Files
android_external_libvpx/vp9/decoder/vp9_decodframe.c
John Koleszar eb939f45b8 Spatial resamping of ZEROMV predictors
This patch allows coding frames using references of different
resolution, in ZEROMV mode. For compound prediction, either
reference may be scaled.

To test, I use the resize_test and enable WRITE_RECON_BUFFER
in vp9_onyxd_if.c. It's also useful to apply this patch to
test/i420_video_source.h:

  --- a/test/i420_video_source.h
  +++ b/test/i420_video_source.h
  @@ -93,6 +93,7 @@ class I420VideoSource : public VideoSource {

     virtual void FillFrame() {
       // Read a frame from input_file.
  +    if (frame_ != 3)
       if (fread(img_->img_data, raw_sz_, 1, input_file_) == 0) {
         limit_ = frame_;
       }

This forces the frame that the resolution changes on to be coded
with no motion, only scaling, and improves the quality of the
result.

Change-Id: I1ee75d19a437ff801192f767fd02a36bcbd1d496
2013-02-26 23:54:23 -08:00

1821 lines
63 KiB
C

/*
* 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_common.h"
#include "vp9/common/vp9_header.h"
#include "vp9/common/vp9_reconintra.h"
#include "vp9/common/vp9_reconinter.h"
#include "vp9/common/vp9_entropy.h"
#include "vp9/decoder/vp9_decodframe.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 "vpx_scale/vpx_scale.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 "vpx_mem/vpx_mem.h"
#include "vp9/decoder/vp9_dboolhuff.h"
#include "vp9/common/vp9_seg_common.h"
#include "vp9/common/vp9_tile_common.h"
#include "vp9_rtcd.h"
#include <assert.h>
#include <stdio.h>
#define COEFCOUNT_TESTING
//#define DEC_DEBUG
#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;
int i;
v = merge_index(v, n - 1, modulus);
if ((m << 1) <= n) {
i = vp9_inv_recenter_nonneg(v + 1, m);
} else {
i = n - 1 - vp9_inv_recenter_nonneg(v + 1, n - 1 - m);
}
return i;
}
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) {
int i;
int q;
VP9_COMMON *const pc = &pbi->common;
for (q = 0; q < QINDEX_RANGE; q++) {
pc->Y1dequant[q][0] = (int16_t)vp9_dc_quant(q, pc->y1dc_delta_q);
pc->UVdequant[q][0] = (int16_t)vp9_dc_uv_quant(q, pc->uvdc_delta_q);
/* all the ac values =; */
for (i = 1; i < 16; i++) {
int rc = vp9_default_zig_zag1d_4x4[i];
pc->Y1dequant[q][rc] = (int16_t)vp9_ac_yquant(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) {
int i;
int qindex;
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)) {
if (xd->mb_segment_abs_delta == SEGMENT_ABSDATA)
/* Abs Value */
qindex = vp9_get_segdata(xd, segment_id, SEG_LVL_ALT_Q);
else {
/* Delta Value */
qindex = pc->base_qindex +
vp9_get_segdata(xd, segment_id, SEG_LVL_ALT_Q);
/* Clamp to valid range */
qindex = (qindex >= 0) ? ((qindex <= MAXQ) ? qindex : MAXQ) : 0;
}
} else
qindex = pc->base_qindex;
xd->q_index = qindex;
/* Set up the block level dequant pointers */
for (i = 0; i < 16; i++) {
xd->block[i].dequant = pc->Y1dequant[qindex];
}
xd->inv_txm4x4_1 = vp9_short_idct4x4llm_1;
xd->inv_txm4x4 = vp9_short_idct4x4llm;
xd->itxm_add = vp9_dequant_idct_add;
xd->dc_only_itxm_add = vp9_dc_only_idct_add_c;
xd->itxm_add_y_block = vp9_dequant_idct_add_y_block;
xd->itxm_add_uv_block = vp9_dequant_idct_add_uv_block;
if (xd->lossless) {
assert(qindex == 0);
xd->inv_txm4x4_1 = vp9_short_inv_walsh4x4_1_x8;
xd->inv_txm4x4 = vp9_short_inv_walsh4x4_x8;
xd->itxm_add = vp9_dequant_idct_add_lossless_c;
xd->dc_only_itxm_add = vp9_dc_only_inv_walsh_add_c;
xd->itxm_add_y_block = vp9_dequant_idct_add_y_block_lossless_c;
xd->itxm_add_uv_block = vp9_dequant_idct_add_uv_block_lossless_c;
}
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,
int mb_row, int mb_col) {
BLOCK_SIZE_TYPE sb_type = xd->mode_info_context->mbmi.sb_type;
if (xd->mode_info_context->mbmi.ref_frame == INTRA_FRAME) {
if (sb_type == BLOCK_SIZE_SB64X64) {
vp9_build_intra_predictors_sb64uv_s(xd);
vp9_build_intra_predictors_sb64y_s(xd);
} else if (sb_type == BLOCK_SIZE_SB32X32) {
vp9_build_intra_predictors_sbuv_s(xd);
vp9_build_intra_predictors_sby_s(xd);
} else {
vp9_build_intra_predictors_mbuv_s(xd);
vp9_build_intra_predictors_mby_s(xd);
}
} else {
if (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,
mb_row, mb_col);
} else if (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,
mb_row, mb_col);
} else {
vp9_build_inter16x16_predictors_mb(xd,
xd->dst.y_buffer,
xd->dst.u_buffer,
xd->dst.v_buffer,
xd->dst.y_stride,
xd->dst.uv_stride,
mb_row, mb_col);
#if CONFIG_COMP_INTERINTRA_PRED
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);
}
#endif
}
}
}
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);
#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->block[0].eob);
} else {
vp9_dequant_idct_add_16x16(xd->qcoeff, xd->block[0].dequant,
xd->predictor, xd->dst.y_buffer,
16, xd->dst.y_stride, xd->block[0].eob);
}
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);
}
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;
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(xd, 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->block[idx].eob);
} else {
vp9_dequant_idct_add_8x8_c(q, dq, pre, dst, 16, stride,
xd->block[idx].eob);
}
}
} else {
vp9_dequant_idct_add_y_block_8x8(xd->qcoeff,
xd->block[0].dequant,
xd->predictor,
xd->dst.y_buffer,
xd->dst.y_stride,
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, &xd->block[16 + i], i8x8mode, b->predictor);
xd->itxm_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, &xd->block[20 + i], i8x8mode, b->predictor);
xd->itxm_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) {
xd->itxm_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);
} 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);
}
#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) {
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(xd, 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,
b->dst_stride, b->eob);
} else {
xd->itxm_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(xd, b, i8x8mode, b->predictor);
xd->itxm_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, b, i8x8mode, b->predictor);
xd->itxm_add(b->qcoeff, b->dequant, b->predictor,
*(b->base_dst) + b->dst, 8, b->dst_stride);
}
} else if (mode == B_PRED) {
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(xd, 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 {
xd->itxm_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);
}
vp9_build_intra_predictors_mbuv(xd);
xd->itxm_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);
} else if (mode == SPLITMV || get_tx_type_4x4(xd, &xd->block[0]) == DCT_DCT) {
xd->itxm_add_y_block(xd->qcoeff,
xd->block[0].dequant,
xd->predictor,
xd->dst.y_buffer,
xd->dst.y_stride,
xd);
xd->itxm_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);
} 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
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,
b->dst_stride, b->eob);
} else {
xd->itxm_add(b->qcoeff, b->dequant, b->predictor,
*(b->base_dst) + b->dst, 16, b->dst_stride);
}
}
xd->itxm_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);
}
}
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->block[0].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);
};
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;
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,
stride, stride, b->eob);
} 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, b->eob);
}
}
} else {
vp9_dequant_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);
}
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);
};
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;
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,
xd->dst.y_stride, xd->dst.y_stride, b->eob);
} else {
xd->itxm_add(
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_dequant_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);
}
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);
};
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;
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, mb_row, mb_col);
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,
mb_row, mb_col);
}
/* 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->block[0].eob);
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);
}
}
} 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 < 24; i++) {
xd->block[i].eob = 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);
// 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, mb_row, mb_col);
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,
mb_row, mb_col);
}
/* 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->block[0].eob);
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);
}
} else {
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 < 24; i++) {
xd->block[i].eob = 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,
int mb_row, unsigned int mb_col,
BOOL_DECODER* const bc) {
int eobtotal = 0;
MB_PREDICTION_MODE mode;
int i;
int tx_size;
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;
if (xd->mode_info_context->mbmi.mb_skip_coeff) {
vp9_reset_mb_tokens_context(xd);
} else if (!bool_error(bc)) {
for (i = 0; i < 24; i++) {
xd->block[i].eob = 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,
&pbi->common);
if (eobtotal == 0 &&
mode != B_PRED &&
mode != SPLITMV &&
mode != I8X8_PRED &&
!bool_error(bc)) {
/* 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, mb_row, mb_col);
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);
if (mode != B_PRED) {
vp9_build_intra_predictors_mby(xd);
}
}
} else {
#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, mb_row, mb_col);
}
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 (i = 0; i < 16; i++) {
for (j = 0; j < 16; j++)
printf("%3d ", xd->dst.y_buffer[i * xd->dst.y_stride + j]);
printf("\n");
}
printf("\n");
printf("final u\n");
for (i = 0; i < 8; i++) {
for (j = 0; j < 8; j++)
printf("%3d ", xd->dst.u_buffer[i * xd->dst.uv_stride + j]);
printf("\n");
}
printf("\n");
printf("final v\n");
for (i = 0; i < 8; i++) {
for (j = 0; j < 8; j++)
printf("%3d ", xd->dst.v_buffer[i * xd->dst.uv_stride + j]);
printf("\n");
}
fflush(stdout);
}
#endif
}
static int get_delta_q(vp9_reader *bc, int prev, int *q_update) {
int ret_val = 0;
if (vp9_read_bit(bc)) {
ret_val = vp9_read_literal(bc, 4);
if (vp9_read_bit(bc))
ret_val = -ret_val;
}
/* Trigger a quantizer update if the delta-q value has changed */
if (ret_val != prev)
*q_update = 1;
return ret_val;
}
#ifdef PACKET_TESTING
#include <stdio.h>
FILE *vpxlog = 0;
#endif
static void set_offsets(VP9D_COMP *pbi, int block_size,
int mb_row, int mb_col) {
VP9_COMMON *const cm = &pbi->common;
MACROBLOCKD *const xd = &pbi->mb;
const int mis = cm->mode_info_stride;
const int idx = mis * mb_row + mb_col;
const int dst_fb_idx = cm->new_fb_idx;
const int recon_y_stride = cm->yv12_fb[dst_fb_idx].y_stride;
const int recon_uv_stride = cm->yv12_fb[dst_fb_idx].uv_stride;
const int recon_yoffset = mb_row * 16 * recon_y_stride + 16 * mb_col;
const int recon_uvoffset = mb_row * 8 * recon_uv_stride + 8 * mb_col;
xd->mode_info_context = cm->mi + idx;
xd->mode_info_context->mbmi.sb_type = block_size >> 5;
xd->prev_mode_info_context = cm->prev_mi + idx;
xd->above_context = cm->above_context + mb_col;
xd->left_context = cm->left_context + (mb_row & 3);
/* Distance of Mb to the various image edges.
* These are specified to 8th pel as they are always compared to
* values that are in 1/8th pel units
*/
block_size >>= 4; // in mb units
xd->mb_to_top_edge = -((mb_row * 16)) << 3;
xd->mb_to_left_edge = -((mb_col * 16) << 3);
xd->mb_to_bottom_edge = ((cm->mb_rows - block_size - mb_row) * 16) << 3;
xd->mb_to_right_edge = ((cm->mb_cols - block_size - mb_col) * 16) << 3;
xd->up_available = (mb_row != 0);
xd->left_available = (mb_col > cm->cur_tile_mb_col_start);
xd->right_available = (mb_col + block_size < cm->cur_tile_mb_col_end);
xd->dst.y_buffer = cm->yv12_fb[dst_fb_idx].y_buffer + recon_yoffset;
xd->dst.u_buffer = cm->yv12_fb[dst_fb_idx].u_buffer + recon_uvoffset;
xd->dst.v_buffer = cm->yv12_fb[dst_fb_idx].v_buffer + recon_uvoffset;
}
static void set_refs(VP9D_COMP *pbi, int block_size,
int mb_row, int mb_col) {
VP9_COMMON *const cm = &pbi->common;
MACROBLOCKD *const xd = &pbi->mb;
MODE_INFO *mi = xd->mode_info_context;
MB_MODE_INFO *const mbmi = &mi->mbmi;
if (mbmi->ref_frame > INTRA_FRAME) {
int ref_fb_idx;
/* Select the appropriate reference frame for this MB */
ref_fb_idx = cm->active_ref_idx[mbmi->ref_frame - 1];
xd->scale_factor[0] = cm->active_ref_scale[mbmi->ref_frame - 1];
xd->scale_factor_uv[0] = cm->active_ref_scale[mbmi->ref_frame - 1];
setup_pred_block(&xd->pre, &cm->yv12_fb[ref_fb_idx], mb_row, mb_col,
&xd->scale_factor[0], &xd->scale_factor_uv[0]);
/* propagate errors from reference frames */
xd->corrupted |= cm->yv12_fb[ref_fb_idx].corrupted;
if (mbmi->second_ref_frame > INTRA_FRAME) {
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];
setup_pred_block(&xd->second_pre, &cm->yv12_fb[second_ref_fb_idx],
mb_row, mb_col,
&xd->scale_factor[1], &xd->scale_factor_uv[1]);
/* propagate errors from reference frames */
xd->corrupted |= cm->yv12_fb[second_ref_fb_idx].corrupted;
}
}
if (mbmi->sb_type) {
const int n_mbs = 1 << mbmi->sb_type;
const int y_mbs = MIN(n_mbs, cm->mb_rows - mb_row);
const int x_mbs = MIN(n_mbs, cm->mb_cols - mb_col);
const int mis = cm->mode_info_stride;
int x, y;
for (y = 0; y < y_mbs; y++) {
for (x = !y; x < x_mbs; x++) {
mi[y * mis + x] = *mi;
}
}
}
}
/* Decode a row of Superblocks (2x2 region of MBs) */
static void decode_sb_row(VP9D_COMP *pbi, VP9_COMMON *pc,
int mb_row, MACROBLOCKD *xd,
BOOL_DECODER* const bc) {
int mb_col;
// For a SB there are 2 left contexts, each pertaining to a MB row within
vpx_memset(pc->left_context, 0, sizeof(pc->left_context));
for (mb_col = pc->cur_tile_mb_col_start;
mb_col < pc->cur_tile_mb_col_end; mb_col += 4) {
if (vp9_read(bc, pc->sb64_coded)) {
set_offsets(pbi, 64, mb_row, mb_col);
vp9_decode_mb_mode_mv(pbi, xd, mb_row, mb_col, bc);
set_refs(pbi, 64, mb_row, mb_col);
decode_superblock64(pbi, xd, mb_row, mb_col, bc);
xd->corrupted |= bool_error(bc);
} else {
int j;
for (j = 0; j < 4; j++) {
const int x_idx_sb = (j & 1) << 1, y_idx_sb = j & 2;
if (mb_row + y_idx_sb >= pc->mb_rows ||
mb_col + x_idx_sb >= pc->mb_cols) {
// MB lies outside frame, skip on to next
continue;
}
xd->sb_index = j;
if (vp9_read(bc, pc->sb32_coded)) {
set_offsets(pbi, 32, mb_row + y_idx_sb, mb_col + x_idx_sb);
vp9_decode_mb_mode_mv(pbi,
xd, mb_row + y_idx_sb, mb_col + x_idx_sb, bc);
set_refs(pbi, 32, mb_row + y_idx_sb, mb_col + x_idx_sb);
decode_superblock32(pbi,
xd, mb_row + y_idx_sb, mb_col + x_idx_sb, bc);
xd->corrupted |= bool_error(bc);
} else {
int i;
// Process the 4 MBs within the SB in the order:
// top-left, top-right, bottom-left, bottom-right
for (i = 0; i < 4; i++) {
const int x_idx = x_idx_sb + (i & 1), y_idx = y_idx_sb + (i >> 1);
if (mb_row + y_idx >= pc->mb_rows ||
mb_col + x_idx >= pc->mb_cols) {
// MB lies outside frame, skip on to next
continue;
}
set_offsets(pbi, 16, mb_row + y_idx, mb_col + x_idx);
xd->mb_index = i;
vp9_decode_mb_mode_mv(pbi, xd, mb_row + y_idx, mb_col + x_idx, bc);
update_blockd_bmi(xd);
set_refs(pbi, 16, mb_row + y_idx, mb_col + x_idx);
decode_macroblock(pbi, xd, mb_row + y_idx, mb_col + x_idx, bc);
/* check if the boolean decoder has suffered an error */
xd->corrupted |= bool_error(bc);
}
}
}
}
}
}
static unsigned int read_partition_size(const unsigned char *cx_size) {
const unsigned int size =
cx_size[0] + (cx_size[1] << 8) + (cx_size[2] << 16);
return size;
}
static int read_is_valid(const unsigned char *start,
size_t len,
const unsigned char *end) {
return start + len > start && start + len <= end;
}
static void setup_token_decoder(VP9D_COMP *pbi,
const unsigned char *cx_data,
BOOL_DECODER* const bool_decoder) {
VP9_COMMON *pc = &pbi->common;
const unsigned char *user_data_end = pbi->Source + pbi->source_sz;
const unsigned char *partition;
ptrdiff_t partition_size;
ptrdiff_t bytes_left;
// Set up pointers to token partition
partition = cx_data;
bytes_left = user_data_end - partition;
partition_size = bytes_left;
/* Validate the calculated partition length. If the buffer
* described by the partition can't be fully read, then restrict
* it to the portion that can be (for EC mode) or throw an error.
*/
if (!read_is_valid(partition, partition_size, user_data_end)) {
vpx_internal_error(&pc->error, VPX_CODEC_CORRUPT_FRAME,
"Truncated packet or corrupt partition "
"%d length", 1);
}
if (vp9_start_decode(bool_decoder,
partition, (unsigned int)partition_size))
vpx_internal_error(&pc->error, VPX_CODEC_MEM_ERROR,
"Failed to allocate bool decoder %d", 1);
}
static void init_frame(VP9D_COMP *pbi) {
VP9_COMMON *const pc = &pbi->common;
MACROBLOCKD *const xd = &pbi->mb;
if (pc->frame_type == KEY_FRAME) {
vp9_setup_past_independence(pc, xd);
/* All buffers are implicitly updated on key frames. */
pbi->refresh_frame_flags = (1 << NUM_REF_FRAMES) - 1;
} else if (pc->error_resilient_mode) {
vp9_setup_past_independence(pc, xd);
}
if (pc->frame_type != KEY_FRAME) {
if (!pc->use_bilinear_mc_filter)
pc->mcomp_filter_type = EIGHTTAP;
else
pc->mcomp_filter_type = BILINEAR;
/* To enable choice of different interpolation filters */
vp9_setup_interp_filters(xd, pc->mcomp_filter_type, pc);
}
xd->mode_info_context = pc->mi;
xd->prev_mode_info_context = pc->prev_mi;
xd->frame_type = pc->frame_type;
xd->mode_info_context->mbmi.mode = DC_PRED;
xd->mode_info_stride = pc->mode_info_stride;
xd->corrupted = 0; /* init without corruption */
xd->fullpixel_mask = 0xffffffff;
if (pc->full_pixel)
xd->fullpixel_mask = 0xfffffff8;
}
static void read_coef_probs_common(BOOL_DECODER* const bc,
vp9_coeff_probs *coef_probs,
int block_types) {
int i, j, k, l, m;
if (vp9_read_bit(bc)) {
for (i = 0; i < block_types; i++) {
for (j = 0; j < REF_TYPES; j++) {
for (k = 0; k < COEF_BANDS; k++) {
for (l = 0; l < PREV_COEF_CONTEXTS; l++) {
if (l >= 3 && k == 0)
continue;
for (m = 0; m < ENTROPY_NODES; m++) {
vp9_prob *const p = coef_probs[i][j][k][l] + m;
if (vp9_read(bc, COEF_UPDATE_PROB)) {
*p = read_prob_diff_update(bc, *p);
}
}
}
}
}
}
}
}
static void read_coef_probs(VP9D_COMP *pbi, BOOL_DECODER* const bc) {
VP9_COMMON *const pc = &pbi->common;
read_coef_probs_common(bc, pc->fc.coef_probs_4x4, BLOCK_TYPES);
if (pbi->common.txfm_mode != ONLY_4X4) {
read_coef_probs_common(bc, pc->fc.coef_probs_8x8, BLOCK_TYPES);
}
if (pbi->common.txfm_mode > ALLOW_8X8) {
read_coef_probs_common(bc, pc->fc.coef_probs_16x16, BLOCK_TYPES);
}
if (pbi->common.txfm_mode > ALLOW_16X16) {
read_coef_probs_common(bc, pc->fc.coef_probs_32x32, BLOCK_TYPES_32X32);
}
}
static void update_frame_size(VP9D_COMP *pbi) {
VP9_COMMON *cm = &pbi->common;
/* our internal buffers are always multiples of 16 */
int width = (cm->Width + 15) & ~15;
int height = (cm->Height + 15) & ~15;
cm->mb_rows = height >> 4;
cm->mb_cols = width >> 4;
cm->MBs = cm->mb_rows * cm->mb_cols;
cm->mode_info_stride = cm->mb_cols + 1;
memset(cm->mip, 0,
(cm->mb_cols + 1) * (cm->mb_rows + 1) * sizeof(MODE_INFO));
vp9_update_mode_info_border(cm, cm->mip);
cm->mi = cm->mip + cm->mode_info_stride + 1;
cm->prev_mi = cm->prev_mip + cm->mode_info_stride + 1;
vp9_update_mode_info_in_image(cm, cm->mi);
}
int vp9_decode_frame(VP9D_COMP *pbi, const unsigned char **p_data_end) {
BOOL_DECODER header_bc, residual_bc;
VP9_COMMON *const pc = &pbi->common;
MACROBLOCKD *const xd = &pbi->mb;
const unsigned char *data = (const unsigned char *)pbi->Source;
const unsigned char *data_end = data + pbi->source_sz;
ptrdiff_t first_partition_length_in_bytes = 0;
int mb_row;
int i, j;
int corrupt_tokens = 0;
// printf("Decoding frame %d\n", pc->current_video_frame);
/* start with no corruption of current frame */
xd->corrupted = 0;
pc->yv12_fb[pc->new_fb_idx].corrupted = 0;
if (data_end - data < 3) {
vpx_internal_error(&pc->error, VPX_CODEC_CORRUPT_FRAME, "Truncated packet");
} else {
pc->last_frame_type = pc->frame_type;
pc->frame_type = (FRAME_TYPE)(data[0] & 1);
pc->version = (data[0] >> 1) & 7;
pc->show_frame = (data[0] >> 4) & 1;
first_partition_length_in_bytes =
(data[0] | (data[1] << 8) | (data[2] << 16)) >> 5;
if ((data + first_partition_length_in_bytes > data_end
|| data + first_partition_length_in_bytes < data))
vpx_internal_error(&pc->error, VPX_CODEC_CORRUPT_FRAME,
"Truncated packet or corrupt partition 0 length");
data += 3;
vp9_setup_version(pc);
if (pc->frame_type == KEY_FRAME) {
/* vet via sync code */
/* When error concealment is enabled we should only check the sync
* code if we have enough bits available
*/
if (data + 3 < data_end) {
if (data[0] != 0x9d || data[1] != 0x01 || data[2] != 0x2a)
vpx_internal_error(&pc->error, VPX_CODEC_UNSUP_BITSTREAM,
"Invalid frame sync code");
}
data += 3;
}
{
const int width = pc->Width;
const int height = pc->Height;
/* If error concealment is enabled we should only parse the new size
* if we have enough data. Otherwise we will end up with the wrong
* size.
*/
if (data + 4 < data_end) {
pc->Width = (data[0] | (data[1] << 8)) & 0x3fff;
pc->horiz_scale = data[1] >> 6;
pc->Height = (data[2] | (data[3] << 8)) & 0x3fff;
pc->vert_scale = data[3] >> 6;
}
data += 4;
if (width != pc->Width || height != pc->Height) {
if (pc->Width <= 0) {
pc->Width = width;
vpx_internal_error(&pc->error, VPX_CODEC_CORRUPT_FRAME,
"Invalid frame width");
}
if (pc->Height <= 0) {
pc->Height = height;
vpx_internal_error(&pc->error, VPX_CODEC_CORRUPT_FRAME,
"Invalid frame height");
}
if (!pbi->initial_width || !pbi->initial_height) {
if (vp9_alloc_frame_buffers(pc, pc->Width, pc->Height))
vpx_internal_error(&pc->error, VPX_CODEC_MEM_ERROR,
"Failed to allocate frame buffers");
pbi->initial_width = pc->Width;
pbi->initial_height = pc->Height;
}
if (pc->Width > pbi->initial_width) {
vpx_internal_error(&pc->error, VPX_CODEC_CORRUPT_FRAME,
"Frame width too large");
}
if (pc->Height > pbi->initial_height) {
vpx_internal_error(&pc->error, VPX_CODEC_CORRUPT_FRAME,
"Frame height too large");
}
update_frame_size(pbi);
}
}
}
if ((!pbi->decoded_key_frame && pc->frame_type != KEY_FRAME) ||
pc->Width == 0 || pc->Height == 0) {
return -1;
}
init_frame(pbi);
/* Reset the frame pointers to the current frame size */
vp8_yv12_realloc_frame_buffer(&pc->yv12_fb[pc->new_fb_idx],
pc->mb_cols * 16, pc->mb_rows * 16,
VP9BORDERINPIXELS);
if (vp9_start_decode(&header_bc, data,
(unsigned int)first_partition_length_in_bytes))
vpx_internal_error(&pc->error, VPX_CODEC_MEM_ERROR,
"Failed to allocate bool decoder 0");
pc->clr_type = (YUV_TYPE)vp9_read_bit(&header_bc);
pc->clamp_type = (CLAMP_TYPE)vp9_read_bit(&header_bc);
pc->error_resilient_mode = vp9_read_bit(&header_bc);
/* Is segmentation enabled */
xd->segmentation_enabled = (unsigned char)vp9_read_bit(&header_bc);
if (xd->segmentation_enabled) {
// Read whether or not the segmentation map is being explicitly
// updated this frame.
xd->update_mb_segmentation_map = (unsigned char)vp9_read_bit(&header_bc);
// If so what method will be used.
if (xd->update_mb_segmentation_map) {
// Which macro block level features are enabled
// Read the probs used to decode the segment id for each macro
// block.
for (i = 0; i < MB_FEATURE_TREE_PROBS; i++) {
xd->mb_segment_tree_probs[i] = vp9_read_bit(&header_bc) ?
(vp9_prob)vp9_read_literal(&header_bc, 8) : 255;
}
// Read the prediction probs needed to decode the segment id
pc->temporal_update = (unsigned char)vp9_read_bit(&header_bc);
for (i = 0; i < PREDICTION_PROBS; i++) {
if (pc->temporal_update) {
pc->segment_pred_probs[i] = vp9_read_bit(&header_bc) ?
(vp9_prob)vp9_read_literal(&header_bc, 8) : 255;
} else {
pc->segment_pred_probs[i] = 255;
}
}
if (pc->temporal_update) {
int count[4];
const vp9_prob *p = xd->mb_segment_tree_probs;
vp9_prob *p_mod = xd->mb_segment_mispred_tree_probs;
count[0] = p[0] * p[1];
count[1] = p[0] * (256 - p[1]);
count[2] = (256 - p[0]) * p[2];
count[3] = (256 - p[0]) * (256 - p[2]);
p_mod[0] = get_binary_prob(count[1], count[2] + count[3]);
p_mod[1] = get_binary_prob(count[0], count[2] + count[3]);
p_mod[2] = get_binary_prob(count[0] + count[1], count[3]);
p_mod[3] = get_binary_prob(count[0] + count[1], count[2]);
}
}
// Is the segment data being updated
xd->update_mb_segmentation_data = (unsigned char)vp9_read_bit(&header_bc);
if (xd->update_mb_segmentation_data) {
int data;
xd->mb_segment_abs_delta = (unsigned char)vp9_read_bit(&header_bc);
vp9_clearall_segfeatures(xd);
// For each segmentation...
for (i = 0; i < MAX_MB_SEGMENTS; i++) {
// For each of the segments features...
for (j = 0; j < SEG_LVL_MAX; j++) {
// Is the feature enabled
if (vp9_read_bit(&header_bc)) {
// Update the feature data and mask
vp9_enable_segfeature(xd, i, j);
data = vp9_decode_unsigned_max(&header_bc,
vp9_seg_feature_data_max(j));
// Is the segment data signed..
if (vp9_is_segfeature_signed(j)) {
if (vp9_read_bit(&header_bc))
data = -data;
}
} else
data = 0;
vp9_set_segdata(xd, i, j, data);
}
}
}
}
// Read common prediction model status flag probability updates for the
// reference frame
if (pc->frame_type == KEY_FRAME) {
// Set the prediction probabilities to defaults
pc->ref_pred_probs[0] = 120;
pc->ref_pred_probs[1] = 80;
pc->ref_pred_probs[2] = 40;
} else {
for (i = 0; i < PREDICTION_PROBS; i++) {
if (vp9_read_bit(&header_bc))
pc->ref_pred_probs[i] = (vp9_prob)vp9_read_literal(&header_bc, 8);
}
}
pc->sb64_coded = vp9_read_literal(&header_bc, 8);
pc->sb32_coded = vp9_read_literal(&header_bc, 8);
xd->lossless = vp9_read_bit(&header_bc);
if (xd->lossless) {
pc->txfm_mode = ONLY_4X4;
} else {
/* Read the loop filter level and type */
pc->txfm_mode = vp9_read_literal(&header_bc, 2);
if (pc->txfm_mode == 3)
pc->txfm_mode += vp9_read_bit(&header_bc);
if (pc->txfm_mode == TX_MODE_SELECT) {
pc->prob_tx[0] = vp9_read_literal(&header_bc, 8);
pc->prob_tx[1] = vp9_read_literal(&header_bc, 8);
pc->prob_tx[2] = vp9_read_literal(&header_bc, 8);
}
}
pc->filter_type = (LOOPFILTERTYPE) vp9_read_bit(&header_bc);
pc->filter_level = vp9_read_literal(&header_bc, 6);
pc->sharpness_level = vp9_read_literal(&header_bc, 3);
/* Read in loop filter deltas applied at the MB level based on mode or ref frame. */
xd->mode_ref_lf_delta_update = 0;
xd->mode_ref_lf_delta_enabled = (unsigned char)vp9_read_bit(&header_bc);
if (xd->mode_ref_lf_delta_enabled) {
/* Do the deltas need to be updated */
xd->mode_ref_lf_delta_update = (unsigned char)vp9_read_bit(&header_bc);
if (xd->mode_ref_lf_delta_update) {
/* Send update */
for (i = 0; i < MAX_REF_LF_DELTAS; i++) {
if (vp9_read_bit(&header_bc)) {
/*sign = vp9_read_bit( &header_bc );*/
xd->ref_lf_deltas[i] = (signed char)vp9_read_literal(&header_bc, 6);
if (vp9_read_bit(&header_bc)) /* Apply sign */
xd->ref_lf_deltas[i] = xd->ref_lf_deltas[i] * -1;
}
}
/* Send update */
for (i = 0; i < MAX_MODE_LF_DELTAS; i++) {
if (vp9_read_bit(&header_bc)) {
/*sign = vp9_read_bit( &header_bc );*/
xd->mode_lf_deltas[i] = (signed char)vp9_read_literal(&header_bc, 6);
if (vp9_read_bit(&header_bc)) /* Apply sign */
xd->mode_lf_deltas[i] = xd->mode_lf_deltas[i] * -1;
}
}
}
}
// Dummy read for now
vp9_read_literal(&header_bc, 2);
/* Read the default quantizers. */
{
int q_update = 0;
pc->base_qindex = vp9_read_literal(&header_bc, QINDEX_BITS);
/* AC 1st order Q = default */
pc->y1dc_delta_q = get_delta_q(&header_bc, pc->y1dc_delta_q, &q_update);
pc->uvdc_delta_q = get_delta_q(&header_bc, pc->uvdc_delta_q, &q_update);
pc->uvac_delta_q = get_delta_q(&header_bc, pc->uvac_delta_q, &q_update);
if (q_update)
vp9_init_de_quantizer(pbi);
/* MB level dequantizer setup */
mb_init_dequantizer(pbi, &pbi->mb);
}
/* Determine if the golden frame or ARF buffer should be updated and how.
* For all non key frames the GF and ARF refresh flags and sign bias
* flags must be set explicitly.
*/
if (pc->frame_type == KEY_FRAME) {
pc->active_ref_idx[0] = pc->new_fb_idx;
pc->active_ref_idx[1] = pc->new_fb_idx;
pc->active_ref_idx[2] = pc->new_fb_idx;
} else {
/* Should the GF or ARF be updated from the current frame */
pbi->refresh_frame_flags = vp9_read_literal(&header_bc, NUM_REF_FRAMES);
/* Select active reference frames */
for (i = 0; i < 3; i++) {
int ref_frame_num = vp9_read_literal(&header_bc, NUM_REF_FRAMES_LG2);
pc->active_ref_idx[i] = pc->ref_frame_map[ref_frame_num];
}
pc->ref_frame_sign_bias[GOLDEN_FRAME] = vp9_read_bit(&header_bc);
pc->ref_frame_sign_bias[ALTREF_FRAME] = vp9_read_bit(&header_bc);
/* Is high precision mv allowed */
xd->allow_high_precision_mv = (unsigned char)vp9_read_bit(&header_bc);
// Read the type of subpel filter to use
if (vp9_read_bit(&header_bc)) {
pc->mcomp_filter_type = SWITCHABLE;
} else {
pc->mcomp_filter_type = vp9_read_literal(&header_bc, 2);
}
#if CONFIG_COMP_INTERINTRA_PRED
pc->use_interintra = vp9_read_bit(&header_bc);
#endif
/* To enable choice of different interploation filters */
vp9_setup_interp_filters(xd, pc->mcomp_filter_type, pc);
}
if (!pc->error_resilient_mode) {
pc->refresh_entropy_probs = vp9_read_bit(&header_bc);
pc->frame_parallel_decoding_mode = vp9_read_bit(&header_bc);
} else {
pc->refresh_entropy_probs = 0;
pc->frame_parallel_decoding_mode = 1;
}
pc->frame_context_idx = vp9_read_literal(&header_bc, NUM_FRAME_CONTEXTS_LG2);
vpx_memcpy(&pc->fc, &pc->frame_contexts[pc->frame_context_idx],
sizeof(pc->fc));
// Read inter mode probability context updates
if (pc->frame_type != KEY_FRAME) {
int i, j;
for (i = 0; i < INTER_MODE_CONTEXTS; i++) {
for (j = 0; j < 4; j++) {
if (vp9_read(&header_bc, 252)) {
pc->fc.vp9_mode_contexts[i][j] =
(vp9_prob)vp9_read_literal(&header_bc, 8);
}
}
}
}
#if CONFIG_NEW_MVREF
// If Key frame reset mv ref id probabilities to defaults
if (pc->frame_type != KEY_FRAME) {
// Read any mv_ref index probability updates
int i, j;
for (i = 0; i < MAX_REF_FRAMES; ++i) {
// Skip the dummy entry for intra ref frame.
if (i == INTRA_FRAME) {
continue;
}
// Read any updates to probabilities
for (j = 0; j < MAX_MV_REF_CANDIDATES - 1; ++j) {
if (vp9_read(&header_bc, VP9_MVREF_UPDATE_PROB)) {
xd->mb_mv_ref_probs[i][j] =
(vp9_prob)vp9_read_literal(&header_bc, 8);
}
}
}
}
#endif
if (0) {
FILE *z = fopen("decodestats.stt", "a");
fprintf(z, "%6d F:%d,R:%d,Q:%d\n",
pc->current_video_frame,
pc->frame_type,
pbi->refresh_frame_flags,
pc->base_qindex);
fclose(z);
}
vp9_copy(pbi->common.fc.pre_coef_probs_4x4,
pbi->common.fc.coef_probs_4x4);
vp9_copy(pbi->common.fc.pre_coef_probs_8x8,
pbi->common.fc.coef_probs_8x8);
vp9_copy(pbi->common.fc.pre_coef_probs_16x16,
pbi->common.fc.coef_probs_16x16);
vp9_copy(pbi->common.fc.pre_coef_probs_32x32,
pbi->common.fc.coef_probs_32x32);
vp9_copy(pbi->common.fc.pre_ymode_prob, pbi->common.fc.ymode_prob);
vp9_copy(pbi->common.fc.pre_sb_ymode_prob, pbi->common.fc.sb_ymode_prob);
vp9_copy(pbi->common.fc.pre_uv_mode_prob, pbi->common.fc.uv_mode_prob);
vp9_copy(pbi->common.fc.pre_bmode_prob, pbi->common.fc.bmode_prob);
vp9_copy(pbi->common.fc.pre_i8x8_mode_prob, pbi->common.fc.i8x8_mode_prob);
vp9_copy(pbi->common.fc.pre_sub_mv_ref_prob, pbi->common.fc.sub_mv_ref_prob);
vp9_copy(pbi->common.fc.pre_mbsplit_prob, pbi->common.fc.mbsplit_prob);
#if CONFIG_COMP_INTERINTRA_PRED
pbi->common.fc.pre_interintra_prob = pbi->common.fc.interintra_prob;
#endif
pbi->common.fc.pre_nmvc = pbi->common.fc.nmvc;
vp9_zero(pbi->common.fc.coef_counts_4x4);
vp9_zero(pbi->common.fc.coef_counts_8x8);
vp9_zero(pbi->common.fc.coef_counts_16x16);
vp9_zero(pbi->common.fc.coef_counts_32x32);
vp9_zero(pbi->common.fc.ymode_counts);
vp9_zero(pbi->common.fc.sb_ymode_counts);
vp9_zero(pbi->common.fc.uv_mode_counts);
vp9_zero(pbi->common.fc.bmode_counts);
vp9_zero(pbi->common.fc.i8x8_mode_counts);
vp9_zero(pbi->common.fc.sub_mv_ref_counts);
vp9_zero(pbi->common.fc.mbsplit_counts);
vp9_zero(pbi->common.fc.NMVcount);
vp9_zero(pbi->common.fc.mv_ref_ct);
#if CONFIG_COMP_INTERINTRA_PRED
vp9_zero(pbi->common.fc.interintra_counts);
#endif
read_coef_probs(pbi, &header_bc);
/* Initialize xd pointers. Any reference should do for xd->pre, so use 0. */
vpx_memcpy(&xd->pre, &pc->yv12_fb[pc->active_ref_idx[0]],
sizeof(YV12_BUFFER_CONFIG));
vpx_memcpy(&xd->dst, &pc->yv12_fb[pc->new_fb_idx],
sizeof(YV12_BUFFER_CONFIG));
// Create the segmentation map structure and set to 0
if (!pc->last_frame_seg_map)
CHECK_MEM_ERROR(pc->last_frame_seg_map,
vpx_calloc((pc->mb_rows * pc->mb_cols), 1));
/* set up frame new frame for intra coded blocks */
vp9_setup_intra_recon(&pc->yv12_fb[pc->new_fb_idx]);
vp9_setup_block_dptrs(xd);
vp9_build_block_doffsets(xd);
/* clear out the coeff buffer */
vpx_memset(xd->qcoeff, 0, sizeof(xd->qcoeff));
/* Read the mb_no_coeff_skip flag */
pc->mb_no_coeff_skip = (int)vp9_read_bit(&header_bc);
vp9_decode_mode_mvs_init(pbi, &header_bc);
/* tile info */
{
const unsigned char *data_ptr = data + first_partition_length_in_bytes;
int tile_row, tile_col, delta_log2_tiles;
vp9_get_tile_n_bits(pc, &pc->log2_tile_columns, &delta_log2_tiles);
while (delta_log2_tiles--) {
if (vp9_read_bit(&header_bc)) {
pc->log2_tile_columns++;
} else {
break;
}
}
pc->log2_tile_rows = vp9_read_bit(&header_bc);
if (pc->log2_tile_rows)
pc->log2_tile_rows += vp9_read_bit(&header_bc);
pc->tile_columns = 1 << pc->log2_tile_columns;
pc->tile_rows = 1 << pc->log2_tile_rows;
vpx_memset(pc->above_context, 0,
sizeof(ENTROPY_CONTEXT_PLANES) * pc->mb_cols);
if (pbi->oxcf.inv_tile_order) {
const int n_cols = pc->tile_columns;
const unsigned char *data_ptr2[4][1 << 6];
BOOL_DECODER UNINITIALIZED_IS_SAFE(bc_bak);
// pre-initialize the offsets, we're going to read in inverse order
data_ptr2[0][0] = data_ptr;
for (tile_row = 0; tile_row < pc->tile_rows; tile_row++) {
if (tile_row) {
int size = data_ptr2[tile_row - 1][n_cols - 1][0] |
(data_ptr2[tile_row - 1][n_cols - 1][1] << 8) |
(data_ptr2[tile_row - 1][n_cols - 1][2] << 16) |
(data_ptr2[tile_row - 1][n_cols - 1][3] << 24);
data_ptr2[tile_row - 1][n_cols - 1] += 4;
data_ptr2[tile_row][0] = data_ptr2[tile_row - 1][n_cols - 1] + size;
}
for (tile_col = 1; tile_col < n_cols; tile_col++) {
int size = data_ptr2[tile_row][tile_col - 1][0] |
(data_ptr2[tile_row][tile_col - 1][1] << 8) |
(data_ptr2[tile_row][tile_col - 1][2] << 16) |
(data_ptr2[tile_row][tile_col - 1][3] << 24);
data_ptr2[tile_row][tile_col - 1] += 4;
data_ptr2[tile_row][tile_col] =
data_ptr2[tile_row][tile_col - 1] + size;
}
}
for (tile_row = 0; tile_row < pc->tile_rows; tile_row++) {
vp9_get_tile_row_offsets(pc, tile_row);
for (tile_col = n_cols - 1; tile_col >= 0; tile_col--) {
vp9_get_tile_col_offsets(pc, tile_col);
setup_token_decoder(pbi, data_ptr2[tile_row][tile_col], &residual_bc);
/* Decode a row of superblocks */
for (mb_row = pc->cur_tile_mb_row_start;
mb_row < pc->cur_tile_mb_row_end; mb_row += 4) {
decode_sb_row(pbi, pc, mb_row, xd, &residual_bc);
}
if (tile_row == pc->tile_rows - 1 && tile_col == n_cols - 1)
bc_bak = residual_bc;
}
}
residual_bc = bc_bak;
} else {
for (tile_row = 0; tile_row < pc->tile_rows; tile_row++) {
vp9_get_tile_row_offsets(pc, tile_row);
for (tile_col = 0; tile_col < pc->tile_columns; tile_col++) {
vp9_get_tile_col_offsets(pc, tile_col);
if (tile_col < pc->tile_columns - 1 || tile_row < pc->tile_rows - 1)
setup_token_decoder(pbi, data_ptr + 4, &residual_bc);
else
setup_token_decoder(pbi, data_ptr, &residual_bc);
/* Decode a row of superblocks */
for (mb_row = pc->cur_tile_mb_row_start;
mb_row < pc->cur_tile_mb_row_end; mb_row += 4) {
decode_sb_row(pbi, pc, mb_row, xd, &residual_bc);
}
if (tile_col < pc->tile_columns - 1 || tile_row < pc->tile_rows - 1) {
int size = data_ptr[0] |
(data_ptr[1] << 8) |
(data_ptr[2] << 16) |
(data_ptr[3] << 24);
data_ptr += 4 + size;
}
}
}
}
}
corrupt_tokens |= xd->corrupted;
/* Collect information about decoder corruption. */
/* 1. Check first boolean decoder for errors. */
pc->yv12_fb[pc->new_fb_idx].corrupted = bool_error(&header_bc);
/* 2. Check the macroblock information */
pc->yv12_fb[pc->new_fb_idx].corrupted |= corrupt_tokens;
if (!pbi->decoded_key_frame) {
if (pc->frame_type == KEY_FRAME &&
!pc->yv12_fb[pc->new_fb_idx].corrupted)
pbi->decoded_key_frame = 1;
else
vpx_internal_error(&pbi->common.error, VPX_CODEC_CORRUPT_FRAME,
"A stream must start with a complete key frame");
}
if (!pc->error_resilient_mode &&
!pc->frame_parallel_decoding_mode)
vp9_adapt_coef_probs(pc);
if (pc->frame_type != KEY_FRAME) {
if (!pc->error_resilient_mode &&
!pc->frame_parallel_decoding_mode) {
vp9_adapt_mode_probs(pc);
vp9_adapt_nmv_probs(pc, xd->allow_high_precision_mv);
vp9_adapt_mode_context(&pbi->common);
}
}
if (pc->refresh_entropy_probs) {
vpx_memcpy(&pc->frame_contexts[pc->frame_context_idx], &pc->fc,
sizeof(pc->fc));
}
#ifdef PACKET_TESTING
{
FILE *f = fopen("decompressor.VP8", "ab");
unsigned int size = residual_bc.pos + header_bc.pos + 8;
fwrite((void *) &size, 4, 1, f);
fwrite((void *) pbi->Source, size, 1, f);
fclose(f);
}
#endif
/* Find the end of the coded buffer */
while (residual_bc.count > CHAR_BIT
&& residual_bc.count < VP9_BD_VALUE_SIZE) {
residual_bc.count -= CHAR_BIT;
residual_bc.user_buffer--;
}
*p_data_end = residual_bc.user_buffer;
return 0;
}