/src/ffmpeg/libavcodec/vp9_mc_template.c

Line

Count

Source

/*

 * VP9 compatible video decoder

 * Copyright (C) 2013 Ronald S. Bultje <rsbultje gmail com>

 * Copyright (C) 2013 Clément Bœsch <u pkh me>

 * This file is part of FFmpeg.

 * FFmpeg is free software; you can redistribute it and/or

 * modify it under the terms of the GNU Lesser General Public

 * License as published by the Free Software Foundation; either

 * version 2.1 of the License, or (at your option) any later version.

 * FFmpeg is distributed in the hope that it will be useful,

 * but WITHOUT ANY WARRANTY; without even the implied warranty of

 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU

 * Lesser General Public License for more details.

 * You should have received a copy of the GNU Lesser General Public

 * License along with FFmpeg; if not, write to the Free Software

 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA

*/

#define ROUNDED_DIV_MVx2(a, b) \

4.41M

    (VP9mv) { .x = ROUNDED_DIV(a.x + b.x, 2), .y = ROUNDED_DIV(a.y + b.y, 2) }

#define ROUNDED_DIV_MVx4(a, b, c, d) \

509k

    (VP9mv) { .x = ROUNDED_DIV(a.x + b.x + c.x + d.x, 4), \

509k

               .y = ROUNDED_DIV(a.y + b.y + c.y + d.y, 4) }

static void FN(inter_pred)(VP9TileData *td)

13.5M

13.5M

    static const uint8_t bwlog_tab[2][N_BS_SIZES] = {

13.5M

        { 0, 0, 1, 1, 1, 2, 2, 2, 3, 3, 3, 4, 4 },

13.5M

        { 1, 1, 2, 2, 2, 3, 3, 3, 4, 4, 4, 4, 4 },

13.5M

};

13.5M

    const VP9Context *s = td->s;

13.5M

    VP9Block *b = td->b;

13.5M

    int row = td->row, col = td->col;

13.5M

    const ProgressFrame *tref1 = &s->s.refs[s->s.h.refidx[b->ref[0]]], *tref2;

13.5M

    const AVFrame *ref1 = tref1->f, *ref2;

13.5M

    int w1 = ref1->width, h1 = ref1->height, w2, h2;

13.5M

    ptrdiff_t ls_y = td->y_stride, ls_uv = td->uv_stride;

13.5M

    int bytesperpixel = BYTES_PER_PIXEL;

13.5M

    if (b->comp) {

8.86M

        tref2 = &s->s.refs[s->s.h.refidx[b->ref[1]]];

8.86M

        ref2 = tref2->f;

8.86M

        w2 = ref2->width;

8.86M

        h2 = ref2->height;

8.86M

    // y inter pred

13.5M

    if (b->bs > BS_8x8) {

4.29M

        VP9mv uvmv;

#if SCALED == 0

2.92M

        if (b->bs == BS_8x4) {

1.08M

            mc_luma_dir(td, mc[3][b->filter][0], td->dst[0], ls_y,

1.08M

                        ref1->data[0], ref1->linesize[0], tref1,

1.08M

                        row << 3, col << 3, &b->mv[0][0],,,,, 8, 4, w1, h1, 0);

1.08M

            mc_luma_dir(td, mc[3][b->filter][0],

1.08M

                        td->dst[0] + 4 * ls_y, ls_y,

1.08M

                        ref1->data[0], ref1->linesize[0], tref1,

1.08M

                        (row << 3) + 4, col << 3, &b->mv[2][0],,,,, 8, 4, w1, h1, 0);

1.08M

            w1 = (w1 + s->ss_h) >> s->ss_h;

1.08M

            if (s->ss_v) {

355k

                h1 = (h1 + 1) >> 1;

355k

                uvmv = ROUNDED_DIV_MVx2(b->mv[0][0], b->mv[2][0]);

355k

                mc_chroma_dir(td, mc[3 + s->ss_h][b->filter][0],

355k

                              td->dst[1], td->dst[2], ls_uv,

355k

                              ref1->data[1], ref1->linesize[1],

355k

                              ref1->data[2], ref1->linesize[2], tref1,

355k

                              row << 2, col << (3 - s->ss_h),

355k

                              &uvmv,,,,, 8 >> s->ss_h, 4, w1, h1, 0);

725k

            } else {

725k

                mc_chroma_dir(td, mc[3 + s->ss_h][b->filter][0],

725k

                              td->dst[1], td->dst[2], ls_uv,

725k

                              ref1->data[1], ref1->linesize[1],

725k

                              ref1->data[2], ref1->linesize[2], tref1,

725k

                              row << 3, col << (3 - s->ss_h),

725k

                              &b->mv[0][0],,,,, 8 >> s->ss_h, 4, w1, h1, 0);

                // BUG for 4:2:2 bs=8x4, libvpx uses the wrong block index

                // to get the motion vector for the bottom 4x4 block

                // https://code.google.com/p/webm/issues/detail?id=993

725k

                if (s->ss_h == 0) {

554k

                    uvmv = b->mv[2][0];

554k

                } else {

171k

                    uvmv = ROUNDED_DIV_MVx2(b->mv[0][0], b->mv[2][0]);

171k

725k

                mc_chroma_dir(td, mc[3 + s->ss_h][b->filter][0],

725k

                              td->dst[1] + 4 * ls_uv, td->dst[2] + 4 * ls_uv, ls_uv,

725k

                              ref1->data[1], ref1->linesize[1],

725k

                              ref1->data[2], ref1->linesize[2], tref1,

725k

                              (row << 3) + 4, col << (3 - s->ss_h),

725k

                              &uvmv,,,,, 8 >> s->ss_h, 4, w1, h1, 0);

725k

1.08M

            if (b->comp) {

657k

                mc_luma_dir(td, mc[3][b->filter][1], td->dst[0], ls_y,

657k

                            ref2->data[0], ref2->linesize[0], tref2,

657k

                            row << 3, col << 3, &b->mv[0][1],,,,, 8, 4, w2, h2, 1);

657k

                mc_luma_dir(td, mc[3][b->filter][1],

657k

                            td->dst[0] + 4 * ls_y, ls_y,

657k

                            ref2->data[0], ref2->linesize[0], tref2,

657k

                            (row << 3) + 4, col << 3, &b->mv[2][1],,,,, 8, 4, w2, h2, 1);

657k

                w2 = (w2 + s->ss_h) >> s->ss_h;

657k

                if (s->ss_v) {

136k

                    h2 = (h2 + 1) >> 1;

136k

                    uvmv = ROUNDED_DIV_MVx2(b->mv[0][1], b->mv[2][1]);

136k

                    mc_chroma_dir(td, mc[3 + s->ss_h][b->filter][1],

136k

                                  td->dst[1], td->dst[2], ls_uv,

136k

                                  ref2->data[1], ref2->linesize[1],

136k

                                  ref2->data[2], ref2->linesize[2], tref2,

136k

                                  row << 2, col << (3 - s->ss_h),

136k

                                  &uvmv,,,,, 8 >> s->ss_h, 4, w2, h2, 1);

520k

                } else {

520k

                    mc_chroma_dir(td, mc[3 + s->ss_h][b->filter][1],

520k

                                  td->dst[1], td->dst[2], ls_uv,

520k

                                  ref2->data[1], ref2->linesize[1],

520k

                                  ref2->data[2], ref2->linesize[2], tref2,

520k

                                  row << 3, col << (3 - s->ss_h),

520k

                                  &b->mv[0][1],,,,, 8 >> s->ss_h, 4, w2, h2, 1);

                    // BUG for 4:2:2 bs=8x4, libvpx uses the wrong block index

                    // to get the motion vector for the bottom 4x4 block

                    // https://code.google.com/p/webm/issues/detail?id=993

520k

                    if (s->ss_h == 0) {

462k

                        uvmv = b->mv[2][1];

462k

                    } else {

57.7k

                        uvmv = ROUNDED_DIV_MVx2(b->mv[0][1], b->mv[2][1]);

57.7k

520k

                    mc_chroma_dir(td, mc[3 + s->ss_h][b->filter][1],

520k

                                  td->dst[1] + 4 * ls_uv, td->dst[2] + 4 * ls_uv, ls_uv,

520k

                                  ref2->data[1], ref2->linesize[1],

520k

                                  ref2->data[2], ref2->linesize[2], tref2,

520k

                                  (row << 3) + 4, col << (3 - s->ss_h),

520k

                                  &uvmv,,,,, 8 >> s->ss_h, 4, w2, h2, 1);

520k

657k

1.84M

        } else if (b->bs == BS_4x8) {

653k

            mc_luma_dir(td, mc[4][b->filter][0], td->dst[0], ls_y,

653k

                        ref1->data[0], ref1->linesize[0], tref1,

653k

                        row << 3, col << 3, &b->mv[0][0],,,,, 4, 8, w1, h1, 0);

653k

            mc_luma_dir(td, mc[4][b->filter][0], td->dst[0] + 4 * bytesperpixel, ls_y,

653k

                        ref1->data[0], ref1->linesize[0], tref1,

653k

                        row << 3, (col << 3) + 4, &b->mv[1][0],,,,, 4, 8, w1, h1, 0);

653k

            h1 = (h1 + s->ss_v) >> s->ss_v;

653k

            if (s->ss_h) {

203k

                w1 = (w1 + 1) >> 1;

203k

                uvmv = ROUNDED_DIV_MVx2(b->mv[0][0], b->mv[1][0]);

203k

                mc_chroma_dir(td, mc[4][b->filter][0],

203k

                              td->dst[1], td->dst[2], ls_uv,

203k

                              ref1->data[1], ref1->linesize[1],

203k

                              ref1->data[2], ref1->linesize[2], tref1,

203k

                              row << (3 - s->ss_v), col << 2,

203k

                              &uvmv,,,,, 4, 8 >> s->ss_v, w1, h1, 0);

449k

            } else {

449k

                mc_chroma_dir(td, mc[4][b->filter][0],

449k

                              td->dst[1], td->dst[2], ls_uv,

449k

                              ref1->data[1], ref1->linesize[1],

449k

                              ref1->data[2], ref1->linesize[2], tref1,

449k

                              row << (3 - s->ss_v), col << 3,

449k

                              &b->mv[0][0],,,,, 4, 8 >> s->ss_v, w1, h1, 0);

449k

                mc_chroma_dir(td, mc[4][b->filter][0],

449k

                              td->dst[1] + 4 * bytesperpixel,

449k

                              td->dst[2] + 4 * bytesperpixel, ls_uv,

449k

                              ref1->data[1], ref1->linesize[1],

449k

                              ref1->data[2], ref1->linesize[2], tref1,

449k

                              row << (3 - s->ss_v), (col << 3) + 4,

449k

                              &b->mv[1][0],,,,, 4, 8 >> s->ss_v, w1, h1, 0);

449k

653k

            if (b->comp) {

397k

                mc_luma_dir(td, mc[4][b->filter][1], td->dst[0], ls_y,

397k

                            ref2->data[0], ref2->linesize[0], tref2,

397k

                            row << 3, col << 3, &b->mv[0][1],,,,, 4, 8, w2, h2, 1);

397k

                mc_luma_dir(td, mc[4][b->filter][1], td->dst[0] + 4 * bytesperpixel, ls_y,

397k

                            ref2->data[0], ref2->linesize[0], tref2,

397k

                            row << 3, (col << 3) + 4, &b->mv[1][1],,,,, 4, 8, w2, h2, 1);

397k

                h2 = (h2 + s->ss_v) >> s->ss_v;

397k

                if (s->ss_h) {

87.9k

                    w2 = (w2 + 1) >> 1;

87.9k

                    uvmv = ROUNDED_DIV_MVx2(b->mv[0][1], b->mv[1][1]);

87.9k

                    mc_chroma_dir(td, mc[4][b->filter][1],

87.9k

                                  td->dst[1], td->dst[2], ls_uv,

87.9k

                                  ref2->data[1], ref2->linesize[1],

87.9k

                                  ref2->data[2], ref2->linesize[2], tref2,

87.9k

                                  row << (3 - s->ss_v), col << 2,

87.9k

                                  &uvmv,,,,, 4, 8 >> s->ss_v, w2, h2, 1);

309k

                } else {

309k

                    mc_chroma_dir(td, mc[4][b->filter][1],

309k

                                  td->dst[1], td->dst[2], ls_uv,

309k

                                  ref2->data[1], ref2->linesize[1],

309k

                                  ref2->data[2], ref2->linesize[2], tref2,

309k

                                  row << (3 - s->ss_v), col << 3,

309k

                                  &b->mv[0][1],,,,, 4, 8 >> s->ss_v, w2, h2, 1);

309k

                    mc_chroma_dir(td, mc[4][b->filter][1],

309k

                                  td->dst[1] + 4 * bytesperpixel,

309k

                                  td->dst[2] + 4 * bytesperpixel, ls_uv,

309k

                                  ref2->data[1], ref2->linesize[1],

309k

                                  ref2->data[2], ref2->linesize[2], tref2,

309k

                                  row << (3 - s->ss_v), (col << 3) + 4,

309k

                                  &b->mv[1][1],,,,, 4, 8 >> s->ss_v, w2, h2, 1);

309k

397k

653k

        } else

1.18M

#endif

1.18M

#if SCALED == 0

1.18M

            av_assert2(b->bs == BS_4x4);

#endif

            // FIXME if two horizontally adjacent blocks have the same MV,

            // do a w8 instead of a w4 call

2.56M

            mc_luma_dir(td, mc[4][b->filter][0], td->dst[0], ls_y,

1.18M

                        ref1->data[0], ref1->linesize[0], tref1,

1.18M

                        row << 3, col << 3, &b->mv[0][0],

1.18M

                        0, 0, 8, 8, 4, 4, w1, h1, 0);

2.56M

            mc_luma_dir(td, mc[4][b->filter][0], td->dst[0] + 4 * bytesperpixel, ls_y,

1.18M

                        ref1->data[0], ref1->linesize[0], tref1,

1.18M

                        row << 3, (col << 3) + 4, &b->mv[1][0],

1.18M

                        4, 0, 8, 8, 4, 4, w1, h1, 0);

2.56M

            mc_luma_dir(td, mc[4][b->filter][0],

1.18M

                        td->dst[0] + 4 * ls_y, ls_y,

1.18M

                        ref1->data[0], ref1->linesize[0], tref1,

1.18M

                        (row << 3) + 4, col << 3, &b->mv[2][0],

1.18M

                        0, 4, 8, 8, 4, 4, w1, h1, 0);

2.56M

            mc_luma_dir(td, mc[4][b->filter][0],

1.18M

                        td->dst[0] + 4 * ls_y + 4 * bytesperpixel, ls_y,

1.18M

                        ref1->data[0], ref1->linesize[0], tref1,

1.18M

                        (row << 3) + 4, (col << 3) + 4, &b->mv[3][0],

1.18M

                        4, 4, 8, 8, 4, 4, w1, h1, 0);

2.56M

            if (s->ss_v) {

1.03M

                h1 = (h1 + 1) >> 1;

1.03M

                if (s->ss_h) {

291k

                    w1 = (w1 + 1) >> 1;

291k

                    uvmv = ROUNDED_DIV_MVx4(b->mv[0][0], b->mv[1][0],

291k

                                            b->mv[2][0], b->mv[3][0]);

291k

                    mc_chroma_dir(td, mc[4][b->filter][0],

291k

                                  td->dst[1], td->dst[2], ls_uv,

291k

                                  ref1->data[1], ref1->linesize[1],

291k

                                  ref1->data[2], ref1->linesize[2], tref1,

291k

                                  row << 2, col << 2,

291k

                                  &uvmv, 0, 0, 4, 4, 4, 4, w1, h1, 0);

747k

                } else {

747k

                    uvmv = ROUNDED_DIV_MVx2(b->mv[0][0], b->mv[2][0]);

747k

                    mc_chroma_dir(td, mc[4][b->filter][0],

747k

                                  td->dst[1], td->dst[2], ls_uv,

747k

                                  ref1->data[1], ref1->linesize[1],

747k

                                  ref1->data[2], ref1->linesize[2], tref1,

747k

                                  row << 2, col << 3,

747k

                                  &uvmv, 0, 0, 8, 4, 4, 4, w1, h1, 0);

747k

                    uvmv = ROUNDED_DIV_MVx2(b->mv[1][0], b->mv[3][0]);

747k

                    mc_chroma_dir(td, mc[4][b->filter][0],

747k

                                  td->dst[1] + 4 * bytesperpixel,

747k

                                  td->dst[2] + 4 * bytesperpixel, ls_uv,

747k

                                  ref1->data[1], ref1->linesize[1],

747k

                                  ref1->data[2], ref1->linesize[2], tref1,

747k

                                  row << 2, (col << 3) + 4,

747k

                                  &uvmv, 4, 0, 8, 4, 4, 4, w1, h1, 0);

747k

1.52M

            } else {

1.52M

                if (s->ss_h) {

280k

                    w1 = (w1 + 1) >> 1;

280k

                    uvmv = ROUNDED_DIV_MVx2(b->mv[0][0], b->mv[1][0]);

280k

                    mc_chroma_dir(td, mc[4][b->filter][0],

280k

                                  td->dst[1], td->dst[2], ls_uv,

280k

                                  ref1->data[1], ref1->linesize[1],

280k

                                  ref1->data[2], ref1->linesize[2], tref1,

280k

                                  row << 3, col << 2,

280k

                                  &uvmv, 0, 0, 4, 8, 4, 4, w1, h1, 0);

                    // BUG libvpx uses wrong block index for 4:2:2 bs=4x4

                    // bottom block

                    // https://code.google.com/p/webm/issues/detail?id=993

280k

                    uvmv = ROUNDED_DIV_MVx2(b->mv[1][0], b->mv[2][0]);

280k

                    mc_chroma_dir(td, mc[4][b->filter][0],

280k

                                  td->dst[1] + 4 * ls_uv, td->dst[2] + 4 * ls_uv, ls_uv,

280k

                                  ref1->data[1], ref1->linesize[1],

280k

                                  ref1->data[2], ref1->linesize[2], tref1,

280k

                                  (row << 3) + 4, col << 2,

280k

                                  &uvmv, 0, 4, 4, 8, 4, 4, w1, h1, 0);

1.24M

                } else {

1.24M

                    mc_chroma_dir(td, mc[4][b->filter][0],

1.24M

                                  td->dst[1], td->dst[2], ls_uv,

1.24M

                                  ref1->data[1], ref1->linesize[1],

1.24M

                                  ref1->data[2], ref1->linesize[2], tref1,

1.24M

                                  row << 3, col << 3,

1.24M

                                  &b->mv[0][0], 0, 0, 8, 8, 4, 4, w1, h1, 0);

1.24M

                    mc_chroma_dir(td, mc[4][b->filter][0],

1.24M

                                  td->dst[1] + 4 * bytesperpixel,

1.24M

                                  td->dst[2] + 4 * bytesperpixel, ls_uv,

1.24M

                                  ref1->data[1], ref1->linesize[1],

1.24M

                                  ref1->data[2], ref1->linesize[2], tref1,

1.24M

                                  row << 3, (col << 3) + 4,

1.24M

                                  &b->mv[1][0], 4, 0, 8, 8, 4, 4, w1, h1, 0);

1.24M

                    mc_chroma_dir(td, mc[4][b->filter][0],

1.24M

                                  td->dst[1] + 4 * ls_uv, td->dst[2] + 4 * ls_uv, ls_uv,

1.24M

                                  ref1->data[1], ref1->linesize[1],

1.24M

                                  ref1->data[2], ref1->linesize[2], tref1,

1.24M

                                  (row << 3) + 4, col << 3,

1.24M

                                  &b->mv[2][0], 0, 4, 8, 8, 4, 4, w1, h1, 0);

1.24M

                    mc_chroma_dir(td, mc[4][b->filter][0],

1.24M

                                  td->dst[1] + 4 * ls_uv + 4 * bytesperpixel,

1.24M

                                  td->dst[2] + 4 * ls_uv + 4 * bytesperpixel, ls_uv,

1.24M

                                  ref1->data[1], ref1->linesize[1],

1.24M

                                  ref1->data[2], ref1->linesize[2], tref1,

1.24M

                                  (row << 3) + 4, (col << 3) + 4,

1.24M

                                  &b->mv[3][0], 4, 4, 8, 8, 4, 4, w1, h1, 0);

1.24M

1.52M

2.56M

            if (b->comp) {

1.95M

                mc_luma_dir(td, mc[4][b->filter][1], td->dst[0], ls_y,

1.95M

                            ref2->data[0], ref2->linesize[0], tref2,

1.95M

                            row << 3, col << 3, &b->mv[0][1], 0, 0, 8, 8, 4, 4, w2, h2, 1);

1.95M

                mc_luma_dir(td, mc[4][b->filter][1], td->dst[0] + 4 * bytesperpixel, ls_y,

1.95M

                            ref2->data[0], ref2->linesize[0], tref2,

1.95M

                            row << 3, (col << 3) + 4, &b->mv[1][1], 4, 0, 8, 8, 4, 4, w2, h2, 1);

1.95M

                mc_luma_dir(td, mc[4][b->filter][1],

1.95M

                            td->dst[0] + 4 * ls_y, ls_y,

1.95M

                            ref2->data[0], ref2->linesize[0], tref2,

1.95M

                            (row << 3) + 4, col << 3, &b->mv[2][1], 0, 4, 8, 8, 4, 4, w2, h2, 1);

1.95M

                mc_luma_dir(td, mc[4][b->filter][1],

1.95M

                            td->dst[0] + 4 * ls_y + 4 * bytesperpixel, ls_y,

1.95M

                            ref2->data[0], ref2->linesize[0], tref2,

1.95M

                            (row << 3) + 4, (col << 3) + 4, &b->mv[3][1], 4, 4, 8, 8, 4, 4, w2, h2, 1);

1.95M

                if (s->ss_v) {

757k

                    h2 = (h2 + 1) >> 1;

757k

                    if (s->ss_h) {

217k

                        w2 = (w2 + 1) >> 1;

217k

                        uvmv = ROUNDED_DIV_MVx4(b->mv[0][1], b->mv[1][1],

217k

                                                b->mv[2][1], b->mv[3][1]);

217k

                        mc_chroma_dir(td, mc[4][b->filter][1],

217k

                                      td->dst[1], td->dst[2], ls_uv,

217k

                                      ref2->data[1], ref2->linesize[1],

217k

                                      ref2->data[2], ref2->linesize[2], tref2,

217k

                                      row << 2, col << 2,

217k

                                      &uvmv, 0, 0, 4, 4, 4, 4, w2, h2, 1);

539k

                    } else {

539k

                        uvmv = ROUNDED_DIV_MVx2(b->mv[0][1], b->mv[2][1]);

539k

                        mc_chroma_dir(td, mc[4][b->filter][1],

539k

                                      td->dst[1], td->dst[2], ls_uv,

539k

                                      ref2->data[1], ref2->linesize[1],

539k

                                      ref2->data[2], ref2->linesize[2], tref2,

539k

                                      row << 2, col << 3,

539k

                                      &uvmv, 0, 0, 8, 4, 4, 4, w2, h2, 1);

539k

                        uvmv = ROUNDED_DIV_MVx2(b->mv[1][1], b->mv[3][1]);

539k

                        mc_chroma_dir(td, mc[4][b->filter][1],

539k

                                      td->dst[1] + 4 * bytesperpixel,

539k

                                      td->dst[2] + 4 * bytesperpixel, ls_uv,

539k

                                      ref2->data[1], ref2->linesize[1],

539k

                                      ref2->data[2], ref2->linesize[2], tref2,

539k

                                      row << 2, (col << 3) + 4,

539k

                                      &uvmv, 4, 0, 8, 4, 4, 4, w2, h2, 1);

539k

1.19M

                } else {

1.19M

                    if (s->ss_h) {

135k

                        w2 = (w2 + 1) >> 1;

135k

                        uvmv = ROUNDED_DIV_MVx2(b->mv[0][1], b->mv[1][1]);

135k

                        mc_chroma_dir(td, mc[4][b->filter][1],

135k

                                      td->dst[1], td->dst[2], ls_uv,

135k

                                      ref2->data[1], ref2->linesize[1],

135k

                                      ref2->data[2], ref2->linesize[2], tref2,

135k

                                      row << 3, col << 2,

135k

                                      &uvmv, 0, 0, 4, 8, 4, 4, w2, h2, 1);

                        // BUG libvpx uses wrong block index for 4:2:2 bs=4x4

                        // bottom block

                        // https://code.google.com/p/webm/issues/detail?id=993

135k

                        uvmv = ROUNDED_DIV_MVx2(b->mv[1][1], b->mv[2][1]);

135k

                        mc_chroma_dir(td, mc[4][b->filter][1],

135k

                                      td->dst[1] + 4 * ls_uv, td->dst[2] + 4 * ls_uv, ls_uv,

135k

                                      ref2->data[1], ref2->linesize[1],

135k

                                      ref2->data[2], ref2->linesize[2], tref2,

135k

                                      (row << 3) + 4, col << 2,

135k

                                      &uvmv, 0, 4, 4, 8, 4, 4, w2, h2, 1);

1.05M

                    } else {

1.05M

                        mc_chroma_dir(td, mc[4][b->filter][1],

1.05M

                                      td->dst[1], td->dst[2], ls_uv,

1.05M

                                      ref2->data[1], ref2->linesize[1],

1.05M

                                      ref2->data[2], ref2->linesize[2], tref2,

1.05M

                                      row << 3, col << 3,

1.05M

                                      &b->mv[0][1], 0, 0, 8, 8, 4, 4, w2, h2, 1);

1.05M

                        mc_chroma_dir(td, mc[4][b->filter][1],

1.05M

                                      td->dst[1] + 4 * bytesperpixel,

1.05M

                                      td->dst[2] + 4 * bytesperpixel, ls_uv,

1.05M

                                      ref2->data[1], ref2->linesize[1],

1.05M

                                      ref2->data[2], ref2->linesize[2], tref2,

1.05M

                                      row << 3, (col << 3) + 4,

1.05M

                                      &b->mv[1][1], 4, 0, 8, 8, 4, 4, w2, h2, 1);

1.05M

                        mc_chroma_dir(td, mc[4][b->filter][1],

1.05M

                                      td->dst[1] + 4 * ls_uv, td->dst[2] + 4 * ls_uv, ls_uv,

1.05M

                                      ref2->data[1], ref2->linesize[1],

1.05M

                                      ref2->data[2], ref2->linesize[2], tref2,

1.05M

                                      (row << 3) + 4, col << 3,

1.05M

                                      &b->mv[2][1], 0, 4, 8, 8, 4, 4, w2, h2, 1);

1.05M

                        mc_chroma_dir(td, mc[4][b->filter][1],

1.05M

                                      td->dst[1] + 4 * ls_uv + 4 * bytesperpixel,

1.05M

                                      td->dst[2] + 4 * ls_uv + 4 * bytesperpixel, ls_uv,

1.05M

                                      ref2->data[1], ref2->linesize[1],

1.05M

                                      ref2->data[2], ref2->linesize[2], tref2,

1.05M

                                      (row << 3) + 4, (col << 3) + 4,

1.05M

                                      &b->mv[3][1], 4, 4, 8, 8, 4, 4, w2, h2, 1);

1.05M

1.19M

1.95M

1.18M

9.25M

    } else {

9.25M

        int bwl = bwlog_tab[0][b->bs];

9.25M

        int bw = ff_vp9_bwh_tab[0][b->bs][0] * 4;

9.25M

        int bh = ff_vp9_bwh_tab[0][b->bs][1] * 4;

9.25M

        int uvbw = ff_vp9_bwh_tab[s->ss_h][b->bs][0] * 4;

9.25M

        int uvbh = ff_vp9_bwh_tab[s->ss_v][b->bs][1] * 4;

9.25M

        mc_luma_dir(td, mc[bwl][b->filter][0], td->dst[0], ls_y,

9.25M

                    ref1->data[0], ref1->linesize[0], tref1,

9.25M

                    row << 3, col << 3, &b->mv[0][0], 0, 0, bw, bh, bw, bh, w1, h1, 0);

9.25M

        w1 = (w1 + s->ss_h) >> s->ss_h;

9.25M

        h1 = (h1 + s->ss_v) >> s->ss_v;

9.25M

        mc_chroma_dir(td, mc[bwl + s->ss_h][b->filter][0],

9.25M

                      td->dst[1], td->dst[2], ls_uv,

9.25M

                      ref1->data[1], ref1->linesize[1],

9.25M

                      ref1->data[2], ref1->linesize[2], tref1,

9.25M

                      row << (3 - s->ss_v), col << (3 - s->ss_h),

9.25M

                      &b->mv[0][0], 0, 0, uvbw, uvbh, uvbw, uvbh, w1, h1, 0);

9.25M

        if (b->comp) {

5.85M

            mc_luma_dir(td, mc[bwl][b->filter][1], td->dst[0], ls_y,

5.85M

                        ref2->data[0], ref2->linesize[0], tref2,

5.85M

                        row << 3, col << 3, &b->mv[0][1], 0, 0, bw, bh, bw, bh, w2, h2, 1);

5.85M

            w2 = (w2 + s->ss_h) >> s->ss_h;

5.85M

            h2 = (h2 + s->ss_v) >> s->ss_v;

5.85M

            mc_chroma_dir(td, mc[bwl + s->ss_h][b->filter][1],

5.85M

                          td->dst[1], td->dst[2], ls_uv,

5.85M

                          ref2->data[1], ref2->linesize[1],

5.85M

                          ref2->data[2], ref2->linesize[2], tref2,

5.85M

                          row << (3 - s->ss_v), col << (3 - s->ss_h),

5.85M

                          &b->mv[0][1], 0, 0, uvbw, uvbh, uvbw, uvbh, w2, h2, 1);

5.85M

9.25M

13.5M

vp9recon.c:inter_pred_scaled_8bpp

Line	Count	Source
31	1.94M	{
32	1.94M	static const uint8_t bwlog_tab[2][N_BS_SIZES] = {
33	1.94M	{ 0, 0, 1, 1, 1, 2, 2, 2, 3, 3, 3, 4, 4 },
34	1.94M	{ 1, 1, 2, 2, 2, 3, 3, 3, 4, 4, 4, 4, 4 },
35	1.94M	};
36	1.94M	const VP9Context *s = td->s;
37	1.94M	VP9Block *b = td->b;
38	1.94M	int row = td->row, col = td->col;
39	1.94M	const ProgressFrame tref1 = &s->s.refs[s->s.h.refidx[b->ref[0]]], tref2;
40	1.94M	const AVFrame ref1 = tref1->f, ref2;
41	1.94M	int w1 = ref1->width, h1 = ref1->height, w2, h2;
42	1.94M	ptrdiff_t ls_y = td->y_stride, ls_uv = td->uv_stride;
43	1.94M	int bytesperpixel = BYTES_PER_PIXEL;
44
45	1.94M	if (b->comp) {
46	1.38M	tref2 = &s->s.refs[s->s.h.refidx[b->ref[1]]];
47	1.38M	ref2 = tref2->f;
48	1.38M	w2 = ref2->width;
49	1.38M	h2 = ref2->height;
50	1.38M	}
51
52		// y inter pred
53	1.94M	if (b->bs > BS_8x8) {
54	531k	VP9mv uvmv;
55
56		#if SCALED == 0
57		if (b->bs == BS_8x4) {
58		mc_luma_dir(td, mc[3][b->filter][0], td->dst[0], ls_y,
59		ref1->data[0], ref1->linesize[0], tref1,
60		row << 3, col << 3, &b->mv[0][0],,,,, 8, 4, w1, h1, 0);
61		mc_luma_dir(td, mc[3][b->filter][0],
62		td->dst[0] + 4 * ls_y, ls_y,
63		ref1->data[0], ref1->linesize[0], tref1,
64		(row << 3) + 4, col << 3, &b->mv[2][0],,,,, 8, 4, w1, h1, 0);
65		w1 = (w1 + s->ss_h) >> s->ss_h;
66		if (s->ss_v) {
67		h1 = (h1 + 1) >> 1;
68		uvmv = ROUNDED_DIV_MVx2(b->mv[0][0], b->mv[2][0]);
69		mc_chroma_dir(td, mc[3 + s->ss_h][b->filter][0],
70		td->dst[1], td->dst[2], ls_uv,
71		ref1->data[1], ref1->linesize[1],
72		ref1->data[2], ref1->linesize[2], tref1,
73		row << 2, col << (3 - s->ss_h),
74		&uvmv,,,,, 8 >> s->ss_h, 4, w1, h1, 0);
75		} else {
76		mc_chroma_dir(td, mc[3 + s->ss_h][b->filter][0],
77		td->dst[1], td->dst[2], ls_uv,
78		ref1->data[1], ref1->linesize[1],
79		ref1->data[2], ref1->linesize[2], tref1,
80		row << 3, col << (3 - s->ss_h),
81		&b->mv[0][0],,,,, 8 >> s->ss_h, 4, w1, h1, 0);
82		// BUG for 4:2:2 bs=8x4, libvpx uses the wrong block index
83		// to get the motion vector for the bottom 4x4 block
84		// https://code.google.com/p/webm/issues/detail?id=993
85		if (s->ss_h == 0) {
86		uvmv = b->mv[2][0];
87		} else {
88		uvmv = ROUNDED_DIV_MVx2(b->mv[0][0], b->mv[2][0]);
89		}
90		mc_chroma_dir(td, mc[3 + s->ss_h][b->filter][0],
91		td->dst[1] + 4 * ls_uv, td->dst[2] + 4 * ls_uv, ls_uv,
92		ref1->data[1], ref1->linesize[1],
93		ref1->data[2], ref1->linesize[2], tref1,
94		(row << 3) + 4, col << (3 - s->ss_h),
95		&uvmv,,,,, 8 >> s->ss_h, 4, w1, h1, 0);
96		}
97
98		if (b->comp) {
99		mc_luma_dir(td, mc[3][b->filter][1], td->dst[0], ls_y,
100		ref2->data[0], ref2->linesize[0], tref2,
101		row << 3, col << 3, &b->mv[0][1],,,,, 8, 4, w2, h2, 1);
102		mc_luma_dir(td, mc[3][b->filter][1],
103		td->dst[0] + 4 * ls_y, ls_y,
104		ref2->data[0], ref2->linesize[0], tref2,
105		(row << 3) + 4, col << 3, &b->mv[2][1],,,,, 8, 4, w2, h2, 1);
106		w2 = (w2 + s->ss_h) >> s->ss_h;
107		if (s->ss_v) {
108		h2 = (h2 + 1) >> 1;
109		uvmv = ROUNDED_DIV_MVx2(b->mv[0][1], b->mv[2][1]);
110		mc_chroma_dir(td, mc[3 + s->ss_h][b->filter][1],
111		td->dst[1], td->dst[2], ls_uv,
112		ref2->data[1], ref2->linesize[1],
113		ref2->data[2], ref2->linesize[2], tref2,
114		row << 2, col << (3 - s->ss_h),
115		&uvmv,,,,, 8 >> s->ss_h, 4, w2, h2, 1);
116		} else {
117		mc_chroma_dir(td, mc[3 + s->ss_h][b->filter][1],
118		td->dst[1], td->dst[2], ls_uv,
119		ref2->data[1], ref2->linesize[1],
120		ref2->data[2], ref2->linesize[2], tref2,
121		row << 3, col << (3 - s->ss_h),
122		&b->mv[0][1],,,,, 8 >> s->ss_h, 4, w2, h2, 1);
123		// BUG for 4:2:2 bs=8x4, libvpx uses the wrong block index
124		// to get the motion vector for the bottom 4x4 block
125		// https://code.google.com/p/webm/issues/detail?id=993
126		if (s->ss_h == 0) {
127		uvmv = b->mv[2][1];
128		} else {
129		uvmv = ROUNDED_DIV_MVx2(b->mv[0][1], b->mv[2][1]);
130		}
131		mc_chroma_dir(td, mc[3 + s->ss_h][b->filter][1],
132		td->dst[1] + 4 * ls_uv, td->dst[2] + 4 * ls_uv, ls_uv,
133		ref2->data[1], ref2->linesize[1],
134		ref2->data[2], ref2->linesize[2], tref2,
135		(row << 3) + 4, col << (3 - s->ss_h),
136		&uvmv,,,,, 8 >> s->ss_h, 4, w2, h2, 1);
137		}
138		}
139		} else if (b->bs == BS_4x8) {
140		mc_luma_dir(td, mc[4][b->filter][0], td->dst[0], ls_y,
141		ref1->data[0], ref1->linesize[0], tref1,
142		row << 3, col << 3, &b->mv[0][0],,,,, 4, 8, w1, h1, 0);
143		mc_luma_dir(td, mc[4][b->filter][0], td->dst[0] + 4 * bytesperpixel, ls_y,
144		ref1->data[0], ref1->linesize[0], tref1,
145		row << 3, (col << 3) + 4, &b->mv[1][0],,,,, 4, 8, w1, h1, 0);
146		h1 = (h1 + s->ss_v) >> s->ss_v;
147		if (s->ss_h) {
148		w1 = (w1 + 1) >> 1;
149		uvmv = ROUNDED_DIV_MVx2(b->mv[0][0], b->mv[1][0]);
150		mc_chroma_dir(td, mc[4][b->filter][0],
151		td->dst[1], td->dst[2], ls_uv,
152		ref1->data[1], ref1->linesize[1],
153		ref1->data[2], ref1->linesize[2], tref1,
154		row << (3 - s->ss_v), col << 2,
155		&uvmv,,,,, 4, 8 >> s->ss_v, w1, h1, 0);
156		} else {
157		mc_chroma_dir(td, mc[4][b->filter][0],
158		td->dst[1], td->dst[2], ls_uv,
159		ref1->data[1], ref1->linesize[1],
160		ref1->data[2], ref1->linesize[2], tref1,
161		row << (3 - s->ss_v), col << 3,
162		&b->mv[0][0],,,,, 4, 8 >> s->ss_v, w1, h1, 0);
163		mc_chroma_dir(td, mc[4][b->filter][0],
164		td->dst[1] + 4 * bytesperpixel,
165		td->dst[2] + 4 * bytesperpixel, ls_uv,
166		ref1->data[1], ref1->linesize[1],
167		ref1->data[2], ref1->linesize[2], tref1,
168		row << (3 - s->ss_v), (col << 3) + 4,
169		&b->mv[1][0],,,,, 4, 8 >> s->ss_v, w1, h1, 0);
170		}
171
172		if (b->comp) {
173		mc_luma_dir(td, mc[4][b->filter][1], td->dst[0], ls_y,
174		ref2->data[0], ref2->linesize[0], tref2,
175		row << 3, col << 3, &b->mv[0][1],,,,, 4, 8, w2, h2, 1);
176		mc_luma_dir(td, mc[4][b->filter][1], td->dst[0] + 4 * bytesperpixel, ls_y,
177		ref2->data[0], ref2->linesize[0], tref2,
178		row << 3, (col << 3) + 4, &b->mv[1][1],,,,, 4, 8, w2, h2, 1);
179		h2 = (h2 + s->ss_v) >> s->ss_v;
180		if (s->ss_h) {
181		w2 = (w2 + 1) >> 1;
182		uvmv = ROUNDED_DIV_MVx2(b->mv[0][1], b->mv[1][1]);
183		mc_chroma_dir(td, mc[4][b->filter][1],
184		td->dst[1], td->dst[2], ls_uv,
185		ref2->data[1], ref2->linesize[1],
186		ref2->data[2], ref2->linesize[2], tref2,
187		row << (3 - s->ss_v), col << 2,
188		&uvmv,,,,, 4, 8 >> s->ss_v, w2, h2, 1);
189		} else {
190		mc_chroma_dir(td, mc[4][b->filter][1],
191		td->dst[1], td->dst[2], ls_uv,
192		ref2->data[1], ref2->linesize[1],
193		ref2->data[2], ref2->linesize[2], tref2,
194		row << (3 - s->ss_v), col << 3,
195		&b->mv[0][1],,,,, 4, 8 >> s->ss_v, w2, h2, 1);
196		mc_chroma_dir(td, mc[4][b->filter][1],
197		td->dst[1] + 4 * bytesperpixel,
198		td->dst[2] + 4 * bytesperpixel, ls_uv,
199		ref2->data[1], ref2->linesize[1],
200		ref2->data[2], ref2->linesize[2], tref2,
201		row << (3 - s->ss_v), (col << 3) + 4,
202		&b->mv[1][1],,,,, 4, 8 >> s->ss_v, w2, h2, 1);
203		}
204		}
205		} else
206		#endif
207	531k	{
208		#if SCALED == 0
209		av_assert2(b->bs == BS_4x4);
210		#endif
211
212		// FIXME if two horizontally adjacent blocks have the same MV,
213		// do a w8 instead of a w4 call
214	531k	mc_luma_dir(td, mc[4][b->filter][0], td->dst[0], ls_y,
215	531k	ref1->data[0], ref1->linesize[0], tref1,
216	531k	row << 3, col << 3, &b->mv[0][0],
217	531k	0, 0, 8, 8, 4, 4, w1, h1, 0);
218	531k	mc_luma_dir(td, mc[4][b->filter][0], td->dst[0] + 4 * bytesperpixel, ls_y,
219	531k	ref1->data[0], ref1->linesize[0], tref1,
220	531k	row << 3, (col << 3) + 4, &b->mv[1][0],
221	531k	4, 0, 8, 8, 4, 4, w1, h1, 0);
222	531k	mc_luma_dir(td, mc[4][b->filter][0],
223	531k	td->dst[0] + 4 * ls_y, ls_y,
224	531k	ref1->data[0], ref1->linesize[0], tref1,
225	531k	(row << 3) + 4, col << 3, &b->mv[2][0],
226	531k	0, 4, 8, 8, 4, 4, w1, h1, 0);
227	531k	mc_luma_dir(td, mc[4][b->filter][0],
228	531k	td->dst[0] + 4 * ls_y + 4 * bytesperpixel, ls_y,
229	531k	ref1->data[0], ref1->linesize[0], tref1,
230	531k	(row << 3) + 4, (col << 3) + 4, &b->mv[3][0],
231	531k	4, 4, 8, 8, 4, 4, w1, h1, 0);
232	531k	if (s->ss_v) {
233	377k	h1 = (h1 + 1) >> 1;
234	377k	if (s->ss_h) {
235	93.2k	w1 = (w1 + 1) >> 1;
236	93.2k	uvmv = ROUNDED_DIV_MVx4(b->mv[0][0], b->mv[1][0],
237	93.2k	b->mv[2][0], b->mv[3][0]);
238	93.2k	mc_chroma_dir(td, mc[4][b->filter][0],
239	93.2k	td->dst[1], td->dst[2], ls_uv,
240	93.2k	ref1->data[1], ref1->linesize[1],
241	93.2k	ref1->data[2], ref1->linesize[2], tref1,
242	93.2k	row << 2, col << 2,
243	93.2k	&uvmv, 0, 0, 4, 4, 4, 4, w1, h1, 0);
244	284k	} else {
245	284k	uvmv = ROUNDED_DIV_MVx2(b->mv[0][0], b->mv[2][0]);
246	284k	mc_chroma_dir(td, mc[4][b->filter][0],
247	284k	td->dst[1], td->dst[2], ls_uv,
248	284k	ref1->data[1], ref1->linesize[1],
249	284k	ref1->data[2], ref1->linesize[2], tref1,
250	284k	row << 2, col << 3,
251	284k	&uvmv, 0, 0, 8, 4, 4, 4, w1, h1, 0);
252	284k	uvmv = ROUNDED_DIV_MVx2(b->mv[1][0], b->mv[3][0]);
253	284k	mc_chroma_dir(td, mc[4][b->filter][0],
254	284k	td->dst[1] + 4 * bytesperpixel,
255	284k	td->dst[2] + 4 * bytesperpixel, ls_uv,
256	284k	ref1->data[1], ref1->linesize[1],
257	284k	ref1->data[2], ref1->linesize[2], tref1,
258	284k	row << 2, (col << 3) + 4,
259	284k	&uvmv, 4, 0, 8, 4, 4, 4, w1, h1, 0);
260	284k	}
261	377k	} else {
262	154k	if (s->ss_h) {
263	76.4k	w1 = (w1 + 1) >> 1;
264	76.4k	uvmv = ROUNDED_DIV_MVx2(b->mv[0][0], b->mv[1][0]);
265	76.4k	mc_chroma_dir(td, mc[4][b->filter][0],
266	76.4k	td->dst[1], td->dst[2], ls_uv,
267	76.4k	ref1->data[1], ref1->linesize[1],
268	76.4k	ref1->data[2], ref1->linesize[2], tref1,
269	76.4k	row << 3, col << 2,
270	76.4k	&uvmv, 0, 0, 4, 8, 4, 4, w1, h1, 0);
271		// BUG libvpx uses wrong block index for 4:2:2 bs=4x4
272		// bottom block
273		// https://code.google.com/p/webm/issues/detail?id=993
274	76.4k	uvmv = ROUNDED_DIV_MVx2(b->mv[1][0], b->mv[2][0]);
275	76.4k	mc_chroma_dir(td, mc[4][b->filter][0],
276	76.4k	td->dst[1] + 4 * ls_uv, td->dst[2] + 4 * ls_uv, ls_uv,
277	76.4k	ref1->data[1], ref1->linesize[1],
278	76.4k	ref1->data[2], ref1->linesize[2], tref1,
279	76.4k	(row << 3) + 4, col << 2,
280	76.4k	&uvmv, 0, 4, 4, 8, 4, 4, w1, h1, 0);
281	77.6k	} else {
282	77.6k	mc_chroma_dir(td, mc[4][b->filter][0],
283	77.6k	td->dst[1], td->dst[2], ls_uv,
284	77.6k	ref1->data[1], ref1->linesize[1],
285	77.6k	ref1->data[2], ref1->linesize[2], tref1,
286	77.6k	row << 3, col << 3,
287	77.6k	&b->mv[0][0], 0, 0, 8, 8, 4, 4, w1, h1, 0);
288	77.6k	mc_chroma_dir(td, mc[4][b->filter][0],
289	77.6k	td->dst[1] + 4 * bytesperpixel,
290	77.6k	td->dst[2] + 4 * bytesperpixel, ls_uv,
291	77.6k	ref1->data[1], ref1->linesize[1],
292	77.6k	ref1->data[2], ref1->linesize[2], tref1,
293	77.6k	row << 3, (col << 3) + 4,
294	77.6k	&b->mv[1][0], 4, 0, 8, 8, 4, 4, w1, h1, 0);
295	77.6k	mc_chroma_dir(td, mc[4][b->filter][0],
296	77.6k	td->dst[1] + 4 * ls_uv, td->dst[2] + 4 * ls_uv, ls_uv,
297	77.6k	ref1->data[1], ref1->linesize[1],
298	77.6k	ref1->data[2], ref1->linesize[2], tref1,
299	77.6k	(row << 3) + 4, col << 3,
300	77.6k	&b->mv[2][0], 0, 4, 8, 8, 4, 4, w1, h1, 0);
301	77.6k	mc_chroma_dir(td, mc[4][b->filter][0],
302	77.6k	td->dst[1] + 4 * ls_uv + 4 * bytesperpixel,
303	77.6k	td->dst[2] + 4 * ls_uv + 4 * bytesperpixel, ls_uv,
304	77.6k	ref1->data[1], ref1->linesize[1],
305	77.6k	ref1->data[2], ref1->linesize[2], tref1,
306	77.6k	(row << 3) + 4, (col << 3) + 4,
307	77.6k	&b->mv[3][0], 4, 4, 8, 8, 4, 4, w1, h1, 0);
308	77.6k	}
309	154k	}
310
311	531k	if (b->comp) {
312	402k	mc_luma_dir(td, mc[4][b->filter][1], td->dst[0], ls_y,
313	402k	ref2->data[0], ref2->linesize[0], tref2,
314	402k	row << 3, col << 3, &b->mv[0][1], 0, 0, 8, 8, 4, 4, w2, h2, 1);
315	402k	mc_luma_dir(td, mc[4][b->filter][1], td->dst[0] + 4 * bytesperpixel, ls_y,
316	402k	ref2->data[0], ref2->linesize[0], tref2,
317	402k	row << 3, (col << 3) + 4, &b->mv[1][1], 4, 0, 8, 8, 4, 4, w2, h2, 1);
318	402k	mc_luma_dir(td, mc[4][b->filter][1],
319	402k	td->dst[0] + 4 * ls_y, ls_y,
320	402k	ref2->data[0], ref2->linesize[0], tref2,
321	402k	(row << 3) + 4, col << 3, &b->mv[2][1], 0, 4, 8, 8, 4, 4, w2, h2, 1);
322	402k	mc_luma_dir(td, mc[4][b->filter][1],
323	402k	td->dst[0] + 4 * ls_y + 4 * bytesperpixel, ls_y,
324	402k	ref2->data[0], ref2->linesize[0], tref2,
325	402k	(row << 3) + 4, (col << 3) + 4, &b->mv[3][1], 4, 4, 8, 8, 4, 4, w2, h2, 1);
326	402k	if (s->ss_v) {
327	300k	h2 = (h2 + 1) >> 1;
328	300k	if (s->ss_h) {
329	65.2k	w2 = (w2 + 1) >> 1;
330	65.2k	uvmv = ROUNDED_DIV_MVx4(b->mv[0][1], b->mv[1][1],
331	65.2k	b->mv[2][1], b->mv[3][1]);
332	65.2k	mc_chroma_dir(td, mc[4][b->filter][1],
333	65.2k	td->dst[1], td->dst[2], ls_uv,
334	65.2k	ref2->data[1], ref2->linesize[1],
335	65.2k	ref2->data[2], ref2->linesize[2], tref2,
336	65.2k	row << 2, col << 2,
337	65.2k	&uvmv, 0, 0, 4, 4, 4, 4, w2, h2, 1);
338	235k	} else {
339	235k	uvmv = ROUNDED_DIV_MVx2(b->mv[0][1], b->mv[2][1]);
340	235k	mc_chroma_dir(td, mc[4][b->filter][1],
341	235k	td->dst[1], td->dst[2], ls_uv,
342	235k	ref2->data[1], ref2->linesize[1],
343	235k	ref2->data[2], ref2->linesize[2], tref2,
344	235k	row << 2, col << 3,
345	235k	&uvmv, 0, 0, 8, 4, 4, 4, w2, h2, 1);
346	235k	uvmv = ROUNDED_DIV_MVx2(b->mv[1][1], b->mv[3][1]);
347	235k	mc_chroma_dir(td, mc[4][b->filter][1],
348	235k	td->dst[1] + 4 * bytesperpixel,
349	235k	td->dst[2] + 4 * bytesperpixel, ls_uv,
350	235k	ref2->data[1], ref2->linesize[1],
351	235k	ref2->data[2], ref2->linesize[2], tref2,
352	235k	row << 2, (col << 3) + 4,
353	235k	&uvmv, 4, 0, 8, 4, 4, 4, w2, h2, 1);
354	235k	}
355	300k	} else {
356	101k	if (s->ss_h) {
357	42.6k	w2 = (w2 + 1) >> 1;
358	42.6k	uvmv = ROUNDED_DIV_MVx2(b->mv[0][1], b->mv[1][1]);
359	42.6k	mc_chroma_dir(td, mc[4][b->filter][1],
360	42.6k	td->dst[1], td->dst[2], ls_uv,
361	42.6k	ref2->data[1], ref2->linesize[1],
362	42.6k	ref2->data[2], ref2->linesize[2], tref2,
363	42.6k	row << 3, col << 2,
364	42.6k	&uvmv, 0, 0, 4, 8, 4, 4, w2, h2, 1);
365		// BUG libvpx uses wrong block index for 4:2:2 bs=4x4
366		// bottom block
367		// https://code.google.com/p/webm/issues/detail?id=993
368	42.6k	uvmv = ROUNDED_DIV_MVx2(b->mv[1][1], b->mv[2][1]);
369	42.6k	mc_chroma_dir(td, mc[4][b->filter][1],
370	42.6k	td->dst[1] + 4 * ls_uv, td->dst[2] + 4 * ls_uv, ls_uv,
371	42.6k	ref2->data[1], ref2->linesize[1],
372	42.6k	ref2->data[2], ref2->linesize[2], tref2,
373	42.6k	(row << 3) + 4, col << 2,
374	42.6k	&uvmv, 0, 4, 4, 8, 4, 4, w2, h2, 1);
375	59.0k	} else {
376	59.0k	mc_chroma_dir(td, mc[4][b->filter][1],
377	59.0k	td->dst[1], td->dst[2], ls_uv,
378	59.0k	ref2->data[1], ref2->linesize[1],
379	59.0k	ref2->data[2], ref2->linesize[2], tref2,
380	59.0k	row << 3, col << 3,
381	59.0k	&b->mv[0][1], 0, 0, 8, 8, 4, 4, w2, h2, 1);
382	59.0k	mc_chroma_dir(td, mc[4][b->filter][1],
383	59.0k	td->dst[1] + 4 * bytesperpixel,
384	59.0k	td->dst[2] + 4 * bytesperpixel, ls_uv,
385	59.0k	ref2->data[1], ref2->linesize[1],
386	59.0k	ref2->data[2], ref2->linesize[2], tref2,
387	59.0k	row << 3, (col << 3) + 4,
388	59.0k	&b->mv[1][1], 4, 0, 8, 8, 4, 4, w2, h2, 1);
389	59.0k	mc_chroma_dir(td, mc[4][b->filter][1],
390	59.0k	td->dst[1] + 4 * ls_uv, td->dst[2] + 4 * ls_uv, ls_uv,
391	59.0k	ref2->data[1], ref2->linesize[1],
392	59.0k	ref2->data[2], ref2->linesize[2], tref2,
393	59.0k	(row << 3) + 4, col << 3,
394	59.0k	&b->mv[2][1], 0, 4, 8, 8, 4, 4, w2, h2, 1);
395	59.0k	mc_chroma_dir(td, mc[4][b->filter][1],
396	59.0k	td->dst[1] + 4 * ls_uv + 4 * bytesperpixel,
397	59.0k	td->dst[2] + 4 * ls_uv + 4 * bytesperpixel, ls_uv,
398	59.0k	ref2->data[1], ref2->linesize[1],
399	59.0k	ref2->data[2], ref2->linesize[2], tref2,
400	59.0k	(row << 3) + 4, (col << 3) + 4,
401	59.0k	&b->mv[3][1], 4, 4, 8, 8, 4, 4, w2, h2, 1);
402	59.0k	}
403	101k	}
404	402k	}
405	531k	}
406	1.41M	} else {
407	1.41M	int bwl = bwlog_tab[0][b->bs];
408	1.41M	int bw = ff_vp9_bwh_tab[0][b->bs][0] * 4;
409	1.41M	int bh = ff_vp9_bwh_tab[0][b->bs][1] * 4;
410	1.41M	int uvbw = ff_vp9_bwh_tab[s->ss_h][b->bs][0] * 4;
411	1.41M	int uvbh = ff_vp9_bwh_tab[s->ss_v][b->bs][1] * 4;
412
413	1.41M	mc_luma_dir(td, mc[bwl][b->filter][0], td->dst[0], ls_y,
414	1.41M	ref1->data[0], ref1->linesize[0], tref1,
415	1.41M	row << 3, col << 3, &b->mv[0][0], 0, 0, bw, bh, bw, bh, w1, h1, 0);
416	1.41M	w1 = (w1 + s->ss_h) >> s->ss_h;
417	1.41M	h1 = (h1 + s->ss_v) >> s->ss_v;
418	1.41M	mc_chroma_dir(td, mc[bwl + s->ss_h][b->filter][0],
419	1.41M	td->dst[1], td->dst[2], ls_uv,
420	1.41M	ref1->data[1], ref1->linesize[1],
421	1.41M	ref1->data[2], ref1->linesize[2], tref1,
422	1.41M	row << (3 - s->ss_v), col << (3 - s->ss_h),
423	1.41M	&b->mv[0][0], 0, 0, uvbw, uvbh, uvbw, uvbh, w1, h1, 0);
424
425	1.41M	if (b->comp) {
426	984k	mc_luma_dir(td, mc[bwl][b->filter][1], td->dst[0], ls_y,
427	984k	ref2->data[0], ref2->linesize[0], tref2,
428	984k	row << 3, col << 3, &b->mv[0][1], 0, 0, bw, bh, bw, bh, w2, h2, 1);
429	984k	w2 = (w2 + s->ss_h) >> s->ss_h;
430	984k	h2 = (h2 + s->ss_v) >> s->ss_v;
431	984k	mc_chroma_dir(td, mc[bwl + s->ss_h][b->filter][1],
432	984k	td->dst[1], td->dst[2], ls_uv,
433	984k	ref2->data[1], ref2->linesize[1],
434	984k	ref2->data[2], ref2->linesize[2], tref2,
435	984k	row << (3 - s->ss_v), col << (3 - s->ss_h),
436	984k	&b->mv[0][1], 0, 0, uvbw, uvbh, uvbw, uvbh, w2, h2, 1);
437	984k	}
438	1.41M	}
439	1.94M	}

vp9recon.c:inter_pred_scaled_16bpp

Line	Count	Source
31	2.73M	{
32	2.73M	static const uint8_t bwlog_tab[2][N_BS_SIZES] = {
33	2.73M	{ 0, 0, 1, 1, 1, 2, 2, 2, 3, 3, 3, 4, 4 },
34	2.73M	{ 1, 1, 2, 2, 2, 3, 3, 3, 4, 4, 4, 4, 4 },
35	2.73M	};
36	2.73M	const VP9Context *s = td->s;
37	2.73M	VP9Block *b = td->b;
38	2.73M	int row = td->row, col = td->col;
39	2.73M	const ProgressFrame tref1 = &s->s.refs[s->s.h.refidx[b->ref[0]]], tref2;
40	2.73M	const AVFrame ref1 = tref1->f, ref2;
41	2.73M	int w1 = ref1->width, h1 = ref1->height, w2, h2;
42	2.73M	ptrdiff_t ls_y = td->y_stride, ls_uv = td->uv_stride;
43	2.73M	int bytesperpixel = BYTES_PER_PIXEL;
44
45	2.73M	if (b->comp) {
46	2.17M	tref2 = &s->s.refs[s->s.h.refidx[b->ref[1]]];
47	2.17M	ref2 = tref2->f;
48	2.17M	w2 = ref2->width;
49	2.17M	h2 = ref2->height;
50	2.17M	}
51
52		// y inter pred
53	2.73M	if (b->bs > BS_8x8) {
54	843k	VP9mv uvmv;
55
56		#if SCALED == 0
57		if (b->bs == BS_8x4) {
58		mc_luma_dir(td, mc[3][b->filter][0], td->dst[0], ls_y,
59		ref1->data[0], ref1->linesize[0], tref1,
60		row << 3, col << 3, &b->mv[0][0],,,,, 8, 4, w1, h1, 0);
61		mc_luma_dir(td, mc[3][b->filter][0],
62		td->dst[0] + 4 * ls_y, ls_y,
63		ref1->data[0], ref1->linesize[0], tref1,
64		(row << 3) + 4, col << 3, &b->mv[2][0],,,,, 8, 4, w1, h1, 0);
65		w1 = (w1 + s->ss_h) >> s->ss_h;
66		if (s->ss_v) {
67		h1 = (h1 + 1) >> 1;
68		uvmv = ROUNDED_DIV_MVx2(b->mv[0][0], b->mv[2][0]);
69		mc_chroma_dir(td, mc[3 + s->ss_h][b->filter][0],
70		td->dst[1], td->dst[2], ls_uv,
71		ref1->data[1], ref1->linesize[1],
72		ref1->data[2], ref1->linesize[2], tref1,
73		row << 2, col << (3 - s->ss_h),
74		&uvmv,,,,, 8 >> s->ss_h, 4, w1, h1, 0);
75		} else {
76		mc_chroma_dir(td, mc[3 + s->ss_h][b->filter][0],
77		td->dst[1], td->dst[2], ls_uv,
78		ref1->data[1], ref1->linesize[1],
79		ref1->data[2], ref1->linesize[2], tref1,
80		row << 3, col << (3 - s->ss_h),
81		&b->mv[0][0],,,,, 8 >> s->ss_h, 4, w1, h1, 0);
82		// BUG for 4:2:2 bs=8x4, libvpx uses the wrong block index
83		// to get the motion vector for the bottom 4x4 block
84		// https://code.google.com/p/webm/issues/detail?id=993
85		if (s->ss_h == 0) {
86		uvmv = b->mv[2][0];
87		} else {
88		uvmv = ROUNDED_DIV_MVx2(b->mv[0][0], b->mv[2][0]);
89		}
90		mc_chroma_dir(td, mc[3 + s->ss_h][b->filter][0],
91		td->dst[1] + 4 * ls_uv, td->dst[2] + 4 * ls_uv, ls_uv,
92		ref1->data[1], ref1->linesize[1],
93		ref1->data[2], ref1->linesize[2], tref1,
94		(row << 3) + 4, col << (3 - s->ss_h),
95		&uvmv,,,,, 8 >> s->ss_h, 4, w1, h1, 0);
96		}
97
98		if (b->comp) {
99		mc_luma_dir(td, mc[3][b->filter][1], td->dst[0], ls_y,
100		ref2->data[0], ref2->linesize[0], tref2,
101		row << 3, col << 3, &b->mv[0][1],,,,, 8, 4, w2, h2, 1);
102		mc_luma_dir(td, mc[3][b->filter][1],
103		td->dst[0] + 4 * ls_y, ls_y,
104		ref2->data[0], ref2->linesize[0], tref2,
105		(row << 3) + 4, col << 3, &b->mv[2][1],,,,, 8, 4, w2, h2, 1);
106		w2 = (w2 + s->ss_h) >> s->ss_h;
107		if (s->ss_v) {
108		h2 = (h2 + 1) >> 1;
109		uvmv = ROUNDED_DIV_MVx2(b->mv[0][1], b->mv[2][1]);
110		mc_chroma_dir(td, mc[3 + s->ss_h][b->filter][1],
111		td->dst[1], td->dst[2], ls_uv,
112		ref2->data[1], ref2->linesize[1],
113		ref2->data[2], ref2->linesize[2], tref2,
114		row << 2, col << (3 - s->ss_h),
115		&uvmv,,,,, 8 >> s->ss_h, 4, w2, h2, 1);
116		} else {
117		mc_chroma_dir(td, mc[3 + s->ss_h][b->filter][1],
118		td->dst[1], td->dst[2], ls_uv,
119		ref2->data[1], ref2->linesize[1],
120		ref2->data[2], ref2->linesize[2], tref2,
121		row << 3, col << (3 - s->ss_h),
122		&b->mv[0][1],,,,, 8 >> s->ss_h, 4, w2, h2, 1);
123		// BUG for 4:2:2 bs=8x4, libvpx uses the wrong block index
124		// to get the motion vector for the bottom 4x4 block
125		// https://code.google.com/p/webm/issues/detail?id=993
126		if (s->ss_h == 0) {
127		uvmv = b->mv[2][1];
128		} else {
129		uvmv = ROUNDED_DIV_MVx2(b->mv[0][1], b->mv[2][1]);
130		}
131		mc_chroma_dir(td, mc[3 + s->ss_h][b->filter][1],
132		td->dst[1] + 4 * ls_uv, td->dst[2] + 4 * ls_uv, ls_uv,
133		ref2->data[1], ref2->linesize[1],
134		ref2->data[2], ref2->linesize[2], tref2,
135		(row << 3) + 4, col << (3 - s->ss_h),
136		&uvmv,,,,, 8 >> s->ss_h, 4, w2, h2, 1);
137		}
138		}
139		} else if (b->bs == BS_4x8) {
140		mc_luma_dir(td, mc[4][b->filter][0], td->dst[0], ls_y,
141		ref1->data[0], ref1->linesize[0], tref1,
142		row << 3, col << 3, &b->mv[0][0],,,,, 4, 8, w1, h1, 0);
143		mc_luma_dir(td, mc[4][b->filter][0], td->dst[0] + 4 * bytesperpixel, ls_y,
144		ref1->data[0], ref1->linesize[0], tref1,
145		row << 3, (col << 3) + 4, &b->mv[1][0],,,,, 4, 8, w1, h1, 0);
146		h1 = (h1 + s->ss_v) >> s->ss_v;
147		if (s->ss_h) {
148		w1 = (w1 + 1) >> 1;
149		uvmv = ROUNDED_DIV_MVx2(b->mv[0][0], b->mv[1][0]);
150		mc_chroma_dir(td, mc[4][b->filter][0],
151		td->dst[1], td->dst[2], ls_uv,
152		ref1->data[1], ref1->linesize[1],
153		ref1->data[2], ref1->linesize[2], tref1,
154		row << (3 - s->ss_v), col << 2,
155		&uvmv,,,,, 4, 8 >> s->ss_v, w1, h1, 0);
156		} else {
157		mc_chroma_dir(td, mc[4][b->filter][0],
158		td->dst[1], td->dst[2], ls_uv,
159		ref1->data[1], ref1->linesize[1],
160		ref1->data[2], ref1->linesize[2], tref1,
161		row << (3 - s->ss_v), col << 3,
162		&b->mv[0][0],,,,, 4, 8 >> s->ss_v, w1, h1, 0);
163		mc_chroma_dir(td, mc[4][b->filter][0],
164		td->dst[1] + 4 * bytesperpixel,
165		td->dst[2] + 4 * bytesperpixel, ls_uv,
166		ref1->data[1], ref1->linesize[1],
167		ref1->data[2], ref1->linesize[2], tref1,
168		row << (3 - s->ss_v), (col << 3) + 4,
169		&b->mv[1][0],,,,, 4, 8 >> s->ss_v, w1, h1, 0);
170		}
171
172		if (b->comp) {
173		mc_luma_dir(td, mc[4][b->filter][1], td->dst[0], ls_y,
174		ref2->data[0], ref2->linesize[0], tref2,
175		row << 3, col << 3, &b->mv[0][1],,,,, 4, 8, w2, h2, 1);
176		mc_luma_dir(td, mc[4][b->filter][1], td->dst[0] + 4 * bytesperpixel, ls_y,
177		ref2->data[0], ref2->linesize[0], tref2,
178		row << 3, (col << 3) + 4, &b->mv[1][1],,,,, 4, 8, w2, h2, 1);
179		h2 = (h2 + s->ss_v) >> s->ss_v;
180		if (s->ss_h) {
181		w2 = (w2 + 1) >> 1;
182		uvmv = ROUNDED_DIV_MVx2(b->mv[0][1], b->mv[1][1]);
183		mc_chroma_dir(td, mc[4][b->filter][1],
184		td->dst[1], td->dst[2], ls_uv,
185		ref2->data[1], ref2->linesize[1],
186		ref2->data[2], ref2->linesize[2], tref2,
187		row << (3 - s->ss_v), col << 2,
188		&uvmv,,,,, 4, 8 >> s->ss_v, w2, h2, 1);
189		} else {
190		mc_chroma_dir(td, mc[4][b->filter][1],
191		td->dst[1], td->dst[2], ls_uv,
192		ref2->data[1], ref2->linesize[1],
193		ref2->data[2], ref2->linesize[2], tref2,
194		row << (3 - s->ss_v), col << 3,
195		&b->mv[0][1],,,,, 4, 8 >> s->ss_v, w2, h2, 1);
196		mc_chroma_dir(td, mc[4][b->filter][1],
197		td->dst[1] + 4 * bytesperpixel,
198		td->dst[2] + 4 * bytesperpixel, ls_uv,
199		ref2->data[1], ref2->linesize[1],
200		ref2->data[2], ref2->linesize[2], tref2,
201		row << (3 - s->ss_v), (col << 3) + 4,
202		&b->mv[1][1],,,,, 4, 8 >> s->ss_v, w2, h2, 1);
203		}
204		}
205		} else
206		#endif
207	843k	{
208		#if SCALED == 0
209		av_assert2(b->bs == BS_4x4);
210		#endif
211
212		// FIXME if two horizontally adjacent blocks have the same MV,
213		// do a w8 instead of a w4 call
214	843k	mc_luma_dir(td, mc[4][b->filter][0], td->dst[0], ls_y,
215	843k	ref1->data[0], ref1->linesize[0], tref1,
216	843k	row << 3, col << 3, &b->mv[0][0],
217	843k	0, 0, 8, 8, 4, 4, w1, h1, 0);
218	843k	mc_luma_dir(td, mc[4][b->filter][0], td->dst[0] + 4 * bytesperpixel, ls_y,
219	843k	ref1->data[0], ref1->linesize[0], tref1,
220	843k	row << 3, (col << 3) + 4, &b->mv[1][0],
221	843k	4, 0, 8, 8, 4, 4, w1, h1, 0);
222	843k	mc_luma_dir(td, mc[4][b->filter][0],
223	843k	td->dst[0] + 4 * ls_y, ls_y,
224	843k	ref1->data[0], ref1->linesize[0], tref1,
225	843k	(row << 3) + 4, col << 3, &b->mv[2][0],
226	843k	0, 4, 8, 8, 4, 4, w1, h1, 0);
227	843k	mc_luma_dir(td, mc[4][b->filter][0],
228	843k	td->dst[0] + 4 * ls_y + 4 * bytesperpixel, ls_y,
229	843k	ref1->data[0], ref1->linesize[0], tref1,
230	843k	(row << 3) + 4, (col << 3) + 4, &b->mv[3][0],
231	843k	4, 4, 8, 8, 4, 4, w1, h1, 0);
232	843k	if (s->ss_v) {
233	331k	h1 = (h1 + 1) >> 1;
234	331k	if (s->ss_h) {
235	68.0k	w1 = (w1 + 1) >> 1;
236	68.0k	uvmv = ROUNDED_DIV_MVx4(b->mv[0][0], b->mv[1][0],
237	68.0k	b->mv[2][0], b->mv[3][0]);
238	68.0k	mc_chroma_dir(td, mc[4][b->filter][0],
239	68.0k	td->dst[1], td->dst[2], ls_uv,
240	68.0k	ref1->data[1], ref1->linesize[1],
241	68.0k	ref1->data[2], ref1->linesize[2], tref1,
242	68.0k	row << 2, col << 2,
243	68.0k	&uvmv, 0, 0, 4, 4, 4, 4, w1, h1, 0);
244	263k	} else {
245	263k	uvmv = ROUNDED_DIV_MVx2(b->mv[0][0], b->mv[2][0]);
246	263k	mc_chroma_dir(td, mc[4][b->filter][0],
247	263k	td->dst[1], td->dst[2], ls_uv,
248	263k	ref1->data[1], ref1->linesize[1],
249	263k	ref1->data[2], ref1->linesize[2], tref1,
250	263k	row << 2, col << 3,
251	263k	&uvmv, 0, 0, 8, 4, 4, 4, w1, h1, 0);
252	263k	uvmv = ROUNDED_DIV_MVx2(b->mv[1][0], b->mv[3][0]);
253	263k	mc_chroma_dir(td, mc[4][b->filter][0],
254	263k	td->dst[1] + 4 * bytesperpixel,
255	263k	td->dst[2] + 4 * bytesperpixel, ls_uv,
256	263k	ref1->data[1], ref1->linesize[1],
257	263k	ref1->data[2], ref1->linesize[2], tref1,
258	263k	row << 2, (col << 3) + 4,
259	263k	&uvmv, 4, 0, 8, 4, 4, 4, w1, h1, 0);
260	263k	}
261	511k	} else {
262	511k	if (s->ss_h) {
263	60.3k	w1 = (w1 + 1) >> 1;
264	60.3k	uvmv = ROUNDED_DIV_MVx2(b->mv[0][0], b->mv[1][0]);
265	60.3k	mc_chroma_dir(td, mc[4][b->filter][0],
266	60.3k	td->dst[1], td->dst[2], ls_uv,
267	60.3k	ref1->data[1], ref1->linesize[1],
268	60.3k	ref1->data[2], ref1->linesize[2], tref1,
269	60.3k	row << 3, col << 2,
270	60.3k	&uvmv, 0, 0, 4, 8, 4, 4, w1, h1, 0);
271		// BUG libvpx uses wrong block index for 4:2:2 bs=4x4
272		// bottom block
273		// https://code.google.com/p/webm/issues/detail?id=993
274	60.3k	uvmv = ROUNDED_DIV_MVx2(b->mv[1][0], b->mv[2][0]);
275	60.3k	mc_chroma_dir(td, mc[4][b->filter][0],
276	60.3k	td->dst[1] + 4 * ls_uv, td->dst[2] + 4 * ls_uv, ls_uv,
277	60.3k	ref1->data[1], ref1->linesize[1],
278	60.3k	ref1->data[2], ref1->linesize[2], tref1,
279	60.3k	(row << 3) + 4, col << 2,
280	60.3k	&uvmv, 0, 4, 4, 8, 4, 4, w1, h1, 0);
281	451k	} else {
282	451k	mc_chroma_dir(td, mc[4][b->filter][0],
283	451k	td->dst[1], td->dst[2], ls_uv,
284	451k	ref1->data[1], ref1->linesize[1],
285	451k	ref1->data[2], ref1->linesize[2], tref1,
286	451k	row << 3, col << 3,
287	451k	&b->mv[0][0], 0, 0, 8, 8, 4, 4, w1, h1, 0);
288	451k	mc_chroma_dir(td, mc[4][b->filter][0],
289	451k	td->dst[1] + 4 * bytesperpixel,
290	451k	td->dst[2] + 4 * bytesperpixel, ls_uv,
291	451k	ref1->data[1], ref1->linesize[1],
292	451k	ref1->data[2], ref1->linesize[2], tref1,
293	451k	row << 3, (col << 3) + 4,
294	451k	&b->mv[1][0], 4, 0, 8, 8, 4, 4, w1, h1, 0);
295	451k	mc_chroma_dir(td, mc[4][b->filter][0],
296	451k	td->dst[1] + 4 * ls_uv, td->dst[2] + 4 * ls_uv, ls_uv,
297	451k	ref1->data[1], ref1->linesize[1],
298	451k	ref1->data[2], ref1->linesize[2], tref1,
299	451k	(row << 3) + 4, col << 3,
300	451k	&b->mv[2][0], 0, 4, 8, 8, 4, 4, w1, h1, 0);
301	451k	mc_chroma_dir(td, mc[4][b->filter][0],
302	451k	td->dst[1] + 4 * ls_uv + 4 * bytesperpixel,
303	451k	td->dst[2] + 4 * ls_uv + 4 * bytesperpixel, ls_uv,
304	451k	ref1->data[1], ref1->linesize[1],
305	451k	ref1->data[2], ref1->linesize[2], tref1,
306	451k	(row << 3) + 4, (col << 3) + 4,
307	451k	&b->mv[3][0], 4, 4, 8, 8, 4, 4, w1, h1, 0);
308	451k	}
309	511k	}
310
311	843k	if (b->comp) {
312	733k	mc_luma_dir(td, mc[4][b->filter][1], td->dst[0], ls_y,
313	733k	ref2->data[0], ref2->linesize[0], tref2,
314	733k	row << 3, col << 3, &b->mv[0][1], 0, 0, 8, 8, 4, 4, w2, h2, 1);
315	733k	mc_luma_dir(td, mc[4][b->filter][1], td->dst[0] + 4 * bytesperpixel, ls_y,
316	733k	ref2->data[0], ref2->linesize[0], tref2,
317	733k	row << 3, (col << 3) + 4, &b->mv[1][1], 4, 0, 8, 8, 4, 4, w2, h2, 1);
318	733k	mc_luma_dir(td, mc[4][b->filter][1],
319	733k	td->dst[0] + 4 * ls_y, ls_y,
320	733k	ref2->data[0], ref2->linesize[0], tref2,
321	733k	(row << 3) + 4, col << 3, &b->mv[2][1], 0, 4, 8, 8, 4, 4, w2, h2, 1);
322	733k	mc_luma_dir(td, mc[4][b->filter][1],
323	733k	td->dst[0] + 4 * ls_y + 4 * bytesperpixel, ls_y,
324	733k	ref2->data[0], ref2->linesize[0], tref2,
325	733k	(row << 3) + 4, (col << 3) + 4, &b->mv[3][1], 4, 4, 8, 8, 4, 4, w2, h2, 1);
326	733k	if (s->ss_v) {
327	294k	h2 = (h2 + 1) >> 1;
328	294k	if (s->ss_h) {
329	56.5k	w2 = (w2 + 1) >> 1;
330	56.5k	uvmv = ROUNDED_DIV_MVx4(b->mv[0][1], b->mv[1][1],
331	56.5k	b->mv[2][1], b->mv[3][1]);
332	56.5k	mc_chroma_dir(td, mc[4][b->filter][1],
333	56.5k	td->dst[1], td->dst[2], ls_uv,
334	56.5k	ref2->data[1], ref2->linesize[1],
335	56.5k	ref2->data[2], ref2->linesize[2], tref2,
336	56.5k	row << 2, col << 2,
337	56.5k	&uvmv, 0, 0, 4, 4, 4, 4, w2, h2, 1);
338	238k	} else {
339	238k	uvmv = ROUNDED_DIV_MVx2(b->mv[0][1], b->mv[2][1]);
340	238k	mc_chroma_dir(td, mc[4][b->filter][1],
341	238k	td->dst[1], td->dst[2], ls_uv,
342	238k	ref2->data[1], ref2->linesize[1],
343	238k	ref2->data[2], ref2->linesize[2], tref2,
344	238k	row << 2, col << 3,
345	238k	&uvmv, 0, 0, 8, 4, 4, 4, w2, h2, 1);
346	238k	uvmv = ROUNDED_DIV_MVx2(b->mv[1][1], b->mv[3][1]);
347	238k	mc_chroma_dir(td, mc[4][b->filter][1],
348	238k	td->dst[1] + 4 * bytesperpixel,
349	238k	td->dst[2] + 4 * bytesperpixel, ls_uv,
350	238k	ref2->data[1], ref2->linesize[1],
351	238k	ref2->data[2], ref2->linesize[2], tref2,
352	238k	row << 2, (col << 3) + 4,
353	238k	&uvmv, 4, 0, 8, 4, 4, 4, w2, h2, 1);
354	238k	}
355	438k	} else {
356	438k	if (s->ss_h) {
357	42.7k	w2 = (w2 + 1) >> 1;
358	42.7k	uvmv = ROUNDED_DIV_MVx2(b->mv[0][1], b->mv[1][1]);
359	42.7k	mc_chroma_dir(td, mc[4][b->filter][1],
360	42.7k	td->dst[1], td->dst[2], ls_uv,
361	42.7k	ref2->data[1], ref2->linesize[1],
362	42.7k	ref2->data[2], ref2->linesize[2], tref2,
363	42.7k	row << 3, col << 2,
364	42.7k	&uvmv, 0, 0, 4, 8, 4, 4, w2, h2, 1);
365		// BUG libvpx uses wrong block index for 4:2:2 bs=4x4
366		// bottom block
367		// https://code.google.com/p/webm/issues/detail?id=993
368	42.7k	uvmv = ROUNDED_DIV_MVx2(b->mv[1][1], b->mv[2][1]);
369	42.7k	mc_chroma_dir(td, mc[4][b->filter][1],
370	42.7k	td->dst[1] + 4 * ls_uv, td->dst[2] + 4 * ls_uv, ls_uv,
371	42.7k	ref2->data[1], ref2->linesize[1],
372	42.7k	ref2->data[2], ref2->linesize[2], tref2,
373	42.7k	(row << 3) + 4, col << 2,
374	42.7k	&uvmv, 0, 4, 4, 8, 4, 4, w2, h2, 1);
375	395k	} else {
376	395k	mc_chroma_dir(td, mc[4][b->filter][1],
377	395k	td->dst[1], td->dst[2], ls_uv,
378	395k	ref2->data[1], ref2->linesize[1],
379	395k	ref2->data[2], ref2->linesize[2], tref2,
380	395k	row << 3, col << 3,
381	395k	&b->mv[0][1], 0, 0, 8, 8, 4, 4, w2, h2, 1);
382	395k	mc_chroma_dir(td, mc[4][b->filter][1],
383	395k	td->dst[1] + 4 * bytesperpixel,
384	395k	td->dst[2] + 4 * bytesperpixel, ls_uv,
385	395k	ref2->data[1], ref2->linesize[1],
386	395k	ref2->data[2], ref2->linesize[2], tref2,
387	395k	row << 3, (col << 3) + 4,
388	395k	&b->mv[1][1], 4, 0, 8, 8, 4, 4, w2, h2, 1);
389	395k	mc_chroma_dir(td, mc[4][b->filter][1],
390	395k	td->dst[1] + 4 * ls_uv, td->dst[2] + 4 * ls_uv, ls_uv,
391	395k	ref2->data[1], ref2->linesize[1],
392	395k	ref2->data[2], ref2->linesize[2], tref2,
393	395k	(row << 3) + 4, col << 3,
394	395k	&b->mv[2][1], 0, 4, 8, 8, 4, 4, w2, h2, 1);
395	395k	mc_chroma_dir(td, mc[4][b->filter][1],
396	395k	td->dst[1] + 4 * ls_uv + 4 * bytesperpixel,
397	395k	td->dst[2] + 4 * ls_uv + 4 * bytesperpixel, ls_uv,
398	395k	ref2->data[1], ref2->linesize[1],
399	395k	ref2->data[2], ref2->linesize[2], tref2,
400	395k	(row << 3) + 4, (col << 3) + 4,
401	395k	&b->mv[3][1], 4, 4, 8, 8, 4, 4, w2, h2, 1);
402	395k	}
403	438k	}
404	733k	}
405	843k	}
406	1.89M	} else {
407	1.89M	int bwl = bwlog_tab[0][b->bs];
408	1.89M	int bw = ff_vp9_bwh_tab[0][b->bs][0] * 4;
409	1.89M	int bh = ff_vp9_bwh_tab[0][b->bs][1] * 4;
410	1.89M	int uvbw = ff_vp9_bwh_tab[s->ss_h][b->bs][0] * 4;
411	1.89M	int uvbh = ff_vp9_bwh_tab[s->ss_v][b->bs][1] * 4;
412
413	1.89M	mc_luma_dir(td, mc[bwl][b->filter][0], td->dst[0], ls_y,
414	1.89M	ref1->data[0], ref1->linesize[0], tref1,
415	1.89M	row << 3, col << 3, &b->mv[0][0], 0, 0, bw, bh, bw, bh, w1, h1, 0);
416	1.89M	w1 = (w1 + s->ss_h) >> s->ss_h;
417	1.89M	h1 = (h1 + s->ss_v) >> s->ss_v;
418	1.89M	mc_chroma_dir(td, mc[bwl + s->ss_h][b->filter][0],
419	1.89M	td->dst[1], td->dst[2], ls_uv,
420	1.89M	ref1->data[1], ref1->linesize[1],
421	1.89M	ref1->data[2], ref1->linesize[2], tref1,
422	1.89M	row << (3 - s->ss_v), col << (3 - s->ss_h),
423	1.89M	&b->mv[0][0], 0, 0, uvbw, uvbh, uvbw, uvbh, w1, h1, 0);
424
425	1.89M	if (b->comp) {
426	1.43M	mc_luma_dir(td, mc[bwl][b->filter][1], td->dst[0], ls_y,
427	1.43M	ref2->data[0], ref2->linesize[0], tref2,
428	1.43M	row << 3, col << 3, &b->mv[0][1], 0, 0, bw, bh, bw, bh, w2, h2, 1);
429	1.43M	w2 = (w2 + s->ss_h) >> s->ss_h;
430	1.43M	h2 = (h2 + s->ss_v) >> s->ss_v;
431	1.43M	mc_chroma_dir(td, mc[bwl + s->ss_h][b->filter][1],
432	1.43M	td->dst[1], td->dst[2], ls_uv,
433	1.43M	ref2->data[1], ref2->linesize[1],
434	1.43M	ref2->data[2], ref2->linesize[2], tref2,
435	1.43M	row << (3 - s->ss_v), col << (3 - s->ss_h),
436	1.43M	&b->mv[0][1], 0, 0, uvbw, uvbh, uvbw, uvbh, w2, h2, 1);
437	1.43M	}
438	1.89M	}
439	2.73M	}

vp9recon.c:inter_pred_8bpp

Line	Count	Source
31	2.24M	{
32	2.24M	static const uint8_t bwlog_tab[2][N_BS_SIZES] = {
33	2.24M	{ 0, 0, 1, 1, 1, 2, 2, 2, 3, 3, 3, 4, 4 },
34	2.24M	{ 1, 1, 2, 2, 2, 3, 3, 3, 4, 4, 4, 4, 4 },
35	2.24M	};
36	2.24M	const VP9Context *s = td->s;
37	2.24M	VP9Block *b = td->b;
38	2.24M	int row = td->row, col = td->col;
39	2.24M	const ProgressFrame tref1 = &s->s.refs[s->s.h.refidx[b->ref[0]]], tref2;
40	2.24M	const AVFrame ref1 = tref1->f, ref2;
41	2.24M	int w1 = ref1->width, h1 = ref1->height, w2, h2;
42	2.24M	ptrdiff_t ls_y = td->y_stride, ls_uv = td->uv_stride;
43	2.24M	int bytesperpixel = BYTES_PER_PIXEL;
44
45	2.24M	if (b->comp) {
46	833k	tref2 = &s->s.refs[s->s.h.refidx[b->ref[1]]];
47	833k	ref2 = tref2->f;
48	833k	w2 = ref2->width;
49	833k	h2 = ref2->height;
50	833k	}
51
52		// y inter pred
53	2.24M	if (b->bs > BS_8x8) {
54	614k	VP9mv uvmv;
55
56	614k	#if SCALED == 0
57	614k	if (b->bs == BS_8x4) {
58	207k	mc_luma_dir(td, mc[3][b->filter][0], td->dst[0], ls_y,
59	207k	ref1->data[0], ref1->linesize[0], tref1,
60	207k	row << 3, col << 3, &b->mv[0][0],,,,, 8, 4, w1, h1, 0);
61	207k	mc_luma_dir(td, mc[3][b->filter][0],
62	207k	td->dst[0] + 4 * ls_y, ls_y,
63	207k	ref1->data[0], ref1->linesize[0], tref1,
64	207k	(row << 3) + 4, col << 3, &b->mv[2][0],,,,, 8, 4, w1, h1, 0);
65	207k	w1 = (w1 + s->ss_h) >> s->ss_h;
66	207k	if (s->ss_v) {
67	86.7k	h1 = (h1 + 1) >> 1;
68	86.7k	uvmv = ROUNDED_DIV_MVx2(b->mv[0][0], b->mv[2][0]);
69	86.7k	mc_chroma_dir(td, mc[3 + s->ss_h][b->filter][0],
70	86.7k	td->dst[1], td->dst[2], ls_uv,
71	86.7k	ref1->data[1], ref1->linesize[1],
72	86.7k	ref1->data[2], ref1->linesize[2], tref1,
73	86.7k	row << 2, col << (3 - s->ss_h),
74	86.7k	&uvmv,,,,, 8 >> s->ss_h, 4, w1, h1, 0);
75	120k	} else {
76	120k	mc_chroma_dir(td, mc[3 + s->ss_h][b->filter][0],
77	120k	td->dst[1], td->dst[2], ls_uv,
78	120k	ref1->data[1], ref1->linesize[1],
79	120k	ref1->data[2], ref1->linesize[2], tref1,
80	120k	row << 3, col << (3 - s->ss_h),
81	120k	&b->mv[0][0],,,,, 8 >> s->ss_h, 4, w1, h1, 0);
82		// BUG for 4:2:2 bs=8x4, libvpx uses the wrong block index
83		// to get the motion vector for the bottom 4x4 block
84		// https://code.google.com/p/webm/issues/detail?id=993
85	120k	if (s->ss_h == 0) {
86	56.4k	uvmv = b->mv[2][0];
87	64.5k	} else {
88	64.5k	uvmv = ROUNDED_DIV_MVx2(b->mv[0][0], b->mv[2][0]);
89	64.5k	}
90	120k	mc_chroma_dir(td, mc[3 + s->ss_h][b->filter][0],
91	120k	td->dst[1] + 4 * ls_uv, td->dst[2] + 4 * ls_uv, ls_uv,
92	120k	ref1->data[1], ref1->linesize[1],
93	120k	ref1->data[2], ref1->linesize[2], tref1,
94	120k	(row << 3) + 4, col << (3 - s->ss_h),
95	120k	&uvmv,,,,, 8 >> s->ss_h, 4, w1, h1, 0);
96	120k	}
97
98	207k	if (b->comp) {
99	80.9k	mc_luma_dir(td, mc[3][b->filter][1], td->dst[0], ls_y,
100	80.9k	ref2->data[0], ref2->linesize[0], tref2,
101	80.9k	row << 3, col << 3, &b->mv[0][1],,,,, 8, 4, w2, h2, 1);
102	80.9k	mc_luma_dir(td, mc[3][b->filter][1],
103	80.9k	td->dst[0] + 4 * ls_y, ls_y,
104	80.9k	ref2->data[0], ref2->linesize[0], tref2,
105	80.9k	(row << 3) + 4, col << 3, &b->mv[2][1],,,,, 8, 4, w2, h2, 1);
106	80.9k	w2 = (w2 + s->ss_h) >> s->ss_h;
107	80.9k	if (s->ss_v) {
108	43.9k	h2 = (h2 + 1) >> 1;
109	43.9k	uvmv = ROUNDED_DIV_MVx2(b->mv[0][1], b->mv[2][1]);
110	43.9k	mc_chroma_dir(td, mc[3 + s->ss_h][b->filter][1],
111	43.9k	td->dst[1], td->dst[2], ls_uv,
112	43.9k	ref2->data[1], ref2->linesize[1],
113	43.9k	ref2->data[2], ref2->linesize[2], tref2,
114	43.9k	row << 2, col << (3 - s->ss_h),
115	43.9k	&uvmv,,,,, 8 >> s->ss_h, 4, w2, h2, 1);
116	43.9k	} else {
117	37.0k	mc_chroma_dir(td, mc[3 + s->ss_h][b->filter][1],
118	37.0k	td->dst[1], td->dst[2], ls_uv,
119	37.0k	ref2->data[1], ref2->linesize[1],
120	37.0k	ref2->data[2], ref2->linesize[2], tref2,
121	37.0k	row << 3, col << (3 - s->ss_h),
122	37.0k	&b->mv[0][1],,,,, 8 >> s->ss_h, 4, w2, h2, 1);
123		// BUG for 4:2:2 bs=8x4, libvpx uses the wrong block index
124		// to get the motion vector for the bottom 4x4 block
125		// https://code.google.com/p/webm/issues/detail?id=993
126	37.0k	if (s->ss_h == 0) {
127	27.3k	uvmv = b->mv[2][1];
128	27.3k	} else {
129	9.68k	uvmv = ROUNDED_DIV_MVx2(b->mv[0][1], b->mv[2][1]);
130	9.68k	}
131	37.0k	mc_chroma_dir(td, mc[3 + s->ss_h][b->filter][1],
132	37.0k	td->dst[1] + 4 * ls_uv, td->dst[2] + 4 * ls_uv, ls_uv,
133	37.0k	ref2->data[1], ref2->linesize[1],
134	37.0k	ref2->data[2], ref2->linesize[2], tref2,
135	37.0k	(row << 3) + 4, col << (3 - s->ss_h),
136	37.0k	&uvmv,,,,, 8 >> s->ss_h, 4, w2, h2, 1);
137	37.0k	}
138	80.9k	}
139	406k	} else if (b->bs == BS_4x8) {
140	173k	mc_luma_dir(td, mc[4][b->filter][0], td->dst[0], ls_y,
141	173k	ref1->data[0], ref1->linesize[0], tref1,
142	173k	row << 3, col << 3, &b->mv[0][0],,,,, 4, 8, w1, h1, 0);
143	173k	mc_luma_dir(td, mc[4][b->filter][0], td->dst[0] + 4 * bytesperpixel, ls_y,
144	173k	ref1->data[0], ref1->linesize[0], tref1,
145	173k	row << 3, (col << 3) + 4, &b->mv[1][0],,,,, 4, 8, w1, h1, 0);
146	173k	h1 = (h1 + s->ss_v) >> s->ss_v;
147	173k	if (s->ss_h) {
148	106k	w1 = (w1 + 1) >> 1;
149	106k	uvmv = ROUNDED_DIV_MVx2(b->mv[0][0], b->mv[1][0]);
150	106k	mc_chroma_dir(td, mc[4][b->filter][0],
151	106k	td->dst[1], td->dst[2], ls_uv,
152	106k	ref1->data[1], ref1->linesize[1],
153	106k	ref1->data[2], ref1->linesize[2], tref1,
154	106k	row << (3 - s->ss_v), col << 2,
155	106k	&uvmv,,,,, 4, 8 >> s->ss_v, w1, h1, 0);
156	106k	} else {
157	67.3k	mc_chroma_dir(td, mc[4][b->filter][0],
158	67.3k	td->dst[1], td->dst[2], ls_uv,
159	67.3k	ref1->data[1], ref1->linesize[1],
160	67.3k	ref1->data[2], ref1->linesize[2], tref1,
161	67.3k	row << (3 - s->ss_v), col << 3,
162	67.3k	&b->mv[0][0],,,,, 4, 8 >> s->ss_v, w1, h1, 0);
163	67.3k	mc_chroma_dir(td, mc[4][b->filter][0],
164	67.3k	td->dst[1] + 4 * bytesperpixel,
165	67.3k	td->dst[2] + 4 * bytesperpixel, ls_uv,
166	67.3k	ref1->data[1], ref1->linesize[1],
167	67.3k	ref1->data[2], ref1->linesize[2], tref1,
168	67.3k	row << (3 - s->ss_v), (col << 3) + 4,
169	67.3k	&b->mv[1][0],,,,, 4, 8 >> s->ss_v, w1, h1, 0);
170	67.3k	}
171
172	173k	if (b->comp) {
173	61.4k	mc_luma_dir(td, mc[4][b->filter][1], td->dst[0], ls_y,
174	61.4k	ref2->data[0], ref2->linesize[0], tref2,
175	61.4k	row << 3, col << 3, &b->mv[0][1],,,,, 4, 8, w2, h2, 1);
176	61.4k	mc_luma_dir(td, mc[4][b->filter][1], td->dst[0] + 4 * bytesperpixel, ls_y,
177	61.4k	ref2->data[0], ref2->linesize[0], tref2,
178	61.4k	row << 3, (col << 3) + 4, &b->mv[1][1],,,,, 4, 8, w2, h2, 1);
179	61.4k	h2 = (h2 + s->ss_v) >> s->ss_v;
180	61.4k	if (s->ss_h) {
181	32.0k	w2 = (w2 + 1) >> 1;
182	32.0k	uvmv = ROUNDED_DIV_MVx2(b->mv[0][1], b->mv[1][1]);
183	32.0k	mc_chroma_dir(td, mc[4][b->filter][1],
184	32.0k	td->dst[1], td->dst[2], ls_uv,
185	32.0k	ref2->data[1], ref2->linesize[1],
186	32.0k	ref2->data[2], ref2->linesize[2], tref2,
187	32.0k	row << (3 - s->ss_v), col << 2,
188	32.0k	&uvmv,,,,, 4, 8 >> s->ss_v, w2, h2, 1);
189	32.0k	} else {
190	29.3k	mc_chroma_dir(td, mc[4][b->filter][1],
191	29.3k	td->dst[1], td->dst[2], ls_uv,
192	29.3k	ref2->data[1], ref2->linesize[1],
193	29.3k	ref2->data[2], ref2->linesize[2], tref2,
194	29.3k	row << (3 - s->ss_v), col << 3,
195	29.3k	&b->mv[0][1],,,,, 4, 8 >> s->ss_v, w2, h2, 1);
196	29.3k	mc_chroma_dir(td, mc[4][b->filter][1],
197	29.3k	td->dst[1] + 4 * bytesperpixel,
198	29.3k	td->dst[2] + 4 * bytesperpixel, ls_uv,
199	29.3k	ref2->data[1], ref2->linesize[1],
200	29.3k	ref2->data[2], ref2->linesize[2], tref2,
201	29.3k	row << (3 - s->ss_v), (col << 3) + 4,
202	29.3k	&b->mv[1][1],,,,, 4, 8 >> s->ss_v, w2, h2, 1);
203	29.3k	}
204	61.4k	}
205	173k	} else
206	232k	#endif
207	232k	{
208	232k	#if SCALED == 0
209	232k	av_assert2(b->bs == BS_4x4);
210	232k	#endif
211
212		// FIXME if two horizontally adjacent blocks have the same MV,
213		// do a w8 instead of a w4 call
214	232k	mc_luma_dir(td, mc[4][b->filter][0], td->dst[0], ls_y,
215	232k	ref1->data[0], ref1->linesize[0], tref1,
216	232k	row << 3, col << 3, &b->mv[0][0],
217	232k	0, 0, 8, 8, 4, 4, w1, h1, 0);
218	232k	mc_luma_dir(td, mc[4][b->filter][0], td->dst[0] + 4 * bytesperpixel, ls_y,
219	232k	ref1->data[0], ref1->linesize[0], tref1,
220	232k	row << 3, (col << 3) + 4, &b->mv[1][0],
221	232k	4, 0, 8, 8, 4, 4, w1, h1, 0);
222	232k	mc_luma_dir(td, mc[4][b->filter][0],
223	232k	td->dst[0] + 4 * ls_y, ls_y,
224	232k	ref1->data[0], ref1->linesize[0], tref1,
225	232k	(row << 3) + 4, col << 3, &b->mv[2][0],
226	232k	0, 4, 8, 8, 4, 4, w1, h1, 0);
227	232k	mc_luma_dir(td, mc[4][b->filter][0],
228	232k	td->dst[0] + 4 * ls_y + 4 * bytesperpixel, ls_y,
229	232k	ref1->data[0], ref1->linesize[0], tref1,
230	232k	(row << 3) + 4, (col << 3) + 4, &b->mv[3][0],
231	232k	4, 4, 8, 8, 4, 4, w1, h1, 0);
232	232k	if (s->ss_v) {
233	79.3k	h1 = (h1 + 1) >> 1;
234	79.3k	if (s->ss_h) {
235	37.9k	w1 = (w1 + 1) >> 1;
236	37.9k	uvmv = ROUNDED_DIV_MVx4(b->mv[0][0], b->mv[1][0],
237	37.9k	b->mv[2][0], b->mv[3][0]);
238	37.9k	mc_chroma_dir(td, mc[4][b->filter][0],
239	37.9k	td->dst[1], td->dst[2], ls_uv,
240	37.9k	ref1->data[1], ref1->linesize[1],
241	37.9k	ref1->data[2], ref1->linesize[2], tref1,
242	37.9k	row << 2, col << 2,
243	37.9k	&uvmv, 0, 0, 4, 4, 4, 4, w1, h1, 0);
244	41.4k	} else {
245	41.4k	uvmv = ROUNDED_DIV_MVx2(b->mv[0][0], b->mv[2][0]);
246	41.4k	mc_chroma_dir(td, mc[4][b->filter][0],
247	41.4k	td->dst[1], td->dst[2], ls_uv,
248	41.4k	ref1->data[1], ref1->linesize[1],
249	41.4k	ref1->data[2], ref1->linesize[2], tref1,
250	41.4k	row << 2, col << 3,
251	41.4k	&uvmv, 0, 0, 8, 4, 4, 4, w1, h1, 0);
252	41.4k	uvmv = ROUNDED_DIV_MVx2(b->mv[1][0], b->mv[3][0]);
253	41.4k	mc_chroma_dir(td, mc[4][b->filter][0],
254	41.4k	td->dst[1] + 4 * bytesperpixel,
255	41.4k	td->dst[2] + 4 * bytesperpixel, ls_uv,
256	41.4k	ref1->data[1], ref1->linesize[1],
257	41.4k	ref1->data[2], ref1->linesize[2], tref1,
258	41.4k	row << 2, (col << 3) + 4,
259	41.4k	&uvmv, 4, 0, 8, 4, 4, 4, w1, h1, 0);
260	41.4k	}
261	153k	} else {
262	153k	if (s->ss_h) {
263	76.9k	w1 = (w1 + 1) >> 1;
264	76.9k	uvmv = ROUNDED_DIV_MVx2(b->mv[0][0], b->mv[1][0]);
265	76.9k	mc_chroma_dir(td, mc[4][b->filter][0],
266	76.9k	td->dst[1], td->dst[2], ls_uv,
267	76.9k	ref1->data[1], ref1->linesize[1],
268	76.9k	ref1->data[2], ref1->linesize[2], tref1,
269	76.9k	row << 3, col << 2,
270	76.9k	&uvmv, 0, 0, 4, 8, 4, 4, w1, h1, 0);
271		// BUG libvpx uses wrong block index for 4:2:2 bs=4x4
272		// bottom block
273		// https://code.google.com/p/webm/issues/detail?id=993
274	76.9k	uvmv = ROUNDED_DIV_MVx2(b->mv[1][0], b->mv[2][0]);
275	76.9k	mc_chroma_dir(td, mc[4][b->filter][0],
276	76.9k	td->dst[1] + 4 * ls_uv, td->dst[2] + 4 * ls_uv, ls_uv,
277	76.9k	ref1->data[1], ref1->linesize[1],
278	76.9k	ref1->data[2], ref1->linesize[2], tref1,
279	76.9k	(row << 3) + 4, col << 2,
280	76.9k	&uvmv, 0, 4, 4, 8, 4, 4, w1, h1, 0);
281	76.9k	} else {
282	76.5k	mc_chroma_dir(td, mc[4][b->filter][0],
283	76.5k	td->dst[1], td->dst[2], ls_uv,
284	76.5k	ref1->data[1], ref1->linesize[1],
285	76.5k	ref1->data[2], ref1->linesize[2], tref1,
286	76.5k	row << 3, col << 3,
287	76.5k	&b->mv[0][0], 0, 0, 8, 8, 4, 4, w1, h1, 0);
288	76.5k	mc_chroma_dir(td, mc[4][b->filter][0],
289	76.5k	td->dst[1] + 4 * bytesperpixel,
290	76.5k	td->dst[2] + 4 * bytesperpixel, ls_uv,
291	76.5k	ref1->data[1], ref1->linesize[1],
292	76.5k	ref1->data[2], ref1->linesize[2], tref1,
293	76.5k	row << 3, (col << 3) + 4,
294	76.5k	&b->mv[1][0], 4, 0, 8, 8, 4, 4, w1, h1, 0);
295	76.5k	mc_chroma_dir(td, mc[4][b->filter][0],
296	76.5k	td->dst[1] + 4 * ls_uv, td->dst[2] + 4 * ls_uv, ls_uv,
297	76.5k	ref1->data[1], ref1->linesize[1],
298	76.5k	ref1->data[2], ref1->linesize[2], tref1,
299	76.5k	(row << 3) + 4, col << 3,
300	76.5k	&b->mv[2][0], 0, 4, 8, 8, 4, 4, w1, h1, 0);
301	76.5k	mc_chroma_dir(td, mc[4][b->filter][0],
302	76.5k	td->dst[1] + 4 * ls_uv + 4 * bytesperpixel,
303	76.5k	td->dst[2] + 4 * ls_uv + 4 * bytesperpixel, ls_uv,
304	76.5k	ref1->data[1], ref1->linesize[1],
305	76.5k	ref1->data[2], ref1->linesize[2], tref1,
306	76.5k	(row << 3) + 4, (col << 3) + 4,
307	76.5k	&b->mv[3][0], 4, 4, 8, 8, 4, 4, w1, h1, 0);
308	76.5k	}
309	153k	}
310
311	232k	if (b->comp) {
312	92.5k	mc_luma_dir(td, mc[4][b->filter][1], td->dst[0], ls_y,
313	92.5k	ref2->data[0], ref2->linesize[0], tref2,
314	92.5k	row << 3, col << 3, &b->mv[0][1], 0, 0, 8, 8, 4, 4, w2, h2, 1);
315	92.5k	mc_luma_dir(td, mc[4][b->filter][1], td->dst[0] + 4 * bytesperpixel, ls_y,
316	92.5k	ref2->data[0], ref2->linesize[0], tref2,
317	92.5k	row << 3, (col << 3) + 4, &b->mv[1][1], 4, 0, 8, 8, 4, 4, w2, h2, 1);
318	92.5k	mc_luma_dir(td, mc[4][b->filter][1],
319	92.5k	td->dst[0] + 4 * ls_y, ls_y,
320	92.5k	ref2->data[0], ref2->linesize[0], tref2,
321	92.5k	(row << 3) + 4, col << 3, &b->mv[2][1], 0, 4, 8, 8, 4, 4, w2, h2, 1);
322	92.5k	mc_luma_dir(td, mc[4][b->filter][1],
323	92.5k	td->dst[0] + 4 * ls_y + 4 * bytesperpixel, ls_y,
324	92.5k	ref2->data[0], ref2->linesize[0], tref2,
325	92.5k	(row << 3) + 4, (col << 3) + 4, &b->mv[3][1], 4, 4, 8, 8, 4, 4, w2, h2, 1);
326	92.5k	if (s->ss_v) {
327	38.1k	h2 = (h2 + 1) >> 1;
328	38.1k	if (s->ss_h) {
329	18.0k	w2 = (w2 + 1) >> 1;
330	18.0k	uvmv = ROUNDED_DIV_MVx4(b->mv[0][1], b->mv[1][1],
331	18.0k	b->mv[2][1], b->mv[3][1]);
332	18.0k	mc_chroma_dir(td, mc[4][b->filter][1],
333	18.0k	td->dst[1], td->dst[2], ls_uv,
334	18.0k	ref2->data[1], ref2->linesize[1],
335	18.0k	ref2->data[2], ref2->linesize[2], tref2,
336	18.0k	row << 2, col << 2,
337	18.0k	&uvmv, 0, 0, 4, 4, 4, 4, w2, h2, 1);
338	20.0k	} else {
339	20.0k	uvmv = ROUNDED_DIV_MVx2(b->mv[0][1], b->mv[2][1]);
340	20.0k	mc_chroma_dir(td, mc[4][b->filter][1],
341	20.0k	td->dst[1], td->dst[2], ls_uv,
342	20.0k	ref2->data[1], ref2->linesize[1],
343	20.0k	ref2->data[2], ref2->linesize[2], tref2,
344	20.0k	row << 2, col << 3,
345	20.0k	&uvmv, 0, 0, 8, 4, 4, 4, w2, h2, 1);
346	20.0k	uvmv = ROUNDED_DIV_MVx2(b->mv[1][1], b->mv[3][1]);
347	20.0k	mc_chroma_dir(td, mc[4][b->filter][1],
348	20.0k	td->dst[1] + 4 * bytesperpixel,
349	20.0k	td->dst[2] + 4 * bytesperpixel, ls_uv,
350	20.0k	ref2->data[1], ref2->linesize[1],
351	20.0k	ref2->data[2], ref2->linesize[2], tref2,
352	20.0k	row << 2, (col << 3) + 4,
353	20.0k	&uvmv, 4, 0, 8, 4, 4, 4, w2, h2, 1);
354	20.0k	}
355	54.4k	} else {
356	54.4k	if (s->ss_h) {
357	12.6k	w2 = (w2 + 1) >> 1;
358	12.6k	uvmv = ROUNDED_DIV_MVx2(b->mv[0][1], b->mv[1][1]);
359	12.6k	mc_chroma_dir(td, mc[4][b->filter][1],
360	12.6k	td->dst[1], td->dst[2], ls_uv,
361	12.6k	ref2->data[1], ref2->linesize[1],
362	12.6k	ref2->data[2], ref2->linesize[2], tref2,
363	12.6k	row << 3, col << 2,
364	12.6k	&uvmv, 0, 0, 4, 8, 4, 4, w2, h2, 1);
365		// BUG libvpx uses wrong block index for 4:2:2 bs=4x4
366		// bottom block
367		// https://code.google.com/p/webm/issues/detail?id=993
368	12.6k	uvmv = ROUNDED_DIV_MVx2(b->mv[1][1], b->mv[2][1]);
369	12.6k	mc_chroma_dir(td, mc[4][b->filter][1],
370	12.6k	td->dst[1] + 4 * ls_uv, td->dst[2] + 4 * ls_uv, ls_uv,
371	12.6k	ref2->data[1], ref2->linesize[1],
372	12.6k	ref2->data[2], ref2->linesize[2], tref2,
373	12.6k	(row << 3) + 4, col << 2,
374	12.6k	&uvmv, 0, 4, 4, 8, 4, 4, w2, h2, 1);
375	41.7k	} else {
376	41.7k	mc_chroma_dir(td, mc[4][b->filter][1],
377	41.7k	td->dst[1], td->dst[2], ls_uv,
378	41.7k	ref2->data[1], ref2->linesize[1],
379	41.7k	ref2->data[2], ref2->linesize[2], tref2,
380	41.7k	row << 3, col << 3,
381	41.7k	&b->mv[0][1], 0, 0, 8, 8, 4, 4, w2, h2, 1);
382	41.7k	mc_chroma_dir(td, mc[4][b->filter][1],
383	41.7k	td->dst[1] + 4 * bytesperpixel,
384	41.7k	td->dst[2] + 4 * bytesperpixel, ls_uv,
385	41.7k	ref2->data[1], ref2->linesize[1],
386	41.7k	ref2->data[2], ref2->linesize[2], tref2,
387	41.7k	row << 3, (col << 3) + 4,
388	41.7k	&b->mv[1][1], 4, 0, 8, 8, 4, 4, w2, h2, 1);
389	41.7k	mc_chroma_dir(td, mc[4][b->filter][1],
390	41.7k	td->dst[1] + 4 * ls_uv, td->dst[2] + 4 * ls_uv, ls_uv,
391	41.7k	ref2->data[1], ref2->linesize[1],
392	41.7k	ref2->data[2], ref2->linesize[2], tref2,
393	41.7k	(row << 3) + 4, col << 3,
394	41.7k	&b->mv[2][1], 0, 4, 8, 8, 4, 4, w2, h2, 1);
395	41.7k	mc_chroma_dir(td, mc[4][b->filter][1],
396	41.7k	td->dst[1] + 4 * ls_uv + 4 * bytesperpixel,
397	41.7k	td->dst[2] + 4 * ls_uv + 4 * bytesperpixel, ls_uv,
398	41.7k	ref2->data[1], ref2->linesize[1],
399	41.7k	ref2->data[2], ref2->linesize[2], tref2,
400	41.7k	(row << 3) + 4, (col << 3) + 4,
401	41.7k	&b->mv[3][1], 4, 4, 8, 8, 4, 4, w2, h2, 1);
402	41.7k	}
403	54.4k	}
404	92.5k	}
405	232k	}
406	1.63M	} else {
407	1.63M	int bwl = bwlog_tab[0][b->bs];
408	1.63M	int bw = ff_vp9_bwh_tab[0][b->bs][0] * 4;
409	1.63M	int bh = ff_vp9_bwh_tab[0][b->bs][1] * 4;
410	1.63M	int uvbw = ff_vp9_bwh_tab[s->ss_h][b->bs][0] * 4;
411	1.63M	int uvbh = ff_vp9_bwh_tab[s->ss_v][b->bs][1] * 4;
412
413	1.63M	mc_luma_dir(td, mc[bwl][b->filter][0], td->dst[0], ls_y,
414	1.63M	ref1->data[0], ref1->linesize[0], tref1,
415	1.63M	row << 3, col << 3, &b->mv[0][0], 0, 0, bw, bh, bw, bh, w1, h1, 0);
416	1.63M	w1 = (w1 + s->ss_h) >> s->ss_h;
417	1.63M	h1 = (h1 + s->ss_v) >> s->ss_v;
418	1.63M	mc_chroma_dir(td, mc[bwl + s->ss_h][b->filter][0],
419	1.63M	td->dst[1], td->dst[2], ls_uv,
420	1.63M	ref1->data[1], ref1->linesize[1],
421	1.63M	ref1->data[2], ref1->linesize[2], tref1,
422	1.63M	row << (3 - s->ss_v), col << (3 - s->ss_h),
423	1.63M	&b->mv[0][0], 0, 0, uvbw, uvbh, uvbw, uvbh, w1, h1, 0);
424
425	1.63M	if (b->comp) {
426	598k	mc_luma_dir(td, mc[bwl][b->filter][1], td->dst[0], ls_y,
427	598k	ref2->data[0], ref2->linesize[0], tref2,
428	598k	row << 3, col << 3, &b->mv[0][1], 0, 0, bw, bh, bw, bh, w2, h2, 1);
429	598k	w2 = (w2 + s->ss_h) >> s->ss_h;
430	598k	h2 = (h2 + s->ss_v) >> s->ss_v;
431	598k	mc_chroma_dir(td, mc[bwl + s->ss_h][b->filter][1],
432	598k	td->dst[1], td->dst[2], ls_uv,
433	598k	ref2->data[1], ref2->linesize[1],
434	598k	ref2->data[2], ref2->linesize[2], tref2,
435	598k	row << (3 - s->ss_v), col << (3 - s->ss_h),
436	598k	&b->mv[0][1], 0, 0, uvbw, uvbh, uvbw, uvbh, w2, h2, 1);
437	598k	}
438	1.63M	}
439	2.24M	}

vp9recon.c:inter_pred_16bpp

Line	Count	Source
31	6.61M	{
32	6.61M	static const uint8_t bwlog_tab[2][N_BS_SIZES] = {
33	6.61M	{ 0, 0, 1, 1, 1, 2, 2, 2, 3, 3, 3, 4, 4 },
34	6.61M	{ 1, 1, 2, 2, 2, 3, 3, 3, 4, 4, 4, 4, 4 },
35	6.61M	};
36	6.61M	const VP9Context *s = td->s;
37	6.61M	VP9Block *b = td->b;
38	6.61M	int row = td->row, col = td->col;
39	6.61M	const ProgressFrame tref1 = &s->s.refs[s->s.h.refidx[b->ref[0]]], tref2;
40	6.61M	const AVFrame ref1 = tref1->f, ref2;
41	6.61M	int w1 = ref1->width, h1 = ref1->height, w2, h2;
42	6.61M	ptrdiff_t ls_y = td->y_stride, ls_uv = td->uv_stride;
43	6.61M	int bytesperpixel = BYTES_PER_PIXEL;
44
45	6.61M	if (b->comp) {
46	4.47M	tref2 = &s->s.refs[s->s.h.refidx[b->ref[1]]];
47	4.47M	ref2 = tref2->f;
48	4.47M	w2 = ref2->width;
49	4.47M	h2 = ref2->height;
50	4.47M	}
51
52		// y inter pred
53	6.61M	if (b->bs > BS_8x8) {
54	2.30M	VP9mv uvmv;
55
56	2.30M	#if SCALED == 0
57	2.30M	if (b->bs == BS_8x4) {
58	873k	mc_luma_dir(td, mc[3][b->filter][0], td->dst[0], ls_y,
59	873k	ref1->data[0], ref1->linesize[0], tref1,
60	873k	row << 3, col << 3, &b->mv[0][0],,,,, 8, 4, w1, h1, 0);
61	873k	mc_luma_dir(td, mc[3][b->filter][0],
62	873k	td->dst[0] + 4 * ls_y, ls_y,
63	873k	ref1->data[0], ref1->linesize[0], tref1,
64	873k	(row << 3) + 4, col << 3, &b->mv[2][0],,,,, 8, 4, w1, h1, 0);
65	873k	w1 = (w1 + s->ss_h) >> s->ss_h;
66	873k	if (s->ss_v) {
67	268k	h1 = (h1 + 1) >> 1;
68	268k	uvmv = ROUNDED_DIV_MVx2(b->mv[0][0], b->mv[2][0]);
69	268k	mc_chroma_dir(td, mc[3 + s->ss_h][b->filter][0],
70	268k	td->dst[1], td->dst[2], ls_uv,
71	268k	ref1->data[1], ref1->linesize[1],
72	268k	ref1->data[2], ref1->linesize[2], tref1,
73	268k	row << 2, col << (3 - s->ss_h),
74	268k	&uvmv,,,,, 8 >> s->ss_h, 4, w1, h1, 0);
75	604k	} else {
76	604k	mc_chroma_dir(td, mc[3 + s->ss_h][b->filter][0],
77	604k	td->dst[1], td->dst[2], ls_uv,
78	604k	ref1->data[1], ref1->linesize[1],
79	604k	ref1->data[2], ref1->linesize[2], tref1,
80	604k	row << 3, col << (3 - s->ss_h),
81	604k	&b->mv[0][0],,,,, 8 >> s->ss_h, 4, w1, h1, 0);
82		// BUG for 4:2:2 bs=8x4, libvpx uses the wrong block index
83		// to get the motion vector for the bottom 4x4 block
84		// https://code.google.com/p/webm/issues/detail?id=993
85	604k	if (s->ss_h == 0) {
86	498k	uvmv = b->mv[2][0];
87	498k	} else {
88	106k	uvmv = ROUNDED_DIV_MVx2(b->mv[0][0], b->mv[2][0]);
89	106k	}
90	604k	mc_chroma_dir(td, mc[3 + s->ss_h][b->filter][0],
91	604k	td->dst[1] + 4 * ls_uv, td->dst[2] + 4 * ls_uv, ls_uv,
92	604k	ref1->data[1], ref1->linesize[1],
93	604k	ref1->data[2], ref1->linesize[2], tref1,
94	604k	(row << 3) + 4, col << (3 - s->ss_h),
95	604k	&uvmv,,,,, 8 >> s->ss_h, 4, w1, h1, 0);
96	604k	}
97
98	873k	if (b->comp) {
99	576k	mc_luma_dir(td, mc[3][b->filter][1], td->dst[0], ls_y,
100	576k	ref2->data[0], ref2->linesize[0], tref2,
101	576k	row << 3, col << 3, &b->mv[0][1],,,,, 8, 4, w2, h2, 1);
102	576k	mc_luma_dir(td, mc[3][b->filter][1],
103	576k	td->dst[0] + 4 * ls_y, ls_y,
104	576k	ref2->data[0], ref2->linesize[0], tref2,
105	576k	(row << 3) + 4, col << 3, &b->mv[2][1],,,,, 8, 4, w2, h2, 1);
106	576k	w2 = (w2 + s->ss_h) >> s->ss_h;
107	576k	if (s->ss_v) {
108	92.8k	h2 = (h2 + 1) >> 1;
109	92.8k	uvmv = ROUNDED_DIV_MVx2(b->mv[0][1], b->mv[2][1]);
110	92.8k	mc_chroma_dir(td, mc[3 + s->ss_h][b->filter][1],
111	92.8k	td->dst[1], td->dst[2], ls_uv,
112	92.8k	ref2->data[1], ref2->linesize[1],
113	92.8k	ref2->data[2], ref2->linesize[2], tref2,
114	92.8k	row << 2, col << (3 - s->ss_h),
115	92.8k	&uvmv,,,,, 8 >> s->ss_h, 4, w2, h2, 1);
116	483k	} else {
117	483k	mc_chroma_dir(td, mc[3 + s->ss_h][b->filter][1],
118	483k	td->dst[1], td->dst[2], ls_uv,
119	483k	ref2->data[1], ref2->linesize[1],
120	483k	ref2->data[2], ref2->linesize[2], tref2,
121	483k	row << 3, col << (3 - s->ss_h),
122	483k	&b->mv[0][1],,,,, 8 >> s->ss_h, 4, w2, h2, 1);
123		// BUG for 4:2:2 bs=8x4, libvpx uses the wrong block index
124		// to get the motion vector for the bottom 4x4 block
125		// https://code.google.com/p/webm/issues/detail?id=993
126	483k	if (s->ss_h == 0) {
127	435k	uvmv = b->mv[2][1];
128	435k	} else {
129	48.1k	uvmv = ROUNDED_DIV_MVx2(b->mv[0][1], b->mv[2][1]);
130	48.1k	}
131	483k	mc_chroma_dir(td, mc[3 + s->ss_h][b->filter][1],
132	483k	td->dst[1] + 4 * ls_uv, td->dst[2] + 4 * ls_uv, ls_uv,
133	483k	ref2->data[1], ref2->linesize[1],
134	483k	ref2->data[2], ref2->linesize[2], tref2,
135	483k	(row << 3) + 4, col << (3 - s->ss_h),
136	483k	&uvmv,,,,, 8 >> s->ss_h, 4, w2, h2, 1);
137	483k	}
138	576k	}
139	1.43M	} else if (b->bs == BS_4x8) {
140	479k	mc_luma_dir(td, mc[4][b->filter][0], td->dst[0], ls_y,
141	479k	ref1->data[0], ref1->linesize[0], tref1,
142	479k	row << 3, col << 3, &b->mv[0][0],,,,, 4, 8, w1, h1, 0);
143	479k	mc_luma_dir(td, mc[4][b->filter][0], td->dst[0] + 4 * bytesperpixel, ls_y,
144	479k	ref1->data[0], ref1->linesize[0], tref1,
145	479k	row << 3, (col << 3) + 4, &b->mv[1][0],,,,, 4, 8, w1, h1, 0);
146	479k	h1 = (h1 + s->ss_v) >> s->ss_v;
147	479k	if (s->ss_h) {
148	96.8k	w1 = (w1 + 1) >> 1;
149	96.8k	uvmv = ROUNDED_DIV_MVx2(b->mv[0][0], b->mv[1][0]);
150	96.8k	mc_chroma_dir(td, mc[4][b->filter][0],
151	96.8k	td->dst[1], td->dst[2], ls_uv,
152	96.8k	ref1->data[1], ref1->linesize[1],
153	96.8k	ref1->data[2], ref1->linesize[2], tref1,
154	96.8k	row << (3 - s->ss_v), col << 2,
155	96.8k	&uvmv,,,,, 4, 8 >> s->ss_v, w1, h1, 0);
156	382k	} else {
157	382k	mc_chroma_dir(td, mc[4][b->filter][0],
158	382k	td->dst[1], td->dst[2], ls_uv,
159	382k	ref1->data[1], ref1->linesize[1],
160	382k	ref1->data[2], ref1->linesize[2], tref1,
161	382k	row << (3 - s->ss_v), col << 3,
162	382k	&b->mv[0][0],,,,, 4, 8 >> s->ss_v, w1, h1, 0);
163	382k	mc_chroma_dir(td, mc[4][b->filter][0],
164	382k	td->dst[1] + 4 * bytesperpixel,
165	382k	td->dst[2] + 4 * bytesperpixel, ls_uv,
166	382k	ref1->data[1], ref1->linesize[1],
167	382k	ref1->data[2], ref1->linesize[2], tref1,
168	382k	row << (3 - s->ss_v), (col << 3) + 4,
169	382k	&b->mv[1][0],,,,, 4, 8 >> s->ss_v, w1, h1, 0);
170	382k	}
171
172	479k	if (b->comp) {
173	335k	mc_luma_dir(td, mc[4][b->filter][1], td->dst[0], ls_y,
174	335k	ref2->data[0], ref2->linesize[0], tref2,
175	335k	row << 3, col << 3, &b->mv[0][1],,,,, 4, 8, w2, h2, 1);
176	335k	mc_luma_dir(td, mc[4][b->filter][1], td->dst[0] + 4 * bytesperpixel, ls_y,
177	335k	ref2->data[0], ref2->linesize[0], tref2,
178	335k	row << 3, (col << 3) + 4, &b->mv[1][1],,,,, 4, 8, w2, h2, 1);
179	335k	h2 = (h2 + s->ss_v) >> s->ss_v;
180	335k	if (s->ss_h) {
181	55.8k	w2 = (w2 + 1) >> 1;
182	55.8k	uvmv = ROUNDED_DIV_MVx2(b->mv[0][1], b->mv[1][1]);
183	55.8k	mc_chroma_dir(td, mc[4][b->filter][1],
184	55.8k	td->dst[1], td->dst[2], ls_uv,
185	55.8k	ref2->data[1], ref2->linesize[1],
186	55.8k	ref2->data[2], ref2->linesize[2], tref2,
187	55.8k	row << (3 - s->ss_v), col << 2,
188	55.8k	&uvmv,,,,, 4, 8 >> s->ss_v, w2, h2, 1);
189	279k	} else {
190	279k	mc_chroma_dir(td, mc[4][b->filter][1],
191	279k	td->dst[1], td->dst[2], ls_uv,
192	279k	ref2->data[1], ref2->linesize[1],
193	279k	ref2->data[2], ref2->linesize[2], tref2,
194	279k	row << (3 - s->ss_v), col << 3,
195	279k	&b->mv[0][1],,,,, 4, 8 >> s->ss_v, w2, h2, 1);
196	279k	mc_chroma_dir(td, mc[4][b->filter][1],
197	279k	td->dst[1] + 4 * bytesperpixel,
198	279k	td->dst[2] + 4 * bytesperpixel, ls_uv,
199	279k	ref2->data[1], ref2->linesize[1],
200	279k	ref2->data[2], ref2->linesize[2], tref2,
201	279k	row << (3 - s->ss_v), (col << 3) + 4,
202	279k	&b->mv[1][1],,,,, 4, 8 >> s->ss_v, w2, h2, 1);
203	279k	}
204	335k	}
205	479k	} else
206	953k	#endif
207	953k	{
208	953k	#if SCALED == 0
209	953k	av_assert2(b->bs == BS_4x4);
210	953k	#endif
211
212		// FIXME if two horizontally adjacent blocks have the same MV,
213		// do a w8 instead of a w4 call
214	953k	mc_luma_dir(td, mc[4][b->filter][0], td->dst[0], ls_y,
215	953k	ref1->data[0], ref1->linesize[0], tref1,
216	953k	row << 3, col << 3, &b->mv[0][0],
217	953k	0, 0, 8, 8, 4, 4, w1, h1, 0);
218	953k	mc_luma_dir(td, mc[4][b->filter][0], td->dst[0] + 4 * bytesperpixel, ls_y,
219	953k	ref1->data[0], ref1->linesize[0], tref1,
220	953k	row << 3, (col << 3) + 4, &b->mv[1][0],
221	953k	4, 0, 8, 8, 4, 4, w1, h1, 0);
222	953k	mc_luma_dir(td, mc[4][b->filter][0],
223	953k	td->dst[0] + 4 * ls_y, ls_y,
224	953k	ref1->data[0], ref1->linesize[0], tref1,
225	953k	(row << 3) + 4, col << 3, &b->mv[2][0],
226	953k	0, 4, 8, 8, 4, 4, w1, h1, 0);
227	953k	mc_luma_dir(td, mc[4][b->filter][0],
228	953k	td->dst[0] + 4 * ls_y + 4 * bytesperpixel, ls_y,
229	953k	ref1->data[0], ref1->linesize[0], tref1,
230	953k	(row << 3) + 4, (col << 3) + 4, &b->mv[3][0],
231	953k	4, 4, 8, 8, 4, 4, w1, h1, 0);
232	953k	if (s->ss_v) {
233	250k	h1 = (h1 + 1) >> 1;
234	250k	if (s->ss_h) {
235	92.2k	w1 = (w1 + 1) >> 1;
236	92.2k	uvmv = ROUNDED_DIV_MVx4(b->mv[0][0], b->mv[1][0],
237	92.2k	b->mv[2][0], b->mv[3][0]);
238	92.2k	mc_chroma_dir(td, mc[4][b->filter][0],
239	92.2k	td->dst[1], td->dst[2], ls_uv,
240	92.2k	ref1->data[1], ref1->linesize[1],
241	92.2k	ref1->data[2], ref1->linesize[2], tref1,
242	92.2k	row << 2, col << 2,
243	92.2k	&uvmv, 0, 0, 4, 4, 4, 4, w1, h1, 0);
244	158k	} else {
245	158k	uvmv = ROUNDED_DIV_MVx2(b->mv[0][0], b->mv[2][0]);
246	158k	mc_chroma_dir(td, mc[4][b->filter][0],
247	158k	td->dst[1], td->dst[2], ls_uv,
248	158k	ref1->data[1], ref1->linesize[1],
249	158k	ref1->data[2], ref1->linesize[2], tref1,
250	158k	row << 2, col << 3,
251	158k	&uvmv, 0, 0, 8, 4, 4, 4, w1, h1, 0);
252	158k	uvmv = ROUNDED_DIV_MVx2(b->mv[1][0], b->mv[3][0]);
253	158k	mc_chroma_dir(td, mc[4][b->filter][0],
254	158k	td->dst[1] + 4 * bytesperpixel,
255	158k	td->dst[2] + 4 * bytesperpixel, ls_uv,
256	158k	ref1->data[1], ref1->linesize[1],
257	158k	ref1->data[2], ref1->linesize[2], tref1,
258	158k	row << 2, (col << 3) + 4,
259	158k	&uvmv, 4, 0, 8, 4, 4, 4, w1, h1, 0);
260	158k	}
261	702k	} else {
262	702k	if (s->ss_h) {
263	67.0k	w1 = (w1 + 1) >> 1;
264	67.0k	uvmv = ROUNDED_DIV_MVx2(b->mv[0][0], b->mv[1][0]);
265	67.0k	mc_chroma_dir(td, mc[4][b->filter][0],
266	67.0k	td->dst[1], td->dst[2], ls_uv,
267	67.0k	ref1->data[1], ref1->linesize[1],
268	67.0k	ref1->data[2], ref1->linesize[2], tref1,
269	67.0k	row << 3, col << 2,
270	67.0k	&uvmv, 0, 0, 4, 8, 4, 4, w1, h1, 0);
271		// BUG libvpx uses wrong block index for 4:2:2 bs=4x4
272		// bottom block
273		// https://code.google.com/p/webm/issues/detail?id=993
274	67.0k	uvmv = ROUNDED_DIV_MVx2(b->mv[1][0], b->mv[2][0]);
275	67.0k	mc_chroma_dir(td, mc[4][b->filter][0],
276	67.0k	td->dst[1] + 4 * ls_uv, td->dst[2] + 4 * ls_uv, ls_uv,
277	67.0k	ref1->data[1], ref1->linesize[1],
278	67.0k	ref1->data[2], ref1->linesize[2], tref1,
279	67.0k	(row << 3) + 4, col << 2,
280	67.0k	&uvmv, 0, 4, 4, 8, 4, 4, w1, h1, 0);
281	635k	} else {
282	635k	mc_chroma_dir(td, mc[4][b->filter][0],
283	635k	td->dst[1], td->dst[2], ls_uv,
284	635k	ref1->data[1], ref1->linesize[1],
285	635k	ref1->data[2], ref1->linesize[2], tref1,
286	635k	row << 3, col << 3,
287	635k	&b->mv[0][0], 0, 0, 8, 8, 4, 4, w1, h1, 0);
288	635k	mc_chroma_dir(td, mc[4][b->filter][0],
289	635k	td->dst[1] + 4 * bytesperpixel,
290	635k	td->dst[2] + 4 * bytesperpixel, ls_uv,
291	635k	ref1->data[1], ref1->linesize[1],
292	635k	ref1->data[2], ref1->linesize[2], tref1,
293	635k	row << 3, (col << 3) + 4,
294	635k	&b->mv[1][0], 4, 0, 8, 8, 4, 4, w1, h1, 0);
295	635k	mc_chroma_dir(td, mc[4][b->filter][0],
296	635k	td->dst[1] + 4 * ls_uv, td->dst[2] + 4 * ls_uv, ls_uv,
297	635k	ref1->data[1], ref1->linesize[1],
298	635k	ref1->data[2], ref1->linesize[2], tref1,
299	635k	(row << 3) + 4, col << 3,
300	635k	&b->mv[2][0], 0, 4, 8, 8, 4, 4, w1, h1, 0);
301	635k	mc_chroma_dir(td, mc[4][b->filter][0],
302	635k	td->dst[1] + 4 * ls_uv + 4 * bytesperpixel,
303	635k	td->dst[2] + 4 * ls_uv + 4 * bytesperpixel, ls_uv,
304	635k	ref1->data[1], ref1->linesize[1],
305	635k	ref1->data[2], ref1->linesize[2], tref1,
306	635k	(row << 3) + 4, (col << 3) + 4,
307	635k	&b->mv[3][0], 4, 4, 8, 8, 4, 4, w1, h1, 0);
308	635k	}
309	702k	}
310
311	953k	if (b->comp) {
312	723k	mc_luma_dir(td, mc[4][b->filter][1], td->dst[0], ls_y,
313	723k	ref2->data[0], ref2->linesize[0], tref2,
314	723k	row << 3, col << 3, &b->mv[0][1], 0, 0, 8, 8, 4, 4, w2, h2, 1);
315	723k	mc_luma_dir(td, mc[4][b->filter][1], td->dst[0] + 4 * bytesperpixel, ls_y,
316	723k	ref2->data[0], ref2->linesize[0], tref2,
317	723k	row << 3, (col << 3) + 4, &b->mv[1][1], 4, 0, 8, 8, 4, 4, w2, h2, 1);
318	723k	mc_luma_dir(td, mc[4][b->filter][1],
319	723k	td->dst[0] + 4 * ls_y, ls_y,
320	723k	ref2->data[0], ref2->linesize[0], tref2,
321	723k	(row << 3) + 4, col << 3, &b->mv[2][1], 0, 4, 8, 8, 4, 4, w2, h2, 1);
322	723k	mc_luma_dir(td, mc[4][b->filter][1],
323	723k	td->dst[0] + 4 * ls_y + 4 * bytesperpixel, ls_y,
324	723k	ref2->data[0], ref2->linesize[0], tref2,
325	723k	(row << 3) + 4, (col << 3) + 4, &b->mv[3][1], 4, 4, 8, 8, 4, 4, w2, h2, 1);
326	723k	if (s->ss_v) {
327	123k	h2 = (h2 + 1) >> 1;
328	123k	if (s->ss_h) {
329	78.1k	w2 = (w2 + 1) >> 1;
330	78.1k	uvmv = ROUNDED_DIV_MVx4(b->mv[0][1], b->mv[1][1],
331	78.1k	b->mv[2][1], b->mv[3][1]);
332	78.1k	mc_chroma_dir(td, mc[4][b->filter][1],
333	78.1k	td->dst[1], td->dst[2], ls_uv,
334	78.1k	ref2->data[1], ref2->linesize[1],
335	78.1k	ref2->data[2], ref2->linesize[2], tref2,
336	78.1k	row << 2, col << 2,
337	78.1k	&uvmv, 0, 0, 4, 4, 4, 4, w2, h2, 1);
338	78.1k	} else {
339	45.3k	uvmv = ROUNDED_DIV_MVx2(b->mv[0][1], b->mv[2][1]);
340	45.3k	mc_chroma_dir(td, mc[4][b->filter][1],
341	45.3k	td->dst[1], td->dst[2], ls_uv,
342	45.3k	ref2->data[1], ref2->linesize[1],
343	45.3k	ref2->data[2], ref2->linesize[2], tref2,
344	45.3k	row << 2, col << 3,
345	45.3k	&uvmv, 0, 0, 8, 4, 4, 4, w2, h2, 1);
346	45.3k	uvmv = ROUNDED_DIV_MVx2(b->mv[1][1], b->mv[3][1]);
347	45.3k	mc_chroma_dir(td, mc[4][b->filter][1],
348	45.3k	td->dst[1] + 4 * bytesperpixel,
349	45.3k	td->dst[2] + 4 * bytesperpixel, ls_uv,
350	45.3k	ref2->data[1], ref2->linesize[1],
351	45.3k	ref2->data[2], ref2->linesize[2], tref2,
352	45.3k	row << 2, (col << 3) + 4,
353	45.3k	&uvmv, 4, 0, 8, 4, 4, 4, w2, h2, 1);
354	45.3k	}
355	600k	} else {
356	600k	if (s->ss_h) {
357	37.5k	w2 = (w2 + 1) >> 1;
358	37.5k	uvmv = ROUNDED_DIV_MVx2(b->mv[0][1], b->mv[1][1]);
359	37.5k	mc_chroma_dir(td, mc[4][b->filter][1],
360	37.5k	td->dst[1], td->dst[2], ls_uv,
361	37.5k	ref2->data[1], ref2->linesize[1],
362	37.5k	ref2->data[2], ref2->linesize[2], tref2,
363	37.5k	row << 3, col << 2,
364	37.5k	&uvmv, 0, 0, 4, 8, 4, 4, w2, h2, 1);
365		// BUG libvpx uses wrong block index for 4:2:2 bs=4x4
366		// bottom block
367		// https://code.google.com/p/webm/issues/detail?id=993
368	37.5k	uvmv = ROUNDED_DIV_MVx2(b->mv[1][1], b->mv[2][1]);
369	37.5k	mc_chroma_dir(td, mc[4][b->filter][1],
370	37.5k	td->dst[1] + 4 * ls_uv, td->dst[2] + 4 * ls_uv, ls_uv,
371	37.5k	ref2->data[1], ref2->linesize[1],
372	37.5k	ref2->data[2], ref2->linesize[2], tref2,
373	37.5k	(row << 3) + 4, col << 2,
374	37.5k	&uvmv, 0, 4, 4, 8, 4, 4, w2, h2, 1);
375	562k	} else {
376	562k	mc_chroma_dir(td, mc[4][b->filter][1],
377	562k	td->dst[1], td->dst[2], ls_uv,
378	562k	ref2->data[1], ref2->linesize[1],
379	562k	ref2->data[2], ref2->linesize[2], tref2,
380	562k	row << 3, col << 3,
381	562k	&b->mv[0][1], 0, 0, 8, 8, 4, 4, w2, h2, 1);
382	562k	mc_chroma_dir(td, mc[4][b->filter][1],
383	562k	td->dst[1] + 4 * bytesperpixel,
384	562k	td->dst[2] + 4 * bytesperpixel, ls_uv,
385	562k	ref2->data[1], ref2->linesize[1],
386	562k	ref2->data[2], ref2->linesize[2], tref2,
387	562k	row << 3, (col << 3) + 4,
388	562k	&b->mv[1][1], 4, 0, 8, 8, 4, 4, w2, h2, 1);
389	562k	mc_chroma_dir(td, mc[4][b->filter][1],
390	562k	td->dst[1] + 4 * ls_uv, td->dst[2] + 4 * ls_uv, ls_uv,
391	562k	ref2->data[1], ref2->linesize[1],
392	562k	ref2->data[2], ref2->linesize[2], tref2,
393	562k	(row << 3) + 4, col << 3,
394	562k	&b->mv[2][1], 0, 4, 8, 8, 4, 4, w2, h2, 1);
395	562k	mc_chroma_dir(td, mc[4][b->filter][1],
396	562k	td->dst[1] + 4 * ls_uv + 4 * bytesperpixel,
397	562k	td->dst[2] + 4 * ls_uv + 4 * bytesperpixel, ls_uv,
398	562k	ref2->data[1], ref2->linesize[1],
399	562k	ref2->data[2], ref2->linesize[2], tref2,
400	562k	(row << 3) + 4, (col << 3) + 4,
401	562k	&b->mv[3][1], 4, 4, 8, 8, 4, 4, w2, h2, 1);
402	562k	}
403	600k	}
404	723k	}
405	953k	}
406	4.31M	} else {
407	4.31M	int bwl = bwlog_tab[0][b->bs];
408	4.31M	int bw = ff_vp9_bwh_tab[0][b->bs][0] * 4;
409	4.31M	int bh = ff_vp9_bwh_tab[0][b->bs][1] * 4;
410	4.31M	int uvbw = ff_vp9_bwh_tab[s->ss_h][b->bs][0] * 4;
411	4.31M	int uvbh = ff_vp9_bwh_tab[s->ss_v][b->bs][1] * 4;
412
413	4.31M	mc_luma_dir(td, mc[bwl][b->filter][0], td->dst[0], ls_y,
414	4.31M	ref1->data[0], ref1->linesize[0], tref1,
415	4.31M	row << 3, col << 3, &b->mv[0][0], 0, 0, bw, bh, bw, bh, w1, h1, 0);
416	4.31M	w1 = (w1 + s->ss_h) >> s->ss_h;
417	4.31M	h1 = (h1 + s->ss_v) >> s->ss_v;
418	4.31M	mc_chroma_dir(td, mc[bwl + s->ss_h][b->filter][0],
419	4.31M	td->dst[1], td->dst[2], ls_uv,
420	4.31M	ref1->data[1], ref1->linesize[1],
421	4.31M	ref1->data[2], ref1->linesize[2], tref1,
422	4.31M	row << (3 - s->ss_v), col << (3 - s->ss_h),
423	4.31M	&b->mv[0][0], 0, 0, uvbw, uvbh, uvbw, uvbh, w1, h1, 0);
424
425	4.31M	if (b->comp) {
426	2.83M	mc_luma_dir(td, mc[bwl][b->filter][1], td->dst[0], ls_y,
427	2.83M	ref2->data[0], ref2->linesize[0], tref2,
428	2.83M	row << 3, col << 3, &b->mv[0][1], 0, 0, bw, bh, bw, bh, w2, h2, 1);
429	2.83M	w2 = (w2 + s->ss_h) >> s->ss_h;
430	2.83M	h2 = (h2 + s->ss_v) >> s->ss_v;
431	2.83M	mc_chroma_dir(td, mc[bwl + s->ss_h][b->filter][1],
432	2.83M	td->dst[1], td->dst[2], ls_uv,
433	2.83M	ref2->data[1], ref2->linesize[1],
434	2.83M	ref2->data[2], ref2->linesize[2], tref2,
435	2.83M	row << (3 - s->ss_v), col << (3 - s->ss_h),
436	2.83M	&b->mv[0][1], 0, 0, uvbw, uvbh, uvbw, uvbh, w2, h2, 1);
437	2.83M	}
438	4.31M	}
439	6.61M	}

Coverage Report

Created: 2025-11-16 07:20