generic-library/ffmpeg
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

vp8: K&R formatting cosmetics

Browse Source 
Signed-off-by: Diego Biurrun <diego@biurrun.de>
This commit is contained in:
Vittorio Giovara 2014-03-28 01:13:54 +01:00 committed by Diego Biurrun
parent 6adf3bc42e
commit 53c20f17c7
6 changed files with 926 additions and 799 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

385
libavcodec/vp8.c
View File

@ -24,12 +24,13 @@
*/ */
#include "libavutil/imgutils.h" #include "libavutil/imgutils.h"
#include "avcodec.h" #include "avcodec.h"
#include "internal.h" #include "internal.h"
#include "vp8.h"
#include "vp8data.h"
#include "rectangle.h" #include "rectangle.h"
#include "thread.h" #include "thread.h"
#include "vp8.h"
#include "vp8data.h"
#if ARCH_ARM #if ARCH_ARM
# include "arm/vp8.h" # include "arm/vp8.h"
@ -91,7 +92,6 @@ static int vp8_ref_frame(VP8Context *s, VP8Frame *dst, VP8Frame *src)
return 0; return 0;
} }
static void vp8_decode_flush_impl(AVCodecContext *avctx, int free_mem) static void vp8_decode_flush_impl(AVCodecContext *avctx, int free_mem)
{ {
VP8Context *s = avctx->priv_data; VP8Context *s = avctx->priv_data;
@ -127,19 +127,22 @@ static int update_dimensions(VP8Context *s, int width, int height)
s->mb_width = (s->avctx->coded_width + 15) / 16; s->mb_width = (s->avctx->coded_width + 15) / 16;
s->mb_height = (s->avctx->coded_height + 15) / 16; s->mb_height = (s->avctx->coded_height + 15) / 16;
s->mb_layout = (avctx->active_thread_type == FF_THREAD_SLICE) && (FFMIN(s->num_coeff_partitions, avctx->thread_count) > 1); s->mb_layout = (avctx->active_thread_type == FF_THREAD_SLICE) &&
(FFMIN(s->num_coeff_partitions, avctx->thread_count) > 1);
if (!s->mb_layout) { // Frame threading and one thread if (!s->mb_layout) { // Frame threading and one thread
s->macroblocks_base = av_mallocz((s->mb_width+s->mb_height*2+1)*sizeof(*s->macroblocks)); s->macroblocks_base = av_mallocz((s->mb_width + s->mb_height * 2 + 1) *
sizeof(*s->macroblocks));
s->intra4x4_pred_mode_top = av_mallocz(s->mb_width * 4); s->intra4x4_pred_mode_top = av_mallocz(s->mb_width * 4);
} } else // Sliced threading
else // Sliced threading s->macroblocks_base = av_mallocz((s->mb_width + 2) * (s->mb_height + 2) *
s->macroblocks_base = av_mallocz((s->mb_width+2)*(s->mb_height+2)*sizeof(*s->macroblocks)); sizeof(*s->macroblocks));
s->top_nnz = av_mallocz(s->mb_width * sizeof(*s->top_nnz)); s->top_nnz = av_mallocz(s->mb_width * sizeof(*s->top_nnz));
s->top_border = av_mallocz((s->mb_width + 1) * sizeof(*s->top_border)); s->top_border = av_mallocz((s->mb_width + 1) * sizeof(*s->top_border));
s->thread_data = av_mallocz(MAX_THREADS * sizeof(VP8ThreadData)); s->thread_data = av_mallocz(MAX_THREADS * sizeof(VP8ThreadData));
for (i = 0; i < MAX_THREADS; i++) { for (i = 0; i < MAX_THREADS; i++) {
s->thread_data[i].filter_strength = av_mallocz(s->mb_width*sizeof(*s->thread_data[0].filter_strength)); s->thread_data[i].filter_strength =
av_mallocz(s->mb_width * sizeof(*s->thread_data[0].filter_strength));
#if HAVE_THREADS #if HAVE_THREADS
pthread_mutex_init(&s->thread_data[i].lock, NULL); pthread_mutex_init(&s->thread_data[i].lock, NULL);
pthread_cond_init(&s->thread_data[i].cond, NULL); pthread_cond_init(&s->thread_data[i].cond, NULL);
@ -248,9 +251,9 @@ static void get_quants(VP8Context *s)
s->qmat[i].luma_qmul[0] = vp8_dc_qlookup[av_clip_uintp2(base_qi + ydc_delta, 7)]; s->qmat[i].luma_qmul[0] = vp8_dc_qlookup[av_clip_uintp2(base_qi + ydc_delta, 7)];
s->qmat[i].luma_qmul[1] = vp8_ac_qlookup[av_clip_uintp2(base_qi, 7)]; s->qmat[i].luma_qmul[1] = vp8_ac_qlookup[av_clip_uintp2(base_qi, 7)];
s->qmat[i].luma_dc_qmul[0] = 2 * vp8_dc_qlookup[av_clip_uintp2(base_qi + y2dc_delta, 7)]; s->qmat[i].luma_dc_qmul[0] = vp8_dc_qlookup[av_clip_uintp2(base_qi + y2dc_delta, 7)] * 2;
/* 101581>>16 is equivalent to 155/100 */ /* 101581>>16 is equivalent to 155/100 */
s->qmat[i].luma_dc_qmul[1] = (101581 * vp8_ac_qlookup[av_clip_uintp2(base_qi + y2ac_delta, 7)]) >> 16; s->qmat[i].luma_dc_qmul[1] = vp8_ac_qlookup[av_clip_uintp2(base_qi + y2ac_delta, 7)] * 101581 >> 16;
s->qmat[i].chroma_qmul[0] = vp8_dc_qlookup[av_clip_uintp2(base_qi + uvdc_delta, 7)]; s->qmat[i].chroma_qmul[0] = vp8_dc_qlookup[av_clip_uintp2(base_qi + uvdc_delta, 7)];
s->qmat[i].chroma_qmul[1] = vp8_ac_qlookup[av_clip_uintp2(base_qi + uvac_delta, 7)]; s->qmat[i].chroma_qmul[1] = vp8_ac_qlookup[av_clip_uintp2(base_qi + uvac_delta, 7)];
@ -317,9 +320,11 @@ static int decode_frame_header(VP8Context *s, const uint8_t *buf, int buf_size)
av_log(s->avctx, AV_LOG_WARNING, "Unknown profile %d\n", s->profile); av_log(s->avctx, AV_LOG_WARNING, "Unknown profile %d\n", s->profile);
if (!s->profile) if (!s->profile)
memcpy(s->put_pixels_tab, s->vp8dsp.put_vp8_epel_pixels_tab, sizeof(s->put_pixels_tab)); memcpy(s->put_pixels_tab, s->vp8dsp.put_vp8_epel_pixels_tab,
sizeof(s->put_pixels_tab));
else // profile 1-3 use bilinear, 4+ aren't defined so whatever else // profile 1-3 use bilinear, 4+ aren't defined so whatever
memcpy(s->put_pixels_tab, s->vp8dsp.put_vp8_bilinear_pixels_tab, sizeof(s->put_pixels_tab)); memcpy(s->put_pixels_tab, s->vp8dsp.put_vp8_bilinear_pixels_tab,
sizeof(s->put_pixels_tab));
if (header_size > buf_size - 7 * s->keyframe) { if (header_size > buf_size - 7 * s->keyframe) {
av_log(s->avctx, AV_LOG_ERROR, "Header size larger than data provided\n"); av_log(s->avctx, AV_LOG_ERROR, "Header size larger than data provided\n");
@ -328,7 +333,8 @@ static int decode_frame_header(VP8Context *s, const uint8_t *buf, int buf_size)
if (s->keyframe) { if (s->keyframe) {
if (AV_RL24(buf) != 0x2a019d) { if (AV_RL24(buf) != 0x2a019d) {
av_log(s->avctx, AV_LOG_ERROR, "Invalid start code 0x%x\n", AV_RL24(buf)); av_log(s->avctx, AV_LOG_ERROR,
"Invalid start code 0x%x\n", AV_RL24(buf));
return AVERROR_INVALIDDATA; return AVERROR_INVALIDDATA;
} }
width = AV_RL16(buf + 3) & 0x3fff; width = AV_RL16(buf + 3) & 0x3fff;
@ -344,11 +350,15 @@ static int decode_frame_header(VP8Context *s, const uint8_t *buf, int buf_size)
s->update_golden = s->update_altref = VP56_FRAME_CURRENT; s->update_golden = s->update_altref = VP56_FRAME_CURRENT;
for (i = 0; i < 4; i++) for (i = 0; i < 4; i++)
for (j = 0; j < 16; j++) for (j = 0; j < 16; j++)
memcpy(s->prob->token[i][j], vp8_token_default_probs[i][vp8_coeff_band[j]], memcpy(s->prob->token[i][j],
vp8_token_default_probs[i][vp8_coeff_band[j]],
sizeof(s->prob->token[i][j])); sizeof(s->prob->token[i][j]));
memcpy(s->prob->pred16x16, vp8_pred16x16_prob_inter, sizeof(s->prob->pred16x16)); memcpy(s->prob->pred16x16, vp8_pred16x16_prob_inter,
memcpy(s->prob->pred8x8c , vp8_pred8x8c_prob_inter , sizeof(s->prob->pred8x8c)); sizeof(s->prob->pred16x16));
memcpy(s->prob->mvc , vp8_mv_default_prob , sizeof(s->prob->mvc)); memcpy(s->prob->pred8x8c, vp8_pred8x8c_prob_inter,
sizeof(s->prob->pred8x8c));
memcpy(s->prob->mvc, vp8_mv_default_prob,
sizeof(s->prob->mvc));
memset(&s->segmentation, 0, sizeof(s->segmentation)); memset(&s->segmentation, 0, sizeof(s->segmentation));
memset(&s->lf_delta, 0, sizeof(s->lf_delta)); memset(&s->lf_delta, 0, sizeof(s->lf_delta));
} }
@ -382,10 +392,9 @@ static int decode_frame_header(VP8Context *s, const uint8_t *buf, int buf_size)
} }
if (!s->macroblocks_base || /* first frame */ if (!s->macroblocks_base || /* first frame */
width != s->avctx->width || height != s->avctx->height) { width != s->avctx->width || height != s->avctx->height)
if ((ret = update_dimensions(s, width, height)) < 0) if ((ret = update_dimensions(s, width, height)) < 0)
return ret; return ret;
}
get_quants(s); get_quants(s);
@ -437,7 +446,8 @@ static int decode_frame_header(VP8Context *s, const uint8_t *buf, int buf_size)
return 0; return 0;
} }
static av_always_inline void clamp_mv(VP8Context *s, VP56mv *dst, const VP56mv *src) static av_always_inline
void clamp_mv(VP8Context *s, VP56mv *dst, const VP56mv *src)
{ {
dst->x = av_clip(src->x, s->mv_min.x, s->mv_max.x); dst->x = av_clip(src->x, s->mv_min.x, s->mv_max.x);
dst->y = av_clip(src->y, s->mv_min.y, s->mv_max.y); dst->y = av_clip(src->y, s->mv_min.y, s->mv_max.y);
@ -495,9 +505,8 @@ int decode_splitmvs(VP8Context *s, VP56RangeCoder *c, VP8Macroblock *mb, int lay
int n, num; int n, num;
VP8Macroblock *top_mb; VP8Macroblock *top_mb;
VP8Macroblock *left_mb = &mb[-1]; VP8Macroblock *left_mb = &mb[-1];
const uint8_t *mbsplits_left = vp8_mbsplits[left_mb->partitioning], const uint8_t *mbsplits_left = vp8_mbsplits[left_mb->partitioning];
*mbsplits_top, const uint8_t *mbsplits_top, *mbsplits_cur, *firstidx;
*mbsplits_cur, *firstidx;
VP56mv *top_mv; VP56mv *top_mv;
VP56mv *left_mv = left_mb->bmv; VP56mv *left_mv = left_mb->bmv;
VP56mv *cur_mv = mb->bmv; VP56mv *cur_mv = mb->bmv;
@ -510,11 +519,10 @@ int decode_splitmvs(VP8Context *s, VP56RangeCoder *c, VP8Macroblock *mb, int lay
top_mv = top_mb->bmv; top_mv = top_mb->bmv;
if (vp56_rac_get_prob_branchy(c, vp8_mbsplit_prob[0])) { if (vp56_rac_get_prob_branchy(c, vp8_mbsplit_prob[0])) {
if (vp56_rac_get_prob_branchy(c, vp8_mbsplit_prob[1])) { if (vp56_rac_get_prob_branchy(c, vp8_mbsplit_prob[1]))
part_idx = VP8_SPLITMVMODE_16x8 + vp56_rac_get_prob(c, vp8_mbsplit_prob[2]); part_idx = VP8_SPLITMVMODE_16x8 + vp56_rac_get_prob(c, vp8_mbsplit_prob[2]);
} else { else
part_idx = VP8_SPLITMVMODE_8x8; part_idx = VP8_SPLITMVMODE_8x8;
}
} else { } else {
part_idx = VP8_SPLITMVMODE_4x4; part_idx = VP8_SPLITMVMODE_4x4;
} }
@ -560,7 +568,8 @@ int decode_splitmvs(VP8Context *s, VP56RangeCoder *c, VP8Macroblock *mb, int lay
} }
static av_always_inline static av_always_inline
void decode_mvs(VP8Context *s, VP8Macroblock *mb, int mb_x, int mb_y, int layout) void decode_mvs(VP8Context *s, VP8Macroblock *mb,
int mb_x, int mb_y, int layout)
{ {
VP8Macroblock *mb_edge[3] = { 0 /* top */, VP8Macroblock *mb_edge[3] = { 0 /* top */,
mb - 1 /* left */, mb - 1 /* left */,
@ -577,8 +586,7 @@ void decode_mvs(VP8Context *s, VP8Macroblock *mb, int mb_x, int mb_y, int layout
if (!layout) { // layout is inlined (s->mb_layout is not) if (!layout) { // layout is inlined (s->mb_layout is not)
mb_edge[0] = mb + 2; mb_edge[0] = mb + 2;
mb_edge[2] = mb + 1; mb_edge[2] = mb + 1;
} } else {
else {
mb_edge[0] = mb - s->mb_width - 1; mb_edge[0] = mb - s->mb_width - 1;
mb_edge[2] = mb - s->mb_width - 2; mb_edge[2] = mb - s->mb_width - 2;
} }
@ -598,7 +606,8 @@ void decode_mvs(VP8Context *s, VP8Macroblock *mb, int mb_x, int mb_y, int layout
if (cur_sign_bias != sign_bias[edge_ref]) { \ if (cur_sign_bias != sign_bias[edge_ref]) { \
/* SWAR negate of the values in mv. */ \ /* SWAR negate of the values in mv. */ \
mv = ~mv; \ mv = ~mv; \
mv = ((mv&0x7fff7fff) + 0x00010001) ^ (mv&0x80008000);\ mv = ((mv & 0x7fff7fff) + \
0x00010001) ^ (mv & 0x80008000); \
} \ } \
if (!n || mv != AV_RN32A(&near_mv[idx])) \ if (!n || mv != AV_RN32A(&near_mv[idx])) \
AV_WN32A(&near_mv[++idx], mv); \ AV_WN32A(&near_mv[++idx], mv); \
@ -617,7 +626,8 @@ void decode_mvs(VP8Context *s, VP8Macroblock *mb, int mb_x, int mb_y, int layout
mb->mode = VP8_MVMODE_MV; mb->mode = VP8_MVMODE_MV;
/* If we have three distinct MVs, merge first and last if they're the same */ /* If we have three distinct MVs, merge first and last if they're the same */
if (cnt[CNT_SPLITMV] && AV_RN32A(&near_mv[1 + VP8_EDGE_TOP]) == AV_RN32A(&near_mv[1 + VP8_EDGE_TOPLEFT])) if (cnt[CNT_SPLITMV] &&
AV_RN32A(&near_mv[1 + VP8_EDGE_TOP]) == AV_RN32A(&near_mv[1 + VP8_EDGE_TOPLEFT]))
cnt[CNT_NEAREST] += 1; cnt[CNT_NEAREST] += 1;
/* Swap near and nearest if necessary */ /* Swap near and nearest if necessary */
@ -628,7 +638,6 @@ void decode_mvs(VP8Context *s, VP8Macroblock *mb, int mb_x, int mb_y, int layout
if (vp56_rac_get_prob_branchy(c, vp8_mode_contexts[cnt[CNT_NEAREST]][1])) { if (vp56_rac_get_prob_branchy(c, vp8_mode_contexts[cnt[CNT_NEAREST]][1])) {
if (vp56_rac_get_prob_branchy(c, vp8_mode_contexts[cnt[CNT_NEAR]][2])) { if (vp56_rac_get_prob_branchy(c, vp8_mode_contexts[cnt[CNT_NEAR]][2])) {
/* Choose the best mv out of 0,0 and the nearest mv */ /* Choose the best mv out of 0,0 and the nearest mv */
clamp_mv(s, &mb->mv, &near_mv[CNT_ZERO + (cnt[CNT_NEAREST] >= cnt[CNT_ZERO])]); clamp_mv(s, &mb->mv, &near_mv[CNT_ZERO + (cnt[CNT_NEAREST] >= cnt[CNT_ZERO])]);
cnt[CNT_SPLITMV] = ((mb_edge[VP8_EDGE_LEFT]->mode == VP8_MVMODE_SPLIT) + cnt[CNT_SPLITMV] = ((mb_edge[VP8_EDGE_LEFT]->mode == VP8_MVMODE_SPLIT) +
@ -688,7 +697,8 @@ void decode_intra4x4_modes(VP8Context *s, VP56RangeCoder *c, VP8Macroblock *mb,
} else { } else {
int i; int i;
for (i = 0; i < 16; i++) for (i = 0; i < 16; i++)
intra4x4[i] = vp8_rac_get_tree(c, vp8_pred4x4_tree, vp8_pred4x4_prob_inter); intra4x4[i] = vp8_rac_get_tree(c, vp8_pred4x4_tree,
vp8_pred4x4_prob_inter);
} }
} }
@ -707,7 +717,8 @@ void decode_mb_mode(VP8Context *s, VP8Macroblock *mb, int mb_x, int mb_y,
mb->skip = s->mbskip_enabled ? vp56_rac_get_prob(c, s->prob->mbskip) : 0; mb->skip = s->mbskip_enabled ? vp56_rac_get_prob(c, s->prob->mbskip) : 0;
if (s->keyframe) { if (s->keyframe) {
mb->mode = vp8_rac_get_tree(c, vp8_pred16x16_tree_intra, vp8_pred16x16_prob_intra); mb->mode = vp8_rac_get_tree(c, vp8_pred16x16_tree_intra,
vp8_pred16x16_prob_intra);
if (mb->mode == MODE_I4x4) { if (mb->mode == MODE_I4x4) {
decode_intra4x4_modes(s, c, mb, mb_x, 1, layout); decode_intra4x4_modes(s, c, mb, mb_x, 1, layout);
@ -720,13 +731,15 @@ void decode_mb_mode(VP8Context *s, VP8Macroblock *mb, int mb_x, int mb_y,
AV_WN32A(s->intra4x4_pred_mode_left, modes); AV_WN32A(s->intra4x4_pred_mode_left, modes);
} }
mb->chroma_pred_mode = vp8_rac_get_tree(c, vp8_pred8x8c_tree, vp8_pred8x8c_prob_intra); mb->chroma_pred_mode = vp8_rac_get_tree(c, vp8_pred8x8c_tree,
vp8_pred8x8c_prob_intra);
mb->ref_frame = VP56_FRAME_CURRENT; mb->ref_frame = VP56_FRAME_CURRENT;
} else if (vp56_rac_get_prob_branchy(c, s->prob->intra)) { } else if (vp56_rac_get_prob_branchy(c, s->prob->intra)) {
// inter MB, 16.2 // inter MB, 16.2
if (vp56_rac_get_prob_branchy(c, s->prob->last)) if (vp56_rac_get_prob_branchy(c, s->prob->last))
mb->ref_frame = vp56_rac_get_prob(c, s->prob->golden) ? mb->ref_frame =
VP56_FRAME_GOLDEN2 /* altref */ : VP56_FRAME_GOLDEN; vp56_rac_get_prob(c, s->prob->golden) ? VP56_FRAME_GOLDEN2 /* altref */
: VP56_FRAME_GOLDEN;
else else
mb->ref_frame = VP56_FRAME_PREVIOUS; mb->ref_frame = VP56_FRAME_PREVIOUS;
s->ref_count[mb->ref_frame - 1]++; s->ref_count[mb->ref_frame - 1]++;
@ -740,7 +753,8 @@ void decode_mb_mode(VP8Context *s, VP8Macroblock *mb, int mb_x, int mb_y,
if (mb->mode == MODE_I4x4) if (mb->mode == MODE_I4x4)
decode_intra4x4_modes(s, c, mb, mb_x, 0, layout); decode_intra4x4_modes(s, c, mb, mb_x, 0, layout);
mb->chroma_pred_mode = vp8_rac_get_tree(c, vp8_pred8x8c_tree, s->prob->pred8x8c); mb->chroma_pred_mode = vp8_rac_get_tree(c, vp8_pred8x8c_tree,
s->prob->pred8x8c);
mb->ref_frame = VP56_FRAME_CURRENT; mb->ref_frame = VP56_FRAME_CURRENT;
mb->partitioning = VP8_SPLITMVMODE_NONE; mb->partitioning = VP8_SPLITMVMODE_NONE;
AV_ZERO32(&mb->bmv[0]); AV_ZERO32(&mb->bmv[0]);
@ -754,12 +768,14 @@ void decode_mb_mode(VP8Context *s, VP8Macroblock *mb, int mb_x, int mb_y,
* @param probs probabilities to use when reading trees from the bitstream * @param probs probabilities to use when reading trees from the bitstream
* @param i initial coeff index, 0 unless a separate DC block is coded * @param i initial coeff index, 0 unless a separate DC block is coded
* @param qmul array holding the dc/ac dequant factor at position 0/1 * @param qmul array holding the dc/ac dequant factor at position 0/1
*
* @return 0 if no coeffs were decoded * @return 0 if no coeffs were decoded
* otherwise, the index of the last coeff decoded plus one * otherwise, the index of the last coeff decoded plus one
*/ */
static int decode_block_coeffs_internal(VP56RangeCoder *r, int16_t block[16], static int decode_block_coeffs_internal(VP56RangeCoder *r, int16_t block[16],
uint8_t probs[16][3][NUM_DCT_TOKENS - 1], uint8_t probs[16][3][NUM_DCT_TOKENS - 1],
int i, uint8_t *token_prob, int16_t qmul[2]) int i, uint8_t *token_prob,
int16_t qmul[2])
{ {
VP56RangeCoder c = *r; VP56RangeCoder c = *r;
goto skip_eob; goto skip_eob;
@ -821,6 +837,7 @@ skip_eob:
* @param zero_nhood the initial prediction context for number of surrounding * @param zero_nhood the initial prediction context for number of surrounding
* all-zero blocks (only left/top, so 0-2) * all-zero blocks (only left/top, so 0-2)
* @param qmul array holding the dc/ac dequant factor at position 0/1 * @param qmul array holding the dc/ac dequant factor at position 0/1
*
* @return 0 if no coeffs were decoded * @return 0 if no coeffs were decoded
* otherwise, the index of the last coeff decoded plus one * otherwise, the index of the last coeff decoded plus one
*/ */
@ -836,8 +853,8 @@ int decode_block_coeffs(VP56RangeCoder *c, int16_t block[16],
} }
static av_always_inline static av_always_inline
void decode_mb_coeffs(VP8Context *s, VP8ThreadData *td, VP56RangeCoder *c, VP8Macroblock *mb, void decode_mb_coeffs(VP8Context *s, VP8ThreadData *td, VP56RangeCoder *c,
uint8_t t_nnz[9], uint8_t l_nnz[9]) VP8Macroblock *mb, uint8_t t_nnz[9], uint8_t l_nnz[9])
{ {
int i, x, y, luma_start = 0, luma_ctx = 3; int i, x, y, luma_start = 0, luma_ctx = 3;
int nnz_pred, nnz, nnz_total = 0; int nnz_pred, nnz, nnz_total = 0;
@ -848,8 +865,8 @@ void decode_mb_coeffs(VP8Context *s, VP8ThreadData *td, VP56RangeCoder *c, VP8Ma
nnz_pred = t_nnz[8] + l_nnz[8]; nnz_pred = t_nnz[8] + l_nnz[8];
// decode DC values and do hadamard // decode DC values and do hadamard
nnz = decode_block_coeffs(c, td->block_dc, s->prob->token[1], 0, nnz_pred, nnz = decode_block_coeffs(c, td->block_dc, s->prob->token[1], 0,
s->qmat[segment].luma_dc_qmul); nnz_pred, s->qmat[segment].luma_dc_qmul);
l_nnz[8] = t_nnz[8] = !!nnz; l_nnz[8] = t_nnz[8] = !!nnz;
if (nnz) { if (nnz) {
nnz_total += nnz; nnz_total += nnz;
@ -867,9 +884,12 @@ void decode_mb_coeffs(VP8Context *s, VP8ThreadData *td, VP56RangeCoder *c, VP8Ma
for (y = 0; y < 4; y++) for (y = 0; y < 4; y++)
for (x = 0; x < 4; x++) { for (x = 0; x < 4; x++) {
nnz_pred = l_nnz[y] + t_nnz[x]; nnz_pred = l_nnz[y] + t_nnz[x];
nnz = decode_block_coeffs(c, td->block[y][x], s->prob->token[luma_ctx], luma_start, nnz = decode_block_coeffs(c, td->block[y][x],
nnz_pred, s->qmat[segment].luma_qmul); s->prob->token[luma_ctx],
// nnz+block_dc may be one more than the actual last index, but we don't care luma_start, nnz_pred,
s->qmat[segment].luma_qmul);
/* nnz+block_dc may be one more than the actual last index,
* but we don't care */
td->non_zero_count_cache[y][x] = nnz + block_dc; td->non_zero_count_cache[y][x] = nnz + block_dc;
t_nnz[x] = l_nnz[y] = !!nnz; t_nnz[x] = l_nnz[y] = !!nnz;
nnz_total += nnz; nnz_total += nnz;
@ -882,8 +902,10 @@ void decode_mb_coeffs(VP8Context *s, VP8ThreadData *td, VP56RangeCoder *c, VP8Ma
for (y = 0; y < 2; y++) for (y = 0; y < 2; y++)
for (x = 0; x < 2; x++) { for (x = 0; x < 2; x++) {
nnz_pred = l_nnz[i + 2 * y] + t_nnz[i + 2 * x]; nnz_pred = l_nnz[i + 2 * y] + t_nnz[i + 2 * x];
nnz = decode_block_coeffs(c, td->block[i][(y<<1)+x], s->prob->token[2], 0, nnz = decode_block_coeffs(c, td->block[i][(y << 1) + x],
nnz_pred, s->qmat[segment].chroma_qmul); s->prob->token[2],
0, nnz_pred,
s->qmat[segment].chroma_qmul);
td->non_zero_count_cache[i][(y << 1) + x] = nnz; td->non_zero_count_cache[i][(y << 1) + x] = nnz;
t_nnz[i + 2 * x] = l_nnz[i + 2 * y] = !!nnz; t_nnz[i + 2 * x] = l_nnz[i + 2 * y] = !!nnz;
nnz_total += nnz; nnz_total += nnz;
@ -897,7 +919,8 @@ void decode_mb_coeffs(VP8Context *s, VP8ThreadData *td, VP56RangeCoder *c, VP8Ma
} }
static av_always_inline static av_always_inline
void backup_mb_border(uint8_t *top_border, uint8_t *src_y, uint8_t *src_cb, uint8_t *src_cr, void backup_mb_border(uint8_t *top_border, uint8_t *src_y,
uint8_t *src_cb, uint8_t *src_cr,
int linesize, int uvlinesize, int simple) int linesize, int uvlinesize, int simple)
{ {
AV_COPY128(top_border, src_y + 15 * linesize); AV_COPY128(top_border, src_y + 15 * linesize);
@ -908,18 +931,21 @@ void backup_mb_border(uint8_t *top_border, uint8_t *src_y, uint8_t *src_cb, uint
} }
static av_always_inline static av_always_inline
void xchg_mb_border(uint8_t *top_border, uint8_t *src_y, uint8_t *src_cb, uint8_t *src_cr, void xchg_mb_border(uint8_t *top_border, uint8_t *src_y, uint8_t *src_cb,
int linesize, int uvlinesize, int mb_x, int mb_y, int mb_width, uint8_t *src_cr, int linesize, int uvlinesize, int mb_x,
int simple, int xchg) int mb_y, int mb_width, int simple, int xchg)
{ {
uint8_t *top_border_m1 = top_border - 32; // for TL prediction uint8_t *top_border_m1 = top_border - 32; // for TL prediction
src_y -= linesize; src_y -= linesize;
src_cb -= uvlinesize; src_cb -= uvlinesize;
src_cr -= uvlinesize; src_cr -= uvlinesize;
#define XCHG(a,b,xchg) do { \ #define XCHG(a, b, xchg) \
if (xchg) AV_SWAP64(b,a); \ do { \
else AV_COPY64(b,a); \ if (xchg) \
AV_SWAP64(b, a); \
else \
AV_COPY64(b, a); \
} while (0) } while (0)
XCHG(top_border_m1 + 8, src_y - 8, xchg); XCHG(top_border_m1 + 8, src_y - 8, xchg);
@ -941,22 +967,20 @@ void xchg_mb_border(uint8_t *top_border, uint8_t *src_y, uint8_t *src_cb, uint8_
static av_always_inline static av_always_inline
int check_dc_pred8x8_mode(int mode, int mb_x, int mb_y) int check_dc_pred8x8_mode(int mode, int mb_x, int mb_y)
{ {
if (!mb_x) { if (!mb_x)
return mb_y ? TOP_DC_PRED8x8 : DC_128_PRED8x8; return mb_y ? TOP_DC_PRED8x8 : DC_128_PRED8x8;
} else { else
return mb_y ? mode : LEFT_DC_PRED8x8; return mb_y ? mode : LEFT_DC_PRED8x8;
} }
}
static av_always_inline static av_always_inline
int check_tm_pred8x8_mode(int mode, int mb_x, int mb_y) int check_tm_pred8x8_mode(int mode, int mb_x, int mb_y)
{ {
if (!mb_x) { if (!mb_x)
return mb_y ? VERT_PRED8x8 : DC_129_PRED8x8; return mb_y ? VERT_PRED8x8 : DC_129_PRED8x8;
} else { else
return mb_y ? mode : HOR_PRED8x8; return mb_y ? mode : HOR_PRED8x8;
} }
}
static av_always_inline static av_always_inline
int check_intra_pred8x8_mode_emuedge(int mode, int mb_x, int mb_y) int check_intra_pred8x8_mode_emuedge(int mode, int mb_x, int mb_y)
@ -968,7 +992,7 @@ int check_intra_pred8x8_mode_emuedge(int mode, int mb_x, int mb_y)
return !mb_y ? DC_127_PRED8x8 : mode; return !mb_y ? DC_127_PRED8x8 : mode;
case HOR_PRED8x8: case HOR_PRED8x8:
return !mb_x ? DC_129_PRED8x8 : mode; return !mb_x ? DC_129_PRED8x8 : mode;
case PLANE_PRED8x8 /*TM*/: case PLANE_PRED8x8: /* TM */
return check_tm_pred8x8_mode(mode, mb_x, mb_y); return check_tm_pred8x8_mode(mode, mb_x, mb_y);
} }
return mode; return mode;
@ -1007,7 +1031,8 @@ int check_intra_pred4x4_mode_emuedge(int mode, int mb_x, int mb_y, int *copy_buf
return !mb_x ? DC_129_PRED : mode; return !mb_x ? DC_129_PRED : mode;
case TM_VP8_PRED: case TM_VP8_PRED:
return check_tm_pred4x4_mode(mode, mb_x, mb_y); return check_tm_pred4x4_mode(mode, mb_x, mb_y);
case DC_PRED: // 4x4 DC doesn't use the same "H.264-style" exceptions as 16x16/8x8 DC case DC_PRED: /* 4x4 DC doesn't use the same "H.264-style" exceptions
* as 16x16/8x8 DC */
case DIAG_DOWN_RIGHT_PRED: case DIAG_DOWN_RIGHT_PRED:
case VERT_RIGHT_PRED: case VERT_RIGHT_PRED:
case HOR_DOWN_PRED: case HOR_DOWN_PRED:
@ -1025,8 +1050,8 @@ void intra_predict(VP8Context *s, VP8ThreadData *td, uint8_t *dst[3],
int x, y, mode, nnz; int x, y, mode, nnz;
uint32_t tr; uint32_t tr;
// for the first row, we need to run xchg_mb_border to init the top edge to 127 /* for the first row, we need to run xchg_mb_border to init the top edge
// otherwise, skip it if we aren't going to deblock * to 127 otherwise, skip it if we aren't going to deblock */
if (mb_y && (s->deblock_filter || !mb_y) && td->thread_nr == 0) if (mb_y && (s->deblock_filter || !mb_y) && td->thread_nr == 0)
xchg_mb_border(s->top_border[mb_x + 1], dst[0], dst[1], dst[2], xchg_mb_border(s->top_border[mb_x + 1], dst[0], dst[1], dst[2],
s->linesize, s->uvlinesize, mb_x, mb_y, s->mb_width, s->linesize, s->uvlinesize, mb_x, mb_y, s->mb_width,
@ -1046,8 +1071,7 @@ void intra_predict(VP8Context *s, VP8ThreadData *td, uint8_t *dst[3],
// if we're on the right edge of the frame, said edge is extended // if we're on the right edge of the frame, said edge is extended
// from the top macroblock // from the top macroblock
if (mb_y && if (mb_y && mb_x == s->mb_width - 1) {
mb_x == s->mb_width-1) {
tr = tr_right[-1] * 0x01010101u; tr = tr_right[-1] * 0x01010101u;
tr_right = (uint8_t *) &tr; tr_right = (uint8_t *) &tr;
} }
@ -1067,7 +1091,9 @@ void intra_predict(VP8Context *s, VP8ThreadData *td, uint8_t *dst[3],
} else if (x == 3) } else if (x == 3)
topright = tr_right; topright = tr_right;
mode = check_intra_pred4x4_mode_emuedge(intra4x4[x], mb_x + x, mb_y + y, &copy); mode = check_intra_pred4x4_mode_emuedge(intra4x4[x],
mb_x + x, mb_y + y,
&copy);
if (copy) { if (copy) {
dst = copy_dst + 12; dst = copy_dst + 12;
linesize = 8; linesize = 8;
@ -1105,9 +1131,11 @@ void intra_predict(VP8Context *s, VP8ThreadData *td, uint8_t *dst[3],
nnz = td->non_zero_count_cache[y][x]; nnz = td->non_zero_count_cache[y][x];
if (nnz) { if (nnz) {
if (nnz == 1) if (nnz == 1)
s->vp8dsp.vp8_idct_dc_add(ptr+4*x, td->block[y][x], s->linesize); s->vp8dsp.vp8_idct_dc_add(ptr + 4 * x,
td->block[y][x], s->linesize);
else else
s->vp8dsp.vp8_idct_add(ptr+4*x, td->block[y][x], s->linesize); s->vp8dsp.vp8_idct_add(ptr + 4 * x,
td->block[y][x], s->linesize);
} }
topright += 4; topright += 4;
} }
@ -1176,15 +1204,18 @@ void vp8_mc_luma(VP8Context *s, VP8ThreadData *td, uint8_t *dst,
s->vdsp.emulated_edge_mc(td->edge_emu_buffer, s->vdsp.emulated_edge_mc(td->edge_emu_buffer,
src - my_idx * linesize - mx_idx, src - my_idx * linesize - mx_idx,
EDGE_EMU_LINESIZE, linesize, EDGE_EMU_LINESIZE, linesize,
block_w + subpel_idx[1][mx], block_h + subpel_idx[1][my], block_w + subpel_idx[1][mx],
x_off - mx_idx, y_off - my_idx, width, height); block_h + subpel_idx[1][my],
x_off - mx_idx, y_off - my_idx,
width, height);
src = td->edge_emu_buffer + mx_idx + EDGE_EMU_LINESIZE * my_idx; src = td->edge_emu_buffer + mx_idx + EDGE_EMU_LINESIZE * my_idx;
src_linesize = EDGE_EMU_LINESIZE; src_linesize = EDGE_EMU_LINESIZE;
} }
mc_func[my_idx][mx_idx](dst, linesize, src, src_linesize, block_h, mx, my); mc_func[my_idx][mx_idx](dst, linesize, src, src_linesize, block_h, mx, my);
} else { } else {
ff_thread_await_progress(ref, (3 + y_off + block_h) >> 4, 0); ff_thread_await_progress(ref, (3 + y_off + block_h) >> 4, 0);
mc_func[0][0](dst, linesize, src + y_off * linesize + x_off, linesize, block_h, 0, 0); mc_func[0][0](dst, linesize, src + y_off * linesize + x_off,
linesize, block_h, 0, 0);
} }
} }
@ -1206,9 +1237,10 @@ void vp8_mc_luma(VP8Context *s, VP8ThreadData *td, uint8_t *dst,
* @param mc_func motion compensation function pointers (bilinear or sixtap MC) * @param mc_func motion compensation function pointers (bilinear or sixtap MC)
*/ */
static av_always_inline static av_always_inline
void vp8_mc_chroma(VP8Context *s, VP8ThreadData *td, uint8_t *dst1, uint8_t *dst2, void vp8_mc_chroma(VP8Context *s, VP8ThreadData *td, uint8_t *dst1,
ThreadFrame *ref, const VP56mv *mv, int x_off, int y_off, uint8_t *dst2, ThreadFrame *ref, const VP56mv *mv,
int block_w, int block_h, int width, int height, ptrdiff_t linesize, int x_off, int y_off, int block_w, int block_h,
int width, int height, ptrdiff_t linesize,
vp8_mc_func mc_func[3][3]) vp8_mc_func mc_func[3][3])
{ {
uint8_t *src1 = ref->f->data[1], *src2 = ref->f->data[2]; uint8_t *src1 = ref->f->data[1], *src2 = ref->f->data[2];
@ -1255,8 +1287,7 @@ void vp8_mc_chroma(VP8Context *s, VP8ThreadData *td, uint8_t *dst1, uint8_t *dst
static av_always_inline static av_always_inline
void vp8_mc_part(VP8Context *s, VP8ThreadData *td, uint8_t *dst[3], void vp8_mc_part(VP8Context *s, VP8ThreadData *td, uint8_t *dst[3],
ThreadFrame *ref_frame, int x_off, int y_off, ThreadFrame *ref_frame, int x_off, int y_off,
int bx_off, int by_off, int bx_off, int by_off, int block_w, int block_h,
int block_w, int block_h,
int width, int height, VP56mv *mv) int width, int height, VP56mv *mv)
{ {
VP56mv uvmv = *mv; VP56mv uvmv = *mv;
@ -1272,10 +1303,14 @@ void vp8_mc_part(VP8Context *s, VP8ThreadData *td, uint8_t *dst[3],
uvmv.x &= ~7; uvmv.x &= ~7;
uvmv.y &= ~7; uvmv.y &= ~7;
} }
x_off >>= 1; y_off >>= 1; x_off >>= 1;
bx_off >>= 1; by_off >>= 1; y_off >>= 1;
width >>= 1; height >>= 1; bx_off >>= 1;
block_w >>= 1; block_h >>= 1; by_off >>= 1;
width >>= 1;
height >>= 1;
block_w >>= 1;
block_h >>= 1;
vp8_mc_chroma(s, td, dst[1] + by_off * s->uvlinesize + bx_off, vp8_mc_chroma(s, td, dst[1] + by_off * s->uvlinesize + bx_off,
dst[2] + by_off * s->uvlinesize + bx_off, ref_frame, dst[2] + by_off * s->uvlinesize + bx_off, ref_frame,
&uvmv, x_off + bx_off, y_off + by_off, &uvmv, x_off + bx_off, y_off + by_off,
@ -1285,7 +1320,9 @@ void vp8_mc_part(VP8Context *s, VP8ThreadData *td, uint8_t *dst[3],
/* Fetch pixels for estimated mv 4 macroblocks ahead. /* Fetch pixels for estimated mv 4 macroblocks ahead.
* Optimized for 64-byte cache lines. Inspired by ffh264 prefetch_motion. */ * Optimized for 64-byte cache lines. Inspired by ffh264 prefetch_motion. */
static av_always_inline void prefetch_motion(VP8Context *s, VP8Macroblock *mb, int mb_x, int mb_y, int mb_xy, int ref) static av_always_inline
void prefetch_motion(VP8Context *s, VP8Macroblock *mb, int mb_x, int mb_y,
int mb_xy, int ref)
{ {
/* Don't prefetch refs that haven't been used very often this frame. */ /* Don't prefetch refs that haven't been used very often this frame. */
if (s->ref_count[ref - 1] > (mb_xy >> 5)) { if (s->ref_count[ref - 1] > (mb_xy >> 5)) {
@ -1336,7 +1373,10 @@ void inter_predict(VP8Context *s, VP8ThreadData *td, uint8_t *dst[3],
} }
/* U/V */ /* U/V */
x_off >>= 1; y_off >>= 1; width >>= 1; height >>= 1; x_off >>= 1;
y_off >>= 1;
width >>= 1;
height >>= 1;
for (y = 0; y < 2; y++) { for (y = 0; y < 2; y++) {
for (x = 0; x < 2; x++) { for (x = 0; x < 2; x++) {
uvmv.x = mb->bmv[2 * y * 4 + 2 * x ].x + uvmv.x = mb->bmv[2 * y * 4 + 2 * x ].x +
@ -1354,8 +1394,8 @@ void inter_predict(VP8Context *s, VP8ThreadData *td, uint8_t *dst[3],
uvmv.y &= ~7; uvmv.y &= ~7;
} }
vp8_mc_chroma(s, td, dst[1] + 4 * y * s->uvlinesize + x * 4, vp8_mc_chroma(s, td, dst[1] + 4 * y * s->uvlinesize + x * 4,
dst[2] + 4*y*s->uvlinesize + x*4, ref, &uvmv, dst[2] + 4 * y * s->uvlinesize + x * 4, ref,
4*x + x_off, 4*y + y_off, 4, 4, &uvmv, 4 * x + x_off, 4 * y + y_off, 4, 4,
width, height, s->uvlinesize, width, height, s->uvlinesize,
s->put_pixels_tab[2]); s->put_pixels_tab[2]);
} }
@ -1387,8 +1427,8 @@ void inter_predict(VP8Context *s, VP8ThreadData *td, uint8_t *dst[3],
} }
} }
static av_always_inline void idct_mb(VP8Context *s, VP8ThreadData *td, static av_always_inline
uint8_t *dst[3], VP8Macroblock *mb) void idct_mb(VP8Context *s, VP8ThreadData *td, uint8_t *dst[3], VP8Macroblock *mb)
{ {
int x, y, ch; int x, y, ch;
@ -1400,9 +1440,13 @@ static av_always_inline void idct_mb(VP8Context *s, VP8ThreadData *td,
if (nnz4 & ~0x01010101) { if (nnz4 & ~0x01010101) {
for (x = 0; x < 4; x++) { for (x = 0; x < 4; x++) {
if ((uint8_t) nnz4 == 1) if ((uint8_t) nnz4 == 1)
s->vp8dsp.vp8_idct_dc_add(y_dst+4*x, td->block[y][x], s->linesize); s->vp8dsp.vp8_idct_dc_add(y_dst + 4 * x,
td->block[y][x],
s->linesize);
else if ((uint8_t) nnz4 > 1) else if ((uint8_t) nnz4 > 1)
s->vp8dsp.vp8_idct_add(y_dst+4*x, td->block[y][x], s->linesize); s->vp8dsp.vp8_idct_add(y_dst + 4 * x,
td->block[y][x],
s->linesize);
nnz4 >>= 8; nnz4 >>= 8;
if (!nnz4) if (!nnz4)
break; break;
@ -1423,9 +1467,13 @@ static av_always_inline void idct_mb(VP8Context *s, VP8ThreadData *td,
for (y = 0; y < 2; y++) { for (y = 0; y < 2; y++) {
for (x = 0; x < 2; x++) { for (x = 0; x < 2; x++) {
if ((uint8_t) nnz4 == 1) if ((uint8_t) nnz4 == 1)
s->vp8dsp.vp8_idct_dc_add(ch_dst+4*x, td->block[4+ch][(y<<1)+x], s->uvlinesize); s->vp8dsp.vp8_idct_dc_add(ch_dst + 4 * x,
td->block[4 + ch][(y << 1) + x],
s->uvlinesize);
else if ((uint8_t) nnz4 > 1) else if ((uint8_t) nnz4 > 1)
s->vp8dsp.vp8_idct_add(ch_dst+4*x, td->block[4+ch][(y<<1)+x], s->uvlinesize); s->vp8dsp.vp8_idct_add(ch_dst + 4 * x,
td->block[4 + ch][(y << 1) + x],
s->uvlinesize);
nnz4 >>= 8; nnz4 >>= 8;
if (!nnz4) if (!nnz4)
goto chroma_idct_end; goto chroma_idct_end;
@ -1436,11 +1484,13 @@ static av_always_inline void idct_mb(VP8Context *s, VP8ThreadData *td,
s->vp8dsp.vp8_idct_dc_add4uv(ch_dst, td->block[4 + ch], s->uvlinesize); s->vp8dsp.vp8_idct_dc_add4uv(ch_dst, td->block[4 + ch], s->uvlinesize);
} }
} }
chroma_idct_end: ; chroma_idct_end:
;
} }
} }
static av_always_inline void filter_level_for_mb(VP8Context *s, VP8Macroblock *mb, VP8FilterStrength *f ) static av_always_inline
void filter_level_for_mb(VP8Context *s, VP8Macroblock *mb, VP8FilterStrength *f)
{ {
int interior_limit, filter_level; int interior_limit, filter_level;
@ -1467,10 +1517,13 @@ static av_always_inline void filter_level_for_mb(VP8Context *s, VP8Macroblock *m
f->filter_level = filter_level; f->filter_level = filter_level;
f->inner_limit = interior_limit; f->inner_limit = interior_limit;
f->inner_filter = !mb->skip || mb->mode == MODE_I4x4 || mb->mode == VP8_MVMODE_SPLIT; f->inner_filter = !mb->skip || mb->mode == MODE_I4x4 ||
mb->mode == VP8_MVMODE_SPLIT;
} }
static av_always_inline void filter_mb(VP8Context *s, uint8_t *dst[3], VP8FilterStrength *f, int mb_x, int mb_y) static av_always_inline
void filter_mb(VP8Context *s, uint8_t *dst[3], VP8FilterStrength *f,
int mb_x, int mb_y)
{ {
int mbedge_lim, bedge_lim, hev_thresh; int mbedge_lim, bedge_lim, hev_thresh;
int filter_level = f->filter_level; int filter_level = f->filter_level;
@ -1540,7 +1593,9 @@ static av_always_inline void filter_mb(VP8Context *s, uint8_t *dst[3], VP8Filter
} }
} }
static av_always_inline void filter_mb_simple(VP8Context *s, uint8_t *dst, VP8FilterStrength *f, int mb_x, int mb_y) static av_always_inline
void filter_mb_simple(VP8Context *s, uint8_t *dst, VP8FilterStrength *f,
int mb_x, int mb_y)
{ {
int mbedge_lim, bedge_lim; int mbedge_lim, bedge_lim;
int filter_level = f->filter_level; int filter_level = f->filter_level;
@ -1581,7 +1636,8 @@ static void vp8_decode_mv_mb_modes(AVCodecContext *avctx, VP8Frame *curframe,
s->mv_min.y = -MARGIN; s->mv_min.y = -MARGIN;
s->mv_max.y = ((s->mb_height - 1) << 6) + MARGIN; s->mv_max.y = ((s->mb_height - 1) << 6) + MARGIN;
for (mb_y = 0; mb_y < s->mb_height; mb_y++) { for (mb_y = 0; mb_y < s->mb_height; mb_y++) {
VP8Macroblock *mb = s->macroblocks_base + ((s->mb_width+1)*(mb_y + 1) + 1); VP8Macroblock *mb = s->macroblocks_base +
((s->mb_width + 1) * (mb_y + 1) + 1);
int mb_xy = mb_y * s->mb_width; int mb_xy = mb_y * s->mb_width;
AV_WN32A(s->intra4x4_pred_mode_left, DC_PRED * 0x01010101); AV_WN32A(s->intra4x4_pred_mode_left, DC_PRED * 0x01010101);
@ -1590,7 +1646,8 @@ static void vp8_decode_mv_mb_modes(AVCodecContext *avctx, VP8Frame *curframe,
s->mv_max.x = ((s->mb_width - 1) << 6) + MARGIN; s->mv_max.x = ((s->mb_width - 1) << 6) + MARGIN;
for (mb_x = 0; mb_x < s->mb_width; mb_x++, mb_xy++, mb++) { for (mb_x = 0; mb_x < s->mb_width; mb_x++, mb_xy++, mb++) {
if (mb_y == 0) if (mb_y == 0)
AV_WN32A((mb-s->mb_width-1)->intra4x4_pred_mode_top, DC_PRED*0x01010101); AV_WN32A((mb - s->mb_width - 1)->intra4x4_pred_mode_top,
DC_PRED * 0x01010101);
decode_mb_mode(s, mb, mb_x, mb_y, curframe->seg_map->data + mb_xy, decode_mb_mode(s, mb, mb_x, mb_y, curframe->seg_map->data + mb_xy,
prev_frame && prev_frame->seg_map ? prev_frame && prev_frame->seg_map ?
prev_frame->seg_map->data + mb_xy : NULL, 1); prev_frame->seg_map->data + mb_xy : NULL, 1);
@ -1622,11 +1679,14 @@ static void vp8_decode_mv_mb_modes(AVCodecContext *avctx, VP8Frame *curframe,
#define update_pos(td, mb_y, mb_x) \ #define update_pos(td, mb_y, mb_x) \
do { \ do { \
int pos = (mb_y << 16) | (mb_x & 0xFFFF); \ int pos = (mb_y << 16) | (mb_x & 0xFFFF); \
int sliced_threading = (avctx->active_thread_type == FF_THREAD_SLICE) && (num_jobs > 1);\ int sliced_threading = (avctx->active_thread_type == FF_THREAD_SLICE) && \
(num_jobs > 1); \
int is_null = (next_td == NULL) || (prev_td == NULL); \ int is_null = (next_td == NULL) || (prev_td == NULL); \
int pos_check = (is_null) ? 1 :\ int pos_check = (is_null) ? 1 \
(next_td != td && pos >= next_td->wait_mb_pos) ||\ : (next_td != td && \
(prev_td != td && pos >= prev_td->wait_mb_pos);\ pos >= next_td->wait_mb_pos) || \
(prev_td != td && \
pos >= prev_td->wait_mb_pos); \
td->thread_mb_pos = pos; \ td->thread_mb_pos = pos; \
if (sliced_threading && pos_check) { \ if (sliced_threading && pos_check) { \
pthread_mutex_lock(&td->lock); \ pthread_mutex_lock(&td->lock); \
@ -1655,10 +1715,14 @@ static void vp8_decode_mb_row_no_filter(AVCodecContext *avctx, void *tdata,
curframe->tf.f->data[1] + 8 * mb_y * s->uvlinesize, curframe->tf.f->data[1] + 8 * mb_y * s->uvlinesize,
curframe->tf.f->data[2] + 8 * mb_y * s->uvlinesize curframe->tf.f->data[2] + 8 * mb_y * s->uvlinesize
}; };
if (mb_y == 0) prev_td = td; if (mb_y == 0)
else prev_td = &s->thread_data[(jobnr + num_jobs - 1)%num_jobs]; prev_td = td;
if (mb_y == s->mb_height-1) next_td = td; else
else next_td = &s->thread_data[(jobnr + 1)%num_jobs]; prev_td = &s->thread_data[(jobnr + num_jobs - 1) % num_jobs];
if (mb_y == s->mb_height - 1)
next_td = td;
else
next_td = &s->thread_data[(jobnr + 1) % num_jobs];
if (s->mb_layout == 1) if (s->mb_layout == 1)
mb = s->macroblocks_base + ((s->mb_width + 1) * (mb_y + 1) + 1); mb = s->macroblocks_base + ((s->mb_width + 1) * (mb_y + 1) + 1);
else { else {
@ -1683,12 +1747,15 @@ static void vp8_decode_mb_row_no_filter(AVCodecContext *avctx, void *tdata,
if (threadnr != 0) { if (threadnr != 0) {
check_thread_pos(td, prev_td, mb_x + 1, mb_y - 1); check_thread_pos(td, prev_td, mb_x + 1, mb_y - 1);
} else { } else {
check_thread_pos(td, prev_td, (s->mb_width+3) + (mb_x+1), mb_y-1); check_thread_pos(td, prev_td,
(s->mb_width + 3) + (mb_x + 1), mb_y - 1);
} }
} }
s->vdsp.prefetch(dst[0] + (mb_x&3)*4*s->linesize + 64, s->linesize, 4); s->vdsp.prefetch(dst[0] + (mb_x & 3) * 4 * s->linesize + 64,
s->vdsp.prefetch(dst[1] + (mb_x&7)*s->uvlinesize + 64, dst[2] - dst[1], 2); s->linesize, 4);
s->vdsp.prefetch(dst[1] + (mb_x & 7) * s->uvlinesize + 64,
dst[2] - dst[1], 2);
if (!s->mb_layout) if (!s->mb_layout)
decode_mb_mode(s, mb, mb_x, mb_y, curframe->seg_map->data + mb_xy, decode_mb_mode(s, mb, mb_x, mb_y, curframe->seg_map->data + mb_xy,
@ -1713,7 +1780,8 @@ static void vp8_decode_mb_row_no_filter(AVCodecContext *avctx, void *tdata,
AV_ZERO64(td->left_nnz); AV_ZERO64(td->left_nnz);
AV_WN64(s->top_nnz[mb_x], 0); // array of 9, so unaligned AV_WN64(s->top_nnz[mb_x], 0); // array of 9, so unaligned
// Reset DC block predictors if they would exist if the mb had coefficients /* Reset DC block predictors if they would exist
* if the mb had coefficients */
if (mb->mode != MODE_I4x4 && mb->mode != VP8_MVMODE_SPLIT) { if (mb->mode != MODE_I4x4 && mb->mode != VP8_MVMODE_SPLIT) {
td->left_nnz[8] = 0; td->left_nnz[8] = 0;
s->top_nnz[mb_x][8] = 0; s->top_nnz[mb_x][8] = 0;
@ -1725,9 +1793,11 @@ static void vp8_decode_mb_row_no_filter(AVCodecContext *avctx, void *tdata,
if (s->deblock_filter && num_jobs != 1 && threadnr == num_jobs - 1) { if (s->deblock_filter && num_jobs != 1 && threadnr == num_jobs - 1) {
if (s->filter.simple) if (s->filter.simple)
backup_mb_border(s->top_border[mb_x+1], dst[0], NULL, NULL, s->linesize, 0, 1); backup_mb_border(s->top_border[mb_x + 1], dst[0],
NULL, NULL, s->linesize, 0, 1);
else else
backup_mb_border(s->top_border[mb_x+1], dst[0], dst[1], dst[2], s->linesize, s->uvlinesize, 0); backup_mb_border(s->top_border[mb_x + 1], dst[0],
dst[1], dst[2], s->linesize, s->uvlinesize, 0);
} }
prefetch_motion(s, mb, mb_x, mb_y, mb_xy, VP56_FRAME_GOLDEN2); prefetch_motion(s, mb, mb_x, mb_y, mb_xy, VP56_FRAME_GOLDEN2);
@ -1766,26 +1836,31 @@ static void vp8_filter_mb_row(AVCodecContext *avctx, void *tdata,
else else
mb = s->macroblocks + (s->mb_height - mb_y - 1) * 2; mb = s->macroblocks + (s->mb_height - mb_y - 1) * 2;
if (mb_y == 0) prev_td = td; if (mb_y == 0)
else prev_td = &s->thread_data[(jobnr + num_jobs - 1)%num_jobs]; prev_td = td;
if (mb_y == s->mb_height-1) next_td = td; else
else next_td = &s->thread_data[(jobnr + 1)%num_jobs]; prev_td = &s->thread_data[(jobnr + num_jobs - 1) % num_jobs];
if (mb_y == s->mb_height - 1)
next_td = td;
else
next_td = &s->thread_data[(jobnr + 1) % num_jobs];
for (mb_x = 0; mb_x < s->mb_width; mb_x++, mb++) { for (mb_x = 0; mb_x < s->mb_width; mb_x++, mb++) {
VP8FilterStrength *f = &td->filter_strength[mb_x]; VP8FilterStrength *f = &td->filter_strength[mb_x];
if (prev_td != td) { if (prev_td != td)
check_thread_pos(td, prev_td, (mb_x+1) + (s->mb_width+3), mb_y-1); check_thread_pos(td, prev_td,
} (mb_x + 1) + (s->mb_width + 3), mb_y - 1);
if (next_td != td) if (next_td != td)
if (next_td != &s->thread_data[0]) { if (next_td != &s->thread_data[0])
check_thread_pos(td, next_td, mb_x + 1, mb_y + 1); check_thread_pos(td, next_td, mb_x + 1, mb_y + 1);
}
if (num_jobs == 1) { if (num_jobs == 1) {
if (s->filter.simple) if (s->filter.simple)
backup_mb_border(s->top_border[mb_x+1], dst[0], NULL, NULL, s->linesize, 0, 1); backup_mb_border(s->top_border[mb_x + 1], dst[0],
NULL, NULL, s->linesize, 0, 1);
else else
backup_mb_border(s->top_border[mb_x+1], dst[0], dst[1], dst[2], s->linesize, s->uvlinesize, 0); backup_mb_border(s->top_border[mb_x + 1], dst[0],
dst[1], dst[2], s->linesize, s->uvlinesize, 0);
} }
if (s->filter.simple) if (s->filter.simple)
@ -1808,9 +1883,11 @@ static int vp8_decode_mb_row_sliced(AVCodecContext *avctx, void *tdata,
VP8ThreadData *next_td = NULL, *prev_td = NULL; VP8ThreadData *next_td = NULL, *prev_td = NULL;
VP8Frame *curframe = s->curframe; VP8Frame *curframe = s->curframe;
int mb_y, num_jobs = s->num_jobs; int mb_y, num_jobs = s->num_jobs;
td->thread_nr = threadnr; td->thread_nr = threadnr;
for (mb_y = jobnr; mb_y < s->mb_height; mb_y += num_jobs) { for (mb_y = jobnr; mb_y < s->mb_height; mb_y += num_jobs) {
if (mb_y >= s->mb_height) break; if (mb_y >= s->mb_height)
break;
td->thread_mb_pos = mb_y << 16; td->thread_mb_pos = mb_y << 16;
vp8_decode_mb_row_no_filter(avctx, tdata, jobnr, threadnr); vp8_decode_mb_row_no_filter(avctx, tdata, jobnr, threadnr);
if (s->deblock_filter) if (s->deblock_filter)
@ -1840,11 +1917,12 @@ int ff_vp8_decode_frame(AVCodecContext *avctx, void *data, int *got_frame,
prev_frame = s->framep[VP56_FRAME_CURRENT]; prev_frame = s->framep[VP56_FRAME_CURRENT];
referenced = s->update_last || s->update_golden == VP56_FRAME_CURRENT referenced = s->update_last || s->update_golden == VP56_FRAME_CURRENT ||
|| s->update_altref == VP56_FRAME_CURRENT; s->update_altref == VP56_FRAME_CURRENT;
skip_thresh = !referenced ? AVDISCARD_NONREF : skip_thresh = !referenced ? AVDISCARD_NONREF
!s->keyframe ? AVDISCARD_NONKEY : AVDISCARD_ALL; : !s->keyframe ? AVDISCARD_NONKEY
: AVDISCARD_ALL;
if (avctx->skip_frame >= skip_thresh) { if (avctx->skip_frame >= skip_thresh) {
s->invisible = 1; s->invisible = 1;
@ -1878,40 +1956,43 @@ int ff_vp8_decode_frame(AVCodecContext *avctx, void *data, int *got_frame,
if (curframe->tf.f->data[0]) if (curframe->tf.f->data[0])
vp8_release_frame(s, curframe); vp8_release_frame(s, curframe);
// Given that arithmetic probabilities are updated every frame, it's quite likely /* Given that arithmetic probabilities are updated every frame, it's quite
// that the values we have on a random interframe are complete junk if we didn't * likely that the values we have on a random interframe are complete
// start decode on a keyframe. So just don't display anything rather than junk. * junk if we didn't start decode on a keyframe. So just don't display
* anything rather than junk. */
if (!s->keyframe && (!s->framep[VP56_FRAME_PREVIOUS] || if (!s->keyframe && (!s->framep[VP56_FRAME_PREVIOUS] ||
!s->framep[VP56_FRAME_GOLDEN] || !s->framep[VP56_FRAME_GOLDEN] ||
!s->framep[VP56_FRAME_GOLDEN2])) { !s->framep[VP56_FRAME_GOLDEN2])) {
av_log(avctx, AV_LOG_WARNING, "Discarding interframe without a prior keyframe!\n"); av_log(avctx, AV_LOG_WARNING,
"Discarding interframe without a prior keyframe!\n");
ret = AVERROR_INVALIDDATA; ret = AVERROR_INVALIDDATA;
goto err; goto err;
} }
curframe->tf.f->key_frame = s->keyframe; curframe->tf.f->key_frame = s->keyframe;
curframe->tf.f->pict_type = s->keyframe ? AV_PICTURE_TYPE_I : AV_PICTURE_TYPE_P; curframe->tf.f->pict_type = s->keyframe ? AV_PICTURE_TYPE_I
: AV_PICTURE_TYPE_P;
if ((ret = vp8_alloc_frame(s, curframe, referenced))) { if ((ret = vp8_alloc_frame(s, curframe, referenced))) {
av_log(avctx, AV_LOG_ERROR, "get_buffer() failed!\n"); av_log(avctx, AV_LOG_ERROR, "get_buffer() failed!\n");
goto err; goto err;
} }
// check if golden and altref are swapped // check if golden and altref are swapped
if (s->update_altref != VP56_FRAME_NONE) { if (s->update_altref != VP56_FRAME_NONE)
s->next_framep[VP56_FRAME_GOLDEN2] = s->framep[s->update_altref]; s->next_framep[VP56_FRAME_GOLDEN2] = s->framep[s->update_altref];
} else { else
s->next_framep[VP56_FRAME_GOLDEN2] = s->framep[VP56_FRAME_GOLDEN2]; s->next_framep[VP56_FRAME_GOLDEN2] = s->framep[VP56_FRAME_GOLDEN2];
}
if (s->update_golden != VP56_FRAME_NONE) { if (s->update_golden != VP56_FRAME_NONE)
s->next_framep[VP56_FRAME_GOLDEN] = s->framep[s->update_golden]; s->next_framep[VP56_FRAME_GOLDEN] = s->framep[s->update_golden];
} else { else
s->next_framep[VP56_FRAME_GOLDEN] = s->framep[VP56_FRAME_GOLDEN]; s->next_framep[VP56_FRAME_GOLDEN] = s->framep[VP56_FRAME_GOLDEN];
}
if (s->update_last) { if (s->update_last)
s->next_framep[VP56_FRAME_PREVIOUS] = curframe; s->next_framep[VP56_FRAME_PREVIOUS] = curframe;
} else { else
s->next_framep[VP56_FRAME_PREVIOUS] = s->framep[VP56_FRAME_PREVIOUS]; s->next_framep[VP56_FRAME_PREVIOUS] = s->framep[VP56_FRAME_PREVIOUS];
}
s->next_framep[VP56_FRAME_CURRENT] = curframe; s->next_framep[VP56_FRAME_CURRENT] = curframe;
ff_thread_finish_setup(avctx); ff_thread_finish_setup(avctx);
@ -1920,15 +2001,16 @@ int ff_vp8_decode_frame(AVCodecContext *avctx, void *data, int *got_frame,
s->uvlinesize = curframe->tf.f->linesize[1]; s->uvlinesize = curframe->tf.f->linesize[1];
memset(s->top_nnz, 0, s->mb_width * sizeof(*s->top_nnz)); memset(s->top_nnz, 0, s->mb_width * sizeof(*s->top_nnz));
/* Zero macroblock structures for top/top-left prediction from outside the frame. */ /* Zero macroblock structures for top/top-left prediction
* from outside the frame. */
if (!s->mb_layout) if (!s->mb_layout)
memset(s->macroblocks + s->mb_height*2 - 1, 0, (s->mb_width+1)*sizeof(*s->macroblocks)); memset(s->macroblocks + s->mb_height * 2 - 1, 0,
(s->mb_width + 1) * sizeof(*s->macroblocks));
if (!s->mb_layout && s->keyframe) if (!s->mb_layout && s->keyframe)
memset(s->intra4x4_pred_mode_top, DC_PRED, s->mb_width * 4); memset(s->intra4x4_pred_mode_top, DC_PRED, s->mb_width * 4);
memset(s->ref_count, 0, sizeof(s->ref_count)); memset(s->ref_count, 0, sizeof(s->ref_count));
if (s->mb_layout == 1) { if (s->mb_layout == 1) {
// Make sure the previous frame has read its segmentation map, // Make sure the previous frame has read its segmentation map,
// if we re-use the same map. // if we re-use the same map.
@ -1951,7 +2033,8 @@ int ff_vp8_decode_frame(AVCodecContext *avctx, void *data, int *got_frame,
s->thread_data[i].thread_mb_pos = 0; s->thread_data[i].thread_mb_pos = 0;
s->thread_data[i].wait_mb_pos = INT_MAX; s->thread_data[i].wait_mb_pos = INT_MAX;
} }
avctx->execute2(avctx, vp8_decode_mb_row_sliced, s->thread_data, NULL, num_jobs); avctx->execute2(avctx, vp8_decode_mb_row_sliced,
s->thread_data, NULL, num_jobs);
ff_thread_report_progress(&curframe->tf, INT_MAX, 0); ff_thread_report_progress(&curframe->tf, INT_MAX, 0);
memcpy(&s->framep[0], &s->next_framep[0], sizeof(s->framep[0]) * 4); memcpy(&s->framep[0], &s->next_framep[0], sizeof(s->framep[0]) * 4);
@ -2033,10 +2116,10 @@ static av_cold int vp8_decode_init_thread_copy(AVCodecContext *avctx)
return 0; return 0;
} }
#define REBASE(pic) \ #define REBASE(pic) pic ? pic - &s_src->frames[0] + &s->frames[0] : NULL
pic ? pic - &s_src->frames[0] + &s->frames[0] : NULL
static int vp8_decode_update_thread_context(AVCodecContext *dst, const AVCodecContext *src) static int vp8_decode_update_thread_context(AVCodecContext *dst,
const AVCodecContext *src)
{ {
VP8Context *s = dst->priv_data, *s_src = src->priv_data; VP8Context *s = dst->priv_data, *s_src = src->priv_data;
int i; int i;

7
libavcodec/vp8.h
View File

@ -28,10 +28,11 @@
#include "libavutil/buffer.h" #include "libavutil/buffer.h"
#include "vp56.h"
#include "vp8dsp.h"
#include "h264pred.h" #include "h264pred.h"
#include "thread.h" #include "thread.h"
#include "vp56.h"
#include "vp8dsp.h"
#if HAVE_PTHREADS #if HAVE_PTHREADS
# include <pthread.h> # include <pthread.h>
#elif HAVE_W32THREADS #elif HAVE_W32THREADS
@ -82,7 +83,7 @@ typedef struct VP8FilterStrength {
typedef struct VP8Macroblock { typedef struct VP8Macroblock {
uint8_t skip; uint8_t skip;
// todo: make it possible to check for at least (i4x4 or split_mv) // TODO: make it possible to check for at least (i4x4 or split_mv)
// in one op. are others needed? // in one op. are others needed?
uint8_t mode; uint8_t mode;
uint8_t ref_frame; uint8_t ref_frame;

3
libavcodec/vp8_parser.c
View File

@ -25,7 +25,8 @@ static int parse(AVCodecParserContext *s,
const uint8_t **poutbuf, int *poutbuf_size, const uint8_t **poutbuf, int *poutbuf_size,
const uint8_t *buf, int buf_size) const uint8_t *buf, int buf_size)
{ {
s->pict_type= (buf[0]&0x01) ? AV_PICTURE_TYPE_P : AV_PICTURE_TYPE_I; s->pict_type = (buf[0] & 0x01) ? AV_PICTURE_TYPE_P
: AV_PICTURE_TYPE_I;
*poutbuf = buf; *poutbuf = buf;
*poutbuf_size = buf_size; *poutbuf_size = buf_size;

116
libavcodec/vp8data.h
View File

@ -30,24 +30,21 @@
#include "vp8.h" #include "vp8.h"
#include "h264pred.h" #include "h264pred.h"
static const uint8_t vp8_pred4x4_mode[] = static const uint8_t vp8_pred4x4_mode[] = {
{
[DC_PRED8x8] = DC_PRED, [DC_PRED8x8] = DC_PRED,
[VERT_PRED8x8] = VERT_PRED, [VERT_PRED8x8] = VERT_PRED,
[HOR_PRED8x8] = HOR_PRED, [HOR_PRED8x8] = HOR_PRED,
[PLANE_PRED8x8] = TM_VP8_PRED, [PLANE_PRED8x8] = TM_VP8_PRED,
}; };
static const int8_t vp8_pred16x16_tree_intra[4][2] = static const int8_t vp8_pred16x16_tree_intra[4][2] = {
{
{ -MODE_I4x4, 1 }, // '0' { -MODE_I4x4, 1 }, // '0'
{ 2, 3 }, { 2, 3 },
{ -DC_PRED8x8, -VERT_PRED8x8 }, // '100', '101' { -DC_PRED8x8, -VERT_PRED8x8 }, // '100', '101'
{ -HOR_PRED8x8, -PLANE_PRED8x8 }, // '110', '111' { -HOR_PRED8x8, -PLANE_PRED8x8 }, // '110', '111'
}; };
static const int8_t vp8_pred16x16_tree_inter[4][2] = static const int8_t vp8_pred16x16_tree_inter[4][2] = {
{
{ -DC_PRED8x8, 1 }, // '0' { -DC_PRED8x8, 1 }, // '0'
{ 2, 3 }, { 2, 3 },
{ -VERT_PRED8x8, -HOR_PRED8x8 }, // '100', '101' { -VERT_PRED8x8, -HOR_PRED8x8 }, // '100', '101'
@ -64,26 +61,26 @@ static const int vp8_mode_contexts[6][4] = {
}; };
static const uint8_t vp8_mbsplits[5][16] = { static const uint8_t vp8_mbsplits[5][16] = {
{ 0, 0, 0, 0, 0, 0, 0, 0, { 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1 },
1, 1, 1, 1, 1, 1, 1, 1 }, { 0, 0, 1, 1, 0, 0, 1, 1, 0, 0, 1, 1, 0, 0, 1, 1 },
{ 0, 0, 1, 1, 0, 0, 1, 1, { 0, 0, 1, 1, 0, 0, 1, 1, 2, 2, 3, 3, 2, 2, 3, 3 },
0, 0, 1, 1, 0, 0, 1, 1 }, { 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15 },
{ 0, 0, 1, 1, 0, 0, 1, 1, { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 }
2, 2, 3, 3, 2, 2, 3, 3 },
{ 0, 1, 2, 3, 4, 5, 6, 7,
8, 9, 10, 11, 12, 13, 14, 15 },
{ 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0 }
}; };
static const uint8_t vp8_mbfirstidx[4][16] = { static const uint8_t vp8_mbfirstidx[4][16] = {
{ 0, 8 }, { 0, 2 }, { 0, 2, 8, 10 }, { 0, 8 },
{ 0, 1, 2, 3, 4, 5, 6, 7, { 0, 2 },
8, 9, 10, 11, 12, 13, 14, 15 } { 0, 2, 8, 10 },
{ 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15 }
}; };
static const uint8_t vp8_mbsplit_count[4] = { 2, 2, 4, 16 }; static const uint8_t vp8_mbsplit_count[4] = {
static const uint8_t vp8_mbsplit_prob[3] = { 110, 111, 150 }; 2, 2, 4, 16
};
static const uint8_t vp8_mbsplit_prob[3] = {
110, 111, 150
};
static const uint8_t vp8_submv_prob[5][3] = { static const uint8_t vp8_submv_prob[5][3] = {
{ 147, 136, 18 }, { 147, 136, 18 },
@ -93,11 +90,14 @@ static const uint8_t vp8_submv_prob[5][3] = {
{ 208, 1, 1 } { 208, 1, 1 }
}; };
static const uint8_t vp8_pred16x16_prob_intra[4] = { 145, 156, 163, 128 }; static const uint8_t vp8_pred16x16_prob_intra[4] = {
static const uint8_t vp8_pred16x16_prob_inter[4] = { 112, 86, 140, 37 }; 145, 156, 163, 128
};
static const uint8_t vp8_pred16x16_prob_inter[4] = {
112, 86, 140, 37
};
static const int8_t vp8_pred4x4_tree[9][2] = static const int8_t vp8_pred4x4_tree[9][2] = {
{
{ -DC_PRED, 1 }, // '0' { -DC_PRED, 1 }, // '0'
{ -TM_VP8_PRED, 2 }, // '10' { -TM_VP8_PRED, 2 }, // '10'
{ -VERT_PRED, 3 }, // '110' { -VERT_PRED, 3 }, // '110'
@ -109,23 +109,23 @@ static const int8_t vp8_pred4x4_tree[9][2] =
{ -HOR_DOWN_PRED, -HOR_UP_PRED }, // '1111110', '1111111' { -HOR_DOWN_PRED, -HOR_UP_PRED }, // '1111110', '1111111'
}; };
static const int8_t vp8_pred8x8c_tree[3][2] = static const int8_t vp8_pred8x8c_tree[3][2] = {
{
{ -DC_PRED8x8, 1 }, // '0' { -DC_PRED8x8, 1 }, // '0'
{ -VERT_PRED8x8, 2 }, // '10 { -VERT_PRED8x8, 2 }, // '10
{ -HOR_PRED8x8, -PLANE_PRED8x8 }, // '110', '111' { -HOR_PRED8x8, -PLANE_PRED8x8 }, // '110', '111'
}; };
static const uint8_t vp8_pred8x8c_prob_intra[3] = { 142, 114, 183 }; static const uint8_t vp8_pred8x8c_prob_intra[3] = {
static const uint8_t vp8_pred8x8c_prob_inter[3] = { 162, 101, 204 }; 142, 114, 183
};
static const uint8_t vp8_pred4x4_prob_inter[9] = static const uint8_t vp8_pred8x8c_prob_inter[3] = {
{ 162, 101, 204
};
static const uint8_t vp8_pred4x4_prob_inter[9] = {
120, 90, 79, 133, 87, 85, 80, 111, 151 120, 90, 79, 133, 87, 85, 80, 111, 151
}; };
static const uint8_t vp8_pred4x4_prob_intra[10][10][9] = static const uint8_t vp8_pred4x4_prob_intra[10][10][9] = {
{
{ {
{ 39, 53, 200, 87, 26, 21, 43, 232, 171 }, { 39, 53, 200, 87, 26, 21, 43, 232, 171 },
{ 56, 34, 51, 104, 114, 102, 29, 93, 77 }, { 56, 34, 51, 104, 114, 102, 29, 93, 77 },
@ -248,22 +248,19 @@ static const uint8_t vp8_pred4x4_prob_intra[10][10][9] =
}, },
}; };
static const int8_t vp8_segmentid_tree[][2] = static const int8_t vp8_segmentid_tree[][2] = {
{
{ 1, 2 }, { 1, 2 },
{ -0, -1 }, // '00', '01' { -0, -1 }, // '00', '01'
{ -2, -3 }, // '10', '11' { -2, -3 }, // '10', '11'
}; };
static const uint8_t vp8_coeff_band[16] = static const uint8_t vp8_coeff_band[16] = {
{
0, 1, 2, 3, 6, 4, 5, 6, 6, 6, 6, 6, 6, 6, 6, 7 0, 1, 2, 3, 6, 4, 5, 6, 6, 6, 6, 6, 6, 6, 6, 7
}; };
/* Inverse of vp8_coeff_band: mappings of bands to coefficient indexes. /* Inverse of vp8_coeff_band: mappings of bands to coefficient indexes.
* Each list is -1-terminated. */ * Each list is -1-terminated. */
static const int8_t vp8_coeff_band_indexes[8][10] = static const int8_t vp8_coeff_band_indexes[8][10] = {
{
{ 0, -1 }, { 0, -1 },
{ 1, -1 }, { 1, -1 },
{ 2, -1 }, { 2, -1 },
@ -274,24 +271,34 @@ static const int8_t vp8_coeff_band_indexes[8][10] =
{ 15, -1 } { 15, -1 }
}; };
static const uint8_t vp8_dct_cat1_prob[] = { 159, 0 }; static const uint8_t vp8_dct_cat1_prob[] = {
static const uint8_t vp8_dct_cat2_prob[] = { 165, 145, 0 }; 159, 0
static const uint8_t vp8_dct_cat3_prob[] = { 173, 148, 140, 0 }; };
static const uint8_t vp8_dct_cat4_prob[] = { 176, 155, 140, 135, 0 }; static const uint8_t vp8_dct_cat2_prob[] = {
static const uint8_t vp8_dct_cat5_prob[] = { 180, 157, 141, 134, 130, 0 }; 165, 145, 0
static const uint8_t vp8_dct_cat6_prob[] = { 254, 254, 243, 230, 196, 177, 153, 140, 133, 130, 129, 0 }; };
static const uint8_t vp8_dct_cat3_prob[] = {
173, 148, 140, 0
};
static const uint8_t vp8_dct_cat4_prob[] = {
176, 155, 140, 135, 0
};
static const uint8_t vp8_dct_cat5_prob[] = {
180, 157, 141, 134, 130, 0
};
static const uint8_t vp8_dct_cat6_prob[] = {
254, 254, 243, 230, 196, 177, 153, 140, 133, 130, 129, 0
};
// only used for cat3 and above; cat 1 and 2 are referenced directly // only used for cat3 and above; cat 1 and 2 are referenced directly
const uint8_t * const ff_vp8_dct_cat_prob[] = const uint8_t *const ff_vp8_dct_cat_prob[] = {
{
vp8_dct_cat3_prob, vp8_dct_cat3_prob,
vp8_dct_cat4_prob, vp8_dct_cat4_prob,
vp8_dct_cat5_prob, vp8_dct_cat5_prob,
vp8_dct_cat6_prob, vp8_dct_cat6_prob,
}; };
static const uint8_t vp8_token_default_probs[4][8][3][NUM_DCT_TOKENS-1] = static const uint8_t vp8_token_default_probs[4][8][3][NUM_DCT_TOKENS - 1] = {
{
{ {
{ {
{ 128, 128, 128, 128, 128, 128, 128, 128, 128, 128, 128 }, { 128, 128, 128, 128, 128, 128, 128, 128, 128, 128, 128 },
@ -462,8 +469,7 @@ static const uint8_t vp8_token_default_probs[4][8][3][NUM_DCT_TOKENS-1] =
}, },
}; };
static const uint8_t vp8_token_update_probs[4][8][3][NUM_DCT_TOKENS-1] = static const uint8_t vp8_token_update_probs[4][8][3][NUM_DCT_TOKENS - 1] = {
{
{ {
{ {
{ 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255 }, { 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255 },
@ -642,8 +648,7 @@ static const uint8_t zigzag_scan[16]={
3 + 1 * 4, 3 + 2 * 4, 2 + 3 * 4, 3 + 3 * 4, 3 + 1 * 4, 3 + 2 * 4, 2 + 3 * 4, 3 + 3 * 4,
}; };
static const uint8_t vp8_dc_qlookup[VP8_MAX_QUANT+1] = static const uint8_t vp8_dc_qlookup[VP8_MAX_QUANT + 1] = {
{
4, 5, 6, 7, 8, 9, 10, 10, 11, 12, 13, 14, 15, 16, 17, 17, 4, 5, 6, 7, 8, 9, 10, 10, 11, 12, 13, 14, 15, 16, 17, 17,
18, 19, 20, 20, 21, 21, 22, 22, 23, 23, 24, 25, 25, 26, 27, 28, 18, 19, 20, 20, 21, 21, 22, 22, 23, 23, 24, 25, 25, 26, 27, 28,
29, 30, 31, 32, 33, 34, 35, 36, 37, 37, 38, 39, 40, 41, 42, 43, 29, 30, 31, 32, 33, 34, 35, 36, 37, 37, 38, 39, 40, 41, 42, 43,
@ -654,8 +659,7 @@ static const uint8_t vp8_dc_qlookup[VP8_MAX_QUANT+1] =
122, 124, 126, 128, 130, 132, 134, 136, 138, 140, 143, 145, 148, 151, 154, 157, 122, 124, 126, 128, 130, 132, 134, 136, 138, 140, 143, 145, 148, 151, 154, 157,
}; };
static const uint16_t vp8_ac_qlookup[VP8_MAX_QUANT+1] = static const uint16_t vp8_ac_qlookup[VP8_MAX_QUANT + 1] = {
{
4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19,
20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35,
36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51,

112
libavcodec/vp8dsp.c
View File

@ -24,9 +24,10 @@
* VP8 compatible video decoder * VP8 compatible video decoder
*/ */
#include "libavutil/common.h"
#include "mathops.h" #include "mathops.h"
#include "vp8dsp.h" #include "vp8dsp.h"
#include "libavutil/common.h"
// TODO: Maybe add dequant // TODO: Maybe add dequant
static void vp8_luma_dc_wht_c(int16_t block[4][4][16], int16_t dc[16]) static void vp8_luma_dc_wht_c(int16_t block[4][4][16], int16_t dc[16])
@ -127,7 +128,8 @@ static void vp8_idct_dc_add_c(uint8_t *dst, int16_t block[16], ptrdiff_t stride)
} }
} }
static void vp8_idct_dc_add4uv_c(uint8_t *dst, int16_t block[4][16], ptrdiff_t stride) static void vp8_idct_dc_add4uv_c(uint8_t *dst, int16_t block[4][16],
ptrdiff_t stride)
{ {
vp8_idct_dc_add_c(dst + stride * 0 + 0, block[0], stride); vp8_idct_dc_add_c(dst + stride * 0 + 0, block[0], stride);
vp8_idct_dc_add_c(dst + stride * 0 + 4, block[1], stride); vp8_idct_dc_add_c(dst + stride * 0 + 4, block[1], stride);
@ -135,7 +137,8 @@ static void vp8_idct_dc_add4uv_c(uint8_t *dst, int16_t block[4][16], ptrdiff_t s
vp8_idct_dc_add_c(dst + stride * 4 + 4, block[3], stride); vp8_idct_dc_add_c(dst + stride * 4 + 4, block[3], stride);
} }
static void vp8_idct_dc_add4y_c(uint8_t *dst, int16_t block[4][16], ptrdiff_t stride) static void vp8_idct_dc_add4y_c(uint8_t *dst, int16_t block[4][16],
ptrdiff_t stride)
{ {
vp8_idct_dc_add_c(dst + 0, block[0], stride); vp8_idct_dc_add_c(dst + 0, block[0], stride);
vp8_idct_dc_add_c(dst + 4, block[1], stride); vp8_idct_dc_add_c(dst + 4, block[1], stride);
@ -156,7 +159,8 @@ static void vp8_idct_dc_add4y_c(uint8_t *dst, int16_t block[4][16], ptrdiff_t st
#define clip_int8(n) (cm[n + 0x80] - 0x80) #define clip_int8(n) (cm[n + 0x80] - 0x80)
static av_always_inline void filter_common(uint8_t *p, ptrdiff_t stride, int is4tap) static av_always_inline void filter_common(uint8_t *p, ptrdiff_t stride,
int is4tap)
{ {
LOAD_PIXELS LOAD_PIXELS
int a, f1, f2; int a, f1, f2;
@ -197,12 +201,17 @@ static av_always_inline int simple_limit(uint8_t *p, ptrdiff_t stride, int flim)
* E - limit at the macroblock edge * E - limit at the macroblock edge
* I - limit for interior difference * I - limit for interior difference
*/ */
static av_always_inline int normal_limit(uint8_t *p, ptrdiff_t stride, int E, int I) static av_always_inline int normal_limit(uint8_t *p, ptrdiff_t stride,
int E, int I)
{ {
LOAD_PIXELS LOAD_PIXELS
return simple_limit(p, stride, E) return simple_limit(p, stride, E) &&
&& FFABS(p3-p2) <= I && FFABS(p2-p1) <= I && FFABS(p1-p0) <= I FFABS(p3 - p2) <= I &&
&& FFABS(q3-q2) <= I && FFABS(q2-q1) <= I && FFABS(q1-q0) <= I; FFABS(p2 - p1) <= I &&
FFABS(p1 - p0) <= I &&
FFABS(q3 - q2) <= I &&
FFABS(q2 - q1) <= I &&
FFABS(q1 - q0) <= I;
} }
// high edge variance // high edge variance
@ -235,11 +244,13 @@ static av_always_inline void filter_mbedge(uint8_t *p, ptrdiff_t stride)
} }
#define LOOP_FILTER(dir, size, stridea, strideb, maybe_inline) \ #define LOOP_FILTER(dir, size, stridea, strideb, maybe_inline) \
static maybe_inline void vp8_ ## dir ## _loop_filter ## size ## _c(uint8_t *dst, ptrdiff_t stride,\ static maybe_inline \
int flim_E, int flim_I, int hev_thresh)\ void vp8_ ## dir ## _loop_filter ## size ## _c(uint8_t *dst, \
ptrdiff_t stride, \
int flim_E, int flim_I, \
int hev_thresh) \
{ \ { \
int i; \ int i; \
\
for (i = 0; i < size; i++) \ for (i = 0; i < size; i++) \
if (normal_limit(dst + i * stridea, strideb, flim_E, flim_I)) { \ if (normal_limit(dst + i * stridea, strideb, flim_E, flim_I)) { \
if (hev(dst + i * stridea, strideb, hev_thresh)) \ if (hev(dst + i * stridea, strideb, hev_thresh)) \
@ -249,11 +260,13 @@ static maybe_inline void vp8_ ## dir ## _loop_filter ## size ## _c(uint8_t *dst,
} \ } \
} \ } \
\ \
static maybe_inline void vp8_ ## dir ## _loop_filter ## size ## _inner_c(uint8_t *dst, ptrdiff_t stride,\ static maybe_inline \
int flim_E, int flim_I, int hev_thresh)\ void vp8_ ## dir ## _loop_filter ## size ## _inner_c(uint8_t *dst, \
ptrdiff_t stride, \
int flim_E, int flim_I, \
int hev_thresh) \
{ \ { \
int i; \ int i; \
\
for (i = 0; i < size; i++) \ for (i = 0; i < size; i++) \
if (normal_limit(dst + i * stridea, strideb, flim_E, flim_I)) { \ if (normal_limit(dst + i * stridea, strideb, flim_E, flim_I)) { \
int hv = hev(dst + i * stridea, strideb, hev_thresh); \ int hv = hev(dst + i * stridea, strideb, hev_thresh); \
@ -269,14 +282,18 @@ LOOP_FILTER(h, 16, stride, 1,)
#define UV_LOOP_FILTER(dir, stridea, strideb) \ #define UV_LOOP_FILTER(dir, stridea, strideb) \
LOOP_FILTER(dir, 8, stridea, strideb, av_always_inline) \ LOOP_FILTER(dir, 8, stridea, strideb, av_always_inline) \
static void vp8_ ## dir ## _loop_filter8uv_c(uint8_t *dstU, uint8_t *dstV, ptrdiff_t stride,\ static void vp8_ ## dir ## _loop_filter8uv_c(uint8_t *dstU, uint8_t *dstV, \
int fE, int fI, int hev_thresh)\ ptrdiff_t stride, int fE, \
int fI, int hev_thresh) \
{ \ { \
vp8_ ## dir ## _loop_filter8_c(dstU, stride, fE, fI, hev_thresh); \ vp8_ ## dir ## _loop_filter8_c(dstU, stride, fE, fI, hev_thresh); \
vp8_ ## dir ## _loop_filter8_c(dstV, stride, fE, fI, hev_thresh); \ vp8_ ## dir ## _loop_filter8_c(dstV, stride, fE, fI, hev_thresh); \
} \ } \
static void vp8_ ## dir ## _loop_filter8uv_inner_c(uint8_t *dstU, uint8_t *dstV, ptrdiff_t stride,\ \
int fE, int fI, int hev_thresh)\ static void vp8_ ## dir ## _loop_filter8uv_inner_c(uint8_t *dstU, \
uint8_t *dstV, \
ptrdiff_t stride, int fE, \
int fI, int hev_thresh) \
{ \ { \
vp8_ ## dir ## _loop_filter8_inner_c(dstU, stride, fE, fI, hev_thresh); \ vp8_ ## dir ## _loop_filter8_inner_c(dstU, stride, fE, fI, hev_thresh); \
vp8_ ## dir ## _loop_filter8_inner_c(dstV, stride, fE, fI, hev_thresh); \ vp8_ ## dir ## _loop_filter8_inner_c(dstV, stride, fE, fI, hev_thresh); \
@ -314,11 +331,13 @@ static const uint8_t subpel_filters[7][6] = {
}; };
#define PUT_PIXELS(WIDTH) \ #define PUT_PIXELS(WIDTH) \
static void put_vp8_pixels ## WIDTH ##_c(uint8_t *dst, ptrdiff_t dststride, uint8_t *src, ptrdiff_t srcstride, int h, int x, int y) { \ static void put_vp8_pixels ## WIDTH ## _c(uint8_t *dst, ptrdiff_t dststride, \
uint8_t *src, ptrdiff_t srcstride, \
int h, int x, int y) \
{ \
int i; \ int i; \
for (i = 0; i < h; i++, dst+= dststride, src+= srcstride) { \ for (i = 0; i < h; i++, dst += dststride, src += srcstride) \
memcpy(dst, src, WIDTH); \ memcpy(dst, src, WIDTH); \
} \
} }
PUT_PIXELS(16) PUT_PIXELS(16)
@ -326,20 +345,24 @@ PUT_PIXELS(8)
PUT_PIXELS(4) PUT_PIXELS(4)
#define FILTER_6TAP(src, F, stride) \ #define FILTER_6TAP(src, F, stride) \
cm[(F[2]*src[x+0*stride] - F[1]*src[x-1*stride] + F[0]*src[x-2*stride] + \ cm[(F[2] * src[x + 0 * stride] - F[1] * src[x - 1 * stride] + \
F[3]*src[x+1*stride] - F[4]*src[x+2*stride] + F[5]*src[x+3*stride] + 64) >> 7] F[0] * src[x - 2 * stride] + F[3] * src[x + 1 * stride] - \
F[4] * src[x + 2 * stride] + F[5] * src[x + 3 * stride] + 64) >> 7]
#define FILTER_4TAP(src, F, stride) \ #define FILTER_4TAP(src, F, stride) \
cm[(F[2] * src[x + 0 * stride] - F[1] * src[x - 1 * stride] + \ cm[(F[2] * src[x + 0 * stride] - F[1] * src[x - 1 * stride] + \
F[3] * src[x + 1 * stride] - F[4] * src[x + 2 * stride] + 64) >> 7] F[3] * src[x + 1 * stride] - F[4] * src[x + 2 * stride] + 64) >> 7]
#define VP8_EPEL_H(SIZE, TAPS) \ #define VP8_EPEL_H(SIZE, TAPS) \
static void put_vp8_epel ## SIZE ## _h ## TAPS ## _c(uint8_t *dst, ptrdiff_t dststride, uint8_t *src, ptrdiff_t srcstride, int h, int mx, int my) \ static void put_vp8_epel ## SIZE ## _h ## TAPS ## _c(uint8_t *dst, \
ptrdiff_t dststride, \
uint8_t *src, \
ptrdiff_t srcstride, \
int h, int mx, int my) \
{ \ { \
const uint8_t *filter = subpel_filters[mx - 1]; \ const uint8_t *filter = subpel_filters[mx - 1]; \
const uint8_t *cm = ff_crop_tab + MAX_NEG_CROP; \ const uint8_t *cm = ff_crop_tab + MAX_NEG_CROP; \
int x, y; \ int x, y; \
\
for (y = 0; y < h; y++) { \ for (y = 0; y < h; y++) { \
for (x = 0; x < SIZE; x++) \ for (x = 0; x < SIZE; x++) \
dst[x] = FILTER_ ## TAPS ## TAP(src, filter, 1); \ dst[x] = FILTER_ ## TAPS ## TAP(src, filter, 1); \
@ -347,13 +370,17 @@ static void put_vp8_epel ## SIZE ## _h ## TAPS ## _c(uint8_t *dst, ptrdiff_t dst
src += srcstride; \ src += srcstride; \
} \ } \
} }
#define VP8_EPEL_V(SIZE, TAPS) \ #define VP8_EPEL_V(SIZE, TAPS) \
static void put_vp8_epel ## SIZE ## _v ## TAPS ## _c(uint8_t *dst, ptrdiff_t dststride, uint8_t *src, ptrdiff_t srcstride, int h, int mx, int my) \ static void put_vp8_epel ## SIZE ## _v ## TAPS ## _c(uint8_t *dst, \
ptrdiff_t dststride, \
uint8_t *src, \
ptrdiff_t srcstride, \
int h, int mx, int my) \
{ \ { \
const uint8_t *filter = subpel_filters[my - 1]; \ const uint8_t *filter = subpel_filters[my - 1]; \
const uint8_t *cm = ff_crop_tab + MAX_NEG_CROP; \ const uint8_t *cm = ff_crop_tab + MAX_NEG_CROP; \
int x, y; \ int x, y; \
\
for (y = 0; y < h; y++) { \ for (y = 0; y < h; y++) { \
for (x = 0; x < SIZE; x++) \ for (x = 0; x < SIZE; x++) \
dst[x] = FILTER_ ## TAPS ## TAP(src, filter, srcstride); \ dst[x] = FILTER_ ## TAPS ## TAP(src, filter, srcstride); \
@ -361,8 +388,15 @@ static void put_vp8_epel ## SIZE ## _v ## TAPS ## _c(uint8_t *dst, ptrdiff_t dst
src += srcstride; \ src += srcstride; \
} \ } \
} }
#define VP8_EPEL_HV(SIZE, HTAPS, VTAPS) \ #define VP8_EPEL_HV(SIZE, HTAPS, VTAPS) \
static void put_vp8_epel ## SIZE ## _h ## HTAPS ## v ## VTAPS ## _c(uint8_t *dst, ptrdiff_t dststride, uint8_t *src, ptrdiff_t srcstride, int h, int mx, int my) \ static void \
put_vp8_epel ## SIZE ## _h ## HTAPS ## v ## VTAPS ## _c(uint8_t *dst, \
ptrdiff_t dststride, \
uint8_t *src, \
ptrdiff_t srcstride, \
int h, int mx, \
int my) \
{ \ { \
const uint8_t *filter = subpel_filters[mx - 1]; \ const uint8_t *filter = subpel_filters[mx - 1]; \
const uint8_t *cm = ff_crop_tab + MAX_NEG_CROP; \ const uint8_t *cm = ff_crop_tab + MAX_NEG_CROP; \
@ -377,7 +411,6 @@ static void put_vp8_epel ## SIZE ## _h ## HTAPS ## v ## VTAPS ## _c(uint8_t *dst
tmp += SIZE; \ tmp += SIZE; \
src += srcstride; \ src += srcstride; \
} \ } \
\
tmp = tmp_array + (2 - (VTAPS == 4)) * SIZE; \ tmp = tmp_array + (2 - (VTAPS == 4)) * SIZE; \
filter = subpel_filters[my - 1]; \ filter = subpel_filters[my - 1]; \
\ \
@ -401,6 +434,7 @@ VP8_EPEL_V(4, 4)
VP8_EPEL_V(16, 6) VP8_EPEL_V(16, 6)
VP8_EPEL_V(8, 6) VP8_EPEL_V(8, 6)
VP8_EPEL_V(4, 6) VP8_EPEL_V(4, 6)
VP8_EPEL_HV(16, 4, 4) VP8_EPEL_HV(16, 4, 4)
VP8_EPEL_HV(8, 4, 4) VP8_EPEL_HV(8, 4, 4)
VP8_EPEL_HV(4, 4, 4) VP8_EPEL_HV(4, 4, 4)
@ -415,11 +449,12 @@ VP8_EPEL_HV(8, 6, 6)
VP8_EPEL_HV(4, 6, 6) VP8_EPEL_HV(4, 6, 6)
#define VP8_BILINEAR(SIZE) \ #define VP8_BILINEAR(SIZE) \
static void put_vp8_bilinear ## SIZE ## _h_c(uint8_t *dst, ptrdiff_t dstride, uint8_t *src, ptrdiff_t sstride, int h, int mx, int my) \ static void put_vp8_bilinear ## SIZE ## _h_c(uint8_t *dst, ptrdiff_t dstride, \
uint8_t *src, ptrdiff_t sstride, \
int h, int mx, int my) \
{ \ { \
int a = 8 - mx, b = mx; \ int a = 8 - mx, b = mx; \
int x, y; \ int x, y; \
\
for (y = 0; y < h; y++) { \ for (y = 0; y < h; y++) { \
for (x = 0; x < SIZE; x++) \ for (x = 0; x < SIZE; x++) \
dst[x] = (a * src[x] + b * src[x + 1] + 4) >> 3; \ dst[x] = (a * src[x] + b * src[x + 1] + 4) >> 3; \
@ -427,11 +462,13 @@ static void put_vp8_bilinear ## SIZE ## _h_c(uint8_t *dst, ptrdiff_t dstride, ui
src += sstride; \ src += sstride; \
} \ } \
} \ } \
static void put_vp8_bilinear ## SIZE ## _v_c(uint8_t *dst, ptrdiff_t dstride, uint8_t *src, ptrdiff_t sstride, int h, int mx, int my) \ \
static void put_vp8_bilinear ## SIZE ## _v_c(uint8_t *dst, ptrdiff_t dstride, \
uint8_t *src, ptrdiff_t sstride, \
int h, int mx, int my) \
{ \ { \
int c = 8 - my, d = my; \ int c = 8 - my, d = my; \
int x, y; \ int x, y; \
\
for (y = 0; y < h; y++) { \ for (y = 0; y < h; y++) { \
for (x = 0; x < SIZE; x++) \ for (x = 0; x < SIZE; x++) \
dst[x] = (c * src[x] + d * src[x + sstride] + 4) >> 3; \ dst[x] = (c * src[x] + d * src[x + sstride] + 4) >> 3; \
@ -440,23 +477,24 @@ static void put_vp8_bilinear ## SIZE ## _v_c(uint8_t *dst, ptrdiff_t dstride, ui
} \ } \
} \ } \
\ \
static void put_vp8_bilinear ## SIZE ## _hv_c(uint8_t *dst, ptrdiff_t dstride, uint8_t *src, ptrdiff_t sstride, int h, int mx, int my) \ static void put_vp8_bilinear ## SIZE ## _hv_c(uint8_t *dst, \
ptrdiff_t dstride, \
uint8_t *src, \
ptrdiff_t sstride, \
int h, int mx, int my) \
{ \ { \
int a = 8 - mx, b = mx; \ int a = 8 - mx, b = mx; \
int c = 8 - my, d = my; \ int c = 8 - my, d = my; \
int x, y; \ int x, y; \
uint8_t tmp_array[(2 * SIZE + 1) * SIZE]; \ uint8_t tmp_array[(2 * SIZE + 1) * SIZE]; \
uint8_t *tmp = tmp_array; \ uint8_t *tmp = tmp_array; \
\
for (y = 0; y < h + 1; y++) { \ for (y = 0; y < h + 1; y++) { \
for (x = 0; x < SIZE; x++) \ for (x = 0; x < SIZE; x++) \
tmp[x] = (a * src[x] + b * src[x + 1] + 4) >> 3; \ tmp[x] = (a * src[x] + b * src[x + 1] + 4) >> 3; \
tmp += SIZE; \ tmp += SIZE; \
src += sstride; \ src += sstride; \
} \ } \
\
tmp = tmp_array; \ tmp = tmp_array; \
\
for (y = 0; y < h; y++) { \ for (y = 0; y < h; y++) { \
for (x = 0; x < SIZE; x++) \ for (x = 0; x < SIZE; x++) \
dst[x] = (c * tmp[x] + d * tmp[x + SIZE] + 4) >> 3; \ dst[x] = (c * tmp[x] + d * tmp[x + SIZE] + 4) >> 3; \