_NMV_UPDATE_PROB);
}
}
}
// Read the referncence frame
static MV_REFERENCE_FRAME read_ref_frame(VP9D_COMP *pbi,
vp9_reader *const bc,
unsigned char segment_id) {
MV_REFERENCE_FRAME ref_frame;
int seg_ref_active;
int seg_ref_count = 0;
VP9_COMMON *const cm = &pbi->common;
MACROBLOCKD *const xd = &pbi->mb;
seg_ref_active = vp9_segfeature_active(xd,
segment_id,
SEG_LVL_REF_FRAME);
// If segment coding enabled does the segment allow for more than one
// possible reference frame
if (seg_ref_active) {
seg_ref_count = vp9_check_segref(xd, segment_id, INTRA_FRAME) +
vp9_check_segref(xd, segment_id, LAST_FRAME) +
vp9_check_segref(xd, segment_id, GOLDEN_FRAME) +
vp9_check_segref(xd, segment_id, ALTREF_FRAME);
}
// Segment reference frame features not available or allows for
// multiple reference frame options
if (!seg_ref_active || (seg_ref_count > 1)) {
// Values used in prediction model coding
unsigned char prediction_flag;
vp9_prob pred_prob;
MV_REFERENCE_FRAME pred_ref;
// Get the context probability the prediction flag
pred_prob = vp9_get_pred_prob(cm, xd, PRED_REF);
// Read the prediction status flag
prediction_flag = (unsigned char)vp9_read(bc, pred_prob);
// Store the prediction flag.
vp9_set_pred_flag(xd, PRED_REF, prediction_flag);
// Get the predicted reference frame.
pred_ref = vp9_get_pred_ref(cm, xd);
// If correctly predicted then use the predicted value
if (prediction_flag) {
ref_frame = pred_ref;
}
// else decode the explicitly coded value
else {
vp9_prob mod_refprobs[PREDICTION_PROBS];
vpx_memcpy(mod_refprobs,
cm->mod_refprobs[pred_ref], sizeof(mod_refprobs));
// If segment coding enabled blank out options that cant occur by
// setting the branch probability to 0.
if (seg_ref_active) {
mod_refprobs[INTRA_FRAME] *=
vp9_check_segref(xd, segment_id, INTRA_FRAME);
mod_refprobs[LAST_FRAME] *=
vp9_check_segref(xd, segment_id, LAST_FRAME);
mod_refprobs[GOLDEN_FRAME] *=
(vp9_check_segref(xd, segment_id, GOLDEN_FRAME) *
vp9_check_segref(xd, segment_id, ALTREF_FRAME));
}
// Default to INTRA_FRAME (value 0)
ref_frame = INTRA_FRAME;
// Do we need to decode the Intra/Inter branch
if (mod_refprobs[0])
ref_frame = (MV_REFERENCE_FRAME) vp9_read(bc, mod_refprobs[0]);
else
ref_frame++;
if (ref_frame) {
// Do we need to decode the Last/Gf_Arf branch
if (mod_refprobs[1])
ref_frame += vp9_read(bc, mod_refprobs[1]);
else
ref_frame++;
if (ref_frame > 1) {
// Do we need to decode the GF/Arf branch
if (mod_refprobs[2])
ref_frame += vp9_read(bc, mod_refprobs[2]);
else {
if (seg_ref_active) {
if ((pred_ref == GOLDEN_FRAME) ||
!vp9_check_segref(xd, segment_id, GOLDEN_FRAME)) {
ref_frame = ALTREF_FRAME;
} else
ref_frame = GOLDEN_FRAME;
} else
ref_frame = (pred_ref == GOLDEN_FRAME)
? ALTREF_FRAME : GOLDEN_FRAME;
}
}
}
}
}
// Segment reference frame features are enabled
else {
// The reference frame for the mb is considered as correclty predicted
// if it is signaled at the segment level for the purposes of the
// common prediction model
vp9_set_pred_flag(xd, PRED_REF, 1);
ref_frame = vp9_get_pred_ref(cm, xd);
}
return (MV_REFERENCE_FRAME)ref_frame;
}
2010-05-18 17:58:33 +02:00
static MB_PREDICTION_MODE read_sb_mv_ref(vp9_reader *bc, const vp9_prob *p) {
return (MB_PREDICTION_MODE) treed_read(bc, vp9_sb_mv_ref_tree, p);
}
2010-05-18 17:58:33 +02:00
static MB_PREDICTION_MODE read_mv_ref(vp9_reader *bc, const vp9_prob *p) {
return (MB_PREDICTION_MODE) treed_read(bc, vp9_mv_ref_tree, p);
2010-05-18 17:58:33 +02:00
}
static B_PREDICTION_MODE sub_mv_ref(vp9_reader *bc, const vp9_prob *p) {
return (B_PREDICTION_MODE) treed_read(bc, vp9_sub_mv_ref_tree, p);
2010-05-18 17:58:33 +02:00
}
#ifdef VPX_MODE_COUNT
unsigned int vp9_mv_cont_count[5][4] = {
{ 0, 0, 0, 0 },
{ 0, 0, 0, 0 },
{ 0, 0, 0, 0 },
{ 0, 0, 0, 0 },
{ 0, 0, 0, 0 }
2010-05-18 17:58:33 +02:00
};
#endif
2010-05-18 17:58:33 +02:00
static const unsigned char mbsplit_fill_count[4] = {8, 8, 4, 1};
static const unsigned char mbsplit_fill_offset[4][16] = {
{ 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15},
{ 0, 1, 4, 5, 8, 9, 12, 13, 2, 3, 6, 7, 10, 11, 14, 15},
{ 0, 1, 4, 5, 2, 3, 6, 7, 8, 9, 12, 13, 10, 11, 14, 15},
{ 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15}
};
2010-05-18 17:58:33 +02:00
static void read_switchable_interp_probs(VP9D_COMP* const pbi,
BOOL_DECODER* const bc) {
VP9_COMMON *const cm = &pbi->common;
int i, j;
for (j = 0; j <= VP9_SWITCHABLE_FILTERS; ++j) {
for (i = 0; i < VP9_SWITCHABLE_FILTERS - 1; ++i) {
cm->fc.switchable_interp_prob[j][i] = vp9_read_literal(bc, 8);
}
}
//printf("DECODER: %d %d\n", cm->fc.switchable_interp_prob[0],
//cm->fc.switchable_interp_prob[1]);
}
2010-05-18 17:58:33 +02:00
static void mb_mode_mv_init(VP9D_COMP *pbi, vp9_reader *bc) {
VP9_COMMON *const cm = &pbi->common;
nmv_context *const nmvc = &pbi->common.fc.nmvc;
MACROBLOCKD *const xd = &pbi->mb;
2010-05-18 17:58:33 +02:00
if (cm->frame_type == KEY_FRAME) {
if (!cm->kf_ymode_probs_update)
cm->kf_ymode_probs_index = vp9_read_literal(bc, 3);
} else {
if (cm->mcomp_filter_type == SWITCHABLE)
read_switchable_interp_probs(pbi, bc);
#if CONFIG_COMP_INTERINTRA_PRED
if (cm->use_interintra) {
if (vp9_read(bc, VP9_UPD_INTERINTRA_PROB))
cm->fc.interintra_prob = (vp9_prob)vp9_read_literal(bc, 8);
}
#endif
// Decode the baseline probabilities for decoding reference frame
cm->prob_intra_coded = (vp9_prob)vp9_read_literal(bc, 8);
cm->prob_last_coded = (vp9_prob)vp9_read_literal(bc, 8);
cm->prob_gf_coded = (vp9_prob)vp9_read_literal(bc, 8);
// Computes a modified set of probabilities for use when reference
// frame prediction fails.
vp9_compute_mod_refprobs(cm);
pbi->common.comp_pred_mode = vp9_read(bc, 128);
if (cm->comp_pred_mode)
cm->comp_pred_mode += vp9_read(bc, 128);
if (cm->comp_pred_mode == HYBRID_PREDICTION) {
int i;
for (i = 0; i < COMP_PRED_CONTEXTS; i++)
cm->prob_comppred[i] = (vp9_prob)vp9_read_literal(bc, 8);
}
2010-05-18 17:58:33 +02:00
if (vp9_read_bit(bc)) {
int i = 0;
2010-05-18 17:58:33 +02:00
do {
cm->fc.ymode_prob[i] = (vp9_prob) vp9_read_literal(bc, 8);
} while (++i < VP9_YMODES - 1);
}
if (vp9_read_bit(bc)) {
int i = 0;
do {
cm->fc.sb_ymode_prob[i] = (vp9_prob) vp9_read_literal(bc, 8);
} while (++i < VP9_I32X32_MODES - 1);
}
read_nmvprobs(bc, nmvc, xd->allow_high_precision_mv);
}
}
2010-05-18 17:58:33 +02:00
// This function either reads the segment id for the current macroblock from
// the bitstream or if the value is temporally predicted asserts the predicted
// value
static void read_mb_segment_id(VP9D_COMP *pbi,
int mb_row, int mb_col,
BOOL_DECODER* const bc) {
VP9_COMMON *const cm = &pbi->common;
MACROBLOCKD *const xd = &pbi->mb;
MODE_INFO *mi = xd->mode_info_context;
MB_MODE_INFO *mbmi = &mi->mbmi;
int index = mb_row * pbi->common.mb_cols + mb_col;
if (xd->segmentation_enabled) {
if (xd->update_mb_segmentation_map) {
// Is temporal coding of the segment id for this mb enabled.
if (cm->temporal_update) {
// Get the context based probability for reading the
// prediction status flag
vp9_prob pred_prob =
vp9_get_pred_prob(cm, xd, PRED_SEG_ID);
// Read the prediction status flag
unsigned char seg_pred_flag =
(unsigned char)vp9_read(bc, pred_prob);
// Store the prediction flag.
vp9_set_pred_flag(xd, PRED_SEG_ID, seg_pred_flag);
// If the value is flagged as correctly predicted
// then use the predicted value
if (seg_pred_flag) {
mbmi->segment_id = vp9_get_pred_mb_segid(cm, xd, index);
}
// Else .... decode it explicitly
else {
read_mb_segid_except(cm, bc, mbmi, xd, mb_row, mb_col);
}
}
// Normal unpredicted coding mode
else {
read_mb_segid(bc, mbmi, xd);
}
if (mbmi->sb_type) {
const int nmbs = 1 << mbmi->sb_type;
const int ymbs = MIN(cm->mb_rows - mb_row, nmbs);
const int xmbs = MIN(cm->mb_cols - mb_col, nmbs);
int x, y;
for (y = 0; y < ymbs; y++) {
for (x = 0; x < xmbs; x++) {
cm->last_frame_seg_map[index + x + y * cm->mb_cols] =
mbmi->segment_id;
}
}
} else {
cm->last_frame_seg_map[index] = mbmi->segment_id;
}
} else {
if (mbmi->sb_type) {
const int nmbs = 1 << mbmi->sb_type;
const int ymbs = MIN(cm->mb_rows - mb_row, nmbs);
const int xmbs = MIN(cm->mb_cols - mb_col, nmbs);
unsigned segment_id = -1;
int x, y;
for (y = 0; y < ymbs; y++) {
for (x = 0; x < xmbs; x++) {
segment_id = MIN(segment_id,
cm->last_frame_seg_map[index + x +
y * cm->mb_cols]);
}
}
mbmi->segment_id = segment_id;
} else {
mbmi->segment_id = cm->last_frame_seg_map[index];
}
}
} else {
// The encoder explicitly sets the segment_id to 0
// when segmentation is disabled
mbmi->segment_id = 0;
}
}
static void read_mb_modes_mv(VP9D_COMP *pbi, MODE_INFO *mi, MB_MODE_INFO *mbmi,
MODE_INFO *prev_mi,
int mb_row, int mb_col,
BOOL_DECODER* const bc) {
VP9_COMMON *const cm = &pbi->common;
nmv_context *const nmvc = &pbi->common.fc.nmvc;
const int mis = pbi->common.mode_info_stride;
MACROBLOCKD *const xd = &pbi->mb;
int_mv *const mv = &mbmi->mv[0];
int mb_to_left_edge;
int mb_to_right_edge;
int mb_to_top_edge;
int mb_to_bottom_edge;
const int mb_size = 1 << mi->mbmi.sb_type;
mb_to_top_edge = xd->mb_to_top_edge;
mb_to_bottom_edge = xd->mb_to_bottom_edge;
mb_to_top_edge -= LEFT_TOP_MARGIN;
mb_to_bottom_edge += RIGHT_BOTTOM_MARGIN;
mbmi->need_to_clamp_mvs = 0;
mbmi->need_to_clamp_secondmv = 0;
mbmi->second_ref_frame = NONE;
/* Distance of Mb to the various image edges.
* These specified to 8th pel as they are always compared to MV values that are in 1/8th pel units
*/
xd->mb_to_left_edge =
mb_to_left_edge = -((mb_col * 16) << 3);
mb_to_left_edge -= LEFT_TOP_MARGIN;
xd->mb_to_right_edge =
mb_to_right_edge = ((pbi->common.mb_cols - mb_size - mb_col) * 16) << 3;
mb_to_right_edge += RIGHT_BOTTOM_MARGIN;
// Make sure the MACROBLOCKD mode info pointer is pointed at the
// correct entry for the current macroblock.
xd->mode_info_context = mi;
xd->prev_mode_info_context = prev_mi;
// Read the macroblock segment id.
read_mb_segment_id(pbi, mb_row, mb_col, bc);
if (pbi->common.mb_no_coeff_skip &&
(!vp9_segfeature_active(xd, mbmi->segment_id, SEG_LVL_SKIP))) {
// Read the macroblock coeff skip flag if this feature is in use,
// else default to 0
mbmi->mb_skip_coeff = vp9_read(bc, vp9_get_pred_prob(cm, xd, PRED_MBSKIP));
} else {
if (vp9_segfeature_active(xd, mbmi->segment_id, SEG_LVL_SKIP)) {
mbmi->mb_skip_coeff = 1;
} else
mbmi->mb_skip_coeff = 0;
}
// Read the reference frame
mbmi->ref_frame = read_ref_frame(pbi, bc, mbmi->segment_id);
/*
if (pbi->common.current_video_frame == 1)
printf("ref frame: %d [%d %d]\n", mbmi->ref_frame, mb_row, mb_col);
*/
// If reference frame is an Inter frame
if (mbmi->ref_frame) {
int_mv nearest, nearby, best_mv;
int_mv nearest_second, nearby_second, best_mv_second;
vp9_prob mv_ref_p [VP9_MVREFS - 1];
int recon_y_stride, recon_yoffset;
int recon_uv_stride, recon_uvoffset;
MV_REFERENCE_FRAME ref_frame = mbmi->ref_frame;
{
int ref_fb_idx;
/* Select the appropriate reference frame for this MB */
ref_fb_idx = cm->active_ref_idx[ref_frame - 1];
recon_y_stride = cm->yv12_fb[ref_fb_idx].y_stride ;
recon_uv_stride = cm->yv12_fb[ref_fb_idx].uv_stride;
recon_yoffset = (mb_row * recon_y_stride * 16) + (mb_col * 16);
recon_uvoffset = (mb_row * recon_uv_stride * 8) + (mb_col * 8);
xd->pre.y_buffer = cm->yv12_fb[ref_fb_idx].y_buffer + recon_yoffset;
xd->pre.u_buffer = cm->yv12_fb[ref_fb_idx].u_buffer + recon_uvoffset;
xd->pre.v_buffer = cm->yv12_fb[ref_fb_idx].v_buffer + recon_uvoffset;
#ifdef DEC_DEBUG
if (dec_debug)
printf("%d %d\n", xd->mode_info_context->mbmi.mv[0].as_mv.row,
xd->mode_info_context->mbmi.mv[0].as_mv.col);
#endif
// if (cm->current_video_frame == 1 && mb_row == 4 && mb_col == 5)
// printf("Dello\n");
[WIP] Add column-based tiling. This patch adds column-based tiling. The idea is to make each tile independently decodable (after reading the common frame header) and also independendly encodable (minus within-frame cost adjustments in the RD loop) to speed-up hardware & software en/decoders if they used multi-threading. Column-based tiling has the added advantage (over other tiling methods) that it minimizes realtime use-case latency, since all threads can start encoding data as soon as the first SB-row worth of data is available to the encoder. There is some test code that does random tile ordering in the decoder, to confirm that each tile is indeed independently decodable from other tiles in the same frame. At tile edges, all contexts assume default values (i.e. 0, 0 motion vector, no coefficients, DC intra4x4 mode), and motion vector search and ordering do not cross tiles in the same frame. t log Tile independence is not maintained between frames ATM, i.e. tile 0 of frame 1 is free to use motion vectors that point into any tile of frame 0. We support 1 (i.e. no tiling), 2 or 4 column-tiles. The loopfilter crosses tile boundaries. I discussed this briefly with Aki and he says that's OK. An in-loop loopfilter would need to do some sync between tile threads, but that shouldn't be a big issue. Resuls: with tiling disabled, we go up slightly because of improved edge use in the intra4x4 prediction. With 2 tiles, we lose about ~1% on derf, ~0.35% on HD and ~0.55% on STD/HD. With 4 tiles, we lose another ~1.5% on derf ~0.77% on HD and ~0.85% on STD/HD. Most of this loss is concentrated in the low-bitrate end of clips, and most of it is because of the loss of edges at tile boundaries and the resulting loss of intra predictors. TODO: - more tiles (perhaps allow row-based tiling also, and max. 8 tiles)? - maybe optionally (for EC purposes), motion vectors themselves should not cross tile edges, or we should emulate such borders as if they were off-frame, to limit error propagation to within one tile only. This doesn't have to be the default behaviour but could be an optional bitstream flag. Change-Id: I5951c3a0742a767b20bc9fb5af685d9892c2c96f
2013-02-01 18:35:28 +01:00
vp9_find_mv_refs(cm, xd, mi, cm->error_resilient_mode ? 0 : prev_mi,
ref_frame, mbmi->ref_mvs[ref_frame],
cm->ref_frame_sign_bias);
vp9_mv_ref_probs(&pbi->common, mv_ref_p,
mbmi->mb_mode_context[ref_frame]);
/*
if (pbi->common.current_video_frame == 1) {
int k = mbmi->mb_mode_context[ref_frame];
printf("vp9_mode_contexts: [%d %d %d %d] %d %d %d %d\n",
mb_row, mb_col, ref_frame, k,
cm->fc.vp9_mode_contexts[k][0],
cm->fc.vp9_mode_contexts[k][1],
cm->fc.vp9_mode_contexts[k][2],
cm->fc.vp9_mode_contexts[k][3]);
}
*/
// Is the segment level skip mode enabled
if (vp9_segfeature_active(xd, mbmi->segment_id, SEG_LVL_SKIP)) {
mbmi->mode = ZEROMV;
} else {
if (mbmi->sb_type)
mbmi->mode = read_sb_mv_ref(bc, mv_ref_p);
else
mbmi->mode = read_mv_ref(bc, mv_ref_p);
vp9_accum_mv_refs(&pbi->common, mbmi->mode,
mbmi->mb_mode_context[ref_frame]);
}
if (mbmi->mode != ZEROMV) {
vp9_find_best_ref_mvs(xd,
pbi->common.error_resilient_mode ||
pbi->common.frame_parallel_decoding_mode ?
0 : xd->pre.y_buffer,
recon_y_stride,
mbmi->ref_mvs[ref_frame],
&nearest, &nearby);
best_mv.as_int = (mbmi->ref_mvs[ref_frame][0]).as_int;
}
#ifdef DEC_DEBUG
if (dec_debug)
printf("[D %d %d] %d %d %d %d\n", ref_frame,
mbmi->mb_mode_context[ref_frame],
mv_ref_p[0], mv_ref_p[1], mv_ref_p[2], mv_ref_p[3]);
#endif
}
if (mbmi->mode >= NEARESTMV && mbmi->mode <= SPLITMV)
{
if (cm->mcomp_filter_type == SWITCHABLE) {
mbmi->interp_filter = vp9_switchable_interp[
treed_read(bc, vp9_switchable_interp_tree,
vp9_get_pred_probs(cm, xd, PRED_SWITCHABLE_INTERP))];
} else {
mbmi->interp_filter = cm->mcomp_filter_type;
}
}
if (cm->comp_pred_mode == COMP_PREDICTION_ONLY ||
(cm->comp_pred_mode == HYBRID_PREDICTION &&
vp9_read(bc, vp9_get_pred_prob(cm, xd, PRED_COMP)))) {
/* Since we have 3 reference frames, we can only have 3 unique
* combinations of combinations of 2 different reference frames
* (A-G, G-L or A-L). In the bitstream, we use this to simply
* derive the second reference frame from the first reference
* frame, by saying it's the next one in the enumerator, and
* if that's > n_refs, then the second reference frame is the
* first one in the enumerator. */
mbmi->second_ref_frame = mbmi->ref_frame + 1;
if (mbmi->second_ref_frame == 4)
mbmi->second_ref_frame = 1;
if (mbmi->second_ref_frame > 0) {
int second_ref_fb_idx;
/* Select the appropriate reference frame for this MB */
second_ref_fb_idx = cm->active_ref_idx[mbmi->second_ref_frame - 1];
xd->second_pre.y_buffer =
cm->yv12_fb[second_ref_fb_idx].y_buffer + recon_yoffset;
xd->second_pre.u_buffer =
cm->yv12_fb[second_ref_fb_idx].u_buffer + recon_uvoffset;
xd->second_pre.v_buffer =
cm->yv12_fb[second_ref_fb_idx].v_buffer + recon_uvoffset;
[WIP] Add column-based tiling. This patch adds column-based tiling. The idea is to make each tile independently decodable (after reading the common frame header) and also independendly encodable (minus within-frame cost adjustments in the RD loop) to speed-up hardware & software en/decoders if they used multi-threading. Column-based tiling has the added advantage (over other tiling methods) that it minimizes realtime use-case latency, since all threads can start encoding data as soon as the first SB-row worth of data is available to the encoder. There is some test code that does random tile ordering in the decoder, to confirm that each tile is indeed independently decodable from other tiles in the same frame. At tile edges, all contexts assume default values (i.e. 0, 0 motion vector, no coefficients, DC intra4x4 mode), and motion vector search and ordering do not cross tiles in the same frame. t log Tile independence is not maintained between frames ATM, i.e. tile 0 of frame 1 is free to use motion vectors that point into any tile of frame 0. We support 1 (i.e. no tiling), 2 or 4 column-tiles. The loopfilter crosses tile boundaries. I discussed this briefly with Aki and he says that's OK. An in-loop loopfilter would need to do some sync between tile threads, but that shouldn't be a big issue. Resuls: with tiling disabled, we go up slightly because of improved edge use in the intra4x4 prediction. With 2 tiles, we lose about ~1% on derf, ~0.35% on HD and ~0.55% on STD/HD. With 4 tiles, we lose another ~1.5% on derf ~0.77% on HD and ~0.85% on STD/HD. Most of this loss is concentrated in the low-bitrate end of clips, and most of it is because of the loss of edges at tile boundaries and the resulting loss of intra predictors. TODO: - more tiles (perhaps allow row-based tiling also, and max. 8 tiles)? - maybe optionally (for EC purposes), motion vectors themselves should not cross tile edges, or we should emulate such borders as if they were off-frame, to limit error propagation to within one tile only. This doesn't have to be the default behaviour but could be an optional bitstream flag. Change-Id: I5951c3a0742a767b20bc9fb5af685d9892c2c96f
2013-02-01 18:35:28 +01:00
vp9_find_mv_refs(cm, xd, mi, cm->error_resilient_mode ? 0 : prev_mi,
mbmi->second_ref_frame,
mbmi->ref_mvs[mbmi->second_ref_frame],
cm->ref_frame_sign_bias);
if (mbmi->mode != ZEROMV) {
vp9_find_best_ref_mvs(xd,
pbi->common.error_resilient_mode ||
pbi->common.frame_parallel_decoding_mode ?
0 : xd->second_pre.y_buffer,
recon_y_stride,
mbmi->ref_mvs[mbmi->second_ref_frame],
&nearest_second,
&nearby_second);
best_mv_second = mbmi->ref_mvs[mbmi->second_ref_frame][0];
}
}
} else {
#if CONFIG_COMP_INTERINTRA_PRED
if (pbi->common.use_interintra &&
mbmi->mode >= NEARESTMV && mbmi->mode < SPLITMV &&
mbmi->second_ref_frame == NONE) {
mbmi->second_ref_frame = (vp9_read(bc, pbi->common.fc.interintra_prob) ?
INTRA_FRAME : NONE);
// printf("-- %d (%d)\n", mbmi->second_ref_frame == INTRA_FRAME,
// pbi->common.fc.interintra_prob);
pbi->common.fc.interintra_counts[
mbmi->second_ref_frame == INTRA_FRAME]++;
if (mbmi->second_ref_frame == INTRA_FRAME) {
mbmi->interintra_mode = (MB_PREDICTION_MODE)read_ymode(
bc, pbi->common.fc.ymode_prob);
pbi->common.fc.ymode_counts[mbmi->interintra_mode]++;
#if SEPARATE_INTERINTRA_UV
mbmi->interintra_uv_mode = (MB_PREDICTION_MODE)read_uv_mode(
bc, pbi->common.fc.uv_mode_prob[mbmi->interintra_mode]);
pbi->common.fc.uv_mode_counts[mbmi->interintra_mode]
[mbmi->interintra_uv_mode]++;
#else
mbmi->interintra_uv_mode = mbmi->interintra_mode;
#endif
// printf("** %d %d\n",
// mbmi->interintra_mode, mbmi->interintra_uv_mode);
}
}
#endif
}
#if CONFIG_NEW_MVREF
// if ((mbmi->mode == NEWMV) || (mbmi->mode == SPLITMV))
if (mbmi->mode == NEWMV) {
int best_index;
MV_REFERENCE_FRAME ref_frame = mbmi->ref_frame;
// Encode the index of the choice.
best_index =
vp9_read_mv_ref_id(bc, xd->mb_mv_ref_probs[ref_frame]);
best_mv.as_int = mbmi->ref_mvs[ref_frame][best_index].as_int;
if (mbmi->second_ref_frame > 0) {
ref_frame = mbmi->second_ref_frame;
// Encode the index of the choice.
best_index =
vp9_read_mv_ref_id(bc, xd->mb_mv_ref_probs[ref_frame]);
best_mv_second.as_int = mbmi->ref_mvs[ref_frame][best_index].as_int;
}
}
#endif
mbmi->uv_mode = DC_PRED;
switch (mbmi->mode) {
case SPLITMV: {
const int s = mbmi->partitioning =
treed_read(bc, vp9_mbsplit_tree, cm->fc.mbsplit_prob);
const int num_p = vp9_mbsplit_count [s];
int j = 0;
cm->fc.mbsplit_counts[s]++;
2010-05-18 17:58:33 +02:00
mbmi->need_to_clamp_mvs = 0;
do { /* for each subset j */
int_mv leftmv, abovemv, second_leftmv, second_abovemv;
int_mv blockmv, secondmv;
int k; /* first block in subset j */
int mv_contz;
int blockmode;
k = vp9_mbsplit_offset[s][j];
[WIP] Add column-based tiling. This patch adds column-based tiling. The idea is to make each tile independently decodable (after reading the common frame header) and also independendly encodable (minus within-frame cost adjustments in the RD loop) to speed-up hardware & software en/decoders if they used multi-threading. Column-based tiling has the added advantage (over other tiling methods) that it minimizes realtime use-case latency, since all threads can start encoding data as soon as the first SB-row worth of data is available to the encoder. There is some test code that does random tile ordering in the decoder, to confirm that each tile is indeed independently decodable from other tiles in the same frame. At tile edges, all contexts assume default values (i.e. 0, 0 motion vector, no coefficients, DC intra4x4 mode), and motion vector search and ordering do not cross tiles in the same frame. t log Tile independence is not maintained between frames ATM, i.e. tile 0 of frame 1 is free to use motion vectors that point into any tile of frame 0. We support 1 (i.e. no tiling), 2 or 4 column-tiles. The loopfilter crosses tile boundaries. I discussed this briefly with Aki and he says that's OK. An in-loop loopfilter would need to do some sync between tile threads, but that shouldn't be a big issue. Resuls: with tiling disabled, we go up slightly because of improved edge use in the intra4x4 prediction. With 2 tiles, we lose about ~1% on derf, ~0.35% on HD and ~0.55% on STD/HD. With 4 tiles, we lose another ~1.5% on derf ~0.77% on HD and ~0.85% on STD/HD. Most of this loss is concentrated in the low-bitrate end of clips, and most of it is because of the loss of edges at tile boundaries and the resulting loss of intra predictors. TODO: - more tiles (perhaps allow row-based tiling also, and max. 8 tiles)? - maybe optionally (for EC purposes), motion vectors themselves should not cross tile edges, or we should emulate such borders as if they were off-frame, to limit error propagation to within one tile only. This doesn't have to be the default behaviour but could be an optional bitstream flag. Change-Id: I5951c3a0742a767b20bc9fb5af685d9892c2c96f
2013-02-01 18:35:28 +01:00
leftmv.as_int = left_block_mv(xd, mi, k);
abovemv.as_int = above_block_mv(mi, k, mis);
second_leftmv.as_int = 0;
second_abovemv.as_int = 0;
if (mbmi->second_ref_frame > 0) {
[WIP] Add column-based tiling. This patch adds column-based tiling. The idea is to make each tile independently decodable (after reading the common frame header) and also independendly encodable (minus within-frame cost adjustments in the RD loop) to speed-up hardware & software en/decoders if they used multi-threading. Column-based tiling has the added advantage (over other tiling methods) that it minimizes realtime use-case latency, since all threads can start encoding data as soon as the first SB-row worth of data is available to the encoder. There is some test code that does random tile ordering in the decoder, to confirm that each tile is indeed independently decodable from other tiles in the same frame. At tile edges, all contexts assume default values (i.e. 0, 0 motion vector, no coefficients, DC intra4x4 mode), and motion vector search and ordering do not cross tiles in the same frame. t log Tile independence is not maintained between frames ATM, i.e. tile 0 of frame 1 is free to use motion vectors that point into any tile of frame 0. We support 1 (i.e. no tiling), 2 or 4 column-tiles. The loopfilter crosses tile boundaries. I discussed this briefly with Aki and he says that's OK. An in-loop loopfilter would need to do some sync between tile threads, but that shouldn't be a big issue. Resuls: with tiling disabled, we go up slightly because of improved edge use in the intra4x4 prediction. With 2 tiles, we lose about ~1% on derf, ~0.35% on HD and ~0.55% on STD/HD. With 4 tiles, we lose another ~1.5% on derf ~0.77% on HD and ~0.85% on STD/HD. Most of this loss is concentrated in the low-bitrate end of clips, and most of it is because of the loss of edges at tile boundaries and the resulting loss of intra predictors. TODO: - more tiles (perhaps allow row-based tiling also, and max. 8 tiles)? - maybe optionally (for EC purposes), motion vectors themselves should not cross tile edges, or we should emulate such borders as if they were off-frame, to limit error propagation to within one tile only. This doesn't have to be the default behaviour but could be an optional bitstream flag. Change-Id: I5951c3a0742a767b20bc9fb5af685d9892c2c96f
2013-02-01 18:35:28 +01:00
second_leftmv.as_int = left_block_second_mv(xd, mi, k);
second_abovemv.as_int = above_block_second_mv(mi, k, mis);
}
mv_contz = vp9_mv_cont(&leftmv, &abovemv);
blockmode = sub_mv_ref(bc, cm->fc.sub_mv_ref_prob [mv_contz]);
cm->fc.sub_mv_ref_counts[mv_contz][blockmode - LEFT4X4]++;
2010-05-18 17:58:33 +02:00
switch (blockmode) {
case NEW4X4:
read_nmv(bc, &blockmv.as_mv, &best_mv.as_mv, nmvc);
read_nmv_fp(bc, &blockmv.as_mv, &best_mv.as_mv, nmvc,
xd->allow_high_precision_mv);
vp9_increment_nmv(&blockmv.as_mv, &best_mv.as_mv,
&cm->fc.NMVcount, xd->allow_high_precision_mv);
blockmv.as_mv.row += best_mv.as_mv.row;
blockmv.as_mv.col += best_mv.as_mv.col;
if (mbmi->second_ref_frame > 0) {
read_nmv(bc, &secondmv.as_mv, &best_mv_second.as_mv, nmvc);
read_nmv_fp(bc, &secondmv.as_mv, &best_mv_second.as_mv, nmvc,
xd->allow_high_precision_mv);
vp9_increment_nmv(&secondmv.as_mv, &best_mv_second.as_mv,
&cm->fc.NMVcount, xd->allow_high_precision_mv);
secondmv.as_mv.row += best_mv_second.as_mv.row;
secondmv.as_mv.col += best_mv_second.as_mv.col;
}
#ifdef VPX_MODE_COUNT
vp9_mv_cont_count[mv_contz][3]++;
#endif
break;
case LEFT4X4:
blockmv.as_int = leftmv.as_int;
if (mbmi->second_ref_frame > 0)
secondmv.as_int = second_leftmv.as_int;
#ifdef VPX_MODE_COUNT
vp9_mv_cont_count[mv_contz][0]++;
#endif
break;
case ABOVE4X4:
blockmv.as_int = abovemv.as_int;
if (mbmi->second_ref_frame > 0)
secondmv.as_int = second_abovemv.as_int;
#ifdef VPX_MODE_COUNT
vp9_mv_cont_count[mv_contz][1]++;
#endif
break;
case ZERO4X4:
blockmv.as_int = 0;
if (mbmi->second_ref_frame > 0)
secondmv.as_int = 0;
#ifdef VPX_MODE_COUNT
vp9_mv_cont_count[mv_contz][2]++;
#endif
break;
default:
break;
}
/* Commenting this section out, not sure why this was needed, and
* there are mismatches with this section in rare cases since it is
* not done in the encoder at all.
mbmi->need_to_clamp_mvs |= check_mv_bounds(&blockmv,
mb_to_left_edge,
mb_to_right_edge,
mb_to_top_edge,
mb_to_bottom_edge);
if (mbmi->second_ref_frame > 0) {
mbmi->need_to_clamp_mvs |= check_mv_bounds(&secondmv,
mb_to_left_edge,
mb_to_right_edge,
mb_to_top_edge,
mb_to_bottom_edge);
}
*/
{
/* Fill (uniform) modes, mvs of jth subset.
Must do it here because ensuing subsets can
refer back to us via "left" or "above". */
const unsigned char *fill_offset;
unsigned int fill_count = mbsplit_fill_count[s];
fill_offset = &mbsplit_fill_offset[s][(unsigned char)j * mbsplit_fill_count[s]];
do {
mi->bmi[ *fill_offset].as_mv.first.as_int = blockmv.as_int;
if (mbmi->second_ref_frame > 0)
mi->bmi[ *fill_offset].as_mv.second.as_int = secondmv.as_int;
fill_offset++;
} while (--fill_count);
}
} while (++j < num_p);
}
mv->as_int = mi->bmi[15].as_mv.first.as_int;
mbmi->mv[1].as_int = mi->bmi[15].as_mv.second.as_int;
break; /* done with SPLITMV */
case NEARMV:
mv->as_int = nearby.as_int;
/* Clip "next_nearest" so that it does not extend to far out of image */
clamp_mv(mv, mb_to_left_edge, mb_to_right_edge,
mb_to_top_edge, mb_to_bottom_edge);
if (mbmi->second_ref_frame > 0) {
mbmi->mv[1].as_int = nearby_second.as_int;
clamp_mv(&mbmi->mv[1], mb_to_left_edge, mb_to_right_edge,
mb_to_top_edge, mb_to_bottom_edge);
}
break;
case NEARESTMV:
mv->as_int = nearest.as_int;
/* Clip "next_nearest" so that it does not extend to far out of image */
clamp_mv(mv, mb_to_left_edge, mb_to_right_edge,
mb_to_top_edge, mb_to_bottom_edge);
if (mbmi->second_ref_frame > 0) {
mbmi->mv[1].as_int = nearest_second.as_int;
clamp_mv(&mbmi->mv[1], mb_to_left_edge, mb_to_right_edge,
mb_to_top_edge, mb_to_bottom_edge);
}
break;
case ZEROMV:
mv->as_int = 0;
if (mbmi->second_ref_frame > 0)
mbmi->mv[1].as_int = 0;
break;
2010-05-18 17:58:33 +02:00
case NEWMV:
read_nmv(bc, &mv->as_mv, &best_mv.as_mv, nmvc);
read_nmv_fp(bc, &mv->as_mv, &best_mv.as_mv, nmvc,
xd->allow_high_precision_mv);
vp9_increment_nmv(&mv->as_mv, &best_mv.as_mv, &cm->fc.NMVcount,
xd->allow_high_precision_mv);
mv->as_mv.row += best_mv.as_mv.row;
mv->as_mv.col += best_mv.as_mv.col;
/* Don't need to check this on NEARMV and NEARESTMV modes
* since those modes clamp the MV. The NEWMV mode does not,
* so signal to the prediction stage whether special
* handling may be required.
*/
mbmi->need_to_clamp_mvs = check_mv_bounds(mv,
mb_to_left_edge,
mb_to_right_edge,
mb_to_top_edge,
mb_to_bottom_edge);
if (mbmi->second_ref_frame > 0) {
read_nmv(bc, &mbmi->mv[1].as_mv, &best_mv_second.as_mv, nmvc);
read_nmv_fp(bc, &mbmi->mv[1].as_mv, &best_mv_second.as_mv, nmvc,
xd->allow_high_precision_mv);
vp9_increment_nmv(&mbmi->mv[1].as_mv, &best_mv_second.as_mv,
&cm->fc.NMVcount, xd->allow_high_precision_mv);
mbmi->mv[1].as_mv.row += best_mv_second.as_mv.row;
mbmi->mv[1].as_mv.col += best_mv_second.as_mv.col;
mbmi->need_to_clamp_secondmv |=
check_mv_bounds(&mbmi->mv[1],
mb_to_left_edge, mb_to_right_edge,
mb_to_top_edge, mb_to_bottom_edge);
}
break;
default:
;
#if CONFIG_DEBUG
assert(0);
#endif
}
} else {
/* required for left and above block mv */
mbmi->mv[0].as_int = 0;
if (mbmi->sb_type) {
mbmi->mode = (MB_PREDICTION_MODE)
read_sb_ymode(bc, pbi->common.fc.sb_ymode_prob);
pbi->common.fc.sb_ymode_counts[mbmi->mode]++;
} else {
mbmi->mode = (MB_PREDICTION_MODE)
read_ymode(bc, pbi->common.fc.ymode_prob);
pbi->common.fc.ymode_counts[mbmi->mode]++;
}
// If MB mode is BPRED read the block modes
if (mbmi->mode == B_PRED) {
int j = 0;
do {
int m;
m = mi->bmi[j].as_mode.first = (B_PREDICTION_MODE)
read_bmode(bc, pbi->common.fc.bmode_prob);
#if CONFIG_NEWBINTRAMODES
if (m == B_CONTEXT_PRED) m -= CONTEXT_PRED_REPLACEMENTS;
#endif
pbi->common.fc.bmode_counts[m]++;
} while (++j < 16);
}
2010-05-18 17:58:33 +02:00
if (mbmi->mode == I8X8_PRED) {
int i;
int mode8x8;
for (i = 0; i < 4; i++) {
int ib = vp9_i8x8_block[i];
mode8x8 = read_i8x8_mode(bc, pbi->common.fc.i8x8_mode_prob);
mi->bmi[ib + 0].as_mode.first = mode8x8;
mi->bmi[ib + 1].as_mode.first = mode8x8;
mi->bmi[ib + 4].as_mode.first = mode8x8;
mi->bmi[ib + 5].as_mode.first = mode8x8;
pbi->common.fc.i8x8_mode_counts[mode8x8]++;
}
} else {
mbmi->uv_mode = (MB_PREDICTION_MODE)read_uv_mode(
bc, pbi->common.fc.uv_mode_prob[mbmi->mode]);
pbi->common.fc.uv_mode_counts[mbmi->mode][mbmi->uv_mode]++;
}
}
/*
if (pbi->common.current_video_frame == 1)
printf("mode: %d skip: %d\n", mbmi->mode, mbmi->mb_skip_coeff);
*/
2010-05-18 17:58:33 +02:00
if (cm->txfm_mode == TX_MODE_SELECT && mbmi->mb_skip_coeff == 0 &&
((mbmi->ref_frame == INTRA_FRAME && mbmi->mode <= I8X8_PRED) ||
(mbmi->ref_frame != INTRA_FRAME && !(mbmi->mode == SPLITMV &&
mbmi->partitioning == PARTITIONING_4X4)))) {
// FIXME(rbultje) code ternary symbol once all experiments are merged
mbmi->txfm_size = vp9_read(bc, cm->prob_tx[0]);
if (mbmi->txfm_size != TX_4X4 && mbmi->mode != I8X8_PRED &&
32x32 transform for superblocks. This adds Debargha's DCT/DWT hybrid and a regular 32x32 DCT, and adds code all over the place to wrap that in the bitstream/encoder/decoder/RD. Some implementation notes (these probably need careful review): - token range is extended by 1 bit, since the value range out of this transform is [-16384,16383]. - the coefficients coming out of the FDCT are manually scaled back by 1 bit, or else they won't fit in int16_t (they are 17 bits). Because of this, the RD error scoring does not right-shift the MSE score by two (unlike for 4x4/8x8/16x16). - to compensate for this loss in precision, the quantizer is halved also. This is currently a little hacky. - FDCT and IDCT is double-only right now. Needs a fixed-point impl. - There are no default probabilities for the 32x32 transform yet; I'm simply using the 16x16 luma ones. A future commit will add newly generated probabilities for all transforms. - No ADST version. I don't think we'll add one for this level; if an ADST is desired, transform-size selection can scale back to 16x16 or lower, and use an ADST at that level. Additional notes specific to Debargha's DWT/DCT hybrid: - coefficient scale is different for the top/left 16x16 (DCT-over-DWT) block than for the rest (DWT pixel differences) of the block. Therefore, RD error scoring isn't easily scalable between coefficient and pixel domain. Thus, unfortunately, we need to compute the RD distortion in the pixel domain until we figure out how to scale these appropriately. Change-Id: I00386f20f35d7fabb19aba94c8162f8aee64ef2b
2012-12-07 23:45:05 +01:00
mbmi->mode != SPLITMV) {
mbmi->txfm_size += vp9_read(bc, cm->prob_tx[1]);
if (mbmi->sb_type && mbmi->txfm_size != TX_8X8)
32x32 transform for superblocks. This adds Debargha's DCT/DWT hybrid and a regular 32x32 DCT, and adds code all over the place to wrap that in the bitstream/encoder/decoder/RD. Some implementation notes (these probably need careful review): - token range is extended by 1 bit, since the value range out of this transform is [-16384,16383]. - the coefficients coming out of the FDCT are manually scaled back by 1 bit, or else they won't fit in int16_t (they are 17 bits). Because of this, the RD error scoring does not right-shift the MSE score by two (unlike for 4x4/8x8/16x16). - to compensate for this loss in precision, the quantizer is halved also. This is currently a little hacky. - FDCT and IDCT is double-only right now. Needs a fixed-point impl. - There are no default probabilities for the 32x32 transform yet; I'm simply using the 16x16 luma ones. A future commit will add newly generated probabilities for all transforms. - No ADST version. I don't think we'll add one for this level; if an ADST is desired, transform-size selection can scale back to 16x16 or lower, and use an ADST at that level. Additional notes specific to Debargha's DWT/DCT hybrid: - coefficient scale is different for the top/left 16x16 (DCT-over-DWT) block than for the rest (DWT pixel differences) of the block. Therefore, RD error scoring isn't easily scalable between coefficient and pixel domain. Thus, unfortunately, we need to compute the RD distortion in the pixel domain until we figure out how to scale these appropriately. Change-Id: I00386f20f35d7fabb19aba94c8162f8aee64ef2b
2012-12-07 23:45:05 +01:00
mbmi->txfm_size += vp9_read(bc, cm->prob_tx[2]);
}
} else if (mbmi->sb_type && cm->txfm_mode >= ALLOW_32X32) {
32x32 transform for superblocks. This adds Debargha's DCT/DWT hybrid and a regular 32x32 DCT, and adds code all over the place to wrap that in the bitstream/encoder/decoder/RD. Some implementation notes (these probably need careful review): - token range is extended by 1 bit, since the value range out of this transform is [-16384,16383]. - the coefficients coming out of the FDCT are manually scaled back by 1 bit, or else they won't fit in int16_t (they are 17 bits). Because of this, the RD error scoring does not right-shift the MSE score by two (unlike for 4x4/8x8/16x16). - to compensate for this loss in precision, the quantizer is halved also. This is currently a little hacky. - FDCT and IDCT is double-only right now. Needs a fixed-point impl. - There are no default probabilities for the 32x32 transform yet; I'm simply using the 16x16 luma ones. A future commit will add newly generated probabilities for all transforms. - No ADST version. I don't think we'll add one for this level; if an ADST is desired, transform-size selection can scale back to 16x16 or lower, and use an ADST at that level. Additional notes specific to Debargha's DWT/DCT hybrid: - coefficient scale is different for the top/left 16x16 (DCT-over-DWT) block than for the rest (DWT pixel differences) of the block. Therefore, RD error scoring isn't easily scalable between coefficient and pixel domain. Thus, unfortunately, we need to compute the RD distortion in the pixel domain until we figure out how to scale these appropriately. Change-Id: I00386f20f35d7fabb19aba94c8162f8aee64ef2b
2012-12-07 23:45:05 +01:00
mbmi->txfm_size = TX_32X32;
} else if (cm->txfm_mode >= ALLOW_16X16 &&
((mbmi->ref_frame == INTRA_FRAME && mbmi->mode <= TM_PRED) ||
(mbmi->ref_frame != INTRA_FRAME && mbmi->mode != SPLITMV))) {
mbmi->txfm_size = TX_16X16;
} else if (cm->txfm_mode >= ALLOW_8X8 &&
(!(mbmi->ref_frame == INTRA_FRAME && mbmi->mode == B_PRED) &&
!(mbmi->ref_frame != INTRA_FRAME && mbmi->mode == SPLITMV &&
mbmi->partitioning == PARTITIONING_4X4))) {
mbmi->txfm_size = TX_8X8;
} else {
mbmi->txfm_size = TX_4X4;
}
}
2010-05-18 17:58:33 +02:00
void vp9_decode_mode_mvs_init(VP9D_COMP* const pbi, BOOL_DECODER* const bc) {
VP9_COMMON *cm = &pbi->common;
vpx_memset(cm->mbskip_pred_probs, 0, sizeof(cm->mbskip_pred_probs));
if (pbi->common.mb_no_coeff_skip) {
int k;
for (k = 0; k < MBSKIP_CONTEXTS; ++k)
cm->mbskip_pred_probs[k] = (vp9_prob)vp9_read_literal(bc, 8);
}
mb_mode_mv_init(pbi, bc);
}
void vp9_decode_mb_mode_mv(VP9D_COMP* const pbi,
MACROBLOCKD* const xd,
int mb_row,
int mb_col,
BOOL_DECODER* const bc) {
MODE_INFO *mi = xd->mode_info_context;
MODE_INFO *prev_mi = xd->prev_mode_info_context;
if (pbi->common.frame_type == KEY_FRAME)
kfread_modes(pbi, mi, mb_row, mb_col, bc);
else
read_mb_modes_mv(pbi, mi, &mi->mbmi, prev_mi, mb_row, mb_col, bc);
}