diff options
Diffstat (limited to 'libavcodec/vc1_mc.c')
-rw-r--r-- | libavcodec/vc1_mc.c | 557 |
1 files changed, 244 insertions, 313 deletions
diff --git a/libavcodec/vc1_mc.c b/libavcodec/vc1_mc.c index e3a75d5..569aad0 100644 --- a/libavcodec/vc1_mc.c +++ b/libavcodec/vc1_mc.c @@ -4,20 +4,20 @@ * Copyright (c) 2006-2007 Konstantin Shishkov * Partly based on vc9.c (c) 2005 Anonymous, Alex Beregszaszi, Michael Niedermayer * - * This file is part of Libav. + * This file is part of FFmpeg. * - * Libav is free software; you can redistribute it and/or + * 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. * - * Libav is distributed in the hope that it will be useful, + * 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 Libav; if not, write to the Free Software + * License along with FFmpeg; if not, write to the Free Software * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA */ @@ -32,6 +32,140 @@ #include "mpegvideo.h" #include "vc1.h" +static av_always_inline void vc1_scale_luma(uint8_t *srcY, + int k, int linesize) +{ + int i, j; + for (j = 0; j < k; j++) { + for (i = 0; i < k; i++) + srcY[i] = ((srcY[i] - 128) >> 1) + 128; + srcY += linesize; + } +} + +static av_always_inline void vc1_scale_chroma(uint8_t *srcU, uint8_t *srcV, + int k, int uvlinesize) +{ + int i, j; + for (j = 0; j < k; j++) { + for (i = 0; i < k; i++) { + srcU[i] = ((srcU[i] - 128) >> 1) + 128; + srcV[i] = ((srcV[i] - 128) >> 1) + 128; + } + srcU += uvlinesize; + srcV += uvlinesize; + } +} + +static av_always_inline void vc1_lut_scale_luma(uint8_t *srcY, + uint8_t *lut1, uint8_t *lut2, + int k, int linesize) +{ + int i, j; + + for (j = 0; j < k; j += 2) { + for (i = 0; i < k; i++) + srcY[i] = lut1[srcY[i]]; + srcY += linesize; + + if (j + 1 == k) + break; + + for (i = 0; i < k; i++) + srcY[i] = lut2[srcY[i]]; + srcY += linesize; + } +} + +static av_always_inline void vc1_lut_scale_chroma(uint8_t *srcU, uint8_t *srcV, + uint8_t *lut1, uint8_t *lut2, + int k, int uvlinesize) +{ + int i, j; + + for (j = 0; j < k; j += 2) { + for (i = 0; i < k; i++) { + srcU[i] = lut1[srcU[i]]; + srcV[i] = lut1[srcV[i]]; + } + srcU += uvlinesize; + srcV += uvlinesize; + + if (j + 1 == k) + break; + + for (i = 0; i < k; i++) { + srcU[i] = lut2[srcU[i]]; + srcV[i] = lut2[srcV[i]]; + } + srcU += uvlinesize; + srcV += uvlinesize; + } +} + +static const uint8_t popcount4[16] = { 0, 1, 1, 2, 1, 2, 2, 3, 1, 2, 2, 3, 2, 3, 3, 4 }; + +static av_always_inline int get_luma_mv(VC1Context *v, int dir, int16_t *tx, int16_t *ty) +{ + MpegEncContext *s = &v->s; + int idx = v->mv_f[dir][s->block_index[0] + v->blocks_off] | + (v->mv_f[dir][s->block_index[1] + v->blocks_off] << 1) | + (v->mv_f[dir][s->block_index[2] + v->blocks_off] << 2) | + (v->mv_f[dir][s->block_index[3] + v->blocks_off] << 3); + static const uint8_t index2[16] = { 0, 0, 0, 0x23, 0, 0x13, 0x03, 0, 0, 0x12, 0x02, 0, 0x01, 0, 0, 0 }; + int opp_count = popcount4[idx]; + + switch (opp_count) { + case 0: + case 4: + *tx = median4(s->mv[dir][0][0], s->mv[dir][1][0], s->mv[dir][2][0], s->mv[dir][3][0]); + *ty = median4(s->mv[dir][0][1], s->mv[dir][1][1], s->mv[dir][2][1], s->mv[dir][3][1]); + break; + case 1: + *tx = mid_pred(s->mv[dir][idx < 2][0], s->mv[dir][1 + (idx < 4)][0], s->mv[dir][2 + (idx < 8)][0]); + *ty = mid_pred(s->mv[dir][idx < 2][1], s->mv[dir][1 + (idx < 4)][1], s->mv[dir][2 + (idx < 8)][1]); + break; + case 3: + *tx = mid_pred(s->mv[dir][idx > 0xd][0], s->mv[dir][1 + (idx > 0xb)][0], s->mv[dir][2 + (idx > 0x7)][0]); + *ty = mid_pred(s->mv[dir][idx > 0xd][1], s->mv[dir][1 + (idx > 0xb)][1], s->mv[dir][2 + (idx > 0x7)][1]); + break; + case 2: + *tx = (s->mv[dir][index2[idx] >> 4][0] + s->mv[dir][index2[idx] & 0xf][0]) / 2; + *ty = (s->mv[dir][index2[idx] >> 4][1] + s->mv[dir][index2[idx] & 0xf][1]) / 2; + break; + } + return opp_count; +} + +static av_always_inline int get_chroma_mv(VC1Context *v, int dir, int16_t *tx, int16_t *ty) +{ + MpegEncContext *s = &v->s; + int idx = !v->mb_type[0][s->block_index[0]] | + (!v->mb_type[0][s->block_index[1]] << 1) | + (!v->mb_type[0][s->block_index[2]] << 2) | + (!v->mb_type[0][s->block_index[3]] << 3); + static const uint8_t index2[16] = { 0, 0, 0, 0x01, 0, 0x02, 0x12, 0, 0, 0x03, 0x13, 0, 0x23, 0, 0, 0 }; + int valid_count = popcount4[idx]; + + switch (valid_count) { + case 4: + *tx = median4(s->mv[dir][0][0], s->mv[dir][1][0], s->mv[dir][2][0], s->mv[dir][3][0]); + *ty = median4(s->mv[dir][0][1], s->mv[dir][1][1], s->mv[dir][2][1], s->mv[dir][3][1]); + break; + case 3: + *tx = mid_pred(s->mv[dir][idx > 0xd][0], s->mv[dir][1 + (idx > 0xb)][0], s->mv[dir][2 + (idx > 0x7)][0]); + *ty = mid_pred(s->mv[dir][idx > 0xd][1], s->mv[dir][1 + (idx > 0xb)][1], s->mv[dir][2 + (idx > 0x7)][1]); + break; + case 2: + *tx = (s->mv[dir][index2[idx] >> 4][0] + s->mv[dir][index2[idx] & 0xf][0]) / 2; + *ty = (s->mv[dir][index2[idx] >> 4][1] + s->mv[dir][index2[idx] & 0xf][1]) / 2; + break; + default: + return 0; + } + return valid_count; +} + /** Do motion compensation over 1 macroblock * Mostly adapted hpel_motion and qpel_motion from mpegvideo.c */ @@ -85,7 +219,7 @@ void ff_vc1_mc_1mv(VC1Context *v, int dir) srcV = s->current_picture.f->data[2]; luty = v->curr_luty; lutuv = v->curr_lutuv; - use_ic = v->curr_use_ic; + use_ic = *v->curr_use_ic; } else { srcY = s->last_picture.f->data[0]; srcU = s->last_picture.f->data[1]; @@ -145,81 +279,51 @@ void ff_vc1_mc_1mv(VC1Context *v, int dir) || s->h_edge_pos < 22 || v_edge_pos < 22 || (unsigned)(src_x - s->mspel) > s->h_edge_pos - (mx&3) - 16 - s->mspel * 3 || (unsigned)(src_y - 1) > v_edge_pos - (my&3) - 16 - 3) { - uint8_t *uvbuf = s->edge_emu_buffer + 19 * s->linesize; + uint8_t *ubuf = s->edge_emu_buffer + 19 * s->linesize; + uint8_t *vbuf = ubuf + 9 * s->uvlinesize; + const int k = 17 + s->mspel * 2; srcY -= s->mspel * (1 + s->linesize); s->vdsp.emulated_edge_mc(s->edge_emu_buffer, srcY, s->linesize, s->linesize, - 17 + s->mspel * 2, 17 + s->mspel * 2, + k, k, src_x - s->mspel, src_y - s->mspel, s->h_edge_pos, v_edge_pos); srcY = s->edge_emu_buffer; - s->vdsp.emulated_edge_mc(uvbuf, srcU, + s->vdsp.emulated_edge_mc(ubuf, srcU, s->uvlinesize, s->uvlinesize, 8 + 1, 8 + 1, - uvsrc_x, uvsrc_y, s->h_edge_pos >> 1, v_edge_pos >> 1); - s->vdsp.emulated_edge_mc(uvbuf + 16, srcV, + uvsrc_x, uvsrc_y, + s->h_edge_pos >> 1, v_edge_pos >> 1); + s->vdsp.emulated_edge_mc(vbuf, srcV, s->uvlinesize, s->uvlinesize, 8 + 1, 8 + 1, - uvsrc_x, uvsrc_y, s->h_edge_pos >> 1, v_edge_pos >> 1); - srcU = uvbuf; - srcV = uvbuf + 16; + uvsrc_x, uvsrc_y, + s->h_edge_pos >> 1, v_edge_pos >> 1); + srcU = ubuf; + srcV = vbuf; /* if we deal with range reduction we need to scale source blocks */ if (v->rangeredfrm) { - int i, j; - uint8_t *src, *src2; - - src = srcY; - for (j = 0; j < 17 + s->mspel * 2; j++) { - for (i = 0; i < 17 + s->mspel * 2; i++) - src[i] = ((src[i] - 128) >> 1) + 128; - src += s->linesize; - } - src = srcU; - src2 = srcV; - for (j = 0; j < 9; j++) { - for (i = 0; i < 9; i++) { - src[i] = ((src[i] - 128) >> 1) + 128; - src2[i] = ((src2[i] - 128) >> 1) + 128; - } - src += s->uvlinesize; - src2 += s->uvlinesize; - } + vc1_scale_luma(srcY, k, s->linesize); + vc1_scale_chroma(srcU, srcV, 9, s->uvlinesize); } /* if we deal with intensity compensation we need to scale source blocks */ if (use_ic) { - int i, j; - uint8_t *src, *src2; - - src = srcY; - for (j = 0; j < 17 + s->mspel * 2; j++) { - int f = v->field_mode ? v->ref_field_type[dir] : ((j + src_y - s->mspel) & 1) ; - for (i = 0; i < 17 + s->mspel * 2; i++) - src[i] = luty[f][src[i]]; - src += s->linesize; - } - src = srcU; - src2 = srcV; - for (j = 0; j < 9; j++) { - int f = v->field_mode ? v->ref_field_type[dir] : ((j + uvsrc_y) & 1); - for (i = 0; i < 9; i++) { - src[i] = lutuv[f][src[i]]; - src2[i] = lutuv[f][src2[i]]; - } - src += s->uvlinesize; - src2 += s->uvlinesize; - } + vc1_lut_scale_luma(srcY, + luty[v->field_mode ? v->ref_field_type[dir] : ((0 + src_y - s->mspel) & 1)], + luty[v->field_mode ? v->ref_field_type[dir] : ((1 + src_y - s->mspel) & 1)], + k, s->linesize); + vc1_lut_scale_chroma(srcU, srcV, + lutuv[v->field_mode ? v->ref_field_type[dir] : ((0 + uvsrc_y) & 1)], + lutuv[v->field_mode ? v->ref_field_type[dir] : ((1 + uvsrc_y) & 1)], + 9, s->uvlinesize); } srcY += s->mspel * (1 + s->linesize); } if (s->mspel) { dxy = ((my & 3) << 2) | (mx & 3); - v->vc1dsp.put_vc1_mspel_pixels_tab[dxy](s->dest[0] , srcY , s->linesize, v->rnd); - v->vc1dsp.put_vc1_mspel_pixels_tab[dxy](s->dest[0] + 8, srcY + 8, s->linesize, v->rnd); - srcY += s->linesize * 8; - v->vc1dsp.put_vc1_mspel_pixels_tab[dxy](s->dest[0] + 8 * s->linesize , srcY , s->linesize, v->rnd); - v->vc1dsp.put_vc1_mspel_pixels_tab[dxy](s->dest[0] + 8 * s->linesize + 8, srcY + 8, s->linesize, v->rnd); + v->vc1dsp.put_vc1_mspel_pixels_tab[0][dxy](s->dest[0], srcY, s->linesize, v->rnd); } else { // hpel mc - always used for luma dxy = (my & 2) | ((mx & 2) >> 1); if (!v->rnd) @@ -241,17 +345,6 @@ void ff_vc1_mc_1mv(VC1Context *v, int dir) } } -static inline int median4(int a, int b, int c, int d) -{ - if (a < b) { - if (c < d) return (FFMIN(b, d) + FFMAX(a, c)) / 2; - else return (FFMIN(b, c) + FFMAX(a, d)) / 2; - } else { - if (c < d) return (FFMIN(a, d) + FFMAX(b, c)) / 2; - else return (FFMIN(a, c) + FFMAX(b, d)) / 2; - } -} - /** Do motion compensation for 4-MV macroblock - luminance block */ void ff_vc1_mc_4mv_luma(VC1Context *v, int n, int dir, int avg) @@ -277,7 +370,7 @@ void ff_vc1_mc_4mv_luma(VC1Context *v, int n, int dir, int avg) if (v->field_mode && (v->cur_field_type != v->ref_field_type[dir]) && v->second_field) { srcY = s->current_picture.f->data[0]; luty = v->curr_luty; - use_ic = v->curr_use_ic; + use_ic = *v->curr_use_ic; } else { srcY = s->last_picture.f->data[0]; luty = v->last_luty; @@ -300,35 +393,10 @@ void ff_vc1_mc_4mv_luma(VC1Context *v, int n, int dir, int avg) } if (s->pict_type == AV_PICTURE_TYPE_P && n == 3 && v->field_mode) { - int same_count = 0, opp_count = 0, k; - int chosen_mv[2][4][2], f; - int tx = 0, ty = 0; - for (k = 0; k < 4; k++) { - f = v->mv_f[0][s->block_index[k] + v->blocks_off]; - chosen_mv[f][f ? opp_count : same_count][0] = s->mv[0][k][0]; - chosen_mv[f][f ? opp_count : same_count][1] = s->mv[0][k][1]; - opp_count += f; - same_count += 1 - f; - } - f = opp_count > same_count; - switch (f ? opp_count : same_count) { - case 4: - tx = median4(chosen_mv[f][0][0], chosen_mv[f][1][0], - chosen_mv[f][2][0], chosen_mv[f][3][0]); - ty = median4(chosen_mv[f][0][1], chosen_mv[f][1][1], - chosen_mv[f][2][1], chosen_mv[f][3][1]); - break; - case 3: - tx = mid_pred(chosen_mv[f][0][0], chosen_mv[f][1][0], chosen_mv[f][2][0]); - ty = mid_pred(chosen_mv[f][0][1], chosen_mv[f][1][1], chosen_mv[f][2][1]); - break; - case 2: - tx = (chosen_mv[f][0][0] + chosen_mv[f][1][0]) / 2; - ty = (chosen_mv[f][0][1] + chosen_mv[f][1][1]) / 2; - break; - } - s->current_picture.motion_val[1][s->block_index[0] + v->blocks_off][0] = tx; - s->current_picture.motion_val[1][s->block_index[0] + v->blocks_off][1] = ty; + int opp_count = get_luma_mv(v, 0, + &s->current_picture.motion_val[1][s->block_index[0] + v->blocks_off][0], + &s->current_picture.motion_val[1][s->block_index[0] + v->blocks_off][1]); + int k, f = opp_count > 2; for (k = 0; k < 4; k++) v->mv_f[1][s->block_index[k] + v->blocks_off] = f; } @@ -384,46 +452,36 @@ void ff_vc1_mc_4mv_luma(VC1Context *v, int n, int dir, int avg) if (v->field_mode && v->ref_field_type[dir]) srcY += s->current_picture_ptr->f->linesize[0]; - if (fieldmv && !(src_y & 1)) - v_edge_pos--; - if (fieldmv && (src_y & 1) && src_y < 4) - src_y--; + if (fieldmv) { + if (!(src_y & 1)) + v_edge_pos--; + else + src_y -= (src_y < 4); + } if (v->rangeredfrm || use_ic || s->h_edge_pos < 13 || v_edge_pos < 23 || (unsigned)(src_x - s->mspel) > s->h_edge_pos - (mx & 3) - 8 - s->mspel * 2 || (unsigned)(src_y - (s->mspel << fieldmv)) > v_edge_pos - (my & 3) - ((8 + s->mspel * 2) << fieldmv)) { + const int k = 9 + s->mspel * 2; + srcY -= s->mspel * (1 + (s->linesize << fieldmv)); /* check emulate edge stride and offset */ s->vdsp.emulated_edge_mc(s->edge_emu_buffer, srcY, s->linesize, s->linesize, - 9 + s->mspel * 2, (9 + s->mspel * 2) << fieldmv, + k, k << fieldmv, src_x - s->mspel, src_y - (s->mspel << fieldmv), s->h_edge_pos, v_edge_pos); srcY = s->edge_emu_buffer; /* if we deal with range reduction we need to scale source blocks */ if (v->rangeredfrm) { - int i, j; - uint8_t *src; - - src = srcY; - for (j = 0; j < 9 + s->mspel * 2; j++) { - for (i = 0; i < 9 + s->mspel * 2; i++) - src[i] = ((src[i] - 128) >> 1) + 128; - src += s->linesize << fieldmv; - } + vc1_scale_luma(srcY, k, s->linesize << fieldmv); } /* if we deal with intensity compensation we need to scale source blocks */ if (use_ic) { - int i, j; - uint8_t *src; - - src = srcY; - for (j = 0; j < 9 + s->mspel * 2; j++) { - int f = v->field_mode ? v->ref_field_type[dir] : (((j<<fieldmv)+src_y - (s->mspel << fieldmv)) & 1); - for (i = 0; i < 9 + s->mspel * 2; i++) - src[i] = luty[f][src[i]]; - src += s->linesize << fieldmv; - } + vc1_lut_scale_luma(srcY, + luty[v->field_mode ? v->ref_field_type[dir] : (((0<<fieldmv)+src_y - (s->mspel << fieldmv)) & 1)], + luty[v->field_mode ? v->ref_field_type[dir] : (((1<<fieldmv)+src_y - (s->mspel << fieldmv)) & 1)], + k, s->linesize << fieldmv); } srcY += s->mspel * (1 + (s->linesize << fieldmv)); } @@ -431,9 +489,9 @@ void ff_vc1_mc_4mv_luma(VC1Context *v, int n, int dir, int avg) if (s->mspel) { dxy = ((my & 3) << 2) | (mx & 3); if (avg) - v->vc1dsp.avg_vc1_mspel_pixels_tab[dxy](s->dest[0] + off, srcY, s->linesize << fieldmv, v->rnd); + v->vc1dsp.avg_vc1_mspel_pixels_tab[1][dxy](s->dest[0] + off, srcY, s->linesize << fieldmv, v->rnd); else - v->vc1dsp.put_vc1_mspel_pixels_tab[dxy](s->dest[0] + off, srcY, s->linesize << fieldmv, v->rnd); + v->vc1dsp.put_vc1_mspel_pixels_tab[1][dxy](s->dest[0] + off, srcY, s->linesize << fieldmv, v->rnd); } else { // hpel mc - always used for luma dxy = (my & 2) | ((mx & 2) >> 1); if (!v->rnd) @@ -443,59 +501,6 @@ void ff_vc1_mc_4mv_luma(VC1Context *v, int n, int dir, int avg) } } -static av_always_inline int get_chroma_mv(int *mvx, int *mvy, int *a, int flag, int *tx, int *ty) -{ - int idx, i; - static const int count[16] = { 0, 1, 1, 2, 1, 2, 2, 3, 1, 2, 2, 3, 2, 3, 3, 4}; - - idx = ((a[3] != flag) << 3) - | ((a[2] != flag) << 2) - | ((a[1] != flag) << 1) - | (a[0] != flag); - if (!idx) { - *tx = median4(mvx[0], mvx[1], mvx[2], mvx[3]); - *ty = median4(mvy[0], mvy[1], mvy[2], mvy[3]); - return 4; - } else if (count[idx] == 1) { - switch (idx) { - case 0x1: - *tx = mid_pred(mvx[1], mvx[2], mvx[3]); - *ty = mid_pred(mvy[1], mvy[2], mvy[3]); - return 3; - case 0x2: - *tx = mid_pred(mvx[0], mvx[2], mvx[3]); - *ty = mid_pred(mvy[0], mvy[2], mvy[3]); - return 3; - case 0x4: - *tx = mid_pred(mvx[0], mvx[1], mvx[3]); - *ty = mid_pred(mvy[0], mvy[1], mvy[3]); - return 3; - case 0x8: - *tx = mid_pred(mvx[0], mvx[1], mvx[2]); - *ty = mid_pred(mvy[0], mvy[1], mvy[2]); - return 3; - } - } else if (count[idx] == 2) { - int t1 = 0, t2 = 0; - for (i = 0; i < 3; i++) - if (!a[i]) { - t1 = i; - break; - } - for (i = t1 + 1; i < 4; i++) - if (!a[i]) { - t2 = i; - break; - } - *tx = (mvx[t1] + mvx[t2]) / 2; - *ty = (mvy[t1] + mvy[t2]) / 2; - return 2; - } else { - return 0; - } - return -1; -} - /** Do motion compensation for 4-MV macroblock - both chroma blocks */ void ff_vc1_mc_4mv_chroma(VC1Context *v, int dir) @@ -504,10 +509,8 @@ void ff_vc1_mc_4mv_chroma(VC1Context *v, int dir) H264ChromaContext *h264chroma = &v->h264chroma; uint8_t *srcU, *srcV; int uvmx, uvmy, uvsrc_x, uvsrc_y; - int k, tx = 0, ty = 0; - int mvx[4], mvy[4], intra[4], mv_f[4]; - int valid_count; - int chroma_ref_type = v->cur_field_type; + int16_t tx, ty; + int chroma_ref_type; int v_edge_pos = s->v_edge_pos >> v->field_mode; uint8_t (*lutuv)[256]; int use_ic; @@ -517,31 +520,19 @@ void ff_vc1_mc_4mv_chroma(VC1Context *v, int dir) if (s->flags & CODEC_FLAG_GRAY) return; - for (k = 0; k < 4; k++) { - mvx[k] = s->mv[dir][k][0]; - mvy[k] = s->mv[dir][k][1]; - intra[k] = v->mb_type[0][s->block_index[k]]; - if (v->field_mode) - mv_f[k] = v->mv_f[dir][s->block_index[k] + v->blocks_off]; - } - /* calculate chroma MV vector from four luma MVs */ - if (!v->field_mode || (v->field_mode && !v->numref)) { - valid_count = get_chroma_mv(mvx, mvy, intra, 0, &tx, &ty); - chroma_ref_type = v->reffield; + if (!v->field_mode || !v->numref) { + int valid_count = get_chroma_mv(v, dir, &tx, &ty); if (!valid_count) { s->current_picture.motion_val[1][s->block_index[0] + v->blocks_off][0] = 0; s->current_picture.motion_val[1][s->block_index[0] + v->blocks_off][1] = 0; v->luma_mv[s->mb_x][0] = v->luma_mv[s->mb_x][1] = 0; return; //no need to do MC for intra blocks } + chroma_ref_type = v->ref_field_type[dir]; } else { - int dominant = 0; - if (mv_f[0] + mv_f[1] + mv_f[2] + mv_f[3] > 2) - dominant = 1; - valid_count = get_chroma_mv(mvx, mvy, mv_f, dominant, &tx, &ty); - if (dominant) - chroma_ref_type = !v->cur_field_type; + int opp_count = get_luma_mv(v, dir, &tx, &ty); + chroma_ref_type = v->cur_field_type ^ (opp_count > 2); } if (v->field_mode && chroma_ref_type == 1 && v->cur_field_type == 1 && !v->s.last_picture.f->data[0]) return; @@ -577,7 +568,7 @@ void ff_vc1_mc_4mv_chroma(VC1Context *v, int dir) srcU = s->current_picture.f->data[1]; srcV = s->current_picture.f->data[2]; lutuv = v->curr_lutuv; - use_ic = v->curr_use_ic; + use_ic = *v->curr_use_ic; } else { srcU = s->last_picture.f->data[1]; srcV = s->last_picture.f->data[2]; @@ -623,36 +614,14 @@ void ff_vc1_mc_4mv_chroma(VC1Context *v, int dir) /* if we deal with range reduction we need to scale source blocks */ if (v->rangeredfrm) { - int i, j; - uint8_t *src, *src2; - - src = srcU; - src2 = srcV; - for (j = 0; j < 9; j++) { - for (i = 0; i < 9; i++) { - src[i] = ((src[i] - 128) >> 1) + 128; - src2[i] = ((src2[i] - 128) >> 1) + 128; - } - src += s->uvlinesize; - src2 += s->uvlinesize; - } + vc1_scale_chroma(srcU, srcV, 9, s->uvlinesize); } /* if we deal with intensity compensation we need to scale source blocks */ if (use_ic) { - int i, j; - uint8_t *src, *src2; - - src = srcU; - src2 = srcV; - for (j = 0; j < 9; j++) { - int f = v->field_mode ? chroma_ref_type : ((j + uvsrc_y) & 1); - for (i = 0; i < 9; i++) { - src[i] = lutuv[f][src[i]]; - src2[i] = lutuv[f][src2[i]]; - } - src += s->uvlinesize; - src2 += s->uvlinesize; - } + vc1_lut_scale_chroma(srcU, srcV, + lutuv[v->field_mode ? chroma_ref_type : ((0 + uvsrc_y) & 1)], + lutuv[v->field_mode ? chroma_ref_type : ((1 + uvsrc_y) & 1)], + 9, s->uvlinesize); } } @@ -679,7 +648,7 @@ void ff_vc1_mc_4mv_chroma4(VC1Context *v, int dir, int dir2, int avg) int uvmx_field[4], uvmy_field[4]; int i, off, tx, ty; int fieldmv = v->blk_mv_type[s->block_index[0]]; - static const int s_rndtblfield[16] = { 0, 0, 1, 2, 4, 4, 5, 6, 2, 2, 3, 8, 6, 6, 7, 12 }; + static const uint8_t s_rndtblfield[16] = { 0, 0, 1, 2, 4, 4, 5, 6, 2, 2, 3, 8, 6, 6, 7, 12 }; int v_dist = fieldmv ? 1 : 4; // vertical offset for lower sub-blocks int v_edge_pos = s->v_edge_pos >> 1; int use_ic; @@ -707,23 +676,29 @@ void ff_vc1_mc_4mv_chroma4(VC1Context *v, int dir, int dir2, int avg) uvsrc_x = av_clip(uvsrc_x, -8, s->avctx->coded_width >> 1); uvsrc_y = av_clip(uvsrc_y, -8, s->avctx->coded_height >> 1); if (i < 2 ? dir : dir2) { - srcU = s->next_picture.f->data[1] + uvsrc_y * s->uvlinesize + uvsrc_x; - srcV = s->next_picture.f->data[2] + uvsrc_y * s->uvlinesize + uvsrc_x; + srcU = s->next_picture.f->data[1]; + srcV = s->next_picture.f->data[2]; lutuv = v->next_lutuv; use_ic = v->next_use_ic; } else { - srcU = s->last_picture.f->data[1] + uvsrc_y * s->uvlinesize + uvsrc_x; - srcV = s->last_picture.f->data[2] + uvsrc_y * s->uvlinesize + uvsrc_x; + srcU = s->last_picture.f->data[1]; + srcV = s->last_picture.f->data[2]; lutuv = v->last_lutuv; use_ic = v->last_use_ic; } + if (!srcU) + return; + srcU += uvsrc_y * s->uvlinesize + uvsrc_x; + srcV += uvsrc_y * s->uvlinesize + uvsrc_x; uvmx_field[i] = (uvmx_field[i] & 3) << 1; uvmy_field[i] = (uvmy_field[i] & 3) << 1; - if (fieldmv && !(uvsrc_y & 1)) - v_edge_pos--; - if (fieldmv && (uvsrc_y & 1) && uvsrc_y < 2) - uvsrc_y--; + if (fieldmv) { + if (!(uvsrc_y & 1)) + v_edge_pos = (s->v_edge_pos >> 1) - 1; + else + uvsrc_y -= (uvsrc_y < 2); + } if (use_ic || s->h_edge_pos < 10 || v_edge_pos < (5 << fieldmv) || (unsigned)uvsrc_x > (s->h_edge_pos >> 1) - 5 @@ -741,20 +716,10 @@ void ff_vc1_mc_4mv_chroma4(VC1Context *v, int dir, int dir2, int avg) /* if we deal with intensity compensation we need to scale source blocks */ if (use_ic) { - int i, j; - uint8_t *src, *src2; - - src = srcU; - src2 = srcV; - for (j = 0; j < 5; j++) { - int f = (uvsrc_y + (j << fieldmv)) & 1; - for (i = 0; i < 5; i++) { - src[i] = lutuv[f][src[i]]; - src2[i] = lutuv[f][src2[i]]; - } - src += s->uvlinesize << fieldmv; - src2 += s->uvlinesize << fieldmv; - } + vc1_lut_scale_chroma(srcU, srcV, + lutuv[(uvsrc_y + (0 << fieldmv)) & 1], + lutuv[(uvsrc_y + (1 << fieldmv)) & 1], + 5, s->uvlinesize << fieldmv); } } if (avg) { @@ -785,7 +750,6 @@ void ff_vc1_interp_mc(VC1Context *v) H264ChromaContext *h264chroma = &v->h264chroma; uint8_t *srcY, *srcU, *srcV; int dxy, mx, my, uvmx, uvmy, src_x, src_y, uvsrc_x, uvsrc_y; - int off, off_uv; int v_edge_pos = s->v_edge_pos >> v->field_mode; int use_ic = v->next_use_ic; @@ -844,93 +808,60 @@ void ff_vc1_interp_mc(VC1Context *v) if (v->rangeredfrm || s->h_edge_pos < 22 || v_edge_pos < 22 || use_ic || (unsigned)(src_x - 1) > s->h_edge_pos - (mx & 3) - 16 - 3 || (unsigned)(src_y - 1) > v_edge_pos - (my & 3) - 16 - 3) { - uint8_t *uvbuf = s->edge_emu_buffer + 19 * s->linesize; + uint8_t *ubuf = s->edge_emu_buffer + 19 * s->linesize; + uint8_t *vbuf = ubuf + 9 * s->uvlinesize; + const int k = 17 + s->mspel * 2; srcY -= s->mspel * (1 + s->linesize); s->vdsp.emulated_edge_mc(s->edge_emu_buffer, srcY, s->linesize, s->linesize, - 17 + s->mspel * 2, 17 + s->mspel * 2, + k, k, src_x - s->mspel, src_y - s->mspel, s->h_edge_pos, v_edge_pos); srcY = s->edge_emu_buffer; - s->vdsp.emulated_edge_mc(uvbuf, srcU, + s->vdsp.emulated_edge_mc(ubuf, srcU, s->uvlinesize, s->uvlinesize, 8 + 1, 8 + 1, - uvsrc_x, uvsrc_y, s->h_edge_pos >> 1, v_edge_pos >> 1); - s->vdsp.emulated_edge_mc(uvbuf + 16, srcV, + uvsrc_x, uvsrc_y, + s->h_edge_pos >> 1, v_edge_pos >> 1); + s->vdsp.emulated_edge_mc(vbuf, srcV, s->uvlinesize, s->uvlinesize, 8 + 1, 8 + 1, - uvsrc_x, uvsrc_y, s->h_edge_pos >> 1, v_edge_pos >> 1); - srcU = uvbuf; - srcV = uvbuf + 16; + uvsrc_x, uvsrc_y, + s->h_edge_pos >> 1, v_edge_pos >> 1); + srcU = ubuf; + srcV = vbuf; /* if we deal with range reduction we need to scale source blocks */ if (v->rangeredfrm) { - int i, j; - uint8_t *src, *src2; - - src = srcY; - for (j = 0; j < 17 + s->mspel * 2; j++) { - for (i = 0; i < 17 + s->mspel * 2; i++) - src[i] = ((src[i] - 128) >> 1) + 128; - src += s->linesize; - } - src = srcU; - src2 = srcV; - for (j = 0; j < 9; j++) { - for (i = 0; i < 9; i++) { - src[i] = ((src[i] - 128) >> 1) + 128; - src2[i] = ((src2[i] - 128) >> 1) + 128; - } - src += s->uvlinesize; - src2 += s->uvlinesize; - } + vc1_scale_luma(srcY, k, s->linesize); + vc1_scale_chroma(srcU, srcV, 9, s->uvlinesize); } if (use_ic) { uint8_t (*luty )[256] = v->next_luty; uint8_t (*lutuv)[256] = v->next_lutuv; - int i, j; - uint8_t *src, *src2; - - src = srcY; - for (j = 0; j < 17 + s->mspel * 2; j++) { - int f = v->field_mode ? v->ref_field_type[1] : ((j+src_y - s->mspel) & 1); - for (i = 0; i < 17 + s->mspel * 2; i++) - src[i] = luty[f][src[i]]; - src += s->linesize; - } - src = srcU; - src2 = srcV; - for (j = 0; j < 9; j++) { - int f = v->field_mode ? v->ref_field_type[1] : ((j+uvsrc_y) & 1); - for (i = 0; i < 9; i++) { - src[i] = lutuv[f][src[i]]; - src2[i] = lutuv[f][src2[i]]; - } - src += s->uvlinesize; - src2 += s->uvlinesize; - } + vc1_lut_scale_luma(srcY, + luty[v->field_mode ? v->ref_field_type[1] : ((0+src_y - s->mspel) & 1)], + luty[v->field_mode ? v->ref_field_type[1] : ((1+src_y - s->mspel) & 1)], + k, s->linesize); + vc1_lut_scale_chroma(srcU, srcV, + lutuv[v->field_mode ? v->ref_field_type[1] : ((0+uvsrc_y) & 1)], + lutuv[v->field_mode ? v->ref_field_type[1] : ((1+uvsrc_y) & 1)], + 9, s->uvlinesize); } srcY += s->mspel * (1 + s->linesize); } - off = 0; - off_uv = 0; - if (s->mspel) { dxy = ((my & 3) << 2) | (mx & 3); - v->vc1dsp.avg_vc1_mspel_pixels_tab[dxy](s->dest[0] + off , srcY , s->linesize, v->rnd); - v->vc1dsp.avg_vc1_mspel_pixels_tab[dxy](s->dest[0] + off + 8, srcY + 8, s->linesize, v->rnd); - srcY += s->linesize * 8; - v->vc1dsp.avg_vc1_mspel_pixels_tab[dxy](s->dest[0] + off + 8 * s->linesize , srcY , s->linesize, v->rnd); - v->vc1dsp.avg_vc1_mspel_pixels_tab[dxy](s->dest[0] + off + 8 * s->linesize + 8, srcY + 8, s->linesize, v->rnd); + v->vc1dsp.avg_vc1_mspel_pixels_tab[0][dxy](s->dest[0], srcY, s->linesize, v->rnd); } else { // hpel mc dxy = (my & 2) | ((mx & 2) >> 1); if (!v->rnd) - s->hdsp.avg_pixels_tab[0][dxy](s->dest[0] + off, srcY, s->linesize, 16); + s->hdsp.avg_pixels_tab[0][dxy](s->dest[0], srcY, s->linesize, 16); else - s->hdsp.avg_no_rnd_pixels_tab[dxy](s->dest[0] + off, srcY, s->linesize, 16); + s->hdsp.avg_no_rnd_pixels_tab[dxy](s->dest[0], srcY, s->linesize, 16); } if (s->flags & CODEC_FLAG_GRAY) return; @@ -938,10 +869,10 @@ void ff_vc1_interp_mc(VC1Context *v) uvmx = (uvmx & 3) << 1; uvmy = (uvmy & 3) << 1; if (!v->rnd) { - h264chroma->avg_h264_chroma_pixels_tab[0](s->dest[1] + off_uv, srcU, s->uvlinesize, 8, uvmx, uvmy); - h264chroma->avg_h264_chroma_pixels_tab[0](s->dest[2] + off_uv, srcV, s->uvlinesize, 8, uvmx, uvmy); + h264chroma->avg_h264_chroma_pixels_tab[0](s->dest[1], srcU, s->uvlinesize, 8, uvmx, uvmy); + h264chroma->avg_h264_chroma_pixels_tab[0](s->dest[2], srcV, s->uvlinesize, 8, uvmx, uvmy); } else { - v->vc1dsp.avg_no_rnd_vc1_chroma_pixels_tab[0](s->dest[1] + off_uv, srcU, s->uvlinesize, 8, uvmx, uvmy); - v->vc1dsp.avg_no_rnd_vc1_chroma_pixels_tab[0](s->dest[2] + off_uv, srcV, s->uvlinesize, 8, uvmx, uvmy); + v->vc1dsp.avg_no_rnd_vc1_chroma_pixels_tab[0](s->dest[1], srcU, s->uvlinesize, 8, uvmx, uvmy); + v->vc1dsp.avg_no_rnd_vc1_chroma_pixels_tab[0](s->dest[2], srcV, s->uvlinesize, 8, uvmx, uvmy); } } |