int h_edge_pos;
int v_edge_pos;
int last_frame_output;
+
+ int mb_x, mb_y;
+ int mb_xy;
+
+ int chroma_pred_mode;
+ int intra16x16_pred_mode;
+
+ int8_t intra4x4_pred_mode_cache[5 * 8];
+ int8_t (*intra4x4_pred_mode);
+
+ unsigned int top_samples_available;
+ unsigned int topright_samples_available;
+ unsigned int left_samples_available;
} SVQ3Context;
#define FULLPEL_MODE 1
for (i = 0; i < 16; i += part_height)
for (j = 0; j < 16; j += part_width) {
- const int b_xy = (4 * sl->mb_x + (j >> 2)) +
- (4 * sl->mb_y + (i >> 2)) * h->b_stride;
+ const int b_xy = (4 * s->mb_x + (j >> 2)) +
+ (4 * s->mb_y + (i >> 2)) * h->b_stride;
int dxy;
- x = 16 * sl->mb_x + j;
- y = 16 * sl->mb_y + i;
+ x = 16 * s->mb_x + j;
+ y = 16 * s->mb_y + i;
k = (j >> 2 & 1) + (i >> 1 & 2) +
(j >> 1 & 4) + (i & 8);
if (IS_INTRA4x4(mb_type)) {
for (i = 0; i < 16; i++) {
uint8_t *const ptr = dest_y + block_offset[i];
- const int dir = sl->intra4x4_pred_mode_cache[scan8[i]];
+ const int dir = s->intra4x4_pred_mode_cache[scan8[i]];
uint8_t *topright;
int nnz, tr;
if (dir == DIAG_DOWN_LEFT_PRED || dir == VERT_LEFT_PRED) {
- const int topright_avail = (sl->topright_samples_available << i) & 0x8000;
- av_assert2(sl->mb_y || linesize <= block_offset[i]);
+ const int topright_avail = (s->topright_samples_available << i) & 0x8000;
+ av_assert2(s->mb_y || linesize <= block_offset[i]);
if (!topright_avail) {
tr = ptr[3 - linesize] * 0x01010101u;
topright = (uint8_t *)&tr;
}
}
} else {
- s->hpc.pred16x16[sl->intra16x16_pred_mode](dest_y, linesize);
+ s->hpc.pred16x16[s->intra16x16_pred_mode](dest_y, linesize);
svq3_luma_dc_dequant_idct_c(sl->mb, sl->mb_luma_dc[0], qscale);
}
}
static void hl_decode_mb(SVQ3Context *s, const H264Context *h, H264SliceContext *sl)
{
- const int mb_x = sl->mb_x;
- const int mb_y = sl->mb_y;
- const int mb_xy = sl->mb_xy;
+ const int mb_x = s->mb_x;
+ const int mb_y = s->mb_y;
+ const int mb_xy = s->mb_xy;
const int mb_type = h->cur_pic.mb_type[mb_xy];
uint8_t *dest_y, *dest_cb, *dest_cr;
int linesize, uvlinesize;
dest_cb = h->cur_pic.f->data[1] + mb_x * 8 + mb_y * sl->uvlinesize * block_h;
dest_cr = h->cur_pic.f->data[2] + mb_x * 8 + mb_y * sl->uvlinesize * block_h;
- s->vdsp.prefetch(dest_y + (sl->mb_x & 3) * 4 * sl->linesize + 64, sl->linesize, 4);
- s->vdsp.prefetch(dest_cb + (sl->mb_x & 7) * sl->uvlinesize + 64, dest_cr - dest_cb, 2);
+ s->vdsp.prefetch(dest_y + (s->mb_x & 3) * 4 * sl->linesize + 64, sl->linesize, 4);
+ s->vdsp.prefetch(dest_cb + (s->mb_x & 7) * sl->uvlinesize + 64, dest_cr - dest_cb, 2);
h->list_counts[mb_xy] = sl->list_count;
uvlinesize = sl->mb_uvlinesize = sl->uvlinesize;
if (IS_INTRA(mb_type)) {
- s->hpc.pred8x8[sl->chroma_pred_mode](dest_cb, uvlinesize);
- s->hpc.pred8x8[sl->chroma_pred_mode](dest_cr, uvlinesize);
+ s->hpc.pred8x8[s->chroma_pred_mode](dest_cb, uvlinesize);
+ s->hpc.pred8x8[s->chroma_pred_mode](dest_cr, uvlinesize);
hl_decode_mb_predict_luma(s, h, sl, mb_type, block_offset, linesize, dest_y);
}
int cbp = 0;
uint32_t vlc;
int8_t *top, *left;
- const int mb_xy = sl->mb_xy;
- const int b_xy = 4 * sl->mb_x + 4 * sl->mb_y * h->b_stride;
+ const int mb_xy = s->mb_xy;
+ const int b_xy = 4 * s->mb_x + 4 * s->mb_y * h->b_stride;
- sl->top_samples_available = (sl->mb_y == 0) ? 0x33FF : 0xFFFF;
- sl->left_samples_available = (sl->mb_x == 0) ? 0x5F5F : 0xFFFF;
- sl->topright_samples_available = 0xFFFF;
+ s->top_samples_available = (s->mb_y == 0) ? 0x33FF : 0xFFFF;
+ s->left_samples_available = (s->mb_x == 0) ? 0x5F5F : 0xFFFF;
+ s->topright_samples_available = 0xFFFF;
if (mb_type == 0) { /* SKIP */
if (h->pict_type == AV_PICTURE_TYPE_P ||
s->next_pic->mb_type[mb_xy] == -1) {
- svq3_mc_dir_part(s, 16 * sl->mb_x, 16 * sl->mb_y, 16, 16,
+ svq3_mc_dir_part(s, 16 * s->mb_x, 16 * s->mb_y, 16, 16,
0, 0, 0, 0, 0, 0);
if (h->pict_type == AV_PICTURE_TYPE_B)
- svq3_mc_dir_part(s, 16 * sl->mb_x, 16 * sl->mb_y, 16, 16,
+ svq3_mc_dir_part(s, 16 * s->mb_x, 16 * s->mb_y, 16, 16,
0, 0, 0, 0, 1, 1);
mb_type = MB_TYPE_SKIP;
*/
for (m = 0; m < 2; m++) {
- if (sl->mb_x > 0 && sl->intra4x4_pred_mode[h->mb2br_xy[mb_xy - 1] + 6] != -1) {
+ if (s->mb_x > 0 && s->intra4x4_pred_mode[h->mb2br_xy[mb_xy - 1] + 6] != -1) {
for (i = 0; i < 4; i++)
AV_COPY32(sl->mv_cache[m][scan8[0] - 1 + i * 8],
h->cur_pic.motion_val[m][b_xy - 1 + i * h->b_stride]);
for (i = 0; i < 4; i++)
AV_ZERO32(sl->mv_cache[m][scan8[0] - 1 + i * 8]);
}
- if (sl->mb_y > 0) {
+ if (s->mb_y > 0) {
memcpy(sl->mv_cache[m][scan8[0] - 1 * 8],
h->cur_pic.motion_val[m][b_xy - h->b_stride],
4 * 2 * sizeof(int16_t));
memset(&sl->ref_cache[m][scan8[0] - 1 * 8],
- (sl->intra4x4_pred_mode[h->mb2br_xy[mb_xy - h->mb_stride]] == -1) ? PART_NOT_AVAILABLE : 1, 4);
+ (s->intra4x4_pred_mode[h->mb2br_xy[mb_xy - h->mb_stride]] == -1) ? PART_NOT_AVAILABLE : 1, 4);
- if (sl->mb_x < h->mb_width - 1) {
+ if (s->mb_x < h->mb_width - 1) {
AV_COPY32(sl->mv_cache[m][scan8[0] + 4 - 1 * 8],
h->cur_pic.motion_val[m][b_xy - h->b_stride + 4]);
sl->ref_cache[m][scan8[0] + 4 - 1 * 8] =
- (sl->intra4x4_pred_mode[h->mb2br_xy[mb_xy - h->mb_stride + 1] + 6] == -1 ||
- sl->intra4x4_pred_mode[h->mb2br_xy[mb_xy - h->mb_stride]] == -1) ? PART_NOT_AVAILABLE : 1;
+ (s->intra4x4_pred_mode[h->mb2br_xy[mb_xy - h->mb_stride + 1] + 6] == -1 ||
+ s->intra4x4_pred_mode[h->mb2br_xy[mb_xy - h->mb_stride]] == -1) ? PART_NOT_AVAILABLE : 1;
} else
sl->ref_cache[m][scan8[0] + 4 - 1 * 8] = PART_NOT_AVAILABLE;
- if (sl->mb_x > 0) {
+ if (s->mb_x > 0) {
AV_COPY32(sl->mv_cache[m][scan8[0] - 1 - 1 * 8],
h->cur_pic.motion_val[m][b_xy - h->b_stride - 1]);
sl->ref_cache[m][scan8[0] - 1 - 1 * 8] =
- (sl->intra4x4_pred_mode[h->mb2br_xy[mb_xy - h->mb_stride - 1] + 3] == -1) ? PART_NOT_AVAILABLE : 1;
+ (s->intra4x4_pred_mode[h->mb2br_xy[mb_xy - h->mb_stride - 1] + 3] == -1) ? PART_NOT_AVAILABLE : 1;
} else
sl->ref_cache[m][scan8[0] - 1 - 1 * 8] = PART_NOT_AVAILABLE;
} else
mb_type = MB_TYPE_16x16;
} else if (mb_type == 8 || mb_type == 33) { /* INTRA4x4 */
- memset(sl->intra4x4_pred_mode_cache, -1, 8 * 5 * sizeof(int8_t));
+ int8_t *i4x4 = s->intra4x4_pred_mode + h->mb2br_xy[s->mb_xy];
+ int8_t *i4x4_cache = s->intra4x4_pred_mode_cache;
+
+ memset(s->intra4x4_pred_mode_cache, -1, 8 * 5 * sizeof(int8_t));
if (mb_type == 8) {
- if (sl->mb_x > 0) {
+ if (s->mb_x > 0) {
for (i = 0; i < 4; i++)
- sl->intra4x4_pred_mode_cache[scan8[0] - 1 + i * 8] = sl->intra4x4_pred_mode[h->mb2br_xy[mb_xy - 1] + 6 - i];
- if (sl->intra4x4_pred_mode_cache[scan8[0] - 1] == -1)
- sl->left_samples_available = 0x5F5F;
+ s->intra4x4_pred_mode_cache[scan8[0] - 1 + i * 8] = s->intra4x4_pred_mode[h->mb2br_xy[mb_xy - 1] + 6 - i];
+ if (s->intra4x4_pred_mode_cache[scan8[0] - 1] == -1)
+ s->left_samples_available = 0x5F5F;
}
- if (sl->mb_y > 0) {
- sl->intra4x4_pred_mode_cache[4 + 8 * 0] = sl->intra4x4_pred_mode[h->mb2br_xy[mb_xy - h->mb_stride] + 0];
- sl->intra4x4_pred_mode_cache[5 + 8 * 0] = sl->intra4x4_pred_mode[h->mb2br_xy[mb_xy - h->mb_stride] + 1];
- sl->intra4x4_pred_mode_cache[6 + 8 * 0] = sl->intra4x4_pred_mode[h->mb2br_xy[mb_xy - h->mb_stride] + 2];
- sl->intra4x4_pred_mode_cache[7 + 8 * 0] = sl->intra4x4_pred_mode[h->mb2br_xy[mb_xy - h->mb_stride] + 3];
-
- if (sl->intra4x4_pred_mode_cache[4 + 8 * 0] == -1)
- sl->top_samples_available = 0x33FF;
+ if (s->mb_y > 0) {
+ s->intra4x4_pred_mode_cache[4 + 8 * 0] = s->intra4x4_pred_mode[h->mb2br_xy[mb_xy - h->mb_stride] + 0];
+ s->intra4x4_pred_mode_cache[5 + 8 * 0] = s->intra4x4_pred_mode[h->mb2br_xy[mb_xy - h->mb_stride] + 1];
+ s->intra4x4_pred_mode_cache[6 + 8 * 0] = s->intra4x4_pred_mode[h->mb2br_xy[mb_xy - h->mb_stride] + 2];
+ s->intra4x4_pred_mode_cache[7 + 8 * 0] = s->intra4x4_pred_mode[h->mb2br_xy[mb_xy - h->mb_stride] + 3];
+
+ if (s->intra4x4_pred_mode_cache[4 + 8 * 0] == -1)
+ s->top_samples_available = 0x33FF;
}
/* decode prediction codes for luma blocks */
return -1;
}
- left = &sl->intra4x4_pred_mode_cache[scan8[i] - 1];
- top = &sl->intra4x4_pred_mode_cache[scan8[i] - 8];
+ left = &s->intra4x4_pred_mode_cache[scan8[i] - 1];
+ top = &s->intra4x4_pred_mode_cache[scan8[i] - 8];
left[1] = svq3_pred_1[top[0] + 1][left[0] + 1][svq3_pred_0[vlc][0]];
left[2] = svq3_pred_1[top[1] + 1][left[1] + 1][svq3_pred_0[vlc][1]];
}
} else { /* mb_type == 33, DC_128_PRED block type */
for (i = 0; i < 4; i++)
- memset(&sl->intra4x4_pred_mode_cache[scan8[0] + 8 * i], DC_PRED, 4);
+ memset(&s->intra4x4_pred_mode_cache[scan8[0] + 8 * i], DC_PRED, 4);
}
- write_back_intra_pred_mode(h, sl);
+ AV_COPY32(i4x4, i4x4_cache + 4 + 8 * 4);
+ i4x4[4] = i4x4_cache[7 + 8 * 3];
+ i4x4[5] = i4x4_cache[7 + 8 * 2];
+ i4x4[6] = i4x4_cache[7 + 8 * 1];
if (mb_type == 8) {
- ff_h264_check_intra4x4_pred_mode(h, sl);
+ ff_h264_check_intra4x4_pred_mode(s->intra4x4_pred_mode_cache,
+ h->avctx, s->top_samples_available,
+ s->left_samples_available);
- sl->top_samples_available = (sl->mb_y == 0) ? 0x33FF : 0xFFFF;
- sl->left_samples_available = (sl->mb_x == 0) ? 0x5F5F : 0xFFFF;
+ s->top_samples_available = (s->mb_y == 0) ? 0x33FF : 0xFFFF;
+ s->left_samples_available = (s->mb_x == 0) ? 0x5F5F : 0xFFFF;
} else {
for (i = 0; i < 4; i++)
- memset(&sl->intra4x4_pred_mode_cache[scan8[0] + 8 * i], DC_128_PRED, 4);
+ memset(&s->intra4x4_pred_mode_cache[scan8[0] + 8 * i], DC_128_PRED, 4);
- sl->top_samples_available = 0x33FF;
- sl->left_samples_available = 0x5F5F;
+ s->top_samples_available = 0x33FF;
+ s->left_samples_available = 0x5F5F;
}
mb_type = MB_TYPE_INTRA4x4;
dir = ff_h264_i_mb_type_info[mb_type - 8].pred_mode;
dir = (dir >> 1) ^ 3 * (dir & 1) ^ 1;
- if ((sl->intra16x16_pred_mode = ff_h264_check_intra_pred_mode(h, sl, dir, 0)) < 0) {
+ if ((s->intra16x16_pred_mode = ff_h264_check_intra_pred_mode(h->avctx, s->top_samples_available,
+ s->left_samples_available, dir, 0)) < 0) {
av_log(h->avctx, AV_LOG_ERROR, "ff_h264_check_intra_pred_mode < 0\n");
- return sl->intra16x16_pred_mode;
+ return s->intra16x16_pred_mode;
}
cbp = ff_h264_i_mb_type_info[mb_type - 8].cbp;
}
}
if (!IS_INTRA4x4(mb_type)) {
- memset(sl->intra4x4_pred_mode + h->mb2br_xy[mb_xy], DC_PRED, 8);
+ memset(s->intra4x4_pred_mode + h->mb2br_xy[mb_xy], DC_PRED, 8);
}
if (!IS_SKIP(mb_type) || h->pict_type == AV_PICTURE_TYPE_B) {
memset(sl->non_zero_count_cache + 8, 0, 14 * 8 * sizeof(uint8_t));
h->cur_pic.mb_type[mb_xy] = mb_type;
if (IS_INTRA(mb_type))
- sl->chroma_pred_mode = ff_h264_check_intra_pred_mode(h, sl, DC_PRED8x8, 1);
+ s->chroma_pred_mode = ff_h264_check_intra_pred_mode(h->avctx, s->top_samples_available,
+ s->left_samples_available, DC_PRED8x8, 1);
return 0;
}
SVQ3Context *s = avctx->priv_data;
H264Context *h = &s->h;
H264SliceContext *sl = &h->slice_ctx[0];
- const int mb_xy = sl->mb_xy;
+ const int mb_xy = s->mb_xy;
int i, header;
unsigned slice_id;
if ((header & 0x9F) == 2) {
i = (h->mb_num < 64) ? 6 : (1 + av_log2(h->mb_num - 1));
sl->mb_skip_run = get_bits(&h->gb, i) -
- (sl->mb_y * h->mb_width + sl->mb_x);
+ (s->mb_y * h->mb_width + s->mb_x);
} else {
skip_bits1(&h->gb);
sl->mb_skip_run = 0;
return AVERROR_INVALIDDATA;
/* reset intra predictors and invalidate motion vector references */
- if (sl->mb_x > 0) {
- memset(sl->intra4x4_pred_mode + h->mb2br_xy[mb_xy - 1] + 3,
+ if (s->mb_x > 0) {
+ memset(s->intra4x4_pred_mode + h->mb2br_xy[mb_xy - 1] + 3,
-1, 4 * sizeof(int8_t));
- memset(sl->intra4x4_pred_mode + h->mb2br_xy[mb_xy - sl->mb_x],
- -1, 8 * sizeof(int8_t) * sl->mb_x);
+ memset(s->intra4x4_pred_mode + h->mb2br_xy[mb_xy - s->mb_x],
+ -1, 8 * sizeof(int8_t) * s->mb_x);
}
- if (sl->mb_y > 0) {
- memset(sl->intra4x4_pred_mode + h->mb2br_xy[mb_xy - h->mb_stride],
- -1, 8 * sizeof(int8_t) * (h->mb_width - sl->mb_x));
+ if (s->mb_y > 0) {
+ memset(s->intra4x4_pred_mode + h->mb2br_xy[mb_xy - h->mb_stride],
+ -1, 8 * sizeof(int8_t) * (h->mb_width - s->mb_x));
- if (sl->mb_x > 0)
- sl->intra4x4_pred_mode[h->mb2br_xy[mb_xy - h->mb_stride - 1] + 3] = -1;
+ if (s->mb_x > 0)
+ s->intra4x4_pred_mode[h->mb2br_xy[mb_xy - h->mb_stride - 1] + 3] = -1;
}
return 0;
}
}
- for (sl->mb_y = 0; sl->mb_y < h->mb_height; sl->mb_y++) {
- for (sl->mb_x = 0; sl->mb_x < h->mb_width; sl->mb_x++) {
+ for (s->mb_y = 0; s->mb_y < h->mb_height; s->mb_y++) {
+ for (s->mb_x = 0; s->mb_x < h->mb_width; s->mb_x++) {
unsigned mb_type;
- sl->mb_xy = sl->mb_x + sl->mb_y * h->mb_stride;
+ s->mb_xy = s->mb_x + s->mb_y * h->mb_stride;
if ((get_bits_left(&h->gb)) <= 7) {
if (((get_bits_count(&h->gb) & 7) == 0 ||
mb_type += 4;
if (mb_type > 33 || svq3_decode_mb(s, mb_type)) {
av_log(h->avctx, AV_LOG_ERROR,
- "error while decoding MB %d %d\n", sl->mb_x, sl->mb_y);
+ "error while decoding MB %d %d\n", s->mb_x, s->mb_y);
return -1;
}
hl_decode_mb(s, h, &h->slice_ctx[0]);
if (h->pict_type != AV_PICTURE_TYPE_B && !h->low_delay)
- h->cur_pic.mb_type[sl->mb_x + sl->mb_y * h->mb_stride] =
+ h->cur_pic.mb_type[s->mb_x + s->mb_y * h->mb_stride] =
(h->pict_type == AV_PICTURE_TYPE_P && mb_type < 8) ? (mb_type - 1) : -1;
}
ff_draw_horiz_band(avctx, s->cur_pic->f,
s->last_pic->f->data[0] ? s->last_pic->f : NULL,
- 16 * sl->mb_y, 16, h->picture_structure, 0,
+ 16 * s->mb_y, 16, h->picture_structure, 0,
h->low_delay);
}
av_freep(&s->next_pic);
av_freep(&s->last_pic);
av_freep(&s->slice_buf);
+ av_freep(&s->intra4x4_pred_mode);
memset(&h->cur_pic, 0, sizeof(h->cur_pic));
projects / ffmpeg / blobdiff