/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) \

3.03M

    (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) \

458k

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

458k

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

static void FN(inter_pred)(VP9TileData *td)

6.91M

6.91M

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

6.91M

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

6.91M

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

6.91M

};

6.91M

    const VP9Context *s = td->s;

6.91M

    VP9Block *b = td->b;

6.91M

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

6.91M

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

6.91M

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

6.91M

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

6.91M

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

6.91M

    int bytesperpixel = BYTES_PER_PIXEL;

6.91M

    if (b->comp) {

4.52M

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

4.52M

        ref2 = tref2->f;

4.52M

        w2 = ref2->width;

4.52M

        h2 = ref2->height;

4.52M

    // y inter pred

6.91M

    if (b->bs > BS_8x8) {

2.28M

        VP9mv uvmv;

#if SCALED == 0

1.29M

        if (b->bs == BS_8x4) {

412k

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

412k

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

412k

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

412k

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

412k

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

412k

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

412k

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

412k

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

412k

            if (s->ss_v) {

143k

                h1 = (h1 + 1) >> 1;

143k

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

143k

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

143k

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

143k

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

143k

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

143k

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

143k

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

268k

            } else {

268k

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

268k

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

268k

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

268k

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

268k

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

268k

                              &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

268k

                if (s->ss_h == 0) {

186k

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

186k

                } else {

81.7k

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

81.7k

268k

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

268k

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

268k

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

268k

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

268k

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

268k

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

268k

412k

            if (b->comp) {

234k

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

234k

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

234k

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

234k

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

234k

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

234k

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

234k

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

234k

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

234k

                if (s->ss_v) {

94.6k

                    h2 = (h2 + 1) >> 1;

94.6k

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

94.6k

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

94.6k

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

94.6k

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

94.6k

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

94.6k

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

94.6k

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

139k

                } else {

139k

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

139k

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

139k

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

139k

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

139k

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

139k

                                  &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

139k

                    if (s->ss_h == 0) {

89.0k

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

89.0k

                    } else {

50.7k

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

50.7k

139k

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

139k

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

139k

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

139k

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

139k

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

139k

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

139k

234k

882k

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

239k

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

239k

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

239k

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

239k

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

239k

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

239k

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

239k

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

239k

            if (s->ss_h) {

70.4k

                w1 = (w1 + 1) >> 1;

70.4k

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

70.4k

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

70.4k

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

70.4k

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

70.4k

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

70.4k

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

70.4k

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

169k

            } else {

169k

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

169k

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

169k

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

169k

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

169k

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

169k

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

169k

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

169k

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

169k

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

169k

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

169k

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

169k

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

169k

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

169k

239k

            if (b->comp) {

133k

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

133k

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

133k

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

133k

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

133k

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

133k

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

133k

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

133k

                if (s->ss_h) {

33.6k

                    w2 = (w2 + 1) >> 1;

33.6k

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

33.6k

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

33.6k

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

33.6k

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

33.6k

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

33.6k

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

33.6k

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

99.3k

                } else {

99.3k

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

99.3k

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

99.3k

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

99.3k

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

99.3k

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

99.3k

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

99.3k

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

99.3k

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

99.3k

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

99.3k

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

99.3k

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

99.3k

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

99.3k

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

99.3k

133k

239k

        } else

642k

#endif

642k

#if SCALED == 0

642k

            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

1.63M

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

642k

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

642k

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

642k

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

1.63M

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

642k

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

642k

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

642k

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

1.63M

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

642k

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

642k

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

642k

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

642k

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

1.63M

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

642k

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

642k

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

642k

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

642k

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

1.63M

            if (s->ss_v) {

663k

                h1 = (h1 + 1) >> 1;

663k

                if (s->ss_h) {

262k

                    w1 = (w1 + 1) >> 1;

262k

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

262k

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

262k

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

262k

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

262k

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

262k

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

262k

                                  row << 2, col << 2,

262k

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

401k

                } else {

401k

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

401k

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

401k

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

401k

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

401k

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

401k

                                  row << 2, col << 3,

401k

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

401k

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

401k

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

401k

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

401k

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

401k

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

401k

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

401k

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

401k

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

401k

968k

            } else {

968k

                if (s->ss_h) {

342k

                    w1 = (w1 + 1) >> 1;

342k

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

342k

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

342k

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

342k

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

342k

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

342k

                                  row << 3, col << 2,

342k

                                  &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

342k

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

342k

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

342k

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

342k

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

342k

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

342k

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

342k

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

626k

                } else {

626k

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

626k

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

626k

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

626k

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

626k

                                  row << 3, col << 3,

626k

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

626k

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

626k

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

626k

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

626k

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

626k

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

626k

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

626k

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

626k

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

626k

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

626k

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

626k

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

626k

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

626k

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

626k

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

626k

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

626k

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

626k

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

626k

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

626k

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

626k

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

626k

968k

1.63M

            if (b->comp) {

1.18M

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

1.18M

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

1.18M

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

1.18M

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

1.18M

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

1.18M

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

1.18M

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

1.18M

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

1.18M

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

1.18M

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

1.18M

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

1.18M

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

1.18M

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

1.18M

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

1.18M

                if (s->ss_v) {

505k

                    h2 = (h2 + 1) >> 1;

505k

                    if (s->ss_h) {

196k

                        w2 = (w2 + 1) >> 1;

196k

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

196k

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

196k

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

196k

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

196k

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

196k

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

196k

                                      row << 2, col << 2,

196k

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

309k

                    } else {

309k

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

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 << 2, col << 3,

309k

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

309k

                        uvmv = ROUNDED_DIV_MVx2(b->mv[1][1], b->mv[3][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 << 2, (col << 3) + 4,

309k

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

309k

676k

                } else {

676k

                    if (s->ss_h) {

226k

                        w2 = (w2 + 1) >> 1;

226k

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

226k

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

226k

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

226k

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

226k

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

226k

                                      row << 3, col << 2,

226k

                                      &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

226k

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

226k

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

226k

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

226k

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

226k

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

226k

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

226k

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

450k

                    } else {

450k

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

450k

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

450k

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

450k

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

450k

                                      row << 3, col << 3,

450k

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

450k

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

450k

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

450k

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

450k

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

450k

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

450k

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

450k

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

450k

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

450k

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

450k

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

450k

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

450k

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

450k

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

450k

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

450k

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

450k

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

450k

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

450k

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

450k

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

450k

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

450k

676k

1.18M

642k

4.63M

    } else {

4.63M

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

4.63M

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

4.63M

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

4.63M

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

4.63M

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

4.63M

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

4.63M

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

4.63M

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

4.63M

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

4.63M

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

4.63M

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

4.63M

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

4.63M

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

4.63M

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

4.63M

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

4.63M

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

4.63M

        if (b->comp) {

2.97M

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

2.97M

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

2.97M

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

2.97M

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

2.97M

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

2.97M

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

2.97M

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

2.97M

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

2.97M

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

2.97M

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

2.97M

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

2.97M

4.63M

6.91M

vp9recon.c:inter_pred_scaled_8bpp

Line	Count	Source
31	1.36M	{
32	1.36M	static const uint8_t bwlog_tab[2][N_BS_SIZES] = {
33	1.36M	{ 0, 0, 1, 1, 1, 2, 2, 2, 3, 3, 3, 4, 4 },
34	1.36M	{ 1, 1, 2, 2, 2, 3, 3, 3, 4, 4, 4, 4, 4 },
35	1.36M	};
36	1.36M	const VP9Context *s = td->s;
37	1.36M	VP9Block *b = td->b;
38	1.36M	int row = td->row, col = td->col;
39	1.36M	const ProgressFrame tref1 = &s->s.refs[s->s.h.refidx[b->ref[0]]], tref2;
40	1.36M	const AVFrame ref1 = tref1->f, ref2;
41	1.36M	int w1 = ref1->width, h1 = ref1->height, w2, h2;
42	1.36M	ptrdiff_t ls_y = td->y_stride, ls_uv = td->uv_stride;
43	1.36M	int bytesperpixel = BYTES_PER_PIXEL;
44
45	1.36M	if (b->comp) {
46	1.12M	tref2 = &s->s.refs[s->s.h.refidx[b->ref[1]]];
47	1.12M	ref2 = tref2->f;
48	1.12M	w2 = ref2->width;
49	1.12M	h2 = ref2->height;
50	1.12M	}
51
52		// y inter pred
53	1.36M	if (b->bs > BS_8x8) {
54	507k	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	507k	{
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	507k	mc_luma_dir(td, mc[4][b->filter][0], td->dst[0], ls_y,
215	507k	ref1->data[0], ref1->linesize[0], tref1,
216	507k	row << 3, col << 3, &b->mv[0][0],
217	507k	0, 0, 8, 8, 4, 4, w1, h1, 0);
218	507k	mc_luma_dir(td, mc[4][b->filter][0], td->dst[0] + 4 * bytesperpixel, ls_y,
219	507k	ref1->data[0], ref1->linesize[0], tref1,
220	507k	row << 3, (col << 3) + 4, &b->mv[1][0],
221	507k	4, 0, 8, 8, 4, 4, w1, h1, 0);
222	507k	mc_luma_dir(td, mc[4][b->filter][0],
223	507k	td->dst[0] + 4 * ls_y, ls_y,
224	507k	ref1->data[0], ref1->linesize[0], tref1,
225	507k	(row << 3) + 4, col << 3, &b->mv[2][0],
226	507k	0, 4, 8, 8, 4, 4, w1, h1, 0);
227	507k	mc_luma_dir(td, mc[4][b->filter][0],
228	507k	td->dst[0] + 4 * ls_y + 4 * bytesperpixel, ls_y,
229	507k	ref1->data[0], ref1->linesize[0], tref1,
230	507k	(row << 3) + 4, (col << 3) + 4, &b->mv[3][0],
231	507k	4, 4, 8, 8, 4, 4, w1, h1, 0);
232	507k	if (s->ss_v) {
233	183k	h1 = (h1 + 1) >> 1;
234	183k	if (s->ss_h) {
235	108k	w1 = (w1 + 1) >> 1;
236	108k	uvmv = ROUNDED_DIV_MVx4(b->mv[0][0], b->mv[1][0],
237	108k	b->mv[2][0], b->mv[3][0]);
238	108k	mc_chroma_dir(td, mc[4][b->filter][0],
239	108k	td->dst[1], td->dst[2], ls_uv,
240	108k	ref1->data[1], ref1->linesize[1],
241	108k	ref1->data[2], ref1->linesize[2], tref1,
242	108k	row << 2, col << 2,
243	108k	&uvmv, 0, 0, 4, 4, 4, 4, w1, h1, 0);
244	108k	} else {
245	74.9k	uvmv = ROUNDED_DIV_MVx2(b->mv[0][0], b->mv[2][0]);
246	74.9k	mc_chroma_dir(td, mc[4][b->filter][0],
247	74.9k	td->dst[1], td->dst[2], ls_uv,
248	74.9k	ref1->data[1], ref1->linesize[1],
249	74.9k	ref1->data[2], ref1->linesize[2], tref1,
250	74.9k	row << 2, col << 3,
251	74.9k	&uvmv, 0, 0, 8, 4, 4, 4, w1, h1, 0);
252	74.9k	uvmv = ROUNDED_DIV_MVx2(b->mv[1][0], b->mv[3][0]);
253	74.9k	mc_chroma_dir(td, mc[4][b->filter][0],
254	74.9k	td->dst[1] + 4 * bytesperpixel,
255	74.9k	td->dst[2] + 4 * bytesperpixel, ls_uv,
256	74.9k	ref1->data[1], ref1->linesize[1],
257	74.9k	ref1->data[2], ref1->linesize[2], tref1,
258	74.9k	row << 2, (col << 3) + 4,
259	74.9k	&uvmv, 4, 0, 8, 4, 4, 4, w1, h1, 0);
260	74.9k	}
261	323k	} else {
262	323k	if (s->ss_h) {
263	65.2k	w1 = (w1 + 1) >> 1;
264	65.2k	uvmv = ROUNDED_DIV_MVx2(b->mv[0][0], b->mv[1][0]);
265	65.2k	mc_chroma_dir(td, mc[4][b->filter][0],
266	65.2k	td->dst[1], td->dst[2], ls_uv,
267	65.2k	ref1->data[1], ref1->linesize[1],
268	65.2k	ref1->data[2], ref1->linesize[2], tref1,
269	65.2k	row << 3, col << 2,
270	65.2k	&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	65.2k	uvmv = ROUNDED_DIV_MVx2(b->mv[1][0], b->mv[2][0]);
275	65.2k	mc_chroma_dir(td, mc[4][b->filter][0],
276	65.2k	td->dst[1] + 4 * ls_uv, td->dst[2] + 4 * ls_uv, ls_uv,
277	65.2k	ref1->data[1], ref1->linesize[1],
278	65.2k	ref1->data[2], ref1->linesize[2], tref1,
279	65.2k	(row << 3) + 4, col << 2,
280	65.2k	&uvmv, 0, 4, 4, 8, 4, 4, w1, h1, 0);
281	258k	} else {
282	258k	mc_chroma_dir(td, mc[4][b->filter][0],
283	258k	td->dst[1], td->dst[2], ls_uv,
284	258k	ref1->data[1], ref1->linesize[1],
285	258k	ref1->data[2], ref1->linesize[2], tref1,
286	258k	row << 3, col << 3,
287	258k	&b->mv[0][0], 0, 0, 8, 8, 4, 4, w1, h1, 0);
288	258k	mc_chroma_dir(td, mc[4][b->filter][0],
289	258k	td->dst[1] + 4 * bytesperpixel,
290	258k	td->dst[2] + 4 * bytesperpixel, ls_uv,
291	258k	ref1->data[1], ref1->linesize[1],
292	258k	ref1->data[2], ref1->linesize[2], tref1,
293	258k	row << 3, (col << 3) + 4,
294	258k	&b->mv[1][0], 4, 0, 8, 8, 4, 4, w1, h1, 0);
295	258k	mc_chroma_dir(td, mc[4][b->filter][0],
296	258k	td->dst[1] + 4 * ls_uv, td->dst[2] + 4 * ls_uv, ls_uv,
297	258k	ref1->data[1], ref1->linesize[1],
298	258k	ref1->data[2], ref1->linesize[2], tref1,
299	258k	(row << 3) + 4, col << 3,
300	258k	&b->mv[2][0], 0, 4, 8, 8, 4, 4, w1, h1, 0);
301	258k	mc_chroma_dir(td, mc[4][b->filter][0],
302	258k	td->dst[1] + 4 * ls_uv + 4 * bytesperpixel,
303	258k	td->dst[2] + 4 * ls_uv + 4 * bytesperpixel, ls_uv,
304	258k	ref1->data[1], ref1->linesize[1],
305	258k	ref1->data[2], ref1->linesize[2], tref1,
306	258k	(row << 3) + 4, (col << 3) + 4,
307	258k	&b->mv[3][0], 4, 4, 8, 8, 4, 4, w1, h1, 0);
308	258k	}
309	323k	}
310
311	507k	if (b->comp) {
312	445k	mc_luma_dir(td, mc[4][b->filter][1], td->dst[0], ls_y,
313	445k	ref2->data[0], ref2->linesize[0], tref2,
314	445k	row << 3, col << 3, &b->mv[0][1], 0, 0, 8, 8, 4, 4, w2, h2, 1);
315	445k	mc_luma_dir(td, mc[4][b->filter][1], td->dst[0] + 4 * bytesperpixel, ls_y,
316	445k	ref2->data[0], ref2->linesize[0], tref2,
317	445k	row << 3, (col << 3) + 4, &b->mv[1][1], 4, 0, 8, 8, 4, 4, w2, h2, 1);
318	445k	mc_luma_dir(td, mc[4][b->filter][1],
319	445k	td->dst[0] + 4 * ls_y, ls_y,
320	445k	ref2->data[0], ref2->linesize[0], tref2,
321	445k	(row << 3) + 4, col << 3, &b->mv[2][1], 0, 4, 8, 8, 4, 4, w2, h2, 1);
322	445k	mc_luma_dir(td, mc[4][b->filter][1],
323	445k	td->dst[0] + 4 * ls_y + 4 * bytesperpixel, ls_y,
324	445k	ref2->data[0], ref2->linesize[0], tref2,
325	445k	(row << 3) + 4, (col << 3) + 4, &b->mv[3][1], 4, 4, 8, 8, 4, 4, w2, h2, 1);
326	445k	if (s->ss_v) {
327	160k	h2 = (h2 + 1) >> 1;
328	160k	if (s->ss_h) {
329	101k	w2 = (w2 + 1) >> 1;
330	101k	uvmv = ROUNDED_DIV_MVx4(b->mv[0][1], b->mv[1][1],
331	101k	b->mv[2][1], b->mv[3][1]);
332	101k	mc_chroma_dir(td, mc[4][b->filter][1],
333	101k	td->dst[1], td->dst[2], ls_uv,
334	101k	ref2->data[1], ref2->linesize[1],
335	101k	ref2->data[2], ref2->linesize[2], tref2,
336	101k	row << 2, col << 2,
337	101k	&uvmv, 0, 0, 4, 4, 4, 4, w2, h2, 1);
338	101k	} else {
339	58.6k	uvmv = ROUNDED_DIV_MVx2(b->mv[0][1], b->mv[2][1]);
340	58.6k	mc_chroma_dir(td, mc[4][b->filter][1],
341	58.6k	td->dst[1], td->dst[2], ls_uv,
342	58.6k	ref2->data[1], ref2->linesize[1],
343	58.6k	ref2->data[2], ref2->linesize[2], tref2,
344	58.6k	row << 2, col << 3,
345	58.6k	&uvmv, 0, 0, 8, 4, 4, 4, w2, h2, 1);
346	58.6k	uvmv = ROUNDED_DIV_MVx2(b->mv[1][1], b->mv[3][1]);
347	58.6k	mc_chroma_dir(td, mc[4][b->filter][1],
348	58.6k	td->dst[1] + 4 * bytesperpixel,
349	58.6k	td->dst[2] + 4 * bytesperpixel, ls_uv,
350	58.6k	ref2->data[1], ref2->linesize[1],
351	58.6k	ref2->data[2], ref2->linesize[2], tref2,
352	58.6k	row << 2, (col << 3) + 4,
353	58.6k	&uvmv, 4, 0, 8, 4, 4, 4, w2, h2, 1);
354	58.6k	}
355	284k	} else {
356	284k	if (s->ss_h) {
357	51.3k	w2 = (w2 + 1) >> 1;
358	51.3k	uvmv = ROUNDED_DIV_MVx2(b->mv[0][1], b->mv[1][1]);
359	51.3k	mc_chroma_dir(td, mc[4][b->filter][1],
360	51.3k	td->dst[1], td->dst[2], ls_uv,
361	51.3k	ref2->data[1], ref2->linesize[1],
362	51.3k	ref2->data[2], ref2->linesize[2], tref2,
363	51.3k	row << 3, col << 2,
364	51.3k	&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	51.3k	uvmv = ROUNDED_DIV_MVx2(b->mv[1][1], b->mv[2][1]);
369	51.3k	mc_chroma_dir(td, mc[4][b->filter][1],
370	51.3k	td->dst[1] + 4 * ls_uv, td->dst[2] + 4 * ls_uv, ls_uv,
371	51.3k	ref2->data[1], ref2->linesize[1],
372	51.3k	ref2->data[2], ref2->linesize[2], tref2,
373	51.3k	(row << 3) + 4, col << 2,
374	51.3k	&uvmv, 0, 4, 4, 8, 4, 4, w2, h2, 1);
375	233k	} else {
376	233k	mc_chroma_dir(td, mc[4][b->filter][1],
377	233k	td->dst[1], td->dst[2], ls_uv,
378	233k	ref2->data[1], ref2->linesize[1],
379	233k	ref2->data[2], ref2->linesize[2], tref2,
380	233k	row << 3, col << 3,
381	233k	&b->mv[0][1], 0, 0, 8, 8, 4, 4, w2, h2, 1);
382	233k	mc_chroma_dir(td, mc[4][b->filter][1],
383	233k	td->dst[1] + 4 * bytesperpixel,
384	233k	td->dst[2] + 4 * bytesperpixel, ls_uv,
385	233k	ref2->data[1], ref2->linesize[1],
386	233k	ref2->data[2], ref2->linesize[2], tref2,
387	233k	row << 3, (col << 3) + 4,
388	233k	&b->mv[1][1], 4, 0, 8, 8, 4, 4, w2, h2, 1);
389	233k	mc_chroma_dir(td, mc[4][b->filter][1],
390	233k	td->dst[1] + 4 * ls_uv, td->dst[2] + 4 * ls_uv, ls_uv,
391	233k	ref2->data[1], ref2->linesize[1],
392	233k	ref2->data[2], ref2->linesize[2], tref2,
393	233k	(row << 3) + 4, col << 3,
394	233k	&b->mv[2][1], 0, 4, 8, 8, 4, 4, w2, h2, 1);
395	233k	mc_chroma_dir(td, mc[4][b->filter][1],
396	233k	td->dst[1] + 4 * ls_uv + 4 * bytesperpixel,
397	233k	td->dst[2] + 4 * ls_uv + 4 * bytesperpixel, ls_uv,
398	233k	ref2->data[1], ref2->linesize[1],
399	233k	ref2->data[2], ref2->linesize[2], tref2,
400	233k	(row << 3) + 4, (col << 3) + 4,
401	233k	&b->mv[3][1], 4, 4, 8, 8, 4, 4, w2, h2, 1);
402	233k	}
403	284k	}
404	445k	}
405	507k	}
406	856k	} else {
407	856k	int bwl = bwlog_tab[0][b->bs];
408	856k	int bw = ff_vp9_bwh_tab[0][b->bs][0] * 4;
409	856k	int bh = ff_vp9_bwh_tab[0][b->bs][1] * 4;
410	856k	int uvbw = ff_vp9_bwh_tab[s->ss_h][b->bs][0] * 4;
411	856k	int uvbh = ff_vp9_bwh_tab[s->ss_v][b->bs][1] * 4;
412
413	856k	mc_luma_dir(td, mc[bwl][b->filter][0], td->dst[0], ls_y,
414	856k	ref1->data[0], ref1->linesize[0], tref1,
415	856k	row << 3, col << 3, &b->mv[0][0], 0, 0, bw, bh, bw, bh, w1, h1, 0);
416	856k	w1 = (w1 + s->ss_h) >> s->ss_h;
417	856k	h1 = (h1 + s->ss_v) >> s->ss_v;
418	856k	mc_chroma_dir(td, mc[bwl + s->ss_h][b->filter][0],
419	856k	td->dst[1], td->dst[2], ls_uv,
420	856k	ref1->data[1], ref1->linesize[1],
421	856k	ref1->data[2], ref1->linesize[2], tref1,
422	856k	row << (3 - s->ss_v), col << (3 - s->ss_h),
423	856k	&b->mv[0][0], 0, 0, uvbw, uvbh, uvbw, uvbh, w1, h1, 0);
424
425	856k	if (b->comp) {
426	675k	mc_luma_dir(td, mc[bwl][b->filter][1], td->dst[0], ls_y,
427	675k	ref2->data[0], ref2->linesize[0], tref2,
428	675k	row << 3, col << 3, &b->mv[0][1], 0, 0, bw, bh, bw, bh, w2, h2, 1);
429	675k	w2 = (w2 + s->ss_h) >> s->ss_h;
430	675k	h2 = (h2 + s->ss_v) >> s->ss_v;
431	675k	mc_chroma_dir(td, mc[bwl + s->ss_h][b->filter][1],
432	675k	td->dst[1], td->dst[2], ls_uv,
433	675k	ref2->data[1], ref2->linesize[1],
434	675k	ref2->data[2], ref2->linesize[2], tref2,
435	675k	row << (3 - s->ss_v), col << (3 - s->ss_h),
436	675k	&b->mv[0][1], 0, 0, uvbw, uvbh, uvbw, uvbh, w2, h2, 1);
437	675k	}
438	856k	}
439	1.36M	}

vp9recon.c:inter_pred_scaled_16bpp

Line	Count	Source
31	1.60M	{
32	1.60M	static const uint8_t bwlog_tab[2][N_BS_SIZES] = {
33	1.60M	{ 0, 0, 1, 1, 1, 2, 2, 2, 3, 3, 3, 4, 4 },
34	1.60M	{ 1, 1, 2, 2, 2, 3, 3, 3, 4, 4, 4, 4, 4 },
35	1.60M	};
36	1.60M	const VP9Context *s = td->s;
37	1.60M	VP9Block *b = td->b;
38	1.60M	int row = td->row, col = td->col;
39	1.60M	const ProgressFrame tref1 = &s->s.refs[s->s.h.refidx[b->ref[0]]], tref2;
40	1.60M	const AVFrame ref1 = tref1->f, ref2;
41	1.60M	int w1 = ref1->width, h1 = ref1->height, w2, h2;
42	1.60M	ptrdiff_t ls_y = td->y_stride, ls_uv = td->uv_stride;
43	1.60M	int bytesperpixel = BYTES_PER_PIXEL;
44
45	1.60M	if (b->comp) {
46	1.22M	tref2 = &s->s.refs[s->s.h.refidx[b->ref[1]]];
47	1.22M	ref2 = tref2->f;
48	1.22M	w2 = ref2->width;
49	1.22M	h2 = ref2->height;
50	1.22M	}
51
52		// y inter pred
53	1.60M	if (b->bs > BS_8x8) {
54	481k	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	481k	{
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	481k	mc_luma_dir(td, mc[4][b->filter][0], td->dst[0], ls_y,
215	481k	ref1->data[0], ref1->linesize[0], tref1,
216	481k	row << 3, col << 3, &b->mv[0][0],
217	481k	0, 0, 8, 8, 4, 4, w1, h1, 0);
218	481k	mc_luma_dir(td, mc[4][b->filter][0], td->dst[0] + 4 * bytesperpixel, ls_y,
219	481k	ref1->data[0], ref1->linesize[0], tref1,
220	481k	row << 3, (col << 3) + 4, &b->mv[1][0],
221	481k	4, 0, 8, 8, 4, 4, w1, h1, 0);
222	481k	mc_luma_dir(td, mc[4][b->filter][0],
223	481k	td->dst[0] + 4 * ls_y, ls_y,
224	481k	ref1->data[0], ref1->linesize[0], tref1,
225	481k	(row << 3) + 4, col << 3, &b->mv[2][0],
226	481k	0, 4, 8, 8, 4, 4, w1, h1, 0);
227	481k	mc_luma_dir(td, mc[4][b->filter][0],
228	481k	td->dst[0] + 4 * ls_y + 4 * bytesperpixel, ls_y,
229	481k	ref1->data[0], ref1->linesize[0], tref1,
230	481k	(row << 3) + 4, (col << 3) + 4, &b->mv[3][0],
231	481k	4, 4, 8, 8, 4, 4, w1, h1, 0);
232	481k	if (s->ss_v) {
233	183k	h1 = (h1 + 1) >> 1;
234	183k	if (s->ss_h) {
235	80.7k	w1 = (w1 + 1) >> 1;
236	80.7k	uvmv = ROUNDED_DIV_MVx4(b->mv[0][0], b->mv[1][0],
237	80.7k	b->mv[2][0], b->mv[3][0]);
238	80.7k	mc_chroma_dir(td, mc[4][b->filter][0],
239	80.7k	td->dst[1], td->dst[2], ls_uv,
240	80.7k	ref1->data[1], ref1->linesize[1],
241	80.7k	ref1->data[2], ref1->linesize[2], tref1,
242	80.7k	row << 2, col << 2,
243	80.7k	&uvmv, 0, 0, 4, 4, 4, 4, w1, h1, 0);
244	102k	} else {
245	102k	uvmv = ROUNDED_DIV_MVx2(b->mv[0][0], b->mv[2][0]);
246	102k	mc_chroma_dir(td, mc[4][b->filter][0],
247	102k	td->dst[1], td->dst[2], ls_uv,
248	102k	ref1->data[1], ref1->linesize[1],
249	102k	ref1->data[2], ref1->linesize[2], tref1,
250	102k	row << 2, col << 3,
251	102k	&uvmv, 0, 0, 8, 4, 4, 4, w1, h1, 0);
252	102k	uvmv = ROUNDED_DIV_MVx2(b->mv[1][0], b->mv[3][0]);
253	102k	mc_chroma_dir(td, mc[4][b->filter][0],
254	102k	td->dst[1] + 4 * bytesperpixel,
255	102k	td->dst[2] + 4 * bytesperpixel, ls_uv,
256	102k	ref1->data[1], ref1->linesize[1],
257	102k	ref1->data[2], ref1->linesize[2], tref1,
258	102k	row << 2, (col << 3) + 4,
259	102k	&uvmv, 4, 0, 8, 4, 4, 4, w1, h1, 0);
260	102k	}
261	298k	} else {
262	298k	if (s->ss_h) {
263	146k	w1 = (w1 + 1) >> 1;
264	146k	uvmv = ROUNDED_DIV_MVx2(b->mv[0][0], b->mv[1][0]);
265	146k	mc_chroma_dir(td, mc[4][b->filter][0],
266	146k	td->dst[1], td->dst[2], ls_uv,
267	146k	ref1->data[1], ref1->linesize[1],
268	146k	ref1->data[2], ref1->linesize[2], tref1,
269	146k	row << 3, col << 2,
270	146k	&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	146k	uvmv = ROUNDED_DIV_MVx2(b->mv[1][0], b->mv[2][0]);
275	146k	mc_chroma_dir(td, mc[4][b->filter][0],
276	146k	td->dst[1] + 4 * ls_uv, td->dst[2] + 4 * ls_uv, ls_uv,
277	146k	ref1->data[1], ref1->linesize[1],
278	146k	ref1->data[2], ref1->linesize[2], tref1,
279	146k	(row << 3) + 4, col << 2,
280	146k	&uvmv, 0, 4, 4, 8, 4, 4, w1, h1, 0);
281	152k	} else {
282	152k	mc_chroma_dir(td, mc[4][b->filter][0],
283	152k	td->dst[1], td->dst[2], ls_uv,
284	152k	ref1->data[1], ref1->linesize[1],
285	152k	ref1->data[2], ref1->linesize[2], tref1,
286	152k	row << 3, col << 3,
287	152k	&b->mv[0][0], 0, 0, 8, 8, 4, 4, w1, h1, 0);
288	152k	mc_chroma_dir(td, mc[4][b->filter][0],
289	152k	td->dst[1] + 4 * bytesperpixel,
290	152k	td->dst[2] + 4 * bytesperpixel, ls_uv,
291	152k	ref1->data[1], ref1->linesize[1],
292	152k	ref1->data[2], ref1->linesize[2], tref1,
293	152k	row << 3, (col << 3) + 4,
294	152k	&b->mv[1][0], 4, 0, 8, 8, 4, 4, w1, h1, 0);
295	152k	mc_chroma_dir(td, mc[4][b->filter][0],
296	152k	td->dst[1] + 4 * ls_uv, td->dst[2] + 4 * ls_uv, ls_uv,
297	152k	ref1->data[1], ref1->linesize[1],
298	152k	ref1->data[2], ref1->linesize[2], tref1,
299	152k	(row << 3) + 4, col << 3,
300	152k	&b->mv[2][0], 0, 4, 8, 8, 4, 4, w1, h1, 0);
301	152k	mc_chroma_dir(td, mc[4][b->filter][0],
302	152k	td->dst[1] + 4 * ls_uv + 4 * bytesperpixel,
303	152k	td->dst[2] + 4 * ls_uv + 4 * bytesperpixel, ls_uv,
304	152k	ref1->data[1], ref1->linesize[1],
305	152k	ref1->data[2], ref1->linesize[2], tref1,
306	152k	(row << 3) + 4, (col << 3) + 4,
307	152k	&b->mv[3][0], 4, 4, 8, 8, 4, 4, w1, h1, 0);
308	152k	}
309	298k	}
310
311	481k	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	153k	h2 = (h2 + 1) >> 1;
328	153k	if (s->ss_h) {
329	65.3k	w2 = (w2 + 1) >> 1;
330	65.3k	uvmv = ROUNDED_DIV_MVx4(b->mv[0][1], b->mv[1][1],
331	65.3k	b->mv[2][1], b->mv[3][1]);
332	65.3k	mc_chroma_dir(td, mc[4][b->filter][1],
333	65.3k	td->dst[1], td->dst[2], ls_uv,
334	65.3k	ref2->data[1], ref2->linesize[1],
335	65.3k	ref2->data[2], ref2->linesize[2], tref2,
336	65.3k	row << 2, col << 2,
337	65.3k	&uvmv, 0, 0, 4, 4, 4, 4, w2, h2, 1);
338	88.1k	} else {
339	88.1k	uvmv = ROUNDED_DIV_MVx2(b->mv[0][1], b->mv[2][1]);
340	88.1k	mc_chroma_dir(td, mc[4][b->filter][1],
341	88.1k	td->dst[1], td->dst[2], ls_uv,
342	88.1k	ref2->data[1], ref2->linesize[1],
343	88.1k	ref2->data[2], ref2->linesize[2], tref2,
344	88.1k	row << 2, col << 3,
345	88.1k	&uvmv, 0, 0, 8, 4, 4, 4, w2, h2, 1);
346	88.1k	uvmv = ROUNDED_DIV_MVx2(b->mv[1][1], b->mv[3][1]);
347	88.1k	mc_chroma_dir(td, mc[4][b->filter][1],
348	88.1k	td->dst[1] + 4 * bytesperpixel,
349	88.1k	td->dst[2] + 4 * bytesperpixel, ls_uv,
350	88.1k	ref2->data[1], ref2->linesize[1],
351	88.1k	ref2->data[2], ref2->linesize[2], tref2,
352	88.1k	row << 2, (col << 3) + 4,
353	88.1k	&uvmv, 4, 0, 8, 4, 4, 4, w2, h2, 1);
354	88.1k	}
355	249k	} else {
356	249k	if (s->ss_h) {
357	114k	w2 = (w2 + 1) >> 1;
358	114k	uvmv = ROUNDED_DIV_MVx2(b->mv[0][1], b->mv[1][1]);
359	114k	mc_chroma_dir(td, mc[4][b->filter][1],
360	114k	td->dst[1], td->dst[2], ls_uv,
361	114k	ref2->data[1], ref2->linesize[1],
362	114k	ref2->data[2], ref2->linesize[2], tref2,
363	114k	row << 3, col << 2,
364	114k	&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	114k	uvmv = ROUNDED_DIV_MVx2(b->mv[1][1], b->mv[2][1]);
369	114k	mc_chroma_dir(td, mc[4][b->filter][1],
370	114k	td->dst[1] + 4 * ls_uv, td->dst[2] + 4 * ls_uv, ls_uv,
371	114k	ref2->data[1], ref2->linesize[1],
372	114k	ref2->data[2], ref2->linesize[2], tref2,
373	114k	(row << 3) + 4, col << 2,
374	114k	&uvmv, 0, 4, 4, 8, 4, 4, w2, h2, 1);
375	134k	} else {
376	134k	mc_chroma_dir(td, mc[4][b->filter][1],
377	134k	td->dst[1], td->dst[2], ls_uv,
378	134k	ref2->data[1], ref2->linesize[1],
379	134k	ref2->data[2], ref2->linesize[2], tref2,
380	134k	row << 3, col << 3,
381	134k	&b->mv[0][1], 0, 0, 8, 8, 4, 4, w2, h2, 1);
382	134k	mc_chroma_dir(td, mc[4][b->filter][1],
383	134k	td->dst[1] + 4 * bytesperpixel,
384	134k	td->dst[2] + 4 * bytesperpixel, ls_uv,
385	134k	ref2->data[1], ref2->linesize[1],
386	134k	ref2->data[2], ref2->linesize[2], tref2,
387	134k	row << 3, (col << 3) + 4,
388	134k	&b->mv[1][1], 4, 0, 8, 8, 4, 4, w2, h2, 1);
389	134k	mc_chroma_dir(td, mc[4][b->filter][1],
390	134k	td->dst[1] + 4 * ls_uv, td->dst[2] + 4 * ls_uv, ls_uv,
391	134k	ref2->data[1], ref2->linesize[1],
392	134k	ref2->data[2], ref2->linesize[2], tref2,
393	134k	(row << 3) + 4, col << 3,
394	134k	&b->mv[2][1], 0, 4, 8, 8, 4, 4, w2, h2, 1);
395	134k	mc_chroma_dir(td, mc[4][b->filter][1],
396	134k	td->dst[1] + 4 * ls_uv + 4 * bytesperpixel,
397	134k	td->dst[2] + 4 * ls_uv + 4 * bytesperpixel, ls_uv,
398	134k	ref2->data[1], ref2->linesize[1],
399	134k	ref2->data[2], ref2->linesize[2], tref2,
400	134k	(row << 3) + 4, (col << 3) + 4,
401	134k	&b->mv[3][1], 4, 4, 8, 8, 4, 4, w2, h2, 1);
402	134k	}
403	249k	}
404	402k	}
405	481k	}
406	1.12M	} else {
407	1.12M	int bwl = bwlog_tab[0][b->bs];
408	1.12M	int bw = ff_vp9_bwh_tab[0][b->bs][0] * 4;
409	1.12M	int bh = ff_vp9_bwh_tab[0][b->bs][1] * 4;
410	1.12M	int uvbw = ff_vp9_bwh_tab[s->ss_h][b->bs][0] * 4;
411	1.12M	int uvbh = ff_vp9_bwh_tab[s->ss_v][b->bs][1] * 4;
412
413	1.12M	mc_luma_dir(td, mc[bwl][b->filter][0], td->dst[0], ls_y,
414	1.12M	ref1->data[0], ref1->linesize[0], tref1,
415	1.12M	row << 3, col << 3, &b->mv[0][0], 0, 0, bw, bh, bw, bh, w1, h1, 0);
416	1.12M	w1 = (w1 + s->ss_h) >> s->ss_h;
417	1.12M	h1 = (h1 + s->ss_v) >> s->ss_v;
418	1.12M	mc_chroma_dir(td, mc[bwl + s->ss_h][b->filter][0],
419	1.12M	td->dst[1], td->dst[2], ls_uv,
420	1.12M	ref1->data[1], ref1->linesize[1],
421	1.12M	ref1->data[2], ref1->linesize[2], tref1,
422	1.12M	row << (3 - s->ss_v), col << (3 - s->ss_h),
423	1.12M	&b->mv[0][0], 0, 0, uvbw, uvbh, uvbw, uvbh, w1, h1, 0);
424
425	1.12M	if (b->comp) {
426	823k	mc_luma_dir(td, mc[bwl][b->filter][1], td->dst[0], ls_y,
427	823k	ref2->data[0], ref2->linesize[0], tref2,
428	823k	row << 3, col << 3, &b->mv[0][1], 0, 0, bw, bh, bw, bh, w2, h2, 1);
429	823k	w2 = (w2 + s->ss_h) >> s->ss_h;
430	823k	h2 = (h2 + s->ss_v) >> s->ss_v;
431	823k	mc_chroma_dir(td, mc[bwl + s->ss_h][b->filter][1],
432	823k	td->dst[1], td->dst[2], ls_uv,
433	823k	ref2->data[1], ref2->linesize[1],
434	823k	ref2->data[2], ref2->linesize[2], tref2,
435	823k	row << (3 - s->ss_v), col << (3 - s->ss_h),
436	823k	&b->mv[0][1], 0, 0, uvbw, uvbh, uvbw, uvbh, w2, h2, 1);
437	823k	}
438	1.12M	}
439	1.60M	}

vp9recon.c:inter_pred_8bpp

Line	Count	Source
31	1.51M	{
32	1.51M	static const uint8_t bwlog_tab[2][N_BS_SIZES] = {
33	1.51M	{ 0, 0, 1, 1, 1, 2, 2, 2, 3, 3, 3, 4, 4 },
34	1.51M	{ 1, 1, 2, 2, 2, 3, 3, 3, 4, 4, 4, 4, 4 },
35	1.51M	};
36	1.51M	const VP9Context *s = td->s;
37	1.51M	VP9Block *b = td->b;
38	1.51M	int row = td->row, col = td->col;
39	1.51M	const ProgressFrame tref1 = &s->s.refs[s->s.h.refidx[b->ref[0]]], tref2;
40	1.51M	const AVFrame ref1 = tref1->f, ref2;
41	1.51M	int w1 = ref1->width, h1 = ref1->height, w2, h2;
42	1.51M	ptrdiff_t ls_y = td->y_stride, ls_uv = td->uv_stride;
43	1.51M	int bytesperpixel = BYTES_PER_PIXEL;
44
45	1.51M	if (b->comp) {
46	1.02M	tref2 = &s->s.refs[s->s.h.refidx[b->ref[1]]];
47	1.02M	ref2 = tref2->f;
48	1.02M	w2 = ref2->width;
49	1.02M	h2 = ref2->height;
50	1.02M	}
51
52		// y inter pred
53	1.51M	if (b->bs > BS_8x8) {
54	522k	VP9mv uvmv;
55
56	522k	#if SCALED == 0
57	522k	if (b->bs == BS_8x4) {
58	140k	mc_luma_dir(td, mc[3][b->filter][0], td->dst[0], ls_y,
59	140k	ref1->data[0], ref1->linesize[0], tref1,
60	140k	row << 3, col << 3, &b->mv[0][0],,,,, 8, 4, w1, h1, 0);
61	140k	mc_luma_dir(td, mc[3][b->filter][0],
62	140k	td->dst[0] + 4 * ls_y, ls_y,
63	140k	ref1->data[0], ref1->linesize[0], tref1,
64	140k	(row << 3) + 4, col << 3, &b->mv[2][0],,,,, 8, 4, w1, h1, 0);
65	140k	w1 = (w1 + s->ss_h) >> s->ss_h;
66	140k	if (s->ss_v) {
67	63.1k	h1 = (h1 + 1) >> 1;
68	63.1k	uvmv = ROUNDED_DIV_MVx2(b->mv[0][0], b->mv[2][0]);
69	63.1k	mc_chroma_dir(td, mc[3 + s->ss_h][b->filter][0],
70	63.1k	td->dst[1], td->dst[2], ls_uv,
71	63.1k	ref1->data[1], ref1->linesize[1],
72	63.1k	ref1->data[2], ref1->linesize[2], tref1,
73	63.1k	row << 2, col << (3 - s->ss_h),
74	63.1k	&uvmv,,,,, 8 >> s->ss_h, 4, w1, h1, 0);
75	77.6k	} else {
76	77.6k	mc_chroma_dir(td, mc[3 + s->ss_h][b->filter][0],
77	77.6k	td->dst[1], td->dst[2], ls_uv,
78	77.6k	ref1->data[1], ref1->linesize[1],
79	77.6k	ref1->data[2], ref1->linesize[2], tref1,
80	77.6k	row << 3, col << (3 - s->ss_h),
81	77.6k	&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	77.6k	if (s->ss_h == 0) {
86	36.5k	uvmv = b->mv[2][0];
87	41.1k	} else {
88	41.1k	uvmv = ROUNDED_DIV_MVx2(b->mv[0][0], b->mv[2][0]);
89	41.1k	}
90	77.6k	mc_chroma_dir(td, mc[3 + s->ss_h][b->filter][0],
91	77.6k	td->dst[1] + 4 * ls_uv, td->dst[2] + 4 * ls_uv, ls_uv,
92	77.6k	ref1->data[1], ref1->linesize[1],
93	77.6k	ref1->data[2], ref1->linesize[2], tref1,
94	77.6k	(row << 3) + 4, col << (3 - s->ss_h),
95	77.6k	&uvmv,,,,, 8 >> s->ss_h, 4, w1, h1, 0);
96	77.6k	}
97
98	140k	if (b->comp) {
99	109k	mc_luma_dir(td, mc[3][b->filter][1], td->dst[0], ls_y,
100	109k	ref2->data[0], ref2->linesize[0], tref2,
101	109k	row << 3, col << 3, &b->mv[0][1],,,,, 8, 4, w2, h2, 1);
102	109k	mc_luma_dir(td, mc[3][b->filter][1],
103	109k	td->dst[0] + 4 * ls_y, ls_y,
104	109k	ref2->data[0], ref2->linesize[0], tref2,
105	109k	(row << 3) + 4, col << 3, &b->mv[2][1],,,,, 8, 4, w2, h2, 1);
106	109k	w2 = (w2 + s->ss_h) >> s->ss_h;
107	109k	if (s->ss_v) {
108	46.9k	h2 = (h2 + 1) >> 1;
109	46.9k	uvmv = ROUNDED_DIV_MVx2(b->mv[0][1], b->mv[2][1]);
110	46.9k	mc_chroma_dir(td, mc[3 + s->ss_h][b->filter][1],
111	46.9k	td->dst[1], td->dst[2], ls_uv,
112	46.9k	ref2->data[1], ref2->linesize[1],
113	46.9k	ref2->data[2], ref2->linesize[2], tref2,
114	46.9k	row << 2, col << (3 - s->ss_h),
115	46.9k	&uvmv,,,,, 8 >> s->ss_h, 4, w2, h2, 1);
116	62.1k	} else {
117	62.1k	mc_chroma_dir(td, mc[3 + s->ss_h][b->filter][1],
118	62.1k	td->dst[1], td->dst[2], ls_uv,
119	62.1k	ref2->data[1], ref2->linesize[1],
120	62.1k	ref2->data[2], ref2->linesize[2], tref2,
121	62.1k	row << 3, col << (3 - s->ss_h),
122	62.1k	&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	62.1k	if (s->ss_h == 0) {
127	26.8k	uvmv = b->mv[2][1];
128	35.2k	} else {
129	35.2k	uvmv = ROUNDED_DIV_MVx2(b->mv[0][1], b->mv[2][1]);
130	35.2k	}
131	62.1k	mc_chroma_dir(td, mc[3 + s->ss_h][b->filter][1],
132	62.1k	td->dst[1] + 4 * ls_uv, td->dst[2] + 4 * ls_uv, ls_uv,
133	62.1k	ref2->data[1], ref2->linesize[1],
134	62.1k	ref2->data[2], ref2->linesize[2], tref2,
135	62.1k	(row << 3) + 4, col << (3 - s->ss_h),
136	62.1k	&uvmv,,,,, 8 >> s->ss_h, 4, w2, h2, 1);
137	62.1k	}
138	109k	}
139	381k	} else if (b->bs == BS_4x8) {
140	77.1k	mc_luma_dir(td, mc[4][b->filter][0], td->dst[0], ls_y,
141	77.1k	ref1->data[0], ref1->linesize[0], tref1,
142	77.1k	row << 3, col << 3, &b->mv[0][0],,,,, 4, 8, w1, h1, 0);
143	77.1k	mc_luma_dir(td, mc[4][b->filter][0], td->dst[0] + 4 * bytesperpixel, ls_y,
144	77.1k	ref1->data[0], ref1->linesize[0], tref1,
145	77.1k	row << 3, (col << 3) + 4, &b->mv[1][0],,,,, 4, 8, w1, h1, 0);
146	77.1k	h1 = (h1 + s->ss_v) >> s->ss_v;
147	77.1k	if (s->ss_h) {
148	23.0k	w1 = (w1 + 1) >> 1;
149	23.0k	uvmv = ROUNDED_DIV_MVx2(b->mv[0][0], b->mv[1][0]);
150	23.0k	mc_chroma_dir(td, mc[4][b->filter][0],
151	23.0k	td->dst[1], td->dst[2], ls_uv,
152	23.0k	ref1->data[1], ref1->linesize[1],
153	23.0k	ref1->data[2], ref1->linesize[2], tref1,
154	23.0k	row << (3 - s->ss_v), col << 2,
155	23.0k	&uvmv,,,,, 4, 8 >> s->ss_v, w1, h1, 0);
156	54.0k	} else {
157	54.0k	mc_chroma_dir(td, mc[4][b->filter][0],
158	54.0k	td->dst[1], td->dst[2], ls_uv,
159	54.0k	ref1->data[1], ref1->linesize[1],
160	54.0k	ref1->data[2], ref1->linesize[2], tref1,
161	54.0k	row << (3 - s->ss_v), col << 3,
162	54.0k	&b->mv[0][0],,,,, 4, 8 >> s->ss_v, w1, h1, 0);
163	54.0k	mc_chroma_dir(td, mc[4][b->filter][0],
164	54.0k	td->dst[1] + 4 * bytesperpixel,
165	54.0k	td->dst[2] + 4 * bytesperpixel, ls_uv,
166	54.0k	ref1->data[1], ref1->linesize[1],
167	54.0k	ref1->data[2], ref1->linesize[2], tref1,
168	54.0k	row << (3 - s->ss_v), (col << 3) + 4,
169	54.0k	&b->mv[1][0],,,,, 4, 8 >> s->ss_v, w1, h1, 0);
170	54.0k	}
171
172	77.1k	if (b->comp) {
173	56.4k	mc_luma_dir(td, mc[4][b->filter][1], td->dst[0], ls_y,
174	56.4k	ref2->data[0], ref2->linesize[0], tref2,
175	56.4k	row << 3, col << 3, &b->mv[0][1],,,,, 4, 8, w2, h2, 1);
176	56.4k	mc_luma_dir(td, mc[4][b->filter][1], td->dst[0] + 4 * bytesperpixel, ls_y,
177	56.4k	ref2->data[0], ref2->linesize[0], tref2,
178	56.4k	row << 3, (col << 3) + 4, &b->mv[1][1],,,,, 4, 8, w2, h2, 1);
179	56.4k	h2 = (h2 + s->ss_v) >> s->ss_v;
180	56.4k	if (s->ss_h) {
181	14.9k	w2 = (w2 + 1) >> 1;
182	14.9k	uvmv = ROUNDED_DIV_MVx2(b->mv[0][1], b->mv[1][1]);
183	14.9k	mc_chroma_dir(td, mc[4][b->filter][1],
184	14.9k	td->dst[1], td->dst[2], ls_uv,
185	14.9k	ref2->data[1], ref2->linesize[1],
186	14.9k	ref2->data[2], ref2->linesize[2], tref2,
187	14.9k	row << (3 - s->ss_v), col << 2,
188	14.9k	&uvmv,,,,, 4, 8 >> s->ss_v, w2, h2, 1);
189	41.4k	} else {
190	41.4k	mc_chroma_dir(td, mc[4][b->filter][1],
191	41.4k	td->dst[1], td->dst[2], ls_uv,
192	41.4k	ref2->data[1], ref2->linesize[1],
193	41.4k	ref2->data[2], ref2->linesize[2], tref2,
194	41.4k	row << (3 - s->ss_v), col << 3,
195	41.4k	&b->mv[0][1],,,,, 4, 8 >> s->ss_v, w2, h2, 1);
196	41.4k	mc_chroma_dir(td, mc[4][b->filter][1],
197	41.4k	td->dst[1] + 4 * bytesperpixel,
198	41.4k	td->dst[2] + 4 * bytesperpixel, ls_uv,
199	41.4k	ref2->data[1], ref2->linesize[1],
200	41.4k	ref2->data[2], ref2->linesize[2], tref2,
201	41.4k	row << (3 - s->ss_v), (col << 3) + 4,
202	41.4k	&b->mv[1][1],,,,, 4, 8 >> s->ss_v, w2, h2, 1);
203	41.4k	}
204	56.4k	}
205	77.1k	} else
206	304k	#endif
207	304k	{
208	304k	#if SCALED == 0
209	304k	av_assert2(b->bs == BS_4x4);
210	304k	#endif
211
212		// FIXME if two horizontally adjacent blocks have the same MV,
213		// do a w8 instead of a w4 call
214	304k	mc_luma_dir(td, mc[4][b->filter][0], td->dst[0], ls_y,
215	304k	ref1->data[0], ref1->linesize[0], tref1,
216	304k	row << 3, col << 3, &b->mv[0][0],
217	304k	0, 0, 8, 8, 4, 4, w1, h1, 0);
218	304k	mc_luma_dir(td, mc[4][b->filter][0], td->dst[0] + 4 * bytesperpixel, ls_y,
219	304k	ref1->data[0], ref1->linesize[0], tref1,
220	304k	row << 3, (col << 3) + 4, &b->mv[1][0],
221	304k	4, 0, 8, 8, 4, 4, w1, h1, 0);
222	304k	mc_luma_dir(td, mc[4][b->filter][0],
223	304k	td->dst[0] + 4 * ls_y, ls_y,
224	304k	ref1->data[0], ref1->linesize[0], tref1,
225	304k	(row << 3) + 4, col << 3, &b->mv[2][0],
226	304k	0, 4, 8, 8, 4, 4, w1, h1, 0);
227	304k	mc_luma_dir(td, mc[4][b->filter][0],
228	304k	td->dst[0] + 4 * ls_y + 4 * bytesperpixel, ls_y,
229	304k	ref1->data[0], ref1->linesize[0], tref1,
230	304k	(row << 3) + 4, (col << 3) + 4, &b->mv[3][0],
231	304k	4, 4, 8, 8, 4, 4, w1, h1, 0);
232	304k	if (s->ss_v) {
233	197k	h1 = (h1 + 1) >> 1;
234	197k	if (s->ss_h) {
235	14.0k	w1 = (w1 + 1) >> 1;
236	14.0k	uvmv = ROUNDED_DIV_MVx4(b->mv[0][0], b->mv[1][0],
237	14.0k	b->mv[2][0], b->mv[3][0]);
238	14.0k	mc_chroma_dir(td, mc[4][b->filter][0],
239	14.0k	td->dst[1], td->dst[2], ls_uv,
240	14.0k	ref1->data[1], ref1->linesize[1],
241	14.0k	ref1->data[2], ref1->linesize[2], tref1,
242	14.0k	row << 2, col << 2,
243	14.0k	&uvmv, 0, 0, 4, 4, 4, 4, w1, h1, 0);
244	183k	} else {
245	183k	uvmv = ROUNDED_DIV_MVx2(b->mv[0][0], b->mv[2][0]);
246	183k	mc_chroma_dir(td, mc[4][b->filter][0],
247	183k	td->dst[1], td->dst[2], ls_uv,
248	183k	ref1->data[1], ref1->linesize[1],
249	183k	ref1->data[2], ref1->linesize[2], tref1,
250	183k	row << 2, col << 3,
251	183k	&uvmv, 0, 0, 8, 4, 4, 4, w1, h1, 0);
252	183k	uvmv = ROUNDED_DIV_MVx2(b->mv[1][0], b->mv[3][0]);
253	183k	mc_chroma_dir(td, mc[4][b->filter][0],
254	183k	td->dst[1] + 4 * bytesperpixel,
255	183k	td->dst[2] + 4 * bytesperpixel, ls_uv,
256	183k	ref1->data[1], ref1->linesize[1],
257	183k	ref1->data[2], ref1->linesize[2], tref1,
258	183k	row << 2, (col << 3) + 4,
259	183k	&uvmv, 4, 0, 8, 4, 4, 4, w1, h1, 0);
260	183k	}
261	197k	} else {
262	107k	if (s->ss_h) {
263	56.1k	w1 = (w1 + 1) >> 1;
264	56.1k	uvmv = ROUNDED_DIV_MVx2(b->mv[0][0], b->mv[1][0]);
265	56.1k	mc_chroma_dir(td, mc[4][b->filter][0],
266	56.1k	td->dst[1], td->dst[2], ls_uv,
267	56.1k	ref1->data[1], ref1->linesize[1],
268	56.1k	ref1->data[2], ref1->linesize[2], tref1,
269	56.1k	row << 3, col << 2,
270	56.1k	&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	56.1k	uvmv = ROUNDED_DIV_MVx2(b->mv[1][0], b->mv[2][0]);
275	56.1k	mc_chroma_dir(td, mc[4][b->filter][0],
276	56.1k	td->dst[1] + 4 * ls_uv, td->dst[2] + 4 * ls_uv, ls_uv,
277	56.1k	ref1->data[1], ref1->linesize[1],
278	56.1k	ref1->data[2], ref1->linesize[2], tref1,
279	56.1k	(row << 3) + 4, col << 2,
280	56.1k	&uvmv, 0, 4, 4, 8, 4, 4, w1, h1, 0);
281	56.1k	} else {
282	50.9k	mc_chroma_dir(td, mc[4][b->filter][0],
283	50.9k	td->dst[1], td->dst[2], ls_uv,
284	50.9k	ref1->data[1], ref1->linesize[1],
285	50.9k	ref1->data[2], ref1->linesize[2], tref1,
286	50.9k	row << 3, col << 3,
287	50.9k	&b->mv[0][0], 0, 0, 8, 8, 4, 4, w1, h1, 0);
288	50.9k	mc_chroma_dir(td, mc[4][b->filter][0],
289	50.9k	td->dst[1] + 4 * bytesperpixel,
290	50.9k	td->dst[2] + 4 * bytesperpixel, ls_uv,
291	50.9k	ref1->data[1], ref1->linesize[1],
292	50.9k	ref1->data[2], ref1->linesize[2], tref1,
293	50.9k	row << 3, (col << 3) + 4,
294	50.9k	&b->mv[1][0], 4, 0, 8, 8, 4, 4, w1, h1, 0);
295	50.9k	mc_chroma_dir(td, mc[4][b->filter][0],
296	50.9k	td->dst[1] + 4 * ls_uv, td->dst[2] + 4 * ls_uv, ls_uv,
297	50.9k	ref1->data[1], ref1->linesize[1],
298	50.9k	ref1->data[2], ref1->linesize[2], tref1,
299	50.9k	(row << 3) + 4, col << 3,
300	50.9k	&b->mv[2][0], 0, 4, 8, 8, 4, 4, w1, h1, 0);
301	50.9k	mc_chroma_dir(td, mc[4][b->filter][0],
302	50.9k	td->dst[1] + 4 * ls_uv + 4 * bytesperpixel,
303	50.9k	td->dst[2] + 4 * ls_uv + 4 * bytesperpixel, ls_uv,
304	50.9k	ref1->data[1], ref1->linesize[1],
305	50.9k	ref1->data[2], ref1->linesize[2], tref1,
306	50.9k	(row << 3) + 4, (col << 3) + 4,
307	50.9k	&b->mv[3][0], 4, 4, 8, 8, 4, 4, w1, h1, 0);
308	50.9k	}
309	107k	}
310
311	304k	if (b->comp) {
312	206k	mc_luma_dir(td, mc[4][b->filter][1], td->dst[0], ls_y,
313	206k	ref2->data[0], ref2->linesize[0], tref2,
314	206k	row << 3, col << 3, &b->mv[0][1], 0, 0, 8, 8, 4, 4, w2, h2, 1);
315	206k	mc_luma_dir(td, mc[4][b->filter][1], td->dst[0] + 4 * bytesperpixel, ls_y,
316	206k	ref2->data[0], ref2->linesize[0], tref2,
317	206k	row << 3, (col << 3) + 4, &b->mv[1][1], 4, 0, 8, 8, 4, 4, w2, h2, 1);
318	206k	mc_luma_dir(td, mc[4][b->filter][1],
319	206k	td->dst[0] + 4 * ls_y, ls_y,
320	206k	ref2->data[0], ref2->linesize[0], tref2,
321	206k	(row << 3) + 4, col << 3, &b->mv[2][1], 0, 4, 8, 8, 4, 4, w2, h2, 1);
322	206k	mc_luma_dir(td, mc[4][b->filter][1],
323	206k	td->dst[0] + 4 * ls_y + 4 * bytesperpixel, ls_y,
324	206k	ref2->data[0], ref2->linesize[0], tref2,
325	206k	(row << 3) + 4, (col << 3) + 4, &b->mv[3][1], 4, 4, 8, 8, 4, 4, w2, h2, 1);
326	206k	if (s->ss_v) {
327	140k	h2 = (h2 + 1) >> 1;
328	140k	if (s->ss_h) {
329	6.61k	w2 = (w2 + 1) >> 1;
330	6.61k	uvmv = ROUNDED_DIV_MVx4(b->mv[0][1], b->mv[1][1],
331	6.61k	b->mv[2][1], b->mv[3][1]);
332	6.61k	mc_chroma_dir(td, mc[4][b->filter][1],
333	6.61k	td->dst[1], td->dst[2], ls_uv,
334	6.61k	ref2->data[1], ref2->linesize[1],
335	6.61k	ref2->data[2], ref2->linesize[2], tref2,
336	6.61k	row << 2, col << 2,
337	6.61k	&uvmv, 0, 0, 4, 4, 4, 4, w2, h2, 1);
338	134k	} else {
339	134k	uvmv = ROUNDED_DIV_MVx2(b->mv[0][1], b->mv[2][1]);
340	134k	mc_chroma_dir(td, mc[4][b->filter][1],
341	134k	td->dst[1], td->dst[2], ls_uv,
342	134k	ref2->data[1], ref2->linesize[1],
343	134k	ref2->data[2], ref2->linesize[2], tref2,
344	134k	row << 2, col << 3,
345	134k	&uvmv, 0, 0, 8, 4, 4, 4, w2, h2, 1);
346	134k	uvmv = ROUNDED_DIV_MVx2(b->mv[1][1], b->mv[3][1]);
347	134k	mc_chroma_dir(td, mc[4][b->filter][1],
348	134k	td->dst[1] + 4 * bytesperpixel,
349	134k	td->dst[2] + 4 * bytesperpixel, ls_uv,
350	134k	ref2->data[1], ref2->linesize[1],
351	134k	ref2->data[2], ref2->linesize[2], tref2,
352	134k	row << 2, (col << 3) + 4,
353	134k	&uvmv, 4, 0, 8, 4, 4, 4, w2, h2, 1);
354	134k	}
355	140k	} else {
356	65.6k	if (s->ss_h) {
357	40.7k	w2 = (w2 + 1) >> 1;
358	40.7k	uvmv = ROUNDED_DIV_MVx2(b->mv[0][1], b->mv[1][1]);
359	40.7k	mc_chroma_dir(td, mc[4][b->filter][1],
360	40.7k	td->dst[1], td->dst[2], ls_uv,
361	40.7k	ref2->data[1], ref2->linesize[1],
362	40.7k	ref2->data[2], ref2->linesize[2], tref2,
363	40.7k	row << 3, col << 2,
364	40.7k	&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	40.7k	uvmv = ROUNDED_DIV_MVx2(b->mv[1][1], b->mv[2][1]);
369	40.7k	mc_chroma_dir(td, mc[4][b->filter][1],
370	40.7k	td->dst[1] + 4 * ls_uv, td->dst[2] + 4 * ls_uv, ls_uv,
371	40.7k	ref2->data[1], ref2->linesize[1],
372	40.7k	ref2->data[2], ref2->linesize[2], tref2,
373	40.7k	(row << 3) + 4, col << 2,
374	40.7k	&uvmv, 0, 4, 4, 8, 4, 4, w2, h2, 1);
375	40.7k	} else {
376	24.8k	mc_chroma_dir(td, mc[4][b->filter][1],
377	24.8k	td->dst[1], td->dst[2], ls_uv,
378	24.8k	ref2->data[1], ref2->linesize[1],
379	24.8k	ref2->data[2], ref2->linesize[2], tref2,
380	24.8k	row << 3, col << 3,
381	24.8k	&b->mv[0][1], 0, 0, 8, 8, 4, 4, w2, h2, 1);
382	24.8k	mc_chroma_dir(td, mc[4][b->filter][1],
383	24.8k	td->dst[1] + 4 * bytesperpixel,
384	24.8k	td->dst[2] + 4 * bytesperpixel, ls_uv,
385	24.8k	ref2->data[1], ref2->linesize[1],
386	24.8k	ref2->data[2], ref2->linesize[2], tref2,
387	24.8k	row << 3, (col << 3) + 4,
388	24.8k	&b->mv[1][1], 4, 0, 8, 8, 4, 4, w2, h2, 1);
389	24.8k	mc_chroma_dir(td, mc[4][b->filter][1],
390	24.8k	td->dst[1] + 4 * ls_uv, td->dst[2] + 4 * ls_uv, ls_uv,
391	24.8k	ref2->data[1], ref2->linesize[1],
392	24.8k	ref2->data[2], ref2->linesize[2], tref2,
393	24.8k	(row << 3) + 4, col << 3,
394	24.8k	&b->mv[2][1], 0, 4, 8, 8, 4, 4, w2, h2, 1);
395	24.8k	mc_chroma_dir(td, mc[4][b->filter][1],
396	24.8k	td->dst[1] + 4 * ls_uv + 4 * bytesperpixel,
397	24.8k	td->dst[2] + 4 * ls_uv + 4 * bytesperpixel, ls_uv,
398	24.8k	ref2->data[1], ref2->linesize[1],
399	24.8k	ref2->data[2], ref2->linesize[2], tref2,
400	24.8k	(row << 3) + 4, (col << 3) + 4,
401	24.8k	&b->mv[3][1], 4, 4, 8, 8, 4, 4, w2, h2, 1);
402	24.8k	}
403	65.6k	}
404	206k	}
405	304k	}
406	993k	} else {
407	993k	int bwl = bwlog_tab[0][b->bs];
408	993k	int bw = ff_vp9_bwh_tab[0][b->bs][0] * 4;
409	993k	int bh = ff_vp9_bwh_tab[0][b->bs][1] * 4;
410	993k	int uvbw = ff_vp9_bwh_tab[s->ss_h][b->bs][0] * 4;
411	993k	int uvbh = ff_vp9_bwh_tab[s->ss_v][b->bs][1] * 4;
412
413	993k	mc_luma_dir(td, mc[bwl][b->filter][0], td->dst[0], ls_y,
414	993k	ref1->data[0], ref1->linesize[0], tref1,
415	993k	row << 3, col << 3, &b->mv[0][0], 0, 0, bw, bh, bw, bh, w1, h1, 0);
416	993k	w1 = (w1 + s->ss_h) >> s->ss_h;
417	993k	h1 = (h1 + s->ss_v) >> s->ss_v;
418	993k	mc_chroma_dir(td, mc[bwl + s->ss_h][b->filter][0],
419	993k	td->dst[1], td->dst[2], ls_uv,
420	993k	ref1->data[1], ref1->linesize[1],
421	993k	ref1->data[2], ref1->linesize[2], tref1,
422	993k	row << (3 - s->ss_v), col << (3 - s->ss_h),
423	993k	&b->mv[0][0], 0, 0, uvbw, uvbh, uvbw, uvbh, w1, h1, 0);
424
425	993k	if (b->comp) {
426	654k	mc_luma_dir(td, mc[bwl][b->filter][1], td->dst[0], ls_y,
427	654k	ref2->data[0], ref2->linesize[0], tref2,
428	654k	row << 3, col << 3, &b->mv[0][1], 0, 0, bw, bh, bw, bh, w2, h2, 1);
429	654k	w2 = (w2 + s->ss_h) >> s->ss_h;
430	654k	h2 = (h2 + s->ss_v) >> s->ss_v;
431	654k	mc_chroma_dir(td, mc[bwl + s->ss_h][b->filter][1],
432	654k	td->dst[1], td->dst[2], ls_uv,
433	654k	ref2->data[1], ref2->linesize[1],
434	654k	ref2->data[2], ref2->linesize[2], tref2,
435	654k	row << (3 - s->ss_v), col << (3 - s->ss_h),
436	654k	&b->mv[0][1], 0, 0, uvbw, uvbh, uvbw, uvbh, w2, h2, 1);
437	654k	}
438	993k	}
439	1.51M	}

vp9recon.c:inter_pred_16bpp

Line	Count	Source
31	2.42M	{
32	2.42M	static const uint8_t bwlog_tab[2][N_BS_SIZES] = {
33	2.42M	{ 0, 0, 1, 1, 1, 2, 2, 2, 3, 3, 3, 4, 4 },
34	2.42M	{ 1, 1, 2, 2, 2, 3, 3, 3, 4, 4, 4, 4, 4 },
35	2.42M	};
36	2.42M	const VP9Context *s = td->s;
37	2.42M	VP9Block *b = td->b;
38	2.42M	int row = td->row, col = td->col;
39	2.42M	const ProgressFrame tref1 = &s->s.refs[s->s.h.refidx[b->ref[0]]], tref2;
40	2.42M	const AVFrame ref1 = tref1->f, ref2;
41	2.42M	int w1 = ref1->width, h1 = ref1->height, w2, h2;
42	2.42M	ptrdiff_t ls_y = td->y_stride, ls_uv = td->uv_stride;
43	2.42M	int bytesperpixel = BYTES_PER_PIXEL;
44
45	2.42M	if (b->comp) {
46	1.15M	tref2 = &s->s.refs[s->s.h.refidx[b->ref[1]]];
47	1.15M	ref2 = tref2->f;
48	1.15M	w2 = ref2->width;
49	1.15M	h2 = ref2->height;
50	1.15M	}
51
52		// y inter pred
53	2.42M	if (b->bs > BS_8x8) {
54	772k	VP9mv uvmv;
55
56	772k	#if SCALED == 0
57	772k	if (b->bs == BS_8x4) {
58	271k	mc_luma_dir(td, mc[3][b->filter][0], td->dst[0], ls_y,
59	271k	ref1->data[0], ref1->linesize[0], tref1,
60	271k	row << 3, col << 3, &b->mv[0][0],,,,, 8, 4, w1, h1, 0);
61	271k	mc_luma_dir(td, mc[3][b->filter][0],
62	271k	td->dst[0] + 4 * ls_y, ls_y,
63	271k	ref1->data[0], ref1->linesize[0], tref1,
64	271k	(row << 3) + 4, col << 3, &b->mv[2][0],,,,, 8, 4, w1, h1, 0);
65	271k	w1 = (w1 + s->ss_h) >> s->ss_h;
66	271k	if (s->ss_v) {
67	80.6k	h1 = (h1 + 1) >> 1;
68	80.6k	uvmv = ROUNDED_DIV_MVx2(b->mv[0][0], b->mv[2][0]);
69	80.6k	mc_chroma_dir(td, mc[3 + s->ss_h][b->filter][0],
70	80.6k	td->dst[1], td->dst[2], ls_uv,
71	80.6k	ref1->data[1], ref1->linesize[1],
72	80.6k	ref1->data[2], ref1->linesize[2], tref1,
73	80.6k	row << 2, col << (3 - s->ss_h),
74	80.6k	&uvmv,,,,, 8 >> s->ss_h, 4, w1, h1, 0);
75	190k	} else {
76	190k	mc_chroma_dir(td, mc[3 + s->ss_h][b->filter][0],
77	190k	td->dst[1], td->dst[2], ls_uv,
78	190k	ref1->data[1], ref1->linesize[1],
79	190k	ref1->data[2], ref1->linesize[2], tref1,
80	190k	row << 3, col << (3 - s->ss_h),
81	190k	&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	190k	if (s->ss_h == 0) {
86	150k	uvmv = b->mv[2][0];
87	150k	} else {
88	40.6k	uvmv = ROUNDED_DIV_MVx2(b->mv[0][0], b->mv[2][0]);
89	40.6k	}
90	190k	mc_chroma_dir(td, mc[3 + s->ss_h][b->filter][0],
91	190k	td->dst[1] + 4 * ls_uv, td->dst[2] + 4 * ls_uv, ls_uv,
92	190k	ref1->data[1], ref1->linesize[1],
93	190k	ref1->data[2], ref1->linesize[2], tref1,
94	190k	(row << 3) + 4, col << (3 - s->ss_h),
95	190k	&uvmv,,,,, 8 >> s->ss_h, 4, w1, h1, 0);
96	190k	}
97
98	271k	if (b->comp) {
99	125k	mc_luma_dir(td, mc[3][b->filter][1], td->dst[0], ls_y,
100	125k	ref2->data[0], ref2->linesize[0], tref2,
101	125k	row << 3, col << 3, &b->mv[0][1],,,,, 8, 4, w2, h2, 1);
102	125k	mc_luma_dir(td, mc[3][b->filter][1],
103	125k	td->dst[0] + 4 * ls_y, ls_y,
104	125k	ref2->data[0], ref2->linesize[0], tref2,
105	125k	(row << 3) + 4, col << 3, &b->mv[2][1],,,,, 8, 4, w2, h2, 1);
106	125k	w2 = (w2 + s->ss_h) >> s->ss_h;
107	125k	if (s->ss_v) {
108	47.7k	h2 = (h2 + 1) >> 1;
109	47.7k	uvmv = ROUNDED_DIV_MVx2(b->mv[0][1], b->mv[2][1]);
110	47.7k	mc_chroma_dir(td, mc[3 + s->ss_h][b->filter][1],
111	47.7k	td->dst[1], td->dst[2], ls_uv,
112	47.7k	ref2->data[1], ref2->linesize[1],
113	47.7k	ref2->data[2], ref2->linesize[2], tref2,
114	47.7k	row << 2, col << (3 - s->ss_h),
115	47.7k	&uvmv,,,,, 8 >> s->ss_h, 4, w2, h2, 1);
116	77.5k	} else {
117	77.5k	mc_chroma_dir(td, mc[3 + s->ss_h][b->filter][1],
118	77.5k	td->dst[1], td->dst[2], ls_uv,
119	77.5k	ref2->data[1], ref2->linesize[1],
120	77.5k	ref2->data[2], ref2->linesize[2], tref2,
121	77.5k	row << 3, col << (3 - s->ss_h),
122	77.5k	&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	77.5k	if (s->ss_h == 0) {
127	62.1k	uvmv = b->mv[2][1];
128	62.1k	} else {
129	15.4k	uvmv = ROUNDED_DIV_MVx2(b->mv[0][1], b->mv[2][1]);
130	15.4k	}
131	77.5k	mc_chroma_dir(td, mc[3 + s->ss_h][b->filter][1],
132	77.5k	td->dst[1] + 4 * ls_uv, td->dst[2] + 4 * ls_uv, ls_uv,
133	77.5k	ref2->data[1], ref2->linesize[1],
134	77.5k	ref2->data[2], ref2->linesize[2], tref2,
135	77.5k	(row << 3) + 4, col << (3 - s->ss_h),
136	77.5k	&uvmv,,,,, 8 >> s->ss_h, 4, w2, h2, 1);
137	77.5k	}
138	125k	}
139	500k	} else if (b->bs == BS_4x8) {
140	162k	mc_luma_dir(td, mc[4][b->filter][0], td->dst[0], ls_y,
141	162k	ref1->data[0], ref1->linesize[0], tref1,
142	162k	row << 3, col << 3, &b->mv[0][0],,,,, 4, 8, w1, h1, 0);
143	162k	mc_luma_dir(td, mc[4][b->filter][0], td->dst[0] + 4 * bytesperpixel, ls_y,
144	162k	ref1->data[0], ref1->linesize[0], tref1,
145	162k	row << 3, (col << 3) + 4, &b->mv[1][0],,,,, 4, 8, w1, h1, 0);
146	162k	h1 = (h1 + s->ss_v) >> s->ss_v;
147	162k	if (s->ss_h) {
148	47.3k	w1 = (w1 + 1) >> 1;
149	47.3k	uvmv = ROUNDED_DIV_MVx2(b->mv[0][0], b->mv[1][0]);
150	47.3k	mc_chroma_dir(td, mc[4][b->filter][0],
151	47.3k	td->dst[1], td->dst[2], ls_uv,
152	47.3k	ref1->data[1], ref1->linesize[1],
153	47.3k	ref1->data[2], ref1->linesize[2], tref1,
154	47.3k	row << (3 - s->ss_v), col << 2,
155	47.3k	&uvmv,,,,, 4, 8 >> s->ss_v, w1, h1, 0);
156	115k	} else {
157	115k	mc_chroma_dir(td, mc[4][b->filter][0],
158	115k	td->dst[1], td->dst[2], ls_uv,
159	115k	ref1->data[1], ref1->linesize[1],
160	115k	ref1->data[2], ref1->linesize[2], tref1,
161	115k	row << (3 - s->ss_v), col << 3,
162	115k	&b->mv[0][0],,,,, 4, 8 >> s->ss_v, w1, h1, 0);
163	115k	mc_chroma_dir(td, mc[4][b->filter][0],
164	115k	td->dst[1] + 4 * bytesperpixel,
165	115k	td->dst[2] + 4 * bytesperpixel, ls_uv,
166	115k	ref1->data[1], ref1->linesize[1],
167	115k	ref1->data[2], ref1->linesize[2], tref1,
168	115k	row << (3 - s->ss_v), (col << 3) + 4,
169	115k	&b->mv[1][0],,,,, 4, 8 >> s->ss_v, w1, h1, 0);
170	115k	}
171
172	162k	if (b->comp) {
173	76.5k	mc_luma_dir(td, mc[4][b->filter][1], td->dst[0], ls_y,
174	76.5k	ref2->data[0], ref2->linesize[0], tref2,
175	76.5k	row << 3, col << 3, &b->mv[0][1],,,,, 4, 8, w2, h2, 1);
176	76.5k	mc_luma_dir(td, mc[4][b->filter][1], td->dst[0] + 4 * bytesperpixel, ls_y,
177	76.5k	ref2->data[0], ref2->linesize[0], tref2,
178	76.5k	row << 3, (col << 3) + 4, &b->mv[1][1],,,,, 4, 8, w2, h2, 1);
179	76.5k	h2 = (h2 + s->ss_v) >> s->ss_v;
180	76.5k	if (s->ss_h) {
181	18.7k	w2 = (w2 + 1) >> 1;
182	18.7k	uvmv = ROUNDED_DIV_MVx2(b->mv[0][1], b->mv[1][1]);
183	18.7k	mc_chroma_dir(td, mc[4][b->filter][1],
184	18.7k	td->dst[1], td->dst[2], ls_uv,
185	18.7k	ref2->data[1], ref2->linesize[1],
186	18.7k	ref2->data[2], ref2->linesize[2], tref2,
187	18.7k	row << (3 - s->ss_v), col << 2,
188	18.7k	&uvmv,,,,, 4, 8 >> s->ss_v, w2, h2, 1);
189	57.8k	} else {
190	57.8k	mc_chroma_dir(td, mc[4][b->filter][1],
191	57.8k	td->dst[1], td->dst[2], ls_uv,
192	57.8k	ref2->data[1], ref2->linesize[1],
193	57.8k	ref2->data[2], ref2->linesize[2], tref2,
194	57.8k	row << (3 - s->ss_v), col << 3,
195	57.8k	&b->mv[0][1],,,,, 4, 8 >> s->ss_v, w2, h2, 1);
196	57.8k	mc_chroma_dir(td, mc[4][b->filter][1],
197	57.8k	td->dst[1] + 4 * bytesperpixel,
198	57.8k	td->dst[2] + 4 * bytesperpixel, ls_uv,
199	57.8k	ref2->data[1], ref2->linesize[1],
200	57.8k	ref2->data[2], ref2->linesize[2], tref2,
201	57.8k	row << (3 - s->ss_v), (col << 3) + 4,
202	57.8k	&b->mv[1][1],,,,, 4, 8 >> s->ss_v, w2, h2, 1);
203	57.8k	}
204	76.5k	}
205	162k	} else
206	338k	#endif
207	338k	{
208	338k	#if SCALED == 0
209	338k	av_assert2(b->bs == BS_4x4);
210	338k	#endif
211
212		// FIXME if two horizontally adjacent blocks have the same MV,
213		// do a w8 instead of a w4 call
214	338k	mc_luma_dir(td, mc[4][b->filter][0], td->dst[0], ls_y,
215	338k	ref1->data[0], ref1->linesize[0], tref1,
216	338k	row << 3, col << 3, &b->mv[0][0],
217	338k	0, 0, 8, 8, 4, 4, w1, h1, 0);
218	338k	mc_luma_dir(td, mc[4][b->filter][0], td->dst[0] + 4 * bytesperpixel, ls_y,
219	338k	ref1->data[0], ref1->linesize[0], tref1,
220	338k	row << 3, (col << 3) + 4, &b->mv[1][0],
221	338k	4, 0, 8, 8, 4, 4, w1, h1, 0);
222	338k	mc_luma_dir(td, mc[4][b->filter][0],
223	338k	td->dst[0] + 4 * ls_y, ls_y,
224	338k	ref1->data[0], ref1->linesize[0], tref1,
225	338k	(row << 3) + 4, col << 3, &b->mv[2][0],
226	338k	0, 4, 8, 8, 4, 4, w1, h1, 0);
227	338k	mc_luma_dir(td, mc[4][b->filter][0],
228	338k	td->dst[0] + 4 * ls_y + 4 * bytesperpixel, ls_y,
229	338k	ref1->data[0], ref1->linesize[0], tref1,
230	338k	(row << 3) + 4, (col << 3) + 4, &b->mv[3][0],
231	338k	4, 4, 8, 8, 4, 4, w1, h1, 0);
232	338k	if (s->ss_v) {
233	99.5k	h1 = (h1 + 1) >> 1;
234	99.5k	if (s->ss_h) {
235	58.8k	w1 = (w1 + 1) >> 1;
236	58.8k	uvmv = ROUNDED_DIV_MVx4(b->mv[0][0], b->mv[1][0],
237	58.8k	b->mv[2][0], b->mv[3][0]);
238	58.8k	mc_chroma_dir(td, mc[4][b->filter][0],
239	58.8k	td->dst[1], td->dst[2], ls_uv,
240	58.8k	ref1->data[1], ref1->linesize[1],
241	58.8k	ref1->data[2], ref1->linesize[2], tref1,
242	58.8k	row << 2, col << 2,
243	58.8k	&uvmv, 0, 0, 4, 4, 4, 4, w1, h1, 0);
244	58.8k	} else {
245	40.6k	uvmv = ROUNDED_DIV_MVx2(b->mv[0][0], b->mv[2][0]);
246	40.6k	mc_chroma_dir(td, mc[4][b->filter][0],
247	40.6k	td->dst[1], td->dst[2], ls_uv,
248	40.6k	ref1->data[1], ref1->linesize[1],
249	40.6k	ref1->data[2], ref1->linesize[2], tref1,
250	40.6k	row << 2, col << 3,
251	40.6k	&uvmv, 0, 0, 8, 4, 4, 4, w1, h1, 0);
252	40.6k	uvmv = ROUNDED_DIV_MVx2(b->mv[1][0], b->mv[3][0]);
253	40.6k	mc_chroma_dir(td, mc[4][b->filter][0],
254	40.6k	td->dst[1] + 4 * bytesperpixel,
255	40.6k	td->dst[2] + 4 * bytesperpixel, ls_uv,
256	40.6k	ref1->data[1], ref1->linesize[1],
257	40.6k	ref1->data[2], ref1->linesize[2], tref1,
258	40.6k	row << 2, (col << 3) + 4,
259	40.6k	&uvmv, 4, 0, 8, 4, 4, 4, w1, h1, 0);
260	40.6k	}
261	238k	} else {
262	238k	if (s->ss_h) {
263	74.3k	w1 = (w1 + 1) >> 1;
264	74.3k	uvmv = ROUNDED_DIV_MVx2(b->mv[0][0], b->mv[1][0]);
265	74.3k	mc_chroma_dir(td, mc[4][b->filter][0],
266	74.3k	td->dst[1], td->dst[2], ls_uv,
267	74.3k	ref1->data[1], ref1->linesize[1],
268	74.3k	ref1->data[2], ref1->linesize[2], tref1,
269	74.3k	row << 3, col << 2,
270	74.3k	&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	74.3k	uvmv = ROUNDED_DIV_MVx2(b->mv[1][0], b->mv[2][0]);
275	74.3k	mc_chroma_dir(td, mc[4][b->filter][0],
276	74.3k	td->dst[1] + 4 * ls_uv, td->dst[2] + 4 * ls_uv, ls_uv,
277	74.3k	ref1->data[1], ref1->linesize[1],
278	74.3k	ref1->data[2], ref1->linesize[2], tref1,
279	74.3k	(row << 3) + 4, col << 2,
280	74.3k	&uvmv, 0, 4, 4, 8, 4, 4, w1, h1, 0);
281	164k	} else {
282	164k	mc_chroma_dir(td, mc[4][b->filter][0],
283	164k	td->dst[1], td->dst[2], ls_uv,
284	164k	ref1->data[1], ref1->linesize[1],
285	164k	ref1->data[2], ref1->linesize[2], tref1,
286	164k	row << 3, col << 3,
287	164k	&b->mv[0][0], 0, 0, 8, 8, 4, 4, w1, h1, 0);
288	164k	mc_chroma_dir(td, mc[4][b->filter][0],
289	164k	td->dst[1] + 4 * bytesperpixel,
290	164k	td->dst[2] + 4 * bytesperpixel, ls_uv,
291	164k	ref1->data[1], ref1->linesize[1],
292	164k	ref1->data[2], ref1->linesize[2], tref1,
293	164k	row << 3, (col << 3) + 4,
294	164k	&b->mv[1][0], 4, 0, 8, 8, 4, 4, w1, h1, 0);
295	164k	mc_chroma_dir(td, mc[4][b->filter][0],
296	164k	td->dst[1] + 4 * ls_uv, td->dst[2] + 4 * ls_uv, ls_uv,
297	164k	ref1->data[1], ref1->linesize[1],
298	164k	ref1->data[2], ref1->linesize[2], tref1,
299	164k	(row << 3) + 4, col << 3,
300	164k	&b->mv[2][0], 0, 4, 8, 8, 4, 4, w1, h1, 0);
301	164k	mc_chroma_dir(td, mc[4][b->filter][0],
302	164k	td->dst[1] + 4 * ls_uv + 4 * bytesperpixel,
303	164k	td->dst[2] + 4 * ls_uv + 4 * bytesperpixel, ls_uv,
304	164k	ref1->data[1], ref1->linesize[1],
305	164k	ref1->data[2], ref1->linesize[2], tref1,
306	164k	(row << 3) + 4, (col << 3) + 4,
307	164k	&b->mv[3][0], 4, 4, 8, 8, 4, 4, w1, h1, 0);
308	164k	}
309	238k	}
310
311	338k	if (b->comp) {
312	128k	mc_luma_dir(td, mc[4][b->filter][1], td->dst[0], ls_y,
313	128k	ref2->data[0], ref2->linesize[0], tref2,
314	128k	row << 3, col << 3, &b->mv[0][1], 0, 0, 8, 8, 4, 4, w2, h2, 1);
315	128k	mc_luma_dir(td, mc[4][b->filter][1], td->dst[0] + 4 * bytesperpixel, ls_y,
316	128k	ref2->data[0], ref2->linesize[0], tref2,
317	128k	row << 3, (col << 3) + 4, &b->mv[1][1], 4, 0, 8, 8, 4, 4, w2, h2, 1);
318	128k	mc_luma_dir(td, mc[4][b->filter][1],
319	128k	td->dst[0] + 4 * ls_y, ls_y,
320	128k	ref2->data[0], ref2->linesize[0], tref2,
321	128k	(row << 3) + 4, col << 3, &b->mv[2][1], 0, 4, 8, 8, 4, 4, w2, h2, 1);
322	128k	mc_luma_dir(td, mc[4][b->filter][1],
323	128k	td->dst[0] + 4 * ls_y + 4 * bytesperpixel, ls_y,
324	128k	ref2->data[0], ref2->linesize[0], tref2,
325	128k	(row << 3) + 4, (col << 3) + 4, &b->mv[3][1], 4, 4, 8, 8, 4, 4, w2, h2, 1);
326	128k	if (s->ss_v) {
327	50.8k	h2 = (h2 + 1) >> 1;
328	50.8k	if (s->ss_h) {
329	22.4k	w2 = (w2 + 1) >> 1;
330	22.4k	uvmv = ROUNDED_DIV_MVx4(b->mv[0][1], b->mv[1][1],
331	22.4k	b->mv[2][1], b->mv[3][1]);
332	22.4k	mc_chroma_dir(td, mc[4][b->filter][1],
333	22.4k	td->dst[1], td->dst[2], ls_uv,
334	22.4k	ref2->data[1], ref2->linesize[1],
335	22.4k	ref2->data[2], ref2->linesize[2], tref2,
336	22.4k	row << 2, col << 2,
337	22.4k	&uvmv, 0, 0, 4, 4, 4, 4, w2, h2, 1);
338	28.4k	} else {
339	28.4k	uvmv = ROUNDED_DIV_MVx2(b->mv[0][1], b->mv[2][1]);
340	28.4k	mc_chroma_dir(td, mc[4][b->filter][1],
341	28.4k	td->dst[1], td->dst[2], ls_uv,
342	28.4k	ref2->data[1], ref2->linesize[1],
343	28.4k	ref2->data[2], ref2->linesize[2], tref2,
344	28.4k	row << 2, col << 3,
345	28.4k	&uvmv, 0, 0, 8, 4, 4, 4, w2, h2, 1);
346	28.4k	uvmv = ROUNDED_DIV_MVx2(b->mv[1][1], b->mv[3][1]);
347	28.4k	mc_chroma_dir(td, mc[4][b->filter][1],
348	28.4k	td->dst[1] + 4 * bytesperpixel,
349	28.4k	td->dst[2] + 4 * bytesperpixel, ls_uv,
350	28.4k	ref2->data[1], ref2->linesize[1],
351	28.4k	ref2->data[2], ref2->linesize[2], tref2,
352	28.4k	row << 2, (col << 3) + 4,
353	28.4k	&uvmv, 4, 0, 8, 4, 4, 4, w2, h2, 1);
354	28.4k	}
355	77.1k	} else {
356	77.1k	if (s->ss_h) {
357	19.9k	w2 = (w2 + 1) >> 1;
358	19.9k	uvmv = ROUNDED_DIV_MVx2(b->mv[0][1], b->mv[1][1]);
359	19.9k	mc_chroma_dir(td, mc[4][b->filter][1],
360	19.9k	td->dst[1], td->dst[2], ls_uv,
361	19.9k	ref2->data[1], ref2->linesize[1],
362	19.9k	ref2->data[2], ref2->linesize[2], tref2,
363	19.9k	row << 3, col << 2,
364	19.9k	&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	19.9k	uvmv = ROUNDED_DIV_MVx2(b->mv[1][1], b->mv[2][1]);
369	19.9k	mc_chroma_dir(td, mc[4][b->filter][1],
370	19.9k	td->dst[1] + 4 * ls_uv, td->dst[2] + 4 * ls_uv, ls_uv,
371	19.9k	ref2->data[1], ref2->linesize[1],
372	19.9k	ref2->data[2], ref2->linesize[2], tref2,
373	19.9k	(row << 3) + 4, col << 2,
374	19.9k	&uvmv, 0, 4, 4, 8, 4, 4, w2, h2, 1);
375	57.2k	} else {
376	57.2k	mc_chroma_dir(td, mc[4][b->filter][1],
377	57.2k	td->dst[1], td->dst[2], ls_uv,
378	57.2k	ref2->data[1], ref2->linesize[1],
379	57.2k	ref2->data[2], ref2->linesize[2], tref2,
380	57.2k	row << 3, col << 3,
381	57.2k	&b->mv[0][1], 0, 0, 8, 8, 4, 4, w2, h2, 1);
382	57.2k	mc_chroma_dir(td, mc[4][b->filter][1],
383	57.2k	td->dst[1] + 4 * bytesperpixel,
384	57.2k	td->dst[2] + 4 * bytesperpixel, ls_uv,
385	57.2k	ref2->data[1], ref2->linesize[1],
386	57.2k	ref2->data[2], ref2->linesize[2], tref2,
387	57.2k	row << 3, (col << 3) + 4,
388	57.2k	&b->mv[1][1], 4, 0, 8, 8, 4, 4, w2, h2, 1);
389	57.2k	mc_chroma_dir(td, mc[4][b->filter][1],
390	57.2k	td->dst[1] + 4 * ls_uv, td->dst[2] + 4 * ls_uv, ls_uv,
391	57.2k	ref2->data[1], ref2->linesize[1],
392	57.2k	ref2->data[2], ref2->linesize[2], tref2,
393	57.2k	(row << 3) + 4, col << 3,
394	57.2k	&b->mv[2][1], 0, 4, 8, 8, 4, 4, w2, h2, 1);
395	57.2k	mc_chroma_dir(td, mc[4][b->filter][1],
396	57.2k	td->dst[1] + 4 * ls_uv + 4 * bytesperpixel,
397	57.2k	td->dst[2] + 4 * ls_uv + 4 * bytesperpixel, ls_uv,
398	57.2k	ref2->data[1], ref2->linesize[1],
399	57.2k	ref2->data[2], ref2->linesize[2], tref2,
400	57.2k	(row << 3) + 4, (col << 3) + 4,
401	57.2k	&b->mv[3][1], 4, 4, 8, 8, 4, 4, w2, h2, 1);
402	57.2k	}
403	77.1k	}
404	128k	}
405	338k	}
406	1.65M	} else {
407	1.65M	int bwl = bwlog_tab[0][b->bs];
408	1.65M	int bw = ff_vp9_bwh_tab[0][b->bs][0] * 4;
409	1.65M	int bh = ff_vp9_bwh_tab[0][b->bs][1] * 4;
410	1.65M	int uvbw = ff_vp9_bwh_tab[s->ss_h][b->bs][0] * 4;
411	1.65M	int uvbh = ff_vp9_bwh_tab[s->ss_v][b->bs][1] * 4;
412
413	1.65M	mc_luma_dir(td, mc[bwl][b->filter][0], td->dst[0], ls_y,
414	1.65M	ref1->data[0], ref1->linesize[0], tref1,
415	1.65M	row << 3, col << 3, &b->mv[0][0], 0, 0, bw, bh, bw, bh, w1, h1, 0);
416	1.65M	w1 = (w1 + s->ss_h) >> s->ss_h;
417	1.65M	h1 = (h1 + s->ss_v) >> s->ss_v;
418	1.65M	mc_chroma_dir(td, mc[bwl + s->ss_h][b->filter][0],
419	1.65M	td->dst[1], td->dst[2], ls_uv,
420	1.65M	ref1->data[1], ref1->linesize[1],
421	1.65M	ref1->data[2], ref1->linesize[2], tref1,
422	1.65M	row << (3 - s->ss_v), col << (3 - s->ss_h),
423	1.65M	&b->mv[0][0], 0, 0, uvbw, uvbh, uvbw, uvbh, w1, h1, 0);
424
425	1.65M	if (b->comp) {
426	822k	mc_luma_dir(td, mc[bwl][b->filter][1], td->dst[0], ls_y,
427	822k	ref2->data[0], ref2->linesize[0], tref2,
428	822k	row << 3, col << 3, &b->mv[0][1], 0, 0, bw, bh, bw, bh, w2, h2, 1);
429	822k	w2 = (w2 + s->ss_h) >> s->ss_h;
430	822k	h2 = (h2 + s->ss_v) >> s->ss_v;
431	822k	mc_chroma_dir(td, mc[bwl + s->ss_h][b->filter][1],
432	822k	td->dst[1], td->dst[2], ls_uv,
433	822k	ref2->data[1], ref2->linesize[1],
434	822k	ref2->data[2], ref2->linesize[2], tref2,
435	822k	row << (3 - s->ss_v), col << (3 - s->ss_h),
436	822k	&b->mv[0][1], 0, 0, uvbw, uvbh, uvbw, uvbh, w2, h2, 1);
437	822k	}
438	1.65M	}
439	2.42M	}

Coverage Report

Created: 2024-09-06 07:53