Skip to content

F
ffmpeg
Commit 53c20f17 authored by Vittorio Giovara's avatar Vittorio Giovara Committed by Diego Biurrun
Browse files
Options

vp8: K&R formatting cosmetics 

Signed-off-by: 's avatarDiego Biurrun <diego@biurrun.de>
parent 6adf3bc4
Hide whitespace changes
Inline Side-by-side
Showing with 934 additions and 807 deletions
libavcodec/vp8.c
View file @ 53c20f17
......@@ -24,12 +24,13 @@
*/
#include "libavutil/imgutils.h"
#include "avcodec.h"
#include "internal.h"
#include "vp8.h"
#include "vp8data.h"
#include "rectangle.h"
#include "thread.h"
#include "vp8.h"
#include "vp8data.h"
#if ARCH_ARM
# include "arm/vp8.h"
......@@ -91,7 +92,6 @@ static int vp8_ref_frame(VP8Context *s, VP8Frame *dst, VP8Frame *src)
return 0;
}
static void vp8_decode_flush_impl(AVCodecContext *avctx, int free_mem)
{
VP8Context *s = avctx->priv_data;
......@@ -124,22 +124,25 @@ static int update_dimensions(VP8Context *s, int width, int height)
return ret;
}
s->mb_width = (s->avctx->coded_width +15) / 16;
s->mb_height = (s->avctx->coded_height+15) / 16;
s->mb_width = (s->avctx->coded_width + 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
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);
}
else // Sliced threading
s->macroblocks_base = av_mallocz((s->mb_width+2)*(s->mb_height+2)*sizeof(*s->macroblocks));
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->thread_data = av_mallocz(MAX_THREADS*sizeof(VP8ThreadData));
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);
} else // Sliced threading
s->macroblocks_base = av_mallocz((s->mb_width + 2) * (s->mb_height + 2) *
sizeof(*s->macroblocks));
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->thread_data = av_mallocz(MAX_THREADS * sizeof(VP8ThreadData));
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
pthread_mutex_init(&s->thread_data[i].lock, NULL);
pthread_cond_init(&s->thread_data[i].cond, NULL);
......@@ -150,7 +153,7 @@ static int update_dimensions(VP8Context *s, int width, int height)
(!s->intra4x4_pred_mode_top && !s->mb_layout))
return AVERROR(ENOMEM);
s->macroblocks = s->macroblocks_base + 1;
s->macroblocks = s->macroblocks_base + 1;
return 0;
}
......@@ -207,13 +210,13 @@ static int setup_partitions(VP8Context *s, const uint8_t *buf, int buf_size)
s->num_coeff_partitions = 1 << vp8_rac_get_uint(&s->c, 2);
buf += 3*(s->num_coeff_partitions-1);
buf_size -= 3*(s->num_coeff_partitions-1);
buf += 3 * (s->num_coeff_partitions - 1);
buf_size -= 3 * (s->num_coeff_partitions - 1);
if (buf_size < 0)
return -1;
for (i = 0; i < s->num_coeff_partitions-1; i++) {
int size = AV_RL24(sizes + 3*i);
for (i = 0; i < s->num_coeff_partitions - 1; i++) {
int size = AV_RL24(sizes + 3 * i);
if (buf_size - size < 0)
return -1;
......@@ -246,13 +249,13 @@ static void get_quants(VP8Context *s)
} else
base_qi = yac_qi;
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_dc_qmul[0] = 2 * vp8_dc_qlookup[av_clip_uintp2(base_qi + y2dc_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_dc_qmul[0] = vp8_dc_qlookup[av_clip_uintp2(base_qi + y2dc_delta, 7)] * 2;
/* 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].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].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[1] = vp8_ac_qlookup[av_clip_uintp2(base_qi + uvac_delta, 7)];
s->qmat[i].luma_dc_qmul[1] = FFMAX(s->qmat[i].luma_dc_qmul[1], 8);
s->qmat[i].chroma_qmul[0] = FFMIN(s->qmat[i].chroma_qmul[0], 132);
......@@ -317,24 +320,27 @@ 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);
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
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");
return AVERROR_INVALIDDATA;
}
if (s->keyframe) {
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;
}
width = AV_RL16(buf+3) & 0x3fff;
height = AV_RL16(buf+5) & 0x3fff;
hscale = buf[4] >> 6;
vscale = buf[6] >> 6;
width = AV_RL16(buf + 3) & 0x3fff;
height = AV_RL16(buf + 5) & 0x3fff;
hscale = buf[4] >> 6;
vscale = buf[6] >> 6;
buf += 7;
buf_size -= 7;
......@@ -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;
for (i = 0; i < 4; i++)
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]));
memcpy(s->prob->pred16x16, vp8_pred16x16_prob_inter, sizeof(s->prob->pred16x16));
memcpy(s->prob->pred8x8c , vp8_pred8x8c_prob_inter , sizeof(s->prob->pred8x8c));
memcpy(s->prob->mvc , vp8_mv_default_prob , sizeof(s->prob->mvc));
memcpy(s->prob->pred16x16, vp8_pred16x16_prob_inter,
sizeof(s->prob->pred16x16));
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->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 */
width != s->avctx->width || height != s->avctx->height) {
width != s->avctx->width || height != s->avctx->height)
if ((ret = update_dimensions(s, width, height)) < 0)
return ret;
}
get_quants(s);
......@@ -405,7 +414,7 @@ static int decode_frame_header(VP8Context *s, const uint8_t *buf, int buf_size)
for (i = 0; i < 4; i++)
for (j = 0; j < 8; j++)
for (k = 0; k < 3; k++)
for (l = 0; l < NUM_DCT_TOKENS-1; l++)
for (l = 0; l < NUM_DCT_TOKENS - 1; l++)
if (vp56_rac_get_prob_branchy(c, vp8_token_update_probs[i][j][k][l])) {
int prob = vp8_rac_get_uint(c, 8);
for (m = 0; vp8_coeff_band_indexes[j][m] >= 0; m++)
......@@ -437,7 +446,8 @@ static int decode_frame_header(VP8Context *s, const uint8_t *buf, int buf_size)
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->y = av_clip(src->y, s->mv_min.y, s->mv_max.y);
......@@ -461,13 +471,13 @@ static int read_mv_component(VP56RangeCoder *c, const uint8_t *p)
x += 8;
} else {
// small_mvtree
const uint8_t *ps = p+2;
const uint8_t *ps = p + 2;
bit = vp56_rac_get_prob(c, *ps);
ps += 1 + 3*bit;
x += 4*bit;
ps += 1 + 3 * bit;
x += 4 * bit;
bit = vp56_rac_get_prob(c, *ps);
ps += 1 + bit;
x += 2*bit;
x += 2 * bit;
x += vp56_rac_get_prob(c, *ps);
}
......@@ -478,10 +488,10 @@ static av_always_inline
const uint8_t *get_submv_prob(uint32_t left, uint32_t top)
{
if (left == top)
return vp8_submv_prob[4-!!left];
return vp8_submv_prob[4 - !!left];
if (!top)
return vp8_submv_prob[2];
return vp8_submv_prob[1-!!left];
return vp8_submv_prob[1 - !!left];
}
/**
......@@ -495,9 +505,8 @@ int decode_splitmvs(VP8Context *s, VP56RangeCoder *c, VP8Macroblock *mb, int lay
int n, num;
VP8Macroblock *top_mb;
VP8Macroblock *left_mb = &mb[-1];
const uint8_t *mbsplits_left = vp8_mbsplits[left_mb->partitioning],
*mbsplits_top,
*mbsplits_cur, *firstidx;
const uint8_t *mbsplits_left = vp8_mbsplits[left_mb->partitioning];
const uint8_t *mbsplits_top, *mbsplits_cur, *firstidx;
VP56mv *top_mv;
VP56mv *left_mv = left_mb->bmv;
VP56mv *cur_mv = mb->bmv;
......@@ -505,23 +514,22 @@ int decode_splitmvs(VP8Context *s, VP56RangeCoder *c, VP8Macroblock *mb, int lay
if (!layout) // layout is inlined, s->mb_layout is not
top_mb = &mb[2];
else
top_mb = &mb[-s->mb_width-1];
top_mb = &mb[-s->mb_width - 1];
mbsplits_top = vp8_mbsplits[top_mb->partitioning];
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[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]);
} else {
else
part_idx = VP8_SPLITMVMODE_8x8;
}
} else {
part_idx = VP8_SPLITMVMODE_4x4;
}
num = vp8_mbsplit_count[part_idx];
mbsplits_cur = vp8_mbsplits[part_idx],
firstidx = vp8_mbfirstidx[part_idx];
num = vp8_mbsplit_count[part_idx];
mbsplits_cur = vp8_mbsplits[part_idx],
firstidx = vp8_mbfirstidx[part_idx];
mb->partitioning = part_idx;
for (n = 0; n < num; n++) {
......@@ -532,7 +540,7 @@ int decode_splitmvs(VP8Context *s, VP56RangeCoder *c, VP8Macroblock *mb, int lay
if (!(k & 3))
left = AV_RN32A(&left_mv[mbsplits_left[k + 3]]);
else
left = AV_RN32A(&cur_mv[mbsplits_cur[k - 1]]);
left = AV_RN32A(&cur_mv[mbsplits_cur[k - 1]]);
if (k <= 3)
above = AV_RN32A(&top_mv[mbsplits_top[k + 12]]);
else
......@@ -560,11 +568,12 @@ int decode_splitmvs(VP8Context *s, VP56RangeCoder *c, VP8Macroblock *mb, int lay
}
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 */,
0 /* top-left */ };
0 /* top-left */ };
enum { CNT_ZERO, CNT_NEAREST, CNT_NEAR, CNT_SPLITMV };
enum { VP8_EDGE_TOP, VP8_EDGE_LEFT, VP8_EDGE_TOPLEFT };
int idx = CNT_ZERO;
......@@ -577,10 +586,9 @@ 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)
mb_edge[0] = mb + 2;
mb_edge[2] = mb + 1;
}
else {
mb_edge[0] = mb - s->mb_width-1;
mb_edge[2] = mb - s->mb_width-2;
} else {
mb_edge[0] = mb - s->mb_width - 1;
mb_edge[2] = mb - s->mb_width - 2;
}
AV_ZERO32(&near_mv[0]);
......@@ -588,24 +596,25 @@ void decode_mvs(VP8Context *s, VP8Macroblock *mb, int mb_x, int mb_y, int layout
AV_ZERO32(&near_mv[2]);
/* Process MB on top, left and top-left */
#define MV_EDGE_CHECK(n)\
{\
VP8Macroblock *edge = mb_edge[n];\
int edge_ref = edge->ref_frame;\
if (edge_ref != VP56_FRAME_CURRENT) {\
uint32_t mv = AV_RN32A(&edge->mv);\
if (mv) {\
if (cur_sign_bias != sign_bias[edge_ref]) {\
/* SWAR negate of the values in mv. */\
mv = ~mv;\
mv = ((mv&0x7fff7fff) + 0x00010001) ^ (mv&0x80008000);\
}\
if (!n || mv != AV_RN32A(&near_mv[idx]))\
AV_WN32A(&near_mv[++idx], mv);\
cnt[idx] += 1 + (n != 2);\
} else\
cnt[CNT_ZERO] += 1 + (n != 2);\
}\
#define MV_EDGE_CHECK(n) \
{ \
VP8Macroblock *edge = mb_edge[n]; \
int edge_ref = edge->ref_frame; \
if (edge_ref != VP56_FRAME_CURRENT) { \
uint32_t mv = AV_RN32A(&edge->mv); \
if (mv) { \
if (cur_sign_bias != sign_bias[edge_ref]) { \
/* SWAR negate of the values in mv. */ \
mv = ~mv; \
mv = ((mv & 0x7fff7fff) + \
0x00010001) ^ (mv & 0x80008000); \
} \
if (!n || mv != AV_RN32A(&near_mv[idx])) \
AV_WN32A(&near_mv[++idx], mv); \
cnt[idx] += 1 + (n != 2); \
} else \
cnt[CNT_ZERO] += 1 + (n != 2); \
} \
}
MV_EDGE_CHECK(0)
......@@ -617,7 +626,8 @@ void decode_mvs(VP8Context *s, VP8Macroblock *mb, int mb_x, int mb_y, int layout
mb->mode = VP8_MVMODE_MV;
/* 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;
/* 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_NEAR]][2])) {
/* 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])]);
cnt[CNT_SPLITMV] = ((mb_edge[VP8_EDGE_LEFT]->mode == VP8_MVMODE_SPLIT) +
......@@ -637,10 +646,10 @@ 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_SPLITMV]][3])) {
mb->mode = VP8_MVMODE_SPLIT;
mb->mv = mb->bmv[decode_splitmvs(s, c, mb, layout) - 1];
mb->mv = mb->bmv[decode_splitmvs(s, c, mb, layout) - 1];
} else {
mb->mv.y += read_mv_component(c, s->prob->mvc[0]);
mb->mv.x += read_mv_component(c, s->prob->mvc[1]);
mb->mv.y += read_mv_component(c, s->prob->mvc[0]);
mb->mv.x += read_mv_component(c, s->prob->mvc[1]);
mb->bmv[0] = mb->mv;
}
} else {
......@@ -670,8 +679,8 @@ void decode_intra4x4_modes(VP8Context *s, VP56RangeCoder *c, VP8Macroblock *mb,
}
if (keyframe) {
int x, y;
uint8_t* top;
uint8_t* const left = s->intra4x4_pred_mode_left;
uint8_t *top;
uint8_t *const left = s->intra4x4_pred_mode_left;
if (layout == 1)
top = mb->intra4x4_pred_mode_top;
else
......@@ -679,16 +688,17 @@ void decode_intra4x4_modes(VP8Context *s, VP56RangeCoder *c, VP8Macroblock *mb,
for (y = 0; y < 4; y++) {
for (x = 0; x < 4; x++) {
const uint8_t *ctx;
ctx = vp8_pred4x4_prob_intra[top[x]][left[y]];
ctx = vp8_pred4x4_prob_intra[top[x]][left[y]];
*intra4x4 = vp8_rac_get_tree(c, vp8_pred4x4_tree, ctx);
left[y] = top[x] = *intra4x4;
left[y] = top[x] = *intra4x4;
intra4x4++;
}
}
} else {
int 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;
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) {
decode_intra4x4_modes(s, c, mb, mb_x, 1, layout);
......@@ -717,19 +728,21 @@ void decode_mb_mode(VP8Context *s, VP8Macroblock *mb, int mb_x, int mb_y,
AV_WN32A(mb->intra4x4_pred_mode_top, modes);
else
AV_WN32A(s->intra4x4_pred_mode_top + 4 * mb_x, modes);
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->ref_frame = VP56_FRAME_CURRENT;
mb->chroma_pred_mode = vp8_rac_get_tree(c, vp8_pred8x8c_tree,
vp8_pred8x8c_prob_intra);
mb->ref_frame = VP56_FRAME_CURRENT;
} else if (vp56_rac_get_prob_branchy(c, s->prob->intra)) {
// inter MB, 16.2
if (vp56_rac_get_prob_branchy(c, s->prob->last))
mb->ref_frame = vp56_rac_get_prob(c, s->prob->golden) ?
VP56_FRAME_GOLDEN2 /* altref */ : VP56_FRAME_GOLDEN;
mb->ref_frame =
vp56_rac_get_prob(c, s->prob->golden) ? VP56_FRAME_GOLDEN2 /* altref */
: VP56_FRAME_GOLDEN;
else
mb->ref_frame = VP56_FRAME_PREVIOUS;
s->ref_count[mb->ref_frame-1]++;
s->ref_count[mb->ref_frame - 1]++;
// motion vectors, 16.3
decode_mvs(s, mb, mb_x, mb_y, layout);
......@@ -740,26 +753,29 @@ void decode_mb_mode(VP8Context *s, VP8Macroblock *mb, int mb_x, int mb_y,
if (mb->mode == MODE_I4x4)
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->ref_frame = VP56_FRAME_CURRENT;
mb->partitioning = VP8_SPLITMVMODE_NONE;
mb->chroma_pred_mode = vp8_rac_get_tree(c, vp8_pred8x8c_tree,
s->prob->pred8x8c);
mb->ref_frame = VP56_FRAME_CURRENT;
mb->partitioning = VP8_SPLITMVMODE_NONE;
AV_ZERO32(&mb->bmv[0]);
}
}
#ifndef decode_block_coeffs_internal
/**
* @param r arithmetic bitstream reader context
* @param r arithmetic bitstream reader context
* @param block destination for block coefficients
* @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 qmul array holding the dc/ac dequant factor at position 0/1
* @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
*
* @return 0 if no coeffs were decoded
* otherwise, the index of the last coeff decoded plus one
*/
static int decode_block_coeffs_internal(VP56RangeCoder *r, int16_t block[16],
uint8_t probs[16][3][NUM_DCT_TOKENS-1],
int i, uint8_t *token_prob, int16_t qmul[2])
uint8_t probs[16][3][NUM_DCT_TOKENS - 1],
int i, uint8_t *token_prob,
int16_t qmul[2])
{
VP56RangeCoder c = *r;
goto skip_eob;
......@@ -778,7 +794,7 @@ skip_eob:
if (!vp56_rac_get_prob_branchy(&c, token_prob[2])) { // DCT_1
coeff = 1;
token_prob = probs[i+1][1];
token_prob = probs[i + 1][1];
} else {
if (!vp56_rac_get_prob_branchy(&c, token_prob[3])) { // DCT 2,3,4
coeff = vp56_rac_get_prob_branchy(&c, token_prob[4]);
......@@ -789,21 +805,21 @@ skip_eob:
// DCT_CAT*
if (!vp56_rac_get_prob_branchy(&c, token_prob[6])) {
if (!vp56_rac_get_prob_branchy(&c, token_prob[7])) { // DCT_CAT1
coeff = 5 + vp56_rac_get_prob(&c, vp8_dct_cat1_prob[0]);
coeff = 5 + vp56_rac_get_prob(&c, vp8_dct_cat1_prob[0]);
} else { // DCT_CAT2
coeff = 7;
coeff += vp56_rac_get_prob(&c, vp8_dct_cat2_prob[0]) << 1;
coeff += vp56_rac_get_prob(&c, vp8_dct_cat2_prob[1]);
}
} else { // DCT_CAT3 and up
int a = vp56_rac_get_prob(&c, token_prob[8]);
int b = vp56_rac_get_prob(&c, token_prob[9+a]);
int cat = (a<<1) + b;
coeff = 3 + (8<<cat);
int a = vp56_rac_get_prob(&c, token_prob[8]);
int b = vp56_rac_get_prob(&c, token_prob[9 + a]);
int cat = (a << 1) + b;
coeff = 3 + (8 << cat);
coeff += vp8_rac_get_coeff(&c, ff_vp8_dct_cat_prob[cat]);
}
}
token_prob = probs[i+1][2];
token_prob = probs[i + 1][2];
}
block[zigzag_scan[i]] = (vp8_rac_get(&c) ? -coeff : coeff) * qmul[!!i];
} while (++i < 16);
......@@ -814,19 +830,20 @@ skip_eob:
#endif
/**
* @param c arithmetic bitstream reader context
* @param block destination for block coefficients
* @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 c arithmetic bitstream reader context
* @param block destination for block coefficients
* @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 zero_nhood the initial prediction context for number of surrounding
* 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
* otherwise, the index of the last coeff decoded plus one
*/
static av_always_inline
int decode_block_coeffs(VP56RangeCoder *c, int16_t block[16],
uint8_t probs[16][3][NUM_DCT_TOKENS-1],
uint8_t probs[16][3][NUM_DCT_TOKENS - 1],
int i, int zero_nhood, int16_t qmul[2])
{
uint8_t *token_prob = probs[i][zero_nhood];
......@@ -836,8 +853,8 @@ int decode_block_coeffs(VP56RangeCoder *c, int16_t block[16],
}
static av_always_inline
void decode_mb_coeffs(VP8Context *s, VP8ThreadData *td, VP56RangeCoder *c, VP8Macroblock *mb,
uint8_t t_nnz[9], uint8_t l_nnz[9])
void decode_mb_coeffs(VP8Context *s, VP8ThreadData *td, VP56RangeCoder *c,
VP8Macroblock *mb, uint8_t t_nnz[9], uint8_t l_nnz[9])
{
int i, x, y, luma_start = 0, luma_ctx = 3;
int nnz_pred, nnz, nnz_total = 0;
......@@ -848,28 +865,31 @@ void decode_mb_coeffs(VP8Context *s, VP8ThreadData *td, VP56RangeCoder *c, VP8Ma
nnz_pred = t_nnz[8] + l_nnz[8];
// decode DC values and do hadamard
nnz = decode_block_coeffs(c, td->block_dc, s->prob->token[1], 0, nnz_pred,
s->qmat[segment].luma_dc_qmul);
nnz = decode_block_coeffs(c, td->block_dc, s->prob->token[1], 0,
nnz_pred, s->qmat[segment].luma_dc_qmul);
l_nnz[8] = t_nnz[8] = !!nnz;
if (nnz) {
nnz_total += nnz;
block_dc = 1;
block_dc = 1;
if (nnz == 1)
s->vp8dsp.vp8_luma_dc_wht_dc(td->block, td->block_dc);
else
s->vp8dsp.vp8_luma_dc_wht(td->block, td->block_dc);
}
luma_start = 1;
luma_ctx = 0;
luma_ctx = 0;
}
// luma blocks
for (y = 0; y < 4; y++)
for (x = 0; x < 4; 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_pred, s->qmat[segment].luma_qmul);
// nnz+block_dc may be one more than the actual last index, but we don't care
nnz = decode_block_coeffs(c, td->block[y][x],
s->prob->token[luma_ctx],
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;
t_nnz[x] = l_nnz[y] = !!nnz;
nnz_total += nnz;
......@@ -881,12 +901,14 @@ void decode_mb_coeffs(VP8Context *s, VP8ThreadData *td, VP56RangeCoder *c, VP8Ma
for (i = 4; i < 6; i++)
for (y = 0; y < 2; y++)
for (x = 0; x < 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_pred, s->qmat[segment].chroma_qmul);
td->non_zero_count_cache[i][(y<<1)+x] = nnz;
t_nnz[i+2*x] = l_nnz[i+2*y] = !!nnz;
nnz_total += nnz;
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_pred,
s->qmat[segment].chroma_qmul);
td->non_zero_count_cache[i][(y << 1) + x] = nnz;
t_nnz[i + 2 * x] = l_nnz[i + 2 * y] = !!nnz;
nnz_total += nnz;
}
// if there were no coded coeffs despite the macroblock not being marked skip,
......@@ -897,65 +919,67 @@ void decode_mb_coeffs(VP8Context *s, VP8ThreadData *td, VP56RangeCoder *c, VP8Ma
}
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)
{
AV_COPY128(top_border, src_y + 15*linesize);
AV_COPY128(top_border, src_y + 15 * linesize);
if (!simple) {
AV_COPY64(top_border+16, src_cb + 7*uvlinesize);
AV_COPY64(top_border+24, src_cr + 7*uvlinesize);
AV_COPY64(top_border + 16, src_cb + 7 * uvlinesize);
AV_COPY64(top_border + 24, src_cr + 7 * uvlinesize);
}
}
static av_always_inline
void xchg_mb_border(uint8_t *top_border, uint8_t *src_y, uint8_t *src_cb, uint8_t *src_cr,
int linesize, int uvlinesize, int mb_x, int mb_y, int mb_width,
int simple, int xchg)
void xchg_mb_border(uint8_t *top_border, uint8_t *src_y, uint8_t *src_cb,
uint8_t *src_cr, int linesize, int uvlinesize, int mb_x,
int mb_y, int mb_width, int simple, int xchg)
{
uint8_t *top_border_m1 = top_border-32; // for TL prediction
src_y -= linesize;
uint8_t *top_border_m1 = top_border - 32; // for TL prediction
src_y -= linesize;
src_cb -= uvlinesize;
src_cr -= uvlinesize;
#define XCHG(a,b,xchg) do { \
if (xchg) AV_SWAP64(b,a); \
else AV_COPY64(b,a); \
#define XCHG(a, b, xchg) \
do { \
if (xchg) \
AV_SWAP64(b, a); \
else \
AV_COPY64(b, a); \
} while (0)
XCHG(top_border_m1+8, src_y-8, xchg);
XCHG(top_border, src_y, xchg);
XCHG(top_border+8, src_y+8, 1);
if (mb_x < mb_width-1)
XCHG(top_border+32, src_y+16, 1);
XCHG(top_border_m1 + 8, src_y - 8, xchg);
XCHG(top_border, src_y, xchg);
XCHG(top_border + 8, src_y + 8, 1);
if (mb_x < mb_width - 1)
XCHG(top_border + 32, src_y + 16, 1);
// only copy chroma for normal loop filter
// or to initialize the top row to 127
if (!simple || !mb_y) {
XCHG(top_border_m1+16, src_cb-8, xchg);
XCHG(top_border_m1+24, src_cr-8, xchg);
XCHG(top_border+16, src_cb, 1);
XCHG(top_border+24, src_cr, 1);
XCHG(top_border_m1 + 16, src_cb - 8, xchg);
XCHG(top_border_m1 + 24, src_cr - 8, xchg);
XCHG(top_border + 16, src_cb, 1);
XCHG(top_border + 24, src_cr, 1);
}
}
static av_always_inline
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;
} else {
else
return mb_y ? mode : LEFT_DC_PRED8x8;
}
}
static av_always_inline
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;
} else {
else
return mb_y ? mode : HOR_PRED8x8;
}
}
static av_always_inline
......@@ -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;
case HOR_PRED8x8:
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 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;
case TM_VP8_PRED:
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 VERT_RIGHT_PRED:
case HOR_DOWN_PRED:
......@@ -1025,10 +1050,10 @@ void intra_predict(VP8Context *s, VP8ThreadData *td, uint8_t *dst[3],
int x, y, mode, nnz;
uint32_t tr;
// for the first row, we need to run xchg_mb_border to init the top edge to 127
// otherwise, skip it if we aren't going to deblock
/* for the first row, we need to run xchg_mb_border to init the top edge
* to 127 otherwise, skip it if we aren't going to deblock */
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->filter.simple, 1);
......@@ -1046,10 +1071,9 @@ 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
// from the top macroblock
if (mb_y &&
mb_x == s->mb_width-1) {
tr = tr_right[-1]*0x01010101u;
tr_right = (uint8_t *)&tr;
if (mb_y && mb_x == s->mb_width - 1) {
tr = tr_right[-1] * 0x01010101u;
tr_right = (uint8_t *) &tr;
}
if (mb->skip)
......@@ -1059,27 +1083,29 @@ void intra_predict(VP8Context *s, VP8ThreadData *td, uint8_t *dst[3],
uint8_t *topright = ptr + 4 - s->linesize;
for (x = 0; x < 4; x++) {
int copy = 0, linesize = s->linesize;
uint8_t *dst = ptr+4*x;
DECLARE_ALIGNED(4, uint8_t, copy_dst)[5*8];
uint8_t *dst = ptr + 4 * x;
DECLARE_ALIGNED(4, uint8_t, copy_dst)[5 * 8];
if ((y == 0 || x == 3) && mb_y == 0) {
topright = tr_top;
} else if (x == 3)
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) {
dst = copy_dst + 12;
dst = copy_dst + 12;
linesize = 8;
if (!(mb_y + y)) {
copy_dst[3] = 127U;
AV_WN32A(copy_dst+4, 127U * 0x01010101U);
AV_WN32A(copy_dst + 4, 127U * 0x01010101U);
} else {
AV_COPY32(copy_dst+4, ptr+4*x-s->linesize);
AV_COPY32(copy_dst + 4, ptr + 4 * x - s->linesize);
if (!(mb_x + x)) {
copy_dst[3] = 129U;
} else {
copy_dst[3] = ptr[4*x-s->linesize-1];
copy_dst[3] = ptr[4 * x - s->linesize - 1];
}
}
if (!(mb_x + x)) {
......@@ -1088,31 +1114,33 @@ void intra_predict(VP8Context *s, VP8ThreadData *td, uint8_t *dst[3],
copy_dst[27] =
copy_dst[35] = 129U;
} else {
copy_dst[11] = ptr[4*x -1];
copy_dst[19] = ptr[4*x+s->linesize -1];
copy_dst[27] = ptr[4*x+s->linesize*2-1];
copy_dst[35] = ptr[4*x+s->linesize*3-1];
copy_dst[11] = ptr[4 * x - 1];
copy_dst[19] = ptr[4 * x + s->linesize - 1];
copy_dst[27] = ptr[4 * x + s->linesize * 2 - 1];
copy_dst[35] = ptr[4 * x + s->linesize * 3 - 1];
}
}
s->hpc.pred4x4[mode](dst, topright, linesize);
if (copy) {
AV_COPY32(ptr+4*x , copy_dst+12);
AV_COPY32(ptr+4*x+s->linesize , copy_dst+20);
AV_COPY32(ptr+4*x+s->linesize*2, copy_dst+28);
AV_COPY32(ptr+4*x+s->linesize*3, copy_dst+36);
AV_COPY32(ptr + 4 * x, copy_dst + 12);
AV_COPY32(ptr + 4 * x + s->linesize, copy_dst + 20);
AV_COPY32(ptr + 4 * x + s->linesize * 2, copy_dst + 28);
AV_COPY32(ptr + 4 * x + s->linesize * 3, copy_dst + 36);
}
nnz = td->non_zero_count_cache[y][x];
if (nnz) {
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
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;
}
ptr += 4*s->linesize;
ptr += 4 * s->linesize;
intra4x4 += 4;
}
}
......@@ -1122,7 +1150,7 @@ void intra_predict(VP8Context *s, VP8ThreadData *td, uint8_t *dst[3],
s->hpc.pred8x8[mode](dst[2], s->uvlinesize);
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->filter.simple, 0);
}
......@@ -1137,18 +1165,18 @@ static const uint8_t subpel_idx[3][8] = {
/**
* luma MC function
*
* @param s VP8 decoding context
* @param dst target buffer for block data at block position
* @param ref reference picture buffer at origin (0, 0)
* @param mv motion vector (relative to block position) to get pixel data from
* @param x_off horizontal position of block from origin (0, 0)
* @param y_off vertical position of block from origin (0, 0)
* @param block_w width of block (16, 8 or 4)
* @param block_h height of block (always same as block_w)
* @param width width of src/dst plane data
* @param height height of src/dst plane data
* @param s VP8 decoding context
* @param dst target buffer for block data at block position
* @param ref reference picture buffer at origin (0, 0)
* @param mv motion vector (relative to block position) to get pixel data from
* @param x_off horizontal position of block from origin (0, 0)
* @param y_off vertical position of block from origin (0, 0)
* @param block_w width of block (16, 8 or 4)
* @param block_h height of block (always same as block_w)
* @param width width of src/dst plane data
* @param height height of src/dst plane data
* @param linesize size of a single line of plane data, including padding
* @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
void vp8_mc_luma(VP8Context *s, VP8ThreadData *td, uint8_t *dst,
......@@ -1162,8 +1190,8 @@ void vp8_mc_luma(VP8Context *s, VP8ThreadData *td, uint8_t *dst,
if (AV_RN32A(mv)) {
int src_linesize = linesize;
int mx = (mv->x << 1)&7, mx_idx = subpel_idx[0][mx];
int my = (mv->y << 1)&7, my_idx = subpel_idx[0][my];
int mx = (mv->x << 1) & 7, mx_idx = subpel_idx[0][mx];
int my = (mv->y << 1) & 7, my_idx = subpel_idx[0][my];
x_off += mv->x >> 2;
y_off += mv->y >> 2;
......@@ -1176,46 +1204,50 @@ void vp8_mc_luma(VP8Context *s, VP8ThreadData *td, uint8_t *dst,
s->vdsp.emulated_edge_mc(td->edge_emu_buffer,
src - my_idx * linesize - mx_idx,
EDGE_EMU_LINESIZE, linesize,
block_w + subpel_idx[1][mx], block_h + subpel_idx[1][my],
x_off - mx_idx, y_off - my_idx, width, height);
block_w + subpel_idx[1][mx],
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_linesize = EDGE_EMU_LINESIZE;
}
mc_func[my_idx][mx_idx](dst, linesize, src, src_linesize, block_h, mx, my);
} else {
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);
}
}
/**
* chroma MC function
*
* @param s VP8 decoding context
* @param dst1 target buffer for block data at block position (U plane)
* @param dst2 target buffer for block data at block position (V plane)
* @param ref reference picture buffer at origin (0, 0)
* @param mv motion vector (relative to block position) to get pixel data from
* @param x_off horizontal position of block from origin (0, 0)
* @param y_off vertical position of block from origin (0, 0)
* @param block_w width of block (16, 8 or 4)
* @param block_h height of block (always same as block_w)
* @param width width of src/dst plane data
* @param height height of src/dst plane data
* @param s VP8 decoding context
* @param dst1 target buffer for block data at block position (U plane)
* @param dst2 target buffer for block data at block position (V plane)
* @param ref reference picture buffer at origin (0, 0)
* @param mv motion vector (relative to block position) to get pixel data from
* @param x_off horizontal position of block from origin (0, 0)
* @param y_off vertical position of block from origin (0, 0)
* @param block_w width of block (16, 8 or 4)
* @param block_h height of block (always same as block_w)
* @param width width of src/dst plane data
* @param height height of src/dst plane data
* @param linesize size of a single line of plane data, including padding
* @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
void vp8_mc_chroma(VP8Context *s, VP8ThreadData *td, uint8_t *dst1, uint8_t *dst2,
ThreadFrame *ref, const VP56mv *mv, int x_off, int y_off,
int block_w, int block_h, int width, int height, ptrdiff_t linesize,
void vp8_mc_chroma(VP8Context *s, VP8ThreadData *td, uint8_t *dst1,
uint8_t *dst2, ThreadFrame *ref, const VP56mv *mv,
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])
{
uint8_t *src1 = ref->f->data[1], *src2 = ref->f->data[2];
if (AV_RN32A(mv)) {
int mx = mv->x&7, mx_idx = subpel_idx[0][mx];
int my = mv->y&7, my_idx = subpel_idx[0][my];
int mx = mv->x & 7, mx_idx = subpel_idx[0][mx];
int my = mv->y & 7, my_idx = subpel_idx[0][my];
x_off += mv->x >> 3;
y_off += mv->y >> 3;
......@@ -1239,7 +1271,7 @@ void vp8_mc_chroma(VP8Context *s, VP8ThreadData *td, uint8_t *dst1, uint8_t *dst
EDGE_EMU_LINESIZE, linesize,
block_w + subpel_idx[1][mx], block_h + subpel_idx[1][my],
x_off - mx_idx, y_off - my_idx, width, height);
src2 = td->edge_emu_buffer + mx_idx + EDGE_EMU_LINESIZE* my_idx;
src2 = td->edge_emu_buffer + mx_idx + EDGE_EMU_LINESIZE * my_idx;
mc_func[my_idx][mx_idx](dst2, linesize, src2, EDGE_EMU_LINESIZE, block_h, mx, my);
} else {
mc_func[my_idx][mx_idx](dst1, linesize, src1, linesize, block_h, mx, my);
......@@ -1255,8 +1287,7 @@ void vp8_mc_chroma(VP8Context *s, VP8ThreadData *td, uint8_t *dst1, uint8_t *dst
static av_always_inline
void vp8_mc_part(VP8Context *s, VP8ThreadData *td, uint8_t *dst[3],
ThreadFrame *ref_frame, int x_off, int y_off,
int bx_off, int by_off,
int block_w, int block_h,
int bx_off, int by_off, int block_w, int block_h,
int width, int height, VP56mv *mv)
{
VP56mv uvmv = *mv;
......@@ -1272,10 +1303,14 @@ void vp8_mc_part(VP8Context *s, VP8ThreadData *td, uint8_t *dst[3],
uvmv.x &= ~7;
uvmv.y &= ~7;
}
x_off >>= 1; y_off >>= 1;
bx_off >>= 1; by_off >>= 1;
width >>= 1; height >>= 1;
block_w >>= 1; block_h >>= 1;
x_off >>= 1;
y_off >>= 1;
bx_off >>= 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,
dst[2] + by_off * s->uvlinesize + bx_off, ref_frame,
&uvmv, x_off + bx_off, y_off + by_off,
......@@ -1284,22 +1319,24 @@ void vp8_mc_part(VP8Context *s, VP8ThreadData *td, uint8_t *dst[3],
}
/* Fetch pixels for estimated mv 4 macroblocks ahead.
* 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)
* 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)
{
/* 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)) {
int x_off = mb_x << 4, y_off = mb_y << 4;
int mx = (mb->mv.x>>2) + x_off + 8;
int my = (mb->mv.y>>2) + y_off;
uint8_t **src= s->framep[ref]->tf.f->data;
int off= mx + (my + (mb_x&3)*4)*s->linesize + 64;
int mx = (mb->mv.x >> 2) + x_off + 8;
int my = (mb->mv.y >> 2) + y_off;
uint8_t **src = s->framep[ref]->tf.f->data;
int off = mx + (my + (mb_x & 3) * 4) * s->linesize + 64;
/* For threading, a ff_thread_await_progress here might be useful, but
* it actually slows down the decoder. Since a bad prefetch doesn't
* generate bad decoder output, we don't run it here. */
s->vdsp.prefetch(src[0]+off, s->linesize, 4);
off= (mx>>1) + ((my>>1) + (mb_x&7))*s->uvlinesize + 64;
s->vdsp.prefetch(src[1]+off, src[2]-src[1], 2);
s->vdsp.prefetch(src[0] + off, s->linesize, 4);
off = (mx >> 1) + ((my >> 1) + (mb_x & 7)) * s->uvlinesize + 64;
s->vdsp.prefetch(src[1] + off, src[2] - src[1], 2);
}
}
......@@ -1311,7 +1348,7 @@ void inter_predict(VP8Context *s, VP8ThreadData *td, uint8_t *dst[3],
VP8Macroblock *mb, int mb_x, int mb_y)
{
int x_off = mb_x << 4, y_off = mb_y << 4;
int width = 16*s->mb_width, height = 16*s->mb_height;
int width = 16 * s->mb_width, height = 16 * s->mb_height;
ThreadFrame *ref = &s->framep[mb->ref_frame]->tf;
VP56mv *bmv = mb->bmv;
......@@ -1327,35 +1364,38 @@ void inter_predict(VP8Context *s, VP8ThreadData *td, uint8_t *dst[3],
/* Y */
for (y = 0; y < 4; y++) {
for (x = 0; x < 4; x++) {
vp8_mc_luma(s, td, dst[0] + 4*y*s->linesize + x*4,
ref, &bmv[4*y + x],
4*x + x_off, 4*y + y_off, 4, 4,
vp8_mc_luma(s, td, dst[0] + 4 * y * s->linesize + x * 4,
ref, &bmv[4 * y + x],
4 * x + x_off, 4 * y + y_off, 4, 4,
width, height, s->linesize,
s->put_pixels_tab[2]);
}
}
/* 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 (x = 0; x < 2; x++) {
uvmv.x = mb->bmv[ 2*y * 4 + 2*x ].x +
mb->bmv[ 2*y * 4 + 2*x+1].x +
mb->bmv[(2*y+1) * 4 + 2*x ].x +
mb->bmv[(2*y+1) * 4 + 2*x+1].x;
uvmv.y = mb->bmv[ 2*y * 4 + 2*x ].y +
mb->bmv[ 2*y * 4 + 2*x+1].y +
mb->bmv[(2*y+1) * 4 + 2*x ].y +
mb->bmv[(2*y+1) * 4 + 2*x+1].y;
uvmv.x = (uvmv.x + 2 + (uvmv.x >> (INT_BIT-1))) >> 2;
uvmv.y = (uvmv.y + 2 + (uvmv.y >> (INT_BIT-1))) >> 2;
uvmv.x = mb->bmv[2 * y * 4 + 2 * x ].x +
mb->bmv[2 * y * 4 + 2 * x + 1].x +
mb->bmv[(2 * y + 1) * 4 + 2 * x ].x +
mb->bmv[(2 * y + 1) * 4 + 2 * x + 1].x;
uvmv.y = mb->bmv[2 * y * 4 + 2 * x ].y +
mb->bmv[2 * y * 4 + 2 * x + 1].y +
mb->bmv[(2 * y + 1) * 4 + 2 * x ].y +
mb->bmv[(2 * y + 1) * 4 + 2 * x + 1].y;
uvmv.x = (uvmv.x + 2 + (uvmv.x >> (INT_BIT - 1))) >> 2;
uvmv.y = (uvmv.y + 2 + (uvmv.y >> (INT_BIT - 1))) >> 2;
if (s->profile == 3) {
uvmv.x &= ~7;
uvmv.y &= ~7;
}
vp8_mc_chroma(s, td, dst[1] + 4*y*s->uvlinesize + x*4,
dst[2] + 4*y*s->uvlinesize + x*4, ref, &uvmv,
4*x + x_off, 4*y + y_off, 4, 4,
vp8_mc_chroma(s, td, dst[1] + 4 * y * s->uvlinesize + x * 4,
dst[2] + 4 * y * s->uvlinesize + x * 4, ref,
&uvmv, 4 * x + x_off, 4 * y + y_off, 4, 4,
width, height, s->uvlinesize,
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,
uint8_t *dst[3], VP8Macroblock *mb)
static av_always_inline
void idct_mb(VP8Context *s, VP8ThreadData *td, uint8_t *dst[3], VP8Macroblock *mb)
{
int x, y, ch;
......@@ -1397,12 +1437,16 @@ static av_always_inline void idct_mb(VP8Context *s, VP8ThreadData *td,
for (y = 0; y < 4; y++) {
uint32_t nnz4 = AV_RL32(td->non_zero_count_cache[y]);
if (nnz4) {
if (nnz4&~0x01010101) {
if (nnz4 & ~0x01010101) {
for (x = 0; x < 4; x++) {
if ((uint8_t)nnz4 == 1)
s->vp8dsp.vp8_idct_dc_add(y_dst+4*x, td->block[y][x], s->linesize);
else if((uint8_t)nnz4 > 1)
s->vp8dsp.vp8_idct_add(y_dst+4*x, td->block[y][x], s->linesize);
if ((uint8_t) nnz4 == 1)
s->vp8dsp.vp8_idct_dc_add(y_dst + 4 * x,
td->block[y][x],
s->linesize);
else if ((uint8_t) nnz4 > 1)
s->vp8dsp.vp8_idct_add(y_dst + 4 * x,
td->block[y][x],
s->linesize);
nnz4 >>= 8;
if (!nnz4)
break;
......@@ -1411,36 +1455,42 @@ static av_always_inline void idct_mb(VP8Context *s, VP8ThreadData *td,
s->vp8dsp.vp8_idct_dc_add4y(y_dst, td->block[y], s->linesize);
}
}
y_dst += 4*s->linesize;
y_dst += 4 * s->linesize;
}
}
for (ch = 0; ch < 2; ch++) {
uint32_t nnz4 = AV_RL32(td->non_zero_count_cache[4+ch]);
uint32_t nnz4 = AV_RL32(td->non_zero_count_cache[4 + ch]);
if (nnz4) {
uint8_t *ch_dst = dst[1+ch];
if (nnz4&~0x01010101) {
uint8_t *ch_dst = dst[1 + ch];
if (nnz4 & ~0x01010101) {
for (y = 0; y < 2; y++) {
for (x = 0; x < 2; x++) {
if ((uint8_t)nnz4 == 1)
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)
s->vp8dsp.vp8_idct_add(ch_dst+4*x, td->block[4+ch][(y<<1)+x], s->uvlinesize);
if ((uint8_t) nnz4 == 1)
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)
s->vp8dsp.vp8_idct_add(ch_dst + 4 * x,
td->block[4 + ch][(y << 1) + x],
s->uvlinesize);
nnz4 >>= 8;
if (!nnz4)
goto chroma_idct_end;
}
ch_dst += 4*s->uvlinesize;
ch_dst += 4 * s->uvlinesize;
}
} else {
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;
......@@ -1467,10 +1517,13 @@ static av_always_inline void filter_level_for_mb(VP8Context *s, VP8Macroblock *m
f->filter_level = filter_level;
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 filter_level = f->filter_level;
......@@ -1492,82 +1545,84 @@ static av_always_inline void filter_mb(VP8Context *s, uint8_t *dst[3], VP8Filter
if (!filter_level)
return;
bedge_lim = 2*filter_level + inner_limit;
bedge_lim = 2 * filter_level + inner_limit;
mbedge_lim = bedge_lim + 4;
hev_thresh = hev_thresh_lut[s->keyframe][filter_level];
if (mb_x) {
s->vp8dsp.vp8_h_loop_filter16y(dst[0], linesize,
s->vp8dsp.vp8_h_loop_filter16y(dst[0], linesize,
mbedge_lim, inner_limit, hev_thresh);
s->vp8dsp.vp8_h_loop_filter8uv(dst[1], dst[2], uvlinesize,
s->vp8dsp.vp8_h_loop_filter8uv(dst[1], dst[2], uvlinesize,
mbedge_lim, inner_limit, hev_thresh);
}
if (inner_filter) {
s->vp8dsp.vp8_h_loop_filter16y_inner(dst[0]+ 4, linesize, bedge_lim,
s->vp8dsp.vp8_h_loop_filter16y_inner(dst[0] + 4, linesize, bedge_lim,
inner_limit, hev_thresh);
s->vp8dsp.vp8_h_loop_filter16y_inner(dst[0]+ 8, linesize, bedge_lim,
s->vp8dsp.vp8_h_loop_filter16y_inner(dst[0] + 8, linesize, bedge_lim,
inner_limit, hev_thresh);
s->vp8dsp.vp8_h_loop_filter16y_inner(dst[0]+12, linesize, bedge_lim,
s->vp8dsp.vp8_h_loop_filter16y_inner(dst[0] + 12, linesize, bedge_lim,
inner_limit, hev_thresh);
s->vp8dsp.vp8_h_loop_filter8uv_inner(dst[1] + 4, dst[2] + 4,
uvlinesize, bedge_lim,
s->vp8dsp.vp8_h_loop_filter8uv_inner(dst[1] + 4, dst[2] + 4,
uvlinesize, bedge_lim,
inner_limit, hev_thresh);
}
if (mb_y) {
s->vp8dsp.vp8_v_loop_filter16y(dst[0], linesize,
s->vp8dsp.vp8_v_loop_filter16y(dst[0], linesize,
mbedge_lim, inner_limit, hev_thresh);
s->vp8dsp.vp8_v_loop_filter8uv(dst[1], dst[2], uvlinesize,
s->vp8dsp.vp8_v_loop_filter8uv(dst[1], dst[2], uvlinesize,
mbedge_lim, inner_limit, hev_thresh);
}
if (inner_filter) {
s->vp8dsp.vp8_v_loop_filter16y_inner(dst[0]+ 4*linesize,
linesize, bedge_lim,
s->vp8dsp.vp8_v_loop_filter16y_inner(dst[0] + 4 * linesize,
linesize, bedge_lim,
inner_limit, hev_thresh);
s->vp8dsp.vp8_v_loop_filter16y_inner(dst[0]+ 8*linesize,
linesize, bedge_lim,
s->vp8dsp.vp8_v_loop_filter16y_inner(dst[0] + 8 * linesize,
linesize, bedge_lim,
inner_limit, hev_thresh);
s->vp8dsp.vp8_v_loop_filter16y_inner(dst[0]+12*linesize,
linesize, bedge_lim,
s->vp8dsp.vp8_v_loop_filter16y_inner(dst[0] + 12 * linesize,
linesize, bedge_lim,
inner_limit, hev_thresh);
s->vp8dsp.vp8_v_loop_filter8uv_inner(dst[1] + 4 * uvlinesize,
dst[2] + 4 * uvlinesize,
uvlinesize, bedge_lim,
s->vp8dsp.vp8_v_loop_filter8uv_inner(dst[1] + 4 * uvlinesize,
dst[2] + 4 * uvlinesize,
uvlinesize, bedge_lim,
inner_limit, hev_thresh);
}
}
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 filter_level = f->filter_level;
int inner_limit = f->inner_limit;
int inner_limit = f->inner_limit;
int inner_filter = f->inner_filter;
int linesize = s->linesize;
int linesize = s->linesize;
if (!filter_level)
return;
bedge_lim = 2*filter_level + inner_limit;
bedge_lim = 2 * filter_level + inner_limit;
mbedge_lim = bedge_lim + 4;
if (mb_x)
s->vp8dsp.vp8_h_loop_filter_simple(dst, linesize, mbedge_lim);
if (inner_filter) {
s->vp8dsp.vp8_h_loop_filter_simple(dst+ 4, linesize, bedge_lim);
s->vp8dsp.vp8_h_loop_filter_simple(dst+ 8, linesize, bedge_lim);
s->vp8dsp.vp8_h_loop_filter_simple(dst+12, linesize, bedge_lim);
s->vp8dsp.vp8_h_loop_filter_simple(dst + 4, linesize, bedge_lim);
s->vp8dsp.vp8_h_loop_filter_simple(dst + 8, linesize, bedge_lim);
s->vp8dsp.vp8_h_loop_filter_simple(dst + 12, linesize, bedge_lim);
}
if (mb_y)
s->vp8dsp.vp8_v_loop_filter_simple(dst, linesize, mbedge_lim);
if (inner_filter) {
s->vp8dsp.vp8_v_loop_filter_simple(dst+ 4*linesize, linesize, bedge_lim);
s->vp8dsp.vp8_v_loop_filter_simple(dst+ 8*linesize, linesize, bedge_lim);
s->vp8dsp.vp8_v_loop_filter_simple(dst+12*linesize, linesize, bedge_lim);
s->vp8dsp.vp8_v_loop_filter_simple(dst + 4 * linesize, linesize, bedge_lim);
s->vp8dsp.vp8_v_loop_filter_simple(dst + 8 * linesize, linesize, bedge_lim);
s->vp8dsp.vp8_v_loop_filter_simple(dst + 12 * linesize, linesize, bedge_lim);
}
}
......@@ -1581,16 +1636,18 @@ static void vp8_decode_mv_mb_modes(AVCodecContext *avctx, VP8Frame *curframe,
s->mv_min.y = -MARGIN;
s->mv_max.y = ((s->mb_height - 1) << 6) + MARGIN;
for (mb_y = 0; mb_y < s->mb_height; mb_y++) {
VP8Macroblock *mb = s->macroblocks_base + ((s->mb_width+1)*(mb_y + 1) + 1);
int mb_xy = mb_y*s->mb_width;
VP8Macroblock *mb = s->macroblocks_base +
((s->mb_width + 1) * (mb_y + 1) + 1);
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);
s->mv_min.x = -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++) {
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,
prev_frame && prev_frame->seg_map ?
prev_frame->seg_map->data + mb_xy : NULL, 1);
......@@ -1603,37 +1660,40 @@ static void vp8_decode_mv_mb_modes(AVCodecContext *avctx, VP8Frame *curframe,
}
#if HAVE_THREADS
#define check_thread_pos(td, otd, mb_x_check, mb_y_check)\
do {\
int tmp = (mb_y_check << 16) | (mb_x_check & 0xFFFF);\
if (otd->thread_mb_pos < tmp) {\
pthread_mutex_lock(&otd->lock);\
td->wait_mb_pos = tmp;\
do {\
if (otd->thread_mb_pos >= tmp)\
break;\
pthread_cond_wait(&otd->cond, &otd->lock);\
} while (1);\
td->wait_mb_pos = INT_MAX;\
pthread_mutex_unlock(&otd->lock);\
}\
} while(0);
#define update_pos(td, mb_y, mb_x)\
do {\
int pos = (mb_y << 16) | (mb_x & 0xFFFF);\
int sliced_threading = (avctx->active_thread_type == FF_THREAD_SLICE) && (num_jobs > 1);\
int is_null = (next_td == NULL) || (prev_td == NULL);\
int pos_check = (is_null) ? 1 :\
(next_td != td && pos >= next_td->wait_mb_pos) ||\
(prev_td != td && pos >= prev_td->wait_mb_pos);\
td->thread_mb_pos = pos;\
if (sliced_threading && pos_check) {\
pthread_mutex_lock(&td->lock);\
pthread_cond_broadcast(&td->cond);\
pthread_mutex_unlock(&td->lock);\
}\
} while(0);
#define check_thread_pos(td, otd, mb_x_check, mb_y_check) \
do { \
int tmp = (mb_y_check << 16) | (mb_x_check & 0xFFFF); \
if (otd->thread_mb_pos < tmp) { \
pthread_mutex_lock(&otd->lock); \
td->wait_mb_pos = tmp; \
do { \
if (otd->thread_mb_pos >= tmp) \
break; \
pthread_cond_wait(&otd->cond, &otd->lock); \
} while (1); \
td->wait_mb_pos = INT_MAX; \
pthread_mutex_unlock(&otd->lock); \
} \
} while (0);
#define update_pos(td, mb_y, mb_x) \
do { \
int pos = (mb_y << 16) | (mb_x & 0xFFFF); \
int sliced_threading = (avctx->active_thread_type == FF_THREAD_SLICE) && \
(num_jobs > 1); \
int is_null = (next_td == NULL) || (prev_td == NULL); \
int pos_check = (is_null) ? 1 \
: (next_td != td && \
pos >= next_td->wait_mb_pos) || \
(prev_td != td && \
pos >= prev_td->wait_mb_pos); \
td->thread_mb_pos = pos; \
if (sliced_threading && pos_check) { \
pthread_mutex_lock(&td->lock); \
pthread_cond_broadcast(&td->cond); \
pthread_mutex_unlock(&td->lock); \
} \
} while (0);
#else
#define check_thread_pos(td, otd, mb_x_check, mb_y_check)
#define update_pos(td, mb_y, mb_x)
......@@ -1644,51 +1704,58 @@ static void vp8_decode_mb_row_no_filter(AVCodecContext *avctx, void *tdata,
{
VP8Context *s = avctx->priv_data;
VP8ThreadData *prev_td, *next_td, *td = &s->thread_data[threadnr];
int mb_y = td->thread_mb_pos>>16;
int mb_x, mb_xy = mb_y*s->mb_width;
int mb_y = td->thread_mb_pos >> 16;
int mb_x, mb_xy = mb_y * s->mb_width;
int num_jobs = s->num_jobs;
VP8Frame *curframe = s->curframe, *prev_frame = s->prev_frame;
VP56RangeCoder *c = &s->coeff_partition[mb_y & (s->num_coeff_partitions-1)];
VP56RangeCoder *c = &s->coeff_partition[mb_y & (s->num_coeff_partitions - 1)];
VP8Macroblock *mb;
uint8_t *dst[3] = {
curframe->tf.f->data[0] + 16*mb_y*s->linesize,
curframe->tf.f->data[1] + 8*mb_y*s->uvlinesize,
curframe->tf.f->data[2] + 8*mb_y*s->uvlinesize
curframe->tf.f->data[0] + 16 * mb_y * s->linesize,
curframe->tf.f->data[1] + 8 * mb_y * s->uvlinesize,
curframe->tf.f->data[2] + 8 * mb_y * s->uvlinesize
};
if (mb_y == 0) prev_td = td;
else 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 (mb_y == 0)
prev_td = td;
else
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)
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 {
// Make sure the previous frame has read its segmentation map,
// if we re-use the same map.
if (prev_frame && s->segmentation.enabled &&
!s->segmentation.update_map)
ff_thread_await_progress(&prev_frame->tf, mb_y, 0);
mb = s->macroblocks + (s->mb_height - mb_y - 1)*2;
mb = s->macroblocks + (s->mb_height - mb_y - 1) * 2;
memset(mb - 1, 0, sizeof(*mb)); // zero left macroblock
AV_WN32A(s->intra4x4_pred_mode_left, DC_PRED*0x01010101);
AV_WN32A(s->intra4x4_pred_mode_left, DC_PRED * 0x01010101);
}
memset(td->left_nnz, 0, sizeof(td->left_nnz));
s->mv_min.x = -MARGIN;
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++) {
// Wait for previous thread to read mb_x+2, and reach mb_y-1.
if (prev_td != td) {
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 {
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[1] + (mb_x&7)*s->uvlinesize + 64, dst[2] - dst[1], 2);
s->vdsp.prefetch(dst[0] + (mb_x & 3) * 4 * s->linesize + 64,
s->linesize, 4);
s->vdsp.prefetch(dst[1] + (mb_x & 7) * s->uvlinesize + 64,
dst[2] - dst[1], 2);
if (!s->mb_layout)
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_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) {
td->left_nnz[8] = 0;
s->top_nnz[mb_x][8] = 0;
......@@ -1723,23 +1791,25 @@ static void vp8_decode_mb_row_no_filter(AVCodecContext *avctx, void *tdata,
if (s->deblock_filter)
filter_level_for_mb(s, mb, &td->filter_strength[mb_x]);
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)
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
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);
dst[0] += 16;
dst[1] += 8;
dst[2] += 8;
dst[0] += 16;
dst[1] += 8;
dst[2] += 8;
s->mv_min.x -= 64;
s->mv_max.x -= 64;
if (mb_x == s->mb_width+1) {
update_pos(td, mb_y, s->mb_width+3);
if (mb_x == s->mb_width + 1) {
update_pos(td, mb_y, s->mb_width + 3);
} else {
update_pos(td, mb_y, mb_x);
}
......@@ -1751,41 +1821,46 @@ static void vp8_filter_mb_row(AVCodecContext *avctx, void *tdata,
{
VP8Context *s = avctx->priv_data;
VP8ThreadData *td = &s->thread_data[threadnr];
int mb_x, mb_y = td->thread_mb_pos>>16, num_jobs = s->num_jobs;
int mb_x, mb_y = td->thread_mb_pos >> 16, num_jobs = s->num_jobs;
AVFrame *curframe = s->curframe->tf.f;
VP8Macroblock *mb;
VP8ThreadData *prev_td, *next_td;
uint8_t *dst[3] = {
curframe->data[0] + 16*mb_y*s->linesize,
curframe->data[1] + 8*mb_y*s->uvlinesize,
curframe->data[2] + 8*mb_y*s->uvlinesize
curframe->data[0] + 16 * mb_y * s->linesize,
curframe->data[1] + 8 * mb_y * s->uvlinesize,
curframe->data[2] + 8 * mb_y * s->uvlinesize
};
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
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;
else 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 (mb_y == 0)
prev_td = td;
else
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++) {
VP8FilterStrength *f = &td->filter_strength[mb_x];
if (prev_td != td) {
check_thread_pos(td, prev_td, (mb_x+1) + (s->mb_width+3), mb_y-1);
}
if (prev_td != td)
check_thread_pos(td, prev_td,
(mb_x + 1) + (s->mb_width + 3), mb_y - 1);
if (next_td != td)
if (next_td != &s->thread_data[0]) {
check_thread_pos(td, next_td, mb_x+1, mb_y+1);
}
if (next_td != &s->thread_data[0])
check_thread_pos(td, next_td, mb_x + 1, mb_y + 1);
if (num_jobs == 1) {
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
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)
......@@ -1796,7 +1871,7 @@ static void vp8_filter_mb_row(AVCodecContext *avctx, void *tdata,
dst[1] += 8;
dst[2] += 8;
update_pos(td, mb_y, (s->mb_width+3) + mb_x);
update_pos(td, mb_y, (s->mb_width + 3) + mb_x);
}
}
......@@ -1808,10 +1883,12 @@ static int vp8_decode_mb_row_sliced(AVCodecContext *avctx, void *tdata,
VP8ThreadData *next_td = NULL, *prev_td = NULL;
VP8Frame *curframe = s->curframe;
int mb_y, num_jobs = s->num_jobs;
td->thread_nr = threadnr;
for (mb_y = jobnr; mb_y < s->mb_height; mb_y += num_jobs) {
if (mb_y >= s->mb_height) break;
td->thread_mb_pos = mb_y<<16;
if (mb_y >= s->mb_height)
break;
td->thread_mb_pos = mb_y << 16;
vp8_decode_mb_row_no_filter(avctx, tdata, jobnr, threadnr);
if (s->deblock_filter)
vp8_filter_mb_row(avctx, tdata, jobnr, threadnr);
......@@ -1840,11 +1917,12 @@ int ff_vp8_decode_frame(AVCodecContext *avctx, void *data, int *got_frame,
prev_frame = s->framep[VP56_FRAME_CURRENT];
referenced = s->update_last || s->update_golden == VP56_FRAME_CURRENT
|| s->update_altref == VP56_FRAME_CURRENT;
referenced = s->update_last || s->update_golden == VP56_FRAME_CURRENT ||
s->update_altref == VP56_FRAME_CURRENT;
skip_thresh = !referenced ? AVDISCARD_NONREF :
!s->keyframe ? AVDISCARD_NONKEY : AVDISCARD_ALL;
skip_thresh = !referenced ? AVDISCARD_NONREF
: !s->keyframe ? AVDISCARD_NONKEY
: AVDISCARD_ALL;
if (avctx->skip_frame >= skip_thresh) {
s->invisible = 1;
......@@ -1858,7 +1936,7 @@ int ff_vp8_decode_frame(AVCodecContext *avctx, void *data, int *got_frame,
if (s->frames[i].tf.f->data[0] &&
&s->frames[i] != prev_frame &&
&s->frames[i] != s->framep[VP56_FRAME_PREVIOUS] &&
&s->frames[i] != s->framep[VP56_FRAME_GOLDEN] &&
&s->frames[i] != s->framep[VP56_FRAME_GOLDEN] &&
&s->frames[i] != s->framep[VP56_FRAME_GOLDEN2])
vp8_release_frame(s, &s->frames[i]);
......@@ -1866,7 +1944,7 @@ int ff_vp8_decode_frame(AVCodecContext *avctx, void *data, int *got_frame,
for (i = 0; i < 5; i++)
if (&s->frames[i] != prev_frame &&
&s->frames[i] != s->framep[VP56_FRAME_PREVIOUS] &&
&s->frames[i] != s->framep[VP56_FRAME_GOLDEN] &&
&s->frames[i] != s->framep[VP56_FRAME_GOLDEN] &&
&s->frames[i] != s->framep[VP56_FRAME_GOLDEN2]) {
curframe = s->framep[VP56_FRAME_CURRENT] = &s->frames[i];
break;
......@@ -1878,57 +1956,61 @@ int ff_vp8_decode_frame(AVCodecContext *avctx, void *data, int *got_frame,
if (curframe->tf.f->data[0])
vp8_release_frame(s, curframe);
// Given that arithmetic probabilities are updated every frame, it's quite likely
// that the values we have on a random interframe are complete junk if we didn't
// start decode on a keyframe. So just don't display anything rather than junk.
/* Given that arithmetic probabilities are updated every frame, it's quite
* likely that the values we have on a random interframe are complete
* 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] ||
!s->framep[VP56_FRAME_GOLDEN] ||
!s->framep[VP56_FRAME_GOLDEN] ||
!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;
goto err;
}
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))) {
av_log(avctx, AV_LOG_ERROR, "get_buffer() failed!\n");
goto err;
}
// check if golden and altref are swapped
if (s->update_altref != VP56_FRAME_NONE) {
s->next_framep[VP56_FRAME_GOLDEN2] = s->framep[s->update_altref];
} else {
s->next_framep[VP56_FRAME_GOLDEN2] = s->framep[VP56_FRAME_GOLDEN2];
}
if (s->update_golden != VP56_FRAME_NONE) {
s->next_framep[VP56_FRAME_GOLDEN] = s->framep[s->update_golden];
} else {
s->next_framep[VP56_FRAME_GOLDEN] = s->framep[VP56_FRAME_GOLDEN];
}
if (s->update_last) {
if (s->update_altref != VP56_FRAME_NONE)
s->next_framep[VP56_FRAME_GOLDEN2] = s->framep[s->update_altref];
else
s->next_framep[VP56_FRAME_GOLDEN2] = s->framep[VP56_FRAME_GOLDEN2];
if (s->update_golden != VP56_FRAME_NONE)
s->next_framep[VP56_FRAME_GOLDEN] = s->framep[s->update_golden];
else
s->next_framep[VP56_FRAME_GOLDEN] = s->framep[VP56_FRAME_GOLDEN];
if (s->update_last)
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_CURRENT] = curframe;
s->next_framep[VP56_FRAME_CURRENT] = curframe;
ff_thread_finish_setup(avctx);
s->linesize = curframe->tf.f->linesize[0];
s->uvlinesize = curframe->tf.f->linesize[1];
memset(s->top_nnz, 0, s->mb_width*sizeof(*s->top_nnz));
/* Zero macroblock structures for top/top-left prediction from outside the frame. */
memset(s->top_nnz, 0, s->mb_width * sizeof(*s->top_nnz));
/* Zero macroblock structures for top/top-left prediction
* from outside the frame. */
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)
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));
if (s->mb_layout == 1) {
// Make sure the previous frame has read its segmentation map,
// if we re-use the same map.
......@@ -1949,9 +2031,10 @@ int ff_vp8_decode_frame(AVCodecContext *avctx, void *data, int *got_frame,
s->mv_max.y = ((s->mb_height - 1) << 6) + MARGIN;
for (i = 0; i < MAX_THREADS; i++) {
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);
memcpy(&s->framep[0], &s->next_framep[0], sizeof(s->framep[0]) * 4);
......@@ -1965,7 +2048,7 @@ skip_decode:
if (!s->invisible) {
if ((ret = av_frame_ref(data, curframe->tf.f)) < 0)
return ret;
*got_frame = 1;
*got_frame = 1;
}
return avpkt->size;
......@@ -2033,10 +2116,10 @@ static av_cold int vp8_decode_init_thread_copy(AVCodecContext *avctx)
return 0;
}
#define REBASE(pic) \
pic ? pic - &s_src->frames[0] + &s->frames[0] : NULL
#define REBASE(pic) 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;
int i;
......@@ -2048,9 +2131,9 @@ static int vp8_decode_update_thread_context(AVCodecContext *dst, const AVCodecCo
s->mb_height = s_src->mb_height;
}
s->prob[0] = s_src->prob[!s_src->update_probabilities];
s->prob[0] = s_src->prob[!s_src->update_probabilities];
s->segmentation = s_src->segmentation;
s->lf_delta = s_src->lf_delta;
s->lf_delta = s_src->lf_delta;
memcpy(s->sign_bias, s_src->sign_bias, sizeof(s->sign_bias));
for (i = 0; i < FF_ARRAY_ELEMS(s_src->frames); i++) {
......
libavcodec/vp8.h
View file @ 53c20f17
......@@ -28,14 +28,15 @@
#include "libavutil/buffer.h"
#include "vp56.h"
#include "vp8dsp.h"
#include "h264pred.h"
#include "thread.h"
#include "vp56.h"
#include "vp8dsp.h"
#if HAVE_PTHREADS
#include <pthread.h>
# include <pthread.h>
#elif HAVE_W32THREADS
#include "compat/w32pthreads.h"
# include "compat/w32pthreads.h"
#endif
#define VP8_MAX_QUANT 127
......@@ -82,7 +83,7 @@ typedef struct VP8FilterStrength {
typedef struct VP8Macroblock {
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?
uint8_t mode;
uint8_t ref_frame;
......@@ -116,7 +117,7 @@ typedef struct VP8ThreadData {
int thread_nr;
#if HAVE_THREADS
pthread_mutex_t lock;
pthread_cond_t cond;
pthread_cond_t cond;
#endif
int thread_mb_pos; // (mb_y << 16) | (mb_x & 0xFFFF)
int wait_mb_pos; // What the current thread is waiting on.
......@@ -203,7 +204,7 @@ typedef struct VP8Context {
* [7] - split mv
* i16x16 modes never have any adjustment
*/
int8_t mode[VP8_MVMODE_SPLIT+1];
int8_t mode[VP8_MVMODE_SPLIT + 1];
/**
* filter strength adjustment for macroblocks that reference:
......@@ -215,7 +216,7 @@ typedef struct VP8Context {
int8_t ref[4];
} lf_delta;
uint8_t (*top_border)[16+8+8];
uint8_t (*top_border)[16 + 8 + 8];
uint8_t (*top_nnz)[9];
VP56RangeCoder c; ///< header context, includes mb modes and motion vectors
......@@ -234,7 +235,7 @@ typedef struct VP8Context {
uint8_t golden;
uint8_t pred16x16[4];
uint8_t pred8x8c[3];
uint8_t token[4][16][3][NUM_DCT_TOKENS-1];
uint8_t token[4][16][3][NUM_DCT_TOKENS - 1];
uint8_t mvc[2][19];
} prob[2];
......
libavcodec/vp8_parser.c
View file @ 53c20f17
......@@ -21,18 +21,19 @@
#include "parser.h"
static int parse(AVCodecParserContext *s,
AVCodecContext *avctx,
const uint8_t **poutbuf, int *poutbuf_size,
const uint8_t *buf, int buf_size)
AVCodecContext *avctx,
const uint8_t **poutbuf, int *poutbuf_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;
return buf_size;
}
AVCodecParser ff_vp8_parser = {
.codec_ids = { AV_CODEC_ID_VP8 },
.parser_parse = parse,
.codec_ids = { AV_CODEC_ID_VP8 },
.parser_parse = parse,
};
libavcodec/vp8data.h
View file @ 53c20f17
......@@ -30,28 +30,25 @@
#include "vp8.h"
#include "h264pred.h"
static const uint8_t vp8_pred4x4_mode[] =
{
static const uint8_t vp8_pred4x4_mode[] = {
[DC_PRED8x8] = DC_PRED,
[VERT_PRED8x8] = VERT_PRED,
[HOR_PRED8x8] = HOR_PRED,
[PLANE_PRED8x8] = TM_VP8_PRED,
};
static const int8_t vp8_pred16x16_tree_intra[4][2] =
{
{ -MODE_I4x4, 1 }, // '0'
{ 2, 3 },
{ -DC_PRED8x8, -VERT_PRED8x8 }, // '100', '101'
{ -HOR_PRED8x8, -PLANE_PRED8x8 }, // '110', '111'
static const int8_t vp8_pred16x16_tree_intra[4][2] = {
{ -MODE_I4x4, 1 }, // '0'
{ 2, 3 },
{ -DC_PRED8x8, -VERT_PRED8x8 }, // '100', '101'
{ -HOR_PRED8x8, -PLANE_PRED8x8 }, // '110', '111'
};
static const int8_t vp8_pred16x16_tree_inter[4][2] =
{
{ -DC_PRED8x8, 1 }, // '0'
{ 2, 3 },
{ -VERT_PRED8x8, -HOR_PRED8x8 }, // '100', '101'
{ -PLANE_PRED8x8, -MODE_I4x4 }, // '110', '111'
static const int8_t vp8_pred16x16_tree_inter[4][2] = {
{ -DC_PRED8x8, 1 }, // '0'
{ 2, 3 },
{ -VERT_PRED8x8, -HOR_PRED8x8 }, // '100', '101'
{ -PLANE_PRED8x8, -MODE_I4x4 }, // '110', '111'
};
static const int vp8_mode_contexts[6][4] = {
......@@ -64,26 +61,26 @@ static const int vp8_mode_contexts[6][4] = {
};
static const uint8_t vp8_mbsplits[5][16] = {
{ 0, 0, 0, 0, 0, 0, 0, 0,
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,
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 }
{ 0, 0, 0, 0, 0, 0, 0, 0, 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, 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] = {
{ 0, 8 }, { 0, 2 }, { 0, 2, 8, 10 },
{ 0, 1, 2, 3, 4, 5, 6, 7,
8, 9, 10, 11, 12, 13, 14, 15 }
{ 0, 8 },
{ 0, 2 },
{ 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_prob[3] = { 110, 111, 150 };
static const uint8_t vp8_mbsplit_count[4] = {
2, 2, 4, 16
};
static const uint8_t vp8_mbsplit_prob[3] = {
110, 111, 150
};
static const uint8_t vp8_submv_prob[5][3] = {
{ 147, 136, 18 },
......@@ -93,39 +90,42 @@ static const uint8_t vp8_submv_prob[5][3] = {
{ 208, 1, 1 }
};
static const uint8_t vp8_pred16x16_prob_intra[4] = { 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] =
{
{ -DC_PRED, 1 }, // '0'
{ -TM_VP8_PRED, 2 }, // '10'
{ -VERT_PRED, 3 }, // '110'
{ 4, 6 },
{ -HOR_PRED, 5 }, // '11100'
{ -DIAG_DOWN_RIGHT_PRED, -VERT_RIGHT_PRED }, // '111010', '111011'
{ -DIAG_DOWN_LEFT_PRED, 7 }, // '11110'
{ -VERT_LEFT_PRED, 8 }, // '111110'
{ -HOR_DOWN_PRED, -HOR_UP_PRED }, // '1111110', '1111111'
static const uint8_t vp8_pred16x16_prob_intra[4] = {
145, 156, 163, 128
};
static const uint8_t vp8_pred16x16_prob_inter[4] = {
112, 86, 140, 37
};
static const int8_t vp8_pred8x8c_tree[3][2] =
{
{ -DC_PRED8x8, 1 }, // '0'
{ -VERT_PRED8x8, 2 }, // '10
{ -HOR_PRED8x8, -PLANE_PRED8x8 }, // '110', '111'
static const int8_t vp8_pred4x4_tree[9][2] = {
{ -DC_PRED, 1 }, // '0'
{ -TM_VP8_PRED, 2 }, // '10'
{ -VERT_PRED, 3 }, // '110'
{ 4, 6 },
{ -HOR_PRED, 5 }, // '11100'
{ -DIAG_DOWN_RIGHT_PRED, -VERT_RIGHT_PRED }, // '111010', '111011'
{ -DIAG_DOWN_LEFT_PRED, 7 }, // '11110'
{ -VERT_LEFT_PRED, 8 }, // '111110'
{ -HOR_DOWN_PRED, -HOR_UP_PRED }, // '1111110', '1111111'
};
static const uint8_t vp8_pred8x8c_prob_intra[3] = { 142, 114, 183 };
static const uint8_t vp8_pred8x8c_prob_inter[3] = { 162, 101, 204 };
static const int8_t vp8_pred8x8c_tree[3][2] = {
{ -DC_PRED8x8, 1 }, // '0'
{ -VERT_PRED8x8, 2 }, // '10
{ -HOR_PRED8x8, -PLANE_PRED8x8 }, // '110', '111'
};
static const uint8_t vp8_pred4x4_prob_inter[9] =
{
static const uint8_t vp8_pred8x8c_prob_intra[3] = {
142, 114, 183
};
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
};
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 },
{ 56, 34, 51, 104, 114, 102, 29, 93, 77 },
......@@ -248,50 +248,57 @@ static const uint8_t vp8_pred4x4_prob_intra[10][10][9] =
},
};
static const int8_t vp8_segmentid_tree[][2] =
{
{ 1, 2 },
{ -0, -1 }, // '00', '01'
{ -2, -3 }, // '10', '11'
static const int8_t vp8_segmentid_tree[][2] = {
{ 1, 2 },
{ -0, -1 }, // '00', '01'
{ -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
};
/* Inverse of vp8_coeff_band: mappings of bands to coefficient indexes.
* Each list is -1-terminated. */
static const int8_t vp8_coeff_band_indexes[8][10] =
{
{0, -1},
{1, -1},
{2, -1},
{3, -1},
{5, -1},
{6, -1},
{4, 7, 8, 9, 10, 11, 12, 13, 14, -1},
{15, -1}
static const int8_t vp8_coeff_band_indexes[8][10] = {
{ 0, -1 },
{ 1, -1 },
{ 2, -1 },
{ 3, -1 },
{ 5, -1 },
{ 6, -1 },
{ 4, 7, 8, 9, 10, 11, 12, 13, 14, -1 },
{ 15, -1 }
};
static const uint8_t vp8_dct_cat1_prob[] = { 159, 0 };
static const uint8_t vp8_dct_cat2_prob[] = { 165, 145, 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 };
static const uint8_t vp8_dct_cat1_prob[] = {
159, 0
};
static const uint8_t vp8_dct_cat2_prob[] = {
165, 145, 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
const uint8_t * const ff_vp8_dct_cat_prob[] =
{
const uint8_t *const ff_vp8_dct_cat_prob[] = {
vp8_dct_cat3_prob,
vp8_dct_cat4_prob,
vp8_dct_cat5_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 },
......@@ -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 },
......@@ -635,15 +641,14 @@ static const uint8_t vp8_token_update_probs[4][8][3][NUM_DCT_TOKENS-1] =
};
// fixme: copied from h264data.h
static const uint8_t zigzag_scan[16]={
0+0*4, 1+0*4, 0+1*4, 0+2*4,
1+1*4, 2+0*4, 3+0*4, 2+1*4,
1+2*4, 0+3*4, 1+3*4, 2+2*4,
3+1*4, 3+2*4, 2+3*4, 3+3*4,
static const uint8_t zigzag_scan[16] = {
0 + 0 * 4, 1 + 0 * 4, 0 + 1 * 4, 0 + 2 * 4,
1 + 1 * 4, 2 + 0 * 4, 3 + 0 * 4, 2 + 1 * 4,
1 + 2 * 4, 0 + 3 * 4, 1 + 3 * 4, 2 + 2 * 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,
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,
......@@ -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,
};
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,
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,
......
libavcodec/vp8dsp.c
View file @ 53c20f17
......@@ -24,9 +24,10 @@
* VP8 compatible video decoder
*/
#include "libavutil/common.h"
#include "mathops.h"
#include "vp8dsp.h"
#include "libavutil/common.h"
// TODO: Maybe add dequant
static void vp8_luma_dc_wht_c(int16_t block[4][4][16], int16_t dc[16])
......@@ -34,26 +35,26 @@ static void vp8_luma_dc_wht_c(int16_t block[4][4][16], int16_t dc[16])
int i, t0, t1, t2, t3;
for (i = 0; i < 4; i++) {
t0 = dc[0*4+i] + dc[3*4+i];
t1 = dc[1*4+i] + dc[2*4+i];
t2 = dc[1*4+i] - dc[2*4+i];
t3 = dc[0*4+i] - dc[3*4+i];
dc[0*4+i] = t0 + t1;
dc[1*4+i] = t3 + t2;
dc[2*4+i] = t0 - t1;
dc[3*4+i] = t3 - t2;
t0 = dc[0 * 4 + i] + dc[3 * 4 + i];
t1 = dc[1 * 4 + i] + dc[2 * 4 + i];
t2 = dc[1 * 4 + i] - dc[2 * 4 + i];
t3 = dc[0 * 4 + i] - dc[3 * 4 + i];
dc[0 * 4 + i] = t0 + t1;
dc[1 * 4 + i] = t3 + t2;
dc[2 * 4 + i] = t0 - t1;
dc[3 * 4 + i] = t3 - t2;
}
for (i = 0; i < 4; i++) {
t0 = dc[i*4+0] + dc[i*4+3] + 3; // rounding
t1 = dc[i*4+1] + dc[i*4+2];
t2 = dc[i*4+1] - dc[i*4+2];
t3 = dc[i*4+0] - dc[i*4+3] + 3; // rounding
dc[i*4+0] = 0;
dc[i*4+1] = 0;
dc[i*4+2] = 0;
dc[i*4+3] = 0;
t0 = dc[i * 4 + 0] + dc[i * 4 + 3] + 3; // rounding
t1 = dc[i * 4 + 1] + dc[i * 4 + 2];
t2 = dc[i * 4 + 1] - dc[i * 4 + 2];
t3 = dc[i * 4 + 0] - dc[i * 4 + 3] + 3; // rounding
dc[i * 4 + 0] = 0;
dc[i * 4 + 1] = 0;
dc[i * 4 + 2] = 0;
dc[i * 4 + 3] = 0;
block[i][0][0] = (t0 + t1) >> 3;
block[i][1][0] = (t3 + t2) >> 3;
......@@ -75,8 +76,8 @@ static void vp8_luma_dc_wht_dc_c(int16_t block[4][4][16], int16_t dc[16])
}
}
#define MUL_20091(a) ((((a)*20091) >> 16) + (a))
#define MUL_35468(a) (((a)*35468) >> 16)
#define MUL_20091(a) ((((a) * 20091) >> 16) + (a))
#define MUL_35468(a) (((a) * 35468) >> 16)
static void vp8_idct_add_c(uint8_t *dst, int16_t block[16], ptrdiff_t stride)
{
......@@ -84,32 +85,32 @@ static void vp8_idct_add_c(uint8_t *dst, int16_t block[16], ptrdiff_t stride)
int16_t tmp[16];
for (i = 0; i < 4; i++) {
t0 = block[0*4+i] + block[2*4+i];
t1 = block[0*4+i] - block[2*4+i];
t2 = MUL_35468(block[1*4+i]) - MUL_20091(block[3*4+i]);
t3 = MUL_20091(block[1*4+i]) + MUL_35468(block[3*4+i]);
block[0*4+i] = 0;
block[1*4+i] = 0;
block[2*4+i] = 0;
block[3*4+i] = 0;
tmp[i*4+0] = t0 + t3;
tmp[i*4+1] = t1 + t2;
tmp[i*4+2] = t1 - t2;
tmp[i*4+3] = t0 - t3;
t0 = block[0 * 4 + i] + block[2 * 4 + i];
t1 = block[0 * 4 + i] - block[2 * 4 + i];
t2 = MUL_35468(block[1 * 4 + i]) - MUL_20091(block[3 * 4 + i]);
t3 = MUL_20091(block[1 * 4 + i]) + MUL_35468(block[3 * 4 + i]);
block[0 * 4 + i] = 0;
block[1 * 4 + i] = 0;
block[2 * 4 + i] = 0;
block[3 * 4 + i] = 0;
tmp[i * 4 + 0] = t0 + t3;
tmp[i * 4 + 1] = t1 + t2;
tmp[i * 4 + 2] = t1 - t2;
tmp[i * 4 + 3] = t0 - t3;
}
for (i = 0; i < 4; i++) {
t0 = tmp[0*4+i] + tmp[2*4+i];
t1 = tmp[0*4+i] - tmp[2*4+i];
t2 = MUL_35468(tmp[1*4+i]) - MUL_20091(tmp[3*4+i]);
t3 = MUL_20091(tmp[1*4+i]) + MUL_35468(tmp[3*4+i]);
t0 = tmp[0 * 4 + i] + tmp[2 * 4 + i];
t1 = tmp[0 * 4 + i] - tmp[2 * 4 + i];
t2 = MUL_35468(tmp[1 * 4 + i]) - MUL_20091(tmp[3 * 4 + i]);
t3 = MUL_20091(tmp[1 * 4 + i]) + MUL_35468(tmp[3 * 4 + i]);
dst[0] = av_clip_uint8(dst[0] + ((t0 + t3 + 4) >> 3));
dst[1] = av_clip_uint8(dst[1] + ((t1 + t2 + 4) >> 3));
dst[2] = av_clip_uint8(dst[2] + ((t1 - t2 + 4) >> 3));
dst[3] = av_clip_uint8(dst[3] + ((t0 - t3 + 4) >> 3));
dst += stride;
dst += stride;
}
}
......@@ -123,46 +124,49 @@ static void vp8_idct_dc_add_c(uint8_t *dst, int16_t block[16], ptrdiff_t stride)
dst[1] = av_clip_uint8(dst[1] + dc);
dst[2] = av_clip_uint8(dst[2] + dc);
dst[3] = av_clip_uint8(dst[3] + dc);
dst += stride;
dst += 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+4, block[1], stride);
vp8_idct_dc_add_c(dst+stride*4+0, block[2], stride);
vp8_idct_dc_add_c(dst+stride*4+4, block[3], 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 * 4 + 0, block[2], 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+ 4, block[1], stride);
vp8_idct_dc_add_c(dst+ 8, block[2], stride);
vp8_idct_dc_add_c(dst+12, block[3], 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 + 8, block[2], stride);
vp8_idct_dc_add_c(dst + 12, block[3], stride);
}
// because I like only having two parameters to pass functions...
#define LOAD_PIXELS\
int av_unused p3 = p[-4*stride];\
int av_unused p2 = p[-3*stride];\
int av_unused p1 = p[-2*stride];\
int av_unused p0 = p[-1*stride];\
int av_unused q0 = p[ 0*stride];\
int av_unused q1 = p[ 1*stride];\
int av_unused q2 = p[ 2*stride];\
int av_unused q3 = p[ 3*stride];
#define clip_int8(n) (cm[n+0x80]-0x80)
static av_always_inline void filter_common(uint8_t *p, ptrdiff_t stride, int is4tap)
#define LOAD_PIXELS \
int av_unused p3 = p[-4 * stride]; \
int av_unused p2 = p[-3 * stride]; \
int av_unused p1 = p[-2 * stride]; \
int av_unused p0 = p[-1 * stride]; \
int av_unused q0 = p[ 0 * stride]; \
int av_unused q1 = p[ 1 * stride]; \
int av_unused q2 = p[ 2 * stride]; \
int av_unused q3 = p[ 3 * stride];
#define clip_int8(n) (cm[n + 0x80] - 0x80)
static av_always_inline void filter_common(uint8_t *p, ptrdiff_t stride,
int is4tap)
{
LOAD_PIXELS
int a, f1, f2;
const uint8_t *cm = ff_crop_tab + MAX_NEG_CROP;
a = 3*(q0 - p0);
a = 3 * (q0 - p0);
if (is4tap)
a += clip_int8(p1 - q1);
......@@ -171,45 +175,50 @@ static av_always_inline void filter_common(uint8_t *p, ptrdiff_t stride, int is4
// We deviate from the spec here with c(a+3) >> 3
// since that's what libvpx does.
f1 = FFMIN(a+4, 127) >> 3;
f2 = FFMIN(a+3, 127) >> 3;
f1 = FFMIN(a + 4, 127) >> 3;
f2 = FFMIN(a + 3, 127) >> 3;
// Despite what the spec says, we do need to clamp here to
// be bitexact with libvpx.
p[-1*stride] = cm[p0 + f2];
p[ 0*stride] = cm[q0 - f1];
p[-1 * stride] = cm[p0 + f2];
p[ 0 * stride] = cm[q0 - f1];
// only used for _inner on blocks without high edge variance
if (!is4tap) {
a = (f1+1)>>1;
p[-2*stride] = cm[p1 + a];
p[ 1*stride] = cm[q1 - a];
a = (f1 + 1) >> 1;
p[-2 * stride] = cm[p1 + a];
p[ 1 * stride] = cm[q1 - a];
}
}
static av_always_inline int simple_limit(uint8_t *p, ptrdiff_t stride, int flim)
{
LOAD_PIXELS
return 2*FFABS(p0-q0) + (FFABS(p1-q1) >> 1) <= flim;
return 2 * FFABS(p0 - q0) + (FFABS(p1 - q1) >> 1) <= flim;
}
/**
* E - limit at the macroblock edge
* 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
return simple_limit(p, stride, E)
&& FFABS(p3-p2) <= I && FFABS(p2-p1) <= I && FFABS(p1-p0) <= I
&& FFABS(q3-q2) <= I && FFABS(q2-q1) <= I && FFABS(q1-q0) <= I;
return simple_limit(p, stride, E) &&
FFABS(p3 - p2) <= I &&
FFABS(p2 - p1) <= I &&
FFABS(p1 - p0) <= I &&
FFABS(q3 - q2) <= I &&
FFABS(q2 - q1) <= I &&
FFABS(q1 - q0) <= I;
}
// high edge variance
static av_always_inline int hev(uint8_t *p, ptrdiff_t stride, int thresh)
{
LOAD_PIXELS
return FFABS(p1-p0) > thresh || FFABS(q1-q0) > thresh;
return FFABS(p1 - p0) > thresh || FFABS(q1 - q0) > thresh;
}
static av_always_inline void filter_mbedge(uint8_t *p, ptrdiff_t stride)
......@@ -219,67 +228,75 @@ static av_always_inline void filter_mbedge(uint8_t *p, ptrdiff_t stride)
LOAD_PIXELS
w = clip_int8(p1-q1);
w = clip_int8(w + 3*(q0-p0));
w = clip_int8(p1 - q1);
w = clip_int8(w + 3 * (q0 - p0));
a0 = (27*w + 63) >> 7;
a1 = (18*w + 63) >> 7;
a2 = ( 9*w + 63) >> 7;
a0 = (27 * w + 63) >> 7;
a1 = (18 * w + 63) >> 7;
a2 = (9 * w + 63) >> 7;
p[-3*stride] = cm[p2 + a2];
p[-2*stride] = cm[p1 + a1];
p[-1*stride] = cm[p0 + a0];
p[ 0*stride] = cm[q0 - a0];
p[ 1*stride] = cm[q1 - a1];
p[ 2*stride] = cm[q2 - a2];
p[-3 * stride] = cm[p2 + a2];
p[-2 * stride] = cm[p1 + a1];
p[-1 * stride] = cm[p0 + a0];
p[ 0 * stride] = cm[q0 - a0];
p[ 1 * stride] = cm[q1 - a1];
p[ 2 * stride] = cm[q2 - a2];
}
#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,\
int flim_E, int flim_I, int hev_thresh)\
{\
int i;\
\
for (i = 0; i < size; i++)\
if (normal_limit(dst+i*stridea, strideb, flim_E, flim_I)) {\
if (hev(dst+i*stridea, strideb, hev_thresh))\
filter_common(dst+i*stridea, strideb, 1);\
else\
filter_mbedge(dst+i*stridea, strideb);\
}\
}\
\
static maybe_inline 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;\
\
for (i = 0; i < size; i++)\
if (normal_limit(dst+i*stridea, strideb, flim_E, flim_I)) {\
int hv = hev(dst+i*stridea, strideb, hev_thresh);\
if (hv) \
filter_common(dst+i*stridea, strideb, 1);\
else \
filter_common(dst+i*stridea, strideb, 0);\
}\
#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, \
int flim_E, int flim_I, \
int hev_thresh) \
{ \
int i; \
for (i = 0; i < size; i++) \
if (normal_limit(dst + i * stridea, strideb, flim_E, flim_I)) { \
if (hev(dst + i * stridea, strideb, hev_thresh)) \
filter_common(dst + i * stridea, strideb, 1); \
else \
filter_mbedge(dst + i * stridea, strideb); \
} \
} \
\
static maybe_inline \
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; \
for (i = 0; i < size; i++) \
if (normal_limit(dst + i * stridea, strideb, flim_E, flim_I)) { \
int hv = hev(dst + i * stridea, strideb, hev_thresh); \
if (hv) \
filter_common(dst + i * stridea, strideb, 1); \
else \
filter_common(dst + i * stridea, strideb, 0); \
} \
}
LOOP_FILTER(v, 16, 1, stride,)
LOOP_FILTER(h, 16, stride, 1,)
#define UV_LOOP_FILTER(dir, stridea, strideb) \
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,\
int fE, int fI, int hev_thresh)\
{\
vp8_ ## dir ## _loop_filter8_c(dstU, 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)\
{\
vp8_ ## dir ## _loop_filter8_inner_c(dstU, stride, fE, fI, hev_thresh);\
vp8_ ## dir ## _loop_filter8_inner_c(dstV, stride, fE, fI, hev_thresh);\
LOOP_FILTER(v, 16, 1, stride, )
LOOP_FILTER(h, 16, stride, 1, )
#define UV_LOOP_FILTER(dir, stridea, strideb) \
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, int fE, \
int fI, int hev_thresh) \
{ \
vp8_ ## dir ## _loop_filter8_c(dstU, 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) \
{ \
vp8_ ## dir ## _loop_filter8_inner_c(dstU, stride, fE, fI, hev_thresh); \
vp8_ ## dir ## _loop_filter8_inner_c(dstV, stride, fE, fI, hev_thresh); \
}
UV_LOOP_FILTER(v, 1, stride)
......@@ -290,8 +307,8 @@ static void vp8_v_loop_filter_simple_c(uint8_t *dst, ptrdiff_t stride, int flim)
int i;
for (i = 0; i < 16; i++)
if (simple_limit(dst+i, stride, flim))
filter_common(dst+i, stride, 1);
if (simple_limit(dst + i, stride, flim))
filter_common(dst + i, stride, 1);
}
static void vp8_h_loop_filter_simple_c(uint8_t *dst, ptrdiff_t stride, int flim)
......@@ -299,94 +316,110 @@ static void vp8_h_loop_filter_simple_c(uint8_t *dst, ptrdiff_t stride, int flim)
int i;
for (i = 0; i < 16; i++)
if (simple_limit(dst+i*stride, 1, flim))
filter_common(dst+i*stride, 1, 1);
if (simple_limit(dst + i * stride, 1, flim))
filter_common(dst + i * stride, 1, 1);
}
static const uint8_t subpel_filters[7][6] = {
{ 0, 6, 123, 12, 1, 0 },
{ 2, 11, 108, 36, 8, 1 },
{ 0, 9, 93, 50, 6, 0 },
{ 3, 16, 77, 77, 16, 3 },
{ 0, 6, 50, 93, 9, 0 },
{ 1, 8, 36, 108, 11, 2 },
{ 0, 1, 12, 123, 6, 0 },
{ 0, 6, 123, 12, 1, 0 },
{ 2, 11, 108, 36, 8, 1 },
{ 0, 9, 93, 50, 6, 0 },
{ 3, 16, 77, 77, 16, 3 },
{ 0, 6, 50, 93, 9, 0 },
{ 1, 8, 36, 108, 11, 2 },
{ 0, 1, 12, 123, 6, 0 },
};
#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) { \
int i; \
for (i = 0; i < h; i++, dst+= dststride, src+= srcstride) { \
memcpy(dst, src, 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) \
{ \
int i; \
for (i = 0; i < h; i++, dst += dststride, src += srcstride) \
memcpy(dst, src, WIDTH); \
}
PUT_PIXELS(16)
PUT_PIXELS(8)
PUT_PIXELS(4)
#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] + \
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) \
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]
#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) \
{ \
const uint8_t *filter = subpel_filters[mx-1]; \
const uint8_t *cm = ff_crop_tab + MAX_NEG_CROP; \
int x, y; \
\
for (y = 0; y < h; y++) { \
for (x = 0; x < SIZE; x++) \
dst[x] = FILTER_ ## TAPS ## TAP(src, filter, 1); \
dst += dststride; \
src += srcstride; \
} \
#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] + 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) \
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]
#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) \
{ \
const uint8_t *filter = subpel_filters[mx - 1]; \
const uint8_t *cm = ff_crop_tab + MAX_NEG_CROP; \
int x, y; \
for (y = 0; y < h; y++) { \
for (x = 0; x < SIZE; x++) \
dst[x] = FILTER_ ## TAPS ## TAP(src, filter, 1); \
dst += dststride; \
src += srcstride; \
} \
}
#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) \
{ \
const uint8_t *filter = subpel_filters[my-1]; \
const uint8_t *cm = ff_crop_tab + MAX_NEG_CROP; \
int x, y; \
\
for (y = 0; y < h; y++) { \
for (x = 0; x < SIZE; x++) \
dst[x] = FILTER_ ## TAPS ## TAP(src, filter, srcstride); \
dst += dststride; \
src += srcstride; \
} \
#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) \
{ \
const uint8_t *filter = subpel_filters[my - 1]; \
const uint8_t *cm = ff_crop_tab + MAX_NEG_CROP; \
int x, y; \
for (y = 0; y < h; y++) { \
for (x = 0; x < SIZE; x++) \
dst[x] = FILTER_ ## TAPS ## TAP(src, filter, srcstride); \
dst += dststride; \
src += srcstride; \
} \
}
#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) \
{ \
const uint8_t *filter = subpel_filters[mx-1]; \
const uint8_t *cm = ff_crop_tab + MAX_NEG_CROP; \
int x, y; \
uint8_t tmp_array[(2*SIZE+VTAPS-1)*SIZE]; \
uint8_t *tmp = tmp_array; \
src -= (2-(VTAPS==4))*srcstride; \
\
for (y = 0; y < h+VTAPS-1; y++) { \
for (x = 0; x < SIZE; x++) \
tmp[x] = FILTER_ ## HTAPS ## TAP(src, filter, 1); \
tmp += SIZE; \
src += srcstride; \
} \
\
tmp = tmp_array + (2-(VTAPS==4))*SIZE; \
filter = subpel_filters[my-1]; \
\
for (y = 0; y < h; y++) { \
for (x = 0; x < SIZE; x++) \
dst[x] = FILTER_ ## VTAPS ## TAP(tmp, filter, SIZE); \
dst += dststride; \
tmp += SIZE; \
} \
#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) \
{ \
const uint8_t *filter = subpel_filters[mx - 1]; \
const uint8_t *cm = ff_crop_tab + MAX_NEG_CROP; \
int x, y; \
uint8_t tmp_array[(2 * SIZE + VTAPS - 1) * SIZE]; \
uint8_t *tmp = tmp_array; \
src -= (2 - (VTAPS == 4)) * srcstride; \
\
for (y = 0; y < h + VTAPS - 1; y++) { \
for (x = 0; x < SIZE; x++) \
tmp[x] = FILTER_ ## HTAPS ## TAP(src, filter, 1); \
tmp += SIZE; \
src += srcstride; \
} \
tmp = tmp_array + (2 - (VTAPS == 4)) * SIZE; \
filter = subpel_filters[my - 1]; \
\
for (y = 0; y < h; y++) { \
for (x = 0; x < SIZE; x++) \
dst[x] = FILTER_ ## VTAPS ## TAP(tmp, filter, SIZE); \
dst += dststride; \
tmp += SIZE; \
} \
}
VP8_EPEL_H(16, 4)
......@@ -401,6 +434,7 @@ VP8_EPEL_V(4, 4)
VP8_EPEL_V(16, 6)
VP8_EPEL_V(8, 6)
VP8_EPEL_V(4, 6)
VP8_EPEL_HV(16, 4, 4)
VP8_EPEL_HV(8, 4, 4)
VP8_EPEL_HV(4, 4, 4)
......@@ -414,73 +448,77 @@ VP8_EPEL_HV(16, 6, 6)
VP8_EPEL_HV(8, 6, 6)
VP8_EPEL_HV(4, 6, 6)
#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) \
{ \
int a = 8-mx, b = mx; \
int x, y; \
\
for (y = 0; y < h; y++) { \
for (x = 0; x < SIZE; x++) \
dst[x] = (a*src[x] + b*src[x+1] + 4) >> 3; \
dst += dstride; \
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) \
{ \
int c = 8-my, d = my; \
int x, y; \
\
for (y = 0; y < h; y++) { \
for (x = 0; x < SIZE; x++) \
dst[x] = (c*src[x] + d*src[x+sstride] + 4) >> 3; \
dst += dstride; \
src += sstride; \
} \
} \
\
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 c = 8-my, d = my; \
int x, y; \
uint8_t tmp_array[(2*SIZE+1)*SIZE]; \
uint8_t *tmp = tmp_array; \
\
for (y = 0; y < h+1; y++) { \
for (x = 0; x < SIZE; x++) \
tmp[x] = (a*src[x] + b*src[x+1] + 4) >> 3; \
tmp += SIZE; \
src += sstride; \
} \
\
tmp = tmp_array; \
\
for (y = 0; y < h; y++) { \
for (x = 0; x < SIZE; x++) \
dst[x] = (c*tmp[x] + d*tmp[x+SIZE] + 4) >> 3; \
dst += dstride; \
tmp += 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) \
{ \
int a = 8 - mx, b = mx; \
int x, y; \
for (y = 0; y < h; y++) { \
for (x = 0; x < SIZE; x++) \
dst[x] = (a * src[x] + b * src[x + 1] + 4) >> 3; \
dst += dstride; \
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) \
{ \
int c = 8 - my, d = my; \
int x, y; \
for (y = 0; y < h; y++) { \
for (x = 0; x < SIZE; x++) \
dst[x] = (c * src[x] + d * src[x + sstride] + 4) >> 3; \
dst += dstride; \
src += sstride; \
} \
} \
\
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 c = 8 - my, d = my; \
int x, y; \
uint8_t tmp_array[(2 * SIZE + 1) * SIZE]; \
uint8_t *tmp = tmp_array; \
for (y = 0; y < h + 1; y++) { \
for (x = 0; x < SIZE; x++) \
tmp[x] = (a * src[x] + b * src[x + 1] + 4) >> 3; \
tmp += SIZE; \
src += sstride; \
} \
tmp = tmp_array; \
for (y = 0; y < h; y++) { \
for (x = 0; x < SIZE; x++) \
dst[x] = (c * tmp[x] + d * tmp[x + SIZE] + 4) >> 3; \
dst += dstride; \
tmp += SIZE; \
} \
}
VP8_BILINEAR(16)
VP8_BILINEAR(8)
VP8_BILINEAR(4)
#define VP8_MC_FUNC(IDX, SIZE) \
dsp->put_vp8_epel_pixels_tab[IDX][0][0] = put_vp8_pixels ## SIZE ## _c; \
dsp->put_vp8_epel_pixels_tab[IDX][0][1] = put_vp8_epel ## SIZE ## _h4_c; \
dsp->put_vp8_epel_pixels_tab[IDX][0][2] = put_vp8_epel ## SIZE ## _h6_c; \
dsp->put_vp8_epel_pixels_tab[IDX][1][0] = put_vp8_epel ## SIZE ## _v4_c; \
#define VP8_MC_FUNC(IDX, SIZE) \
dsp->put_vp8_epel_pixels_tab[IDX][0][0] = put_vp8_pixels ## SIZE ## _c; \
dsp->put_vp8_epel_pixels_tab[IDX][0][1] = put_vp8_epel ## SIZE ## _h4_c; \
dsp->put_vp8_epel_pixels_tab[IDX][0][2] = put_vp8_epel ## SIZE ## _h6_c; \
dsp->put_vp8_epel_pixels_tab[IDX][1][0] = put_vp8_epel ## SIZE ## _v4_c; \
dsp->put_vp8_epel_pixels_tab[IDX][1][1] = put_vp8_epel ## SIZE ## _h4v4_c; \
dsp->put_vp8_epel_pixels_tab[IDX][1][2] = put_vp8_epel ## SIZE ## _h6v4_c; \
dsp->put_vp8_epel_pixels_tab[IDX][2][0] = put_vp8_epel ## SIZE ## _v6_c; \
dsp->put_vp8_epel_pixels_tab[IDX][2][0] = put_vp8_epel ## SIZE ## _v6_c; \
dsp->put_vp8_epel_pixels_tab[IDX][2][1] = put_vp8_epel ## SIZE ## _h4v6_c; \
dsp->put_vp8_epel_pixels_tab[IDX][2][2] = put_vp8_epel ## SIZE ## _h6v6_c
#define VP8_BILINEAR_MC_FUNC(IDX, SIZE) \
#define VP8_BILINEAR_MC_FUNC(IDX, SIZE) \
dsp->put_vp8_bilinear_pixels_tab[IDX][0][0] = put_vp8_pixels ## SIZE ## _c; \
dsp->put_vp8_bilinear_pixels_tab[IDX][0][1] = put_vp8_bilinear ## SIZE ## _h_c; \
dsp->put_vp8_bilinear_pixels_tab[IDX][0][2] = put_vp8_bilinear ## SIZE ## _h_c; \
......
libavcodec/vp8dsp.h
View file @ 53c20f17
......@@ -30,8 +30,8 @@
#include <stddef.h>
#include <stdint.h>
typedef void (*vp8_mc_func)(uint8_t *dst/*align 8*/, ptrdiff_t dstStride,
uint8_t *src/*align 1*/, ptrdiff_t srcStride,
typedef void (*vp8_mc_func)(uint8_t *dst /* align 8 */, ptrdiff_t dstStride,
uint8_t *src /* align 1 */, ptrdiff_t srcStride,
int h, int x, int y);
typedef struct VP8DSPContext {
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment