vp56_rac_get_prob_branchy(&s->c,
s->prob.segpred[s->above_segpred_ctx[col] +
s->left_segpred_ctx[row7]]))) {
- uint8_t *refsegmap = s->frames[LAST_FRAME].segmentation_map;
- int pred = MAX_SEGMENT - 1;
- int x;
-
- for (y = 0; y < h4; y++)
- for (x = 0; x < w4; x++)
- pred = FFMIN(pred,
- refsegmap[(y + row) * 8 * s->sb_cols + x + col]);
- b->seg_id = pred;
+ if (!s->errorres) {
+ uint8_t *refsegmap = s->frames[LAST_FRAME].segmentation_map;
+ int pred = MAX_SEGMENT - 1;
+ int x;
+
+ if (!s->last_uses_2pass)
+ ff_thread_await_progress(&s->frames[LAST_FRAME].tf, row >> 3, 0);
+
+ for (y = 0; y < h4; y++)
+ for (x = 0; x < w4; x++)
+ pred = FFMIN(pred,
+ refsegmap[(y + row) * 8 * s->sb_cols + x + col]);
+ b->seg_id = pred;
+ } else {
+ b->seg_id = 0;
+ }
memset(&s->above_segpred_ctx[col], 1, w4);
memset(&s->left_segpred_ctx[row7], 1, h4);
uint8_t *dst, ptrdiff_t dst_stride,
const uint8_t *ref,
ptrdiff_t ref_stride,
+ ThreadFrame *ref_frame,
ptrdiff_t y, ptrdiff_t x,
const VP56mv *mv,
int bw, int bh, int w, int h)
{
int mx = mv->x, my = mv->y;
+ int th;
y += my >> 3;
x += mx >> 3;
ref += y * ref_stride + x;
mx &= 7;
my &= 7;
+
+ // we use +7 because the last 7 pixels of each sbrow can be changed in
+ // the longest loopfilter of the next sbrow
+ th = (y + bh + 4 * !!my + 7) >> 6;
+ ff_thread_await_progress(ref_frame, FFMAX(th, 0), 0);
+
// FIXME bilinear filter only needs 0/1 pixels, not 3/4
+ // The arm/aarch64 _hv filters read one more row than what actually is
+ // needed, so switch to emulated edge one pixel sooner vertically
+ // (!!my * 5) than horizontally (!!mx * 4).
if (x < !!mx * 3 || y < !!my * 3 ||
- x + !!mx * 4 > w - bw || y + !!my * 4 > h - bh) {
+ x + !!mx * 4 > w - bw || y + !!my * 5 > h - bh) {
s->vdsp.emulated_edge_mc(s->edge_emu_buffer,
ref - !!my * 3 * ref_stride - !!mx * 3,
80,
ref = s->edge_emu_buffer + !!my * 3 * 80 + !!mx * 3;
ref_stride = 80;
}
- mc[!!mx][!!my](dst, ref, dst_stride, ref_stride, bh, mx << 1, my << 1);
+ mc[!!mx][!!my](dst, dst_stride, ref, ref_stride, bh, mx << 1, my << 1);
}
static av_always_inline void mc_chroma_dir(VP9Context *s, vp9_mc_func(*mc)[2],
ptrdiff_t src_stride_u,
const uint8_t *ref_v,
ptrdiff_t src_stride_v,
+ ThreadFrame *ref_frame,
ptrdiff_t y, ptrdiff_t x,
const VP56mv *mv,
int bw, int bh, int w, int h)
{
int mx = mv->x, my = mv->y;
+ int th;
y += my >> 4;
x += mx >> 4;
ref_v += y * src_stride_v + x;
mx &= 15;
my &= 15;
+
+ // we use +7 because the last 7 pixels of each sbrow can be changed in
+ // the longest loopfilter of the next sbrow
+ th = (y + bh + 4 * !!my + 7) >> 5;
+ ff_thread_await_progress(ref_frame, FFMAX(th, 0), 0);
+
// FIXME bilinear filter only needs 0/1 pixels, not 3/4
+ // The arm/aarch64 _hv filters read one more row than what actually is
+ // needed, so switch to emulated edge one pixel sooner vertically
+ // (!!my * 5) than horizontally (!!mx * 4).
if (x < !!mx * 3 || y < !!my * 3 ||
- x + !!mx * 4 > w - bw || y + !!my * 4 > h - bh) {
+ x + !!mx * 4 > w - bw || y + !!my * 5 > h - bh) {
s->vdsp.emulated_edge_mc(s->edge_emu_buffer,
ref_u - !!my * 3 * src_stride_u - !!mx * 3,
80,
bw + !!mx * 7, bh + !!my * 7,
x - !!mx * 3, y - !!my * 3, w, h);
ref_u = s->edge_emu_buffer + !!my * 3 * 80 + !!mx * 3;
- mc[!!mx][!!my](dst_u, ref_u, dst_stride, 80, bh, mx, my);
+ mc[!!mx][!!my](dst_u, dst_stride, ref_u, 80, bh, mx, my);
s->vdsp.emulated_edge_mc(s->edge_emu_buffer,
ref_v - !!my * 3 * src_stride_v - !!mx * 3,
bw + !!mx * 7, bh + !!my * 7,
x - !!mx * 3, y - !!my * 3, w, h);
ref_v = s->edge_emu_buffer + !!my * 3 * 80 + !!mx * 3;
- mc[!!mx][!!my](dst_v, ref_v, dst_stride, 80, bh, mx, my);
+ mc[!!mx][!!my](dst_v, dst_stride, ref_v, 80, bh, mx, my);
} else {
- mc[!!mx][!!my](dst_u, ref_u, dst_stride, src_stride_u, bh, mx, my);
- mc[!!mx][!!my](dst_v, ref_v, dst_stride, src_stride_v, bh, mx, my);
+ mc[!!mx][!!my](dst_u, dst_stride, ref_u, src_stride_u, bh, mx, my);
+ mc[!!mx][!!my](dst_v, dst_stride, ref_v, src_stride_v, bh, mx, my);
}
}
if (b->bs > BS_8x8) {
if (b->bs == BS_8x4) {
mc_luma_dir(s, s->dsp.mc[3][b->filter][0], b->dst[0], ls_y,
- ref1->data[0], ref1->linesize[0],
+ ref1->data[0], ref1->linesize[0], tref1,
row << 3, col << 3, &b->mv[0][0], 8, 4, w, h);
mc_luma_dir(s, s->dsp.mc[3][b->filter][0],
b->dst[0] + 4 * ls_y, ls_y,
- ref1->data[0], ref1->linesize[0],
+ ref1->data[0], ref1->linesize[0], tref1,
(row << 3) + 4, col << 3, &b->mv[2][0], 8, 4, w, h);
if (b->comp) {
mc_luma_dir(s, s->dsp.mc[3][b->filter][1], b->dst[0], ls_y,
- ref2->data[0], ref2->linesize[0],
+ ref2->data[0], ref2->linesize[0], tref2,
row << 3, col << 3, &b->mv[0][1], 8, 4, w, h);
mc_luma_dir(s, s->dsp.mc[3][b->filter][1],
b->dst[0] + 4 * ls_y, ls_y,
- ref2->data[0], ref2->linesize[0],
+ ref2->data[0], ref2->linesize[0], tref2,
(row << 3) + 4, col << 3, &b->mv[2][1], 8, 4, w, h);
}
} else if (b->bs == BS_4x8) {
mc_luma_dir(s, s->dsp.mc[4][b->filter][0], b->dst[0], ls_y,
- ref1->data[0], ref1->linesize[0],
+ ref1->data[0], ref1->linesize[0], tref1,
row << 3, col << 3, &b->mv[0][0], 4, 8, w, h);
mc_luma_dir(s, s->dsp.mc[4][b->filter][0], b->dst[0] + 4, ls_y,
- ref1->data[0], ref1->linesize[0],
+ ref1->data[0], ref1->linesize[0], tref1,
row << 3, (col << 3) + 4, &b->mv[1][0], 4, 8, w, h);
if (b->comp) {
mc_luma_dir(s, s->dsp.mc[4][b->filter][1], b->dst[0], ls_y,
- ref2->data[0], ref2->linesize[0],
+ ref2->data[0], ref2->linesize[0], tref2,
row << 3, col << 3, &b->mv[0][1], 4, 8, w, h);
mc_luma_dir(s, s->dsp.mc[4][b->filter][1], b->dst[0] + 4, ls_y,
- ref2->data[0], ref2->linesize[0],
+ ref2->data[0], ref2->linesize[0], tref2,
row << 3, (col << 3) + 4, &b->mv[1][1], 4, 8, w, h);
}
} else {
// FIXME if two horizontally adjacent blocks have the same MV,
// do a w8 instead of a w4 call
mc_luma_dir(s, s->dsp.mc[4][b->filter][0], b->dst[0], ls_y,
- ref1->data[0], ref1->linesize[0],
+ ref1->data[0], ref1->linesize[0], tref1,
row << 3, col << 3, &b->mv[0][0], 4, 4, w, h);
mc_luma_dir(s, s->dsp.mc[4][b->filter][0], b->dst[0] + 4, ls_y,
- ref1->data[0], ref1->linesize[0],
+ ref1->data[0], ref1->linesize[0], tref1,
row << 3, (col << 3) + 4, &b->mv[1][0], 4, 4, w, h);
mc_luma_dir(s, s->dsp.mc[4][b->filter][0],
b->dst[0] + 4 * ls_y, ls_y,
- ref1->data[0], ref1->linesize[0],
+ ref1->data[0], ref1->linesize[0], tref1,
(row << 3) + 4, col << 3, &b->mv[2][0], 4, 4, w, h);
mc_luma_dir(s, s->dsp.mc[4][b->filter][0],
b->dst[0] + 4 * ls_y + 4, ls_y,
- ref1->data[0], ref1->linesize[0],
+ ref1->data[0], ref1->linesize[0], tref1,
(row << 3) + 4, (col << 3) + 4, &b->mv[3][0], 4, 4, w, h);
if (b->comp) {
mc_luma_dir(s, s->dsp.mc[4][b->filter][1], b->dst[0], ls_y,
- ref2->data[0], ref2->linesize[0],
+ ref2->data[0], ref2->linesize[0], tref2,
row << 3, col << 3, &b->mv[0][1], 4, 4, w, h);
mc_luma_dir(s, s->dsp.mc[4][b->filter][1], b->dst[0] + 4, ls_y,
- ref2->data[0], ref2->linesize[0],
+ ref2->data[0], ref2->linesize[0], tref2,
row << 3, (col << 3) + 4, &b->mv[1][1], 4, 4, w, h);
mc_luma_dir(s, s->dsp.mc[4][b->filter][1],
b->dst[0] + 4 * ls_y, ls_y,
- ref2->data[0], ref2->linesize[0],
+ ref2->data[0], ref2->linesize[0], tref2,
(row << 3) + 4, col << 3, &b->mv[2][1], 4, 4, w, h);
mc_luma_dir(s, s->dsp.mc[4][b->filter][1],
b->dst[0] + 4 * ls_y + 4, ls_y,
- ref2->data[0], ref2->linesize[0],
+ ref2->data[0], ref2->linesize[0], tref2,
(row << 3) + 4, (col << 3) + 4, &b->mv[3][1], 4, 4, w, h);
}
}
int bh = bwh_tab[0][b->bs][1] * 4;
mc_luma_dir(s, s->dsp.mc[bwl][b->filter][0], b->dst[0], ls_y,
- ref1->data[0], ref1->linesize[0],
+ ref1->data[0], ref1->linesize[0], tref1,
row << 3, col << 3, &b->mv[0][0], bw, bh, w, h);
if (b->comp)
mc_luma_dir(s, s->dsp.mc[bwl][b->filter][1], b->dst[0], ls_y,
- ref2->data[0], ref2->linesize[0],
+ ref2->data[0], ref2->linesize[0], tref2,
row << 3, col << 3, &b->mv[0][1], bw, bh, w, h);
}
mc_chroma_dir(s, s->dsp.mc[bwl][b->filter][0],
b->dst[1], b->dst[2], ls_uv,
ref1->data[1], ref1->linesize[1],
- ref1->data[2], ref1->linesize[2],
+ ref1->data[2], ref1->linesize[2], tref1,
row << 2, col << 2, &mvuv, bw, bh, w, h);
if (b->comp) {
mc_chroma_dir(s, s->dsp.mc[bwl][b->filter][1],
b->dst[1], b->dst[2], ls_uv,
ref2->data[1], ref2->linesize[1],
- ref2->data[2], ref2->linesize[2],
+ ref2->data[2], ref2->linesize[2], tref2,
row << 2, col << 2, &mvuv, bw, bh, w, h);
}
}
s->max_mv.x = 128 + (s->cols - col - w4) * 64;
s->max_mv.y = 128 + (s->rows - row - h4) * 64;
- b->bs = bs;
- decode_mode(s, b);
- b->uvtx = b->tx - (w4 * 2 == (1 << b->tx) || h4 * 2 == (1 << b->tx));
+ if (s->pass < 2) {
+ b->bs = bs;
+ b->bl = bl;
+ b->bp = bp;
+ decode_mode(s, b);
+ b->uvtx = b->tx - (w4 * 2 == (1 << b->tx) || h4 * 2 == (1 << b->tx));
- if (!b->skip) {
- if ((ret = decode_coeffs(avctx)) < 0)
- return ret;
- } else {
- int pl;
+ if (!b->skip) {
+ if ((ret = decode_coeffs(avctx)) < 0)
+ return ret;
+ } else {
+ int pl;
- memset(&s->above_y_nnz_ctx[col * 2], 0, w4 * 2);
- memset(&s->left_y_nnz_ctx[(row & 7) << 1], 0, h4 * 2);
- for (pl = 0; pl < 2; pl++) {
- memset(&s->above_uv_nnz_ctx[pl][col], 0, w4);
- memset(&s->left_uv_nnz_ctx[pl][row & 7], 0, h4);
+ memset(&s->above_y_nnz_ctx[col * 2], 0, w4 * 2);
+ memset(&s->left_y_nnz_ctx[(row & 7) << 1], 0, h4 * 2);
+ for (pl = 0; pl < 2; pl++) {
+ memset(&s->above_uv_nnz_ctx[pl][col], 0, w4);
+ memset(&s->left_uv_nnz_ctx[pl][row & 7], 0, h4);
+ }
+ }
+
+ if (s->pass == 1) {
+ s->b++;
+ s->block += w4 * h4 * 64;
+ s->uvblock[0] += w4 * h4 * 16;
+ s->uvblock[1] += w4 * h4 * 16;
+ s->eob += w4 * h4 * 4;
+ s->uveob[0] += w4 * h4;
+ s->uveob[1] += w4 * h4;
+
+ return 0;
}
}
av_assert2(n <= 4);
if (w & bw) {
s->dsp.mc[n][0][0][0][0](f->data[0] + yoff + o,
- s->tmp_y + o,
f->linesize[0],
+ s->tmp_y + o,
64, h, 0, 0);
o += bw;
}
av_assert2(n <= 4);
if (w & bw) {
s->dsp.mc[n][0][0][0][0](f->data[1] + uvoff + o,
- s->tmp_uv[0] + o,
f->linesize[1],
+ s->tmp_uv[0] + o,
32, h, 0, 0);
s->dsp.mc[n][0][0][0][0](f->data[2] + uvoff + o,
- s->tmp_uv[1] + o,
f->linesize[2],
+ s->tmp_uv[1] + o,
32, h, 0, 0);
o += bw;
}
}
}
+ if (s->pass == 2) {
+ s->b++;
+ s->block += w4 * h4 * 64;
+ s->uvblock[0] += w4 * h4 * 16;
+ s->uvblock[1] += w4 * h4 * 16;
+ s->eob += w4 * h4 * 4;
+ s->uveob[0] += w4 * h4;
+ s->uveob[1] += w4 * h4;
+ }
+
return 0;
}