X-Git-Url: https://git.sesse.net/?a=blobdiff_plain;f=libavcodec%2Fvc1dec.c;h=f7f6a9f1216704af1a80931b2dc4396f842a3520;hb=0026e356d044e72b6e743b234708b8b8af457ac0;hp=7b0b8c92074b88459af1f1c9c2beab7cbffb82ab;hpb=36ef5369ee9b336febc2c270f8718cec4476cb85;p=ffmpeg diff --git a/libavcodec/vc1dec.c b/libavcodec/vc1dec.c index 7b0b8c92074..f7f6a9f1216 100644 --- a/libavcodec/vc1dec.c +++ b/libavcodec/vc1dec.c @@ -27,17 +27,20 @@ */ #include "internal.h" -#include "dsputil.h" #include "avcodec.h" +#include "error_resilience.h" +#include "mpeg_er.h" +#include "mpegutils.h" #include "mpegvideo.h" #include "h263.h" +#include "h264chroma.h" +#include "qpeldsp.h" #include "vc1.h" #include "vc1data.h" #include "vc1acdata.h" #include "msmpeg4data.h" #include "unary.h" #include "mathops.h" -#include "vdpau_internal.h" #undef NDEBUG #include @@ -72,6 +75,16 @@ enum Imode { }; /** @} */ //imode defines +static void init_block_index(VC1Context *v) +{ + MpegEncContext *s = &v->s; + ff_init_block_index(s); + if (v->field_mode && !(v->second_field ^ v->tff)) { + s->dest[0] += s->current_picture_ptr->f->linesize[0]; + s->dest[1] += s->current_picture_ptr->f->linesize[1]; + s->dest[2] += s->current_picture_ptr->f->linesize[2]; + } +} /** @} */ //Bitplane group @@ -79,7 +92,7 @@ static void vc1_put_signed_blocks_clamped(VC1Context *v) { MpegEncContext *s = &v->s; int topleft_mb_pos, top_mb_pos; - int stride_y, fieldtx; + int stride_y, fieldtx = 0; int v_dist; /* The put pixels loop is always one MB row behind the decoding loop, @@ -92,51 +105,53 @@ static void vc1_put_signed_blocks_clamped(VC1Context *v) if (!s->first_slice_line) { if (s->mb_x) { topleft_mb_pos = (s->mb_y - 1) * s->mb_stride + s->mb_x - 1; - fieldtx = v->fieldtx_plane[topleft_mb_pos]; + if (v->fcm == ILACE_FRAME) + fieldtx = v->fieldtx_plane[topleft_mb_pos]; stride_y = s->linesize << fieldtx; v_dist = (16 - fieldtx) >> (fieldtx == 0); - s->dsp.put_signed_pixels_clamped(v->block[v->topleft_blk_idx][0], - s->dest[0] - 16 * s->linesize - 16, - stride_y); - s->dsp.put_signed_pixels_clamped(v->block[v->topleft_blk_idx][1], - s->dest[0] - 16 * s->linesize - 8, - stride_y); - s->dsp.put_signed_pixels_clamped(v->block[v->topleft_blk_idx][2], - s->dest[0] - v_dist * s->linesize - 16, - stride_y); - s->dsp.put_signed_pixels_clamped(v->block[v->topleft_blk_idx][3], - s->dest[0] - v_dist * s->linesize - 8, - stride_y); - s->dsp.put_signed_pixels_clamped(v->block[v->topleft_blk_idx][4], - s->dest[1] - 8 * s->uvlinesize - 8, - s->uvlinesize); - s->dsp.put_signed_pixels_clamped(v->block[v->topleft_blk_idx][5], - s->dest[2] - 8 * s->uvlinesize - 8, - s->uvlinesize); + s->idsp.put_signed_pixels_clamped(v->block[v->topleft_blk_idx][0], + s->dest[0] - 16 * s->linesize - 16, + stride_y); + s->idsp.put_signed_pixels_clamped(v->block[v->topleft_blk_idx][1], + s->dest[0] - 16 * s->linesize - 8, + stride_y); + s->idsp.put_signed_pixels_clamped(v->block[v->topleft_blk_idx][2], + s->dest[0] - v_dist * s->linesize - 16, + stride_y); + s->idsp.put_signed_pixels_clamped(v->block[v->topleft_blk_idx][3], + s->dest[0] - v_dist * s->linesize - 8, + stride_y); + s->idsp.put_signed_pixels_clamped(v->block[v->topleft_blk_idx][4], + s->dest[1] - 8 * s->uvlinesize - 8, + s->uvlinesize); + s->idsp.put_signed_pixels_clamped(v->block[v->topleft_blk_idx][5], + s->dest[2] - 8 * s->uvlinesize - 8, + s->uvlinesize); } if (s->mb_x == s->mb_width - 1) { top_mb_pos = (s->mb_y - 1) * s->mb_stride + s->mb_x; - fieldtx = v->fieldtx_plane[top_mb_pos]; + if (v->fcm == ILACE_FRAME) + fieldtx = v->fieldtx_plane[top_mb_pos]; stride_y = s->linesize << fieldtx; v_dist = fieldtx ? 15 : 8; - s->dsp.put_signed_pixels_clamped(v->block[v->top_blk_idx][0], - s->dest[0] - 16 * s->linesize, - stride_y); - s->dsp.put_signed_pixels_clamped(v->block[v->top_blk_idx][1], - s->dest[0] - 16 * s->linesize + 8, - stride_y); - s->dsp.put_signed_pixels_clamped(v->block[v->top_blk_idx][2], - s->dest[0] - v_dist * s->linesize, - stride_y); - s->dsp.put_signed_pixels_clamped(v->block[v->top_blk_idx][3], - s->dest[0] - v_dist * s->linesize + 8, - stride_y); - s->dsp.put_signed_pixels_clamped(v->block[v->top_blk_idx][4], - s->dest[1] - 8 * s->uvlinesize, - s->uvlinesize); - s->dsp.put_signed_pixels_clamped(v->block[v->top_blk_idx][5], - s->dest[2] - 8 * s->uvlinesize, - s->uvlinesize); + s->idsp.put_signed_pixels_clamped(v->block[v->top_blk_idx][0], + s->dest[0] - 16 * s->linesize, + stride_y); + s->idsp.put_signed_pixels_clamped(v->block[v->top_blk_idx][1], + s->dest[0] - 16 * s->linesize + 8, + stride_y); + s->idsp.put_signed_pixels_clamped(v->block[v->top_blk_idx][2], + s->dest[0] - v_dist * s->linesize, + stride_y); + s->idsp.put_signed_pixels_clamped(v->block[v->top_blk_idx][3], + s->dest[0] - v_dist * s->linesize + 8, + stride_y); + s->idsp.put_signed_pixels_clamped(v->block[v->top_blk_idx][4], + s->dest[1] - 8 * s->uvlinesize, + s->uvlinesize); + s->idsp.put_signed_pixels_clamped(v->block[v->top_blk_idx][5], + s->dest[2] - 8 * s->uvlinesize, + s->uvlinesize); } } @@ -330,15 +345,17 @@ static void vc1_smooth_overlap_filter_iblk(VC1Context *v) static void vc1_mc_1mv(VC1Context *v, int dir) { MpegEncContext *s = &v->s; - DSPContext *dsp = &v->s.dsp; + 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 i; + uint8_t (*luty)[256], (*lutuv)[256]; + int use_ic; if ((!v->field_mode || (v->ref_field_type[dir] == 1 && v->cur_field_type == 1)) && - !v->s.last_picture.f.data[0]) + !v->s.last_picture.f->data[0]) return; mx = s->mv[dir][0][0]; @@ -346,8 +363,10 @@ static void vc1_mc_1mv(VC1Context *v, int dir) // store motion vectors for further use in B frames if (s->pict_type == AV_PICTURE_TYPE_P) { - s->current_picture.f.motion_val[1][s->block_index[0] + v->blocks_off][0] = mx; - s->current_picture.f.motion_val[1][s->block_index[0] + v->blocks_off][1] = my; + for (i = 0; i < 4; i++) { + s->current_picture.motion_val[1][s->block_index[i] + v->blocks_off][0] = mx; + s->current_picture.motion_val[1][s->block_index[i] + v->blocks_off][1] = my; + } } uvmx = (mx + ((mx & 3) == 3)) >> 1; @@ -366,32 +385,34 @@ static void vc1_mc_1mv(VC1Context *v, int dir) uvmx = uvmx + ((uvmx < 0) ? (uvmx & 1) : -(uvmx & 1)); uvmy = uvmy + ((uvmy < 0) ? (uvmy & 1) : -(uvmy & 1)); } - if (v->field_mode) { // interlaced field picture - if (!dir) { - if ((v->cur_field_type != v->ref_field_type[dir]) && v->cur_field_type) { - srcY = s->current_picture.f.data[0]; - srcU = s->current_picture.f.data[1]; - srcV = s->current_picture.f.data[2]; - } else { - srcY = s->last_picture.f.data[0]; - srcU = s->last_picture.f.data[1]; - srcV = s->last_picture.f.data[2]; - } + if (!dir) { + if (v->field_mode && (v->cur_field_type != v->ref_field_type[dir]) && v->second_field) { + srcY = s->current_picture.f->data[0]; + srcU = s->current_picture.f->data[1]; + srcV = s->current_picture.f->data[2]; + luty = v->curr_luty; + lutuv = v->curr_lutuv; + use_ic = v->curr_use_ic; } else { - srcY = s->next_picture.f.data[0]; - srcU = s->next_picture.f.data[1]; - srcV = s->next_picture.f.data[2]; + srcY = s->last_picture.f->data[0]; + srcU = s->last_picture.f->data[1]; + srcV = s->last_picture.f->data[2]; + luty = v->last_luty; + lutuv = v->last_lutuv; + use_ic = v->last_use_ic; } } else { - if (!dir) { - srcY = s->last_picture.f.data[0]; - srcU = s->last_picture.f.data[1]; - srcV = s->last_picture.f.data[2]; - } else { - srcY = s->next_picture.f.data[0]; - srcU = s->next_picture.f.data[1]; - srcV = s->next_picture.f.data[2]; - } + srcY = s->next_picture.f->data[0]; + srcU = s->next_picture.f->data[1]; + srcV = s->next_picture.f->data[2]; + luty = v->next_luty; + lutuv = v->next_lutuv; + use_ic = v->next_use_ic; + } + + if (!srcY || !srcU) { + av_log(v->s.avctx, AV_LOG_ERROR, "Referenced frame missing.\n"); + return; } src_x = s->mb_x * 16 + (mx >> 2); @@ -416,9 +437,9 @@ static void vc1_mc_1mv(VC1Context *v, int dir) srcV += uvsrc_y * s->uvlinesize + uvsrc_x; if (v->field_mode && v->ref_field_type[dir]) { - srcY += s->current_picture_ptr->f.linesize[0]; - srcU += s->current_picture_ptr->f.linesize[1]; - srcV += s->current_picture_ptr->f.linesize[2]; + srcY += s->current_picture_ptr->f->linesize[0]; + srcU += s->current_picture_ptr->f->linesize[1]; + srcV += s->current_picture_ptr->f->linesize[2]; } /* for grayscale we should not try to read from unknown area */ @@ -427,22 +448,27 @@ static void vc1_mc_1mv(VC1Context *v, int dir) srcV = s->edge_emu_buffer + 18 * s->linesize; } - if (v->rangeredfrm || (v->mv_mode == MV_PMODE_INTENSITY_COMP) + if (v->rangeredfrm || use_ic || 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 - s->mspel) > v_edge_pos - (my&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; srcY -= s->mspel * (1 + s->linesize); - s->dsp.emulated_edge_mc(s->edge_emu_buffer, srcY, s->linesize, - 17 + s->mspel * 2, 17 + s->mspel * 2, - src_x - s->mspel, src_y - s->mspel, - s->h_edge_pos, v_edge_pos); + s->vdsp.emulated_edge_mc(s->edge_emu_buffer, srcY, + s->linesize, s->linesize, + 17 + s->mspel * 2, 17 + s->mspel * 2, + src_x - s->mspel, src_y - s->mspel, + s->h_edge_pos, v_edge_pos); srcY = s->edge_emu_buffer; - s->dsp.emulated_edge_mc(uvbuf , srcU, s->uvlinesize, 8 + 1, 8 + 1, - uvsrc_x, uvsrc_y, s->h_edge_pos >> 1, v_edge_pos >> 1); - s->dsp.emulated_edge_mc(uvbuf + 16, srcV, 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, 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, + 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; /* if we deal with range reduction we need to scale source blocks */ @@ -468,22 +494,24 @@ static void vc1_mc_1mv(VC1Context *v, int dir) } } /* if we deal with intensity compensation we need to scale source blocks */ - if (v->mv_mode == MV_PMODE_INTENSITY_COMP) { + 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] = v->luty[src[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] = v->lutuv[src[i]]; - src2[i] = v->lutuv[src2[i]]; + src[i] = lutuv[f][src[i]]; + src2[i] = lutuv[f][src2[i]]; } src += s->uvlinesize; src2 += s->uvlinesize; @@ -492,26 +520,19 @@ static void vc1_mc_1mv(VC1Context *v, int dir) srcY += s->mspel * (1 + s->linesize); } - if (v->field_mode && v->cur_field_type) { - off = s->current_picture_ptr->f.linesize[0]; - off_uv = s->current_picture_ptr->f.linesize[1]; - } else { - off = 0; - off_uv = 0; - } if (s->mspel) { dxy = ((my & 3) << 2) | (mx & 3); - v->vc1dsp.put_vc1_mspel_pixels_tab[dxy](s->dest[0] + off , srcY , s->linesize, v->rnd); - v->vc1dsp.put_vc1_mspel_pixels_tab[dxy](s->dest[0] + off + 8, srcY + 8, s->linesize, v->rnd); + 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] + off + 8 * s->linesize , srcY , s->linesize, v->rnd); - v->vc1dsp.put_vc1_mspel_pixels_tab[dxy](s->dest[0] + off + 8 * s->linesize + 8, srcY + 8, s->linesize, v->rnd); + 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); } else { // hpel mc - always used for luma dxy = (my & 2) | ((mx & 2) >> 1); if (!v->rnd) - dsp->put_pixels_tab[0][dxy](s->dest[0] + off, srcY, s->linesize, 16); + s->hdsp.put_pixels_tab[0][dxy](s->dest[0], srcY, s->linesize, 16); else - dsp->put_no_rnd_pixels_tab[0][dxy](s->dest[0] + off, srcY, s->linesize, 16); + s->hdsp.put_no_rnd_pixels_tab[0][dxy](s->dest[0], srcY, s->linesize, 16); } if (s->flags & CODEC_FLAG_GRAY) return; @@ -519,11 +540,11 @@ static void vc1_mc_1mv(VC1Context *v, int dir) uvmx = (uvmx & 3) << 1; uvmy = (uvmy & 3) << 1; if (!v->rnd) { - dsp->put_h264_chroma_pixels_tab[0](s->dest[1] + off_uv, srcU, s->uvlinesize, 8, uvmx, uvmy); - dsp->put_h264_chroma_pixels_tab[0](s->dest[2] + off_uv, srcV, s->uvlinesize, 8, uvmx, uvmy); + h264chroma->put_h264_chroma_pixels_tab[0](s->dest[1], srcU, s->uvlinesize, 8, uvmx, uvmy); + h264chroma->put_h264_chroma_pixels_tab[0](s->dest[2], srcV, s->uvlinesize, 8, uvmx, uvmy); } else { - v->vc1dsp.put_no_rnd_vc1_chroma_pixels_tab[0](s->dest[1] + off_uv, srcU, s->uvlinesize, 8, uvmx, uvmy); - v->vc1dsp.put_no_rnd_vc1_chroma_pixels_tab[0](s->dest[2] + off_uv, srcV, s->uvlinesize, 8, uvmx, uvmy); + v->vc1dsp.put_no_rnd_vc1_chroma_pixels_tab[0](s->dest[1], srcU, s->uvlinesize, 8, uvmx, uvmy); + v->vc1dsp.put_no_rnd_vc1_chroma_pixels_tab[0](s->dest[2], srcV, s->uvlinesize, 8, uvmx, uvmy); } } @@ -540,34 +561,45 @@ static inline int median4(int a, int b, int c, int d) /** Do motion compensation for 4-MV macroblock - luminance block */ -static void vc1_mc_4mv_luma(VC1Context *v, int n, int dir) +static void vc1_mc_4mv_luma(VC1Context *v, int n, int dir, int avg) { MpegEncContext *s = &v->s; - DSPContext *dsp = &v->s.dsp; uint8_t *srcY; int dxy, mx, my, src_x, src_y; int off; int fieldmv = (v->fcm == ILACE_FRAME) ? v->blk_mv_type[s->block_index[n]] : 0; int v_edge_pos = s->v_edge_pos >> v->field_mode; + uint8_t (*luty)[256]; + int use_ic; if ((!v->field_mode || (v->ref_field_type[dir] == 1 && v->cur_field_type == 1)) && - !v->s.last_picture.f.data[0]) + !v->s.last_picture.f->data[0]) return; mx = s->mv[dir][n][0]; my = s->mv[dir][n][1]; if (!dir) { - if (v->field_mode) { - if ((v->cur_field_type != v->ref_field_type[dir]) && v->cur_field_type) - srcY = s->current_picture.f.data[0]; - else - srcY = s->last_picture.f.data[0]; - } else - srcY = s->last_picture.f.data[0]; - } else - srcY = s->next_picture.f.data[0]; + 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; + } else { + srcY = s->last_picture.f->data[0]; + luty = v->last_luty; + use_ic = v->last_use_ic; + } + } else { + srcY = s->next_picture.f->data[0]; + luty = v->next_luty; + use_ic = v->next_use_ic; + } + + if (!srcY) { + av_log(v->s.avctx, AV_LOG_ERROR, "Referenced frame missing.\n"); + return; + } if (v->field_mode) { if (v->cur_field_type != v->ref_field_type[dir]) @@ -602,8 +634,8 @@ static void vc1_mc_4mv_luma(VC1Context *v, int n, int dir) ty = (chosen_mv[f][0][1] + chosen_mv[f][1][1]) / 2; break; } - s->current_picture.f.motion_val[1][s->block_index[0] + v->blocks_off][0] = tx; - s->current_picture.f.motion_val[1][s->block_index[0] + v->blocks_off][1] = ty; + 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; for (k = 0; k < 4; k++) v->mv_f[1][s->block_index[k] + v->blocks_off] = f; } @@ -612,6 +644,10 @@ static void vc1_mc_4mv_luma(VC1Context *v, int n, int dir) int qx, qy; int width = s->avctx->coded_width; int height = s->avctx->coded_height >> 1; + if (s->pict_type == AV_PICTURE_TYPE_P) { + s->current_picture.motion_val[1][s->block_index[n] + v->blocks_off][0] = mx; + s->current_picture.motion_val[1][s->block_index[n] + v->blocks_off][1] = my; + } qx = (s->mb_x * 16) + (mx >> 2); qy = (s->mb_y * 8) + (my >> 3); @@ -629,8 +665,6 @@ static void vc1_mc_4mv_luma(VC1Context *v, int n, int dir) off = ((n > 1) ? s->linesize : 0) + (n & 1) * 8; else off = s->linesize * 4 * (n & 2) + (n & 1) * 8; - if (v->field_mode && v->cur_field_type) - off += s->current_picture_ptr->f.linesize[0]; src_x = s->mb_x * 16 + (n & 1) * 8 + (mx >> 2); if (!fieldmv) @@ -655,22 +689,23 @@ static void vc1_mc_4mv_luma(VC1Context *v, int n, int dir) srcY += src_y * s->linesize + src_x; if (v->field_mode && v->ref_field_type[dir]) - srcY += s->current_picture_ptr->f.linesize[0]; + 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 (v->rangeredfrm || (v->mv_mode == MV_PMODE_INTENSITY_COMP) + 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)) { srcY -= s->mspel * (1 + (s->linesize << fieldmv)); /* check emulate edge stride and offset */ - s->dsp.emulated_edge_mc(s->edge_emu_buffer, srcY, s->linesize, - 9 + s->mspel * 2, (9 + s->mspel * 2) << fieldmv, - src_x - s->mspel, src_y - (s->mspel << fieldmv), - s->h_edge_pos, v_edge_pos); + s->vdsp.emulated_edge_mc(s->edge_emu_buffer, srcY, + s->linesize, s->linesize, + 9 + s->mspel * 2, (9 + s->mspel * 2) << 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) { @@ -685,14 +720,15 @@ static void vc1_mc_4mv_luma(VC1Context *v, int n, int dir) } } /* if we deal with intensity compensation we need to scale source blocks */ - if (v->mv_mode == MV_PMODE_INTENSITY_COMP) { + 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<mspel << fieldmv)) & 1); for (i = 0; i < 9 + s->mspel * 2; i++) - src[i] = v->luty[src[i]]; + src[i] = luty[f][src[i]]; src += s->linesize << fieldmv; } } @@ -701,13 +737,16 @@ static void vc1_mc_4mv_luma(VC1Context *v, int n, int dir) if (s->mspel) { dxy = ((my & 3) << 2) | (mx & 3); - v->vc1dsp.put_vc1_mspel_pixels_tab[dxy](s->dest[0] + off, srcY, s->linesize << fieldmv, v->rnd); + if (avg) + v->vc1dsp.avg_vc1_mspel_pixels_tab[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); } else { // hpel mc - always used for luma dxy = (my & 2) | ((mx & 2) >> 1); if (!v->rnd) - dsp->put_pixels_tab[1][dxy](s->dest[0] + off, srcY, s->linesize, 8); + s->hdsp.put_pixels_tab[1][dxy](s->dest[0] + off, srcY, s->linesize, 8); else - dsp->put_no_rnd_pixels_tab[1][dxy](s->dest[0] + off, srcY, s->linesize, 8); + s->hdsp.put_no_rnd_pixels_tab[1][dxy](s->dest[0] + off, srcY, s->linesize, 8); } } @@ -769,16 +808,18 @@ static av_always_inline int get_chroma_mv(int *mvx, int *mvy, int *a, int flag, static void vc1_mc_4mv_chroma(VC1Context *v, int dir) { MpegEncContext *s = &v->s; - DSPContext *dsp = &v->s.dsp; + 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, off = 0; + int chroma_ref_type = v->cur_field_type; int v_edge_pos = s->v_edge_pos >> v->field_mode; + uint8_t (*lutuv)[256]; + int use_ic; - if (!v->field_mode && !v->s.last_picture.f.data[0]) + if (!v->field_mode && !v->s.last_picture.f->data[0]) return; if (s->flags & CODEC_FLAG_GRAY) return; @@ -794,9 +835,10 @@ static void vc1_mc_4mv_chroma(VC1Context *v, int dir) /* 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 (!valid_count) { - s->current_picture.f.motion_val[1][s->block_index[0] + v->blocks_off][0] = 0; - s->current_picture.f.motion_val[1][s->block_index[0] + v->blocks_off][1] = 0; + 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 } @@ -808,10 +850,10 @@ static void vc1_mc_4mv_chroma(VC1Context *v, int dir) if (dominant) chroma_ref_type = !v->cur_field_type; } - if (v->field_mode && chroma_ref_type == 1 && v->cur_field_type == 1 && !v->s.last_picture.f.data[0]) + if (v->field_mode && chroma_ref_type == 1 && v->cur_field_type == 1 && !v->s.last_picture.f->data[0]) return; - s->current_picture.f.motion_val[1][s->block_index[0] + v->blocks_off][0] = tx; - s->current_picture.f.motion_val[1][s->block_index[0] + v->blocks_off][1] = ty; + 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; uvmx = (tx + ((tx & 3) == 3)) >> 1; uvmy = (ty + ((ty & 3) == 3)) >> 1; @@ -838,41 +880,51 @@ static void vc1_mc_4mv_chroma(VC1Context *v, int dir) } if (!dir) { - if (v->field_mode) { - if ((v->cur_field_type != chroma_ref_type) && v->cur_field_type) { - srcU = s->current_picture.f.data[1] + uvsrc_y * s->uvlinesize + uvsrc_x; - srcV = s->current_picture.f.data[2] + uvsrc_y * s->uvlinesize + uvsrc_x; - } 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; - } + if (v->field_mode && (v->cur_field_type != chroma_ref_type) && v->second_field) { + srcU = s->current_picture.f->data[1]; + srcV = s->current_picture.f->data[2]; + lutuv = v->curr_lutuv; + use_ic = v->curr_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; } } else { - 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; } + if (!srcU) { + av_log(v->s.avctx, AV_LOG_ERROR, "Referenced frame missing.\n"); + return; + } + + srcU += uvsrc_y * s->uvlinesize + uvsrc_x; + srcV += uvsrc_y * s->uvlinesize + uvsrc_x; + if (v->field_mode) { if (chroma_ref_type) { - srcU += s->current_picture_ptr->f.linesize[1]; - srcV += s->current_picture_ptr->f.linesize[2]; + srcU += s->current_picture_ptr->f->linesize[1]; + srcV += s->current_picture_ptr->f->linesize[2]; } - off = v->cur_field_type ? s->current_picture_ptr->f.linesize[1] : 0; } - if (v->rangeredfrm || (v->mv_mode == MV_PMODE_INTENSITY_COMP) + if (v->rangeredfrm || use_ic || s->h_edge_pos < 18 || v_edge_pos < 18 || (unsigned)uvsrc_x > (s->h_edge_pos >> 1) - 9 || (unsigned)uvsrc_y > (v_edge_pos >> 1) - 9) { - s->dsp.emulated_edge_mc(s->edge_emu_buffer , srcU, s->uvlinesize, - 8 + 1, 8 + 1, uvsrc_x, uvsrc_y, - s->h_edge_pos >> 1, v_edge_pos >> 1); - s->dsp.emulated_edge_mc(s->edge_emu_buffer + 16, srcV, s->uvlinesize, - 8 + 1, 8 + 1, uvsrc_x, uvsrc_y, - s->h_edge_pos >> 1, v_edge_pos >> 1); + s->vdsp.emulated_edge_mc(s->edge_emu_buffer, 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(s->edge_emu_buffer + 16, srcV, + s->uvlinesize, s->uvlinesize, + 8 + 1, 8 + 1, uvsrc_x, uvsrc_y, + s->h_edge_pos >> 1, v_edge_pos >> 1); srcU = s->edge_emu_buffer; srcV = s->edge_emu_buffer + 16; @@ -893,16 +945,17 @@ static void vc1_mc_4mv_chroma(VC1Context *v, int dir) } } /* if we deal with intensity compensation we need to scale source blocks */ - if (v->mv_mode == MV_PMODE_INTENSITY_COMP) { + 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] = v->lutuv[src[i]]; - src2[i] = v->lutuv[src2[i]]; + src[i] = lutuv[f][src[i]]; + src2[i] = lutuv[f][src2[i]]; } src += s->uvlinesize; src2 += s->uvlinesize; @@ -914,20 +967,20 @@ static void vc1_mc_4mv_chroma(VC1Context *v, int dir) uvmx = (uvmx & 3) << 1; uvmy = (uvmy & 3) << 1; if (!v->rnd) { - dsp->put_h264_chroma_pixels_tab[0](s->dest[1] + off, srcU, s->uvlinesize, 8, uvmx, uvmy); - dsp->put_h264_chroma_pixels_tab[0](s->dest[2] + off, srcV, s->uvlinesize, 8, uvmx, uvmy); + h264chroma->put_h264_chroma_pixels_tab[0](s->dest[1], srcU, s->uvlinesize, 8, uvmx, uvmy); + h264chroma->put_h264_chroma_pixels_tab[0](s->dest[2], srcV, s->uvlinesize, 8, uvmx, uvmy); } else { - v->vc1dsp.put_no_rnd_vc1_chroma_pixels_tab[0](s->dest[1] + off, srcU, s->uvlinesize, 8, uvmx, uvmy); - v->vc1dsp.put_no_rnd_vc1_chroma_pixels_tab[0](s->dest[2] + off, srcV, s->uvlinesize, 8, uvmx, uvmy); + v->vc1dsp.put_no_rnd_vc1_chroma_pixels_tab[0](s->dest[1], srcU, s->uvlinesize, 8, uvmx, uvmy); + v->vc1dsp.put_no_rnd_vc1_chroma_pixels_tab[0](s->dest[2], srcV, s->uvlinesize, 8, uvmx, uvmy); } } -/** Do motion compensation for 4-MV field chroma macroblock (both U and V) +/** Do motion compensation for 4-MV interlaced frame chroma macroblock (both U and V) */ -static void vc1_mc_4mv_chroma4(VC1Context *v) +static void vc1_mc_4mv_chroma4(VC1Context *v, int dir, int dir2, int avg) { MpegEncContext *s = &v->s; - DSPContext *dsp = &v->s.dsp; + H264ChromaContext *h264chroma = &v->h264chroma; uint8_t *srcU, *srcV; int uvsrc_x, uvsrc_y; int uvmx_field[4], uvmy_field[4]; @@ -936,16 +989,17 @@ static void vc1_mc_4mv_chroma4(VC1Context *v) static const int 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; + uint8_t (*lutuv)[256]; - if (!v->s.last_picture.f.data[0]) - return; if (s->flags & CODEC_FLAG_GRAY) return; for (i = 0; i < 4; i++) { - tx = s->mv[0][i][0]; + int d = i < 2 ? dir: dir2; + tx = s->mv[d][i][0]; uvmx_field[i] = (tx + ((tx & 3) == 3)) >> 1; - ty = s->mv[0][i][1]; + ty = s->mv[d][i][1]; if (fieldmv) uvmy_field[i] = (ty >> 4) * 8 + s_rndtblfield[ty & 0xF]; else @@ -959,8 +1013,17 @@ static void vc1_mc_4mv_chroma4(VC1Context *v) // FIXME: implement proper pull-back (see vc1cropmv.c, vc1CROPMV_ChromaPullBack()) uvsrc_x = av_clip(uvsrc_x, -8, s->avctx->coded_width >> 1); uvsrc_y = av_clip(uvsrc_y, -8, s->avctx->coded_height >> 1); - 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; + 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; + 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; + lutuv = v->last_lutuv; + use_ic = v->last_use_ic; + } uvmx_field[i] = (uvmx_field[i] & 3) << 1; uvmy_field[i] = (uvmy_field[i] & 3) << 1; @@ -968,42 +1031,55 @@ static void vc1_mc_4mv_chroma4(VC1Context *v) v_edge_pos--; if (fieldmv && (uvsrc_y & 1) && uvsrc_y < 2) uvsrc_y--; - if ((v->mv_mode == MV_PMODE_INTENSITY_COMP) + if (use_ic || s->h_edge_pos < 10 || v_edge_pos < (5 << fieldmv) || (unsigned)uvsrc_x > (s->h_edge_pos >> 1) - 5 || (unsigned)uvsrc_y > v_edge_pos - (5 << fieldmv)) { - s->dsp.emulated_edge_mc(s->edge_emu_buffer, srcU, s->uvlinesize, - 5, (5 << fieldmv), uvsrc_x, uvsrc_y, - s->h_edge_pos >> 1, v_edge_pos); - s->dsp.emulated_edge_mc(s->edge_emu_buffer + 16, srcV, s->uvlinesize, - 5, (5 << fieldmv), uvsrc_x, uvsrc_y, - s->h_edge_pos >> 1, v_edge_pos); + s->vdsp.emulated_edge_mc(s->edge_emu_buffer, srcU, + s->uvlinesize, s->uvlinesize, + 5, (5 << fieldmv), uvsrc_x, uvsrc_y, + s->h_edge_pos >> 1, v_edge_pos); + s->vdsp.emulated_edge_mc(s->edge_emu_buffer + 16, srcV, + s->uvlinesize, s->uvlinesize, + 5, (5 << fieldmv), uvsrc_x, uvsrc_y, + s->h_edge_pos >> 1, v_edge_pos); srcU = s->edge_emu_buffer; srcV = s->edge_emu_buffer + 16; /* if we deal with intensity compensation we need to scale source blocks */ - if (v->mv_mode == MV_PMODE_INTENSITY_COMP) { + 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] = v->lutuv[src[i]]; - src2[i] = v->lutuv[src2[i]]; + src[i] = lutuv[f][src[i]]; + src2[i] = lutuv[f][src2[i]]; } - src += s->uvlinesize << 1; - src2 += s->uvlinesize << 1; + src += s->uvlinesize << fieldmv; + src2 += s->uvlinesize << fieldmv; } } } - if (!v->rnd) { - dsp->put_h264_chroma_pixels_tab[1](s->dest[1] + off, srcU, s->uvlinesize << fieldmv, 4, uvmx_field[i], uvmy_field[i]); - dsp->put_h264_chroma_pixels_tab[1](s->dest[2] + off, srcV, s->uvlinesize << fieldmv, 4, uvmx_field[i], uvmy_field[i]); + if (avg) { + if (!v->rnd) { + h264chroma->avg_h264_chroma_pixels_tab[1](s->dest[1] + off, srcU, s->uvlinesize << fieldmv, 4, uvmx_field[i], uvmy_field[i]); + h264chroma->avg_h264_chroma_pixels_tab[1](s->dest[2] + off, srcV, s->uvlinesize << fieldmv, 4, uvmx_field[i], uvmy_field[i]); + } else { + v->vc1dsp.avg_no_rnd_vc1_chroma_pixels_tab[1](s->dest[1] + off, srcU, s->uvlinesize << fieldmv, 4, uvmx_field[i], uvmy_field[i]); + v->vc1dsp.avg_no_rnd_vc1_chroma_pixels_tab[1](s->dest[2] + off, srcV, s->uvlinesize << fieldmv, 4, uvmx_field[i], uvmy_field[i]); + } } else { - v->vc1dsp.put_no_rnd_vc1_chroma_pixels_tab[1](s->dest[1] + off, srcU, s->uvlinesize << fieldmv, 4, uvmx_field[i], uvmy_field[i]); - v->vc1dsp.put_no_rnd_vc1_chroma_pixels_tab[1](s->dest[2] + off, srcV, s->uvlinesize << fieldmv, 4, uvmx_field[i], uvmy_field[i]); + if (!v->rnd) { + h264chroma->put_h264_chroma_pixels_tab[1](s->dest[1] + off, srcU, s->uvlinesize << fieldmv, 4, uvmx_field[i], uvmy_field[i]); + h264chroma->put_h264_chroma_pixels_tab[1](s->dest[2] + off, srcV, s->uvlinesize << fieldmv, 4, uvmx_field[i], uvmy_field[i]); + } else { + v->vc1dsp.put_no_rnd_vc1_chroma_pixels_tab[1](s->dest[1] + off, srcU, s->uvlinesize << fieldmv, 4, uvmx_field[i], uvmy_field[i]); + v->vc1dsp.put_no_rnd_vc1_chroma_pixels_tab[1](s->dest[2] + off, srcV, s->uvlinesize << fieldmv, 4, uvmx_field[i], uvmy_field[i]); + } } } } @@ -1132,8 +1208,12 @@ static av_always_inline void get_mvdata_interlaced(VC1Context *v, int *dmv_x, *dmv_x = get_bits(gb, v->k_x); *dmv_y = get_bits(gb, v->k_y); if (v->numref) { - *pred_flag = *dmv_y & 1; - *dmv_y = (*dmv_y + *pred_flag) >> 1; + if (pred_flag) { + *pred_flag = *dmv_y & 1; + *dmv_y = (*dmv_y + *pred_flag) >> 1; + } else { + *dmv_y = (*dmv_y + (*dmv_y & 1)) >> 1; + } } } else { @@ -1159,7 +1239,7 @@ static av_always_inline void get_mvdata_interlaced(VC1Context *v, int *dmv_x, *dmv_y = (sign ^ ((val >> 1) + offs_tab[index1 >> v->numref])) - sign; } else *dmv_y = 0; - if (v->numref) + if (v->numref && pred_flag) *pred_flag = index1 & 1; } } @@ -1349,7 +1429,7 @@ static inline void vc1_pred_mv(VC1Context *v, int n, int dmv_x, int dmv_y, int px, py; int sum; int mixedmv_pic, num_samefield = 0, num_oppfield = 0; - int opposit, a_f, b_f, c_f; + int opposite, a_f, b_f, c_f; int16_t field_predA[2]; int16_t field_predB[2]; int16_t field_predC[2]; @@ -1369,30 +1449,30 @@ static inline void vc1_pred_mv(VC1Context *v, int n, int dmv_x, int dmv_y, xy = s->block_index[n]; if (s->mb_intra) { - s->mv[0][n][0] = s->current_picture.f.motion_val[0][xy + v->blocks_off][0] = 0; - s->mv[0][n][1] = s->current_picture.f.motion_val[0][xy + v->blocks_off][1] = 0; - s->current_picture.f.motion_val[1][xy + v->blocks_off][0] = 0; - s->current_picture.f.motion_val[1][xy + v->blocks_off][1] = 0; + s->mv[0][n][0] = s->current_picture.motion_val[0][xy + v->blocks_off][0] = 0; + s->mv[0][n][1] = s->current_picture.motion_val[0][xy + v->blocks_off][1] = 0; + s->current_picture.motion_val[1][xy + v->blocks_off][0] = 0; + s->current_picture.motion_val[1][xy + v->blocks_off][1] = 0; if (mv1) { /* duplicate motion data for 1-MV block */ - s->current_picture.f.motion_val[0][xy + 1 + v->blocks_off][0] = 0; - s->current_picture.f.motion_val[0][xy + 1 + v->blocks_off][1] = 0; - s->current_picture.f.motion_val[0][xy + wrap + v->blocks_off][0] = 0; - s->current_picture.f.motion_val[0][xy + wrap + v->blocks_off][1] = 0; - s->current_picture.f.motion_val[0][xy + wrap + 1 + v->blocks_off][0] = 0; - s->current_picture.f.motion_val[0][xy + wrap + 1 + v->blocks_off][1] = 0; + s->current_picture.motion_val[0][xy + 1 + v->blocks_off][0] = 0; + s->current_picture.motion_val[0][xy + 1 + v->blocks_off][1] = 0; + s->current_picture.motion_val[0][xy + wrap + v->blocks_off][0] = 0; + s->current_picture.motion_val[0][xy + wrap + v->blocks_off][1] = 0; + s->current_picture.motion_val[0][xy + wrap + 1 + v->blocks_off][0] = 0; + s->current_picture.motion_val[0][xy + wrap + 1 + v->blocks_off][1] = 0; v->luma_mv[s->mb_x][0] = v->luma_mv[s->mb_x][1] = 0; - s->current_picture.f.motion_val[1][xy + 1 + v->blocks_off][0] = 0; - s->current_picture.f.motion_val[1][xy + 1 + v->blocks_off][1] = 0; - s->current_picture.f.motion_val[1][xy + wrap][0] = 0; - s->current_picture.f.motion_val[1][xy + wrap + v->blocks_off][1] = 0; - s->current_picture.f.motion_val[1][xy + wrap + 1 + v->blocks_off][0] = 0; - s->current_picture.f.motion_val[1][xy + wrap + 1 + v->blocks_off][1] = 0; + s->current_picture.motion_val[1][xy + 1 + v->blocks_off][0] = 0; + s->current_picture.motion_val[1][xy + 1 + v->blocks_off][1] = 0; + s->current_picture.motion_val[1][xy + wrap][0] = 0; + s->current_picture.motion_val[1][xy + wrap + v->blocks_off][1] = 0; + s->current_picture.motion_val[1][xy + wrap + 1 + v->blocks_off][0] = 0; + s->current_picture.motion_val[1][xy + wrap + 1 + v->blocks_off][1] = 0; } return; } - C = s->current_picture.f.motion_val[dir][xy - 1 + v->blocks_off]; - A = s->current_picture.f.motion_val[dir][xy - wrap + v->blocks_off]; + C = s->current_picture.motion_val[dir][xy - 1 + v->blocks_off]; + A = s->current_picture.motion_val[dir][xy - wrap + v->blocks_off]; if (mv1) { if (v->field_mode && mixedmv_pic) off = (s->mb_x == (s->mb_width - 1)) ? -2 : 2; @@ -1414,7 +1494,7 @@ static inline void vc1_pred_mv(VC1Context *v, int n, int dmv_x, int dmv_y, off = -1; } } - B = s->current_picture.f.motion_val[dir][xy - wrap + off + v->blocks_off]; + B = s->current_picture.motion_val[dir][xy - wrap + off + v->blocks_off]; a_valid = !s->first_slice_line || (n == 2 || n == 3); b_valid = a_valid && (s->mb_width > 1); @@ -1457,13 +1537,19 @@ static inline void vc1_pred_mv(VC1Context *v, int n, int dmv_x, int dmv_y, } if (v->field_mode) { - if (num_samefield <= num_oppfield) - opposit = 1 - pred_flag; - else - opposit = pred_flag; + if (!v->numref) + // REFFIELD determines if the last field or the second-last field is + // to be used as reference + opposite = 1 - v->reffield; + else { + if (num_samefield <= num_oppfield) + opposite = 1 - pred_flag; + else + opposite = pred_flag; + } } else - opposit = 0; - if (opposit) { + opposite = 0; + if (opposite) { if (a_valid && !a_f) { field_predA[0] = scaleforopp(v, field_predA[0], 0, dir); field_predA[1] = scaleforopp(v, field_predA[1], 1, dir); @@ -1566,24 +1652,20 @@ static inline void vc1_pred_mv(VC1Context *v, int n, int dmv_x, int dmv_y, } } - if (v->field_mode && !s->quarter_sample) { - r_x <<= 1; - r_y <<= 1; - } if (v->field_mode && v->numref) r_y >>= 1; if (v->field_mode && v->cur_field_type && v->ref_field_type[dir] == 0) y_bias = 1; /* store MV using signed modulus of MV range defined in 4.11 */ - s->mv[dir][n][0] = s->current_picture.f.motion_val[dir][xy + v->blocks_off][0] = ((px + dmv_x + r_x) & ((r_x << 1) - 1)) - r_x; - s->mv[dir][n][1] = s->current_picture.f.motion_val[dir][xy + v->blocks_off][1] = ((py + dmv_y + r_y - y_bias) & ((r_y << 1) - 1)) - r_y + y_bias; + s->mv[dir][n][0] = s->current_picture.motion_val[dir][xy + v->blocks_off][0] = ((px + dmv_x + r_x) & ((r_x << 1) - 1)) - r_x; + s->mv[dir][n][1] = s->current_picture.motion_val[dir][xy + v->blocks_off][1] = ((py + dmv_y + r_y - y_bias) & ((r_y << 1) - 1)) - r_y + y_bias; if (mv1) { /* duplicate motion data for 1-MV block */ - s->current_picture.f.motion_val[dir][xy + 1 + v->blocks_off][0] = s->current_picture.f.motion_val[dir][xy + v->blocks_off][0]; - s->current_picture.f.motion_val[dir][xy + 1 + v->blocks_off][1] = s->current_picture.f.motion_val[dir][xy + v->blocks_off][1]; - s->current_picture.f.motion_val[dir][xy + wrap + v->blocks_off][0] = s->current_picture.f.motion_val[dir][xy + v->blocks_off][0]; - s->current_picture.f.motion_val[dir][xy + wrap + v->blocks_off][1] = s->current_picture.f.motion_val[dir][xy + v->blocks_off][1]; - s->current_picture.f.motion_val[dir][xy + wrap + 1 + v->blocks_off][0] = s->current_picture.f.motion_val[dir][xy + v->blocks_off][0]; - s->current_picture.f.motion_val[dir][xy + wrap + 1 + v->blocks_off][1] = s->current_picture.f.motion_val[dir][xy + v->blocks_off][1]; + s->current_picture.motion_val[dir][xy + 1 + v->blocks_off][0] = s->current_picture.motion_val[dir][xy + v->blocks_off][0]; + s->current_picture.motion_val[dir][xy + 1 + v->blocks_off][1] = s->current_picture.motion_val[dir][xy + v->blocks_off][1]; + s->current_picture.motion_val[dir][xy + wrap + v->blocks_off][0] = s->current_picture.motion_val[dir][xy + v->blocks_off][0]; + s->current_picture.motion_val[dir][xy + wrap + v->blocks_off][1] = s->current_picture.motion_val[dir][xy + v->blocks_off][1]; + s->current_picture.motion_val[dir][xy + wrap + 1 + v->blocks_off][0] = s->current_picture.motion_val[dir][xy + v->blocks_off][0]; + s->current_picture.motion_val[dir][xy + wrap + 1 + v->blocks_off][1] = s->current_picture.motion_val[dir][xy + v->blocks_off][1]; v->mv_f[dir][xy + 1 + v->blocks_off] = v->mv_f[dir][xy + v->blocks_off]; v->mv_f[dir][xy + wrap + v->blocks_off] = v->mv_f[dir][xy + wrap + 1 + v->blocks_off] = v->mv_f[dir][xy + v->blocks_off]; } @@ -1592,7 +1674,7 @@ static inline void vc1_pred_mv(VC1Context *v, int n, int dmv_x, int dmv_y, /** Predict and set motion vector for interlaced frame picture MBs */ static inline void vc1_pred_mv_intfr(VC1Context *v, int n, int dmv_x, int dmv_y, - int mvn, int r_x, int r_y, uint8_t* is_intra) + int mvn, int r_x, int r_y, uint8_t* is_intra, int dir) { MpegEncContext *s = &v->s; int xy, wrap, off = 0; @@ -1607,24 +1689,24 @@ static inline void vc1_pred_mv_intfr(VC1Context *v, int n, int dmv_x, int dmv_y, xy = s->block_index[n]; if (s->mb_intra) { - s->mv[0][n][0] = s->current_picture.f.motion_val[0][xy][0] = 0; - s->mv[0][n][1] = s->current_picture.f.motion_val[0][xy][1] = 0; - s->current_picture.f.motion_val[1][xy][0] = 0; - s->current_picture.f.motion_val[1][xy][1] = 0; + s->mv[0][n][0] = s->current_picture.motion_val[0][xy][0] = 0; + s->mv[0][n][1] = s->current_picture.motion_val[0][xy][1] = 0; + s->current_picture.motion_val[1][xy][0] = 0; + s->current_picture.motion_val[1][xy][1] = 0; if (mvn == 1) { /* duplicate motion data for 1-MV block */ - s->current_picture.f.motion_val[0][xy + 1][0] = 0; - s->current_picture.f.motion_val[0][xy + 1][1] = 0; - s->current_picture.f.motion_val[0][xy + wrap][0] = 0; - s->current_picture.f.motion_val[0][xy + wrap][1] = 0; - s->current_picture.f.motion_val[0][xy + wrap + 1][0] = 0; - s->current_picture.f.motion_val[0][xy + wrap + 1][1] = 0; + s->current_picture.motion_val[0][xy + 1][0] = 0; + s->current_picture.motion_val[0][xy + 1][1] = 0; + s->current_picture.motion_val[0][xy + wrap][0] = 0; + s->current_picture.motion_val[0][xy + wrap][1] = 0; + s->current_picture.motion_val[0][xy + wrap + 1][0] = 0; + s->current_picture.motion_val[0][xy + wrap + 1][1] = 0; v->luma_mv[s->mb_x][0] = v->luma_mv[s->mb_x][1] = 0; - s->current_picture.f.motion_val[1][xy + 1][0] = 0; - s->current_picture.f.motion_val[1][xy + 1][1] = 0; - s->current_picture.f.motion_val[1][xy + wrap][0] = 0; - s->current_picture.f.motion_val[1][xy + wrap][1] = 0; - s->current_picture.f.motion_val[1][xy + wrap + 1][0] = 0; - s->current_picture.f.motion_val[1][xy + wrap + 1][1] = 0; + s->current_picture.motion_val[1][xy + 1][0] = 0; + s->current_picture.motion_val[1][xy + 1][1] = 0; + s->current_picture.motion_val[1][xy + wrap][0] = 0; + s->current_picture.motion_val[1][xy + wrap][1] = 0; + s->current_picture.motion_val[1][xy + wrap + 1][0] = 0; + s->current_picture.motion_val[1][xy + wrap + 1][1] = 0; } return; } @@ -1634,14 +1716,14 @@ static inline void vc1_pred_mv_intfr(VC1Context *v, int n, int dmv_x, int dmv_y, if (s->mb_x || (n == 1) || (n == 3)) { if ((v->blk_mv_type[xy]) // current block (MB) has a field MV || (!v->blk_mv_type[xy] && !v->blk_mv_type[xy - 1])) { // or both have frame MV - A[0] = s->current_picture.f.motion_val[0][xy - 1][0]; - A[1] = s->current_picture.f.motion_val[0][xy - 1][1]; + A[0] = s->current_picture.motion_val[dir][xy - 1][0]; + A[1] = s->current_picture.motion_val[dir][xy - 1][1]; a_valid = 1; } else { // current block has frame mv and cand. has field MV (so average) - A[0] = (s->current_picture.f.motion_val[0][xy - 1][0] - + s->current_picture.f.motion_val[0][xy - 1 + off * wrap][0] + 1) >> 1; - A[1] = (s->current_picture.f.motion_val[0][xy - 1][1] - + s->current_picture.f.motion_val[0][xy - 1 + off * wrap][1] + 1) >> 1; + A[0] = (s->current_picture.motion_val[dir][xy - 1][0] + + s->current_picture.motion_val[dir][xy - 1 + off * wrap][0] + 1) >> 1; + A[1] = (s->current_picture.motion_val[dir][xy - 1][1] + + s->current_picture.motion_val[dir][xy - 1 + off * wrap][1] + 1) >> 1; a_valid = 1; } if (!(n & 1) && v->is_intra[s->mb_x - 1]) { @@ -1661,11 +1743,11 @@ static inline void vc1_pred_mv_intfr(VC1Context *v, int n, int dmv_x, int dmv_y, if (v->blk_mv_type[pos_b] && v->blk_mv_type[xy]) { n_adj = (n & 2) | (n & 1); } - B[0] = s->current_picture.f.motion_val[0][s->block_index[n_adj] - 2 * wrap][0]; - B[1] = s->current_picture.f.motion_val[0][s->block_index[n_adj] - 2 * wrap][1]; + B[0] = s->current_picture.motion_val[dir][s->block_index[n_adj] - 2 * wrap][0]; + B[1] = s->current_picture.motion_val[dir][s->block_index[n_adj] - 2 * wrap][1]; if (v->blk_mv_type[pos_b] && !v->blk_mv_type[xy]) { - B[0] = (B[0] + s->current_picture.f.motion_val[0][s->block_index[n_adj ^ 2] - 2 * wrap][0] + 1) >> 1; - B[1] = (B[1] + s->current_picture.f.motion_val[0][s->block_index[n_adj ^ 2] - 2 * wrap][1] + 1) >> 1; + B[0] = (B[0] + s->current_picture.motion_val[dir][s->block_index[n_adj ^ 2] - 2 * wrap][0] + 1) >> 1; + B[1] = (B[1] + s->current_picture.motion_val[dir][s->block_index[n_adj ^ 2] - 2 * wrap][1] + 1) >> 1; } } if (s->mb_width > 1) { @@ -1676,11 +1758,11 @@ static inline void vc1_pred_mv_intfr(VC1Context *v, int n, int dmv_x, int dmv_y, if (v->blk_mv_type[pos_c] && v->blk_mv_type[xy]) { n_adj = n & 2; } - C[0] = s->current_picture.f.motion_val[0][s->block_index[n_adj] - 2 * wrap + 2][0]; - C[1] = s->current_picture.f.motion_val[0][s->block_index[n_adj] - 2 * wrap + 2][1]; + C[0] = s->current_picture.motion_val[dir][s->block_index[n_adj] - 2 * wrap + 2][0]; + C[1] = s->current_picture.motion_val[dir][s->block_index[n_adj] - 2 * wrap + 2][1]; if (v->blk_mv_type[pos_c] && !v->blk_mv_type[xy]) { - C[0] = (1 + C[0] + (s->current_picture.f.motion_val[0][s->block_index[n_adj ^ 2] - 2 * wrap + 2][0])) >> 1; - C[1] = (1 + C[1] + (s->current_picture.f.motion_val[0][s->block_index[n_adj ^ 2] - 2 * wrap + 2][1])) >> 1; + C[0] = (1 + C[0] + (s->current_picture.motion_val[dir][s->block_index[n_adj ^ 2] - 2 * wrap + 2][0])) >> 1; + C[1] = (1 + C[1] + (s->current_picture.motion_val[dir][s->block_index[n_adj ^ 2] - 2 * wrap + 2][1])) >> 1; } if (s->mb_x == s->mb_width - 1) { if (!v->is_intra[s->mb_x - s->mb_stride - 1]) { @@ -1690,11 +1772,11 @@ static inline void vc1_pred_mv_intfr(VC1Context *v, int n, int dmv_x, int dmv_y, if (v->blk_mv_type[pos_c] && v->blk_mv_type[xy]) { n_adj = n | 1; } - C[0] = s->current_picture.f.motion_val[0][s->block_index[n_adj] - 2 * wrap - 2][0]; - C[1] = s->current_picture.f.motion_val[0][s->block_index[n_adj] - 2 * wrap - 2][1]; + C[0] = s->current_picture.motion_val[dir][s->block_index[n_adj] - 2 * wrap - 2][0]; + C[1] = s->current_picture.motion_val[dir][s->block_index[n_adj] - 2 * wrap - 2][1]; if (v->blk_mv_type[pos_c] && !v->blk_mv_type[xy]) { - C[0] = (1 + C[0] + s->current_picture.f.motion_val[0][s->block_index[1] - 2 * wrap - 2][0]) >> 1; - C[1] = (1 + C[1] + s->current_picture.f.motion_val[0][s->block_index[1] - 2 * wrap - 2][1]) >> 1; + C[0] = (1 + C[0] + s->current_picture.motion_val[dir][s->block_index[1] - 2 * wrap - 2][0]) >> 1; + C[1] = (1 + C[1] + s->current_picture.motion_val[dir][s->block_index[1] - 2 * wrap - 2][1]) >> 1; } } else c_valid = 0; @@ -1705,12 +1787,12 @@ static inline void vc1_pred_mv_intfr(VC1Context *v, int n, int dmv_x, int dmv_y, } else { pos_b = s->block_index[1]; b_valid = 1; - B[0] = s->current_picture.f.motion_val[0][pos_b][0]; - B[1] = s->current_picture.f.motion_val[0][pos_b][1]; + B[0] = s->current_picture.motion_val[dir][pos_b][0]; + B[1] = s->current_picture.motion_val[dir][pos_b][1]; pos_c = s->block_index[0]; c_valid = 1; - C[0] = s->current_picture.f.motion_val[0][pos_c][0]; - C[1] = s->current_picture.f.motion_val[0][pos_c][1]; + C[0] = s->current_picture.motion_val[dir][pos_c][0]; + C[1] = s->current_picture.motion_val[dir][pos_c][1]; } total_valid = a_valid + b_valid + c_valid; @@ -1788,7 +1870,8 @@ static inline void vc1_pred_mv_intfr(VC1Context *v, int n, int dmv_x, int dmv_y, } else if (c_valid) { px = C[0]; py = C[1]; - } + } else + px = py = 0; } } else if (total_valid == 1) { px = (a_valid) ? A[0] : ((b_valid) ? B[0] : C[0]); @@ -1798,20 +1881,20 @@ static inline void vc1_pred_mv_intfr(VC1Context *v, int n, int dmv_x, int dmv_y, } /* store MV using signed modulus of MV range defined in 4.11 */ - s->mv[0][n][0] = s->current_picture.f.motion_val[0][xy][0] = ((px + dmv_x + r_x) & ((r_x << 1) - 1)) - r_x; - s->mv[0][n][1] = s->current_picture.f.motion_val[0][xy][1] = ((py + dmv_y + r_y) & ((r_y << 1) - 1)) - r_y; + s->mv[dir][n][0] = s->current_picture.motion_val[dir][xy][0] = ((px + dmv_x + r_x) & ((r_x << 1) - 1)) - r_x; + s->mv[dir][n][1] = s->current_picture.motion_val[dir][xy][1] = ((py + dmv_y + r_y) & ((r_y << 1) - 1)) - r_y; if (mvn == 1) { /* duplicate motion data for 1-MV block */ - s->current_picture.f.motion_val[0][xy + 1 ][0] = s->current_picture.f.motion_val[0][xy][0]; - s->current_picture.f.motion_val[0][xy + 1 ][1] = s->current_picture.f.motion_val[0][xy][1]; - s->current_picture.f.motion_val[0][xy + wrap ][0] = s->current_picture.f.motion_val[0][xy][0]; - s->current_picture.f.motion_val[0][xy + wrap ][1] = s->current_picture.f.motion_val[0][xy][1]; - s->current_picture.f.motion_val[0][xy + wrap + 1][0] = s->current_picture.f.motion_val[0][xy][0]; - s->current_picture.f.motion_val[0][xy + wrap + 1][1] = s->current_picture.f.motion_val[0][xy][1]; + s->current_picture.motion_val[dir][xy + 1 ][0] = s->current_picture.motion_val[dir][xy][0]; + s->current_picture.motion_val[dir][xy + 1 ][1] = s->current_picture.motion_val[dir][xy][1]; + s->current_picture.motion_val[dir][xy + wrap ][0] = s->current_picture.motion_val[dir][xy][0]; + s->current_picture.motion_val[dir][xy + wrap ][1] = s->current_picture.motion_val[dir][xy][1]; + s->current_picture.motion_val[dir][xy + wrap + 1][0] = s->current_picture.motion_val[dir][xy][0]; + s->current_picture.motion_val[dir][xy + wrap + 1][1] = s->current_picture.motion_val[dir][xy][1]; } else if (mvn == 2) { /* duplicate motion data for 2-Field MV block */ - s->current_picture.f.motion_val[0][xy + 1][0] = s->current_picture.f.motion_val[0][xy][0]; - s->current_picture.f.motion_val[0][xy + 1][1] = s->current_picture.f.motion_val[0][xy][1]; - s->mv[0][n + 1][0] = s->mv[0][n][0]; - s->mv[0][n + 1][1] = s->mv[0][n][1]; + s->current_picture.motion_val[dir][xy + 1][0] = s->current_picture.motion_val[dir][xy][0]; + s->current_picture.motion_val[dir][xy + 1][1] = s->current_picture.motion_val[dir][xy][1]; + s->mv[dir][n + 1][0] = s->mv[dir][n][0]; + s->mv[dir][n + 1][1] = s->mv[dir][n][1]; } } @@ -1820,13 +1903,14 @@ static inline void vc1_pred_mv_intfr(VC1Context *v, int n, int dmv_x, int dmv_y, static void vc1_interp_mc(VC1Context *v) { MpegEncContext *s = &v->s; - DSPContext *dsp = &v->s.dsp; + 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; - if (!v->field_mode && !v->s.next_picture.f.data[0]) + if (!v->field_mode && !v->s.next_picture.f->data[0]) return; mx = s->mv[1][0][0]; @@ -1842,9 +1926,9 @@ static void vc1_interp_mc(VC1Context *v) uvmx = uvmx + ((uvmx < 0) ? -(uvmx & 1) : (uvmx & 1)); uvmy = uvmy + ((uvmy < 0) ? -(uvmy & 1) : (uvmy & 1)); } - srcY = s->next_picture.f.data[0]; - srcU = s->next_picture.f.data[1]; - srcV = s->next_picture.f.data[2]; + srcY = s->next_picture.f->data[0]; + srcU = s->next_picture.f->data[1]; + srcV = s->next_picture.f->data[2]; src_x = s->mb_x * 16 + (mx >> 2); src_y = s->mb_y * 16 + (my >> 2); @@ -1868,9 +1952,9 @@ static void vc1_interp_mc(VC1Context *v) srcV += uvsrc_y * s->uvlinesize + uvsrc_x; if (v->field_mode && v->ref_field_type[1]) { - srcY += s->current_picture_ptr->f.linesize[0]; - srcU += s->current_picture_ptr->f.linesize[1]; - srcV += s->current_picture_ptr->f.linesize[2]; + srcY += s->current_picture_ptr->f->linesize[0]; + srcU += s->current_picture_ptr->f->linesize[1]; + srcV += s->current_picture_ptr->f->linesize[2]; } /* for grayscale we should not try to read from unknown area */ @@ -1879,21 +1963,26 @@ static void vc1_interp_mc(VC1Context *v) srcV = s->edge_emu_buffer + 18 * s->linesize; } - if (v->rangeredfrm || s->h_edge_pos < 22 || v_edge_pos < 22 + 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; srcY -= s->mspel * (1 + s->linesize); - s->dsp.emulated_edge_mc(s->edge_emu_buffer, srcY, s->linesize, - 17 + s->mspel * 2, 17 + s->mspel * 2, - src_x - s->mspel, src_y - s->mspel, - s->h_edge_pos, v_edge_pos); + s->vdsp.emulated_edge_mc(s->edge_emu_buffer, srcY, + s->linesize, s->linesize, + 17 + s->mspel * 2, 17 + s->mspel * 2, + src_x - s->mspel, src_y - s->mspel, + s->h_edge_pos, v_edge_pos); srcY = s->edge_emu_buffer; - s->dsp.emulated_edge_mc(uvbuf , srcU, s->uvlinesize, 8 + 1, 8 + 1, - uvsrc_x, uvsrc_y, s->h_edge_pos >> 1, v_edge_pos >> 1); - s->dsp.emulated_edge_mc(uvbuf + 16, srcV, 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, 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, + 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; /* if we deal with range reduction we need to scale source blocks */ @@ -1918,16 +2007,37 @@ static void vc1_interp_mc(VC1Context *v) src2 += 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; + } + } srcY += s->mspel * (1 + s->linesize); } - if (v->field_mode && v->cur_field_type) { - off = s->current_picture_ptr->f.linesize[0]; - off_uv = s->current_picture_ptr->f.linesize[1]; - } else { - off = 0; - off_uv = 0; - } + off = 0; + off_uv = 0; if (s->mspel) { dxy = ((my & 3) << 2) | (mx & 3); @@ -1940,9 +2050,9 @@ static void vc1_interp_mc(VC1Context *v) dxy = (my & 2) | ((mx & 2) >> 1); if (!v->rnd) - dsp->avg_pixels_tab[0][dxy](s->dest[0] + off, srcY, s->linesize, 16); + s->hdsp.avg_pixels_tab[0][dxy](s->dest[0] + off, srcY, s->linesize, 16); else - dsp->avg_no_rnd_pixels_tab[0][dxy](s->dest[0] + off, srcY, s->linesize, 16); + s->hdsp.avg_no_rnd_pixels_tab[dxy](s->dest[0] + off, srcY, s->linesize, 16); } if (s->flags & CODEC_FLAG_GRAY) return; @@ -1950,8 +2060,8 @@ static void vc1_interp_mc(VC1Context *v) uvmx = (uvmx & 3) << 1; uvmy = (uvmy & 3) << 1; if (!v->rnd) { - dsp->avg_h264_chroma_pixels_tab[0](s->dest[1] + off_uv, srcU, s->uvlinesize, 8, uvmx, uvmy); - dsp->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] + 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); } 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); @@ -1982,30 +2092,18 @@ static av_always_inline int scale_mv(int value, int bfrac, int inv, int qs) static inline void vc1_b_mc(VC1Context *v, int dmv_x[2], int dmv_y[2], int direct, int mode) { - if (v->use_ic) { - v->mv_mode2 = v->mv_mode; - v->mv_mode = MV_PMODE_INTENSITY_COMP; - } if (direct) { vc1_mc_1mv(v, 0); vc1_interp_mc(v); - if (v->use_ic) - v->mv_mode = v->mv_mode2; return; } if (mode == BMV_TYPE_INTERPOLATED) { vc1_mc_1mv(v, 0); vc1_interp_mc(v); - if (v->use_ic) - v->mv_mode = v->mv_mode2; return; } - if (v->use_ic && (mode == BMV_TYPE_BACKWARD)) - v->mv_mode = v->mv_mode2; vc1_mc_1mv(v, (mode == BMV_TYPE_BACKWARD)); - if (v->use_ic) - v->mv_mode = v->mv_mode2; } static inline void vc1_pred_b_mv(VC1Context *v, int dmv_x[2], int dmv_y[2], @@ -2031,17 +2129,17 @@ static inline void vc1_pred_b_mv(VC1Context *v, int dmv_x[2], int dmv_y[2], xy = s->block_index[0]; if (s->mb_intra) { - s->current_picture.f.motion_val[0][xy + v->blocks_off][0] = - s->current_picture.f.motion_val[0][xy + v->blocks_off][1] = - s->current_picture.f.motion_val[1][xy + v->blocks_off][0] = - s->current_picture.f.motion_val[1][xy + v->blocks_off][1] = 0; + s->current_picture.motion_val[0][xy + v->blocks_off][0] = + s->current_picture.motion_val[0][xy + v->blocks_off][1] = + s->current_picture.motion_val[1][xy + v->blocks_off][0] = + s->current_picture.motion_val[1][xy + v->blocks_off][1] = 0; return; } if (!v->field_mode) { - s->mv[0][0][0] = scale_mv(s->next_picture.f.motion_val[1][xy][0], v->bfraction, 0, s->quarter_sample); - s->mv[0][0][1] = scale_mv(s->next_picture.f.motion_val[1][xy][1], v->bfraction, 0, s->quarter_sample); - s->mv[1][0][0] = scale_mv(s->next_picture.f.motion_val[1][xy][0], v->bfraction, 1, s->quarter_sample); - s->mv[1][0][1] = scale_mv(s->next_picture.f.motion_val[1][xy][1], v->bfraction, 1, s->quarter_sample); + s->mv[0][0][0] = scale_mv(s->next_picture.motion_val[1][xy][0], v->bfraction, 0, s->quarter_sample); + s->mv[0][0][1] = scale_mv(s->next_picture.motion_val[1][xy][1], v->bfraction, 0, s->quarter_sample); + s->mv[1][0][0] = scale_mv(s->next_picture.motion_val[1][xy][0], v->bfraction, 1, s->quarter_sample); + s->mv[1][0][1] = scale_mv(s->next_picture.motion_val[1][xy][1], v->bfraction, 1, s->quarter_sample); /* Pullback predicted motion vectors as specified in 8.4.5.4 */ s->mv[0][0][0] = av_clip(s->mv[0][0][0], -60 - (s->mb_x << 6), (s->mb_width << 6) - 4 - (s->mb_x << 6)); @@ -2050,18 +2148,18 @@ static inline void vc1_pred_b_mv(VC1Context *v, int dmv_x[2], int dmv_y[2], s->mv[1][0][1] = av_clip(s->mv[1][0][1], -60 - (s->mb_y << 6), (s->mb_height << 6) - 4 - (s->mb_y << 6)); } if (direct) { - s->current_picture.f.motion_val[0][xy + v->blocks_off][0] = s->mv[0][0][0]; - s->current_picture.f.motion_val[0][xy + v->blocks_off][1] = s->mv[0][0][1]; - s->current_picture.f.motion_val[1][xy + v->blocks_off][0] = s->mv[1][0][0]; - s->current_picture.f.motion_val[1][xy + v->blocks_off][1] = s->mv[1][0][1]; + s->current_picture.motion_val[0][xy + v->blocks_off][0] = s->mv[0][0][0]; + s->current_picture.motion_val[0][xy + v->blocks_off][1] = s->mv[0][0][1]; + s->current_picture.motion_val[1][xy + v->blocks_off][0] = s->mv[1][0][0]; + s->current_picture.motion_val[1][xy + v->blocks_off][1] = s->mv[1][0][1]; return; } if ((mvtype == BMV_TYPE_FORWARD) || (mvtype == BMV_TYPE_INTERPOLATED)) { - C = s->current_picture.f.motion_val[0][xy - 2]; - A = s->current_picture.f.motion_val[0][xy - wrap * 2]; + C = s->current_picture.motion_val[0][xy - 2]; + A = s->current_picture.motion_val[0][xy - wrap * 2]; off = (s->mb_x == (s->mb_width - 1)) ? -2 : 2; - B = s->current_picture.f.motion_val[0][xy - wrap * 2 + off]; + B = s->current_picture.motion_val[0][xy - wrap * 2 + off]; if (!s->mb_x) C[0] = C[1] = 0; if (!s->first_slice_line) { // predictor A is not out of bounds @@ -2136,10 +2234,10 @@ static inline void vc1_pred_b_mv(VC1Context *v, int dmv_x[2], int dmv_y[2], s->mv[0][0][1] = ((py + dmv_y[0] + r_y) & ((r_y << 1) - 1)) - r_y; } if ((mvtype == BMV_TYPE_BACKWARD) || (mvtype == BMV_TYPE_INTERPOLATED)) { - C = s->current_picture.f.motion_val[1][xy - 2]; - A = s->current_picture.f.motion_val[1][xy - wrap * 2]; + C = s->current_picture.motion_val[1][xy - 2]; + A = s->current_picture.motion_val[1][xy - wrap * 2]; off = (s->mb_x == (s->mb_width - 1)) ? -2 : 2; - B = s->current_picture.f.motion_val[1][xy - wrap * 2 + off]; + B = s->current_picture.motion_val[1][xy - wrap * 2 + off]; if (!s->mb_x) C[0] = C[1] = 0; @@ -2215,10 +2313,10 @@ static inline void vc1_pred_b_mv(VC1Context *v, int dmv_x[2], int dmv_y[2], s->mv[1][0][0] = ((px + dmv_x[1] + r_x) & ((r_x << 1) - 1)) - r_x; s->mv[1][0][1] = ((py + dmv_y[1] + r_y) & ((r_y << 1) - 1)) - r_y; } - s->current_picture.f.motion_val[0][xy][0] = s->mv[0][0][0]; - s->current_picture.f.motion_val[0][xy][1] = s->mv[0][0][1]; - s->current_picture.f.motion_val[1][xy][0] = s->mv[1][0][0]; - s->current_picture.f.motion_val[1][xy][1] = s->mv[1][0][1]; + s->current_picture.motion_val[0][xy][0] = s->mv[0][0][0]; + s->current_picture.motion_val[0][xy][1] = s->mv[0][0][1]; + s->current_picture.motion_val[1][xy][0] = s->mv[1][0][0]; + s->current_picture.motion_val[1][xy][1] = s->mv[1][0][1]; } static inline void vc1_pred_b_mv_intfi(VC1Context *v, int n, int *dmv_x, int *dmv_y, int mv1, int *pred_flag) @@ -2229,14 +2327,14 @@ static inline void vc1_pred_b_mv_intfi(VC1Context *v, int n, int *dmv_x, int *dm if (v->bmvtype == BMV_TYPE_DIRECT) { int total_opp, k, f; - if (s->next_picture.f.mb_type[mb_pos + v->mb_off] != MB_TYPE_INTRA) { - s->mv[0][0][0] = scale_mv(s->next_picture.f.motion_val[1][s->block_index[0] + v->blocks_off][0], + if (s->next_picture.mb_type[mb_pos + v->mb_off] != MB_TYPE_INTRA) { + s->mv[0][0][0] = scale_mv(s->next_picture.motion_val[1][s->block_index[0] + v->blocks_off][0], v->bfraction, 0, s->quarter_sample); - s->mv[0][0][1] = scale_mv(s->next_picture.f.motion_val[1][s->block_index[0] + v->blocks_off][1], + s->mv[0][0][1] = scale_mv(s->next_picture.motion_val[1][s->block_index[0] + v->blocks_off][1], v->bfraction, 0, s->quarter_sample); - s->mv[1][0][0] = scale_mv(s->next_picture.f.motion_val[1][s->block_index[0] + v->blocks_off][0], + s->mv[1][0][0] = scale_mv(s->next_picture.motion_val[1][s->block_index[0] + v->blocks_off][0], v->bfraction, 1, s->quarter_sample); - s->mv[1][0][1] = scale_mv(s->next_picture.f.motion_val[1][s->block_index[0] + v->blocks_off][1], + s->mv[1][0][1] = scale_mv(s->next_picture.motion_val[1][s->block_index[0] + v->blocks_off][1], v->bfraction, 1, s->quarter_sample); total_opp = v->mv_f_next[0][s->block_index[0] + v->blocks_off] @@ -2251,10 +2349,10 @@ static inline void vc1_pred_b_mv_intfi(VC1Context *v, int n, int *dmv_x, int *dm } v->ref_field_type[0] = v->ref_field_type[1] = v->cur_field_type ^ f; for (k = 0; k < 4; k++) { - s->current_picture.f.motion_val[0][s->block_index[k] + v->blocks_off][0] = s->mv[0][0][0]; - s->current_picture.f.motion_val[0][s->block_index[k] + v->blocks_off][1] = s->mv[0][0][1]; - s->current_picture.f.motion_val[1][s->block_index[k] + v->blocks_off][0] = s->mv[1][0][0]; - s->current_picture.f.motion_val[1][s->block_index[k] + v->blocks_off][1] = s->mv[1][0][1]; + s->current_picture.motion_val[0][s->block_index[k] + v->blocks_off][0] = s->mv[0][0][0]; + s->current_picture.motion_val[0][s->block_index[k] + v->blocks_off][1] = s->mv[0][0][1]; + s->current_picture.motion_val[1][s->block_index[k] + v->blocks_off][0] = s->mv[1][0][0]; + s->current_picture.motion_val[1][s->block_index[k] + v->blocks_off][1] = s->mv[1][0][1]; v->mv_f[0][s->block_index[k] + v->blocks_off] = f; v->mv_f[1][s->block_index[k] + v->blocks_off] = f; } @@ -2372,17 +2470,17 @@ static inline int vc1_pred_dc(MpegEncContext *s, int overlap, int pq, int n, b = dc_val[ - 1 - wrap]; a = dc_val[ - wrap]; /* scale predictors if needed */ - q1 = s->current_picture.f.qscale_table[mb_pos]; + q1 = s->current_picture.qscale_table[mb_pos]; dqscale_index = s->y_dc_scale_table[q1] - 1; if (dqscale_index < 0) return 0; if (c_avail && (n != 1 && n != 3)) { - q2 = s->current_picture.f.qscale_table[mb_pos - 1]; + q2 = s->current_picture.qscale_table[mb_pos - 1]; if (q2 && q2 != q1) c = (c * s->y_dc_scale_table[q2] * ff_vc1_dqscale[dqscale_index] + 0x20000) >> 18; } if (a_avail && (n != 2 && n != 3)) { - q2 = s->current_picture.f.qscale_table[mb_pos - s->mb_stride]; + q2 = s->current_picture.qscale_table[mb_pos - s->mb_stride]; if (q2 && q2 != q1) a = (a * s->y_dc_scale_table[q2] * ff_vc1_dqscale[dqscale_index] + 0x20000) >> 18; } @@ -2392,7 +2490,7 @@ static inline int vc1_pred_dc(MpegEncContext *s, int overlap, int pq, int n, off--; if (n != 2) off -= s->mb_stride; - q2 = s->current_picture.f.qscale_table[off]; + q2 = s->current_picture.qscale_table[off]; if (q2 && q2 != q1) b = (b * s->y_dc_scale_table[q2] * ff_vc1_dqscale[dqscale_index] + 0x20000) >> 18; } @@ -2531,7 +2629,7 @@ static void vc1_decode_ac_coeff(VC1Context *v, int *last, int *skip, * @param coded are AC coeffs present or not * @param codingset set of VLC to decode data */ -static int vc1_decode_i_block(VC1Context *v, DCTELEM block[64], int n, +static int vc1_decode_i_block(VC1Context *v, int16_t block[64], int n, int coded, int codingset) { GetBitContext *gb = &v->s.gb; @@ -2694,7 +2792,7 @@ not_coded: * @param codingset set of VLC to decode data * @param mquant quantizer value for this macroblock */ -static int vc1_decode_i_block_adv(VC1Context *v, DCTELEM block[64], int n, +static int vc1_decode_i_block_adv(VC1Context *v, int16_t block[64], int n, int coded, int codingset, int mquant) { GetBitContext *gb = &v->s.gb; @@ -2763,11 +2861,11 @@ static int vc1_decode_i_block_adv(VC1Context *v, DCTELEM block[64], int n, else // top ac_val -= 16 * s->block_wrap[n]; - q1 = s->current_picture.f.qscale_table[mb_pos]; + q1 = s->current_picture.qscale_table[mb_pos]; if ( dc_pred_dir && c_avail && mb_pos) - q2 = s->current_picture.f.qscale_table[mb_pos - 1]; + q2 = s->current_picture.qscale_table[mb_pos - 1]; if (!dc_pred_dir && a_avail && mb_pos >= s->mb_stride) - q2 = s->current_picture.f.qscale_table[mb_pos - s->mb_stride]; + q2 = s->current_picture.qscale_table[mb_pos - s->mb_stride]; if ( dc_pred_dir && n == 1) q2 = q1; if (!dc_pred_dir && n == 2) @@ -2906,7 +3004,7 @@ static int vc1_decode_i_block_adv(VC1Context *v, DCTELEM block[64], int n, * @param mquant block quantizer * @param codingset set of VLC to decode data */ -static int vc1_decode_intra_block(VC1Context *v, DCTELEM block[64], int n, +static int vc1_decode_intra_block(VC1Context *v, int16_t block[64], int n, int coded, int mquant, int codingset) { GetBitContext *gb = &v->s.gb; @@ -2922,7 +3020,7 @@ static int vc1_decode_intra_block(VC1Context *v, DCTELEM block[64], int n, int scale; int q1, q2 = 0; - s->dsp.clear_block(block); + s->bdsp.clear_block(block); /* XXX: Guard against dumb values of mquant */ mquant = (mquant < 1) ? 0 : ((mquant > 31) ? 31 : mquant); @@ -2986,11 +3084,11 @@ static int vc1_decode_intra_block(VC1Context *v, DCTELEM block[64], int n, else //top ac_val -= 16 * s->block_wrap[n]; - q1 = s->current_picture.f.qscale_table[mb_pos]; + q1 = s->current_picture.qscale_table[mb_pos]; if (dc_pred_dir && c_avail && mb_pos) - q2 = s->current_picture.f.qscale_table[mb_pos - 1]; + q2 = s->current_picture.qscale_table[mb_pos - 1]; if (!dc_pred_dir && a_avail && mb_pos >= s->mb_stride) - q2 = s->current_picture.f.qscale_table[mb_pos - s->mb_stride]; + q2 = s->current_picture.qscale_table[mb_pos - s->mb_stride]; if ( dc_pred_dir && n == 1) q2 = q1; if (!dc_pred_dir && n == 2) @@ -3116,7 +3214,7 @@ static int vc1_decode_intra_block(VC1Context *v, DCTELEM block[64], int n, /** Decode P block */ -static int vc1_decode_p_block(VC1Context *v, DCTELEM block[64], int n, +static int vc1_decode_p_block(VC1Context *v, int16_t block[64], int n, int mquant, int ttmb, int first_block, uint8_t *dst, int linesize, int skip_block, int *ttmb_out) @@ -3129,7 +3227,7 @@ static int vc1_decode_p_block(VC1Context *v, DCTELEM block[64], int n, int ttblk = ttmb & 7; int pat = 0; - s->dsp.clear_block(block); + s->bdsp.clear_block(block); if (ttmb == -1) { ttblk = ff_vc1_ttblk_to_tt[v->tt_index][get_vlc2(gb, ff_vc1_ttblk_vlc[v->tt_index].table, VC1_TTBLK_VLC_BITS, 1)]; @@ -3182,7 +3280,7 @@ static int vc1_decode_p_block(VC1Context *v, DCTELEM block[64], int n, v->vc1dsp.vc1_inv_trans_8x8_dc(dst, linesize, block); else { v->vc1dsp.vc1_inv_trans_8x8(block); - s->dsp.add_pixels_clamped(block, dst, linesize); + s->idsp.add_pixels_clamped(block, dst, linesize); } } break; @@ -3308,7 +3406,7 @@ static av_always_inline void vc1_apply_p_v_loop_filter(VC1Context *v, int block_ bottom_is_intra = (block_num < 2) ? (mb_is_intra >> ((block_num + 2) * 4)) : (v->is_intra[s->mb_x] >> ((block_num - 2) * 4)); mv_stride = s->b8_stride; - mv = &s->current_picture.f.motion_val[0][s->block_index[block_num] - 2 * mv_stride]; + mv = &s->current_picture.motion_val[0][s->block_index[block_num] - 2 * mv_stride]; } if (bottom_is_intra & 1 || block_is_intra & 1 || @@ -3370,7 +3468,7 @@ static av_always_inline void vc1_apply_p_h_loop_filter(VC1Context *v, int block_ : (mb_cbp >> ((block_num + 1) * 4)); right_is_intra = (block_num & 1) ? (v->is_intra[s->mb_x - s->mb_stride] >> ((block_num - 1) * 4)) : (mb_is_intra >> ((block_num + 1) * 4)); - mv = &s->current_picture.f.motion_val[0][s->block_index[block_num] - s->b8_stride * 2 - 2]; + mv = &s->current_picture.motion_val[0][s->block_index[block_num] - s->b8_stride * 2 - 2]; } if (block_is_intra & 1 || right_is_intra & 1 || mv[0][0] != mv[1][0] || mv[0][1] != mv[1][1]) { v->vc1dsp.vc1_h_loop_filter8(dst, linesize, v->pq); @@ -3464,10 +3562,10 @@ static int vc1_decode_p_mb(VC1Context *v) GET_MVDATA(dmv_x, dmv_y); if (s->mb_intra) { - s->current_picture.f.motion_val[1][s->block_index[0]][0] = 0; - s->current_picture.f.motion_val[1][s->block_index[0]][1] = 0; + s->current_picture.motion_val[1][s->block_index[0]][0] = 0; + s->current_picture.motion_val[1][s->block_index[0]][1] = 0; } - s->current_picture.f.mb_type[mb_pos] = s->mb_intra ? MB_TYPE_INTRA : MB_TYPE_16x16; + s->current_picture.mb_type[mb_pos] = s->mb_intra ? MB_TYPE_INTRA : MB_TYPE_16x16; vc1_pred_mv(v, 0, dmv_x, dmv_y, 1, v->range_x, v->range_y, v->mb_type[0], 0, 0); /* FIXME Set DC val for inter block ? */ @@ -3484,7 +3582,7 @@ static int vc1_decode_p_mb(VC1Context *v) mquant = v->pq; cbp = 0; } - s->current_picture.f.qscale_table[mb_pos] = mquant; + s->current_picture.qscale_table[mb_pos] = mquant; if (!v->ttmbf && !s->mb_intra && mb_has_coeffs) ttmb = get_vlc2(gb, ff_vc1_ttmb_vlc[v->tt_index].table, @@ -3513,7 +3611,10 @@ static int vc1_decode_p_mb(VC1Context *v) if (v->rangeredfrm) for (j = 0; j < 64; j++) s->block[i][j] <<= 1; - s->dsp.put_signed_pixels_clamped(s->block[i], s->dest[dst_idx] + off, i & 4 ? s->uvlinesize : s->linesize); + s->idsp.put_signed_pixels_clamped(s->block[i], + s->dest[dst_idx] + off, + i & 4 ? s->uvlinesize + : s->linesize); if (v->pq >= 9 && v->overlap) { if (v->c_avail) v->vc1dsp.vc1_h_overlap(s->dest[dst_idx] + off, i & 4 ? s->uvlinesize : s->linesize); @@ -3538,8 +3639,8 @@ static int vc1_decode_p_mb(VC1Context *v) v->mb_type[0][s->block_index[i]] = 0; s->dc_val[0][s->block_index[i]] = 0; } - s->current_picture.f.mb_type[mb_pos] = MB_TYPE_SKIP; - s->current_picture.f.qscale_table[mb_pos] = 0; + s->current_picture.mb_type[mb_pos] = MB_TYPE_SKIP; + s->current_picture.qscale_table[mb_pos] = 0; vc1_pred_mv(v, 0, 0, 0, 1, v->range_x, v->range_y, v->mb_type[0], 0, 0); vc1_mc_1mv(v, 0); } @@ -3562,7 +3663,7 @@ static int vc1_decode_p_mb(VC1Context *v) } vc1_pred_mv(v, i, dmv_x, dmv_y, 0, v->range_x, v->range_y, v->mb_type[0], 0, 0); if (!s->mb_intra) - vc1_mc_4mv_luma(v, i, 0); + vc1_mc_4mv_luma(v, i, 0, 0); intra_count += s->mb_intra; is_intra[i] = s->mb_intra; is_coded[i] = mb_has_coeffs; @@ -3582,7 +3683,7 @@ static int vc1_decode_p_mb(VC1Context *v) if (!intra_count && !coded_inter) goto end; GET_MQUANT(); - s->current_picture.f.qscale_table[mb_pos] = mquant; + s->current_picture.qscale_table[mb_pos] = mquant; /* test if block is intra and has pred */ { int intrapred = 0; @@ -3621,8 +3722,10 @@ static int vc1_decode_p_mb(VC1Context *v) if (v->rangeredfrm) for (j = 0; j < 64; j++) s->block[i][j] <<= 1; - s->dsp.put_signed_pixels_clamped(s->block[i], s->dest[dst_idx] + off, - (i & 4) ? s->uvlinesize : s->linesize); + s->idsp.put_signed_pixels_clamped(s->block[i], + s->dest[dst_idx] + off, + (i & 4) ? s->uvlinesize + : s->linesize); if (v->pq >= 9 && v->overlap) { if (v->c_avail) v->vc1dsp.vc1_h_overlap(s->dest[dst_idx] + off, i & 4 ? s->uvlinesize : s->linesize); @@ -3645,17 +3748,17 @@ static int vc1_decode_p_mb(VC1Context *v) } } else { // skipped MB s->mb_intra = 0; - s->current_picture.f.qscale_table[mb_pos] = 0; + s->current_picture.qscale_table[mb_pos] = 0; for (i = 0; i < 6; i++) { v->mb_type[0][s->block_index[i]] = 0; s->dc_val[0][s->block_index[i]] = 0; } for (i = 0; i < 4; i++) { vc1_pred_mv(v, i, 0, 0, 0, v->range_x, v->range_y, v->mb_type[0], 0, 0); - vc1_mc_4mv_luma(v, i, 0); + vc1_mc_4mv_luma(v, i, 0, 0); } vc1_mc_4mv_chroma(v, 0); - s->current_picture.f.qscale_table[mb_pos] = 0; + s->current_picture.qscale_table[mb_pos] = 0; } } end: @@ -3730,9 +3833,11 @@ static int vc1_decode_p_mb_intfr(VC1Context *v) break; } if (ff_vc1_mbmode_intfrp[v->fourmvswitch][idx_mbmode][0] == MV_PMODE_INTFR_INTRA) { // intra MB - s->current_picture.f.motion_val[1][s->block_index[0]][0] = 0; - s->current_picture.f.motion_val[1][s->block_index[0]][1] = 0; - s->current_picture.f.mb_type[mb_pos] = MB_TYPE_INTRA; + for (i = 0; i < 4; i++) { + s->current_picture.motion_val[1][s->block_index[i]][0] = 0; + s->current_picture.motion_val[1][s->block_index[i]][1] = 0; + } + s->current_picture.mb_type[mb_pos] = MB_TYPE_INTRA; s->mb_intra = v->is_intra[s->mb_x] = 1; for (i = 0; i < 6; i++) v->mb_type[0][s->block_index[i]] = 1; @@ -3742,7 +3847,7 @@ static int vc1_decode_p_mb_intfr(VC1Context *v) cbp = 1 + get_vlc2(&v->s.gb, v->cbpcy_vlc->table, VC1_CBPCY_P_VLC_BITS, 2); v->s.ac_pred = v->acpred_plane[mb_pos] = get_bits1(gb); GET_MQUANT(); - s->current_picture.f.qscale_table[mb_pos] = mquant; + s->current_picture.qscale_table[mb_pos] = mquant; /* Set DC scale - y and c use the same (not sure if necessary here) */ s->y_dc_scale = s->y_dc_scale_table[mquant]; s->c_dc_scale = s->c_dc_scale_table[mquant]; @@ -3769,7 +3874,9 @@ static int vc1_decode_p_mb_intfr(VC1Context *v) stride_y = s->uvlinesize; off = 0; } - s->dsp.put_signed_pixels_clamped(s->block[i], s->dest[dst_idx] + off, stride_y); + s->idsp.put_signed_pixels_clamped(s->block[i], + s->dest[dst_idx] + off, + stride_y); //TODO: loop filter } @@ -3800,10 +3907,10 @@ static int vc1_decode_p_mb_intfr(VC1Context *v) if (val) { get_mvdata_interlaced(v, &dmv_x, &dmv_y, 0); } - vc1_pred_mv_intfr(v, i, dmv_x, dmv_y, 0, v->range_x, v->range_y, v->mb_type[0]); - vc1_mc_4mv_luma(v, i, 0); + vc1_pred_mv_intfr(v, i, dmv_x, dmv_y, 0, v->range_x, v->range_y, v->mb_type[0], 0); + vc1_mc_4mv_luma(v, i, 0, 0); } else if (i == 4) { - vc1_mc_4mv_chroma4(v); + vc1_mc_4mv_chroma4(v, 0, 0, 0); } } } else if (twomv) { @@ -3812,29 +3919,29 @@ static int vc1_decode_p_mb_intfr(VC1Context *v) if (mvbp & 2) { get_mvdata_interlaced(v, &dmv_x, &dmv_y, 0); } - vc1_pred_mv_intfr(v, 0, dmv_x, dmv_y, 2, v->range_x, v->range_y, v->mb_type[0]); - vc1_mc_4mv_luma(v, 0, 0); - vc1_mc_4mv_luma(v, 1, 0); + vc1_pred_mv_intfr(v, 0, dmv_x, dmv_y, 2, v->range_x, v->range_y, v->mb_type[0], 0); + vc1_mc_4mv_luma(v, 0, 0, 0); + vc1_mc_4mv_luma(v, 1, 0, 0); dmv_x = dmv_y = 0; if (mvbp & 1) { get_mvdata_interlaced(v, &dmv_x, &dmv_y, 0); } - vc1_pred_mv_intfr(v, 2, dmv_x, dmv_y, 2, v->range_x, v->range_y, v->mb_type[0]); - vc1_mc_4mv_luma(v, 2, 0); - vc1_mc_4mv_luma(v, 3, 0); - vc1_mc_4mv_chroma4(v); + vc1_pred_mv_intfr(v, 2, dmv_x, dmv_y, 2, v->range_x, v->range_y, v->mb_type[0], 0); + vc1_mc_4mv_luma(v, 2, 0, 0); + vc1_mc_4mv_luma(v, 3, 0, 0); + vc1_mc_4mv_chroma4(v, 0, 0, 0); } else { mvbp = ff_vc1_mbmode_intfrp[v->fourmvswitch][idx_mbmode][2]; dmv_x = dmv_y = 0; if (mvbp) { get_mvdata_interlaced(v, &dmv_x, &dmv_y, 0); } - vc1_pred_mv_intfr(v, 0, dmv_x, dmv_y, 1, v->range_x, v->range_y, v->mb_type[0]); + vc1_pred_mv_intfr(v, 0, dmv_x, dmv_y, 1, v->range_x, v->range_y, v->mb_type[0], 0); vc1_mc_1mv(v, 0); } if (cbp) GET_MQUANT(); // p. 227 - s->current_picture.f.qscale_table[mb_pos] = mquant; + s->current_picture.qscale_table[mb_pos] = mquant; if (!v->ttmbf && cbp) ttmb = get_vlc2(gb, ff_vc1_ttmb_vlc[v->tt_index].table, VC1_TTMB_VLC_BITS, 2); for (i = 0; i < 6; i++) { @@ -3863,13 +3970,13 @@ static int vc1_decode_p_mb_intfr(VC1Context *v) v->mb_type[0][s->block_index[i]] = 0; s->dc_val[0][s->block_index[i]] = 0; } - s->current_picture.f.mb_type[mb_pos] = MB_TYPE_SKIP; - s->current_picture.f.qscale_table[mb_pos] = 0; + s->current_picture.mb_type[mb_pos] = MB_TYPE_SKIP; + s->current_picture.qscale_table[mb_pos] = 0; v->blk_mv_type[s->block_index[0]] = 0; v->blk_mv_type[s->block_index[1]] = 0; v->blk_mv_type[s->block_index[2]] = 0; v->blk_mv_type[s->block_index[3]] = 0; - vc1_pred_mv_intfr(v, 0, 0, 0, 1, v->range_x, v->range_y, v->mb_type[0]); + vc1_pred_mv_intfr(v, 0, 0, 0, 1, v->range_x, v->range_y, v->mb_type[0], 0); vc1_mc_1mv(v, 0); } if (s->mb_x == s->mb_width - 1) @@ -3901,11 +4008,11 @@ static int vc1_decode_p_mb_intfi(VC1Context *v) idx_mbmode = get_vlc2(gb, v->mbmode_vlc->table, VC1_IF_MBMODE_VLC_BITS, 2); if (idx_mbmode <= 1) { // intra MB s->mb_intra = v->is_intra[s->mb_x] = 1; - s->current_picture.f.motion_val[1][s->block_index[0] + v->blocks_off][0] = 0; - s->current_picture.f.motion_val[1][s->block_index[0] + v->blocks_off][1] = 0; - s->current_picture.f.mb_type[mb_pos + v->mb_off] = MB_TYPE_INTRA; + 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; + s->current_picture.mb_type[mb_pos + v->mb_off] = MB_TYPE_INTRA; GET_MQUANT(); - s->current_picture.f.qscale_table[mb_pos] = mquant; + s->current_picture.qscale_table[mb_pos] = mquant; /* Set DC scale - y and c use the same (not sure if necessary here) */ s->y_dc_scale = s->y_dc_scale_table[mquant]; s->c_dc_scale = s->c_dc_scale_table[mquant]; @@ -3931,13 +4038,15 @@ static int vc1_decode_p_mb_intfi(VC1Context *v) continue; v->vc1dsp.vc1_inv_trans_8x8(s->block[i]); off = (i & 4) ? 0 : ((i & 1) * 8 + (i & 2) * 4 * s->linesize); - off += v->cur_field_type ? ((i & 4) ? s->current_picture_ptr->f.linesize[1] : s->current_picture_ptr->f.linesize[0]) : 0; - s->dsp.put_signed_pixels_clamped(s->block[i], s->dest[dst_idx] + off, (i & 4) ? s->uvlinesize : s->linesize); + s->idsp.put_signed_pixels_clamped(s->block[i], + s->dest[dst_idx] + off, + (i & 4) ? s->uvlinesize + : s->linesize); // TODO: loop filter } } else { s->mb_intra = v->is_intra[s->mb_x] = 0; - s->current_picture.f.mb_type[mb_pos + v->mb_off] = MB_TYPE_16x16; + s->current_picture.mb_type[mb_pos + v->mb_off] = MB_TYPE_16x16; for (i = 0; i < 6; i++) v->mb_type[0][s->block_index[i]] = 0; if (idx_mbmode <= 5) { // 1-MV dmv_x = dmv_y = pred_flag = 0; @@ -3957,7 +4066,7 @@ static int vc1_decode_p_mb_intfi(VC1Context *v) get_mvdata_interlaced(v, &dmv_x, &dmv_y, &pred_flag); } vc1_pred_mv(v, i, dmv_x, dmv_y, 0, v->range_x, v->range_y, v->mb_type[0], pred_flag, 0); - vc1_mc_4mv_luma(v, i, 0); + vc1_mc_4mv_luma(v, i, 0, 0); } else if (i == 4) vc1_mc_4mv_chroma(v, 0); } @@ -3968,7 +4077,7 @@ static int vc1_decode_p_mb_intfi(VC1Context *v) if (cbp) { GET_MQUANT(); } - s->current_picture.f.qscale_table[mb_pos] = mquant; + s->current_picture.qscale_table[mb_pos] = mquant; if (!v->ttmbf && cbp) { ttmb = get_vlc2(gb, ff_vc1_ttmb_vlc[v->tt_index].table, VC1_TTMB_VLC_BITS, 2); } @@ -3978,8 +4087,6 @@ static int vc1_decode_p_mb_intfi(VC1Context *v) dst_idx += i >> 2; val = ((cbp >> (5 - i)) & 1); off = (i & 4) ? 0 : (i & 1) * 8 + (i & 2) * 4 * s->linesize; - if (v->cur_field_type) - off += (i & 4) ? s->current_picture_ptr->f.linesize[1] : s->current_picture_ptr->f.linesize[0]; if (val) { pat = vc1_decode_p_block(v, s->block[i], i, mquant, ttmb, first_block, s->dest[dst_idx] + off, @@ -4034,7 +4141,7 @@ static void vc1_decode_b_mb(VC1Context *v) v->mb_type[0][s->block_index[i]] = 0; s->dc_val[0][s->block_index[i]] = 0; } - s->current_picture.f.qscale_table[mb_pos] = 0; + s->current_picture.qscale_table[mb_pos] = 0; if (!direct) { if (!skipped) { @@ -4071,7 +4178,7 @@ static void vc1_decode_b_mb(VC1Context *v) cbp = get_vlc2(&v->s.gb, v->cbpcy_vlc->table, VC1_CBPCY_P_VLC_BITS, 2); GET_MQUANT(); s->mb_intra = 0; - s->current_picture.f.qscale_table[mb_pos] = mquant; + s->current_picture.qscale_table[mb_pos] = mquant; if (!v->ttmbf) ttmb = get_vlc2(gb, ff_vc1_ttmb_vlc[v->tt_index].table, VC1_TTMB_VLC_BITS, 2); dmv_x[0] = dmv_y[0] = dmv_x[1] = dmv_y[1] = 0; @@ -4086,7 +4193,7 @@ static void vc1_decode_b_mb(VC1Context *v) } if (s->mb_intra && !mb_has_coeffs) { GET_MQUANT(); - s->current_picture.f.qscale_table[mb_pos] = mquant; + s->current_picture.qscale_table[mb_pos] = mquant; s->ac_pred = get_bits1(gb); cbp = 0; vc1_pred_b_mv(v, dmv_x, dmv_y, direct, bmvtype); @@ -4108,7 +4215,7 @@ static void vc1_decode_b_mb(VC1Context *v) s->ac_pred = get_bits1(gb); cbp = get_vlc2(&v->s.gb, v->cbpcy_vlc->table, VC1_CBPCY_P_VLC_BITS, 2); GET_MQUANT(); - s->current_picture.f.qscale_table[mb_pos] = mquant; + s->current_picture.qscale_table[mb_pos] = mquant; if (!v->ttmbf && !s->mb_intra && mb_has_coeffs) ttmb = get_vlc2(gb, ff_vc1_ttmb_vlc[v->tt_index].table, VC1_TTMB_VLC_BITS, 2); } @@ -4136,7 +4243,10 @@ static void vc1_decode_b_mb(VC1Context *v) if (v->rangeredfrm) for (j = 0; j < 64; j++) s->block[i][j] <<= 1; - s->dsp.put_signed_pixels_clamped(s->block[i], s->dest[dst_idx] + off, i & 4 ? s->uvlinesize : s->linesize); + s->idsp.put_signed_pixels_clamped(s->block[i], + s->dest[dst_idx] + off, + i & 4 ? s->uvlinesize + : s->linesize); } else if (val) { vc1_decode_p_block(v, s->block[i], i, mquant, ttmb, first_block, s->dest[dst_idx] + off, @@ -4175,11 +4285,11 @@ static void vc1_decode_b_mb_intfi(VC1Context *v) idx_mbmode = get_vlc2(gb, v->mbmode_vlc->table, VC1_IF_MBMODE_VLC_BITS, 2); if (idx_mbmode <= 1) { // intra MB s->mb_intra = v->is_intra[s->mb_x] = 1; - s->current_picture.f.motion_val[1][s->block_index[0]][0] = 0; - s->current_picture.f.motion_val[1][s->block_index[0]][1] = 0; - s->current_picture.f.mb_type[mb_pos + v->mb_off] = MB_TYPE_INTRA; + s->current_picture.motion_val[1][s->block_index[0]][0] = 0; + s->current_picture.motion_val[1][s->block_index[0]][1] = 0; + s->current_picture.mb_type[mb_pos + v->mb_off] = MB_TYPE_INTRA; GET_MQUANT(); - s->current_picture.f.qscale_table[mb_pos] = mquant; + s->current_picture.qscale_table[mb_pos] = mquant; /* Set DC scale - y and c use the same (not sure if necessary here) */ s->y_dc_scale = s->y_dc_scale_table[mquant]; s->c_dc_scale = s->c_dc_scale_table[mquant]; @@ -4208,13 +4318,15 @@ static void vc1_decode_b_mb_intfi(VC1Context *v) for (j = 0; j < 64; j++) s->block[i][j] <<= 1; off = (i & 4) ? 0 : ((i & 1) * 8 + (i & 2) * 4 * s->linesize); - off += v->cur_field_type ? ((i & 4) ? s->current_picture_ptr->f.linesize[1] : s->current_picture_ptr->f.linesize[0]) : 0; - s->dsp.put_signed_pixels_clamped(s->block[i], s->dest[dst_idx] + off, (i & 4) ? s->uvlinesize : s->linesize); + s->idsp.put_signed_pixels_clamped(s->block[i], + s->dest[dst_idx] + off, + (i & 4) ? s->uvlinesize + : s->linesize); // TODO: yet to perform loop filter } } else { s->mb_intra = v->is_intra[s->mb_x] = 0; - s->current_picture.f.mb_type[mb_pos + v->mb_off] = MB_TYPE_16x16; + s->current_picture.mb_type[mb_pos + v->mb_off] = MB_TYPE_16x16; for (i = 0; i < 6; i++) v->mb_type[0][s->block_index[i]] = 0; if (v->fmb_is_raw) fwd = v->forward_mb_plane[mb_pos] = get_bits1(gb); @@ -4269,7 +4381,7 @@ static void vc1_decode_b_mb_intfi(VC1Context *v) &pred_flag[bmvtype == BMV_TYPE_BACKWARD]); } vc1_pred_b_mv_intfi(v, i, dmv_x, dmv_y, 0, pred_flag); - vc1_mc_4mv_luma(v, i, bmvtype == BMV_TYPE_BACKWARD); + vc1_mc_4mv_luma(v, i, bmvtype == BMV_TYPE_BACKWARD, 0); } else if (i == 4) vc1_mc_4mv_chroma(v, bmvtype == BMV_TYPE_BACKWARD); } @@ -4280,7 +4392,7 @@ static void vc1_decode_b_mb_intfi(VC1Context *v) if (cbp) { GET_MQUANT(); } - s->current_picture.f.qscale_table[mb_pos] = mquant; + s->current_picture.qscale_table[mb_pos] = mquant; if (!v->ttmbf && cbp) { ttmb = get_vlc2(gb, ff_vc1_ttmb_vlc[v->tt_index].table, VC1_TTMB_VLC_BITS, 2); } @@ -4290,8 +4402,6 @@ static void vc1_decode_b_mb_intfi(VC1Context *v) dst_idx += i >> 2; val = ((cbp >> (5 - i)) & 1); off = (i & 4) ? 0 : (i & 1) * 8 + (i & 2) * 4 * s->linesize; - if (v->cur_field_type) - off += (i & 4) ? s->current_picture_ptr->f.linesize[1] : s->current_picture_ptr->f.linesize[0]; if (val) { vc1_decode_p_block(v, s->block[i], i, mquant, ttmb, first_block, s->dest[dst_idx] + off, @@ -4305,6 +4415,352 @@ static void vc1_decode_b_mb_intfi(VC1Context *v) } } +/** Decode one B-frame MB (in interlaced frame B picture) + */ +static int vc1_decode_b_mb_intfr(VC1Context *v) +{ + MpegEncContext *s = &v->s; + GetBitContext *gb = &s->gb; + int i, j; + int mb_pos = s->mb_x + s->mb_y * s->mb_stride; + int cbp = 0; /* cbp decoding stuff */ + int mqdiff, mquant; /* MB quantization */ + int ttmb = v->ttfrm; /* MB Transform type */ + int mvsw = 0; /* motion vector switch */ + int mb_has_coeffs = 1; /* last_flag */ + int dmv_x, dmv_y; /* Differential MV components */ + int val; /* temp value */ + int first_block = 1; + int dst_idx, off; + int skipped, direct, twomv = 0; + int block_cbp = 0, pat, block_tt = 0; + int idx_mbmode = 0, mvbp; + int stride_y, fieldtx; + int bmvtype = BMV_TYPE_BACKWARD; + int dir, dir2; + + mquant = v->pq; /* Lossy initialization */ + s->mb_intra = 0; + if (v->skip_is_raw) + skipped = get_bits1(gb); + else + skipped = v->s.mbskip_table[mb_pos]; + + if (!skipped) { + idx_mbmode = get_vlc2(gb, v->mbmode_vlc->table, VC1_INTFR_NON4MV_MBMODE_VLC_BITS, 2); + if (ff_vc1_mbmode_intfrp[0][idx_mbmode][0] == MV_PMODE_INTFR_2MV_FIELD) { + twomv = 1; + v->blk_mv_type[s->block_index[0]] = 1; + v->blk_mv_type[s->block_index[1]] = 1; + v->blk_mv_type[s->block_index[2]] = 1; + v->blk_mv_type[s->block_index[3]] = 1; + } else { + v->blk_mv_type[s->block_index[0]] = 0; + v->blk_mv_type[s->block_index[1]] = 0; + v->blk_mv_type[s->block_index[2]] = 0; + v->blk_mv_type[s->block_index[3]] = 0; + } + } + + if (v->dmb_is_raw) + direct = get_bits1(gb); + else + direct = v->direct_mb_plane[mb_pos]; + + if (direct) { + s->mv[0][0][0] = s->current_picture.motion_val[0][s->block_index[0]][0] = scale_mv(s->next_picture.motion_val[1][s->block_index[0]][0], v->bfraction, 0, s->quarter_sample); + s->mv[0][0][1] = s->current_picture.motion_val[0][s->block_index[0]][1] = scale_mv(s->next_picture.motion_val[1][s->block_index[0]][1], v->bfraction, 0, s->quarter_sample); + s->mv[1][0][0] = s->current_picture.motion_val[1][s->block_index[0]][0] = scale_mv(s->next_picture.motion_val[1][s->block_index[0]][0], v->bfraction, 1, s->quarter_sample); + s->mv[1][0][1] = s->current_picture.motion_val[1][s->block_index[0]][1] = scale_mv(s->next_picture.motion_val[1][s->block_index[0]][1], v->bfraction, 1, s->quarter_sample); + + if (twomv) { + s->mv[0][2][0] = s->current_picture.motion_val[0][s->block_index[2]][0] = scale_mv(s->next_picture.motion_val[1][s->block_index[2]][0], v->bfraction, 0, s->quarter_sample); + s->mv[0][2][1] = s->current_picture.motion_val[0][s->block_index[2]][1] = scale_mv(s->next_picture.motion_val[1][s->block_index[2]][1], v->bfraction, 0, s->quarter_sample); + s->mv[1][2][0] = s->current_picture.motion_val[1][s->block_index[2]][0] = scale_mv(s->next_picture.motion_val[1][s->block_index[2]][0], v->bfraction, 1, s->quarter_sample); + s->mv[1][2][1] = s->current_picture.motion_val[1][s->block_index[2]][1] = scale_mv(s->next_picture.motion_val[1][s->block_index[2]][1], v->bfraction, 1, s->quarter_sample); + + for (i = 1; i < 4; i += 2) { + s->mv[0][i][0] = s->current_picture.motion_val[0][s->block_index[i]][0] = s->mv[0][i-1][0]; + s->mv[0][i][1] = s->current_picture.motion_val[0][s->block_index[i]][1] = s->mv[0][i-1][1]; + s->mv[1][i][0] = s->current_picture.motion_val[1][s->block_index[i]][0] = s->mv[1][i-1][0]; + s->mv[1][i][1] = s->current_picture.motion_val[1][s->block_index[i]][1] = s->mv[1][i-1][1]; + } + } else { + for (i = 1; i < 4; i++) { + s->mv[0][i][0] = s->current_picture.motion_val[0][s->block_index[i]][0] = s->mv[0][0][0]; + s->mv[0][i][1] = s->current_picture.motion_val[0][s->block_index[i]][1] = s->mv[0][0][1]; + s->mv[1][i][0] = s->current_picture.motion_val[1][s->block_index[i]][0] = s->mv[1][0][0]; + s->mv[1][i][1] = s->current_picture.motion_val[1][s->block_index[i]][1] = s->mv[1][0][1]; + } + } + } + + if (ff_vc1_mbmode_intfrp[0][idx_mbmode][0] == MV_PMODE_INTFR_INTRA) { // intra MB + for (i = 0; i < 4; i++) { + s->mv[0][i][0] = s->current_picture.motion_val[0][s->block_index[i]][0] = 0; + s->mv[0][i][1] = s->current_picture.motion_val[0][s->block_index[i]][1] = 0; + s->mv[1][i][0] = s->current_picture.motion_val[1][s->block_index[i]][0] = 0; + s->mv[1][i][1] = s->current_picture.motion_val[1][s->block_index[i]][1] = 0; + } + s->current_picture.mb_type[mb_pos] = MB_TYPE_INTRA; + s->mb_intra = v->is_intra[s->mb_x] = 1; + for (i = 0; i < 6; i++) + v->mb_type[0][s->block_index[i]] = 1; + fieldtx = v->fieldtx_plane[mb_pos] = get_bits1(gb); + mb_has_coeffs = get_bits1(gb); + if (mb_has_coeffs) + cbp = 1 + get_vlc2(&v->s.gb, v->cbpcy_vlc->table, VC1_CBPCY_P_VLC_BITS, 2); + v->s.ac_pred = v->acpred_plane[mb_pos] = get_bits1(gb); + GET_MQUANT(); + s->current_picture.qscale_table[mb_pos] = mquant; + /* Set DC scale - y and c use the same (not sure if necessary here) */ + s->y_dc_scale = s->y_dc_scale_table[mquant]; + s->c_dc_scale = s->c_dc_scale_table[mquant]; + dst_idx = 0; + for (i = 0; i < 6; i++) { + s->dc_val[0][s->block_index[i]] = 0; + dst_idx += i >> 2; + val = ((cbp >> (5 - i)) & 1); + v->mb_type[0][s->block_index[i]] = s->mb_intra; + v->a_avail = v->c_avail = 0; + if (i == 2 || i == 3 || !s->first_slice_line) + v->a_avail = v->mb_type[0][s->block_index[i] - s->block_wrap[i]]; + if (i == 1 || i == 3 || s->mb_x) + v->c_avail = v->mb_type[0][s->block_index[i] - 1]; + + vc1_decode_intra_block(v, s->block[i], i, val, mquant, + (i & 4) ? v->codingset2 : v->codingset); + if (i > 3 && (s->flags & CODEC_FLAG_GRAY)) + continue; + v->vc1dsp.vc1_inv_trans_8x8(s->block[i]); + if (i < 4) { + stride_y = s->linesize << fieldtx; + off = (fieldtx) ? ((i & 1) * 8) + ((i & 2) >> 1) * s->linesize : (i & 1) * 8 + 4 * (i & 2) * s->linesize; + } else { + stride_y = s->uvlinesize; + off = 0; + } + s->idsp.put_signed_pixels_clamped(s->block[i], + s->dest[dst_idx] + off, + stride_y); + } + } else { + s->mb_intra = v->is_intra[s->mb_x] = 0; + if (!direct) { + if (skipped || !s->mb_intra) { + bmvtype = decode012(gb); + switch (bmvtype) { + case 0: + bmvtype = (v->bfraction >= (B_FRACTION_DEN/2)) ? BMV_TYPE_BACKWARD : BMV_TYPE_FORWARD; + break; + case 1: + bmvtype = (v->bfraction >= (B_FRACTION_DEN/2)) ? BMV_TYPE_FORWARD : BMV_TYPE_BACKWARD; + break; + case 2: + bmvtype = BMV_TYPE_INTERPOLATED; + } + } + + if (twomv && bmvtype != BMV_TYPE_INTERPOLATED) + mvsw = get_bits1(gb); + } + + if (!skipped) { // inter MB + mb_has_coeffs = ff_vc1_mbmode_intfrp[0][idx_mbmode][3]; + if (mb_has_coeffs) + cbp = 1 + get_vlc2(&v->s.gb, v->cbpcy_vlc->table, VC1_CBPCY_P_VLC_BITS, 2); + if (!direct) { + if (bmvtype == BMV_TYPE_INTERPOLATED & twomv) { + v->fourmvbp = get_vlc2(gb, v->fourmvbp_vlc->table, VC1_4MV_BLOCK_PATTERN_VLC_BITS, 1); + } else if (bmvtype == BMV_TYPE_INTERPOLATED | twomv) { + v->twomvbp = get_vlc2(gb, v->twomvbp_vlc->table, VC1_2MV_BLOCK_PATTERN_VLC_BITS, 1); + } + } + + for (i = 0; i < 6; i++) + v->mb_type[0][s->block_index[i]] = 0; + fieldtx = v->fieldtx_plane[mb_pos] = ff_vc1_mbmode_intfrp[0][idx_mbmode][1]; + /* for all motion vector read MVDATA and motion compensate each block */ + dst_idx = 0; + if (direct) { + if (twomv) { + for (i = 0; i < 4; i++) { + vc1_mc_4mv_luma(v, i, 0, 0); + vc1_mc_4mv_luma(v, i, 1, 1); + } + vc1_mc_4mv_chroma4(v, 0, 0, 0); + vc1_mc_4mv_chroma4(v, 1, 1, 1); + } else { + vc1_mc_1mv(v, 0); + vc1_interp_mc(v); + } + } else if (twomv && bmvtype == BMV_TYPE_INTERPOLATED) { + mvbp = v->fourmvbp; + for (i = 0; i < 4; i++) { + dir = i==1 || i==3; + dmv_x = dmv_y = 0; + val = ((mvbp >> (3 - i)) & 1); + if (val) + get_mvdata_interlaced(v, &dmv_x, &dmv_y, 0); + j = i > 1 ? 2 : 0; + vc1_pred_mv_intfr(v, j, dmv_x, dmv_y, 2, v->range_x, v->range_y, v->mb_type[0], dir); + vc1_mc_4mv_luma(v, j, dir, dir); + vc1_mc_4mv_luma(v, j+1, dir, dir); + } + + vc1_mc_4mv_chroma4(v, 0, 0, 0); + vc1_mc_4mv_chroma4(v, 1, 1, 1); + } else if (bmvtype == BMV_TYPE_INTERPOLATED) { + mvbp = v->twomvbp; + dmv_x = dmv_y = 0; + if (mvbp & 2) + get_mvdata_interlaced(v, &dmv_x, &dmv_y, 0); + + vc1_pred_mv_intfr(v, 0, dmv_x, dmv_y, 1, v->range_x, v->range_y, v->mb_type[0], 0); + vc1_mc_1mv(v, 0); + + dmv_x = dmv_y = 0; + if (mvbp & 1) + get_mvdata_interlaced(v, &dmv_x, &dmv_y, 0); + + vc1_pred_mv_intfr(v, 0, dmv_x, dmv_y, 1, v->range_x, v->range_y, v->mb_type[0], 1); + vc1_interp_mc(v); + } else if (twomv) { + dir = bmvtype == BMV_TYPE_BACKWARD; + dir2 = dir; + if (mvsw) + dir2 = !dir; + mvbp = v->twomvbp; + dmv_x = dmv_y = 0; + if (mvbp & 2) + get_mvdata_interlaced(v, &dmv_x, &dmv_y, 0); + vc1_pred_mv_intfr(v, 0, dmv_x, dmv_y, 2, v->range_x, v->range_y, v->mb_type[0], dir); + + dmv_x = dmv_y = 0; + if (mvbp & 1) + get_mvdata_interlaced(v, &dmv_x, &dmv_y, 0); + vc1_pred_mv_intfr(v, 2, dmv_x, dmv_y, 2, v->range_x, v->range_y, v->mb_type[0], dir2); + + if (mvsw) { + for (i = 0; i < 2; i++) { + s->mv[dir][i+2][0] = s->mv[dir][i][0] = s->current_picture.motion_val[dir][s->block_index[i+2]][0] = s->current_picture.motion_val[dir][s->block_index[i]][0]; + s->mv[dir][i+2][1] = s->mv[dir][i][1] = s->current_picture.motion_val[dir][s->block_index[i+2]][1] = s->current_picture.motion_val[dir][s->block_index[i]][1]; + s->mv[dir2][i+2][0] = s->mv[dir2][i][0] = s->current_picture.motion_val[dir2][s->block_index[i]][0] = s->current_picture.motion_val[dir2][s->block_index[i+2]][0]; + s->mv[dir2][i+2][1] = s->mv[dir2][i][1] = s->current_picture.motion_val[dir2][s->block_index[i]][1] = s->current_picture.motion_val[dir2][s->block_index[i+2]][1]; + } + } else { + vc1_pred_mv_intfr(v, 0, 0, 0, 2, v->range_x, v->range_y, v->mb_type[0], !dir); + vc1_pred_mv_intfr(v, 2, 0, 0, 2, v->range_x, v->range_y, v->mb_type[0], !dir); + } + + vc1_mc_4mv_luma(v, 0, dir, 0); + vc1_mc_4mv_luma(v, 1, dir, 0); + vc1_mc_4mv_luma(v, 2, dir2, 0); + vc1_mc_4mv_luma(v, 3, dir2, 0); + vc1_mc_4mv_chroma4(v, dir, dir2, 0); + } else { + dir = bmvtype == BMV_TYPE_BACKWARD; + + mvbp = ff_vc1_mbmode_intfrp[0][idx_mbmode][2]; + dmv_x = dmv_y = 0; + if (mvbp) + get_mvdata_interlaced(v, &dmv_x, &dmv_y, 0); + + vc1_pred_mv_intfr(v, 0, dmv_x, dmv_y, 1, v->range_x, v->range_y, v->mb_type[0], dir); + v->blk_mv_type[s->block_index[0]] = 1; + v->blk_mv_type[s->block_index[1]] = 1; + v->blk_mv_type[s->block_index[2]] = 1; + v->blk_mv_type[s->block_index[3]] = 1; + vc1_pred_mv_intfr(v, 0, 0, 0, 2, v->range_x, v->range_y, 0, !dir); + for (i = 0; i < 2; i++) { + s->mv[!dir][i+2][0] = s->mv[!dir][i][0] = s->current_picture.motion_val[!dir][s->block_index[i+2]][0] = s->current_picture.motion_val[!dir][s->block_index[i]][0]; + s->mv[!dir][i+2][1] = s->mv[!dir][i][1] = s->current_picture.motion_val[!dir][s->block_index[i+2]][1] = s->current_picture.motion_val[!dir][s->block_index[i]][1]; + } + vc1_mc_1mv(v, dir); + } + + if (cbp) + GET_MQUANT(); // p. 227 + s->current_picture.qscale_table[mb_pos] = mquant; + if (!v->ttmbf && cbp) + ttmb = get_vlc2(gb, ff_vc1_ttmb_vlc[v->tt_index].table, VC1_TTMB_VLC_BITS, 2); + for (i = 0; i < 6; i++) { + s->dc_val[0][s->block_index[i]] = 0; + dst_idx += i >> 2; + val = ((cbp >> (5 - i)) & 1); + if (!fieldtx) + off = (i & 4) ? 0 : ((i & 1) * 8 + (i & 2) * 4 * s->linesize); + else + off = (i & 4) ? 0 : ((i & 1) * 8 + ((i > 1) * s->linesize)); + if (val) { + pat = vc1_decode_p_block(v, s->block[i], i, mquant, ttmb, + first_block, s->dest[dst_idx] + off, + (i & 4) ? s->uvlinesize : (s->linesize << fieldtx), + (i & 4) && (s->flags & CODEC_FLAG_GRAY), &block_tt); + block_cbp |= pat << (i << 2); + if (!v->ttmbf && ttmb < 8) + ttmb = -1; + first_block = 0; + } + } + + } else { // skipped + dir = 0; + for (i = 0; i < 6; i++) { + v->mb_type[0][s->block_index[i]] = 0; + s->dc_val[0][s->block_index[i]] = 0; + } + s->current_picture.mb_type[mb_pos] = MB_TYPE_SKIP; + s->current_picture.qscale_table[mb_pos] = 0; + v->blk_mv_type[s->block_index[0]] = 0; + v->blk_mv_type[s->block_index[1]] = 0; + v->blk_mv_type[s->block_index[2]] = 0; + v->blk_mv_type[s->block_index[3]] = 0; + + if (!direct) { + if (bmvtype == BMV_TYPE_INTERPOLATED) { + vc1_pred_mv_intfr(v, 0, 0, 0, 1, v->range_x, v->range_y, v->mb_type[0], 0); + vc1_pred_mv_intfr(v, 0, 0, 0, 1, v->range_x, v->range_y, v->mb_type[0], 1); + } else { + dir = bmvtype == BMV_TYPE_BACKWARD; + vc1_pred_mv_intfr(v, 0, 0, 0, 1, v->range_x, v->range_y, v->mb_type[0], dir); + if (mvsw) { + int dir2 = dir; + if (mvsw) + dir2 = !dir; + for (i = 0; i < 2; i++) { + s->mv[dir][i+2][0] = s->mv[dir][i][0] = s->current_picture.motion_val[dir][s->block_index[i+2]][0] = s->current_picture.motion_val[dir][s->block_index[i]][0]; + s->mv[dir][i+2][1] = s->mv[dir][i][1] = s->current_picture.motion_val[dir][s->block_index[i+2]][1] = s->current_picture.motion_val[dir][s->block_index[i]][1]; + s->mv[dir2][i+2][0] = s->mv[dir2][i][0] = s->current_picture.motion_val[dir2][s->block_index[i]][0] = s->current_picture.motion_val[dir2][s->block_index[i+2]][0]; + s->mv[dir2][i+2][1] = s->mv[dir2][i][1] = s->current_picture.motion_val[dir2][s->block_index[i]][1] = s->current_picture.motion_val[dir2][s->block_index[i+2]][1]; + } + } else { + v->blk_mv_type[s->block_index[0]] = 1; + v->blk_mv_type[s->block_index[1]] = 1; + v->blk_mv_type[s->block_index[2]] = 1; + v->blk_mv_type[s->block_index[3]] = 1; + vc1_pred_mv_intfr(v, 0, 0, 0, 2, v->range_x, v->range_y, 0, !dir); + for (i = 0; i < 2; i++) { + s->mv[!dir][i+2][0] = s->mv[!dir][i][0] = s->current_picture.motion_val[!dir][s->block_index[i+2]][0] = s->current_picture.motion_val[!dir][s->block_index[i]][0]; + s->mv[!dir][i+2][1] = s->mv[!dir][i][1] = s->current_picture.motion_val[!dir][s->block_index[i+2]][1] = s->current_picture.motion_val[!dir][s->block_index[i]][1]; + } + } + } + } + + vc1_mc_1mv(v, dir); + if (direct || bmvtype == BMV_TYPE_INTERPOLATED) { + vc1_interp_mc(v); + } + } + } + if (s->mb_x == s->mb_width - 1) + memmove(v->is_intra_base, v->is_intra, sizeof(v->is_intra_base[0]) * s->mb_stride); + v->cbp[s->mb_x] = block_cbp; + v->ttblk[s->mb_x] = block_tt; + return 0; +} + /** Decode blocks of I-frame */ static void vc1_decode_i_blocks(VC1Context *v) @@ -4348,10 +4804,10 @@ static void vc1_decode_i_blocks(VC1Context *v) s->mb_x = s->mb_y = 0; s->mb_intra = 1; s->first_slice_line = 1; - for (s->mb_y = 0; s->mb_y < s->mb_height; s->mb_y++) { + for (s->mb_y = 0; s->mb_y < s->end_mb_y; s->mb_y++) { s->mb_x = 0; - ff_init_block_index(s); - for (; s->mb_x < s->mb_width; s->mb_x++) { + init_block_index(v); + for (; s->mb_x < v->end_mb_x; s->mb_x++) { uint8_t *dst[6]; ff_update_block_index(s); dst[0] = s->dest[0]; @@ -4360,12 +4816,12 @@ static void vc1_decode_i_blocks(VC1Context *v) dst[3] = dst[2] + 8; dst[4] = s->dest[1]; dst[5] = s->dest[2]; - s->dsp.clear_blocks(s->block[0]); + s->bdsp.clear_blocks(s->block[0]); mb_pos = s->mb_x + s->mb_y * s->mb_width; - s->current_picture.f.mb_type[mb_pos] = MB_TYPE_INTRA; - s->current_picture.f.qscale_table[mb_pos] = v->pq; - s->current_picture.f.motion_val[1][s->block_index[0]][0] = 0; - s->current_picture.f.motion_val[1][s->block_index[0]][1] = 0; + s->current_picture.mb_type[mb_pos] = MB_TYPE_INTRA; + s->current_picture.qscale_table[mb_pos] = v->pq; + s->current_picture.motion_val[1][s->block_index[0]][0] = 0; + s->current_picture.motion_val[1][s->block_index[0]][1] = 0; // do actual MB decoding and displaying cbp = get_vlc2(&v->s.gb, ff_msmp4_mb_i_vlc.table, MB_INTRA_VLC_BITS, 2); @@ -4390,12 +4846,16 @@ static void vc1_decode_i_blocks(VC1Context *v) if (v->rangeredfrm) for (j = 0; j < 64; j++) s->block[k][j] <<= 1; - s->dsp.put_signed_pixels_clamped(s->block[k], dst[k], k & 4 ? s->uvlinesize : s->linesize); + s->idsp.put_signed_pixels_clamped(s->block[k], dst[k], + k & 4 ? s->uvlinesize + : s->linesize); } else { if (v->rangeredfrm) for (j = 0; j < 64; j++) s->block[k][j] = (s->block[k][j] - 64) << 1; - s->dsp.put_pixels_clamped(s->block[k], dst[k], k & 4 ? s->uvlinesize : s->linesize); + s->idsp.put_pixels_clamped(s->block[k], dst[k], + k & 4 ? s->uvlinesize + : s->linesize); } } @@ -4424,22 +4884,25 @@ static void vc1_decode_i_blocks(VC1Context *v) if (v->s.loop_filter) vc1_loop_filter_iblk(v, v->pq); if (get_bits_count(&s->gb) > v->bits) { - ff_er_add_slice(s, 0, 0, s->mb_x, s->mb_y, ER_MB_ERROR); + ff_er_add_slice(&s->er, 0, 0, s->mb_x, s->mb_y, ER_MB_ERROR); av_log(s->avctx, AV_LOG_ERROR, "Bits overconsumption: %i > %i\n", get_bits_count(&s->gb), v->bits); return; } } if (!v->s.loop_filter) - ff_draw_horiz_band(s, s->mb_y * 16, 16); + ff_mpeg_draw_horiz_band(s, s->mb_y * 16, 16); else if (s->mb_y) - ff_draw_horiz_band(s, (s->mb_y - 1) * 16, 16); + ff_mpeg_draw_horiz_band(s, (s->mb_y - 1) * 16, 16); s->first_slice_line = 0; } if (v->s.loop_filter) - ff_draw_horiz_band(s, (s->mb_height - 1) * 16, 16); - ff_er_add_slice(s, 0, 0, s->mb_width - 1, s->mb_height - 1, ER_MB_END); + ff_mpeg_draw_horiz_band(s, (s->end_mb_y - 1) * 16, 16); + + /* This is intentionally mb_height and not end_mb_y - unlike in advanced + * profile, these only differ are when decoding MSS2 rectangles. */ + ff_er_add_slice(&s->er, 0, 0, s->mb_width - 1, s->mb_height - 1, ER_MB_END); } /** Decode blocks of I-frame for advanced profile @@ -4487,21 +4950,21 @@ static void vc1_decode_i_blocks_adv(VC1Context *v) s->mb_y = s->start_mb_y; if (s->start_mb_y) { s->mb_x = 0; - ff_init_block_index(s); + init_block_index(v); memset(&s->coded_block[s->block_index[0] - s->b8_stride], 0, (1 + s->b8_stride) * sizeof(*s->coded_block)); } for (; s->mb_y < s->end_mb_y; s->mb_y++) { s->mb_x = 0; - ff_init_block_index(s); + init_block_index(v); for (;s->mb_x < s->mb_width; s->mb_x++) { - DCTELEM (*block)[64] = v->block[v->cur_blk_idx]; + int16_t (*block)[64] = v->block[v->cur_blk_idx]; ff_update_block_index(s); - s->dsp.clear_blocks(block[0]); + s->bdsp.clear_blocks(block[0]); mb_pos = s->mb_x + s->mb_y * s->mb_stride; - s->current_picture.f.mb_type[mb_pos + v->mb_off] = MB_TYPE_INTRA; - s->current_picture.f.motion_val[1][s->block_index[0] + v->blocks_off][0] = 0; - s->current_picture.f.motion_val[1][s->block_index[0] + v->blocks_off][1] = 0; + s->current_picture.mb_type[mb_pos + v->mb_off] = MB_TYPE_INTRA; + 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; // do actual MB decoding and displaying if (v->fieldtx_is_raw) @@ -4517,7 +4980,7 @@ static void vc1_decode_i_blocks_adv(VC1Context *v) GET_MQUANT(); - s->current_picture.f.qscale_table[mb_pos] = mquant; + s->current_picture.qscale_table[mb_pos] = mquant; /* Set DC scale - y and c use the same */ s->y_dc_scale = s->y_dc_scale_table[mquant]; s->c_dc_scale = s->c_dc_scale_table[mquant]; @@ -4549,22 +5012,23 @@ static void vc1_decode_i_blocks_adv(VC1Context *v) if (get_bits_count(&s->gb) > v->bits) { // TODO: may need modification to handle slice coding - ff_er_add_slice(s, 0, s->start_mb_y, s->mb_x, s->mb_y, ER_MB_ERROR); + ff_er_add_slice(&s->er, 0, s->start_mb_y, s->mb_x, s->mb_y, ER_MB_ERROR); av_log(s->avctx, AV_LOG_ERROR, "Bits overconsumption: %i > %i\n", get_bits_count(&s->gb), v->bits); return; } } if (!v->s.loop_filter) - ff_draw_horiz_band(s, s->mb_y * 16, 16); + ff_mpeg_draw_horiz_band(s, s->mb_y * 16, 16); else if (s->mb_y) - ff_draw_horiz_band(s, (s->mb_y-1) * 16, 16); + ff_mpeg_draw_horiz_band(s, (s->mb_y-1) * 16, 16); s->first_slice_line = 0; } /* raw bottom MB row */ s->mb_x = 0; - ff_init_block_index(s); + init_block_index(v); + for (;s->mb_x < s->mb_width; s->mb_x++) { ff_update_block_index(s); vc1_put_signed_blocks_clamped(v); @@ -4572,8 +5036,8 @@ static void vc1_decode_i_blocks_adv(VC1Context *v) vc1_loop_filter_iblk_delayed(v, v->pq); } if (v->s.loop_filter) - ff_draw_horiz_band(s, (s->end_mb_y-1)*16, 16); - ff_er_add_slice(s, 0, s->start_mb_y << v->field_mode, s->mb_width - 1, + ff_mpeg_draw_horiz_band(s, (s->end_mb_y-1)*16, 16); + ff_er_add_slice(&s->er, 0, s->start_mb_y << v->field_mode, s->mb_width - 1, (s->end_mb_y << v->field_mode) - 1, ER_MB_END); } @@ -4607,12 +5071,13 @@ static void vc1_decode_p_blocks(VC1Context *v) break; } - apply_loop_filter = s->loop_filter && !(s->avctx->skip_loop_filter >= AVDISCARD_NONKEY); + apply_loop_filter = s->loop_filter && !(s->avctx->skip_loop_filter >= AVDISCARD_NONKEY) && + v->fcm == PROGRESSIVE; s->first_slice_line = 1; memset(v->cbp_base, 0, sizeof(v->cbp_base[0])*2*s->mb_stride); for (s->mb_y = s->start_mb_y; s->mb_y < s->end_mb_y; s->mb_y++) { s->mb_x = 0; - ff_init_block_index(s); + init_block_index(v); for (; s->mb_x < s->mb_width; s->mb_x++) { ff_update_block_index(s); @@ -4621,11 +5086,11 @@ static void vc1_decode_p_blocks(VC1Context *v) else if (v->fcm == ILACE_FRAME) vc1_decode_p_mb_intfr(v); else vc1_decode_p_mb(v); - if (s->mb_y != s->start_mb_y && apply_loop_filter && v->fcm == PROGRESSIVE) + if (s->mb_y != s->start_mb_y && apply_loop_filter) vc1_apply_p_loop_filter(v); if (get_bits_count(&s->gb) > v->bits || get_bits_count(&s->gb) < 0) { // TODO: may need modification to handle slice coding - ff_er_add_slice(s, 0, s->start_mb_y, s->mb_x, s->mb_y, ER_MB_ERROR); + ff_er_add_slice(&s->er, 0, s->start_mb_y, s->mb_x, s->mb_y, ER_MB_ERROR); av_log(s->avctx, AV_LOG_ERROR, "Bits overconsumption: %i > %i at %ix%i\n", get_bits_count(&s->gb), v->bits, s->mb_x, s->mb_y); return; @@ -4635,20 +5100,20 @@ static void vc1_decode_p_blocks(VC1Context *v) memmove(v->ttblk_base, v->ttblk, sizeof(v->ttblk_base[0]) * s->mb_stride); memmove(v->is_intra_base, v->is_intra, sizeof(v->is_intra_base[0]) * s->mb_stride); memmove(v->luma_mv_base, v->luma_mv, sizeof(v->luma_mv_base[0]) * s->mb_stride); - if (s->mb_y != s->start_mb_y) ff_draw_horiz_band(s, (s->mb_y - 1) * 16, 16); + if (s->mb_y != s->start_mb_y) ff_mpeg_draw_horiz_band(s, (s->mb_y - 1) * 16, 16); s->first_slice_line = 0; } if (apply_loop_filter) { s->mb_x = 0; - ff_init_block_index(s); + init_block_index(v); for (; s->mb_x < s->mb_width; s->mb_x++) { ff_update_block_index(s); vc1_apply_p_loop_filter(v); } } if (s->end_mb_y >= s->start_mb_y) - ff_draw_horiz_band(s, (s->end_mb_y - 1) * 16, 16); - ff_er_add_slice(s, 0, s->start_mb_y << v->field_mode, s->mb_width - 1, + ff_mpeg_draw_horiz_band(s, (s->end_mb_y - 1) * 16, 16); + ff_er_add_slice(&s->er, 0, s->start_mb_y << v->field_mode, s->mb_width - 1, (s->end_mb_y << v->field_mode) - 1, ER_MB_END); } @@ -4684,17 +5149,19 @@ static void vc1_decode_b_blocks(VC1Context *v) s->first_slice_line = 1; for (s->mb_y = s->start_mb_y; s->mb_y < s->end_mb_y; s->mb_y++) { s->mb_x = 0; - ff_init_block_index(s); + init_block_index(v); for (; s->mb_x < s->mb_width; s->mb_x++) { ff_update_block_index(s); if (v->fcm == ILACE_FIELD) vc1_decode_b_mb_intfi(v); + else if (v->fcm == ILACE_FRAME) + vc1_decode_b_mb_intfr(v); else vc1_decode_b_mb(v); if (get_bits_count(&s->gb) > v->bits || get_bits_count(&s->gb) < 0) { // TODO: may need modification to handle slice coding - ff_er_add_slice(s, 0, s->start_mb_y, s->mb_x, s->mb_y, ER_MB_ERROR); + ff_er_add_slice(&s->er, 0, s->start_mb_y, s->mb_x, s->mb_y, ER_MB_ERROR); av_log(s->avctx, AV_LOG_ERROR, "Bits overconsumption: %i > %i at %ix%i\n", get_bits_count(&s->gb), v->bits, s->mb_x, s->mb_y); return; @@ -4702,14 +5169,14 @@ static void vc1_decode_b_blocks(VC1Context *v) if (v->s.loop_filter) vc1_loop_filter_iblk(v, v->pq); } if (!v->s.loop_filter) - ff_draw_horiz_band(s, s->mb_y * 16, 16); + ff_mpeg_draw_horiz_band(s, s->mb_y * 16, 16); else if (s->mb_y) - ff_draw_horiz_band(s, (s->mb_y - 1) * 16, 16); + ff_mpeg_draw_horiz_band(s, (s->mb_y - 1) * 16, 16); s->first_slice_line = 0; } if (v->s.loop_filter) - ff_draw_horiz_band(s, (s->end_mb_y - 1) * 16, 16); - ff_er_add_slice(s, 0, s->start_mb_y << v->field_mode, s->mb_width - 1, + ff_mpeg_draw_horiz_band(s, (s->end_mb_y - 1) * 16, 16); + ff_er_add_slice(&s->er, 0, s->start_mb_y << v->field_mode, s->mb_width - 1, (s->end_mb_y << v->field_mode) - 1, ER_MB_END); } @@ -4717,22 +5184,25 @@ static void vc1_decode_skip_blocks(VC1Context *v) { MpegEncContext *s = &v->s; - ff_er_add_slice(s, 0, s->start_mb_y, s->mb_width - 1, s->end_mb_y - 1, ER_MB_END); + if (!v->s.last_picture.f->data[0]) + return; + + ff_er_add_slice(&s->er, 0, s->start_mb_y, s->mb_width - 1, s->end_mb_y - 1, ER_MB_END); s->first_slice_line = 1; for (s->mb_y = s->start_mb_y; s->mb_y < s->end_mb_y; s->mb_y++) { s->mb_x = 0; - ff_init_block_index(s); + init_block_index(v); ff_update_block_index(s); - memcpy(s->dest[0], s->last_picture.f.data[0] + s->mb_y * 16 * s->linesize, s->linesize * 16); - memcpy(s->dest[1], s->last_picture.f.data[1] + s->mb_y * 8 * s->uvlinesize, s->uvlinesize * 8); - memcpy(s->dest[2], s->last_picture.f.data[2] + s->mb_y * 8 * s->uvlinesize, s->uvlinesize * 8); - ff_draw_horiz_band(s, s->mb_y * 16, 16); + memcpy(s->dest[0], s->last_picture.f->data[0] + s->mb_y * 16 * s->linesize, s->linesize * 16); + memcpy(s->dest[1], s->last_picture.f->data[1] + s->mb_y * 8 * s->uvlinesize, s->uvlinesize * 8); + memcpy(s->dest[2], s->last_picture.f->data[2] + s->mb_y * 8 * s->uvlinesize, s->uvlinesize * 8); + ff_mpeg_draw_horiz_band(s, s->mb_y * 16, 16); s->first_slice_line = 0; } s->pict_type = AV_PICTURE_TYPE_P; } -static void vc1_decode_blocks(VC1Context *v) +void ff_vc1_decode_blocks(VC1Context *v) { v->s.esc3_level_length = 0; @@ -4837,7 +5307,7 @@ static void vc1_parse_sprites(VC1Context *v, GetBitContext* gb, SpriteData* sd) for (sprite = 0; sprite <= v->two_sprites; sprite++) { vc1_sprite_parse_transform(gb, sd->coefs[sprite]); if (sd->coefs[sprite][1] || sd->coefs[sprite][3]) - av_log_ask_for_sample(avctx, "Rotation coefficients are not zero"); + avpriv_request_sample(avctx, "Non-zero rotation coefficients"); av_log(avctx, AV_LOG_DEBUG, sprite ? "S2:" : "S1:"); for (i = 0; i < 7; i++) av_log(avctx, AV_LOG_DEBUG, " %d.%.3d", @@ -4920,19 +5390,19 @@ static void vc1_draw_sprites(VC1Context *v, SpriteData* sd) int width = v->output_width>>!!plane; for (row = 0; row < v->output_height>>!!plane; row++) { - uint8_t *dst = v->sprite_output_frame.data[plane] + - v->sprite_output_frame.linesize[plane] * row; + uint8_t *dst = v->sprite_output_frame->data[plane] + + v->sprite_output_frame->linesize[plane] * row; for (sprite = 0; sprite <= v->two_sprites; sprite++) { - uint8_t *iplane = s->current_picture.f.data[plane]; - int iline = s->current_picture.f.linesize[plane]; + uint8_t *iplane = s->current_picture.f->data[plane]; + int iline = s->current_picture.f->linesize[plane]; int ycoord = yoff[sprite] + yadv[sprite] * row; int yline = ycoord >> 16; int next_line; ysub[sprite] = ycoord & 0xFFFF; if (sprite) { - iplane = s->last_picture.f.data[plane]; - iline = s->last_picture.f.linesize[plane]; + iplane = s->last_picture.f->data[plane]; + iline = s->last_picture.f->linesize[plane]; } next_line = FFMIN(yline + 1, (v->sprite_height >> !!plane) - 1) * iline; if (!(xoff[sprite] & 0xFFFF) && xadv[sprite] == 1 << 16) { @@ -5001,22 +5471,18 @@ static int vc1_decode_sprites(VC1Context *v, GetBitContext* gb) vc1_parse_sprites(v, gb, &sd); - if (!s->current_picture.f.data[0]) { + if (!s->current_picture.f->data[0]) { av_log(avctx, AV_LOG_ERROR, "Got no sprites\n"); return -1; } - if (v->two_sprites && (!s->last_picture_ptr || !s->last_picture.f.data[0])) { + if (v->two_sprites && (!s->last_picture_ptr || !s->last_picture.f->data[0])) { av_log(avctx, AV_LOG_WARNING, "Need two sprites, only got one\n"); v->two_sprites = 0; } - if (v->sprite_output_frame.data[0]) - avctx->release_buffer(avctx, &v->sprite_output_frame); - - v->sprite_output_frame.buffer_hints = FF_BUFFER_HINTS_VALID; - v->sprite_output_frame.reference = 0; - if (avctx->get_buffer(avctx, &v->sprite_output_frame) < 0) { + av_frame_unref(v->sprite_output_frame); + if (ff_get_buffer(avctx, v->sprite_output_frame, 0) < 0) { av_log(avctx, AV_LOG_ERROR, "get_buffer() failed\n"); return -1; } @@ -5030,7 +5496,7 @@ static void vc1_sprite_flush(AVCodecContext *avctx) { VC1Context *v = avctx->priv_data; MpegEncContext *s = &v->s; - AVFrame *f = &s->current_picture.f; + AVFrame *f = s->current_picture.f; int plane, i; /* Windows Media Image codecs have a convergence interval of two keyframes. @@ -5046,18 +5512,19 @@ static void vc1_sprite_flush(AVCodecContext *avctx) #endif -static av_cold int vc1_decode_init_alloc_tables(VC1Context *v) +av_cold int ff_vc1_decode_init_alloc_tables(VC1Context *v) { MpegEncContext *s = &v->s; int i; + int mb_height = FFALIGN(s->mb_height, 2); /* Allocate mb bitplanes */ - v->mv_type_mb_plane = av_malloc (s->mb_stride * s->mb_height); - v->direct_mb_plane = av_malloc (s->mb_stride * s->mb_height); - v->forward_mb_plane = av_malloc (s->mb_stride * s->mb_height); - v->fieldtx_plane = av_mallocz(s->mb_stride * s->mb_height); - v->acpred_plane = av_malloc (s->mb_stride * s->mb_height); - v->over_flags_plane = av_malloc (s->mb_stride * s->mb_height); + v->mv_type_mb_plane = av_malloc (s->mb_stride * mb_height); + v->direct_mb_plane = av_malloc (s->mb_stride * mb_height); + v->forward_mb_plane = av_malloc (s->mb_stride * mb_height); + v->fieldtx_plane = av_mallocz(s->mb_stride * mb_height); + v->acpred_plane = av_malloc (s->mb_stride * mb_height); + v->over_flags_plane = av_malloc (s->mb_stride * mb_height); v->n_allocated_blks = s->mb_width + 2; v->block = av_malloc(sizeof(*v->block) * v->n_allocated_blks); @@ -5071,23 +5538,20 @@ static av_cold int vc1_decode_init_alloc_tables(VC1Context *v) v->luma_mv = v->luma_mv_base + s->mb_stride; /* allocate block type info in that way so it could be used with s->block_index[] */ - v->mb_type_base = av_malloc(s->b8_stride * (s->mb_height * 2 + 1) + s->mb_stride * (s->mb_height + 1) * 2); + v->mb_type_base = av_malloc(s->b8_stride * (mb_height * 2 + 1) + s->mb_stride * (mb_height + 1) * 2); v->mb_type[0] = v->mb_type_base + s->b8_stride + 1; - v->mb_type[1] = v->mb_type_base + s->b8_stride * (s->mb_height * 2 + 1) + s->mb_stride + 1; - v->mb_type[2] = v->mb_type[1] + s->mb_stride * (s->mb_height + 1); + v->mb_type[1] = v->mb_type_base + s->b8_stride * (mb_height * 2 + 1) + s->mb_stride + 1; + v->mb_type[2] = v->mb_type[1] + s->mb_stride * (mb_height + 1); /* allocate memory to store block level MV info */ - v->blk_mv_type_base = av_mallocz( s->b8_stride * (s->mb_height * 2 + 1) + s->mb_stride * (s->mb_height + 1) * 2); + v->blk_mv_type_base = av_mallocz( s->b8_stride * (mb_height * 2 + 1) + s->mb_stride * (mb_height + 1) * 2); v->blk_mv_type = v->blk_mv_type_base + s->b8_stride + 1; - v->mv_f_base = av_mallocz(2 * (s->b8_stride * (s->mb_height * 2 + 1) + s->mb_stride * (s->mb_height + 1) * 2)); + v->mv_f_base = av_mallocz(2 * (s->b8_stride * (mb_height * 2 + 1) + s->mb_stride * (mb_height + 1) * 2)); v->mv_f[0] = v->mv_f_base + s->b8_stride + 1; - v->mv_f[1] = v->mv_f[0] + (s->b8_stride * (s->mb_height * 2 + 1) + s->mb_stride * (s->mb_height + 1) * 2); - v->mv_f_last_base = av_mallocz(2 * (s->b8_stride * (s->mb_height * 2 + 1) + s->mb_stride * (s->mb_height + 1) * 2)); - v->mv_f_last[0] = v->mv_f_last_base + s->b8_stride + 1; - v->mv_f_last[1] = v->mv_f_last[0] + (s->b8_stride * (s->mb_height * 2 + 1) + s->mb_stride * (s->mb_height + 1) * 2); - v->mv_f_next_base = av_mallocz(2 * (s->b8_stride * (s->mb_height * 2 + 1) + s->mb_stride * (s->mb_height + 1) * 2)); + v->mv_f[1] = v->mv_f[0] + (s->b8_stride * (mb_height * 2 + 1) + s->mb_stride * (mb_height + 1) * 2); + v->mv_f_next_base = av_mallocz(2 * (s->b8_stride * (mb_height * 2 + 1) + s->mb_stride * (mb_height + 1) * 2)); v->mv_f_next[0] = v->mv_f_next_base + s->b8_stride + 1; - v->mv_f_next[1] = v->mv_f_next[0] + (s->b8_stride * (s->mb_height * 2 + 1) + s->mb_stride * (s->mb_height + 1) * 2); + v->mv_f_next[1] = v->mv_f_next[0] + (s->b8_stride * (mb_height * 2 + 1) + s->mb_stride * (mb_height + 1) * 2); /* Init coded blocks info */ if (v->profile == PROFILE_ADVANCED) { @@ -5106,12 +5570,38 @@ static av_cold int vc1_decode_init_alloc_tables(VC1Context *v) if (!v->mv_type_mb_plane || !v->direct_mb_plane || !v->acpred_plane || !v->over_flags_plane || !v->block || !v->cbp_base || !v->ttblk_base || !v->is_intra_base || !v->luma_mv_base || - !v->mb_type_base) - return -1; + !v->mb_type_base) { + av_freep(&v->mv_type_mb_plane); + av_freep(&v->direct_mb_plane); + av_freep(&v->acpred_plane); + av_freep(&v->over_flags_plane); + av_freep(&v->block); + av_freep(&v->cbp_base); + av_freep(&v->ttblk_base); + av_freep(&v->is_intra_base); + av_freep(&v->luma_mv_base); + av_freep(&v->mb_type_base); + return AVERROR(ENOMEM); + } return 0; } +av_cold void ff_vc1_init_transposed_scantables(VC1Context *v) +{ + int i; + for (i = 0; i < 64; i++) { +#define transpose(x) ((x >> 3) | ((x & 7) << 3)) + v->zz_8x8[0][i] = transpose(ff_wmv1_scantable[0][i]); + v->zz_8x8[1][i] = transpose(ff_wmv1_scantable[1][i]); + v->zz_8x8[2][i] = transpose(ff_wmv1_scantable[2][i]); + v->zz_8x8[3][i] = transpose(ff_wmv1_scantable[3][i]); + v->zzi_8x8[i] = transpose(ff_vc1_adv_interlaced_8x8_zz[i]); + } + v->left_blk_sh = 0; + v->top_blk_sh = 3; +} + /** Initialize a VC1/WMV3 decoder * @todo TODO: Handle VC-1 IDUs (Transport level?) * @todo TODO: Decypher remaining bits in extra_data @@ -5121,7 +5611,6 @@ static av_cold int vc1_decode_init(AVCodecContext *avctx) VC1Context *v = avctx->priv_data; MpegEncContext *s = &v->s; GetBitContext gb; - int i; /* save the container output size for WMImage */ v->output_width = avctx->width; @@ -5130,21 +5619,16 @@ static av_cold int vc1_decode_init(AVCodecContext *avctx) if (!avctx->extradata_size || !avctx->extradata) return -1; if (!(avctx->flags & CODEC_FLAG_GRAY)) - avctx->pix_fmt = avctx->get_format(avctx, avctx->codec->pix_fmts); + avctx->pix_fmt = ff_get_format(avctx, avctx->codec->pix_fmts); else - avctx->pix_fmt = PIX_FMT_GRAY8; - avctx->hwaccel = ff_find_hwaccel(avctx->codec->id, avctx->pix_fmt); + avctx->pix_fmt = AV_PIX_FMT_GRAY8; v->s.avctx = avctx; - avctx->flags |= CODEC_FLAG_EMU_EDGE; - v->s.flags |= CODEC_FLAG_EMU_EDGE; - - if (avctx->idct_algo == FF_IDCT_AUTO) { - avctx->idct_algo = FF_IDCT_WMV2; - } if (ff_vc1_init_common(v) < 0) return -1; - ff_vc1dsp_init(&v->vc1dsp); + ff_blockdsp_init(&s->bdsp, avctx); + ff_h264chroma_init(&v->h264chroma, 8); + ff_qpeldsp_init(&s->qdsp); if (avctx->codec_id == AV_CODEC_ID_WMV3 || avctx->codec_id == AV_CODEC_ID_WMV3IMAGE) { int count = 0; @@ -5211,9 +5695,13 @@ static av_cold int vc1_decode_init(AVCodecContext *avctx) av_log(avctx, AV_LOG_ERROR, "Incomplete extradata\n"); return -1; } - v->res_sprite = (avctx->codec_tag == MKTAG('W','V','P','2')); + v->res_sprite = (avctx->codec_id == AV_CODEC_ID_VC1IMAGE); } + v->sprite_output_frame = av_frame_alloc(); + if (!v->sprite_output_frame) + return AVERROR(ENOMEM); + avctx->profile = v->profile; if (v->profile == PROFILE_ADVANCED) avctx->level = v->level; @@ -5224,16 +5712,7 @@ static av_cold int vc1_decode_init(AVCodecContext *avctx) s->mb_height = (avctx->coded_height + 15) >> 4; if (v->profile == PROFILE_ADVANCED || v->res_fasttx) { - for (i = 0; i < 64; i++) { -#define transpose(x) ((x >> 3) | ((x & 7) << 3)) - v->zz_8x8[0][i] = transpose(ff_wmv1_scantable[0][i]); - v->zz_8x8[1][i] = transpose(ff_wmv1_scantable[1][i]); - v->zz_8x8[2][i] = transpose(ff_wmv1_scantable[2][i]); - v->zz_8x8[3][i] = transpose(ff_wmv1_scantable[3][i]); - v->zzi_8x8[i] = transpose(ff_vc1_adv_interlaced_8x8_zz[i]); - } - v->left_blk_sh = 0; - v->top_blk_sh = 3; + ff_vc1_init_transposed_scantables(v); } else { memcpy(v->zz_8x8, ff_wmv1_scantable, 4*64); v->left_blk_sh = 3; @@ -5259,14 +5738,13 @@ static av_cold int vc1_decode_init(AVCodecContext *avctx) /** Close a VC1/WMV3 decoder * @warning Initial try at using MpegEncContext stuff */ -static av_cold int vc1_decode_end(AVCodecContext *avctx) +av_cold int ff_vc1_decode_end(AVCodecContext *avctx) { VC1Context *v = avctx->priv_data; int i; - if ((avctx->codec_id == AV_CODEC_ID_WMV3IMAGE || avctx->codec_id == AV_CODEC_ID_VC1IMAGE) - && v->sprite_output_frame.data[0]) - avctx->release_buffer(avctx, &v->sprite_output_frame); + av_frame_free(&v->sprite_output_frame); + for (i = 0; i < 4; i++) av_freep(&v->sr_rows[i >> 1][i & 1]); av_freep(&v->hrd_rate); @@ -5281,7 +5759,6 @@ static av_cold int vc1_decode_end(AVCodecContext *avctx) av_freep(&v->mb_type_base); av_freep(&v->blk_mv_type_base); av_freep(&v->mv_f_base); - av_freep(&v->mv_f_last_base); av_freep(&v->mv_f_next_base); av_freep(&v->block); av_freep(&v->cbp_base); @@ -5297,10 +5774,10 @@ static av_cold int vc1_decode_end(AVCodecContext *avctx) * @todo TODO: Handle VC-1 IDUs (Transport level?) */ static int vc1_decode_frame(AVCodecContext *avctx, void *data, - int *data_size, AVPacket *avpkt) + int *got_frame, AVPacket *avpkt) { const uint8_t *buf = avpkt->data; - int buf_size = avpkt->size, n_slices = 0, i; + int buf_size = avpkt->size, n_slices = 0, i, ret; VC1Context *v = avctx->priv_data; MpegEncContext *s = &v->s; AVFrame *pict = data; @@ -5317,22 +5794,16 @@ static int vc1_decode_frame(AVCodecContext *avctx, void *data, if (buf_size == 0 || (buf_size == 4 && AV_RB32(buf) == VC1_CODE_ENDOFSEQ)) { /* special case for last picture */ if (s->low_delay == 0 && s->next_picture_ptr) { - *pict = s->next_picture_ptr->f; + if ((ret = av_frame_ref(pict, s->next_picture_ptr->f)) < 0) + return ret; s->next_picture_ptr = NULL; - *data_size = sizeof(AVFrame); + *got_frame = 1; } return 0; } - if (s->avctx->codec->capabilities&CODEC_CAP_HWACCEL_VDPAU) { - if (v->profile < PROFILE_ADVANCED) - avctx->pix_fmt = PIX_FMT_VDPAU_WMV3; - else - avctx->pix_fmt = PIX_FMT_VDPAU_VC1; - } - //for advanced profile we may need to parse and unescape data if (avctx->codec_id == AV_CODEC_ID_VC1 || avctx->codec_id == AV_CODEC_ID_VC1IMAGE) { int buf_size2 = 0; @@ -5349,16 +5820,16 @@ static int vc1_decode_frame(AVCodecContext *avctx, void *data, if (size <= 0) continue; switch (AV_RB32(start)) { case VC1_CODE_FRAME: - if (avctx->hwaccel || - s->avctx->codec->capabilities&CODEC_CAP_HWACCEL_VDPAU) + if (avctx->hwaccel) buf_start = start; buf_size2 = vc1_unescape_buffer(start + 4, size, buf2); break; case VC1_CODE_FIELD: { int buf_size3; - slices = av_realloc(slices, sizeof(*slices) * (n_slices+1)); - if (!slices) + tmp = av_realloc(slices, sizeof(*slices) * (n_slices+1)); + if (!tmp) goto err; + slices = tmp; slices[n_slices].buf = av_mallocz(buf_size + FF_INPUT_BUFFER_PADDING_SIZE); if (!slices[n_slices].buf) goto err; @@ -5380,9 +5851,10 @@ static int vc1_decode_frame(AVCodecContext *avctx, void *data, break; case VC1_CODE_SLICE: { int buf_size3; - slices = av_realloc(slices, sizeof(*slices) * (n_slices+1)); - if (!slices) + tmp = av_realloc(slices, sizeof(*slices) * (n_slices+1)); + if (!tmp) goto err; + slices = tmp; slices[n_slices].buf = av_mallocz(buf_size + FF_INPUT_BUFFER_PADDING_SIZE); if (!slices[n_slices].buf) goto err; @@ -5448,12 +5920,16 @@ static int vc1_decode_frame(AVCodecContext *avctx, void *data, if (s->context_initialized && (s->width != avctx->coded_width || s->height != avctx->coded_height)) { - vc1_decode_end(avctx); + ff_vc1_decode_end(avctx); } if (!s->context_initialized) { - if (ff_msmpeg4_decode_init(avctx) < 0 || vc1_decode_init_alloc_tables(v) < 0) - return -1; + if (ff_msmpeg4_decode_init(avctx) < 0) + goto err; + if (ff_vc1_decode_init_alloc_tables(v) < 0) { + ff_MPV_common_end(s); + goto err; + } s->low_delay = !avctx->has_b_frames || v->res_sprite; @@ -5463,26 +5939,19 @@ static int vc1_decode_frame(AVCodecContext *avctx, void *data, } } - /* We need to set current_picture_ptr before reading the header, - * otherwise we cannot store anything in there. */ - if (s->current_picture_ptr == NULL || s->current_picture_ptr->f.data[0]) { - int i = ff_find_unused_picture(s, 0); - if (i < 0) - goto err; - s->current_picture_ptr = &s->picture[i]; - } - // do parse frame header v->pic_header_flag = 0; + v->first_pic_header_flag = 1; if (v->profile < PROFILE_ADVANCED) { - if (ff_vc1_parse_frame_header(v, &s->gb) == -1) { + if (ff_vc1_parse_frame_header(v, &s->gb) < 0) { goto err; } } else { - if (ff_vc1_parse_frame_header_adv(v, &s->gb) == -1) { + if (ff_vc1_parse_frame_header_adv(v, &s->gb) < 0) { goto err; } } + v->first_pic_header_flag = 0; if ((avctx->codec_id == AV_CODEC_ID_WMV3IMAGE || avctx->codec_id == AV_CODEC_ID_VC1IMAGE) && s->pict_type != AV_PICTURE_TYPE_I) { @@ -5490,25 +5959,13 @@ static int vc1_decode_frame(AVCodecContext *avctx, void *data, goto err; } - // process pulldown flags - s->current_picture_ptr->f.repeat_pict = 0; - // Pulldown flags are only valid when 'broadcast' has been set. - // So ticks_per_frame will be 2 - if (v->rff) { - // repeat field - s->current_picture_ptr->f.repeat_pict = 1; - } else if (v->rptfrm) { - // repeat frames - s->current_picture_ptr->f.repeat_pict = v->rptfrm * 2; - } - // for skipping the frame - s->current_picture.f.pict_type = s->pict_type; - s->current_picture.f.key_frame = s->pict_type == AV_PICTURE_TYPE_I; + s->current_picture.f->pict_type = s->pict_type; + s->current_picture.f->key_frame = s->pict_type == AV_PICTURE_TYPE_I; /* skip B-frames if we don't have reference frames */ - if (s->last_picture_ptr == NULL && (s->pict_type == AV_PICTURE_TYPE_B || s->dropable)) { - goto err; + if (s->last_picture_ptr == NULL && (s->pict_type == AV_PICTURE_TYPE_B || s->droppable)) { + goto end; } if ((avctx->skip_frame >= AVDISCARD_NONREF && s->pict_type == AV_PICTURE_TYPE_B) || (avctx->skip_frame >= AVDISCARD_NONKEY && s->pict_type != AV_PICTURE_TYPE_I) || @@ -5527,13 +5984,22 @@ static int vc1_decode_frame(AVCodecContext *avctx, void *data, goto err; } - s->me.qpel_put = s->dsp.put_qpel_pixels_tab; - s->me.qpel_avg = s->dsp.avg_qpel_pixels_tab; + // process pulldown flags + s->current_picture_ptr->f->repeat_pict = 0; + // Pulldown flags are only valid when 'broadcast' has been set. + // So ticks_per_frame will be 2 + if (v->rff) { + // repeat field + s->current_picture_ptr->f->repeat_pict = 1; + } else if (v->rptfrm) { + // repeat frames + s->current_picture_ptr->f->repeat_pict = v->rptfrm * 2; + } + + s->me.qpel_put = s->qdsp.put_qpel_pixels_tab; + s->me.qpel_avg = s->qdsp.avg_qpel_pixels_tab; - if ((CONFIG_VC1_VDPAU_DECODER) - &&s->avctx->codec->capabilities&CODEC_CAP_HWACCEL_VDPAU) - ff_vdpau_vc1_decode_picture(s, buf_start, (buf + buf_size) - buf_start); - else if (avctx->hwaccel) { + if (avctx->hwaccel) { if (avctx->hwaccel->start_frame(avctx, buf, buf_size) < 0) goto err; if (avctx->hwaccel->decode_slice(avctx, buf_start, (buf + buf_size) - buf_start) < 0) @@ -5541,26 +6007,26 @@ static int vc1_decode_frame(AVCodecContext *avctx, void *data, if (avctx->hwaccel->end_frame(avctx) < 0) goto err; } else { - ff_er_frame_start(s); + int header_ret = 0; + + ff_mpeg_er_frame_start(s); v->bits = buf_size * 8; + v->end_mb_x = s->mb_width; if (v->field_mode) { - uint8_t *tmp[2]; - s->current_picture.f.linesize[0] <<= 1; - s->current_picture.f.linesize[1] <<= 1; - s->current_picture.f.linesize[2] <<= 1; + s->current_picture.f->linesize[0] <<= 1; + s->current_picture.f->linesize[1] <<= 1; + s->current_picture.f->linesize[2] <<= 1; s->linesize <<= 1; s->uvlinesize <<= 1; - tmp[0] = v->mv_f_last[0]; - tmp[1] = v->mv_f_last[1]; - v->mv_f_last[0] = v->mv_f_next[0]; - v->mv_f_last[1] = v->mv_f_next[1]; - v->mv_f_next[0] = v->mv_f[0]; - v->mv_f_next[1] = v->mv_f[1]; - v->mv_f[0] = tmp[0]; - v->mv_f[1] = tmp[1]; } mb_height = s->mb_height >> v->field_mode; + + if (!mb_height) { + av_log(v->s.avctx, AV_LOG_ERROR, "Invalid mb_height.\n"); + goto err; + } + for (i = 0; i <= n_slices; i++) { if (i > 0 && slices[i - 1].mby_start >= mb_height) { if (v->field_mode <= 0) { @@ -5580,43 +6046,51 @@ static int vc1_decode_frame(AVCodecContext *avctx, void *data, if (i) { v->pic_header_flag = 0; if (v->field_mode && i == n_slices1 + 2) { - if (ff_vc1_parse_frame_header_adv(v, &s->gb) < 0) { + if ((header_ret = ff_vc1_parse_frame_header_adv(v, &s->gb)) < 0) { av_log(v->s.avctx, AV_LOG_ERROR, "Field header damaged\n"); + if (avctx->err_recognition & AV_EF_EXPLODE) + goto err; continue; } } else if (get_bits1(&s->gb)) { v->pic_header_flag = 1; - if (ff_vc1_parse_frame_header_adv(v, &s->gb) < 0) { + if ((header_ret = ff_vc1_parse_frame_header_adv(v, &s->gb)) < 0) { av_log(v->s.avctx, AV_LOG_ERROR, "Slice header damaged\n"); + if (avctx->err_recognition & AV_EF_EXPLODE) + goto err; continue; } } } + if (header_ret < 0) + continue; s->start_mb_y = (i == 0) ? 0 : FFMAX(0, slices[i-1].mby_start % mb_height); if (!v->field_mode || v->second_field) s->end_mb_y = (i == n_slices ) ? mb_height : FFMIN(mb_height, slices[i].mby_start % mb_height); else s->end_mb_y = (i <= n_slices1 + 1) ? mb_height : FFMIN(mb_height, slices[i].mby_start % mb_height); - vc1_decode_blocks(v); + ff_vc1_decode_blocks(v); if (i != n_slices) s->gb = slices[i].gb; } if (v->field_mode) { v->second_field = 0; - if (s->pict_type == AV_PICTURE_TYPE_B) { - memcpy(v->mv_f_base, v->mv_f_next_base, - 2 * (s->b8_stride * (s->mb_height * 2 + 1) + s->mb_stride * (s->mb_height + 1) * 2)); - } - s->current_picture.f.linesize[0] >>= 1; - s->current_picture.f.linesize[1] >>= 1; - s->current_picture.f.linesize[2] >>= 1; + s->current_picture.f->linesize[0] >>= 1; + s->current_picture.f->linesize[1] >>= 1; + s->current_picture.f->linesize[2] >>= 1; s->linesize >>= 1; s->uvlinesize >>= 1; + if (v->s.pict_type != AV_PICTURE_TYPE_BI && v->s.pict_type != AV_PICTURE_TYPE_B) { + FFSWAP(uint8_t *, v->mv_f_next[0], v->mv_f[0]); + FFSWAP(uint8_t *, v->mv_f_next[1], v->mv_f[1]); + } } -//av_log(s->avctx, AV_LOG_INFO, "Consumed %i/%i bits\n", get_bits_count(&s->gb), s->gb.size_in_bits); + av_dlog(s->avctx, "Consumed %i/%i bits\n", + get_bits_count(&s->gb), s->gb.size_in_bits); // if (get_bits_count(&s->gb) > buf_size * 8) // return -1; - ff_er_frame_end(s); + if (!v->field_mode) + ff_er_frame_end(&s->er); } ff_MPV_frame_end(s); @@ -5631,17 +6105,20 @@ image: if (vc1_decode_sprites(v, &s->gb)) goto err; #endif - *pict = v->sprite_output_frame; - *data_size = sizeof(AVFrame); + if ((ret = av_frame_ref(pict, v->sprite_output_frame)) < 0) + goto err; + *got_frame = 1; } else { if (s->pict_type == AV_PICTURE_TYPE_B || s->low_delay) { - *pict = s->current_picture_ptr->f; + if ((ret = av_frame_ref(pict, s->current_picture_ptr->f)) < 0) + goto err; + ff_print_debug_info(s, s->current_picture_ptr); + *got_frame = 1; } else if (s->last_picture_ptr != NULL) { - *pict = s->last_picture_ptr->f; - } - if (s->last_picture_ptr || s->low_delay) { - *data_size = sizeof(AVFrame); - ff_print_debug_info(s, pict); + if ((ret = av_frame_ref(pict, s->last_picture_ptr->f)) < 0) + goto err; + ff_print_debug_info(s, s->last_picture_ptr); + *got_frame = 1; } } @@ -5669,64 +6146,48 @@ static const AVProfile profiles[] = { { FF_PROFILE_UNKNOWN }, }; +static const enum AVPixelFormat vc1_hwaccel_pixfmt_list_420[] = { +#if CONFIG_VC1_DXVA2_HWACCEL + AV_PIX_FMT_DXVA2_VLD, +#endif +#if CONFIG_VC1_VAAPI_HWACCEL + AV_PIX_FMT_VAAPI_VLD, +#endif +#if CONFIG_VC1_VDPAU_HWACCEL + AV_PIX_FMT_VDPAU, +#endif + AV_PIX_FMT_YUV420P, + AV_PIX_FMT_NONE +}; + AVCodec ff_vc1_decoder = { .name = "vc1", + .long_name = NULL_IF_CONFIG_SMALL("SMPTE VC-1"), .type = AVMEDIA_TYPE_VIDEO, .id = AV_CODEC_ID_VC1, .priv_data_size = sizeof(VC1Context), .init = vc1_decode_init, - .close = vc1_decode_end, + .close = ff_vc1_decode_end, .decode = vc1_decode_frame, + .flush = ff_mpeg_flush, .capabilities = CODEC_CAP_DR1 | CODEC_CAP_DELAY, - .long_name = NULL_IF_CONFIG_SMALL("SMPTE VC-1"), - .pix_fmts = ff_hwaccel_pixfmt_list_420, + .pix_fmts = vc1_hwaccel_pixfmt_list_420, .profiles = NULL_IF_CONFIG_SMALL(profiles) }; #if CONFIG_WMV3_DECODER AVCodec ff_wmv3_decoder = { .name = "wmv3", - .type = AVMEDIA_TYPE_VIDEO, - .id = AV_CODEC_ID_WMV3, - .priv_data_size = sizeof(VC1Context), - .init = vc1_decode_init, - .close = vc1_decode_end, - .decode = vc1_decode_frame, - .capabilities = CODEC_CAP_DR1 | CODEC_CAP_DELAY, .long_name = NULL_IF_CONFIG_SMALL("Windows Media Video 9"), - .pix_fmts = ff_hwaccel_pixfmt_list_420, - .profiles = NULL_IF_CONFIG_SMALL(profiles) -}; -#endif - -#if CONFIG_WMV3_VDPAU_DECODER -AVCodec ff_wmv3_vdpau_decoder = { - .name = "wmv3_vdpau", .type = AVMEDIA_TYPE_VIDEO, .id = AV_CODEC_ID_WMV3, .priv_data_size = sizeof(VC1Context), .init = vc1_decode_init, - .close = vc1_decode_end, - .decode = vc1_decode_frame, - .capabilities = CODEC_CAP_DR1 | CODEC_CAP_DELAY | CODEC_CAP_HWACCEL_VDPAU, - .long_name = NULL_IF_CONFIG_SMALL("Windows Media Video 9 VDPAU"), - .pix_fmts = (const enum PixelFormat[]){ PIX_FMT_VDPAU_WMV3, PIX_FMT_NONE }, - .profiles = NULL_IF_CONFIG_SMALL(profiles) -}; -#endif - -#if CONFIG_VC1_VDPAU_DECODER -AVCodec ff_vc1_vdpau_decoder = { - .name = "vc1_vdpau", - .type = AVMEDIA_TYPE_VIDEO, - .id = AV_CODEC_ID_VC1, - .priv_data_size = sizeof(VC1Context), - .init = vc1_decode_init, - .close = vc1_decode_end, + .close = ff_vc1_decode_end, .decode = vc1_decode_frame, - .capabilities = CODEC_CAP_DR1 | CODEC_CAP_DELAY | CODEC_CAP_HWACCEL_VDPAU, - .long_name = NULL_IF_CONFIG_SMALL("SMPTE VC-1 VDPAU"), - .pix_fmts = (const enum PixelFormat[]){ PIX_FMT_VDPAU_VC1, PIX_FMT_NONE }, + .flush = ff_mpeg_flush, + .capabilities = CODEC_CAP_DR1 | CODEC_CAP_DELAY, + .pix_fmts = vc1_hwaccel_pixfmt_list_420, .profiles = NULL_IF_CONFIG_SMALL(profiles) }; #endif @@ -5734,31 +6195,37 @@ AVCodec ff_vc1_vdpau_decoder = { #if CONFIG_WMV3IMAGE_DECODER AVCodec ff_wmv3image_decoder = { .name = "wmv3image", + .long_name = NULL_IF_CONFIG_SMALL("Windows Media Video 9 Image"), .type = AVMEDIA_TYPE_VIDEO, .id = AV_CODEC_ID_WMV3IMAGE, .priv_data_size = sizeof(VC1Context), .init = vc1_decode_init, - .close = vc1_decode_end, + .close = ff_vc1_decode_end, .decode = vc1_decode_frame, .capabilities = CODEC_CAP_DR1, .flush = vc1_sprite_flush, - .long_name = NULL_IF_CONFIG_SMALL("Windows Media Video 9 Image"), - .pix_fmts = ff_pixfmt_list_420 + .pix_fmts = (const enum AVPixelFormat[]) { + AV_PIX_FMT_YUV420P, + AV_PIX_FMT_NONE + }, }; #endif #if CONFIG_VC1IMAGE_DECODER AVCodec ff_vc1image_decoder = { .name = "vc1image", + .long_name = NULL_IF_CONFIG_SMALL("Windows Media Video 9 Image v2"), .type = AVMEDIA_TYPE_VIDEO, .id = AV_CODEC_ID_VC1IMAGE, .priv_data_size = sizeof(VC1Context), .init = vc1_decode_init, - .close = vc1_decode_end, + .close = ff_vc1_decode_end, .decode = vc1_decode_frame, .capabilities = CODEC_CAP_DR1, .flush = vc1_sprite_flush, - .long_name = NULL_IF_CONFIG_SMALL("Windows Media Video 9 Image v2"), - .pix_fmts = ff_pixfmt_list_420 + .pix_fmts = (const enum AVPixelFormat[]) { + AV_PIX_FMT_YUV420P, + AV_PIX_FMT_NONE + }, }; #endif