From 3ada94ba287e345cc50ce8ba34ea412a52a1ac67 Mon Sep 17 00:00:00 2001 From: Benoit Fouet Date: Tue, 3 Jul 2007 09:55:56 +0000 Subject: [PATCH] Description: split mpegvideo.c Originally committed as revision 9467 to svn://svn.ffmpeg.org/ffmpeg/trunk --- libavcodec/Makefile | 1 + libavcodec/mpegvideo.c | 4576 +-------------------------------- libavcodec/mpegvideo_common.h | 836 ++++++ libavcodec/mpegvideo_enc.c | 3796 +++++++++++++++++++++++++++ 4 files changed, 4637 insertions(+), 4572 deletions(-) create mode 100644 libavcodec/mpegvideo_common.h create mode 100644 libavcodec/mpegvideo_enc.c diff --git a/libavcodec/Makefile b/libavcodec/Makefile index b9200f9476..19e5925be3 100644 --- a/libavcodec/Makefile +++ b/libavcodec/Makefile @@ -67,6 +67,7 @@ OBJS-$(CONFIG_DVVIDEO_DECODER) += dv.o OBJS-$(CONFIG_DVVIDEO_ENCODER) += dv.o OBJS-$(CONFIG_DXA_DECODER) += dxa.o OBJS-$(CONFIG_EIGHTBPS_DECODER) += 8bps.o +OBJS-$(CONFIG_ENCODERS) += mpegvideo_enc.o OBJS-$(CONFIG_FFV1_DECODER) += ffv1.o golomb.o OBJS-$(CONFIG_FFV1_ENCODER) += ffv1.o OBJS-$(CONFIG_FFVHUFF_DECODER) += huffyuv.o diff --git a/libavcodec/mpegvideo.c b/libavcodec/mpegvideo.c index 9233ff3258..c71a96a342 100644 --- a/libavcodec/mpegvideo.c +++ b/libavcodec/mpegvideo.c @@ -30,6 +30,7 @@ #include "avcodec.h" #include "dsputil.h" #include "mpegvideo.h" +#include "mpegvideo_common.h" #include "mjpegenc.h" #include "msmpeg4.h" #include "faandct.h" @@ -38,9 +39,6 @@ //#undef NDEBUG //#include -#ifdef CONFIG_ENCODERS -static int encode_picture(MpegEncContext *s, int picture_number); -#endif //CONFIG_ENCODERS static void dct_unquantize_mpeg1_intra_c(MpegEncContext *s, DCTELEM *block, int n, int qscale); static void dct_unquantize_mpeg1_inter_c(MpegEncContext *s, @@ -56,13 +54,6 @@ static void dct_unquantize_h263_intra_c(MpegEncContext *s, static void dct_unquantize_h263_inter_c(MpegEncContext *s, DCTELEM *block, int n, int qscale); static void draw_edges_c(uint8_t *buf, int wrap, int width, int height, int w); -#ifdef CONFIG_ENCODERS -static int dct_quantize_c(MpegEncContext *s, DCTELEM *block, int n, int qscale, int *overflow); -static int dct_quantize_trellis_c(MpegEncContext *s, DCTELEM *block, int n, int qscale, int *overflow); -static int dct_quantize_refine(MpegEncContext *s, DCTELEM *block, int16_t *weight, DCTELEM *orig, int n, int qscale); -static int sse_mb(MpegEncContext *s); -static void denoise_dct_c(MpegEncContext *s, DCTELEM *block); -#endif //CONFIG_ENCODERS #ifdef HAVE_XVMC extern int XVMC_field_start(MpegEncContext*s, AVCodecContext *avctx); @@ -79,20 +70,6 @@ void (*draw_edges)(uint8_t *buf, int wrap, int width, int height, int w)= draw_e //#define DEBUG -#ifdef CONFIG_ENCODERS -static const uint16_t aanscales[64] = { - /* precomputed values scaled up by 14 bits */ - 16384, 22725, 21407, 19266, 16384, 12873, 8867, 4520, - 22725, 31521, 29692, 26722, 22725, 17855, 12299, 6270, - 21407, 29692, 27969, 25172, 21407, 16819, 11585, 5906, - 19266, 26722, 25172, 22654, 19266, 15137, 10426, 5315, - 16384, 22725, 21407, 19266, 16384, 12873, 8867, 4520, - 12873, 17855, 16819, 15137, 12873, 10114, 6967, 3552, - 8867 , 12299, 11585, 10426, 8867, 6967, 4799, 2446, - 4520 , 6270, 5906, 5315, 4520, 3552, 2446, 1247 -}; -#endif //CONFIG_ENCODERS - static const uint8_t h263_chroma_roundtab[16] = { // 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 2, 2, @@ -103,94 +80,6 @@ static const uint8_t ff_default_chroma_qscale_table[32]={ 0, 1, 2, 3, 4, 5, 6, 7, 8, 9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31 }; -#ifdef CONFIG_ENCODERS -static uint8_t default_mv_penalty[MAX_FCODE+1][MAX_MV*2+1]; -static uint8_t default_fcode_tab[MAX_MV*2+1]; - -enum PixelFormat ff_yuv420p_list[2]= {PIX_FMT_YUV420P, -1}; - -static void convert_matrix(DSPContext *dsp, int (*qmat)[64], uint16_t (*qmat16)[2][64], - const uint16_t *quant_matrix, int bias, int qmin, int qmax, int intra) -{ - int qscale; - int shift=0; - - for(qscale=qmin; qscale<=qmax; qscale++){ - int i; - if (dsp->fdct == ff_jpeg_fdct_islow -#ifdef FAAN_POSTSCALE - || dsp->fdct == ff_faandct -#endif - ) { - for(i=0;i<64;i++) { - const int j= dsp->idct_permutation[i]; - /* 16 <= qscale * quant_matrix[i] <= 7905 */ - /* 19952 <= aanscales[i] * qscale * quant_matrix[i] <= 249205026 */ - /* (1<<36)/19952 >= (1<<36)/(aanscales[i] * qscale * quant_matrix[i]) >= (1<<36)/249205026 */ - /* 3444240 >= (1<<36)/(aanscales[i] * qscale * quant_matrix[i]) >= 275 */ - - qmat[qscale][i] = (int)((UINT64_C(1) << QMAT_SHIFT) / - (qscale * quant_matrix[j])); - } - } else if (dsp->fdct == fdct_ifast -#ifndef FAAN_POSTSCALE - || dsp->fdct == ff_faandct -#endif - ) { - for(i=0;i<64;i++) { - const int j= dsp->idct_permutation[i]; - /* 16 <= qscale * quant_matrix[i] <= 7905 */ - /* 19952 <= aanscales[i] * qscale * quant_matrix[i] <= 249205026 */ - /* (1<<36)/19952 >= (1<<36)/(aanscales[i] * qscale * quant_matrix[i]) >= (1<<36)/249205026 */ - /* 3444240 >= (1<<36)/(aanscales[i] * qscale * quant_matrix[i]) >= 275 */ - - qmat[qscale][i] = (int)((UINT64_C(1) << (QMAT_SHIFT + 14)) / - (aanscales[i] * qscale * quant_matrix[j])); - } - } else { - for(i=0;i<64;i++) { - const int j= dsp->idct_permutation[i]; - /* We can safely suppose that 16 <= quant_matrix[i] <= 255 - So 16 <= qscale * quant_matrix[i] <= 7905 - so (1<<19) / 16 >= (1<<19) / (qscale * quant_matrix[i]) >= (1<<19) / 7905 - so 32768 >= (1<<19) / (qscale * quant_matrix[i]) >= 67 - */ - qmat[qscale][i] = (int)((UINT64_C(1) << QMAT_SHIFT) / (qscale * quant_matrix[j])); -// qmat [qscale][i] = (1 << QMAT_SHIFT_MMX) / (qscale * quant_matrix[i]); - qmat16[qscale][0][i] = (1 << QMAT_SHIFT_MMX) / (qscale * quant_matrix[j]); - - if(qmat16[qscale][0][i]==0 || qmat16[qscale][0][i]==128*256) qmat16[qscale][0][i]=128*256-1; - qmat16[qscale][1][i]= ROUNDED_DIV(bias<<(16-QUANT_BIAS_SHIFT), qmat16[qscale][0][i]); - } - } - - for(i=intra; i<64; i++){ - int64_t max= 8191; - if (dsp->fdct == fdct_ifast -#ifndef FAAN_POSTSCALE - || dsp->fdct == ff_faandct -#endif - ) { - max= (8191LL*aanscales[i]) >> 14; - } - while(((max * qmat[qscale][i]) >> shift) > INT_MAX){ - shift++; - } - } - } - if(shift){ - av_log(NULL, AV_LOG_INFO, "Warning, QMAT_SHIFT is larger than %d, overflows possible\n", QMAT_SHIFT - shift); - } -} - -static inline void update_qscale(MpegEncContext *s){ - s->qscale= (s->lambda*139 + FF_LAMBDA_SCALE*64) >> (FF_LAMBDA_SHIFT + 7); - s->qscale= av_clip(s->qscale, s->avctx->qmin, s->avctx->qmax); - - s->lambda2= (s->lambda*s->lambda + FF_LAMBDA_SCALE/2) >> FF_LAMBDA_SHIFT; -} -#endif //CONFIG_ENCODERS - void ff_init_scantable(uint8_t *permutation, ScanTable *st, const uint8_t *src_scantable){ int i; int end; @@ -215,20 +104,6 @@ void ff_init_scantable(uint8_t *permutation, ScanTable *st, const uint8_t *src_s } } -#ifdef CONFIG_ENCODERS -void ff_write_quant_matrix(PutBitContext *pb, uint16_t *matrix){ - int i; - - if(matrix){ - put_bits(pb, 1, 1); - for(i=0;i<64;i++) { - put_bits(pb, 8, matrix[ ff_zigzag_direct[i] ]); - } - }else - put_bits(pb, 1, 0); -} -#endif //CONFIG_ENCODERS - const uint8_t *ff_find_start_code(const uint8_t * restrict p, const uint8_t *end, uint32_t * restrict state){ int i; @@ -316,57 +191,16 @@ int DCT_common_init(MpegEncContext *s) return 0; } -static void copy_picture(Picture *dst, Picture *src){ +void copy_picture(Picture *dst, Picture *src){ *dst = *src; dst->type= FF_BUFFER_TYPE_COPY; } -#ifdef CONFIG_ENCODERS -static void copy_picture_attributes(MpegEncContext *s, AVFrame *dst, AVFrame *src){ - int i; - - dst->pict_type = src->pict_type; - dst->quality = src->quality; - dst->coded_picture_number = src->coded_picture_number; - dst->display_picture_number = src->display_picture_number; -// dst->reference = src->reference; - dst->pts = src->pts; - dst->interlaced_frame = src->interlaced_frame; - dst->top_field_first = src->top_field_first; - - if(s->avctx->me_threshold){ - if(!src->motion_val[0]) - av_log(s->avctx, AV_LOG_ERROR, "AVFrame.motion_val not set!\n"); - if(!src->mb_type) - av_log(s->avctx, AV_LOG_ERROR, "AVFrame.mb_type not set!\n"); - if(!src->ref_index[0]) - av_log(s->avctx, AV_LOG_ERROR, "AVFrame.ref_index not set!\n"); - if(src->motion_subsample_log2 != dst->motion_subsample_log2) - av_log(s->avctx, AV_LOG_ERROR, "AVFrame.motion_subsample_log2 doesn't match! (%d!=%d)\n", - src->motion_subsample_log2, dst->motion_subsample_log2); - - memcpy(dst->mb_type, src->mb_type, s->mb_stride * s->mb_height * sizeof(dst->mb_type[0])); - - for(i=0; i<2; i++){ - int stride= ((16*s->mb_width )>>src->motion_subsample_log2) + 1; - int height= ((16*s->mb_height)>>src->motion_subsample_log2); - - if(src->motion_val[i] && src->motion_val[i] != dst->motion_val[i]){ - memcpy(dst->motion_val[i], src->motion_val[i], 2*stride*height*sizeof(int16_t)); - } - if(src->ref_index[i] && src->ref_index[i] != dst->ref_index[i]){ - memcpy(dst->ref_index[i], src->ref_index[i], s->b8_stride*2*s->mb_height*sizeof(int8_t)); - } - } - } -} -#endif - /** * allocates a Picture * The pixels are allocated/set by calling get_buffer() if shared=0 */ -static int alloc_picture(MpegEncContext *s, Picture *pic, int shared){ +int alloc_picture(MpegEncContext *s, Picture *pic, int shared){ const int big_mb_num= s->mb_stride*(s->mb_height+1) + 1; //the +1 is needed so memset(,,stride*height) does not sig11 const int mb_array_size= s->mb_stride*s->mb_height; const int b8_array_size= s->b8_stride*s->mb_height*2; @@ -563,30 +397,11 @@ void ff_update_duplicate_context(MpegEncContext *dst, MpegEncContext *src){ //STOP_TIMER("update_duplicate_context") //about 10k cycles / 0.01 sec for 1000frames on 1ghz with 2 threads } -#ifdef CONFIG_ENCODERS -static void update_duplicate_context_after_me(MpegEncContext *dst, MpegEncContext *src){ -#define COPY(a) dst->a= src->a - COPY(pict_type); - COPY(current_picture); - COPY(f_code); - COPY(b_code); - COPY(qscale); - COPY(lambda); - COPY(lambda2); - COPY(picture_in_gop_number); - COPY(gop_picture_number); - COPY(frame_pred_frame_dct); //FIXME don't set in encode_header - COPY(progressive_frame); //FIXME don't set in encode_header - COPY(partitioned_frame); //FIXME don't set in encode_header -#undef COPY -} -#endif - /** * sets the given MpegEncContext to common defaults (same for encoding and decoding). * the changed fields will not depend upon the prior state of the MpegEncContext. */ -static void MPV_common_defaults(MpegEncContext *s){ +void MPV_common_defaults(MpegEncContext *s){ s->y_dc_scale_table= s->c_dc_scale_table= ff_mpeg1_dc_scale_table; s->chroma_qscale_table= ff_default_chroma_qscale_table; @@ -612,24 +427,6 @@ void MPV_decode_defaults(MpegEncContext *s){ MPV_common_defaults(s); } -/** - * sets the given MpegEncContext to defaults for encoding. - * the changed fields will not depend upon the prior state of the MpegEncContext. - */ - -#ifdef CONFIG_ENCODERS -static void MPV_encode_defaults(MpegEncContext *s){ - int i; - MPV_common_defaults(s); - - for(i=-16; i<16; i++){ - default_fcode_tab[i + MAX_MV]= 1; - } - s->me.mv_penalty= default_mv_penalty; - s->fcode_tab= default_fcode_tab; -} -#endif //CONFIG_ENCODERS - /** * init common structure for both encoder and decoder. * this assumes that some variables like width/height are already set @@ -907,509 +704,6 @@ void MPV_common_end(MpegEncContext *s) avcodec_default_free_buffers(s->avctx); } -#ifdef CONFIG_ENCODERS - -/* init video encoder */ -int MPV_encode_init(AVCodecContext *avctx) -{ - MpegEncContext *s = avctx->priv_data; - int i; - int chroma_h_shift, chroma_v_shift; - - MPV_encode_defaults(s); - - switch (avctx->codec_id) { - case CODEC_ID_MPEG2VIDEO: - if(avctx->pix_fmt != PIX_FMT_YUV420P && avctx->pix_fmt != PIX_FMT_YUV422P){ - av_log(avctx, AV_LOG_ERROR, "only YUV420 and YUV422 are supported\n"); - return -1; - } - break; - case CODEC_ID_LJPEG: - case CODEC_ID_MJPEG: - if(avctx->pix_fmt != PIX_FMT_YUVJ420P && avctx->pix_fmt != PIX_FMT_YUVJ422P && - ((avctx->pix_fmt != PIX_FMT_YUV420P && avctx->pix_fmt != PIX_FMT_YUV422P) || avctx->strict_std_compliance>FF_COMPLIANCE_INOFFICIAL)){ - av_log(avctx, AV_LOG_ERROR, "colorspace not supported in jpeg\n"); - return -1; - } - break; - default: - if(avctx->pix_fmt != PIX_FMT_YUV420P){ - av_log(avctx, AV_LOG_ERROR, "only YUV420 is supported\n"); - return -1; - } - } - - switch (avctx->pix_fmt) { - case PIX_FMT_YUVJ422P: - case PIX_FMT_YUV422P: - s->chroma_format = CHROMA_422; - break; - case PIX_FMT_YUVJ420P: - case PIX_FMT_YUV420P: - default: - s->chroma_format = CHROMA_420; - break; - } - - s->bit_rate = avctx->bit_rate; - s->width = avctx->width; - s->height = avctx->height; - if(avctx->gop_size > 600 && avctx->strict_std_compliance>FF_COMPLIANCE_EXPERIMENTAL){ - av_log(avctx, AV_LOG_ERROR, "Warning keyframe interval too large! reducing it ...\n"); - avctx->gop_size=600; - } - s->gop_size = avctx->gop_size; - s->avctx = avctx; - s->flags= avctx->flags; - s->flags2= avctx->flags2; - s->max_b_frames= avctx->max_b_frames; - s->codec_id= avctx->codec->id; - s->luma_elim_threshold = avctx->luma_elim_threshold; - s->chroma_elim_threshold= avctx->chroma_elim_threshold; - s->strict_std_compliance= avctx->strict_std_compliance; - s->data_partitioning= avctx->flags & CODEC_FLAG_PART; - s->quarter_sample= (avctx->flags & CODEC_FLAG_QPEL)!=0; - s->mpeg_quant= avctx->mpeg_quant; - s->rtp_mode= !!avctx->rtp_payload_size; - s->intra_dc_precision= avctx->intra_dc_precision; - s->user_specified_pts = AV_NOPTS_VALUE; - - if (s->gop_size <= 1) { - s->intra_only = 1; - s->gop_size = 12; - } else { - s->intra_only = 0; - } - - s->me_method = avctx->me_method; - - /* Fixed QSCALE */ - s->fixed_qscale = !!(avctx->flags & CODEC_FLAG_QSCALE); - - s->adaptive_quant= ( s->avctx->lumi_masking - || s->avctx->dark_masking - || s->avctx->temporal_cplx_masking - || s->avctx->spatial_cplx_masking - || s->avctx->p_masking - || s->avctx->border_masking - || (s->flags&CODEC_FLAG_QP_RD)) - && !s->fixed_qscale; - - s->obmc= !!(s->flags & CODEC_FLAG_OBMC); - s->loop_filter= !!(s->flags & CODEC_FLAG_LOOP_FILTER); - s->alternate_scan= !!(s->flags & CODEC_FLAG_ALT_SCAN); - s->intra_vlc_format= !!(s->flags2 & CODEC_FLAG2_INTRA_VLC); - s->q_scale_type= !!(s->flags2 & CODEC_FLAG2_NON_LINEAR_QUANT); - - if(avctx->rc_max_rate && !avctx->rc_buffer_size){ - av_log(avctx, AV_LOG_ERROR, "a vbv buffer size is needed, for encoding with a maximum bitrate\n"); - return -1; - } - - if(avctx->rc_min_rate && avctx->rc_max_rate != avctx->rc_min_rate){ - av_log(avctx, AV_LOG_INFO, "Warning min_rate > 0 but min_rate != max_rate isn't recommended!\n"); - } - - if(avctx->rc_min_rate && avctx->rc_min_rate > avctx->bit_rate){ - av_log(avctx, AV_LOG_ERROR, "bitrate below min bitrate\n"); - return -1; - } - - if(avctx->rc_max_rate && avctx->rc_max_rate < avctx->bit_rate){ - av_log(avctx, AV_LOG_INFO, "bitrate above max bitrate\n"); - return -1; - } - - if(avctx->rc_buffer_size && avctx->bit_rate*av_q2d(avctx->time_base) > avctx->rc_buffer_size){ - av_log(avctx, AV_LOG_ERROR, "VBV buffer too small for bitrate\n"); - return -1; - } - - if(avctx->bit_rate*av_q2d(avctx->time_base) > avctx->bit_rate_tolerance){ - av_log(avctx, AV_LOG_ERROR, "bitrate tolerance too small for bitrate\n"); - return -1; - } - - if( s->avctx->rc_max_rate && s->avctx->rc_min_rate == s->avctx->rc_max_rate - && (s->codec_id == CODEC_ID_MPEG1VIDEO || s->codec_id == CODEC_ID_MPEG2VIDEO) - && 90000LL * (avctx->rc_buffer_size-1) > s->avctx->rc_max_rate*0xFFFFLL){ - - av_log(avctx, AV_LOG_INFO, "Warning vbv_delay will be set to 0xFFFF (=VBR) as the specified vbv buffer is too large for the given bitrate!\n"); - } - - if((s->flags & CODEC_FLAG_4MV) && s->codec_id != CODEC_ID_MPEG4 - && s->codec_id != CODEC_ID_H263 && s->codec_id != CODEC_ID_H263P && s->codec_id != CODEC_ID_FLV1){ - av_log(avctx, AV_LOG_ERROR, "4MV not supported by codec\n"); - return -1; - } - - if(s->obmc && s->avctx->mb_decision != FF_MB_DECISION_SIMPLE){ - av_log(avctx, AV_LOG_ERROR, "OBMC is only supported with simple mb decision\n"); - return -1; - } - - if(s->obmc && s->codec_id != CODEC_ID_H263 && s->codec_id != CODEC_ID_H263P){ - av_log(avctx, AV_LOG_ERROR, "OBMC is only supported with H263(+)\n"); - return -1; - } - - if(s->quarter_sample && s->codec_id != CODEC_ID_MPEG4){ - av_log(avctx, AV_LOG_ERROR, "qpel not supported by codec\n"); - return -1; - } - - if(s->data_partitioning && s->codec_id != CODEC_ID_MPEG4){ - av_log(avctx, AV_LOG_ERROR, "data partitioning not supported by codec\n"); - return -1; - } - - if(s->max_b_frames && s->codec_id != CODEC_ID_MPEG4 && s->codec_id != CODEC_ID_MPEG1VIDEO && s->codec_id != CODEC_ID_MPEG2VIDEO){ - av_log(avctx, AV_LOG_ERROR, "b frames not supported by codec\n"); - return -1; - } - - if((s->flags & (CODEC_FLAG_INTERLACED_DCT|CODEC_FLAG_INTERLACED_ME|CODEC_FLAG_ALT_SCAN)) - && s->codec_id != CODEC_ID_MPEG4 && s->codec_id != CODEC_ID_MPEG2VIDEO){ - av_log(avctx, AV_LOG_ERROR, "interlacing not supported by codec\n"); - return -1; - } - - if(s->mpeg_quant && s->codec_id != CODEC_ID_MPEG4){ //FIXME mpeg2 uses that too - av_log(avctx, AV_LOG_ERROR, "mpeg2 style quantization not supported by codec\n"); - return -1; - } - - if((s->flags & CODEC_FLAG_CBP_RD) && !(s->flags & CODEC_FLAG_TRELLIS_QUANT)){ - av_log(avctx, AV_LOG_ERROR, "CBP RD needs trellis quant\n"); - return -1; - } - - if((s->flags & CODEC_FLAG_QP_RD) && s->avctx->mb_decision != FF_MB_DECISION_RD){ - av_log(avctx, AV_LOG_ERROR, "QP RD needs mbd=2\n"); - return -1; - } - - if(s->avctx->scenechange_threshold < 1000000000 && (s->flags & CODEC_FLAG_CLOSED_GOP)){ - av_log(avctx, AV_LOG_ERROR, "closed gop with scene change detection arent supported yet, set threshold to 1000000000\n"); - return -1; - } - - if((s->flags2 & CODEC_FLAG2_INTRA_VLC) && s->codec_id != CODEC_ID_MPEG2VIDEO){ - av_log(avctx, AV_LOG_ERROR, "intra vlc table not supported by codec\n"); - return -1; - } - - if(s->flags & CODEC_FLAG_LOW_DELAY){ - if (s->codec_id != CODEC_ID_MPEG2VIDEO && s->codec_id != CODEC_ID_MPEG1VIDEO){ - av_log(avctx, AV_LOG_ERROR, "low delay forcing is only available for mpeg1/2\n"); - return -1; - } - if (s->max_b_frames != 0){ - av_log(avctx, AV_LOG_ERROR, "b frames cannot be used with low delay\n"); - return -1; - } - } - - if(s->q_scale_type == 1){ - if(s->codec_id != CODEC_ID_MPEG2VIDEO){ - av_log(avctx, AV_LOG_ERROR, "non linear quant is only available for mpeg2\n"); - return -1; - } - if(avctx->qmax > 12){ - av_log(avctx, AV_LOG_ERROR, "non linear quant only supports qmax <= 12 currently\n"); - return -1; - } - } - - if(s->avctx->thread_count > 1 && s->codec_id != CODEC_ID_MPEG4 - && s->codec_id != CODEC_ID_MPEG1VIDEO && s->codec_id != CODEC_ID_MPEG2VIDEO - && (s->codec_id != CODEC_ID_H263P || !(s->flags & CODEC_FLAG_H263P_SLICE_STRUCT))){ - av_log(avctx, AV_LOG_ERROR, "multi threaded encoding not supported by codec\n"); - return -1; - } - - if(s->avctx->thread_count > 1) - s->rtp_mode= 1; - - if(!avctx->time_base.den || !avctx->time_base.num){ - av_log(avctx, AV_LOG_ERROR, "framerate not set\n"); - return -1; - } - - i= (INT_MAX/2+128)>>8; - if(avctx->me_threshold >= i){ - av_log(avctx, AV_LOG_ERROR, "me_threshold too large, max is %d\n", i - 1); - return -1; - } - if(avctx->mb_threshold >= i){ - av_log(avctx, AV_LOG_ERROR, "mb_threshold too large, max is %d\n", i - 1); - return -1; - } - - if(avctx->b_frame_strategy && (avctx->flags&CODEC_FLAG_PASS2)){ - av_log(avctx, AV_LOG_INFO, "notice: b_frame_strategy only affects the first pass\n"); - avctx->b_frame_strategy = 0; - } - - i= ff_gcd(avctx->time_base.den, avctx->time_base.num); - if(i > 1){ - av_log(avctx, AV_LOG_INFO, "removing common factors from framerate\n"); - avctx->time_base.den /= i; - avctx->time_base.num /= i; -// return -1; - } - - if(s->codec_id==CODEC_ID_MJPEG){ - s->intra_quant_bias= 1<<(QUANT_BIAS_SHIFT-1); //(a + x/2)/x - s->inter_quant_bias= 0; - }else if(s->mpeg_quant || s->codec_id==CODEC_ID_MPEG1VIDEO || s->codec_id==CODEC_ID_MPEG2VIDEO){ - s->intra_quant_bias= 3<<(QUANT_BIAS_SHIFT-3); //(a + x*3/8)/x - s->inter_quant_bias= 0; - }else{ - s->intra_quant_bias=0; - s->inter_quant_bias=-(1<<(QUANT_BIAS_SHIFT-2)); //(a - x/4)/x - } - - if(avctx->intra_quant_bias != FF_DEFAULT_QUANT_BIAS) - s->intra_quant_bias= avctx->intra_quant_bias; - if(avctx->inter_quant_bias != FF_DEFAULT_QUANT_BIAS) - s->inter_quant_bias= avctx->inter_quant_bias; - - avcodec_get_chroma_sub_sample(avctx->pix_fmt, &chroma_h_shift, &chroma_v_shift); - - if(avctx->codec_id == CODEC_ID_MPEG4 && s->avctx->time_base.den > (1<<16)-1){ - av_log(avctx, AV_LOG_ERROR, "timebase not supported by mpeg 4 standard\n"); - return -1; - } - s->time_increment_bits = av_log2(s->avctx->time_base.den - 1) + 1; - - switch(avctx->codec->id) { - case CODEC_ID_MPEG1VIDEO: - s->out_format = FMT_MPEG1; - s->low_delay= !!(s->flags & CODEC_FLAG_LOW_DELAY); - avctx->delay= s->low_delay ? 0 : (s->max_b_frames + 1); - break; - case CODEC_ID_MPEG2VIDEO: - s->out_format = FMT_MPEG1; - s->low_delay= !!(s->flags & CODEC_FLAG_LOW_DELAY); - avctx->delay= s->low_delay ? 0 : (s->max_b_frames + 1); - s->rtp_mode= 1; - break; - case CODEC_ID_LJPEG: - case CODEC_ID_MJPEG: - s->out_format = FMT_MJPEG; - s->intra_only = 1; /* force intra only for jpeg */ - s->mjpeg_vsample[0] = 2; - s->mjpeg_vsample[1] = 2>>chroma_v_shift; - s->mjpeg_vsample[2] = 2>>chroma_v_shift; - s->mjpeg_hsample[0] = 2; - s->mjpeg_hsample[1] = 2>>chroma_h_shift; - s->mjpeg_hsample[2] = 2>>chroma_h_shift; - if (!(ENABLE_MJPEG_ENCODER || ENABLE_LJPEG_ENCODER) - || ff_mjpeg_encode_init(s) < 0) - return -1; - avctx->delay=0; - s->low_delay=1; - break; - case CODEC_ID_H261: - if (!ENABLE_H261_ENCODER) return -1; - if (ff_h261_get_picture_format(s->width, s->height) < 0) { - av_log(avctx, AV_LOG_ERROR, "The specified picture size of %dx%d is not valid for the H.261 codec.\nValid sizes are 176x144, 352x288\n", s->width, s->height); - return -1; - } - s->out_format = FMT_H261; - avctx->delay=0; - s->low_delay=1; - break; - case CODEC_ID_H263: - if (h263_get_picture_format(s->width, s->height) == 7) { - av_log(avctx, AV_LOG_INFO, "The specified picture size of %dx%d is not valid for the H.263 codec.\nValid sizes are 128x96, 176x144, 352x288, 704x576, and 1408x1152. Try H.263+.\n", s->width, s->height); - return -1; - } - s->out_format = FMT_H263; - s->obmc= (avctx->flags & CODEC_FLAG_OBMC) ? 1:0; - avctx->delay=0; - s->low_delay=1; - break; - case CODEC_ID_H263P: - s->out_format = FMT_H263; - s->h263_plus = 1; - /* Fx */ - s->umvplus = (avctx->flags & CODEC_FLAG_H263P_UMV) ? 1:0; - s->h263_aic= (avctx->flags & CODEC_FLAG_AC_PRED) ? 1:0; - s->modified_quant= s->h263_aic; - s->alt_inter_vlc= (avctx->flags & CODEC_FLAG_H263P_AIV) ? 1:0; - s->obmc= (avctx->flags & CODEC_FLAG_OBMC) ? 1:0; - s->loop_filter= (avctx->flags & CODEC_FLAG_LOOP_FILTER) ? 1:0; - s->unrestricted_mv= s->obmc || s->loop_filter || s->umvplus; - s->h263_slice_structured= (s->flags & CODEC_FLAG_H263P_SLICE_STRUCT) ? 1:0; - - /* /Fx */ - /* These are just to be sure */ - avctx->delay=0; - s->low_delay=1; - break; - case CODEC_ID_FLV1: - s->out_format = FMT_H263; - s->h263_flv = 2; /* format = 1; 11-bit codes */ - s->unrestricted_mv = 1; - s->rtp_mode=0; /* don't allow GOB */ - avctx->delay=0; - s->low_delay=1; - break; - case CODEC_ID_RV10: - s->out_format = FMT_H263; - avctx->delay=0; - s->low_delay=1; - break; - case CODEC_ID_RV20: - s->out_format = FMT_H263; - avctx->delay=0; - s->low_delay=1; - s->modified_quant=1; - s->h263_aic=1; - s->h263_plus=1; - s->loop_filter=1; - s->unrestricted_mv= s->obmc || s->loop_filter || s->umvplus; - break; - case CODEC_ID_MPEG4: - s->out_format = FMT_H263; - s->h263_pred = 1; - s->unrestricted_mv = 1; - s->low_delay= s->max_b_frames ? 0 : 1; - avctx->delay= s->low_delay ? 0 : (s->max_b_frames + 1); - break; - case CODEC_ID_MSMPEG4V1: - s->out_format = FMT_H263; - s->h263_msmpeg4 = 1; - s->h263_pred = 1; - s->unrestricted_mv = 1; - s->msmpeg4_version= 1; - avctx->delay=0; - s->low_delay=1; - break; - case CODEC_ID_MSMPEG4V2: - s->out_format = FMT_H263; - s->h263_msmpeg4 = 1; - s->h263_pred = 1; - s->unrestricted_mv = 1; - s->msmpeg4_version= 2; - avctx->delay=0; - s->low_delay=1; - break; - case CODEC_ID_MSMPEG4V3: - s->out_format = FMT_H263; - s->h263_msmpeg4 = 1; - s->h263_pred = 1; - s->unrestricted_mv = 1; - s->msmpeg4_version= 3; - s->flipflop_rounding=1; - avctx->delay=0; - s->low_delay=1; - break; - case CODEC_ID_WMV1: - s->out_format = FMT_H263; - s->h263_msmpeg4 = 1; - s->h263_pred = 1; - s->unrestricted_mv = 1; - s->msmpeg4_version= 4; - s->flipflop_rounding=1; - avctx->delay=0; - s->low_delay=1; - break; - case CODEC_ID_WMV2: - s->out_format = FMT_H263; - s->h263_msmpeg4 = 1; - s->h263_pred = 1; - s->unrestricted_mv = 1; - s->msmpeg4_version= 5; - s->flipflop_rounding=1; - avctx->delay=0; - s->low_delay=1; - break; - default: - return -1; - } - - avctx->has_b_frames= !s->low_delay; - - s->encoding = 1; - - /* init */ - if (MPV_common_init(s) < 0) - return -1; - - if(s->modified_quant) - s->chroma_qscale_table= ff_h263_chroma_qscale_table; - s->progressive_frame= - s->progressive_sequence= !(avctx->flags & (CODEC_FLAG_INTERLACED_DCT|CODEC_FLAG_INTERLACED_ME|CODEC_FLAG_ALT_SCAN)); - s->quant_precision=5; - - ff_set_cmp(&s->dsp, s->dsp.ildct_cmp, s->avctx->ildct_cmp); - ff_set_cmp(&s->dsp, s->dsp.frame_skip_cmp, s->avctx->frame_skip_cmp); - - if (ENABLE_H261_ENCODER && s->out_format == FMT_H261) - ff_h261_encode_init(s); - if (s->out_format == FMT_H263) - h263_encode_init(s); - if (ENABLE_MSMPEG4_ENCODER && s->msmpeg4_version) - ff_msmpeg4_encode_init(s); - if (s->out_format == FMT_MPEG1) - ff_mpeg1_encode_init(s); - - /* init q matrix */ - for(i=0;i<64;i++) { - int j= s->dsp.idct_permutation[i]; - if(s->codec_id==CODEC_ID_MPEG4 && s->mpeg_quant){ - s->intra_matrix[j] = ff_mpeg4_default_intra_matrix[i]; - s->inter_matrix[j] = ff_mpeg4_default_non_intra_matrix[i]; - }else if(s->out_format == FMT_H263 || s->out_format == FMT_H261){ - s->intra_matrix[j] = - s->inter_matrix[j] = ff_mpeg1_default_non_intra_matrix[i]; - }else - { /* mpeg1/2 */ - s->intra_matrix[j] = ff_mpeg1_default_intra_matrix[i]; - s->inter_matrix[j] = ff_mpeg1_default_non_intra_matrix[i]; - } - if(s->avctx->intra_matrix) - s->intra_matrix[j] = s->avctx->intra_matrix[i]; - if(s->avctx->inter_matrix) - s->inter_matrix[j] = s->avctx->inter_matrix[i]; - } - - /* precompute matrix */ - /* for mjpeg, we do include qscale in the matrix */ - if (s->out_format != FMT_MJPEG) { - convert_matrix(&s->dsp, s->q_intra_matrix, s->q_intra_matrix16, - s->intra_matrix, s->intra_quant_bias, avctx->qmin, 31, 1); - convert_matrix(&s->dsp, s->q_inter_matrix, s->q_inter_matrix16, - s->inter_matrix, s->inter_quant_bias, avctx->qmin, 31, 0); - } - - if(ff_rate_control_init(s) < 0) - return -1; - - return 0; -} - -int MPV_encode_end(AVCodecContext *avctx) -{ - MpegEncContext *s = avctx->priv_data; - - ff_rate_control_uninit(s); - - MPV_common_end(s); - if ((ENABLE_MJPEG_ENCODER || ENABLE_LJPEG_ENCODER) && s->out_format == FMT_MJPEG) - ff_mjpeg_encode_close(s); - - av_freep(&avctx->extradata); - - return 0; -} - -#endif //CONFIG_ENCODERS - void init_rl(RLTable *rl, uint8_t static_store[2][2*MAX_RUN + MAX_LEVEL + 3]) { int8_t max_level[MAX_RUN+1], max_run[MAX_LEVEL+1]; @@ -2048,751 +1342,6 @@ v= (int)(128 + r*sin(theta*3.141592/180)); } } -#ifdef CONFIG_ENCODERS - -static int get_sae(uint8_t *src, int ref, int stride){ - int x,y; - int acc=0; - - for(y=0; y<16; y++){ - for(x=0; x<16; x++){ - acc+= FFABS(src[x+y*stride] - ref); - } - } - - return acc; -} - -static int get_intra_count(MpegEncContext *s, uint8_t *src, uint8_t *ref, int stride){ - int x, y, w, h; - int acc=0; - - w= s->width &~15; - h= s->height&~15; - - for(y=0; ydsp.sad[0](NULL, src + offset, ref + offset, stride, 16); - int mean= (s->dsp.pix_sum(src + offset, stride) + 128)>>8; - int sae = get_sae(src + offset, mean, stride); - - acc+= sae + 500 < sad; - } - } - return acc; -} - - -static int load_input_picture(MpegEncContext *s, AVFrame *pic_arg){ - AVFrame *pic=NULL; - int64_t pts; - int i; - const int encoding_delay= s->max_b_frames; - int direct=1; - - if(pic_arg){ - pts= pic_arg->pts; - pic_arg->display_picture_number= s->input_picture_number++; - - if(pts != AV_NOPTS_VALUE){ - if(s->user_specified_pts != AV_NOPTS_VALUE){ - int64_t time= pts; - int64_t last= s->user_specified_pts; - - if(time <= last){ - av_log(s->avctx, AV_LOG_ERROR, "Error, Invalid timestamp=%"PRId64", last=%"PRId64"\n", pts, s->user_specified_pts); - return -1; - } - } - s->user_specified_pts= pts; - }else{ - if(s->user_specified_pts != AV_NOPTS_VALUE){ - s->user_specified_pts= - pts= s->user_specified_pts + 1; - av_log(s->avctx, AV_LOG_INFO, "Warning: AVFrame.pts=? trying to guess (%"PRId64")\n", pts); - }else{ - pts= pic_arg->display_picture_number; - } - } - } - - if(pic_arg){ - if(encoding_delay && !(s->flags&CODEC_FLAG_INPUT_PRESERVED)) direct=0; - if(pic_arg->linesize[0] != s->linesize) direct=0; - if(pic_arg->linesize[1] != s->uvlinesize) direct=0; - if(pic_arg->linesize[2] != s->uvlinesize) direct=0; - -// av_log(AV_LOG_DEBUG, "%d %d %d %d\n",pic_arg->linesize[0], pic_arg->linesize[1], s->linesize, s->uvlinesize); - - if(direct){ - i= ff_find_unused_picture(s, 1); - - pic= (AVFrame*)&s->picture[i]; - pic->reference= 3; - - for(i=0; i<4; i++){ - pic->data[i]= pic_arg->data[i]; - pic->linesize[i]= pic_arg->linesize[i]; - } - alloc_picture(s, (Picture*)pic, 1); - }else{ - i= ff_find_unused_picture(s, 0); - - pic= (AVFrame*)&s->picture[i]; - pic->reference= 3; - - alloc_picture(s, (Picture*)pic, 0); - - if( pic->data[0] + INPLACE_OFFSET == pic_arg->data[0] - && pic->data[1] + INPLACE_OFFSET == pic_arg->data[1] - && pic->data[2] + INPLACE_OFFSET == pic_arg->data[2]){ - // empty - }else{ - int h_chroma_shift, v_chroma_shift; - avcodec_get_chroma_sub_sample(s->avctx->pix_fmt, &h_chroma_shift, &v_chroma_shift); - - for(i=0; i<3; i++){ - int src_stride= pic_arg->linesize[i]; - int dst_stride= i ? s->uvlinesize : s->linesize; - int h_shift= i ? h_chroma_shift : 0; - int v_shift= i ? v_chroma_shift : 0; - int w= s->width >>h_shift; - int h= s->height>>v_shift; - uint8_t *src= pic_arg->data[i]; - uint8_t *dst= pic->data[i]; - - if(!s->avctx->rc_buffer_size) - dst +=INPLACE_OFFSET; - - if(src_stride==dst_stride) - memcpy(dst, src, src_stride*h); - else{ - while(h--){ - memcpy(dst, src, w); - dst += dst_stride; - src += src_stride; - } - } - } - } - } - copy_picture_attributes(s, pic, pic_arg); - pic->pts= pts; //we set this here to avoid modifiying pic_arg - } - - /* shift buffer entries */ - for(i=1; iencoding_delay+1*/; i++) - s->input_picture[i-1]= s->input_picture[i]; - - s->input_picture[encoding_delay]= (Picture*)pic; - - return 0; -} - -static int skip_check(MpegEncContext *s, Picture *p, Picture *ref){ - int x, y, plane; - int score=0; - int64_t score64=0; - - for(plane=0; plane<3; plane++){ - const int stride= p->linesize[plane]; - const int bw= plane ? 1 : 2; - for(y=0; ymb_height*bw; y++){ - for(x=0; xmb_width*bw; x++){ - int off= p->type == FF_BUFFER_TYPE_SHARED ? 0: 16; - int v= s->dsp.frame_skip_cmp[1](s, p->data[plane] + 8*(x + y*stride)+off, ref->data[plane] + 8*(x + y*stride), stride, 8); - - switch(s->avctx->frame_skip_exp){ - case 0: score= FFMAX(score, v); break; - case 1: score+= FFABS(v);break; - case 2: score+= v*v;break; - case 3: score64+= FFABS(v*v*(int64_t)v);break; - case 4: score64+= v*v*(int64_t)(v*v);break; - } - } - } - } - - if(score) score64= score; - - if(score64 < s->avctx->frame_skip_threshold) - return 1; - if(score64 < ((s->avctx->frame_skip_factor * (int64_t)s->lambda)>>8)) - return 1; - return 0; -} - -static int estimate_best_b_count(MpegEncContext *s){ - AVCodec *codec= avcodec_find_encoder(s->avctx->codec_id); - AVCodecContext *c= avcodec_alloc_context(); - AVFrame input[FF_MAX_B_FRAMES+2]; - const int scale= s->avctx->brd_scale; - int i, j, out_size, p_lambda, b_lambda, lambda2; - int outbuf_size= s->width * s->height; //FIXME - uint8_t *outbuf= av_malloc(outbuf_size); - int64_t best_rd= INT64_MAX; - int best_b_count= -1; - - assert(scale>=0 && scale <=3); - -// emms_c(); - p_lambda= s->last_lambda_for[P_TYPE]; //s->next_picture_ptr->quality; - b_lambda= s->last_lambda_for[B_TYPE]; //p_lambda *FFABS(s->avctx->b_quant_factor) + s->avctx->b_quant_offset; - if(!b_lambda) b_lambda= p_lambda; //FIXME we should do this somewhere else - lambda2= (b_lambda*b_lambda + (1<> FF_LAMBDA_SHIFT; - - c->width = s->width >> scale; - c->height= s->height>> scale; - c->flags= CODEC_FLAG_QSCALE | CODEC_FLAG_PSNR | CODEC_FLAG_INPUT_PRESERVED /*| CODEC_FLAG_EMU_EDGE*/; - c->flags|= s->avctx->flags & CODEC_FLAG_QPEL; - c->mb_decision= s->avctx->mb_decision; - c->me_cmp= s->avctx->me_cmp; - c->mb_cmp= s->avctx->mb_cmp; - c->me_sub_cmp= s->avctx->me_sub_cmp; - c->pix_fmt = PIX_FMT_YUV420P; - c->time_base= s->avctx->time_base; - c->max_b_frames= s->max_b_frames; - - if (avcodec_open(c, codec) < 0) - return -1; - - for(i=0; imax_b_frames+2; i++){ - int ysize= c->width*c->height; - int csize= (c->width/2)*(c->height/2); - Picture pre_input, *pre_input_ptr= i ? s->input_picture[i-1] : s->next_picture_ptr; - - avcodec_get_frame_defaults(&input[i]); - input[i].data[0]= av_malloc(ysize + 2*csize); - input[i].data[1]= input[i].data[0] + ysize; - input[i].data[2]= input[i].data[1] + csize; - input[i].linesize[0]= c->width; - input[i].linesize[1]= - input[i].linesize[2]= c->width/2; - - if(pre_input_ptr && (!i || s->input_picture[i-1])) { - pre_input= *pre_input_ptr; - - if(pre_input.type != FF_BUFFER_TYPE_SHARED && i) { - pre_input.data[0]+=INPLACE_OFFSET; - pre_input.data[1]+=INPLACE_OFFSET; - pre_input.data[2]+=INPLACE_OFFSET; - } - - s->dsp.shrink[scale](input[i].data[0], input[i].linesize[0], pre_input.data[0], pre_input.linesize[0], c->width, c->height); - s->dsp.shrink[scale](input[i].data[1], input[i].linesize[1], pre_input.data[1], pre_input.linesize[1], c->width>>1, c->height>>1); - s->dsp.shrink[scale](input[i].data[2], input[i].linesize[2], pre_input.data[2], pre_input.linesize[2], c->width>>1, c->height>>1); - } - } - - for(j=0; jmax_b_frames+1; j++){ - int64_t rd=0; - - if(!s->input_picture[j]) - break; - - c->error[0]= c->error[1]= c->error[2]= 0; - - input[0].pict_type= I_TYPE; - input[0].quality= 1 * FF_QP2LAMBDA; - out_size = avcodec_encode_video(c, outbuf, outbuf_size, &input[0]); -// rd += (out_size * lambda2) >> FF_LAMBDA_SHIFT; - - for(i=0; imax_b_frames+1; i++){ - int is_p= i % (j+1) == j || i==s->max_b_frames; - - input[i+1].pict_type= is_p ? P_TYPE : B_TYPE; - input[i+1].quality= is_p ? p_lambda : b_lambda; - out_size = avcodec_encode_video(c, outbuf, outbuf_size, &input[i+1]); - rd += (out_size * lambda2) >> (FF_LAMBDA_SHIFT - 3); - } - - /* get the delayed frames */ - while(out_size){ - out_size = avcodec_encode_video(c, outbuf, outbuf_size, NULL); - rd += (out_size * lambda2) >> (FF_LAMBDA_SHIFT - 3); - } - - rd += c->error[0] + c->error[1] + c->error[2]; - - if(rd < best_rd){ - best_rd= rd; - best_b_count= j; - } - } - - av_freep(&outbuf); - avcodec_close(c); - av_freep(&c); - - for(i=0; imax_b_frames+2; i++){ - av_freep(&input[i].data[0]); - } - - return best_b_count; -} - -static void select_input_picture(MpegEncContext *s){ - int i; - - for(i=1; ireordered_input_picture[i-1]= s->reordered_input_picture[i]; - s->reordered_input_picture[MAX_PICTURE_COUNT-1]= NULL; - - /* set next picture type & ordering */ - if(s->reordered_input_picture[0]==NULL && s->input_picture[0]){ - if(/*s->picture_in_gop_number >= s->gop_size ||*/ s->next_picture_ptr==NULL || s->intra_only){ - s->reordered_input_picture[0]= s->input_picture[0]; - s->reordered_input_picture[0]->pict_type= I_TYPE; - s->reordered_input_picture[0]->coded_picture_number= s->coded_picture_number++; - }else{ - int b_frames; - - if(s->avctx->frame_skip_threshold || s->avctx->frame_skip_factor){ - if(s->picture_in_gop_number < s->gop_size && skip_check(s, s->input_picture[0], s->next_picture_ptr)){ - //FIXME check that te gop check above is +-1 correct -//av_log(NULL, AV_LOG_DEBUG, "skip %p %"PRId64"\n", s->input_picture[0]->data[0], s->input_picture[0]->pts); - - if(s->input_picture[0]->type == FF_BUFFER_TYPE_SHARED){ - for(i=0; i<4; i++) - s->input_picture[0]->data[i]= NULL; - s->input_picture[0]->type= 0; - }else{ - assert( s->input_picture[0]->type==FF_BUFFER_TYPE_USER - || s->input_picture[0]->type==FF_BUFFER_TYPE_INTERNAL); - - s->avctx->release_buffer(s->avctx, (AVFrame*)s->input_picture[0]); - } - - emms_c(); - ff_vbv_update(s, 0); - - goto no_output_pic; - } - } - - if(s->flags&CODEC_FLAG_PASS2){ - for(i=0; imax_b_frames+1; i++){ - int pict_num= s->input_picture[0]->display_picture_number + i; - - if(pict_num >= s->rc_context.num_entries) - break; - if(!s->input_picture[i]){ - s->rc_context.entry[pict_num-1].new_pict_type = P_TYPE; - break; - } - - s->input_picture[i]->pict_type= - s->rc_context.entry[pict_num].new_pict_type; - } - } - - if(s->avctx->b_frame_strategy==0){ - b_frames= s->max_b_frames; - while(b_frames && !s->input_picture[b_frames]) b_frames--; - }else if(s->avctx->b_frame_strategy==1){ - for(i=1; imax_b_frames+1; i++){ - if(s->input_picture[i] && s->input_picture[i]->b_frame_score==0){ - s->input_picture[i]->b_frame_score= - get_intra_count(s, s->input_picture[i ]->data[0], - s->input_picture[i-1]->data[0], s->linesize) + 1; - } - } - for(i=0; imax_b_frames+1; i++){ - if(s->input_picture[i]==NULL || s->input_picture[i]->b_frame_score - 1 > s->mb_num/s->avctx->b_sensitivity) break; - } - - b_frames= FFMAX(0, i-1); - - /* reset scores */ - for(i=0; iinput_picture[i]->b_frame_score=0; - } - }else if(s->avctx->b_frame_strategy==2){ - b_frames= estimate_best_b_count(s); - }else{ - av_log(s->avctx, AV_LOG_ERROR, "illegal b frame strategy\n"); - b_frames=0; - } - - emms_c(); -//static int b_count=0; -//b_count+= b_frames; -//av_log(s->avctx, AV_LOG_DEBUG, "b_frames: %d\n", b_count); - - for(i= b_frames - 1; i>=0; i--){ - int type= s->input_picture[i]->pict_type; - if(type && type != B_TYPE) - b_frames= i; - } - if(s->input_picture[b_frames]->pict_type == B_TYPE && b_frames == s->max_b_frames){ - av_log(s->avctx, AV_LOG_ERROR, "warning, too many b frames in a row\n"); - } - - if(s->picture_in_gop_number + b_frames >= s->gop_size){ - if((s->flags2 & CODEC_FLAG2_STRICT_GOP) && s->gop_size > s->picture_in_gop_number){ - b_frames= s->gop_size - s->picture_in_gop_number - 1; - }else{ - if(s->flags & CODEC_FLAG_CLOSED_GOP) - b_frames=0; - s->input_picture[b_frames]->pict_type= I_TYPE; - } - } - - if( (s->flags & CODEC_FLAG_CLOSED_GOP) - && b_frames - && s->input_picture[b_frames]->pict_type== I_TYPE) - b_frames--; - - s->reordered_input_picture[0]= s->input_picture[b_frames]; - if(s->reordered_input_picture[0]->pict_type != I_TYPE) - s->reordered_input_picture[0]->pict_type= P_TYPE; - s->reordered_input_picture[0]->coded_picture_number= s->coded_picture_number++; - for(i=0; ireordered_input_picture[i+1]= s->input_picture[i]; - s->reordered_input_picture[i+1]->pict_type= B_TYPE; - s->reordered_input_picture[i+1]->coded_picture_number= s->coded_picture_number++; - } - } - } -no_output_pic: - if(s->reordered_input_picture[0]){ - s->reordered_input_picture[0]->reference= s->reordered_input_picture[0]->pict_type!=B_TYPE ? 3 : 0; - - copy_picture(&s->new_picture, s->reordered_input_picture[0]); - - if(s->reordered_input_picture[0]->type == FF_BUFFER_TYPE_SHARED || s->avctx->rc_buffer_size){ - // input is a shared pix, so we can't modifiy it -> alloc a new one & ensure that the shared one is reuseable - - int i= ff_find_unused_picture(s, 0); - Picture *pic= &s->picture[i]; - - pic->reference = s->reordered_input_picture[0]->reference; - alloc_picture(s, pic, 0); - - /* mark us unused / free shared pic */ - if(s->reordered_input_picture[0]->type == FF_BUFFER_TYPE_INTERNAL) - s->avctx->release_buffer(s->avctx, (AVFrame*)s->reordered_input_picture[0]); - for(i=0; i<4; i++) - s->reordered_input_picture[0]->data[i]= NULL; - s->reordered_input_picture[0]->type= 0; - - copy_picture_attributes(s, (AVFrame*)pic, (AVFrame*)s->reordered_input_picture[0]); - - s->current_picture_ptr= pic; - }else{ - // input is not a shared pix -> reuse buffer for current_pix - - assert( s->reordered_input_picture[0]->type==FF_BUFFER_TYPE_USER - || s->reordered_input_picture[0]->type==FF_BUFFER_TYPE_INTERNAL); - - s->current_picture_ptr= s->reordered_input_picture[0]; - for(i=0; i<4; i++){ - s->new_picture.data[i]+= INPLACE_OFFSET; - } - } - copy_picture(&s->current_picture, s->current_picture_ptr); - - s->picture_number= s->new_picture.display_picture_number; -//printf("dpn:%d\n", s->picture_number); - }else{ - memset(&s->new_picture, 0, sizeof(Picture)); - } -} - -int MPV_encode_picture(AVCodecContext *avctx, - unsigned char *buf, int buf_size, void *data) -{ - MpegEncContext *s = avctx->priv_data; - AVFrame *pic_arg = data; - int i, stuffing_count; - - for(i=0; ithread_count; i++){ - int start_y= s->thread_context[i]->start_mb_y; - int end_y= s->thread_context[i]-> end_mb_y; - int h= s->mb_height; - uint8_t *start= buf + (size_t)(((int64_t) buf_size)*start_y/h); - uint8_t *end = buf + (size_t)(((int64_t) buf_size)* end_y/h); - - init_put_bits(&s->thread_context[i]->pb, start, end - start); - } - - s->picture_in_gop_number++; - - if(load_input_picture(s, pic_arg) < 0) - return -1; - - select_input_picture(s); - - /* output? */ - if(s->new_picture.data[0]){ - s->pict_type= s->new_picture.pict_type; -//emms_c(); -//printf("qs:%f %f %d\n", s->new_picture.quality, s->current_picture.quality, s->qscale); - MPV_frame_start(s, avctx); -vbv_retry: - if (encode_picture(s, s->picture_number) < 0) - return -1; - - avctx->real_pict_num = s->picture_number; - avctx->header_bits = s->header_bits; - avctx->mv_bits = s->mv_bits; - avctx->misc_bits = s->misc_bits; - avctx->i_tex_bits = s->i_tex_bits; - avctx->p_tex_bits = s->p_tex_bits; - avctx->i_count = s->i_count; - avctx->p_count = s->mb_num - s->i_count - s->skip_count; //FIXME f/b_count in avctx - avctx->skip_count = s->skip_count; - - MPV_frame_end(s); - - if (ENABLE_MJPEG_ENCODER && s->out_format == FMT_MJPEG) - ff_mjpeg_encode_picture_trailer(s); - - if(avctx->rc_buffer_size){ - RateControlContext *rcc= &s->rc_context; - int max_size= rcc->buffer_index/3; - - if(put_bits_count(&s->pb) > max_size && s->lambda < s->avctx->lmax){ - s->next_lambda= FFMAX(s->lambda+1, s->lambda*(s->qscale+1) / s->qscale); - if(s->adaptive_quant){ - int i; - for(i=0; imb_height*s->mb_stride; i++) - s->lambda_table[i]= FFMAX(s->lambda_table[i]+1, s->lambda_table[i]*(s->qscale+1) / s->qscale); - } - s->mb_skipped = 0; //done in MPV_frame_start() - if(s->pict_type==P_TYPE){ //done in encode_picture() so we must undo it - if(s->flipflop_rounding || s->codec_id == CODEC_ID_H263P || s->codec_id == CODEC_ID_MPEG4) - s->no_rounding ^= 1; - } - if(s->pict_type!=B_TYPE){ - s->time_base= s->last_time_base; - s->last_non_b_time= s->time - s->pp_time; - } -// av_log(NULL, AV_LOG_ERROR, "R:%d ", s->next_lambda); - for(i=0; ithread_count; i++){ - PutBitContext *pb= &s->thread_context[i]->pb; - init_put_bits(pb, pb->buf, pb->buf_end - pb->buf); - } - goto vbv_retry; - } - - assert(s->avctx->rc_max_rate); - } - - if(s->flags&CODEC_FLAG_PASS1) - ff_write_pass1_stats(s); - - for(i=0; i<4; i++){ - s->current_picture_ptr->error[i]= s->current_picture.error[i]; - avctx->error[i] += s->current_picture_ptr->error[i]; - } - - if(s->flags&CODEC_FLAG_PASS1) - assert(avctx->header_bits + avctx->mv_bits + avctx->misc_bits + avctx->i_tex_bits + avctx->p_tex_bits == put_bits_count(&s->pb)); - flush_put_bits(&s->pb); - s->frame_bits = put_bits_count(&s->pb); - - stuffing_count= ff_vbv_update(s, s->frame_bits); - if(stuffing_count){ - if(s->pb.buf_end - s->pb.buf - (put_bits_count(&s->pb)>>3) < stuffing_count + 50){ - av_log(s->avctx, AV_LOG_ERROR, "stuffing too large\n"); - return -1; - } - - switch(s->codec_id){ - case CODEC_ID_MPEG1VIDEO: - case CODEC_ID_MPEG2VIDEO: - while(stuffing_count--){ - put_bits(&s->pb, 8, 0); - } - break; - case CODEC_ID_MPEG4: - put_bits(&s->pb, 16, 0); - put_bits(&s->pb, 16, 0x1C3); - stuffing_count -= 4; - while(stuffing_count--){ - put_bits(&s->pb, 8, 0xFF); - } - break; - default: - av_log(s->avctx, AV_LOG_ERROR, "vbv buffer overflow\n"); - } - flush_put_bits(&s->pb); - s->frame_bits = put_bits_count(&s->pb); - } - - /* update mpeg1/2 vbv_delay for CBR */ - if(s->avctx->rc_max_rate && s->avctx->rc_min_rate == s->avctx->rc_max_rate && s->out_format == FMT_MPEG1 - && 90000LL * (avctx->rc_buffer_size-1) <= s->avctx->rc_max_rate*0xFFFFLL){ - int vbv_delay; - - assert(s->repeat_first_field==0); - - vbv_delay= lrintf(90000 * s->rc_context.buffer_index / s->avctx->rc_max_rate); - assert(vbv_delay < 0xFFFF); - - s->vbv_delay_ptr[0] &= 0xF8; - s->vbv_delay_ptr[0] |= vbv_delay>>13; - s->vbv_delay_ptr[1] = vbv_delay>>5; - s->vbv_delay_ptr[2] &= 0x07; - s->vbv_delay_ptr[2] |= vbv_delay<<3; - } - s->total_bits += s->frame_bits; - avctx->frame_bits = s->frame_bits; - }else{ - assert((pbBufPtr(&s->pb) == s->pb.buf)); - s->frame_bits=0; - } - assert((s->frame_bits&7)==0); - - return s->frame_bits/8; -} - -#endif //CONFIG_ENCODERS - -static inline void gmc1_motion(MpegEncContext *s, - uint8_t *dest_y, uint8_t *dest_cb, uint8_t *dest_cr, - uint8_t **ref_picture) -{ - uint8_t *ptr; - int offset, src_x, src_y, linesize, uvlinesize; - int motion_x, motion_y; - int emu=0; - - motion_x= s->sprite_offset[0][0]; - motion_y= s->sprite_offset[0][1]; - src_x = s->mb_x * 16 + (motion_x >> (s->sprite_warping_accuracy+1)); - src_y = s->mb_y * 16 + (motion_y >> (s->sprite_warping_accuracy+1)); - motion_x<<=(3-s->sprite_warping_accuracy); - motion_y<<=(3-s->sprite_warping_accuracy); - src_x = av_clip(src_x, -16, s->width); - if (src_x == s->width) - motion_x =0; - src_y = av_clip(src_y, -16, s->height); - if (src_y == s->height) - motion_y =0; - - linesize = s->linesize; - uvlinesize = s->uvlinesize; - - ptr = ref_picture[0] + (src_y * linesize) + src_x; - - if(s->flags&CODEC_FLAG_EMU_EDGE){ - if( (unsigned)src_x >= s->h_edge_pos - 17 - || (unsigned)src_y >= s->v_edge_pos - 17){ - ff_emulated_edge_mc(s->edge_emu_buffer, ptr, linesize, 17, 17, src_x, src_y, s->h_edge_pos, s->v_edge_pos); - ptr= s->edge_emu_buffer; - } - } - - if((motion_x|motion_y)&7){ - s->dsp.gmc1(dest_y , ptr , linesize, 16, motion_x&15, motion_y&15, 128 - s->no_rounding); - s->dsp.gmc1(dest_y+8, ptr+8, linesize, 16, motion_x&15, motion_y&15, 128 - s->no_rounding); - }else{ - int dxy; - - dxy= ((motion_x>>3)&1) | ((motion_y>>2)&2); - if (s->no_rounding){ - s->dsp.put_no_rnd_pixels_tab[0][dxy](dest_y, ptr, linesize, 16); - }else{ - s->dsp.put_pixels_tab [0][dxy](dest_y, ptr, linesize, 16); - } - } - - if(s->flags&CODEC_FLAG_GRAY) return; - - motion_x= s->sprite_offset[1][0]; - motion_y= s->sprite_offset[1][1]; - src_x = s->mb_x * 8 + (motion_x >> (s->sprite_warping_accuracy+1)); - src_y = s->mb_y * 8 + (motion_y >> (s->sprite_warping_accuracy+1)); - motion_x<<=(3-s->sprite_warping_accuracy); - motion_y<<=(3-s->sprite_warping_accuracy); - src_x = av_clip(src_x, -8, s->width>>1); - if (src_x == s->width>>1) - motion_x =0; - src_y = av_clip(src_y, -8, s->height>>1); - if (src_y == s->height>>1) - motion_y =0; - - offset = (src_y * uvlinesize) + src_x; - ptr = ref_picture[1] + offset; - if(s->flags&CODEC_FLAG_EMU_EDGE){ - if( (unsigned)src_x >= (s->h_edge_pos>>1) - 9 - || (unsigned)src_y >= (s->v_edge_pos>>1) - 9){ - ff_emulated_edge_mc(s->edge_emu_buffer, ptr, uvlinesize, 9, 9, src_x, src_y, s->h_edge_pos>>1, s->v_edge_pos>>1); - ptr= s->edge_emu_buffer; - emu=1; - } - } - s->dsp.gmc1(dest_cb, ptr, uvlinesize, 8, motion_x&15, motion_y&15, 128 - s->no_rounding); - - ptr = ref_picture[2] + offset; - if(emu){ - ff_emulated_edge_mc(s->edge_emu_buffer, ptr, uvlinesize, 9, 9, src_x, src_y, s->h_edge_pos>>1, s->v_edge_pos>>1); - ptr= s->edge_emu_buffer; - } - s->dsp.gmc1(dest_cr, ptr, uvlinesize, 8, motion_x&15, motion_y&15, 128 - s->no_rounding); - - return; -} - -static inline void gmc_motion(MpegEncContext *s, - uint8_t *dest_y, uint8_t *dest_cb, uint8_t *dest_cr, - uint8_t **ref_picture) -{ - uint8_t *ptr; - int linesize, uvlinesize; - const int a= s->sprite_warping_accuracy; - int ox, oy; - - linesize = s->linesize; - uvlinesize = s->uvlinesize; - - ptr = ref_picture[0]; - - ox= s->sprite_offset[0][0] + s->sprite_delta[0][0]*s->mb_x*16 + s->sprite_delta[0][1]*s->mb_y*16; - oy= s->sprite_offset[0][1] + s->sprite_delta[1][0]*s->mb_x*16 + s->sprite_delta[1][1]*s->mb_y*16; - - s->dsp.gmc(dest_y, ptr, linesize, 16, - ox, - oy, - s->sprite_delta[0][0], s->sprite_delta[0][1], - s->sprite_delta[1][0], s->sprite_delta[1][1], - a+1, (1<<(2*a+1)) - s->no_rounding, - s->h_edge_pos, s->v_edge_pos); - s->dsp.gmc(dest_y+8, ptr, linesize, 16, - ox + s->sprite_delta[0][0]*8, - oy + s->sprite_delta[1][0]*8, - s->sprite_delta[0][0], s->sprite_delta[0][1], - s->sprite_delta[1][0], s->sprite_delta[1][1], - a+1, (1<<(2*a+1)) - s->no_rounding, - s->h_edge_pos, s->v_edge_pos); - - if(s->flags&CODEC_FLAG_GRAY) return; - - ox= s->sprite_offset[1][0] + s->sprite_delta[0][0]*s->mb_x*8 + s->sprite_delta[0][1]*s->mb_y*8; - oy= s->sprite_offset[1][1] + s->sprite_delta[1][0]*s->mb_x*8 + s->sprite_delta[1][1]*s->mb_y*8; - - ptr = ref_picture[1]; - s->dsp.gmc(dest_cb, ptr, uvlinesize, 8, - ox, - oy, - s->sprite_delta[0][0], s->sprite_delta[0][1], - s->sprite_delta[1][0], s->sprite_delta[1][1], - a+1, (1<<(2*a+1)) - s->no_rounding, - s->h_edge_pos>>1, s->v_edge_pos>>1); - - ptr = ref_picture[2]; - s->dsp.gmc(dest_cr, ptr, uvlinesize, 8, - ox, - oy, - s->sprite_delta[0][0], s->sprite_delta[0][1], - s->sprite_delta[1][0], s->sprite_delta[1][1], - a+1, (1<<(2*a+1)) - s->no_rounding, - s->h_edge_pos>>1, s->v_edge_pos>>1); -} - /** * Copies a rectangular area of samples to a temporary buffer and replicates the boarder samples. * @param buf destination buffer @@ -2864,46 +1413,6 @@ void ff_emulated_edge_mc(uint8_t *buf, uint8_t *src, int linesize, int block_w, } } -static inline int hpel_motion(MpegEncContext *s, - uint8_t *dest, uint8_t *src, - int field_based, int field_select, - int src_x, int src_y, - int width, int height, int stride, - int h_edge_pos, int v_edge_pos, - int w, int h, op_pixels_func *pix_op, - int motion_x, int motion_y) -{ - int dxy; - int emu=0; - - dxy = ((motion_y & 1) << 1) | (motion_x & 1); - src_x += motion_x >> 1; - src_y += motion_y >> 1; - - /* WARNING: do no forget half pels */ - src_x = av_clip(src_x, -16, width); //FIXME unneeded for emu? - if (src_x == width) - dxy &= ~1; - src_y = av_clip(src_y, -16, height); - if (src_y == height) - dxy &= ~2; - src += src_y * stride + src_x; - - if(s->unrestricted_mv && (s->flags&CODEC_FLAG_EMU_EDGE)){ - if( (unsigned)src_x > h_edge_pos - (motion_x&1) - w - || (unsigned)src_y > v_edge_pos - (motion_y&1) - h){ - ff_emulated_edge_mc(s->edge_emu_buffer, src, s->linesize, w+1, (h+1)<v_edge_pos); - src= s->edge_emu_buffer; - emu=1; - } - } - if(field_select) - src += s->linesize; - pix_op[dxy](dest, src, stride, h); - return emu; -} - static inline int hpel_motion_lowres(MpegEncContext *s, uint8_t *dest, uint8_t *src, int field_based, int field_select, @@ -2946,121 +1455,6 @@ static inline int hpel_motion_lowres(MpegEncContext *s, return emu; } -/* apply one mpeg motion vector to the three components */ -static av_always_inline void mpeg_motion(MpegEncContext *s, - uint8_t *dest_y, uint8_t *dest_cb, uint8_t *dest_cr, - int field_based, int bottom_field, int field_select, - uint8_t **ref_picture, op_pixels_func (*pix_op)[4], - int motion_x, int motion_y, int h) -{ - uint8_t *ptr_y, *ptr_cb, *ptr_cr; - int dxy, uvdxy, mx, my, src_x, src_y, uvsrc_x, uvsrc_y, v_edge_pos, uvlinesize, linesize; - -#if 0 -if(s->quarter_sample) -{ - motion_x>>=1; - motion_y>>=1; -} -#endif - - v_edge_pos = s->v_edge_pos >> field_based; - linesize = s->current_picture.linesize[0] << field_based; - uvlinesize = s->current_picture.linesize[1] << field_based; - - dxy = ((motion_y & 1) << 1) | (motion_x & 1); - src_x = s->mb_x* 16 + (motion_x >> 1); - src_y =(s->mb_y<<(4-field_based)) + (motion_y >> 1); - - if (s->out_format == FMT_H263) { - if((s->workaround_bugs & FF_BUG_HPEL_CHROMA) && field_based){ - mx = (motion_x>>1)|(motion_x&1); - my = motion_y >>1; - uvdxy = ((my & 1) << 1) | (mx & 1); - uvsrc_x = s->mb_x* 8 + (mx >> 1); - uvsrc_y = (s->mb_y<<(3-field_based)) + (my >> 1); - }else{ - uvdxy = dxy | (motion_y & 2) | ((motion_x & 2) >> 1); - uvsrc_x = src_x>>1; - uvsrc_y = src_y>>1; - } - }else if(s->out_format == FMT_H261){//even chroma mv's are full pel in H261 - mx = motion_x / 4; - my = motion_y / 4; - uvdxy = 0; - uvsrc_x = s->mb_x*8 + mx; - uvsrc_y = s->mb_y*8 + my; - } else { - if(s->chroma_y_shift){ - mx = motion_x / 2; - my = motion_y / 2; - uvdxy = ((my & 1) << 1) | (mx & 1); - uvsrc_x = s->mb_x* 8 + (mx >> 1); - uvsrc_y = (s->mb_y<<(3-field_based)) + (my >> 1); - } else { - if(s->chroma_x_shift){ - //Chroma422 - mx = motion_x / 2; - uvdxy = ((motion_y & 1) << 1) | (mx & 1); - uvsrc_x = s->mb_x* 8 + (mx >> 1); - uvsrc_y = src_y; - } else { - //Chroma444 - uvdxy = dxy; - uvsrc_x = src_x; - uvsrc_y = src_y; - } - } - } - - ptr_y = ref_picture[0] + src_y * linesize + src_x; - ptr_cb = ref_picture[1] + uvsrc_y * uvlinesize + uvsrc_x; - ptr_cr = ref_picture[2] + uvsrc_y * uvlinesize + uvsrc_x; - - if( (unsigned)src_x > s->h_edge_pos - (motion_x&1) - 16 - || (unsigned)src_y > v_edge_pos - (motion_y&1) - h){ - if(s->codec_id == CODEC_ID_MPEG2VIDEO || - s->codec_id == CODEC_ID_MPEG1VIDEO){ - av_log(s->avctx,AV_LOG_DEBUG,"MPEG motion vector out of boundary\n"); - return ; - } - ff_emulated_edge_mc(s->edge_emu_buffer, ptr_y, s->linesize, 17, 17+field_based, - src_x, src_y<h_edge_pos, s->v_edge_pos); - ptr_y = s->edge_emu_buffer; - if(!(s->flags&CODEC_FLAG_GRAY)){ - uint8_t *uvbuf= s->edge_emu_buffer+18*s->linesize; - ff_emulated_edge_mc(uvbuf , ptr_cb, s->uvlinesize, 9, 9+field_based, - uvsrc_x, uvsrc_y<h_edge_pos>>1, s->v_edge_pos>>1); - ff_emulated_edge_mc(uvbuf+16, ptr_cr, s->uvlinesize, 9, 9+field_based, - uvsrc_x, uvsrc_y<h_edge_pos>>1, s->v_edge_pos>>1); - ptr_cb= uvbuf; - ptr_cr= uvbuf+16; - } - } - - if(bottom_field){ //FIXME use this for field pix too instead of the obnoxious hack which changes picture.data - dest_y += s->linesize; - dest_cb+= s->uvlinesize; - dest_cr+= s->uvlinesize; - } - - if(field_select){ - ptr_y += s->linesize; - ptr_cb+= s->uvlinesize; - ptr_cr+= s->uvlinesize; - } - - pix_op[0][dxy](dest_y, ptr_y, linesize, h); - - if(!(s->flags&CODEC_FLAG_GRAY)){ - pix_op[s->chroma_x_shift][uvdxy](dest_cb, ptr_cb, uvlinesize, h >> s->chroma_y_shift); - pix_op[s->chroma_x_shift][uvdxy](dest_cr, ptr_cr, uvlinesize, h >> s->chroma_y_shift); - } - if((ENABLE_H261_ENCODER || ENABLE_H261_DECODER) && s->out_format == FMT_H261){ - ff_h261_loop_filter(s); - } -} - /* apply one mpeg motion vector to the three components */ static av_always_inline void mpeg_motion_lowres(MpegEncContext *s, uint8_t *dest_y, uint8_t *dest_cb, uint8_t *dest_cr, @@ -3158,175 +1552,6 @@ static av_always_inline void mpeg_motion_lowres(MpegEncContext *s, //FIXME h261 lowres loop filter } -//FIXME move to dsputil, avg variant, 16x16 version -static inline void put_obmc(uint8_t *dst, uint8_t *src[5], int stride){ - int x; - uint8_t * const top = src[1]; - uint8_t * const left = src[2]; - uint8_t * const mid = src[0]; - uint8_t * const right = src[3]; - uint8_t * const bottom= src[4]; -#define OBMC_FILTER(x, t, l, m, r, b)\ - dst[x]= (t*top[x] + l*left[x] + m*mid[x] + r*right[x] + b*bottom[x] + 4)>>3 -#define OBMC_FILTER4(x, t, l, m, r, b)\ - OBMC_FILTER(x , t, l, m, r, b);\ - OBMC_FILTER(x+1 , t, l, m, r, b);\ - OBMC_FILTER(x +stride, t, l, m, r, b);\ - OBMC_FILTER(x+1+stride, t, l, m, r, b); - - x=0; - OBMC_FILTER (x , 2, 2, 4, 0, 0); - OBMC_FILTER (x+1, 2, 1, 5, 0, 0); - OBMC_FILTER4(x+2, 2, 1, 5, 0, 0); - OBMC_FILTER4(x+4, 2, 0, 5, 1, 0); - OBMC_FILTER (x+6, 2, 0, 5, 1, 0); - OBMC_FILTER (x+7, 2, 0, 4, 2, 0); - x+= stride; - OBMC_FILTER (x , 1, 2, 5, 0, 0); - OBMC_FILTER (x+1, 1, 2, 5, 0, 0); - OBMC_FILTER (x+6, 1, 0, 5, 2, 0); - OBMC_FILTER (x+7, 1, 0, 5, 2, 0); - x+= stride; - OBMC_FILTER4(x , 1, 2, 5, 0, 0); - OBMC_FILTER4(x+2, 1, 1, 6, 0, 0); - OBMC_FILTER4(x+4, 1, 0, 6, 1, 0); - OBMC_FILTER4(x+6, 1, 0, 5, 2, 0); - x+= 2*stride; - OBMC_FILTER4(x , 0, 2, 5, 0, 1); - OBMC_FILTER4(x+2, 0, 1, 6, 0, 1); - OBMC_FILTER4(x+4, 0, 0, 6, 1, 1); - OBMC_FILTER4(x+6, 0, 0, 5, 2, 1); - x+= 2*stride; - OBMC_FILTER (x , 0, 2, 5, 0, 1); - OBMC_FILTER (x+1, 0, 2, 5, 0, 1); - OBMC_FILTER4(x+2, 0, 1, 5, 0, 2); - OBMC_FILTER4(x+4, 0, 0, 5, 1, 2); - OBMC_FILTER (x+6, 0, 0, 5, 2, 1); - OBMC_FILTER (x+7, 0, 0, 5, 2, 1); - x+= stride; - OBMC_FILTER (x , 0, 2, 4, 0, 2); - OBMC_FILTER (x+1, 0, 1, 5, 0, 2); - OBMC_FILTER (x+6, 0, 0, 5, 1, 2); - OBMC_FILTER (x+7, 0, 0, 4, 2, 2); -} - -/* obmc for 1 8x8 luma block */ -static inline void obmc_motion(MpegEncContext *s, - uint8_t *dest, uint8_t *src, - int src_x, int src_y, - op_pixels_func *pix_op, - int16_t mv[5][2]/* mid top left right bottom*/) -#define MID 0 -{ - int i; - uint8_t *ptr[5]; - - assert(s->quarter_sample==0); - - for(i=0; i<5; i++){ - if(i && mv[i][0]==mv[MID][0] && mv[i][1]==mv[MID][1]){ - ptr[i]= ptr[MID]; - }else{ - ptr[i]= s->obmc_scratchpad + 8*(i&1) + s->linesize*8*(i>>1); - hpel_motion(s, ptr[i], src, 0, 0, - src_x, src_y, - s->width, s->height, s->linesize, - s->h_edge_pos, s->v_edge_pos, - 8, 8, pix_op, - mv[i][0], mv[i][1]); - } - } - - put_obmc(dest, ptr, s->linesize); -} - -static inline void qpel_motion(MpegEncContext *s, - uint8_t *dest_y, uint8_t *dest_cb, uint8_t *dest_cr, - int field_based, int bottom_field, int field_select, - uint8_t **ref_picture, op_pixels_func (*pix_op)[4], - qpel_mc_func (*qpix_op)[16], - int motion_x, int motion_y, int h) -{ - uint8_t *ptr_y, *ptr_cb, *ptr_cr; - int dxy, uvdxy, mx, my, src_x, src_y, uvsrc_x, uvsrc_y, v_edge_pos, linesize, uvlinesize; - - dxy = ((motion_y & 3) << 2) | (motion_x & 3); - src_x = s->mb_x * 16 + (motion_x >> 2); - src_y = s->mb_y * (16 >> field_based) + (motion_y >> 2); - - v_edge_pos = s->v_edge_pos >> field_based; - linesize = s->linesize << field_based; - uvlinesize = s->uvlinesize << field_based; - - if(field_based){ - mx= motion_x/2; - my= motion_y>>1; - }else if(s->workaround_bugs&FF_BUG_QPEL_CHROMA2){ - static const int rtab[8]= {0,0,1,1,0,0,0,1}; - mx= (motion_x>>1) + rtab[motion_x&7]; - my= (motion_y>>1) + rtab[motion_y&7]; - }else if(s->workaround_bugs&FF_BUG_QPEL_CHROMA){ - mx= (motion_x>>1)|(motion_x&1); - my= (motion_y>>1)|(motion_y&1); - }else{ - mx= motion_x/2; - my= motion_y/2; - } - mx= (mx>>1)|(mx&1); - my= (my>>1)|(my&1); - - uvdxy= (mx&1) | ((my&1)<<1); - mx>>=1; - my>>=1; - - uvsrc_x = s->mb_x * 8 + mx; - uvsrc_y = s->mb_y * (8 >> field_based) + my; - - ptr_y = ref_picture[0] + src_y * linesize + src_x; - ptr_cb = ref_picture[1] + uvsrc_y * uvlinesize + uvsrc_x; - ptr_cr = ref_picture[2] + uvsrc_y * uvlinesize + uvsrc_x; - - if( (unsigned)src_x > s->h_edge_pos - (motion_x&3) - 16 - || (unsigned)src_y > v_edge_pos - (motion_y&3) - h ){ - ff_emulated_edge_mc(s->edge_emu_buffer, ptr_y, s->linesize, 17, 17+field_based, - src_x, src_y<h_edge_pos, s->v_edge_pos); - ptr_y= s->edge_emu_buffer; - if(!(s->flags&CODEC_FLAG_GRAY)){ - uint8_t *uvbuf= s->edge_emu_buffer + 18*s->linesize; - ff_emulated_edge_mc(uvbuf, ptr_cb, s->uvlinesize, 9, 9 + field_based, - uvsrc_x, uvsrc_y<h_edge_pos>>1, s->v_edge_pos>>1); - ff_emulated_edge_mc(uvbuf + 16, ptr_cr, s->uvlinesize, 9, 9 + field_based, - uvsrc_x, uvsrc_y<h_edge_pos>>1, s->v_edge_pos>>1); - ptr_cb= uvbuf; - ptr_cr= uvbuf + 16; - } - } - - if(!field_based) - qpix_op[0][dxy](dest_y, ptr_y, linesize); - else{ - if(bottom_field){ - dest_y += s->linesize; - dest_cb+= s->uvlinesize; - dest_cr+= s->uvlinesize; - } - - if(field_select){ - ptr_y += s->linesize; - ptr_cb += s->uvlinesize; - ptr_cr += s->uvlinesize; - } - //damn interlaced mode - //FIXME boundary mirroring is not exactly correct here - qpix_op[1][dxy](dest_y , ptr_y , linesize); - qpix_op[1][dxy](dest_y+8, ptr_y+8, linesize); - } - if(!(s->flags&CODEC_FLAG_GRAY)){ - pix_op[1][uvdxy](dest_cr, ptr_cr, uvlinesize, h >> 1); - pix_op[1][uvdxy](dest_cb, ptr_cb, uvlinesize, h >> 1); - } -} - inline int ff_h263_round_chroma(int x){ if (x >= 0) return (h263_chroma_roundtab[x & 0xf] + ((x >> 3) & ~1)); @@ -3336,55 +1561,6 @@ inline int ff_h263_round_chroma(int x){ } } -/** - * h263 chorma 4mv motion compensation. - */ -static inline void chroma_4mv_motion(MpegEncContext *s, - uint8_t *dest_cb, uint8_t *dest_cr, - uint8_t **ref_picture, - op_pixels_func *pix_op, - int mx, int my){ - int dxy, emu=0, src_x, src_y, offset; - uint8_t *ptr; - - /* In case of 8X8, we construct a single chroma motion vector - with a special rounding */ - mx= ff_h263_round_chroma(mx); - my= ff_h263_round_chroma(my); - - dxy = ((my & 1) << 1) | (mx & 1); - mx >>= 1; - my >>= 1; - - src_x = s->mb_x * 8 + mx; - src_y = s->mb_y * 8 + my; - src_x = av_clip(src_x, -8, s->width/2); - if (src_x == s->width/2) - dxy &= ~1; - src_y = av_clip(src_y, -8, s->height/2); - if (src_y == s->height/2) - dxy &= ~2; - - offset = (src_y * (s->uvlinesize)) + src_x; - ptr = ref_picture[1] + offset; - if(s->flags&CODEC_FLAG_EMU_EDGE){ - if( (unsigned)src_x > (s->h_edge_pos>>1) - (dxy &1) - 8 - || (unsigned)src_y > (s->v_edge_pos>>1) - (dxy>>1) - 8){ - ff_emulated_edge_mc(s->edge_emu_buffer, ptr, s->uvlinesize, 9, 9, src_x, src_y, s->h_edge_pos>>1, s->v_edge_pos>>1); - ptr= s->edge_emu_buffer; - emu=1; - } - } - pix_op[dxy](dest_cb, ptr, s->uvlinesize, 8); - - ptr = ref_picture[2] + offset; - if(emu){ - ff_emulated_edge_mc(s->edge_emu_buffer, ptr, s->uvlinesize, 9, 9, src_x, src_y, s->h_edge_pos>>1, s->v_edge_pos>>1); - ptr= s->edge_emu_buffer; - } - pix_op[dxy](dest_cr, ptr, s->uvlinesize, 8); -} - static inline void chroma_4mv_motion_lowres(MpegEncContext *s, uint8_t *dest_cb, uint8_t *dest_cr, uint8_t **ref_picture, @@ -3435,269 +1611,6 @@ static inline void chroma_4mv_motion_lowres(MpegEncContext *s, pix_op[lowres](dest_cr, ptr, s->uvlinesize, block_s, sx, sy); } -static inline void prefetch_motion(MpegEncContext *s, uint8_t **pix, int dir){ - /* fetch pixels for estimated mv 4 macroblocks ahead - * optimized for 64byte cache lines */ - const int shift = s->quarter_sample ? 2 : 1; - const int mx= (s->mv[dir][0][0]>>shift) + 16*s->mb_x + 8; - const int my= (s->mv[dir][0][1]>>shift) + 16*s->mb_y; - int off= mx + (my + (s->mb_x&3)*4)*s->linesize + 64; - s->dsp.prefetch(pix[0]+off, s->linesize, 4); - off= (mx>>1) + ((my>>1) + (s->mb_x&7))*s->uvlinesize + 64; - s->dsp.prefetch(pix[1]+off, pix[2]-pix[1], 2); -} - -/** - * motion compensation of a single macroblock - * @param s context - * @param dest_y luma destination pointer - * @param dest_cb chroma cb/u destination pointer - * @param dest_cr chroma cr/v destination pointer - * @param dir direction (0->forward, 1->backward) - * @param ref_picture array[3] of pointers to the 3 planes of the reference picture - * @param pic_op halfpel motion compensation function (average or put normally) - * @param pic_op qpel motion compensation function (average or put normally) - * the motion vectors are taken from s->mv and the MV type from s->mv_type - */ -static inline void MPV_motion(MpegEncContext *s, - uint8_t *dest_y, uint8_t *dest_cb, uint8_t *dest_cr, - int dir, uint8_t **ref_picture, - op_pixels_func (*pix_op)[4], qpel_mc_func (*qpix_op)[16]) -{ - int dxy, mx, my, src_x, src_y, motion_x, motion_y; - int mb_x, mb_y, i; - uint8_t *ptr, *dest; - - mb_x = s->mb_x; - mb_y = s->mb_y; - - prefetch_motion(s, ref_picture, dir); - - if(s->obmc && s->pict_type != B_TYPE){ - int16_t mv_cache[4][4][2]; - const int xy= s->mb_x + s->mb_y*s->mb_stride; - const int mot_stride= s->b8_stride; - const int mot_xy= mb_x*2 + mb_y*2*mot_stride; - - assert(!s->mb_skipped); - - memcpy(mv_cache[1][1], s->current_picture.motion_val[0][mot_xy ], sizeof(int16_t)*4); - memcpy(mv_cache[2][1], s->current_picture.motion_val[0][mot_xy+mot_stride], sizeof(int16_t)*4); - memcpy(mv_cache[3][1], s->current_picture.motion_val[0][mot_xy+mot_stride], sizeof(int16_t)*4); - - if(mb_y==0 || IS_INTRA(s->current_picture.mb_type[xy-s->mb_stride])){ - memcpy(mv_cache[0][1], mv_cache[1][1], sizeof(int16_t)*4); - }else{ - memcpy(mv_cache[0][1], s->current_picture.motion_val[0][mot_xy-mot_stride], sizeof(int16_t)*4); - } - - if(mb_x==0 || IS_INTRA(s->current_picture.mb_type[xy-1])){ - *(int32_t*)mv_cache[1][0]= *(int32_t*)mv_cache[1][1]; - *(int32_t*)mv_cache[2][0]= *(int32_t*)mv_cache[2][1]; - }else{ - *(int32_t*)mv_cache[1][0]= *(int32_t*)s->current_picture.motion_val[0][mot_xy-1]; - *(int32_t*)mv_cache[2][0]= *(int32_t*)s->current_picture.motion_val[0][mot_xy-1+mot_stride]; - } - - if(mb_x+1>=s->mb_width || IS_INTRA(s->current_picture.mb_type[xy+1])){ - *(int32_t*)mv_cache[1][3]= *(int32_t*)mv_cache[1][2]; - *(int32_t*)mv_cache[2][3]= *(int32_t*)mv_cache[2][2]; - }else{ - *(int32_t*)mv_cache[1][3]= *(int32_t*)s->current_picture.motion_val[0][mot_xy+2]; - *(int32_t*)mv_cache[2][3]= *(int32_t*)s->current_picture.motion_val[0][mot_xy+2+mot_stride]; - } - - mx = 0; - my = 0; - for(i=0;i<4;i++) { - const int x= (i&1)+1; - const int y= (i>>1)+1; - int16_t mv[5][2]= { - {mv_cache[y][x ][0], mv_cache[y][x ][1]}, - {mv_cache[y-1][x][0], mv_cache[y-1][x][1]}, - {mv_cache[y][x-1][0], mv_cache[y][x-1][1]}, - {mv_cache[y][x+1][0], mv_cache[y][x+1][1]}, - {mv_cache[y+1][x][0], mv_cache[y+1][x][1]}}; - //FIXME cleanup - obmc_motion(s, dest_y + ((i & 1) * 8) + (i >> 1) * 8 * s->linesize, - ref_picture[0], - mb_x * 16 + (i & 1) * 8, mb_y * 16 + (i >>1) * 8, - pix_op[1], - mv); - - mx += mv[0][0]; - my += mv[0][1]; - } - if(!(s->flags&CODEC_FLAG_GRAY)) - chroma_4mv_motion(s, dest_cb, dest_cr, ref_picture, pix_op[1], mx, my); - - return; - } - - switch(s->mv_type) { - case MV_TYPE_16X16: - if(s->mcsel){ - if(s->real_sprite_warping_points==1){ - gmc1_motion(s, dest_y, dest_cb, dest_cr, - ref_picture); - }else{ - gmc_motion(s, dest_y, dest_cb, dest_cr, - ref_picture); - } - }else if(s->quarter_sample){ - qpel_motion(s, dest_y, dest_cb, dest_cr, - 0, 0, 0, - ref_picture, pix_op, qpix_op, - s->mv[dir][0][0], s->mv[dir][0][1], 16); - }else if(ENABLE_WMV2 && s->mspel){ - ff_mspel_motion(s, dest_y, dest_cb, dest_cr, - ref_picture, pix_op, - s->mv[dir][0][0], s->mv[dir][0][1], 16); - }else - { - mpeg_motion(s, dest_y, dest_cb, dest_cr, - 0, 0, 0, - ref_picture, pix_op, - s->mv[dir][0][0], s->mv[dir][0][1], 16); - } - break; - case MV_TYPE_8X8: - mx = 0; - my = 0; - if(s->quarter_sample){ - for(i=0;i<4;i++) { - motion_x = s->mv[dir][i][0]; - motion_y = s->mv[dir][i][1]; - - dxy = ((motion_y & 3) << 2) | (motion_x & 3); - src_x = mb_x * 16 + (motion_x >> 2) + (i & 1) * 8; - src_y = mb_y * 16 + (motion_y >> 2) + (i >>1) * 8; - - /* WARNING: do no forget half pels */ - src_x = av_clip(src_x, -16, s->width); - if (src_x == s->width) - dxy &= ~3; - src_y = av_clip(src_y, -16, s->height); - if (src_y == s->height) - dxy &= ~12; - - ptr = ref_picture[0] + (src_y * s->linesize) + (src_x); - if(s->flags&CODEC_FLAG_EMU_EDGE){ - if( (unsigned)src_x > s->h_edge_pos - (motion_x&3) - 8 - || (unsigned)src_y > s->v_edge_pos - (motion_y&3) - 8 ){ - ff_emulated_edge_mc(s->edge_emu_buffer, ptr, s->linesize, 9, 9, src_x, src_y, s->h_edge_pos, s->v_edge_pos); - ptr= s->edge_emu_buffer; - } - } - dest = dest_y + ((i & 1) * 8) + (i >> 1) * 8 * s->linesize; - qpix_op[1][dxy](dest, ptr, s->linesize); - - mx += s->mv[dir][i][0]/2; - my += s->mv[dir][i][1]/2; - } - }else{ - for(i=0;i<4;i++) { - hpel_motion(s, dest_y + ((i & 1) * 8) + (i >> 1) * 8 * s->linesize, - ref_picture[0], 0, 0, - mb_x * 16 + (i & 1) * 8, mb_y * 16 + (i >>1) * 8, - s->width, s->height, s->linesize, - s->h_edge_pos, s->v_edge_pos, - 8, 8, pix_op[1], - s->mv[dir][i][0], s->mv[dir][i][1]); - - mx += s->mv[dir][i][0]; - my += s->mv[dir][i][1]; - } - } - - if(!(s->flags&CODEC_FLAG_GRAY)) - chroma_4mv_motion(s, dest_cb, dest_cr, ref_picture, pix_op[1], mx, my); - break; - case MV_TYPE_FIELD: - if (s->picture_structure == PICT_FRAME) { - if(s->quarter_sample){ - for(i=0; i<2; i++){ - qpel_motion(s, dest_y, dest_cb, dest_cr, - 1, i, s->field_select[dir][i], - ref_picture, pix_op, qpix_op, - s->mv[dir][i][0], s->mv[dir][i][1], 8); - } - }else{ - /* top field */ - mpeg_motion(s, dest_y, dest_cb, dest_cr, - 1, 0, s->field_select[dir][0], - ref_picture, pix_op, - s->mv[dir][0][0], s->mv[dir][0][1], 8); - /* bottom field */ - mpeg_motion(s, dest_y, dest_cb, dest_cr, - 1, 1, s->field_select[dir][1], - ref_picture, pix_op, - s->mv[dir][1][0], s->mv[dir][1][1], 8); - } - } else { - if(s->picture_structure != s->field_select[dir][0] + 1 && s->pict_type != B_TYPE && !s->first_field){ - ref_picture= s->current_picture_ptr->data; - } - - mpeg_motion(s, dest_y, dest_cb, dest_cr, - 0, 0, s->field_select[dir][0], - ref_picture, pix_op, - s->mv[dir][0][0], s->mv[dir][0][1], 16); - } - break; - case MV_TYPE_16X8: - for(i=0; i<2; i++){ - uint8_t ** ref2picture; - - if(s->picture_structure == s->field_select[dir][i] + 1 || s->pict_type == B_TYPE || s->first_field){ - ref2picture= ref_picture; - }else{ - ref2picture= s->current_picture_ptr->data; - } - - mpeg_motion(s, dest_y, dest_cb, dest_cr, - 0, 0, s->field_select[dir][i], - ref2picture, pix_op, - s->mv[dir][i][0], s->mv[dir][i][1] + 16*i, 8); - - dest_y += 16*s->linesize; - dest_cb+= (16>>s->chroma_y_shift)*s->uvlinesize; - dest_cr+= (16>>s->chroma_y_shift)*s->uvlinesize; - } - break; - case MV_TYPE_DMV: - if(s->picture_structure == PICT_FRAME){ - for(i=0; i<2; i++){ - int j; - for(j=0; j<2; j++){ - mpeg_motion(s, dest_y, dest_cb, dest_cr, - 1, j, j^i, - ref_picture, pix_op, - s->mv[dir][2*i + j][0], s->mv[dir][2*i + j][1], 8); - } - pix_op = s->dsp.avg_pixels_tab; - } - }else{ - for(i=0; i<2; i++){ - mpeg_motion(s, dest_y, dest_cb, dest_cr, - 0, 0, s->picture_structure != i+1, - ref_picture, pix_op, - s->mv[dir][2*i][0],s->mv[dir][2*i][1],16); - - // after put we make avg of the same block - pix_op=s->dsp.avg_pixels_tab; - - //opposite parity is always in the same frame if this is second field - if(!s->first_field){ - ref_picture = s->current_picture_ptr->data; - } - } - } - break; - default: assert(0); - } -} - /** * motion compensation of a single macroblock * @param s context @@ -4140,90 +2053,6 @@ void MPV_decode_mb(MpegEncContext *s, DCTELEM block[12][64]){ else MPV_decode_mb_internal(s, block, 0); } -#ifdef CONFIG_ENCODERS - -static inline void dct_single_coeff_elimination(MpegEncContext *s, int n, int threshold) -{ - static const char tab[64]= - {3,2,2,1,1,1,1,1, - 1,1,1,1,1,1,1,1, - 1,1,1,1,1,1,1,1, - 0,0,0,0,0,0,0,0, - 0,0,0,0,0,0,0,0, - 0,0,0,0,0,0,0,0, - 0,0,0,0,0,0,0,0, - 0,0,0,0,0,0,0,0}; - int score=0; - int run=0; - int i; - DCTELEM *block= s->block[n]; - const int last_index= s->block_last_index[n]; - int skip_dc; - - if(threshold<0){ - skip_dc=0; - threshold= -threshold; - }else - skip_dc=1; - - /* are all which we could set to zero are allready zero? */ - if(last_index<=skip_dc - 1) return; - - for(i=0; i<=last_index; i++){ - const int j = s->intra_scantable.permutated[i]; - const int level = FFABS(block[j]); - if(level==1){ - if(skip_dc && i==0) continue; - score+= tab[run]; - run=0; - }else if(level>1){ - return; - }else{ - run++; - } - } - if(score >= threshold) return; - for(i=skip_dc; i<=last_index; i++){ - const int j = s->intra_scantable.permutated[i]; - block[j]=0; - } - if(block[0]) s->block_last_index[n]= 0; - else s->block_last_index[n]= -1; -} - -static inline void clip_coeffs(MpegEncContext *s, DCTELEM *block, int last_index) -{ - int i; - const int maxlevel= s->max_qcoeff; - const int minlevel= s->min_qcoeff; - int overflow=0; - - if(s->mb_intra){ - i=1; //skip clipping of intra dc - }else - i=0; - - for(;i<=last_index; i++){ - const int j= s->intra_scantable.permutated[i]; - int level = block[j]; - - if (level>maxlevel){ - level=maxlevel; - overflow++; - }else if(levelavctx->mb_decision == FF_MB_DECISION_SIMPLE) - av_log(s->avctx, AV_LOG_INFO, "warning, clipping %d dct coefficients to %d..%d\n", overflow, minlevel, maxlevel); -} - -#endif //CONFIG_ENCODERS - /** * * @param h is the normal height, this will be reduced automatically if needed for the last row @@ -4292,326 +2121,6 @@ void ff_init_block_index(MpegEncContext *s){ //FIXME maybe rename } } -#ifdef CONFIG_ENCODERS - -static void get_vissual_weight(int16_t *weight, uint8_t *ptr, int stride){ - int x, y; -//FIXME optimize - for(y=0; y<8; y++){ - for(x=0; x<8; x++){ - int x2, y2; - int sum=0; - int sqr=0; - int count=0; - - for(y2= FFMAX(y-1, 0); y2 < FFMIN(8, y+2); y2++){ - for(x2= FFMAX(x-1, 0); x2 < FFMIN(8, x+2); x2++){ - int v= ptr[x2 + y2*stride]; - sum += v; - sqr += v*v; - count++; - } - } - weight[x + 8*y]= (36*ff_sqrt(count*sqr - sum*sum)) / count; - } - } -} - -static av_always_inline void encode_mb_internal(MpegEncContext *s, int motion_x, int motion_y, int mb_block_height, int mb_block_count) -{ - int16_t weight[8][64]; - DCTELEM orig[8][64]; - const int mb_x= s->mb_x; - const int mb_y= s->mb_y; - int i; - int skip_dct[8]; - int dct_offset = s->linesize*8; //default for progressive frames - uint8_t *ptr_y, *ptr_cb, *ptr_cr; - int wrap_y, wrap_c; - - for(i=0; iskipdct; - - if(s->adaptive_quant){ - const int last_qp= s->qscale; - const int mb_xy= mb_x + mb_y*s->mb_stride; - - s->lambda= s->lambda_table[mb_xy]; - update_qscale(s); - - if(!(s->flags&CODEC_FLAG_QP_RD)){ - s->qscale= s->current_picture_ptr->qscale_table[mb_xy]; - s->dquant= s->qscale - last_qp; - - if(s->out_format==FMT_H263){ - s->dquant= av_clip(s->dquant, -2, 2); - - if(s->codec_id==CODEC_ID_MPEG4){ - if(!s->mb_intra){ - if(s->pict_type == B_TYPE){ - if(s->dquant&1 || s->mv_dir&MV_DIRECT) - s->dquant= 0; - } - if(s->mv_type==MV_TYPE_8X8) - s->dquant=0; - } - } - } - } - ff_set_qscale(s, last_qp + s->dquant); - }else if(s->flags&CODEC_FLAG_QP_RD) - ff_set_qscale(s, s->qscale + s->dquant); - - wrap_y = s->linesize; - wrap_c = s->uvlinesize; - ptr_y = s->new_picture.data[0] + (mb_y * 16 * wrap_y) + mb_x * 16; - ptr_cb = s->new_picture.data[1] + (mb_y * mb_block_height * wrap_c) + mb_x * 8; - ptr_cr = s->new_picture.data[2] + (mb_y * mb_block_height * wrap_c) + mb_x * 8; - - if(mb_x*16+16 > s->width || mb_y*16+16 > s->height){ - uint8_t *ebuf= s->edge_emu_buffer + 32; - ff_emulated_edge_mc(ebuf , ptr_y , wrap_y,16,16,mb_x*16,mb_y*16, s->width , s->height); - ptr_y= ebuf; - ff_emulated_edge_mc(ebuf+18*wrap_y , ptr_cb, wrap_c, 8, mb_block_height, mb_x*8, mb_y*8, s->width>>1, s->height>>1); - ptr_cb= ebuf+18*wrap_y; - ff_emulated_edge_mc(ebuf+18*wrap_y+8, ptr_cr, wrap_c, 8, mb_block_height, mb_x*8, mb_y*8, s->width>>1, s->height>>1); - ptr_cr= ebuf+18*wrap_y+8; - } - - if (s->mb_intra) { - if(s->flags&CODEC_FLAG_INTERLACED_DCT){ - int progressive_score, interlaced_score; - - s->interlaced_dct=0; - progressive_score= s->dsp.ildct_cmp[4](s, ptr_y , NULL, wrap_y, 8) - +s->dsp.ildct_cmp[4](s, ptr_y + wrap_y*8, NULL, wrap_y, 8) - 400; - - if(progressive_score > 0){ - interlaced_score = s->dsp.ildct_cmp[4](s, ptr_y , NULL, wrap_y*2, 8) - +s->dsp.ildct_cmp[4](s, ptr_y + wrap_y , NULL, wrap_y*2, 8); - if(progressive_score > interlaced_score){ - s->interlaced_dct=1; - - dct_offset= wrap_y; - wrap_y<<=1; - if (s->chroma_format == CHROMA_422) - wrap_c<<=1; - } - } - } - - s->dsp.get_pixels(s->block[0], ptr_y , wrap_y); - s->dsp.get_pixels(s->block[1], ptr_y + 8, wrap_y); - s->dsp.get_pixels(s->block[2], ptr_y + dct_offset , wrap_y); - s->dsp.get_pixels(s->block[3], ptr_y + dct_offset + 8, wrap_y); - - if(s->flags&CODEC_FLAG_GRAY){ - skip_dct[4]= 1; - skip_dct[5]= 1; - }else{ - s->dsp.get_pixels(s->block[4], ptr_cb, wrap_c); - s->dsp.get_pixels(s->block[5], ptr_cr, wrap_c); - if(!s->chroma_y_shift){ /* 422 */ - s->dsp.get_pixels(s->block[6], ptr_cb + (dct_offset>>1), wrap_c); - s->dsp.get_pixels(s->block[7], ptr_cr + (dct_offset>>1), wrap_c); - } - } - }else{ - op_pixels_func (*op_pix)[4]; - qpel_mc_func (*op_qpix)[16]; - uint8_t *dest_y, *dest_cb, *dest_cr; - - dest_y = s->dest[0]; - dest_cb = s->dest[1]; - dest_cr = s->dest[2]; - - if ((!s->no_rounding) || s->pict_type==B_TYPE){ - op_pix = s->dsp.put_pixels_tab; - op_qpix= s->dsp.put_qpel_pixels_tab; - }else{ - op_pix = s->dsp.put_no_rnd_pixels_tab; - op_qpix= s->dsp.put_no_rnd_qpel_pixels_tab; - } - - if (s->mv_dir & MV_DIR_FORWARD) { - MPV_motion(s, dest_y, dest_cb, dest_cr, 0, s->last_picture.data, op_pix, op_qpix); - op_pix = s->dsp.avg_pixels_tab; - op_qpix= s->dsp.avg_qpel_pixels_tab; - } - if (s->mv_dir & MV_DIR_BACKWARD) { - MPV_motion(s, dest_y, dest_cb, dest_cr, 1, s->next_picture.data, op_pix, op_qpix); - } - - if(s->flags&CODEC_FLAG_INTERLACED_DCT){ - int progressive_score, interlaced_score; - - s->interlaced_dct=0; - progressive_score= s->dsp.ildct_cmp[0](s, dest_y , ptr_y , wrap_y, 8) - +s->dsp.ildct_cmp[0](s, dest_y + wrap_y*8, ptr_y + wrap_y*8, wrap_y, 8) - 400; - - if(s->avctx->ildct_cmp == FF_CMP_VSSE) progressive_score -= 400; - - if(progressive_score>0){ - interlaced_score = s->dsp.ildct_cmp[0](s, dest_y , ptr_y , wrap_y*2, 8) - +s->dsp.ildct_cmp[0](s, dest_y + wrap_y , ptr_y + wrap_y , wrap_y*2, 8); - - if(progressive_score > interlaced_score){ - s->interlaced_dct=1; - - dct_offset= wrap_y; - wrap_y<<=1; - if (s->chroma_format == CHROMA_422) - wrap_c<<=1; - } - } - } - - s->dsp.diff_pixels(s->block[0], ptr_y , dest_y , wrap_y); - s->dsp.diff_pixels(s->block[1], ptr_y + 8, dest_y + 8, wrap_y); - s->dsp.diff_pixels(s->block[2], ptr_y + dct_offset , dest_y + dct_offset , wrap_y); - s->dsp.diff_pixels(s->block[3], ptr_y + dct_offset + 8, dest_y + dct_offset + 8, wrap_y); - - if(s->flags&CODEC_FLAG_GRAY){ - skip_dct[4]= 1; - skip_dct[5]= 1; - }else{ - s->dsp.diff_pixels(s->block[4], ptr_cb, dest_cb, wrap_c); - s->dsp.diff_pixels(s->block[5], ptr_cr, dest_cr, wrap_c); - if(!s->chroma_y_shift){ /* 422 */ - s->dsp.diff_pixels(s->block[6], ptr_cb + (dct_offset>>1), dest_cb + (dct_offset>>1), wrap_c); - s->dsp.diff_pixels(s->block[7], ptr_cr + (dct_offset>>1), dest_cr + (dct_offset>>1), wrap_c); - } - } - /* pre quantization */ - if(s->current_picture.mc_mb_var[s->mb_stride*mb_y+ mb_x]<2*s->qscale*s->qscale){ - //FIXME optimize - if(s->dsp.sad[1](NULL, ptr_y , dest_y , wrap_y, 8) < 20*s->qscale) skip_dct[0]= 1; - if(s->dsp.sad[1](NULL, ptr_y + 8, dest_y + 8, wrap_y, 8) < 20*s->qscale) skip_dct[1]= 1; - if(s->dsp.sad[1](NULL, ptr_y +dct_offset , dest_y +dct_offset , wrap_y, 8) < 20*s->qscale) skip_dct[2]= 1; - if(s->dsp.sad[1](NULL, ptr_y +dct_offset+ 8, dest_y +dct_offset+ 8, wrap_y, 8) < 20*s->qscale) skip_dct[3]= 1; - if(s->dsp.sad[1](NULL, ptr_cb , dest_cb , wrap_c, 8) < 20*s->qscale) skip_dct[4]= 1; - if(s->dsp.sad[1](NULL, ptr_cr , dest_cr , wrap_c, 8) < 20*s->qscale) skip_dct[5]= 1; - if(!s->chroma_y_shift){ /* 422 */ - if(s->dsp.sad[1](NULL, ptr_cb +(dct_offset>>1), dest_cb +(dct_offset>>1), wrap_c, 8) < 20*s->qscale) skip_dct[6]= 1; - if(s->dsp.sad[1](NULL, ptr_cr +(dct_offset>>1), dest_cr +(dct_offset>>1), wrap_c, 8) < 20*s->qscale) skip_dct[7]= 1; - } - } - } - - if(s->avctx->quantizer_noise_shaping){ - if(!skip_dct[0]) get_vissual_weight(weight[0], ptr_y , wrap_y); - if(!skip_dct[1]) get_vissual_weight(weight[1], ptr_y + 8, wrap_y); - if(!skip_dct[2]) get_vissual_weight(weight[2], ptr_y + dct_offset , wrap_y); - if(!skip_dct[3]) get_vissual_weight(weight[3], ptr_y + dct_offset + 8, wrap_y); - if(!skip_dct[4]) get_vissual_weight(weight[4], ptr_cb , wrap_c); - if(!skip_dct[5]) get_vissual_weight(weight[5], ptr_cr , wrap_c); - if(!s->chroma_y_shift){ /* 422 */ - if(!skip_dct[6]) get_vissual_weight(weight[6], ptr_cb + (dct_offset>>1), wrap_c); - if(!skip_dct[7]) get_vissual_weight(weight[7], ptr_cr + (dct_offset>>1), wrap_c); - } - memcpy(orig[0], s->block[0], sizeof(DCTELEM)*64*mb_block_count); - } - - /* DCT & quantize */ - assert(s->out_format!=FMT_MJPEG || s->qscale==8); - { - for(i=0;iblock_last_index[i] = s->dct_quantize(s, s->block[i], i, s->qscale, &overflow); - // FIXME we could decide to change to quantizer instead of clipping - // JS: I don't think that would be a good idea it could lower quality instead - // of improve it. Just INTRADC clipping deserves changes in quantizer - if (overflow) clip_coeffs(s, s->block[i], s->block_last_index[i]); - }else - s->block_last_index[i]= -1; - } - if(s->avctx->quantizer_noise_shaping){ - for(i=0;iblock_last_index[i] = dct_quantize_refine(s, s->block[i], weight[i], orig[i], i, s->qscale); - } - } - } - - if(s->luma_elim_threshold && !s->mb_intra) - for(i=0; i<4; i++) - dct_single_coeff_elimination(s, i, s->luma_elim_threshold); - if(s->chroma_elim_threshold && !s->mb_intra) - for(i=4; ichroma_elim_threshold); - - if(s->flags & CODEC_FLAG_CBP_RD){ - for(i=0;iblock_last_index[i] == -1) - s->coded_score[i]= INT_MAX/256; - } - } - } - - if((s->flags&CODEC_FLAG_GRAY) && s->mb_intra){ - s->block_last_index[4]= - s->block_last_index[5]= 0; - s->block[4][0]= - s->block[5][0]= (1024 + s->c_dc_scale/2)/ s->c_dc_scale; - } - - //non c quantize code returns incorrect block_last_index FIXME - if(s->alternate_scan && s->dct_quantize != dct_quantize_c){ - for(i=0; iblock_last_index[i]>0){ - for(j=63; j>0; j--){ - if(s->block[i][ s->intra_scantable.permutated[j] ]) break; - } - s->block_last_index[i]= j; - } - } - } - - /* huffman encode */ - switch(s->codec_id){ //FIXME funct ptr could be slightly faster - case CODEC_ID_MPEG1VIDEO: - case CODEC_ID_MPEG2VIDEO: - mpeg1_encode_mb(s, s->block, motion_x, motion_y); break; - case CODEC_ID_MPEG4: - mpeg4_encode_mb(s, s->block, motion_x, motion_y); break; - case CODEC_ID_MSMPEG4V2: - case CODEC_ID_MSMPEG4V3: - case CODEC_ID_WMV1: - if (ENABLE_MSMPEG4_ENCODER) - msmpeg4_encode_mb(s, s->block, motion_x, motion_y); - break; - case CODEC_ID_WMV2: - if (ENABLE_WMV2_ENCODER) - ff_wmv2_encode_mb(s, s->block, motion_x, motion_y); - break; - case CODEC_ID_H261: - if (ENABLE_H261_ENCODER) - ff_h261_encode_mb(s, s->block, motion_x, motion_y); - break; - case CODEC_ID_H263: - case CODEC_ID_H263P: - case CODEC_ID_FLV1: - case CODEC_ID_RV10: - case CODEC_ID_RV20: - h263_encode_mb(s, s->block, motion_x, motion_y); break; - case CODEC_ID_MJPEG: - if (ENABLE_MJPEG_ENCODER) - ff_mjpeg_encode_mb(s, s->block); - break; - default: - assert(0); - } -} - -static av_always_inline void encode_mb(MpegEncContext *s, int motion_x, int motion_y) -{ - if (s->chroma_format == CHROMA_420) encode_mb_internal(s, motion_x, motion_y, 8, 6); - else encode_mb_internal(s, motion_x, motion_y, 16, 8); -} - -#endif //CONFIG_ENCODERS - void ff_mpeg_flush(AVCodecContext *avctx){ int i; MpegEncContext *s = avctx->priv_data; @@ -4638,1970 +2147,6 @@ void ff_mpeg_flush(AVCodecContext *avctx){ s->pp_time=0; } -#ifdef CONFIG_ENCODERS -void ff_copy_bits(PutBitContext *pb, uint8_t *src, int length) -{ - const uint16_t *srcw= (uint16_t*)src; - int words= length>>4; - int bits= length&15; - int i; - - if(length==0) return; - - if(words < 16){ - for(i=0; i>(16-bits)); -} - -static inline void copy_context_before_encode(MpegEncContext *d, MpegEncContext *s, int type){ - int i; - - memcpy(d->last_mv, s->last_mv, 2*2*2*sizeof(int)); //FIXME is memcpy faster then a loop? - - /* mpeg1 */ - d->mb_skip_run= s->mb_skip_run; - for(i=0; i<3; i++) - d->last_dc[i]= s->last_dc[i]; - - /* statistics */ - d->mv_bits= s->mv_bits; - d->i_tex_bits= s->i_tex_bits; - d->p_tex_bits= s->p_tex_bits; - d->i_count= s->i_count; - d->f_count= s->f_count; - d->b_count= s->b_count; - d->skip_count= s->skip_count; - d->misc_bits= s->misc_bits; - d->last_bits= 0; - - d->mb_skipped= 0; - d->qscale= s->qscale; - d->dquant= s->dquant; -} - -static inline void copy_context_after_encode(MpegEncContext *d, MpegEncContext *s, int type){ - int i; - - memcpy(d->mv, s->mv, 2*4*2*sizeof(int)); - memcpy(d->last_mv, s->last_mv, 2*2*2*sizeof(int)); //FIXME is memcpy faster then a loop? - - /* mpeg1 */ - d->mb_skip_run= s->mb_skip_run; - for(i=0; i<3; i++) - d->last_dc[i]= s->last_dc[i]; - - /* statistics */ - d->mv_bits= s->mv_bits; - d->i_tex_bits= s->i_tex_bits; - d->p_tex_bits= s->p_tex_bits; - d->i_count= s->i_count; - d->f_count= s->f_count; - d->b_count= s->b_count; - d->skip_count= s->skip_count; - d->misc_bits= s->misc_bits; - - d->mb_intra= s->mb_intra; - d->mb_skipped= s->mb_skipped; - d->mv_type= s->mv_type; - d->mv_dir= s->mv_dir; - d->pb= s->pb; - if(s->data_partitioning){ - d->pb2= s->pb2; - d->tex_pb= s->tex_pb; - } - d->block= s->block; - for(i=0; i<8; i++) - d->block_last_index[i]= s->block_last_index[i]; - d->interlaced_dct= s->interlaced_dct; - d->qscale= s->qscale; -} - -static inline void encode_mb_hq(MpegEncContext *s, MpegEncContext *backup, MpegEncContext *best, int type, - PutBitContext pb[2], PutBitContext pb2[2], PutBitContext tex_pb[2], - int *dmin, int *next_block, int motion_x, int motion_y) -{ - int score; - uint8_t *dest_backup[3]; - - copy_context_before_encode(s, backup, type); - - s->block= s->blocks[*next_block]; - s->pb= pb[*next_block]; - if(s->data_partitioning){ - s->pb2 = pb2 [*next_block]; - s->tex_pb= tex_pb[*next_block]; - } - - if(*next_block){ - memcpy(dest_backup, s->dest, sizeof(s->dest)); - s->dest[0] = s->rd_scratchpad; - s->dest[1] = s->rd_scratchpad + 16*s->linesize; - s->dest[2] = s->rd_scratchpad + 16*s->linesize + 8; - assert(s->linesize >= 32); //FIXME - } - - encode_mb(s, motion_x, motion_y); - - score= put_bits_count(&s->pb); - if(s->data_partitioning){ - score+= put_bits_count(&s->pb2); - score+= put_bits_count(&s->tex_pb); - } - - if(s->avctx->mb_decision == FF_MB_DECISION_RD){ - MPV_decode_mb(s, s->block); - - score *= s->lambda2; - score += sse_mb(s) << FF_LAMBDA_SHIFT; - } - - if(*next_block){ - memcpy(s->dest, dest_backup, sizeof(s->dest)); - } - - if(score<*dmin){ - *dmin= score; - *next_block^=1; - - copy_context_after_encode(best, s, type); - } -} - -static int sse(MpegEncContext *s, uint8_t *src1, uint8_t *src2, int w, int h, int stride){ - uint32_t *sq = ff_squareTbl + 256; - int acc=0; - int x,y; - - if(w==16 && h==16) - return s->dsp.sse[0](NULL, src1, src2, stride, 16); - else if(w==8 && h==8) - return s->dsp.sse[1](NULL, src1, src2, stride, 8); - - for(y=0; y=0); - - return acc; -} - -static int sse_mb(MpegEncContext *s){ - int w= 16; - int h= 16; - - if(s->mb_x*16 + 16 > s->width ) w= s->width - s->mb_x*16; - if(s->mb_y*16 + 16 > s->height) h= s->height- s->mb_y*16; - - if(w==16 && h==16) - if(s->avctx->mb_cmp == FF_CMP_NSSE){ - return s->dsp.nsse[0](s, s->new_picture.data[0] + s->mb_x*16 + s->mb_y*s->linesize*16, s->dest[0], s->linesize, 16) - +s->dsp.nsse[1](s, s->new_picture.data[1] + s->mb_x*8 + s->mb_y*s->uvlinesize*8,s->dest[1], s->uvlinesize, 8) - +s->dsp.nsse[1](s, s->new_picture.data[2] + s->mb_x*8 + s->mb_y*s->uvlinesize*8,s->dest[2], s->uvlinesize, 8); - }else{ - return s->dsp.sse[0](NULL, s->new_picture.data[0] + s->mb_x*16 + s->mb_y*s->linesize*16, s->dest[0], s->linesize, 16) - +s->dsp.sse[1](NULL, s->new_picture.data[1] + s->mb_x*8 + s->mb_y*s->uvlinesize*8,s->dest[1], s->uvlinesize, 8) - +s->dsp.sse[1](NULL, s->new_picture.data[2] + s->mb_x*8 + s->mb_y*s->uvlinesize*8,s->dest[2], s->uvlinesize, 8); - } - else - return sse(s, s->new_picture.data[0] + s->mb_x*16 + s->mb_y*s->linesize*16, s->dest[0], w, h, s->linesize) - +sse(s, s->new_picture.data[1] + s->mb_x*8 + s->mb_y*s->uvlinesize*8,s->dest[1], w>>1, h>>1, s->uvlinesize) - +sse(s, s->new_picture.data[2] + s->mb_x*8 + s->mb_y*s->uvlinesize*8,s->dest[2], w>>1, h>>1, s->uvlinesize); -} - -static int pre_estimate_motion_thread(AVCodecContext *c, void *arg){ - MpegEncContext *s= arg; - - - s->me.pre_pass=1; - s->me.dia_size= s->avctx->pre_dia_size; - s->first_slice_line=1; - for(s->mb_y= s->end_mb_y-1; s->mb_y >= s->start_mb_y; s->mb_y--) { - for(s->mb_x=s->mb_width-1; s->mb_x >=0 ;s->mb_x--) { - ff_pre_estimate_p_frame_motion(s, s->mb_x, s->mb_y); - } - s->first_slice_line=0; - } - - s->me.pre_pass=0; - - return 0; -} - -static int estimate_motion_thread(AVCodecContext *c, void *arg){ - MpegEncContext *s= arg; - - ff_check_alignment(); - - s->me.dia_size= s->avctx->dia_size; - 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; //for block init below - ff_init_block_index(s); - for(s->mb_x=0; s->mb_x < s->mb_width; s->mb_x++) { - s->block_index[0]+=2; - s->block_index[1]+=2; - s->block_index[2]+=2; - s->block_index[3]+=2; - - /* compute motion vector & mb_type and store in context */ - if(s->pict_type==B_TYPE) - ff_estimate_b_frame_motion(s, s->mb_x, s->mb_y); - else - ff_estimate_p_frame_motion(s, s->mb_x, s->mb_y); - } - s->first_slice_line=0; - } - return 0; -} - -static int mb_var_thread(AVCodecContext *c, void *arg){ - MpegEncContext *s= arg; - int mb_x, mb_y; - - ff_check_alignment(); - - for(mb_y=s->start_mb_y; mb_y < s->end_mb_y; mb_y++) { - for(mb_x=0; mb_x < s->mb_width; mb_x++) { - int xx = mb_x * 16; - int yy = mb_y * 16; - uint8_t *pix = s->new_picture.data[0] + (yy * s->linesize) + xx; - int varc; - int sum = s->dsp.pix_sum(pix, s->linesize); - - varc = (s->dsp.pix_norm1(pix, s->linesize) - (((unsigned)(sum*sum))>>8) + 500 + 128)>>8; - - s->current_picture.mb_var [s->mb_stride * mb_y + mb_x] = varc; - s->current_picture.mb_mean[s->mb_stride * mb_y + mb_x] = (sum+128)>>8; - s->me.mb_var_sum_temp += varc; - } - } - return 0; -} - -static void write_slice_end(MpegEncContext *s){ - if(s->codec_id==CODEC_ID_MPEG4){ - if(s->partitioned_frame){ - ff_mpeg4_merge_partitions(s); - } - - ff_mpeg4_stuffing(&s->pb); - }else if(ENABLE_MJPEG_ENCODER && s->out_format == FMT_MJPEG){ - ff_mjpeg_encode_stuffing(&s->pb); - } - - align_put_bits(&s->pb); - flush_put_bits(&s->pb); - - if((s->flags&CODEC_FLAG_PASS1) && !s->partitioned_frame) - s->misc_bits+= get_bits_diff(s); -} - -static int encode_thread(AVCodecContext *c, void *arg){ - MpegEncContext *s= arg; - int mb_x, mb_y, pdif = 0; - int i, j; - MpegEncContext best_s, backup_s; - uint8_t bit_buf[2][MAX_MB_BYTES]; - uint8_t bit_buf2[2][MAX_MB_BYTES]; - uint8_t bit_buf_tex[2][MAX_MB_BYTES]; - PutBitContext pb[2], pb2[2], tex_pb[2]; -//printf("%d->%d\n", s->resync_mb_y, s->end_mb_y); - - ff_check_alignment(); - - for(i=0; i<2; i++){ - init_put_bits(&pb [i], bit_buf [i], MAX_MB_BYTES); - init_put_bits(&pb2 [i], bit_buf2 [i], MAX_MB_BYTES); - init_put_bits(&tex_pb[i], bit_buf_tex[i], MAX_MB_BYTES); - } - - s->last_bits= put_bits_count(&s->pb); - s->mv_bits=0; - s->misc_bits=0; - s->i_tex_bits=0; - s->p_tex_bits=0; - s->i_count=0; - s->f_count=0; - s->b_count=0; - s->skip_count=0; - - for(i=0; i<3; i++){ - /* init last dc values */ - /* note: quant matrix value (8) is implied here */ - s->last_dc[i] = 128 << s->intra_dc_precision; - - s->current_picture.error[i] = 0; - } - s->mb_skip_run = 0; - memset(s->last_mv, 0, sizeof(s->last_mv)); - - s->last_mv_dir = 0; - - switch(s->codec_id){ - case CODEC_ID_H263: - case CODEC_ID_H263P: - case CODEC_ID_FLV1: - s->gob_index = ff_h263_get_gob_height(s); - break; - case CODEC_ID_MPEG4: - if(s->partitioned_frame) - ff_mpeg4_init_partitions(s); - break; - } - - s->resync_mb_x=0; - s->resync_mb_y=0; - s->first_slice_line = 1; - s->ptr_lastgob = s->pb.buf; - for(mb_y= s->start_mb_y; mb_y < s->end_mb_y; mb_y++) { -// printf("row %d at %X\n", s->mb_y, (int)s); - s->mb_x=0; - s->mb_y= mb_y; - - ff_set_qscale(s, s->qscale); - ff_init_block_index(s); - - for(mb_x=0; mb_x < s->mb_width; mb_x++) { - int xy= mb_y*s->mb_stride + mb_x; // removed const, H261 needs to adjust this - int mb_type= s->mb_type[xy]; -// int d; - int dmin= INT_MAX; - int dir; - - if(s->pb.buf_end - s->pb.buf - (put_bits_count(&s->pb)>>3) < MAX_MB_BYTES){ - av_log(s->avctx, AV_LOG_ERROR, "encoded frame too large\n"); - return -1; - } - if(s->data_partitioning){ - if( s->pb2 .buf_end - s->pb2 .buf - (put_bits_count(&s-> pb2)>>3) < MAX_MB_BYTES - || s->tex_pb.buf_end - s->tex_pb.buf - (put_bits_count(&s->tex_pb )>>3) < MAX_MB_BYTES){ - av_log(s->avctx, AV_LOG_ERROR, "encoded frame too large\n"); - return -1; - } - } - - s->mb_x = mb_x; - s->mb_y = mb_y; // moved into loop, can get changed by H.261 - ff_update_block_index(s); - - if(ENABLE_H261_ENCODER && s->codec_id == CODEC_ID_H261){ - ff_h261_reorder_mb_index(s); - xy= s->mb_y*s->mb_stride + s->mb_x; - mb_type= s->mb_type[xy]; - } - - /* write gob / video packet header */ - if(s->rtp_mode){ - int current_packet_size, is_gob_start; - - current_packet_size= ((put_bits_count(&s->pb)+7)>>3) - (s->ptr_lastgob - s->pb.buf); - - is_gob_start= s->avctx->rtp_payload_size && current_packet_size >= s->avctx->rtp_payload_size && mb_y + mb_x>0; - - if(s->start_mb_y == mb_y && mb_y > 0 && mb_x==0) is_gob_start=1; - - switch(s->codec_id){ - case CODEC_ID_H263: - case CODEC_ID_H263P: - if(!s->h263_slice_structured) - if(s->mb_x || s->mb_y%s->gob_index) is_gob_start=0; - break; - case CODEC_ID_MPEG2VIDEO: - if(s->mb_x==0 && s->mb_y!=0) is_gob_start=1; - case CODEC_ID_MPEG1VIDEO: - if(s->mb_skip_run) is_gob_start=0; - break; - } - - if(is_gob_start){ - if(s->start_mb_y != mb_y || mb_x!=0){ - write_slice_end(s); - - if(s->codec_id==CODEC_ID_MPEG4 && s->partitioned_frame){ - ff_mpeg4_init_partitions(s); - } - } - - assert((put_bits_count(&s->pb)&7) == 0); - current_packet_size= pbBufPtr(&s->pb) - s->ptr_lastgob; - - if(s->avctx->error_rate && s->resync_mb_x + s->resync_mb_y > 0){ - int r= put_bits_count(&s->pb)/8 + s->picture_number + 16 + s->mb_x + s->mb_y; - int d= 100 / s->avctx->error_rate; - if(r % d == 0){ - current_packet_size=0; -#ifndef ALT_BITSTREAM_WRITER - s->pb.buf_ptr= s->ptr_lastgob; -#endif - assert(pbBufPtr(&s->pb) == s->ptr_lastgob); - } - } - - if (s->avctx->rtp_callback){ - int number_mb = (mb_y - s->resync_mb_y)*s->mb_width + mb_x - s->resync_mb_x; - s->avctx->rtp_callback(s->avctx, s->ptr_lastgob, current_packet_size, number_mb); - } - - switch(s->codec_id){ - case CODEC_ID_MPEG4: - ff_mpeg4_encode_video_packet_header(s); - ff_mpeg4_clean_buffers(s); - break; - case CODEC_ID_MPEG1VIDEO: - case CODEC_ID_MPEG2VIDEO: - ff_mpeg1_encode_slice_header(s); - ff_mpeg1_clean_buffers(s); - break; - case CODEC_ID_H263: - case CODEC_ID_H263P: - h263_encode_gob_header(s, mb_y); - break; - } - - if(s->flags&CODEC_FLAG_PASS1){ - int bits= put_bits_count(&s->pb); - s->misc_bits+= bits - s->last_bits; - s->last_bits= bits; - } - - s->ptr_lastgob += current_packet_size; - s->first_slice_line=1; - s->resync_mb_x=mb_x; - s->resync_mb_y=mb_y; - } - } - - if( (s->resync_mb_x == s->mb_x) - && s->resync_mb_y+1 == s->mb_y){ - s->first_slice_line=0; - } - - s->mb_skipped=0; - s->dquant=0; //only for QP_RD - - if(mb_type & (mb_type-1) || (s->flags & CODEC_FLAG_QP_RD)){ // more than 1 MB type possible or CODEC_FLAG_QP_RD - int next_block=0; - int pb_bits_count, pb2_bits_count, tex_pb_bits_count; - - copy_context_before_encode(&backup_s, s, -1); - backup_s.pb= s->pb; - best_s.data_partitioning= s->data_partitioning; - best_s.partitioned_frame= s->partitioned_frame; - if(s->data_partitioning){ - backup_s.pb2= s->pb2; - backup_s.tex_pb= s->tex_pb; - } - - if(mb_type&CANDIDATE_MB_TYPE_INTER){ - s->mv_dir = MV_DIR_FORWARD; - s->mv_type = MV_TYPE_16X16; - s->mb_intra= 0; - s->mv[0][0][0] = s->p_mv_table[xy][0]; - s->mv[0][0][1] = s->p_mv_table[xy][1]; - encode_mb_hq(s, &backup_s, &best_s, CANDIDATE_MB_TYPE_INTER, pb, pb2, tex_pb, - &dmin, &next_block, s->mv[0][0][0], s->mv[0][0][1]); - } - if(mb_type&CANDIDATE_MB_TYPE_INTER_I){ - s->mv_dir = MV_DIR_FORWARD; - s->mv_type = MV_TYPE_FIELD; - s->mb_intra= 0; - for(i=0; i<2; i++){ - j= s->field_select[0][i] = s->p_field_select_table[i][xy]; - s->mv[0][i][0] = s->p_field_mv_table[i][j][xy][0]; - s->mv[0][i][1] = s->p_field_mv_table[i][j][xy][1]; - } - encode_mb_hq(s, &backup_s, &best_s, CANDIDATE_MB_TYPE_INTER_I, pb, pb2, tex_pb, - &dmin, &next_block, 0, 0); - } - if(mb_type&CANDIDATE_MB_TYPE_SKIPPED){ - s->mv_dir = MV_DIR_FORWARD; - s->mv_type = MV_TYPE_16X16; - s->mb_intra= 0; - s->mv[0][0][0] = 0; - s->mv[0][0][1] = 0; - encode_mb_hq(s, &backup_s, &best_s, CANDIDATE_MB_TYPE_SKIPPED, pb, pb2, tex_pb, - &dmin, &next_block, s->mv[0][0][0], s->mv[0][0][1]); - } - if(mb_type&CANDIDATE_MB_TYPE_INTER4V){ - s->mv_dir = MV_DIR_FORWARD; - s->mv_type = MV_TYPE_8X8; - s->mb_intra= 0; - for(i=0; i<4; i++){ - s->mv[0][i][0] = s->current_picture.motion_val[0][s->block_index[i]][0]; - s->mv[0][i][1] = s->current_picture.motion_val[0][s->block_index[i]][1]; - } - encode_mb_hq(s, &backup_s, &best_s, CANDIDATE_MB_TYPE_INTER4V, pb, pb2, tex_pb, - &dmin, &next_block, 0, 0); - } - if(mb_type&CANDIDATE_MB_TYPE_FORWARD){ - s->mv_dir = MV_DIR_FORWARD; - s->mv_type = MV_TYPE_16X16; - s->mb_intra= 0; - s->mv[0][0][0] = s->b_forw_mv_table[xy][0]; - s->mv[0][0][1] = s->b_forw_mv_table[xy][1]; - encode_mb_hq(s, &backup_s, &best_s, CANDIDATE_MB_TYPE_FORWARD, pb, pb2, tex_pb, - &dmin, &next_block, s->mv[0][0][0], s->mv[0][0][1]); - } - if(mb_type&CANDIDATE_MB_TYPE_BACKWARD){ - s->mv_dir = MV_DIR_BACKWARD; - s->mv_type = MV_TYPE_16X16; - s->mb_intra= 0; - s->mv[1][0][0] = s->b_back_mv_table[xy][0]; - s->mv[1][0][1] = s->b_back_mv_table[xy][1]; - encode_mb_hq(s, &backup_s, &best_s, CANDIDATE_MB_TYPE_BACKWARD, pb, pb2, tex_pb, - &dmin, &next_block, s->mv[1][0][0], s->mv[1][0][1]); - } - if(mb_type&CANDIDATE_MB_TYPE_BIDIR){ - s->mv_dir = MV_DIR_FORWARD | MV_DIR_BACKWARD; - s->mv_type = MV_TYPE_16X16; - s->mb_intra= 0; - s->mv[0][0][0] = s->b_bidir_forw_mv_table[xy][0]; - s->mv[0][0][1] = s->b_bidir_forw_mv_table[xy][1]; - s->mv[1][0][0] = s->b_bidir_back_mv_table[xy][0]; - s->mv[1][0][1] = s->b_bidir_back_mv_table[xy][1]; - encode_mb_hq(s, &backup_s, &best_s, CANDIDATE_MB_TYPE_BIDIR, pb, pb2, tex_pb, - &dmin, &next_block, 0, 0); - } - if(mb_type&CANDIDATE_MB_TYPE_FORWARD_I){ - s->mv_dir = MV_DIR_FORWARD; - s->mv_type = MV_TYPE_FIELD; - s->mb_intra= 0; - for(i=0; i<2; i++){ - j= s->field_select[0][i] = s->b_field_select_table[0][i][xy]; - s->mv[0][i][0] = s->b_field_mv_table[0][i][j][xy][0]; - s->mv[0][i][1] = s->b_field_mv_table[0][i][j][xy][1]; - } - encode_mb_hq(s, &backup_s, &best_s, CANDIDATE_MB_TYPE_FORWARD_I, pb, pb2, tex_pb, - &dmin, &next_block, 0, 0); - } - if(mb_type&CANDIDATE_MB_TYPE_BACKWARD_I){ - s->mv_dir = MV_DIR_BACKWARD; - s->mv_type = MV_TYPE_FIELD; - s->mb_intra= 0; - for(i=0; i<2; i++){ - j= s->field_select[1][i] = s->b_field_select_table[1][i][xy]; - s->mv[1][i][0] = s->b_field_mv_table[1][i][j][xy][0]; - s->mv[1][i][1] = s->b_field_mv_table[1][i][j][xy][1]; - } - encode_mb_hq(s, &backup_s, &best_s, CANDIDATE_MB_TYPE_BACKWARD_I, pb, pb2, tex_pb, - &dmin, &next_block, 0, 0); - } - if(mb_type&CANDIDATE_MB_TYPE_BIDIR_I){ - s->mv_dir = MV_DIR_FORWARD | MV_DIR_BACKWARD; - s->mv_type = MV_TYPE_FIELD; - s->mb_intra= 0; - for(dir=0; dir<2; dir++){ - for(i=0; i<2; i++){ - j= s->field_select[dir][i] = s->b_field_select_table[dir][i][xy]; - s->mv[dir][i][0] = s->b_field_mv_table[dir][i][j][xy][0]; - s->mv[dir][i][1] = s->b_field_mv_table[dir][i][j][xy][1]; - } - } - encode_mb_hq(s, &backup_s, &best_s, CANDIDATE_MB_TYPE_BIDIR_I, pb, pb2, tex_pb, - &dmin, &next_block, 0, 0); - } - if(mb_type&CANDIDATE_MB_TYPE_INTRA){ - s->mv_dir = 0; - s->mv_type = MV_TYPE_16X16; - s->mb_intra= 1; - s->mv[0][0][0] = 0; - s->mv[0][0][1] = 0; - encode_mb_hq(s, &backup_s, &best_s, CANDIDATE_MB_TYPE_INTRA, pb, pb2, tex_pb, - &dmin, &next_block, 0, 0); - if(s->h263_pred || s->h263_aic){ - if(best_s.mb_intra) - s->mbintra_table[mb_x + mb_y*s->mb_stride]=1; - else - ff_clean_intra_table_entries(s); //old mode? - } - } - - if((s->flags & CODEC_FLAG_QP_RD) && dmin < INT_MAX){ - if(best_s.mv_type==MV_TYPE_16X16){ //FIXME move 4mv after QPRD - const int last_qp= backup_s.qscale; - int qpi, qp, dc[6]; - DCTELEM ac[6][16]; - const int mvdir= (best_s.mv_dir&MV_DIR_BACKWARD) ? 1 : 0; - static const int dquant_tab[4]={-1,1,-2,2}; - - assert(backup_s.dquant == 0); - - //FIXME intra - s->mv_dir= best_s.mv_dir; - s->mv_type = MV_TYPE_16X16; - s->mb_intra= best_s.mb_intra; - s->mv[0][0][0] = best_s.mv[0][0][0]; - s->mv[0][0][1] = best_s.mv[0][0][1]; - s->mv[1][0][0] = best_s.mv[1][0][0]; - s->mv[1][0][1] = best_s.mv[1][0][1]; - - qpi = s->pict_type == B_TYPE ? 2 : 0; - for(; qpi<4; qpi++){ - int dquant= dquant_tab[qpi]; - qp= last_qp + dquant; - if(qp < s->avctx->qmin || qp > s->avctx->qmax) - continue; - backup_s.dquant= dquant; - if(s->mb_intra && s->dc_val[0]){ - for(i=0; i<6; i++){ - dc[i]= s->dc_val[0][ s->block_index[i] ]; - memcpy(ac[i], s->ac_val[0][s->block_index[i]], sizeof(DCTELEM)*16); - } - } - - encode_mb_hq(s, &backup_s, &best_s, CANDIDATE_MB_TYPE_INTER /* wrong but unused */, pb, pb2, tex_pb, - &dmin, &next_block, s->mv[mvdir][0][0], s->mv[mvdir][0][1]); - if(best_s.qscale != qp){ - if(s->mb_intra && s->dc_val[0]){ - for(i=0; i<6; i++){ - s->dc_val[0][ s->block_index[i] ]= dc[i]; - memcpy(s->ac_val[0][s->block_index[i]], ac[i], sizeof(DCTELEM)*16); - } - } - } - } - } - } - if(mb_type&CANDIDATE_MB_TYPE_DIRECT){ - int mx= s->b_direct_mv_table[xy][0]; - int my= s->b_direct_mv_table[xy][1]; - - backup_s.dquant = 0; - s->mv_dir = MV_DIR_FORWARD | MV_DIR_BACKWARD | MV_DIRECT; - s->mb_intra= 0; - ff_mpeg4_set_direct_mv(s, mx, my); - encode_mb_hq(s, &backup_s, &best_s, CANDIDATE_MB_TYPE_DIRECT, pb, pb2, tex_pb, - &dmin, &next_block, mx, my); - } - if(mb_type&CANDIDATE_MB_TYPE_DIRECT0){ - backup_s.dquant = 0; - s->mv_dir = MV_DIR_FORWARD | MV_DIR_BACKWARD | MV_DIRECT; - s->mb_intra= 0; - ff_mpeg4_set_direct_mv(s, 0, 0); - encode_mb_hq(s, &backup_s, &best_s, CANDIDATE_MB_TYPE_DIRECT, pb, pb2, tex_pb, - &dmin, &next_block, 0, 0); - } - if(!best_s.mb_intra && s->flags2&CODEC_FLAG2_SKIP_RD){ - int coded=0; - for(i=0; i<6; i++) - coded |= s->block_last_index[i]; - if(coded){ - int mx,my; - memcpy(s->mv, best_s.mv, sizeof(s->mv)); - if(best_s.mv_dir & MV_DIRECT){ - mx=my=0; //FIXME find the one we actually used - ff_mpeg4_set_direct_mv(s, mx, my); - }else if(best_s.mv_dir&MV_DIR_BACKWARD){ - mx= s->mv[1][0][0]; - my= s->mv[1][0][1]; - }else{ - mx= s->mv[0][0][0]; - my= s->mv[0][0][1]; - } - - s->mv_dir= best_s.mv_dir; - s->mv_type = best_s.mv_type; - s->mb_intra= 0; -/* s->mv[0][0][0] = best_s.mv[0][0][0]; - s->mv[0][0][1] = best_s.mv[0][0][1]; - s->mv[1][0][0] = best_s.mv[1][0][0]; - s->mv[1][0][1] = best_s.mv[1][0][1];*/ - backup_s.dquant= 0; - s->skipdct=1; - encode_mb_hq(s, &backup_s, &best_s, CANDIDATE_MB_TYPE_INTER /* wrong but unused */, pb, pb2, tex_pb, - &dmin, &next_block, mx, my); - s->skipdct=0; - } - } - - s->current_picture.qscale_table[xy]= best_s.qscale; - - copy_context_after_encode(s, &best_s, -1); - - pb_bits_count= put_bits_count(&s->pb); - flush_put_bits(&s->pb); - ff_copy_bits(&backup_s.pb, bit_buf[next_block^1], pb_bits_count); - s->pb= backup_s.pb; - - if(s->data_partitioning){ - pb2_bits_count= put_bits_count(&s->pb2); - flush_put_bits(&s->pb2); - ff_copy_bits(&backup_s.pb2, bit_buf2[next_block^1], pb2_bits_count); - s->pb2= backup_s.pb2; - - tex_pb_bits_count= put_bits_count(&s->tex_pb); - flush_put_bits(&s->tex_pb); - ff_copy_bits(&backup_s.tex_pb, bit_buf_tex[next_block^1], tex_pb_bits_count); - s->tex_pb= backup_s.tex_pb; - } - s->last_bits= put_bits_count(&s->pb); - - if (s->out_format == FMT_H263 && s->pict_type!=B_TYPE) - ff_h263_update_motion_val(s); - - if(next_block==0){ //FIXME 16 vs linesize16 - s->dsp.put_pixels_tab[0][0](s->dest[0], s->rd_scratchpad , s->linesize ,16); - s->dsp.put_pixels_tab[1][0](s->dest[1], s->rd_scratchpad + 16*s->linesize , s->uvlinesize, 8); - s->dsp.put_pixels_tab[1][0](s->dest[2], s->rd_scratchpad + 16*s->linesize + 8, s->uvlinesize, 8); - } - - if(s->avctx->mb_decision == FF_MB_DECISION_BITS) - MPV_decode_mb(s, s->block); - } else { - int motion_x = 0, motion_y = 0; - s->mv_type=MV_TYPE_16X16; - // only one MB-Type possible - - switch(mb_type){ - case CANDIDATE_MB_TYPE_INTRA: - s->mv_dir = 0; - s->mb_intra= 1; - motion_x= s->mv[0][0][0] = 0; - motion_y= s->mv[0][0][1] = 0; - break; - case CANDIDATE_MB_TYPE_INTER: - s->mv_dir = MV_DIR_FORWARD; - s->mb_intra= 0; - motion_x= s->mv[0][0][0] = s->p_mv_table[xy][0]; - motion_y= s->mv[0][0][1] = s->p_mv_table[xy][1]; - break; - case CANDIDATE_MB_TYPE_INTER_I: - s->mv_dir = MV_DIR_FORWARD; - s->mv_type = MV_TYPE_FIELD; - s->mb_intra= 0; - for(i=0; i<2; i++){ - j= s->field_select[0][i] = s->p_field_select_table[i][xy]; - s->mv[0][i][0] = s->p_field_mv_table[i][j][xy][0]; - s->mv[0][i][1] = s->p_field_mv_table[i][j][xy][1]; - } - break; - case CANDIDATE_MB_TYPE_INTER4V: - s->mv_dir = MV_DIR_FORWARD; - s->mv_type = MV_TYPE_8X8; - s->mb_intra= 0; - for(i=0; i<4; i++){ - s->mv[0][i][0] = s->current_picture.motion_val[0][s->block_index[i]][0]; - s->mv[0][i][1] = s->current_picture.motion_val[0][s->block_index[i]][1]; - } - break; - case CANDIDATE_MB_TYPE_DIRECT: - s->mv_dir = MV_DIR_FORWARD | MV_DIR_BACKWARD | MV_DIRECT; - s->mb_intra= 0; - motion_x=s->b_direct_mv_table[xy][0]; - motion_y=s->b_direct_mv_table[xy][1]; - ff_mpeg4_set_direct_mv(s, motion_x, motion_y); - break; - case CANDIDATE_MB_TYPE_DIRECT0: - s->mv_dir = MV_DIR_FORWARD | MV_DIR_BACKWARD | MV_DIRECT; - s->mb_intra= 0; - ff_mpeg4_set_direct_mv(s, 0, 0); - break; - case CANDIDATE_MB_TYPE_BIDIR: - s->mv_dir = MV_DIR_FORWARD | MV_DIR_BACKWARD; - s->mb_intra= 0; - s->mv[0][0][0] = s->b_bidir_forw_mv_table[xy][0]; - s->mv[0][0][1] = s->b_bidir_forw_mv_table[xy][1]; - s->mv[1][0][0] = s->b_bidir_back_mv_table[xy][0]; - s->mv[1][0][1] = s->b_bidir_back_mv_table[xy][1]; - break; - case CANDIDATE_MB_TYPE_BACKWARD: - s->mv_dir = MV_DIR_BACKWARD; - s->mb_intra= 0; - motion_x= s->mv[1][0][0] = s->b_back_mv_table[xy][0]; - motion_y= s->mv[1][0][1] = s->b_back_mv_table[xy][1]; - break; - case CANDIDATE_MB_TYPE_FORWARD: - s->mv_dir = MV_DIR_FORWARD; - s->mb_intra= 0; - motion_x= s->mv[0][0][0] = s->b_forw_mv_table[xy][0]; - motion_y= s->mv[0][0][1] = s->b_forw_mv_table[xy][1]; -// printf(" %d %d ", motion_x, motion_y); - break; - case CANDIDATE_MB_TYPE_FORWARD_I: - s->mv_dir = MV_DIR_FORWARD; - s->mv_type = MV_TYPE_FIELD; - s->mb_intra= 0; - for(i=0; i<2; i++){ - j= s->field_select[0][i] = s->b_field_select_table[0][i][xy]; - s->mv[0][i][0] = s->b_field_mv_table[0][i][j][xy][0]; - s->mv[0][i][1] = s->b_field_mv_table[0][i][j][xy][1]; - } - break; - case CANDIDATE_MB_TYPE_BACKWARD_I: - s->mv_dir = MV_DIR_BACKWARD; - s->mv_type = MV_TYPE_FIELD; - s->mb_intra= 0; - for(i=0; i<2; i++){ - j= s->field_select[1][i] = s->b_field_select_table[1][i][xy]; - s->mv[1][i][0] = s->b_field_mv_table[1][i][j][xy][0]; - s->mv[1][i][1] = s->b_field_mv_table[1][i][j][xy][1]; - } - break; - case CANDIDATE_MB_TYPE_BIDIR_I: - s->mv_dir = MV_DIR_FORWARD | MV_DIR_BACKWARD; - s->mv_type = MV_TYPE_FIELD; - s->mb_intra= 0; - for(dir=0; dir<2; dir++){ - for(i=0; i<2; i++){ - j= s->field_select[dir][i] = s->b_field_select_table[dir][i][xy]; - s->mv[dir][i][0] = s->b_field_mv_table[dir][i][j][xy][0]; - s->mv[dir][i][1] = s->b_field_mv_table[dir][i][j][xy][1]; - } - } - break; - default: - av_log(s->avctx, AV_LOG_ERROR, "illegal MB type\n"); - } - - encode_mb(s, motion_x, motion_y); - - // RAL: Update last macroblock type - s->last_mv_dir = s->mv_dir; - - if (s->out_format == FMT_H263 && s->pict_type!=B_TYPE) - ff_h263_update_motion_val(s); - - MPV_decode_mb(s, s->block); - } - - /* clean the MV table in IPS frames for direct mode in B frames */ - if(s->mb_intra /* && I,P,S_TYPE */){ - s->p_mv_table[xy][0]=0; - s->p_mv_table[xy][1]=0; - } - - if(s->flags&CODEC_FLAG_PSNR){ - int w= 16; - int h= 16; - - if(s->mb_x*16 + 16 > s->width ) w= s->width - s->mb_x*16; - if(s->mb_y*16 + 16 > s->height) h= s->height- s->mb_y*16; - - s->current_picture.error[0] += sse( - s, s->new_picture.data[0] + s->mb_x*16 + s->mb_y*s->linesize*16, - s->dest[0], w, h, s->linesize); - s->current_picture.error[1] += sse( - s, s->new_picture.data[1] + s->mb_x*8 + s->mb_y*s->uvlinesize*8, - s->dest[1], w>>1, h>>1, s->uvlinesize); - s->current_picture.error[2] += sse( - s, s->new_picture .data[2] + s->mb_x*8 + s->mb_y*s->uvlinesize*8, - s->dest[2], w>>1, h>>1, s->uvlinesize); - } - if(s->loop_filter){ - if(s->out_format == FMT_H263) - ff_h263_loop_filter(s); - } -//printf("MB %d %d bits\n", s->mb_x+s->mb_y*s->mb_stride, put_bits_count(&s->pb)); - } - } - - //not beautiful here but we must write it before flushing so it has to be here - if (ENABLE_MSMPEG4_ENCODER && s->msmpeg4_version && s->msmpeg4_version<4 && s->pict_type == I_TYPE) - msmpeg4_encode_ext_header(s); - - write_slice_end(s); - - /* Send the last GOB if RTP */ - if (s->avctx->rtp_callback) { - int number_mb = (mb_y - s->resync_mb_y)*s->mb_width - s->resync_mb_x; - pdif = pbBufPtr(&s->pb) - s->ptr_lastgob; - /* Call the RTP callback to send the last GOB */ - emms_c(); - s->avctx->rtp_callback(s->avctx, s->ptr_lastgob, pdif, number_mb); - } - - return 0; -} - -#define MERGE(field) dst->field += src->field; src->field=0 -static void merge_context_after_me(MpegEncContext *dst, MpegEncContext *src){ - MERGE(me.scene_change_score); - MERGE(me.mc_mb_var_sum_temp); - MERGE(me.mb_var_sum_temp); -} - -static void merge_context_after_encode(MpegEncContext *dst, MpegEncContext *src){ - int i; - - MERGE(dct_count[0]); //note, the other dct vars are not part of the context - MERGE(dct_count[1]); - MERGE(mv_bits); - MERGE(i_tex_bits); - MERGE(p_tex_bits); - MERGE(i_count); - MERGE(f_count); - MERGE(b_count); - MERGE(skip_count); - MERGE(misc_bits); - MERGE(error_count); - MERGE(padding_bug_score); - MERGE(current_picture.error[0]); - MERGE(current_picture.error[1]); - MERGE(current_picture.error[2]); - - if(dst->avctx->noise_reduction){ - for(i=0; i<64; i++){ - MERGE(dct_error_sum[0][i]); - MERGE(dct_error_sum[1][i]); - } - } - - assert(put_bits_count(&src->pb) % 8 ==0); - assert(put_bits_count(&dst->pb) % 8 ==0); - ff_copy_bits(&dst->pb, src->pb.buf, put_bits_count(&src->pb)); - flush_put_bits(&dst->pb); -} - -static int estimate_qp(MpegEncContext *s, int dry_run){ - if (s->next_lambda){ - s->current_picture_ptr->quality= - s->current_picture.quality = s->next_lambda; - if(!dry_run) s->next_lambda= 0; - } else if (!s->fixed_qscale) { - s->current_picture_ptr->quality= - s->current_picture.quality = ff_rate_estimate_qscale(s, dry_run); - if (s->current_picture.quality < 0) - return -1; - } - - if(s->adaptive_quant){ - switch(s->codec_id){ - case CODEC_ID_MPEG4: - ff_clean_mpeg4_qscales(s); - break; - case CODEC_ID_H263: - case CODEC_ID_H263P: - case CODEC_ID_FLV1: - ff_clean_h263_qscales(s); - break; - } - - s->lambda= s->lambda_table[0]; - //FIXME broken - }else - s->lambda= s->current_picture.quality; -//printf("%d %d\n", s->avctx->global_quality, s->current_picture.quality); - update_qscale(s); - return 0; -} - -static int encode_picture(MpegEncContext *s, int picture_number) -{ - int i; - int bits; - - s->picture_number = picture_number; - - /* Reset the average MB variance */ - s->me.mb_var_sum_temp = - s->me.mc_mb_var_sum_temp = 0; - - /* we need to initialize some time vars before we can encode b-frames */ - // RAL: Condition added for MPEG1VIDEO - if (s->codec_id == CODEC_ID_MPEG1VIDEO || s->codec_id == CODEC_ID_MPEG2VIDEO || (s->h263_pred && !s->h263_msmpeg4)) - ff_set_mpeg4_time(s, s->picture_number); //FIXME rename and use has_b_frames or similar - - s->me.scene_change_score=0; - -// s->lambda= s->current_picture_ptr->quality; //FIXME qscale / ... stuff for ME ratedistoration - - if(s->pict_type==I_TYPE){ - if(s->msmpeg4_version >= 3) s->no_rounding=1; - else s->no_rounding=0; - }else if(s->pict_type!=B_TYPE){ - if(s->flipflop_rounding || s->codec_id == CODEC_ID_H263P || s->codec_id == CODEC_ID_MPEG4) - s->no_rounding ^= 1; - } - - if(s->flags & CODEC_FLAG_PASS2){ - if (estimate_qp(s,1) < 0) - return -1; - ff_get_2pass_fcode(s); - }else if(!(s->flags & CODEC_FLAG_QSCALE)){ - if(s->pict_type==B_TYPE) - s->lambda= s->last_lambda_for[s->pict_type]; - else - s->lambda= s->last_lambda_for[s->last_non_b_pict_type]; - update_qscale(s); - } - - s->mb_intra=0; //for the rate distortion & bit compare functions - for(i=1; iavctx->thread_count; i++){ - ff_update_duplicate_context(s->thread_context[i], s); - } - - ff_init_me(s); - - /* Estimate motion for every MB */ - if(s->pict_type != I_TYPE){ - s->lambda = (s->lambda * s->avctx->me_penalty_compensation + 128)>>8; - s->lambda2= (s->lambda2* (int64_t)s->avctx->me_penalty_compensation + 128)>>8; - if(s->pict_type != B_TYPE && s->avctx->me_threshold==0){ - if((s->avctx->pre_me && s->last_non_b_pict_type==I_TYPE) || s->avctx->pre_me==2){ - s->avctx->execute(s->avctx, pre_estimate_motion_thread, (void**)&(s->thread_context[0]), NULL, s->avctx->thread_count); - } - } - - s->avctx->execute(s->avctx, estimate_motion_thread, (void**)&(s->thread_context[0]), NULL, s->avctx->thread_count); - }else /* if(s->pict_type == I_TYPE) */{ - /* I-Frame */ - for(i=0; imb_stride*s->mb_height; i++) - s->mb_type[i]= CANDIDATE_MB_TYPE_INTRA; - - if(!s->fixed_qscale){ - /* finding spatial complexity for I-frame rate control */ - s->avctx->execute(s->avctx, mb_var_thread, (void**)&(s->thread_context[0]), NULL, s->avctx->thread_count); - } - } - for(i=1; iavctx->thread_count; i++){ - merge_context_after_me(s, s->thread_context[i]); - } - s->current_picture.mc_mb_var_sum= s->current_picture_ptr->mc_mb_var_sum= s->me.mc_mb_var_sum_temp; - s->current_picture. mb_var_sum= s->current_picture_ptr-> mb_var_sum= s->me. mb_var_sum_temp; - emms_c(); - - if(s->me.scene_change_score > s->avctx->scenechange_threshold && s->pict_type == P_TYPE){ - s->pict_type= I_TYPE; - for(i=0; imb_stride*s->mb_height; i++) - s->mb_type[i]= CANDIDATE_MB_TYPE_INTRA; -//printf("Scene change detected, encoding as I Frame %d %d\n", s->current_picture.mb_var_sum, s->current_picture.mc_mb_var_sum); - } - - if(!s->umvplus){ - if(s->pict_type==P_TYPE || s->pict_type==S_TYPE) { - s->f_code= ff_get_best_fcode(s, s->p_mv_table, CANDIDATE_MB_TYPE_INTER); - - if(s->flags & CODEC_FLAG_INTERLACED_ME){ - int a,b; - a= ff_get_best_fcode(s, s->p_field_mv_table[0][0], CANDIDATE_MB_TYPE_INTER_I); //FIXME field_select - b= ff_get_best_fcode(s, s->p_field_mv_table[1][1], CANDIDATE_MB_TYPE_INTER_I); - s->f_code= FFMAX(s->f_code, FFMAX(a,b)); - } - - ff_fix_long_p_mvs(s); - ff_fix_long_mvs(s, NULL, 0, s->p_mv_table, s->f_code, CANDIDATE_MB_TYPE_INTER, 0); - if(s->flags & CODEC_FLAG_INTERLACED_ME){ - int j; - for(i=0; i<2; i++){ - for(j=0; j<2; j++) - ff_fix_long_mvs(s, s->p_field_select_table[i], j, - s->p_field_mv_table[i][j], s->f_code, CANDIDATE_MB_TYPE_INTER_I, 0); - } - } - } - - if(s->pict_type==B_TYPE){ - int a, b; - - a = ff_get_best_fcode(s, s->b_forw_mv_table, CANDIDATE_MB_TYPE_FORWARD); - b = ff_get_best_fcode(s, s->b_bidir_forw_mv_table, CANDIDATE_MB_TYPE_BIDIR); - s->f_code = FFMAX(a, b); - - a = ff_get_best_fcode(s, s->b_back_mv_table, CANDIDATE_MB_TYPE_BACKWARD); - b = ff_get_best_fcode(s, s->b_bidir_back_mv_table, CANDIDATE_MB_TYPE_BIDIR); - s->b_code = FFMAX(a, b); - - ff_fix_long_mvs(s, NULL, 0, s->b_forw_mv_table, s->f_code, CANDIDATE_MB_TYPE_FORWARD, 1); - ff_fix_long_mvs(s, NULL, 0, s->b_back_mv_table, s->b_code, CANDIDATE_MB_TYPE_BACKWARD, 1); - ff_fix_long_mvs(s, NULL, 0, s->b_bidir_forw_mv_table, s->f_code, CANDIDATE_MB_TYPE_BIDIR, 1); - ff_fix_long_mvs(s, NULL, 0, s->b_bidir_back_mv_table, s->b_code, CANDIDATE_MB_TYPE_BIDIR, 1); - if(s->flags & CODEC_FLAG_INTERLACED_ME){ - int dir, j; - for(dir=0; dir<2; dir++){ - for(i=0; i<2; i++){ - for(j=0; j<2; j++){ - int type= dir ? (CANDIDATE_MB_TYPE_BACKWARD_I|CANDIDATE_MB_TYPE_BIDIR_I) - : (CANDIDATE_MB_TYPE_FORWARD_I |CANDIDATE_MB_TYPE_BIDIR_I); - ff_fix_long_mvs(s, s->b_field_select_table[dir][i], j, - s->b_field_mv_table[dir][i][j], dir ? s->b_code : s->f_code, type, 1); - } - } - } - } - } - } - - if (estimate_qp(s, 0) < 0) - return -1; - - if(s->qscale < 3 && s->max_qcoeff<=128 && s->pict_type==I_TYPE && !(s->flags & CODEC_FLAG_QSCALE)) - s->qscale= 3; //reduce clipping problems - - if (s->out_format == FMT_MJPEG) { - /* for mjpeg, we do include qscale in the matrix */ - s->intra_matrix[0] = ff_mpeg1_default_intra_matrix[0]; - for(i=1;i<64;i++){ - int j= s->dsp.idct_permutation[i]; - - s->intra_matrix[j] = av_clip_uint8((ff_mpeg1_default_intra_matrix[i] * s->qscale) >> 3); - } - convert_matrix(&s->dsp, s->q_intra_matrix, s->q_intra_matrix16, - s->intra_matrix, s->intra_quant_bias, 8, 8, 1); - s->qscale= 8; - } - - //FIXME var duplication - s->current_picture_ptr->key_frame= - s->current_picture.key_frame= s->pict_type == I_TYPE; //FIXME pic_ptr - s->current_picture_ptr->pict_type= - s->current_picture.pict_type= s->pict_type; - - if(s->current_picture.key_frame) - s->picture_in_gop_number=0; - - s->last_bits= put_bits_count(&s->pb); - switch(s->out_format) { - case FMT_MJPEG: - if (ENABLE_MJPEG_ENCODER) - ff_mjpeg_encode_picture_header(s); - break; - case FMT_H261: - if (ENABLE_H261_ENCODER) - ff_h261_encode_picture_header(s, picture_number); - break; - case FMT_H263: - if (ENABLE_WMV2_ENCODER && s->codec_id == CODEC_ID_WMV2) - ff_wmv2_encode_picture_header(s, picture_number); - else if (ENABLE_MSMPEG4_ENCODER && s->h263_msmpeg4) - msmpeg4_encode_picture_header(s, picture_number); - else if (s->h263_pred) - mpeg4_encode_picture_header(s, picture_number); - else if (ENABLE_RV10_ENCODER && s->codec_id == CODEC_ID_RV10) - rv10_encode_picture_header(s, picture_number); - else if (ENABLE_RV20_ENCODER && s->codec_id == CODEC_ID_RV20) - rv20_encode_picture_header(s, picture_number); - else if (s->codec_id == CODEC_ID_FLV1) - ff_flv_encode_picture_header(s, picture_number); - else - h263_encode_picture_header(s, picture_number); - break; - case FMT_MPEG1: - mpeg1_encode_picture_header(s, picture_number); - break; - case FMT_H264: - break; - default: - assert(0); - } - bits= put_bits_count(&s->pb); - s->header_bits= bits - s->last_bits; - - for(i=1; iavctx->thread_count; i++){ - update_duplicate_context_after_me(s->thread_context[i], s); - } - s->avctx->execute(s->avctx, encode_thread, (void**)&(s->thread_context[0]), NULL, s->avctx->thread_count); - for(i=1; iavctx->thread_count; i++){ - merge_context_after_encode(s, s->thread_context[i]); - } - emms_c(); - return 0; -} - -static void denoise_dct_c(MpegEncContext *s, DCTELEM *block){ - const int intra= s->mb_intra; - int i; - - s->dct_count[intra]++; - - for(i=0; i<64; i++){ - int level= block[i]; - - if(level){ - if(level>0){ - s->dct_error_sum[intra][i] += level; - level -= s->dct_offset[intra][i]; - if(level<0) level=0; - }else{ - s->dct_error_sum[intra][i] -= level; - level += s->dct_offset[intra][i]; - if(level>0) level=0; - } - block[i]= level; - } - } -} - -static int dct_quantize_trellis_c(MpegEncContext *s, - DCTELEM *block, int n, - int qscale, int *overflow){ - const int *qmat; - const uint8_t *scantable= s->intra_scantable.scantable; - const uint8_t *perm_scantable= s->intra_scantable.permutated; - int max=0; - unsigned int threshold1, threshold2; - int bias=0; - int run_tab[65]; - int level_tab[65]; - int score_tab[65]; - int survivor[65]; - int survivor_count; - int last_run=0; - int last_level=0; - int last_score= 0; - int last_i; - int coeff[2][64]; - int coeff_count[64]; - int qmul, qadd, start_i, last_non_zero, i, dc; - const int esc_length= s->ac_esc_length; - uint8_t * length; - uint8_t * last_length; - const int lambda= s->lambda2 >> (FF_LAMBDA_SHIFT - 6); - - s->dsp.fdct (block); - - if(s->dct_error_sum) - s->denoise_dct(s, block); - qmul= qscale*16; - qadd= ((qscale-1)|1)*8; - - if (s->mb_intra) { - int q; - if (!s->h263_aic) { - if (n < 4) - q = s->y_dc_scale; - else - q = s->c_dc_scale; - q = q << 3; - } else{ - /* For AIC we skip quant/dequant of INTRADC */ - q = 1 << 3; - qadd=0; - } - - /* note: block[0] is assumed to be positive */ - block[0] = (block[0] + (q >> 1)) / q; - start_i = 1; - last_non_zero = 0; - qmat = s->q_intra_matrix[qscale]; - if(s->mpeg_quant || s->out_format == FMT_MPEG1) - bias= 1<<(QMAT_SHIFT-1); - length = s->intra_ac_vlc_length; - last_length= s->intra_ac_vlc_last_length; - } else { - start_i = 0; - last_non_zero = -1; - qmat = s->q_inter_matrix[qscale]; - length = s->inter_ac_vlc_length; - last_length= s->inter_ac_vlc_last_length; - } - last_i= start_i; - - threshold1= (1<=start_i; i--) { - const int j = scantable[i]; - int level = block[j] * qmat[j]; - - if(((unsigned)(level+threshold1))>threshold2){ - last_non_zero = i; - break; - } - } - - for(i=start_i; i<=last_non_zero; i++) { - const int j = scantable[i]; - int level = block[j] * qmat[j]; - -// if( bias+level >= (1<<(QMAT_SHIFT - 3)) -// || bias-level >= (1<<(QMAT_SHIFT - 3))){ - if(((unsigned)(level+threshold1))>threshold2){ - if(level>0){ - level= (bias + level)>>QMAT_SHIFT; - coeff[0][i]= level; - coeff[1][i]= level-1; -// coeff[2][k]= level-2; - }else{ - level= (bias - level)>>QMAT_SHIFT; - coeff[0][i]= -level; - coeff[1][i]= -level+1; -// coeff[2][k]= -level+2; - } - coeff_count[i]= FFMIN(level, 2); - assert(coeff_count[i]); - max |=level; - }else{ - coeff[0][i]= (level>>31)|1; - coeff_count[i]= 1; - } - } - - *overflow= s->max_qcoeff < max; //overflow might have happened - - if(last_non_zero < start_i){ - memset(block + start_i, 0, (64-start_i)*sizeof(DCTELEM)); - return last_non_zero; - } - - score_tab[start_i]= 0; - survivor[0]= start_i; - survivor_count= 1; - - for(i=start_i; i<=last_non_zero; i++){ - int level_index, j; - const int dct_coeff= FFABS(block[ scantable[i] ]); - const int zero_distoration= dct_coeff*dct_coeff; - int best_score=256*256*256*120; - for(level_index=0; level_index < coeff_count[i]; level_index++){ - int distoration; - int level= coeff[level_index][i]; - const int alevel= FFABS(level); - int unquant_coeff; - - assert(level); - - if(s->out_format == FMT_H263){ - unquant_coeff= alevel*qmul + qadd; - }else{ //MPEG1 - j= s->dsp.idct_permutation[ scantable[i] ]; //FIXME optimize - if(s->mb_intra){ - unquant_coeff = (int)( alevel * qscale * s->intra_matrix[j]) >> 3; - unquant_coeff = (unquant_coeff - 1) | 1; - }else{ - unquant_coeff = ((( alevel << 1) + 1) * qscale * ((int) s->inter_matrix[j])) >> 4; - unquant_coeff = (unquant_coeff - 1) | 1; - } - unquant_coeff<<= 3; - } - - distoration= (unquant_coeff - dct_coeff) * (unquant_coeff - dct_coeff) - zero_distoration; - level+=64; - if((level&(~127)) == 0){ - for(j=survivor_count-1; j>=0; j--){ - int run= i - survivor[j]; - int score= distoration + length[UNI_AC_ENC_INDEX(run, level)]*lambda; - score += score_tab[i-run]; - - if(score < best_score){ - best_score= score; - run_tab[i+1]= run; - level_tab[i+1]= level-64; - } - } - - if(s->out_format == FMT_H263){ - for(j=survivor_count-1; j>=0; j--){ - int run= i - survivor[j]; - int score= distoration + last_length[UNI_AC_ENC_INDEX(run, level)]*lambda; - score += score_tab[i-run]; - if(score < last_score){ - last_score= score; - last_run= run; - last_level= level-64; - last_i= i+1; - } - } - } - }else{ - distoration += esc_length*lambda; - for(j=survivor_count-1; j>=0; j--){ - int run= i - survivor[j]; - int score= distoration + score_tab[i-run]; - - if(score < best_score){ - best_score= score; - run_tab[i+1]= run; - level_tab[i+1]= level-64; - } - } - - if(s->out_format == FMT_H263){ - for(j=survivor_count-1; j>=0; j--){ - int run= i - survivor[j]; - int score= distoration + score_tab[i-run]; - if(score < last_score){ - last_score= score; - last_run= run; - last_level= level-64; - last_i= i+1; - } - } - } - } - } - - score_tab[i+1]= best_score; - - //Note: there is a vlc code in mpeg4 which is 1 bit shorter then another one with a shorter run and the same level - if(last_non_zero <= 27){ - for(; survivor_count; survivor_count--){ - if(score_tab[ survivor[survivor_count-1] ] <= best_score) - break; - } - }else{ - for(; survivor_count; survivor_count--){ - if(score_tab[ survivor[survivor_count-1] ] <= best_score + lambda) - break; - } - } - - survivor[ survivor_count++ ]= i+1; - } - - if(s->out_format != FMT_H263){ - last_score= 256*256*256*120; - for(i= survivor[0]; i<=last_non_zero + 1; i++){ - int score= score_tab[i]; - if(i) score += lambda*2; //FIXME exacter? - - if(score < last_score){ - last_score= score; - last_i= i; - last_level= level_tab[i]; - last_run= run_tab[i]; - } - } - } - - s->coded_score[n] = last_score; - - dc= FFABS(block[0]); - last_non_zero= last_i - 1; - memset(block + start_i, 0, (64-start_i)*sizeof(DCTELEM)); - - if(last_non_zero < start_i) - return last_non_zero; - - if(last_non_zero == 0 && start_i == 0){ - int best_level= 0; - int best_score= dc * dc; - - for(i=0; iout_format == FMT_H263){ - unquant_coeff= (alevel*qmul + qadd)>>3; - }else{ //MPEG1 - unquant_coeff = ((( alevel << 1) + 1) * qscale * ((int) s->inter_matrix[0])) >> 4; - unquant_coeff = (unquant_coeff - 1) | 1; - } - unquant_coeff = (unquant_coeff + 4) >> 3; - unquant_coeff<<= 3 + 3; - - distortion= (unquant_coeff - dc) * (unquant_coeff - dc); - level+=64; - if((level&(~127)) == 0) score= distortion + last_length[UNI_AC_ENC_INDEX(0, level)]*lambda; - else score= distortion + esc_length*lambda; - - if(score < best_score){ - best_score= score; - best_level= level - 64; - } - } - block[0]= best_level; - s->coded_score[n] = best_score - dc*dc; - if(best_level == 0) return -1; - else return last_non_zero; - } - - i= last_i; - assert(last_level); - - block[ perm_scantable[last_non_zero] ]= last_level; - i -= last_run + 1; - - for(; i>start_i; i -= run_tab[i] + 1){ - block[ perm_scantable[i-1] ]= level_tab[i]; - } - - return last_non_zero; -} - -//#define REFINE_STATS 1 -static int16_t basis[64][64]; - -static void build_basis(uint8_t *perm){ - int i, j, x, y; - emms_c(); - for(i=0; i<8; i++){ - for(j=0; j<8; j++){ - for(y=0; y<8; y++){ - for(x=0; x<8; x++){ - double s= 0.25*(1<intra_scantable.scantable; - const uint8_t *perm_scantable= s->intra_scantable.permutated; -// unsigned int threshold1, threshold2; -// int bias=0; - int run_tab[65]; - int prev_run=0; - int prev_level=0; - int qmul, qadd, start_i, last_non_zero, i, dc; - uint8_t * length; - uint8_t * last_length; - int lambda; - int rle_index, run, q = 1, sum; //q is only used when s->mb_intra is true -#ifdef REFINE_STATS -static int count=0; -static int after_last=0; -static int to_zero=0; -static int from_zero=0; -static int raise=0; -static int lower=0; -static int messed_sign=0; -#endif - - if(basis[0][0] == 0) - build_basis(s->dsp.idct_permutation); - - qmul= qscale*2; - qadd= (qscale-1)|1; - if (s->mb_intra) { - if (!s->h263_aic) { - if (n < 4) - q = s->y_dc_scale; - else - q = s->c_dc_scale; - } else{ - /* For AIC we skip quant/dequant of INTRADC */ - q = 1; - qadd=0; - } - q <<= RECON_SHIFT-3; - /* note: block[0] is assumed to be positive */ - dc= block[0]*q; -// block[0] = (block[0] + (q >> 1)) / q; - start_i = 1; - qmat = s->q_intra_matrix[qscale]; -// if(s->mpeg_quant || s->out_format == FMT_MPEG1) -// bias= 1<<(QMAT_SHIFT-1); - length = s->intra_ac_vlc_length; - last_length= s->intra_ac_vlc_last_length; - } else { - dc= 0; - start_i = 0; - qmat = s->q_inter_matrix[qscale]; - length = s->inter_ac_vlc_length; - last_length= s->inter_ac_vlc_last_length; - } - last_non_zero = s->block_last_index[n]; - -#ifdef REFINE_STATS -{START_TIMER -#endif - dc += (1<<(RECON_SHIFT-1)); - for(i=0; i<64; i++){ - rem[i]= dc - (orig[i]<0); - assert(w<(1<<6)); - sum += w*w; - } - lambda= sum*(uint64_t)s->lambda2 >> (FF_LAMBDA_SHIFT - 6 + 6 + 6 + 6); -#ifdef REFINE_STATS -{START_TIMER -#endif - run=0; - rle_index=0; - for(i=start_i; i<=last_non_zero; i++){ - int j= perm_scantable[i]; - const int level= block[j]; - int coeff; - - if(level){ - if(level<0) coeff= qmul*level - qadd; - else coeff= qmul*level + qadd; - run_tab[rle_index++]=run; - run=0; - - s->dsp.add_8x8basis(rem, basis[j], coeff); - }else{ - run++; - } - } -#ifdef REFINE_STATS -if(last_non_zero>0){ -STOP_TIMER("init rem[]") -} -} - -{START_TIMER -#endif - for(;;){ - int best_score=s->dsp.try_8x8basis(rem, weight, basis[0], 0); - int best_coeff=0; - int best_change=0; - int run2, best_unquant_change=0, analyze_gradient; -#ifdef REFINE_STATS -{START_TIMER -#endif - analyze_gradient = last_non_zero > 2 || s->avctx->quantizer_noise_shaping >= 3; - - if(analyze_gradient){ -#ifdef REFINE_STATS -{START_TIMER -#endif - for(i=0; i<64; i++){ - int w= weight[i]; - - d1[i] = (rem[i]*w*w + (1<<(RECON_SHIFT+12-1)))>>(RECON_SHIFT+12); - } -#ifdef REFINE_STATS -STOP_TIMER("rem*w*w")} -{START_TIMER -#endif - s->dsp.fdct(d1); -#ifdef REFINE_STATS -STOP_TIMER("dct")} -#endif - } - - if(start_i){ - const int level= block[0]; - int change, old_coeff; - - assert(s->mb_intra); - - old_coeff= q*level; - - for(change=-1; change<=1; change+=2){ - int new_level= level + change; - int score, new_coeff; - - new_coeff= q*new_level; - if(new_coeff >= 2048 || new_coeff < 0) - continue; - - score= s->dsp.try_8x8basis(rem, weight, basis[0], new_coeff - old_coeff); - if(scoreavctx->quantizer_noise_shaping < 3 && i > last_non_zero + 1) - break; - - if(level){ - if(level<0) old_coeff= qmul*level - qadd; - else old_coeff= qmul*level + qadd; - run2= run_tab[rle_index++]; //FIXME ! maybe after last - }else{ - old_coeff=0; - run2--; - assert(run2>=0 || i >= last_non_zero ); - } - - for(change=-1; change<=1; change+=2){ - int new_level= level + change; - int score, new_coeff, unquant_change; - - score=0; - if(s->avctx->quantizer_noise_shaping < 2 && FFABS(new_level) > FFABS(level)) - continue; - - if(new_level){ - if(new_level<0) new_coeff= qmul*new_level - qadd; - else new_coeff= qmul*new_level + qadd; - if(new_coeff >= 2048 || new_coeff <= -2048) - continue; - //FIXME check for overflow - - if(level){ - if(level < 63 && level > -63){ - if(i < last_non_zero) - score += length[UNI_AC_ENC_INDEX(run, new_level+64)] - - length[UNI_AC_ENC_INDEX(run, level+64)]; - else - score += last_length[UNI_AC_ENC_INDEX(run, new_level+64)] - - last_length[UNI_AC_ENC_INDEX(run, level+64)]; - } - }else{ - assert(FFABS(new_level)==1); - - if(analyze_gradient){ - int g= d1[ scantable[i] ]; - if(g && (g^new_level) >= 0) - continue; - } - - if(i < last_non_zero){ - int next_i= i + run2 + 1; - int next_level= block[ perm_scantable[next_i] ] + 64; - - if(next_level&(~127)) - next_level= 0; - - if(next_i < last_non_zero) - score += length[UNI_AC_ENC_INDEX(run, 65)] - + length[UNI_AC_ENC_INDEX(run2, next_level)] - - length[UNI_AC_ENC_INDEX(run + run2 + 1, next_level)]; - else - score += length[UNI_AC_ENC_INDEX(run, 65)] - + last_length[UNI_AC_ENC_INDEX(run2, next_level)] - - last_length[UNI_AC_ENC_INDEX(run + run2 + 1, next_level)]; - }else{ - score += last_length[UNI_AC_ENC_INDEX(run, 65)]; - if(prev_level){ - score += length[UNI_AC_ENC_INDEX(prev_run, prev_level)] - - last_length[UNI_AC_ENC_INDEX(prev_run, prev_level)]; - } - } - } - }else{ - new_coeff=0; - assert(FFABS(level)==1); - - if(i < last_non_zero){ - int next_i= i + run2 + 1; - int next_level= block[ perm_scantable[next_i] ] + 64; - - if(next_level&(~127)) - next_level= 0; - - if(next_i < last_non_zero) - score += length[UNI_AC_ENC_INDEX(run + run2 + 1, next_level)] - - length[UNI_AC_ENC_INDEX(run2, next_level)] - - length[UNI_AC_ENC_INDEX(run, 65)]; - else - score += last_length[UNI_AC_ENC_INDEX(run + run2 + 1, next_level)] - - last_length[UNI_AC_ENC_INDEX(run2, next_level)] - - length[UNI_AC_ENC_INDEX(run, 65)]; - }else{ - score += -last_length[UNI_AC_ENC_INDEX(run, 65)]; - if(prev_level){ - score += last_length[UNI_AC_ENC_INDEX(prev_run, prev_level)] - - length[UNI_AC_ENC_INDEX(prev_run, prev_level)]; - } - } - } - - score *= lambda; - - unquant_change= new_coeff - old_coeff; - assert((score < 100*lambda && score > -100*lambda) || lambda==0); - - score+= s->dsp.try_8x8basis(rem, weight, basis[j], unquant_change); - if(score last_non_zero){ - last_non_zero= best_coeff; - assert(block[j]); -#ifdef REFINE_STATS -after_last++; -#endif - }else{ -#ifdef REFINE_STATS -if(block[j]){ - if(block[j] - best_change){ - if(FFABS(block[j]) > FFABS(block[j] - best_change)){ - raise++; - }else{ - lower++; - } - }else{ - from_zero++; - } -}else{ - to_zero++; -} -#endif - for(; last_non_zero>=start_i; last_non_zero--){ - if(block[perm_scantable[last_non_zero]]) - break; - } - } -#ifdef REFINE_STATS -count++; -if(256*256*256*64 % count == 0){ - printf("after_last:%d to_zero:%d from_zero:%d raise:%d lower:%d sign:%d xyp:%d/%d/%d\n", after_last, to_zero, from_zero, raise, lower, messed_sign, s->mb_x, s->mb_y, s->picture_number); -} -#endif - run=0; - rle_index=0; - for(i=start_i; i<=last_non_zero; i++){ - int j= perm_scantable[i]; - const int level= block[j]; - - if(level){ - run_tab[rle_index++]=run; - run=0; - }else{ - run++; - } - } - - s->dsp.add_8x8basis(rem, basis[j], best_unquant_change); - }else{ - break; - } - } -#ifdef REFINE_STATS -if(last_non_zero>0){ -STOP_TIMER("iterative search") -} -} -#endif - - return last_non_zero; -} - -static int dct_quantize_c(MpegEncContext *s, - DCTELEM *block, int n, - int qscale, int *overflow) -{ - int i, j, level, last_non_zero, q, start_i; - const int *qmat; - const uint8_t *scantable= s->intra_scantable.scantable; - int bias; - int max=0; - unsigned int threshold1, threshold2; - - s->dsp.fdct (block); - - if(s->dct_error_sum) - s->denoise_dct(s, block); - - if (s->mb_intra) { - if (!s->h263_aic) { - if (n < 4) - q = s->y_dc_scale; - else - q = s->c_dc_scale; - q = q << 3; - } else - /* For AIC we skip quant/dequant of INTRADC */ - q = 1 << 3; - - /* note: block[0] is assumed to be positive */ - block[0] = (block[0] + (q >> 1)) / q; - start_i = 1; - last_non_zero = 0; - qmat = s->q_intra_matrix[qscale]; - bias= s->intra_quant_bias<<(QMAT_SHIFT - QUANT_BIAS_SHIFT); - } else { - start_i = 0; - last_non_zero = -1; - qmat = s->q_inter_matrix[qscale]; - bias= s->inter_quant_bias<<(QMAT_SHIFT - QUANT_BIAS_SHIFT); - } - threshold1= (1<=start_i;i--) { - j = scantable[i]; - level = block[j] * qmat[j]; - - if(((unsigned)(level+threshold1))>threshold2){ - last_non_zero = i; - break; - }else{ - block[j]=0; - } - } - for(i=start_i; i<=last_non_zero; i++) { - j = scantable[i]; - level = block[j] * qmat[j]; - -// if( bias+level >= (1<= (1<threshold2){ - if(level>0){ - level= (bias + level)>>QMAT_SHIFT; - block[j]= level; - }else{ - level= (bias - level)>>QMAT_SHIFT; - block[j]= -level; - } - max |=level; - }else{ - block[j]=0; - } - } - *overflow= s->max_qcoeff < max; //overflow might have happened - - /* we need this permutation so that we correct the IDCT, we only permute the !=0 elements */ - if (s->dsp.idct_permutation_type != FF_NO_IDCT_PERM) - ff_block_permute(block, s->dsp.idct_permutation, scantable, last_non_zero); - - return last_non_zero; -} - -#endif //CONFIG_ENCODERS - static void dct_unquantize_mpeg1_intra_c(MpegEncContext *s, DCTELEM *block, int n, int qscale) { @@ -6820,116 +2365,3 @@ static void dct_unquantize_h263_inter_c(MpegEncContext *s, } } -#ifdef CONFIG_ENCODERS -AVCodec h263_encoder = { - "h263", - CODEC_TYPE_VIDEO, - CODEC_ID_H263, - sizeof(MpegEncContext), - MPV_encode_init, - MPV_encode_picture, - MPV_encode_end, - .pix_fmts= (enum PixelFormat[]){PIX_FMT_YUV420P, -1}, -}; - -AVCodec h263p_encoder = { - "h263p", - CODEC_TYPE_VIDEO, - CODEC_ID_H263P, - sizeof(MpegEncContext), - MPV_encode_init, - MPV_encode_picture, - MPV_encode_end, - .pix_fmts= (enum PixelFormat[]){PIX_FMT_YUV420P, -1}, -}; - -AVCodec flv_encoder = { - "flv", - CODEC_TYPE_VIDEO, - CODEC_ID_FLV1, - sizeof(MpegEncContext), - MPV_encode_init, - MPV_encode_picture, - MPV_encode_end, - .pix_fmts= (enum PixelFormat[]){PIX_FMT_YUV420P, -1}, -}; - -AVCodec rv10_encoder = { - "rv10", - CODEC_TYPE_VIDEO, - CODEC_ID_RV10, - sizeof(MpegEncContext), - MPV_encode_init, - MPV_encode_picture, - MPV_encode_end, - .pix_fmts= (enum PixelFormat[]){PIX_FMT_YUV420P, -1}, -}; - -AVCodec rv20_encoder = { - "rv20", - CODEC_TYPE_VIDEO, - CODEC_ID_RV20, - sizeof(MpegEncContext), - MPV_encode_init, - MPV_encode_picture, - MPV_encode_end, - .pix_fmts= (enum PixelFormat[]){PIX_FMT_YUV420P, -1}, -}; - -AVCodec mpeg4_encoder = { - "mpeg4", - CODEC_TYPE_VIDEO, - CODEC_ID_MPEG4, - sizeof(MpegEncContext), - MPV_encode_init, - MPV_encode_picture, - MPV_encode_end, - .pix_fmts= (enum PixelFormat[]){PIX_FMT_YUV420P, -1}, - .capabilities= CODEC_CAP_DELAY, -}; - -AVCodec msmpeg4v1_encoder = { - "msmpeg4v1", - CODEC_TYPE_VIDEO, - CODEC_ID_MSMPEG4V1, - sizeof(MpegEncContext), - MPV_encode_init, - MPV_encode_picture, - MPV_encode_end, - .pix_fmts= (enum PixelFormat[]){PIX_FMT_YUV420P, -1}, -}; - -AVCodec msmpeg4v2_encoder = { - "msmpeg4v2", - CODEC_TYPE_VIDEO, - CODEC_ID_MSMPEG4V2, - sizeof(MpegEncContext), - MPV_encode_init, - MPV_encode_picture, - MPV_encode_end, - .pix_fmts= (enum PixelFormat[]){PIX_FMT_YUV420P, -1}, -}; - -AVCodec msmpeg4v3_encoder = { - "msmpeg4", - CODEC_TYPE_VIDEO, - CODEC_ID_MSMPEG4V3, - sizeof(MpegEncContext), - MPV_encode_init, - MPV_encode_picture, - MPV_encode_end, - .pix_fmts= (enum PixelFormat[]){PIX_FMT_YUV420P, -1}, -}; - -AVCodec wmv1_encoder = { - "wmv1", - CODEC_TYPE_VIDEO, - CODEC_ID_WMV1, - sizeof(MpegEncContext), - MPV_encode_init, - MPV_encode_picture, - MPV_encode_end, - .pix_fmts= (enum PixelFormat[]){PIX_FMT_YUV420P, -1}, -}; - -#endif //CONFIG_ENCODERS diff --git a/libavcodec/mpegvideo_common.h b/libavcodec/mpegvideo_common.h new file mode 100644 index 0000000000..7fc1ee52b7 --- /dev/null +++ b/libavcodec/mpegvideo_common.h @@ -0,0 +1,836 @@ +/* + * The simplest mpeg encoder (well, it was the simplest!) + * Copyright (c) 2000,2001 Fabrice Bellard. + * Copyright (c) 2002-2004 Michael Niedermayer + * + * This file is part of FFmpeg. + * + * FFmpeg is free software; you can redistribute it and/or + * modify it under the terms of the GNU Lesser General Public + * License as published by the Free Software Foundation; either + * version 2.1 of the License, or (at your option) any later version. + * + * FFmpeg is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * Lesser General Public License for more details. + * + * You should have received a copy of the GNU Lesser General Public + * License along with FFmpeg; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA + * + * 4MV & hq & b-frame encoding stuff by Michael Niedermayer + */ + +/** + * @file mpegvideo_common.h + * The simplest mpeg encoder (well, it was the simplest!). + */ + +#ifndef MPEGVIDEO_COMMON_H +#define MPEGVIDEO_COMMON_H + +#include "avcodec.h" +#include "dsputil.h" +#include "mpegvideo.h" +#include "mjpegenc.h" +#include "msmpeg4.h" +#include "faandct.h" +#include + +int dct_quantize_c(MpegEncContext *s, DCTELEM *block, int n, int qscale, int *overflow); +int dct_quantize_trellis_c(MpegEncContext *s, DCTELEM *block, int n, int qscale, int *overflow); +void denoise_dct_c(MpegEncContext *s, DCTELEM *block); +void copy_picture(Picture *dst, Picture *src); + +/** + * allocates a Picture + * The pixels are allocated/set by calling get_buffer() if shared=0 + */ +int alloc_picture(MpegEncContext *s, Picture *pic, int shared); + +/** + * sets the given MpegEncContext to common defaults (same for encoding and decoding). + * the changed fields will not depend upon the prior state of the MpegEncContext. + */ +void MPV_common_defaults(MpegEncContext *s); + +static inline void gmc1_motion(MpegEncContext *s, + uint8_t *dest_y, uint8_t *dest_cb, uint8_t *dest_cr, + uint8_t **ref_picture) +{ + uint8_t *ptr; + int offset, src_x, src_y, linesize, uvlinesize; + int motion_x, motion_y; + int emu=0; + + motion_x= s->sprite_offset[0][0]; + motion_y= s->sprite_offset[0][1]; + src_x = s->mb_x * 16 + (motion_x >> (s->sprite_warping_accuracy+1)); + src_y = s->mb_y * 16 + (motion_y >> (s->sprite_warping_accuracy+1)); + motion_x<<=(3-s->sprite_warping_accuracy); + motion_y<<=(3-s->sprite_warping_accuracy); + src_x = av_clip(src_x, -16, s->width); + if (src_x == s->width) + motion_x =0; + src_y = av_clip(src_y, -16, s->height); + if (src_y == s->height) + motion_y =0; + + linesize = s->linesize; + uvlinesize = s->uvlinesize; + + ptr = ref_picture[0] + (src_y * linesize) + src_x; + + if(s->flags&CODEC_FLAG_EMU_EDGE){ + if( (unsigned)src_x >= s->h_edge_pos - 17 + || (unsigned)src_y >= s->v_edge_pos - 17){ + ff_emulated_edge_mc(s->edge_emu_buffer, ptr, linesize, 17, 17, src_x, src_y, s->h_edge_pos, s->v_edge_pos); + ptr= s->edge_emu_buffer; + } + } + + if((motion_x|motion_y)&7){ + s->dsp.gmc1(dest_y , ptr , linesize, 16, motion_x&15, motion_y&15, 128 - s->no_rounding); + s->dsp.gmc1(dest_y+8, ptr+8, linesize, 16, motion_x&15, motion_y&15, 128 - s->no_rounding); + }else{ + int dxy; + + dxy= ((motion_x>>3)&1) | ((motion_y>>2)&2); + if (s->no_rounding){ + s->dsp.put_no_rnd_pixels_tab[0][dxy](dest_y, ptr, linesize, 16); + }else{ + s->dsp.put_pixels_tab [0][dxy](dest_y, ptr, linesize, 16); + } + } + + if(s->flags&CODEC_FLAG_GRAY) return; + + motion_x= s->sprite_offset[1][0]; + motion_y= s->sprite_offset[1][1]; + src_x = s->mb_x * 8 + (motion_x >> (s->sprite_warping_accuracy+1)); + src_y = s->mb_y * 8 + (motion_y >> (s->sprite_warping_accuracy+1)); + motion_x<<=(3-s->sprite_warping_accuracy); + motion_y<<=(3-s->sprite_warping_accuracy); + src_x = av_clip(src_x, -8, s->width>>1); + if (src_x == s->width>>1) + motion_x =0; + src_y = av_clip(src_y, -8, s->height>>1); + if (src_y == s->height>>1) + motion_y =0; + + offset = (src_y * uvlinesize) + src_x; + ptr = ref_picture[1] + offset; + if(s->flags&CODEC_FLAG_EMU_EDGE){ + if( (unsigned)src_x >= (s->h_edge_pos>>1) - 9 + || (unsigned)src_y >= (s->v_edge_pos>>1) - 9){ + ff_emulated_edge_mc(s->edge_emu_buffer, ptr, uvlinesize, 9, 9, src_x, src_y, s->h_edge_pos>>1, s->v_edge_pos>>1); + ptr= s->edge_emu_buffer; + emu=1; + } + } + s->dsp.gmc1(dest_cb, ptr, uvlinesize, 8, motion_x&15, motion_y&15, 128 - s->no_rounding); + + ptr = ref_picture[2] + offset; + if(emu){ + ff_emulated_edge_mc(s->edge_emu_buffer, ptr, uvlinesize, 9, 9, src_x, src_y, s->h_edge_pos>>1, s->v_edge_pos>>1); + ptr= s->edge_emu_buffer; + } + s->dsp.gmc1(dest_cr, ptr, uvlinesize, 8, motion_x&15, motion_y&15, 128 - s->no_rounding); + + return; +} + +static inline void gmc_motion(MpegEncContext *s, + uint8_t *dest_y, uint8_t *dest_cb, uint8_t *dest_cr, + uint8_t **ref_picture) +{ + uint8_t *ptr; + int linesize, uvlinesize; + const int a= s->sprite_warping_accuracy; + int ox, oy; + + linesize = s->linesize; + uvlinesize = s->uvlinesize; + + ptr = ref_picture[0]; + + ox= s->sprite_offset[0][0] + s->sprite_delta[0][0]*s->mb_x*16 + s->sprite_delta[0][1]*s->mb_y*16; + oy= s->sprite_offset[0][1] + s->sprite_delta[1][0]*s->mb_x*16 + s->sprite_delta[1][1]*s->mb_y*16; + + s->dsp.gmc(dest_y, ptr, linesize, 16, + ox, + oy, + s->sprite_delta[0][0], s->sprite_delta[0][1], + s->sprite_delta[1][0], s->sprite_delta[1][1], + a+1, (1<<(2*a+1)) - s->no_rounding, + s->h_edge_pos, s->v_edge_pos); + s->dsp.gmc(dest_y+8, ptr, linesize, 16, + ox + s->sprite_delta[0][0]*8, + oy + s->sprite_delta[1][0]*8, + s->sprite_delta[0][0], s->sprite_delta[0][1], + s->sprite_delta[1][0], s->sprite_delta[1][1], + a+1, (1<<(2*a+1)) - s->no_rounding, + s->h_edge_pos, s->v_edge_pos); + + if(s->flags&CODEC_FLAG_GRAY) return; + + ox= s->sprite_offset[1][0] + s->sprite_delta[0][0]*s->mb_x*8 + s->sprite_delta[0][1]*s->mb_y*8; + oy= s->sprite_offset[1][1] + s->sprite_delta[1][0]*s->mb_x*8 + s->sprite_delta[1][1]*s->mb_y*8; + + ptr = ref_picture[1]; + s->dsp.gmc(dest_cb, ptr, uvlinesize, 8, + ox, + oy, + s->sprite_delta[0][0], s->sprite_delta[0][1], + s->sprite_delta[1][0], s->sprite_delta[1][1], + a+1, (1<<(2*a+1)) - s->no_rounding, + s->h_edge_pos>>1, s->v_edge_pos>>1); + + ptr = ref_picture[2]; + s->dsp.gmc(dest_cr, ptr, uvlinesize, 8, + ox, + oy, + s->sprite_delta[0][0], s->sprite_delta[0][1], + s->sprite_delta[1][0], s->sprite_delta[1][1], + a+1, (1<<(2*a+1)) - s->no_rounding, + s->h_edge_pos>>1, s->v_edge_pos>>1); +} + +static inline int hpel_motion(MpegEncContext *s, + uint8_t *dest, uint8_t *src, + int field_based, int field_select, + int src_x, int src_y, + int width, int height, int stride, + int h_edge_pos, int v_edge_pos, + int w, int h, op_pixels_func *pix_op, + int motion_x, int motion_y) +{ + int dxy; + int emu=0; + + dxy = ((motion_y & 1) << 1) | (motion_x & 1); + src_x += motion_x >> 1; + src_y += motion_y >> 1; + + /* WARNING: do no forget half pels */ + src_x = av_clip(src_x, -16, width); //FIXME unneeded for emu? + if (src_x == width) + dxy &= ~1; + src_y = av_clip(src_y, -16, height); + if (src_y == height) + dxy &= ~2; + src += src_y * stride + src_x; + + if(s->unrestricted_mv && (s->flags&CODEC_FLAG_EMU_EDGE)){ + if( (unsigned)src_x > h_edge_pos - (motion_x&1) - w + || (unsigned)src_y > v_edge_pos - (motion_y&1) - h){ + ff_emulated_edge_mc(s->edge_emu_buffer, src, s->linesize, w+1, (h+1)<v_edge_pos); + src= s->edge_emu_buffer; + emu=1; + } + } + if(field_select) + src += s->linesize; + pix_op[dxy](dest, src, stride, h); + return emu; +} + +/* apply one mpeg motion vector to the three components */ +static av_always_inline void mpeg_motion(MpegEncContext *s, + uint8_t *dest_y, uint8_t *dest_cb, uint8_t *dest_cr, + int field_based, int bottom_field, int field_select, + uint8_t **ref_picture, op_pixels_func (*pix_op)[4], + int motion_x, int motion_y, int h) +{ + uint8_t *ptr_y, *ptr_cb, *ptr_cr; + int dxy, uvdxy, mx, my, src_x, src_y, uvsrc_x, uvsrc_y, v_edge_pos, uvlinesize, linesize; + +#if 0 +if(s->quarter_sample) +{ + motion_x>>=1; + motion_y>>=1; +} +#endif + + v_edge_pos = s->v_edge_pos >> field_based; + linesize = s->current_picture.linesize[0] << field_based; + uvlinesize = s->current_picture.linesize[1] << field_based; + + dxy = ((motion_y & 1) << 1) | (motion_x & 1); + src_x = s->mb_x* 16 + (motion_x >> 1); + src_y =(s->mb_y<<(4-field_based)) + (motion_y >> 1); + + if (s->out_format == FMT_H263) { + if((s->workaround_bugs & FF_BUG_HPEL_CHROMA) && field_based){ + mx = (motion_x>>1)|(motion_x&1); + my = motion_y >>1; + uvdxy = ((my & 1) << 1) | (mx & 1); + uvsrc_x = s->mb_x* 8 + (mx >> 1); + uvsrc_y = (s->mb_y<<(3-field_based)) + (my >> 1); + }else{ + uvdxy = dxy | (motion_y & 2) | ((motion_x & 2) >> 1); + uvsrc_x = src_x>>1; + uvsrc_y = src_y>>1; + } + }else if(s->out_format == FMT_H261){//even chroma mv's are full pel in H261 + mx = motion_x / 4; + my = motion_y / 4; + uvdxy = 0; + uvsrc_x = s->mb_x*8 + mx; + uvsrc_y = s->mb_y*8 + my; + } else { + if(s->chroma_y_shift){ + mx = motion_x / 2; + my = motion_y / 2; + uvdxy = ((my & 1) << 1) | (mx & 1); + uvsrc_x = s->mb_x* 8 + (mx >> 1); + uvsrc_y = (s->mb_y<<(3-field_based)) + (my >> 1); + } else { + if(s->chroma_x_shift){ + //Chroma422 + mx = motion_x / 2; + uvdxy = ((motion_y & 1) << 1) | (mx & 1); + uvsrc_x = s->mb_x* 8 + (mx >> 1); + uvsrc_y = src_y; + } else { + //Chroma444 + uvdxy = dxy; + uvsrc_x = src_x; + uvsrc_y = src_y; + } + } + } + + ptr_y = ref_picture[0] + src_y * linesize + src_x; + ptr_cb = ref_picture[1] + uvsrc_y * uvlinesize + uvsrc_x; + ptr_cr = ref_picture[2] + uvsrc_y * uvlinesize + uvsrc_x; + + if( (unsigned)src_x > s->h_edge_pos - (motion_x&1) - 16 + || (unsigned)src_y > v_edge_pos - (motion_y&1) - h){ + if(s->codec_id == CODEC_ID_MPEG2VIDEO || + s->codec_id == CODEC_ID_MPEG1VIDEO){ + av_log(s->avctx,AV_LOG_DEBUG,"MPEG motion vector out of boundary\n"); + return ; + } + ff_emulated_edge_mc(s->edge_emu_buffer, ptr_y, s->linesize, 17, 17+field_based, + src_x, src_y<h_edge_pos, s->v_edge_pos); + ptr_y = s->edge_emu_buffer; + if(!(s->flags&CODEC_FLAG_GRAY)){ + uint8_t *uvbuf= s->edge_emu_buffer+18*s->linesize; + ff_emulated_edge_mc(uvbuf , ptr_cb, s->uvlinesize, 9, 9+field_based, + uvsrc_x, uvsrc_y<h_edge_pos>>1, s->v_edge_pos>>1); + ff_emulated_edge_mc(uvbuf+16, ptr_cr, s->uvlinesize, 9, 9+field_based, + uvsrc_x, uvsrc_y<h_edge_pos>>1, s->v_edge_pos>>1); + ptr_cb= uvbuf; + ptr_cr= uvbuf+16; + } + } + + if(bottom_field){ //FIXME use this for field pix too instead of the obnoxious hack which changes picture.data + dest_y += s->linesize; + dest_cb+= s->uvlinesize; + dest_cr+= s->uvlinesize; + } + + if(field_select){ + ptr_y += s->linesize; + ptr_cb+= s->uvlinesize; + ptr_cr+= s->uvlinesize; + } + + pix_op[0][dxy](dest_y, ptr_y, linesize, h); + + if(!(s->flags&CODEC_FLAG_GRAY)){ + pix_op[s->chroma_x_shift][uvdxy](dest_cb, ptr_cb, uvlinesize, h >> s->chroma_y_shift); + pix_op[s->chroma_x_shift][uvdxy](dest_cr, ptr_cr, uvlinesize, h >> s->chroma_y_shift); + } + if((ENABLE_H261_ENCODER || ENABLE_H261_DECODER) && s->out_format == FMT_H261){ + ff_h261_loop_filter(s); + } +} + +//FIXME move to dsputil, avg variant, 16x16 version +static inline void put_obmc(uint8_t *dst, uint8_t *src[5], int stride){ + int x; + uint8_t * const top = src[1]; + uint8_t * const left = src[2]; + uint8_t * const mid = src[0]; + uint8_t * const right = src[3]; + uint8_t * const bottom= src[4]; +#define OBMC_FILTER(x, t, l, m, r, b)\ + dst[x]= (t*top[x] + l*left[x] + m*mid[x] + r*right[x] + b*bottom[x] + 4)>>3 +#define OBMC_FILTER4(x, t, l, m, r, b)\ + OBMC_FILTER(x , t, l, m, r, b);\ + OBMC_FILTER(x+1 , t, l, m, r, b);\ + OBMC_FILTER(x +stride, t, l, m, r, b);\ + OBMC_FILTER(x+1+stride, t, l, m, r, b); + + x=0; + OBMC_FILTER (x , 2, 2, 4, 0, 0); + OBMC_FILTER (x+1, 2, 1, 5, 0, 0); + OBMC_FILTER4(x+2, 2, 1, 5, 0, 0); + OBMC_FILTER4(x+4, 2, 0, 5, 1, 0); + OBMC_FILTER (x+6, 2, 0, 5, 1, 0); + OBMC_FILTER (x+7, 2, 0, 4, 2, 0); + x+= stride; + OBMC_FILTER (x , 1, 2, 5, 0, 0); + OBMC_FILTER (x+1, 1, 2, 5, 0, 0); + OBMC_FILTER (x+6, 1, 0, 5, 2, 0); + OBMC_FILTER (x+7, 1, 0, 5, 2, 0); + x+= stride; + OBMC_FILTER4(x , 1, 2, 5, 0, 0); + OBMC_FILTER4(x+2, 1, 1, 6, 0, 0); + OBMC_FILTER4(x+4, 1, 0, 6, 1, 0); + OBMC_FILTER4(x+6, 1, 0, 5, 2, 0); + x+= 2*stride; + OBMC_FILTER4(x , 0, 2, 5, 0, 1); + OBMC_FILTER4(x+2, 0, 1, 6, 0, 1); + OBMC_FILTER4(x+4, 0, 0, 6, 1, 1); + OBMC_FILTER4(x+6, 0, 0, 5, 2, 1); + x+= 2*stride; + OBMC_FILTER (x , 0, 2, 5, 0, 1); + OBMC_FILTER (x+1, 0, 2, 5, 0, 1); + OBMC_FILTER4(x+2, 0, 1, 5, 0, 2); + OBMC_FILTER4(x+4, 0, 0, 5, 1, 2); + OBMC_FILTER (x+6, 0, 0, 5, 2, 1); + OBMC_FILTER (x+7, 0, 0, 5, 2, 1); + x+= stride; + OBMC_FILTER (x , 0, 2, 4, 0, 2); + OBMC_FILTER (x+1, 0, 1, 5, 0, 2); + OBMC_FILTER (x+6, 0, 0, 5, 1, 2); + OBMC_FILTER (x+7, 0, 0, 4, 2, 2); +} + +/* obmc for 1 8x8 luma block */ +static inline void obmc_motion(MpegEncContext *s, + uint8_t *dest, uint8_t *src, + int src_x, int src_y, + op_pixels_func *pix_op, + int16_t mv[5][2]/* mid top left right bottom*/) +#define MID 0 +{ + int i; + uint8_t *ptr[5]; + + assert(s->quarter_sample==0); + + for(i=0; i<5; i++){ + if(i && mv[i][0]==mv[MID][0] && mv[i][1]==mv[MID][1]){ + ptr[i]= ptr[MID]; + }else{ + ptr[i]= s->obmc_scratchpad + 8*(i&1) + s->linesize*8*(i>>1); + hpel_motion(s, ptr[i], src, 0, 0, + src_x, src_y, + s->width, s->height, s->linesize, + s->h_edge_pos, s->v_edge_pos, + 8, 8, pix_op, + mv[i][0], mv[i][1]); + } + } + + put_obmc(dest, ptr, s->linesize); +} + +static inline void qpel_motion(MpegEncContext *s, + uint8_t *dest_y, uint8_t *dest_cb, uint8_t *dest_cr, + int field_based, int bottom_field, int field_select, + uint8_t **ref_picture, op_pixels_func (*pix_op)[4], + qpel_mc_func (*qpix_op)[16], + int motion_x, int motion_y, int h) +{ + uint8_t *ptr_y, *ptr_cb, *ptr_cr; + int dxy, uvdxy, mx, my, src_x, src_y, uvsrc_x, uvsrc_y, v_edge_pos, linesize, uvlinesize; + + dxy = ((motion_y & 3) << 2) | (motion_x & 3); + src_x = s->mb_x * 16 + (motion_x >> 2); + src_y = s->mb_y * (16 >> field_based) + (motion_y >> 2); + + v_edge_pos = s->v_edge_pos >> field_based; + linesize = s->linesize << field_based; + uvlinesize = s->uvlinesize << field_based; + + if(field_based){ + mx= motion_x/2; + my= motion_y>>1; + }else if(s->workaround_bugs&FF_BUG_QPEL_CHROMA2){ + static const int rtab[8]= {0,0,1,1,0,0,0,1}; + mx= (motion_x>>1) + rtab[motion_x&7]; + my= (motion_y>>1) + rtab[motion_y&7]; + }else if(s->workaround_bugs&FF_BUG_QPEL_CHROMA){ + mx= (motion_x>>1)|(motion_x&1); + my= (motion_y>>1)|(motion_y&1); + }else{ + mx= motion_x/2; + my= motion_y/2; + } + mx= (mx>>1)|(mx&1); + my= (my>>1)|(my&1); + + uvdxy= (mx&1) | ((my&1)<<1); + mx>>=1; + my>>=1; + + uvsrc_x = s->mb_x * 8 + mx; + uvsrc_y = s->mb_y * (8 >> field_based) + my; + + ptr_y = ref_picture[0] + src_y * linesize + src_x; + ptr_cb = ref_picture[1] + uvsrc_y * uvlinesize + uvsrc_x; + ptr_cr = ref_picture[2] + uvsrc_y * uvlinesize + uvsrc_x; + + if( (unsigned)src_x > s->h_edge_pos - (motion_x&3) - 16 + || (unsigned)src_y > v_edge_pos - (motion_y&3) - h ){ + ff_emulated_edge_mc(s->edge_emu_buffer, ptr_y, s->linesize, 17, 17+field_based, + src_x, src_y<h_edge_pos, s->v_edge_pos); + ptr_y= s->edge_emu_buffer; + if(!(s->flags&CODEC_FLAG_GRAY)){ + uint8_t *uvbuf= s->edge_emu_buffer + 18*s->linesize; + ff_emulated_edge_mc(uvbuf, ptr_cb, s->uvlinesize, 9, 9 + field_based, + uvsrc_x, uvsrc_y<h_edge_pos>>1, s->v_edge_pos>>1); + ff_emulated_edge_mc(uvbuf + 16, ptr_cr, s->uvlinesize, 9, 9 + field_based, + uvsrc_x, uvsrc_y<h_edge_pos>>1, s->v_edge_pos>>1); + ptr_cb= uvbuf; + ptr_cr= uvbuf + 16; + } + } + + if(!field_based) + qpix_op[0][dxy](dest_y, ptr_y, linesize); + else{ + if(bottom_field){ + dest_y += s->linesize; + dest_cb+= s->uvlinesize; + dest_cr+= s->uvlinesize; + } + + if(field_select){ + ptr_y += s->linesize; + ptr_cb += s->uvlinesize; + ptr_cr += s->uvlinesize; + } + //damn interlaced mode + //FIXME boundary mirroring is not exactly correct here + qpix_op[1][dxy](dest_y , ptr_y , linesize); + qpix_op[1][dxy](dest_y+8, ptr_y+8, linesize); + } + if(!(s->flags&CODEC_FLAG_GRAY)){ + pix_op[1][uvdxy](dest_cr, ptr_cr, uvlinesize, h >> 1); + pix_op[1][uvdxy](dest_cb, ptr_cb, uvlinesize, h >> 1); + } +} + +/** + * h263 chorma 4mv motion compensation. + */ +static inline void chroma_4mv_motion(MpegEncContext *s, + uint8_t *dest_cb, uint8_t *dest_cr, + uint8_t **ref_picture, + op_pixels_func *pix_op, + int mx, int my){ + int dxy, emu=0, src_x, src_y, offset; + uint8_t *ptr; + + /* In case of 8X8, we construct a single chroma motion vector + with a special rounding */ + mx= ff_h263_round_chroma(mx); + my= ff_h263_round_chroma(my); + + dxy = ((my & 1) << 1) | (mx & 1); + mx >>= 1; + my >>= 1; + + src_x = s->mb_x * 8 + mx; + src_y = s->mb_y * 8 + my; + src_x = av_clip(src_x, -8, s->width/2); + if (src_x == s->width/2) + dxy &= ~1; + src_y = av_clip(src_y, -8, s->height/2); + if (src_y == s->height/2) + dxy &= ~2; + + offset = (src_y * (s->uvlinesize)) + src_x; + ptr = ref_picture[1] + offset; + if(s->flags&CODEC_FLAG_EMU_EDGE){ + if( (unsigned)src_x > (s->h_edge_pos>>1) - (dxy &1) - 8 + || (unsigned)src_y > (s->v_edge_pos>>1) - (dxy>>1) - 8){ + ff_emulated_edge_mc(s->edge_emu_buffer, ptr, s->uvlinesize, 9, 9, src_x, src_y, s->h_edge_pos>>1, s->v_edge_pos>>1); + ptr= s->edge_emu_buffer; + emu=1; + } + } + pix_op[dxy](dest_cb, ptr, s->uvlinesize, 8); + + ptr = ref_picture[2] + offset; + if(emu){ + ff_emulated_edge_mc(s->edge_emu_buffer, ptr, s->uvlinesize, 9, 9, src_x, src_y, s->h_edge_pos>>1, s->v_edge_pos>>1); + ptr= s->edge_emu_buffer; + } + pix_op[dxy](dest_cr, ptr, s->uvlinesize, 8); +} + +static inline void prefetch_motion(MpegEncContext *s, uint8_t **pix, int dir){ + /* fetch pixels for estimated mv 4 macroblocks ahead + * optimized for 64byte cache lines */ + const int shift = s->quarter_sample ? 2 : 1; + const int mx= (s->mv[dir][0][0]>>shift) + 16*s->mb_x + 8; + const int my= (s->mv[dir][0][1]>>shift) + 16*s->mb_y; + int off= mx + (my + (s->mb_x&3)*4)*s->linesize + 64; + s->dsp.prefetch(pix[0]+off, s->linesize, 4); + off= (mx>>1) + ((my>>1) + (s->mb_x&7))*s->uvlinesize + 64; + s->dsp.prefetch(pix[1]+off, pix[2]-pix[1], 2); +} + +/** + * motion compensation of a single macroblock + * @param s context + * @param dest_y luma destination pointer + * @param dest_cb chroma cb/u destination pointer + * @param dest_cr chroma cr/v destination pointer + * @param dir direction (0->forward, 1->backward) + * @param ref_picture array[3] of pointers to the 3 planes of the reference picture + * @param pic_op halfpel motion compensation function (average or put normally) + * @param pic_op qpel motion compensation function (average or put normally) + * the motion vectors are taken from s->mv and the MV type from s->mv_type + */ +static inline void MPV_motion(MpegEncContext *s, + uint8_t *dest_y, uint8_t *dest_cb, uint8_t *dest_cr, + int dir, uint8_t **ref_picture, + op_pixels_func (*pix_op)[4], qpel_mc_func (*qpix_op)[16]) +{ + int dxy, mx, my, src_x, src_y, motion_x, motion_y; + int mb_x, mb_y, i; + uint8_t *ptr, *dest; + + mb_x = s->mb_x; + mb_y = s->mb_y; + + prefetch_motion(s, ref_picture, dir); + + if(s->obmc && s->pict_type != B_TYPE){ + int16_t mv_cache[4][4][2]; + const int xy= s->mb_x + s->mb_y*s->mb_stride; + const int mot_stride= s->b8_stride; + const int mot_xy= mb_x*2 + mb_y*2*mot_stride; + + assert(!s->mb_skipped); + + memcpy(mv_cache[1][1], s->current_picture.motion_val[0][mot_xy ], sizeof(int16_t)*4); + memcpy(mv_cache[2][1], s->current_picture.motion_val[0][mot_xy+mot_stride], sizeof(int16_t)*4); + memcpy(mv_cache[3][1], s->current_picture.motion_val[0][mot_xy+mot_stride], sizeof(int16_t)*4); + + if(mb_y==0 || IS_INTRA(s->current_picture.mb_type[xy-s->mb_stride])){ + memcpy(mv_cache[0][1], mv_cache[1][1], sizeof(int16_t)*4); + }else{ + memcpy(mv_cache[0][1], s->current_picture.motion_val[0][mot_xy-mot_stride], sizeof(int16_t)*4); + } + + if(mb_x==0 || IS_INTRA(s->current_picture.mb_type[xy-1])){ + *(int32_t*)mv_cache[1][0]= *(int32_t*)mv_cache[1][1]; + *(int32_t*)mv_cache[2][0]= *(int32_t*)mv_cache[2][1]; + }else{ + *(int32_t*)mv_cache[1][0]= *(int32_t*)s->current_picture.motion_val[0][mot_xy-1]; + *(int32_t*)mv_cache[2][0]= *(int32_t*)s->current_picture.motion_val[0][mot_xy-1+mot_stride]; + } + + if(mb_x+1>=s->mb_width || IS_INTRA(s->current_picture.mb_type[xy+1])){ + *(int32_t*)mv_cache[1][3]= *(int32_t*)mv_cache[1][2]; + *(int32_t*)mv_cache[2][3]= *(int32_t*)mv_cache[2][2]; + }else{ + *(int32_t*)mv_cache[1][3]= *(int32_t*)s->current_picture.motion_val[0][mot_xy+2]; + *(int32_t*)mv_cache[2][3]= *(int32_t*)s->current_picture.motion_val[0][mot_xy+2+mot_stride]; + } + + mx = 0; + my = 0; + for(i=0;i<4;i++) { + const int x= (i&1)+1; + const int y= (i>>1)+1; + int16_t mv[5][2]= { + {mv_cache[y][x ][0], mv_cache[y][x ][1]}, + {mv_cache[y-1][x][0], mv_cache[y-1][x][1]}, + {mv_cache[y][x-1][0], mv_cache[y][x-1][1]}, + {mv_cache[y][x+1][0], mv_cache[y][x+1][1]}, + {mv_cache[y+1][x][0], mv_cache[y+1][x][1]}}; + //FIXME cleanup + obmc_motion(s, dest_y + ((i & 1) * 8) + (i >> 1) * 8 * s->linesize, + ref_picture[0], + mb_x * 16 + (i & 1) * 8, mb_y * 16 + (i >>1) * 8, + pix_op[1], + mv); + + mx += mv[0][0]; + my += mv[0][1]; + } + if(!(s->flags&CODEC_FLAG_GRAY)) + chroma_4mv_motion(s, dest_cb, dest_cr, ref_picture, pix_op[1], mx, my); + + return; + } + + switch(s->mv_type) { + case MV_TYPE_16X16: + if(s->mcsel){ + if(s->real_sprite_warping_points==1){ + gmc1_motion(s, dest_y, dest_cb, dest_cr, + ref_picture); + }else{ + gmc_motion(s, dest_y, dest_cb, dest_cr, + ref_picture); + } + }else if(s->quarter_sample){ + qpel_motion(s, dest_y, dest_cb, dest_cr, + 0, 0, 0, + ref_picture, pix_op, qpix_op, + s->mv[dir][0][0], s->mv[dir][0][1], 16); + }else if(ENABLE_WMV2 && s->mspel){ + ff_mspel_motion(s, dest_y, dest_cb, dest_cr, + ref_picture, pix_op, + s->mv[dir][0][0], s->mv[dir][0][1], 16); + }else + { + mpeg_motion(s, dest_y, dest_cb, dest_cr, + 0, 0, 0, + ref_picture, pix_op, + s->mv[dir][0][0], s->mv[dir][0][1], 16); + } + break; + case MV_TYPE_8X8: + mx = 0; + my = 0; + if(s->quarter_sample){ + for(i=0;i<4;i++) { + motion_x = s->mv[dir][i][0]; + motion_y = s->mv[dir][i][1]; + + dxy = ((motion_y & 3) << 2) | (motion_x & 3); + src_x = mb_x * 16 + (motion_x >> 2) + (i & 1) * 8; + src_y = mb_y * 16 + (motion_y >> 2) + (i >>1) * 8; + + /* WARNING: do no forget half pels */ + src_x = av_clip(src_x, -16, s->width); + if (src_x == s->width) + dxy &= ~3; + src_y = av_clip(src_y, -16, s->height); + if (src_y == s->height) + dxy &= ~12; + + ptr = ref_picture[0] + (src_y * s->linesize) + (src_x); + if(s->flags&CODEC_FLAG_EMU_EDGE){ + if( (unsigned)src_x > s->h_edge_pos - (motion_x&3) - 8 + || (unsigned)src_y > s->v_edge_pos - (motion_y&3) - 8 ){ + ff_emulated_edge_mc(s->edge_emu_buffer, ptr, s->linesize, 9, 9, src_x, src_y, s->h_edge_pos, s->v_edge_pos); + ptr= s->edge_emu_buffer; + } + } + dest = dest_y + ((i & 1) * 8) + (i >> 1) * 8 * s->linesize; + qpix_op[1][dxy](dest, ptr, s->linesize); + + mx += s->mv[dir][i][0]/2; + my += s->mv[dir][i][1]/2; + } + }else{ + for(i=0;i<4;i++) { + hpel_motion(s, dest_y + ((i & 1) * 8) + (i >> 1) * 8 * s->linesize, + ref_picture[0], 0, 0, + mb_x * 16 + (i & 1) * 8, mb_y * 16 + (i >>1) * 8, + s->width, s->height, s->linesize, + s->h_edge_pos, s->v_edge_pos, + 8, 8, pix_op[1], + s->mv[dir][i][0], s->mv[dir][i][1]); + + mx += s->mv[dir][i][0]; + my += s->mv[dir][i][1]; + } + } + + if(!(s->flags&CODEC_FLAG_GRAY)) + chroma_4mv_motion(s, dest_cb, dest_cr, ref_picture, pix_op[1], mx, my); + break; + case MV_TYPE_FIELD: + if (s->picture_structure == PICT_FRAME) { + if(s->quarter_sample){ + for(i=0; i<2; i++){ + qpel_motion(s, dest_y, dest_cb, dest_cr, + 1, i, s->field_select[dir][i], + ref_picture, pix_op, qpix_op, + s->mv[dir][i][0], s->mv[dir][i][1], 8); + } + }else{ + /* top field */ + mpeg_motion(s, dest_y, dest_cb, dest_cr, + 1, 0, s->field_select[dir][0], + ref_picture, pix_op, + s->mv[dir][0][0], s->mv[dir][0][1], 8); + /* bottom field */ + mpeg_motion(s, dest_y, dest_cb, dest_cr, + 1, 1, s->field_select[dir][1], + ref_picture, pix_op, + s->mv[dir][1][0], s->mv[dir][1][1], 8); + } + } else { + if(s->picture_structure != s->field_select[dir][0] + 1 && s->pict_type != B_TYPE && !s->first_field){ + ref_picture= s->current_picture_ptr->data; + } + + mpeg_motion(s, dest_y, dest_cb, dest_cr, + 0, 0, s->field_select[dir][0], + ref_picture, pix_op, + s->mv[dir][0][0], s->mv[dir][0][1], 16); + } + break; + case MV_TYPE_16X8: + for(i=0; i<2; i++){ + uint8_t ** ref2picture; + + if(s->picture_structure == s->field_select[dir][i] + 1 || s->pict_type == B_TYPE || s->first_field){ + ref2picture= ref_picture; + }else{ + ref2picture= s->current_picture_ptr->data; + } + + mpeg_motion(s, dest_y, dest_cb, dest_cr, + 0, 0, s->field_select[dir][i], + ref2picture, pix_op, + s->mv[dir][i][0], s->mv[dir][i][1] + 16*i, 8); + + dest_y += 16*s->linesize; + dest_cb+= (16>>s->chroma_y_shift)*s->uvlinesize; + dest_cr+= (16>>s->chroma_y_shift)*s->uvlinesize; + } + break; + case MV_TYPE_DMV: + if(s->picture_structure == PICT_FRAME){ + for(i=0; i<2; i++){ + int j; + for(j=0; j<2; j++){ + mpeg_motion(s, dest_y, dest_cb, dest_cr, + 1, j, j^i, + ref_picture, pix_op, + s->mv[dir][2*i + j][0], s->mv[dir][2*i + j][1], 8); + } + pix_op = s->dsp.avg_pixels_tab; + } + }else{ + for(i=0; i<2; i++){ + mpeg_motion(s, dest_y, dest_cb, dest_cr, + 0, 0, s->picture_structure != i+1, + ref_picture, pix_op, + s->mv[dir][2*i][0],s->mv[dir][2*i][1],16); + + // after put we make avg of the same block + pix_op=s->dsp.avg_pixels_tab; + + //opposite parity is always in the same frame if this is second field + if(!s->first_field){ + ref_picture = s->current_picture_ptr->data; + } + } + } + break; + default: assert(0); + } +} + +#endif /* MPEGVIDEO_COMMON_H */ diff --git a/libavcodec/mpegvideo_enc.c b/libavcodec/mpegvideo_enc.c new file mode 100644 index 0000000000..c418b47df0 --- /dev/null +++ b/libavcodec/mpegvideo_enc.c @@ -0,0 +1,3796 @@ +/* + * The simplest mpeg encoder (well, it was the simplest!) + * Copyright (c) 2000,2001 Fabrice Bellard. + * Copyright (c) 2002-2004 Michael Niedermayer + * + * This file is part of FFmpeg. + * + * FFmpeg is free software; you can redistribute it and/or + * modify it under the terms of the GNU Lesser General Public + * License as published by the Free Software Foundation; either + * version 2.1 of the License, or (at your option) any later version. + * + * FFmpeg is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * Lesser General Public License for more details. + * + * You should have received a copy of the GNU Lesser General Public + * License along with FFmpeg; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA + * + * 4MV & hq & b-frame encoding stuff by Michael Niedermayer + */ + +/** + * @file mpegvideo_enc.c + * The simplest mpeg encoder (well, it was the simplest!). + */ + +#include "avcodec.h" +#include "dsputil.h" +#include "mpegvideo.h" +#include "mpegvideo_common.h" +#include "mjpegenc.h" +#include "msmpeg4.h" +#include "faandct.h" +#include + +//#undef NDEBUG +//#include + +static int encode_picture(MpegEncContext *s, int picture_number); +static int dct_quantize_refine(MpegEncContext *s, DCTELEM *block, int16_t *weight, DCTELEM *orig, int n, int qscale); +static int sse_mb(MpegEncContext *s); + +/* enable all paranoid tests for rounding, overflows, etc... */ +//#define PARANOID + +//#define DEBUG + +static const uint16_t aanscales[64] = { + /* precomputed values scaled up by 14 bits */ + 16384, 22725, 21407, 19266, 16384, 12873, 8867, 4520, + 22725, 31521, 29692, 26722, 22725, 17855, 12299, 6270, + 21407, 29692, 27969, 25172, 21407, 16819, 11585, 5906, + 19266, 26722, 25172, 22654, 19266, 15137, 10426, 5315, + 16384, 22725, 21407, 19266, 16384, 12873, 8867, 4520, + 12873, 17855, 16819, 15137, 12873, 10114, 6967, 3552, + 8867 , 12299, 11585, 10426, 8867, 6967, 4799, 2446, + 4520 , 6270, 5906, 5315, 4520, 3552, 2446, 1247 +}; + +static uint8_t default_mv_penalty[MAX_FCODE+1][MAX_MV*2+1]; +static uint8_t default_fcode_tab[MAX_MV*2+1]; + +enum PixelFormat ff_yuv420p_list[2]= {PIX_FMT_YUV420P, -1}; + +static void convert_matrix(DSPContext *dsp, int (*qmat)[64], uint16_t (*qmat16)[2][64], + const uint16_t *quant_matrix, int bias, int qmin, int qmax, int intra) +{ + int qscale; + int shift=0; + + for(qscale=qmin; qscale<=qmax; qscale++){ + int i; + if (dsp->fdct == ff_jpeg_fdct_islow +#ifdef FAAN_POSTSCALE + || dsp->fdct == ff_faandct +#endif + ) { + for(i=0;i<64;i++) { + const int j= dsp->idct_permutation[i]; + /* 16 <= qscale * quant_matrix[i] <= 7905 */ + /* 19952 <= aanscales[i] * qscale * quant_matrix[i] <= 249205026 */ + /* (1<<36)/19952 >= (1<<36)/(aanscales[i] * qscale * quant_matrix[i]) >= (1<<36)/249205026 */ + /* 3444240 >= (1<<36)/(aanscales[i] * qscale * quant_matrix[i]) >= 275 */ + + qmat[qscale][i] = (int)((UINT64_C(1) << QMAT_SHIFT) / + (qscale * quant_matrix[j])); + } + } else if (dsp->fdct == fdct_ifast +#ifndef FAAN_POSTSCALE + || dsp->fdct == ff_faandct +#endif + ) { + for(i=0;i<64;i++) { + const int j= dsp->idct_permutation[i]; + /* 16 <= qscale * quant_matrix[i] <= 7905 */ + /* 19952 <= aanscales[i] * qscale * quant_matrix[i] <= 249205026 */ + /* (1<<36)/19952 >= (1<<36)/(aanscales[i] * qscale * quant_matrix[i]) >= (1<<36)/249205026 */ + /* 3444240 >= (1<<36)/(aanscales[i] * qscale * quant_matrix[i]) >= 275 */ + + qmat[qscale][i] = (int)((UINT64_C(1) << (QMAT_SHIFT + 14)) / + (aanscales[i] * qscale * quant_matrix[j])); + } + } else { + for(i=0;i<64;i++) { + const int j= dsp->idct_permutation[i]; + /* We can safely suppose that 16 <= quant_matrix[i] <= 255 + So 16 <= qscale * quant_matrix[i] <= 7905 + so (1<<19) / 16 >= (1<<19) / (qscale * quant_matrix[i]) >= (1<<19) / 7905 + so 32768 >= (1<<19) / (qscale * quant_matrix[i]) >= 67 + */ + qmat[qscale][i] = (int)((UINT64_C(1) << QMAT_SHIFT) / (qscale * quant_matrix[j])); +// qmat [qscale][i] = (1 << QMAT_SHIFT_MMX) / (qscale * quant_matrix[i]); + qmat16[qscale][0][i] = (1 << QMAT_SHIFT_MMX) / (qscale * quant_matrix[j]); + + if(qmat16[qscale][0][i]==0 || qmat16[qscale][0][i]==128*256) qmat16[qscale][0][i]=128*256-1; + qmat16[qscale][1][i]= ROUNDED_DIV(bias<<(16-QUANT_BIAS_SHIFT), qmat16[qscale][0][i]); + } + } + + for(i=intra; i<64; i++){ + int64_t max= 8191; + if (dsp->fdct == fdct_ifast +#ifndef FAAN_POSTSCALE + || dsp->fdct == ff_faandct +#endif + ) { + max= (8191LL*aanscales[i]) >> 14; + } + while(((max * qmat[qscale][i]) >> shift) > INT_MAX){ + shift++; + } + } + } + if(shift){ + av_log(NULL, AV_LOG_INFO, "Warning, QMAT_SHIFT is larger than %d, overflows possible\n", QMAT_SHIFT - shift); + } +} + +static inline void update_qscale(MpegEncContext *s){ + s->qscale= (s->lambda*139 + FF_LAMBDA_SCALE*64) >> (FF_LAMBDA_SHIFT + 7); + s->qscale= av_clip(s->qscale, s->avctx->qmin, s->avctx->qmax); + + s->lambda2= (s->lambda*s->lambda + FF_LAMBDA_SCALE/2) >> FF_LAMBDA_SHIFT; +} + +void ff_write_quant_matrix(PutBitContext *pb, uint16_t *matrix){ + int i; + + if(matrix){ + put_bits(pb, 1, 1); + for(i=0;i<64;i++) { + put_bits(pb, 8, matrix[ ff_zigzag_direct[i] ]); + } + }else + put_bits(pb, 1, 0); +} + +static void copy_picture_attributes(MpegEncContext *s, AVFrame *dst, AVFrame *src){ + int i; + + dst->pict_type = src->pict_type; + dst->quality = src->quality; + dst->coded_picture_number = src->coded_picture_number; + dst->display_picture_number = src->display_picture_number; +// dst->reference = src->reference; + dst->pts = src->pts; + dst->interlaced_frame = src->interlaced_frame; + dst->top_field_first = src->top_field_first; + + if(s->avctx->me_threshold){ + if(!src->motion_val[0]) + av_log(s->avctx, AV_LOG_ERROR, "AVFrame.motion_val not set!\n"); + if(!src->mb_type) + av_log(s->avctx, AV_LOG_ERROR, "AVFrame.mb_type not set!\n"); + if(!src->ref_index[0]) + av_log(s->avctx, AV_LOG_ERROR, "AVFrame.ref_index not set!\n"); + if(src->motion_subsample_log2 != dst->motion_subsample_log2) + av_log(s->avctx, AV_LOG_ERROR, "AVFrame.motion_subsample_log2 doesn't match! (%d!=%d)\n", + src->motion_subsample_log2, dst->motion_subsample_log2); + + memcpy(dst->mb_type, src->mb_type, s->mb_stride * s->mb_height * sizeof(dst->mb_type[0])); + + for(i=0; i<2; i++){ + int stride= ((16*s->mb_width )>>src->motion_subsample_log2) + 1; + int height= ((16*s->mb_height)>>src->motion_subsample_log2); + + if(src->motion_val[i] && src->motion_val[i] != dst->motion_val[i]){ + memcpy(dst->motion_val[i], src->motion_val[i], 2*stride*height*sizeof(int16_t)); + } + if(src->ref_index[i] && src->ref_index[i] != dst->ref_index[i]){ + memcpy(dst->ref_index[i], src->ref_index[i], s->b8_stride*2*s->mb_height*sizeof(int8_t)); + } + } + } +} + +static void update_duplicate_context_after_me(MpegEncContext *dst, MpegEncContext *src){ +#define COPY(a) dst->a= src->a + COPY(pict_type); + COPY(current_picture); + COPY(f_code); + COPY(b_code); + COPY(qscale); + COPY(lambda); + COPY(lambda2); + COPY(picture_in_gop_number); + COPY(gop_picture_number); + COPY(frame_pred_frame_dct); //FIXME don't set in encode_header + COPY(progressive_frame); //FIXME don't set in encode_header + COPY(partitioned_frame); //FIXME don't set in encode_header +#undef COPY +} + +/** + * sets the given MpegEncContext to defaults for encoding. + * the changed fields will not depend upon the prior state of the MpegEncContext. + */ +static void MPV_encode_defaults(MpegEncContext *s){ + int i; + MPV_common_defaults(s); + + for(i=-16; i<16; i++){ + default_fcode_tab[i + MAX_MV]= 1; + } + s->me.mv_penalty= default_mv_penalty; + s->fcode_tab= default_fcode_tab; +} + +/* init video encoder */ +int MPV_encode_init(AVCodecContext *avctx) +{ + MpegEncContext *s = avctx->priv_data; + int i; + int chroma_h_shift, chroma_v_shift; + + MPV_encode_defaults(s); + + switch (avctx->codec_id) { + case CODEC_ID_MPEG2VIDEO: + if(avctx->pix_fmt != PIX_FMT_YUV420P && avctx->pix_fmt != PIX_FMT_YUV422P){ + av_log(avctx, AV_LOG_ERROR, "only YUV420 and YUV422 are supported\n"); + return -1; + } + break; + case CODEC_ID_LJPEG: + case CODEC_ID_MJPEG: + if(avctx->pix_fmt != PIX_FMT_YUVJ420P && avctx->pix_fmt != PIX_FMT_YUVJ422P && + ((avctx->pix_fmt != PIX_FMT_YUV420P && avctx->pix_fmt != PIX_FMT_YUV422P) || avctx->strict_std_compliance>FF_COMPLIANCE_INOFFICIAL)){ + av_log(avctx, AV_LOG_ERROR, "colorspace not supported in jpeg\n"); + return -1; + } + break; + default: + if(avctx->pix_fmt != PIX_FMT_YUV420P){ + av_log(avctx, AV_LOG_ERROR, "only YUV420 is supported\n"); + return -1; + } + } + + switch (avctx->pix_fmt) { + case PIX_FMT_YUVJ422P: + case PIX_FMT_YUV422P: + s->chroma_format = CHROMA_422; + break; + case PIX_FMT_YUVJ420P: + case PIX_FMT_YUV420P: + default: + s->chroma_format = CHROMA_420; + break; + } + + s->bit_rate = avctx->bit_rate; + s->width = avctx->width; + s->height = avctx->height; + if(avctx->gop_size > 600 && avctx->strict_std_compliance>FF_COMPLIANCE_EXPERIMENTAL){ + av_log(avctx, AV_LOG_ERROR, "Warning keyframe interval too large! reducing it ...\n"); + avctx->gop_size=600; + } + s->gop_size = avctx->gop_size; + s->avctx = avctx; + s->flags= avctx->flags; + s->flags2= avctx->flags2; + s->max_b_frames= avctx->max_b_frames; + s->codec_id= avctx->codec->id; + s->luma_elim_threshold = avctx->luma_elim_threshold; + s->chroma_elim_threshold= avctx->chroma_elim_threshold; + s->strict_std_compliance= avctx->strict_std_compliance; + s->data_partitioning= avctx->flags & CODEC_FLAG_PART; + s->quarter_sample= (avctx->flags & CODEC_FLAG_QPEL)!=0; + s->mpeg_quant= avctx->mpeg_quant; + s->rtp_mode= !!avctx->rtp_payload_size; + s->intra_dc_precision= avctx->intra_dc_precision; + s->user_specified_pts = AV_NOPTS_VALUE; + + if (s->gop_size <= 1) { + s->intra_only = 1; + s->gop_size = 12; + } else { + s->intra_only = 0; + } + + s->me_method = avctx->me_method; + + /* Fixed QSCALE */ + s->fixed_qscale = !!(avctx->flags & CODEC_FLAG_QSCALE); + + s->adaptive_quant= ( s->avctx->lumi_masking + || s->avctx->dark_masking + || s->avctx->temporal_cplx_masking + || s->avctx->spatial_cplx_masking + || s->avctx->p_masking + || s->avctx->border_masking + || (s->flags&CODEC_FLAG_QP_RD)) + && !s->fixed_qscale; + + s->obmc= !!(s->flags & CODEC_FLAG_OBMC); + s->loop_filter= !!(s->flags & CODEC_FLAG_LOOP_FILTER); + s->alternate_scan= !!(s->flags & CODEC_FLAG_ALT_SCAN); + s->intra_vlc_format= !!(s->flags2 & CODEC_FLAG2_INTRA_VLC); + s->q_scale_type= !!(s->flags2 & CODEC_FLAG2_NON_LINEAR_QUANT); + + if(avctx->rc_max_rate && !avctx->rc_buffer_size){ + av_log(avctx, AV_LOG_ERROR, "a vbv buffer size is needed, for encoding with a maximum bitrate\n"); + return -1; + } + + if(avctx->rc_min_rate && avctx->rc_max_rate != avctx->rc_min_rate){ + av_log(avctx, AV_LOG_INFO, "Warning min_rate > 0 but min_rate != max_rate isn't recommended!\n"); + } + + if(avctx->rc_min_rate && avctx->rc_min_rate > avctx->bit_rate){ + av_log(avctx, AV_LOG_ERROR, "bitrate below min bitrate\n"); + return -1; + } + + if(avctx->rc_max_rate && avctx->rc_max_rate < avctx->bit_rate){ + av_log(avctx, AV_LOG_INFO, "bitrate above max bitrate\n"); + return -1; + } + + if(avctx->rc_buffer_size && avctx->bit_rate*av_q2d(avctx->time_base) > avctx->rc_buffer_size){ + av_log(avctx, AV_LOG_ERROR, "VBV buffer too small for bitrate\n"); + return -1; + } + + if(avctx->bit_rate*av_q2d(avctx->time_base) > avctx->bit_rate_tolerance){ + av_log(avctx, AV_LOG_ERROR, "bitrate tolerance too small for bitrate\n"); + return -1; + } + + if( s->avctx->rc_max_rate && s->avctx->rc_min_rate == s->avctx->rc_max_rate + && (s->codec_id == CODEC_ID_MPEG1VIDEO || s->codec_id == CODEC_ID_MPEG2VIDEO) + && 90000LL * (avctx->rc_buffer_size-1) > s->avctx->rc_max_rate*0xFFFFLL){ + + av_log(avctx, AV_LOG_INFO, "Warning vbv_delay will be set to 0xFFFF (=VBR) as the specified vbv buffer is too large for the given bitrate!\n"); + } + + if((s->flags & CODEC_FLAG_4MV) && s->codec_id != CODEC_ID_MPEG4 + && s->codec_id != CODEC_ID_H263 && s->codec_id != CODEC_ID_H263P && s->codec_id != CODEC_ID_FLV1){ + av_log(avctx, AV_LOG_ERROR, "4MV not supported by codec\n"); + return -1; + } + + if(s->obmc && s->avctx->mb_decision != FF_MB_DECISION_SIMPLE){ + av_log(avctx, AV_LOG_ERROR, "OBMC is only supported with simple mb decision\n"); + return -1; + } + + if(s->obmc && s->codec_id != CODEC_ID_H263 && s->codec_id != CODEC_ID_H263P){ + av_log(avctx, AV_LOG_ERROR, "OBMC is only supported with H263(+)\n"); + return -1; + } + + if(s->quarter_sample && s->codec_id != CODEC_ID_MPEG4){ + av_log(avctx, AV_LOG_ERROR, "qpel not supported by codec\n"); + return -1; + } + + if(s->data_partitioning && s->codec_id != CODEC_ID_MPEG4){ + av_log(avctx, AV_LOG_ERROR, "data partitioning not supported by codec\n"); + return -1; + } + + if(s->max_b_frames && s->codec_id != CODEC_ID_MPEG4 && s->codec_id != CODEC_ID_MPEG1VIDEO && s->codec_id != CODEC_ID_MPEG2VIDEO){ + av_log(avctx, AV_LOG_ERROR, "b frames not supported by codec\n"); + return -1; + } + + if((s->flags & (CODEC_FLAG_INTERLACED_DCT|CODEC_FLAG_INTERLACED_ME|CODEC_FLAG_ALT_SCAN)) + && s->codec_id != CODEC_ID_MPEG4 && s->codec_id != CODEC_ID_MPEG2VIDEO){ + av_log(avctx, AV_LOG_ERROR, "interlacing not supported by codec\n"); + return -1; + } + + if(s->mpeg_quant && s->codec_id != CODEC_ID_MPEG4){ //FIXME mpeg2 uses that too + av_log(avctx, AV_LOG_ERROR, "mpeg2 style quantization not supported by codec\n"); + return -1; + } + + if((s->flags & CODEC_FLAG_CBP_RD) && !(s->flags & CODEC_FLAG_TRELLIS_QUANT)){ + av_log(avctx, AV_LOG_ERROR, "CBP RD needs trellis quant\n"); + return -1; + } + + if((s->flags & CODEC_FLAG_QP_RD) && s->avctx->mb_decision != FF_MB_DECISION_RD){ + av_log(avctx, AV_LOG_ERROR, "QP RD needs mbd=2\n"); + return -1; + } + + if(s->avctx->scenechange_threshold < 1000000000 && (s->flags & CODEC_FLAG_CLOSED_GOP)){ + av_log(avctx, AV_LOG_ERROR, "closed gop with scene change detection arent supported yet, set threshold to 1000000000\n"); + return -1; + } + + if((s->flags2 & CODEC_FLAG2_INTRA_VLC) && s->codec_id != CODEC_ID_MPEG2VIDEO){ + av_log(avctx, AV_LOG_ERROR, "intra vlc table not supported by codec\n"); + return -1; + } + + if(s->flags & CODEC_FLAG_LOW_DELAY){ + if (s->codec_id != CODEC_ID_MPEG2VIDEO && s->codec_id != CODEC_ID_MPEG1VIDEO){ + av_log(avctx, AV_LOG_ERROR, "low delay forcing is only available for mpeg1/2\n"); + return -1; + } + if (s->max_b_frames != 0){ + av_log(avctx, AV_LOG_ERROR, "b frames cannot be used with low delay\n"); + return -1; + } + } + + if(s->q_scale_type == 1){ + if(s->codec_id != CODEC_ID_MPEG2VIDEO){ + av_log(avctx, AV_LOG_ERROR, "non linear quant is only available for mpeg2\n"); + return -1; + } + if(avctx->qmax > 12){ + av_log(avctx, AV_LOG_ERROR, "non linear quant only supports qmax <= 12 currently\n"); + return -1; + } + } + + if(s->avctx->thread_count > 1 && s->codec_id != CODEC_ID_MPEG4 + && s->codec_id != CODEC_ID_MPEG1VIDEO && s->codec_id != CODEC_ID_MPEG2VIDEO + && (s->codec_id != CODEC_ID_H263P || !(s->flags & CODEC_FLAG_H263P_SLICE_STRUCT))){ + av_log(avctx, AV_LOG_ERROR, "multi threaded encoding not supported by codec\n"); + return -1; + } + + if(s->avctx->thread_count > 1) + s->rtp_mode= 1; + + if(!avctx->time_base.den || !avctx->time_base.num){ + av_log(avctx, AV_LOG_ERROR, "framerate not set\n"); + return -1; + } + + i= (INT_MAX/2+128)>>8; + if(avctx->me_threshold >= i){ + av_log(avctx, AV_LOG_ERROR, "me_threshold too large, max is %d\n", i - 1); + return -1; + } + if(avctx->mb_threshold >= i){ + av_log(avctx, AV_LOG_ERROR, "mb_threshold too large, max is %d\n", i - 1); + return -1; + } + + if(avctx->b_frame_strategy && (avctx->flags&CODEC_FLAG_PASS2)){ + av_log(avctx, AV_LOG_INFO, "notice: b_frame_strategy only affects the first pass\n"); + avctx->b_frame_strategy = 0; + } + + i= ff_gcd(avctx->time_base.den, avctx->time_base.num); + if(i > 1){ + av_log(avctx, AV_LOG_INFO, "removing common factors from framerate\n"); + avctx->time_base.den /= i; + avctx->time_base.num /= i; +// return -1; + } + + if(s->codec_id==CODEC_ID_MJPEG){ + s->intra_quant_bias= 1<<(QUANT_BIAS_SHIFT-1); //(a + x/2)/x + s->inter_quant_bias= 0; + }else if(s->mpeg_quant || s->codec_id==CODEC_ID_MPEG1VIDEO || s->codec_id==CODEC_ID_MPEG2VIDEO){ + s->intra_quant_bias= 3<<(QUANT_BIAS_SHIFT-3); //(a + x*3/8)/x + s->inter_quant_bias= 0; + }else{ + s->intra_quant_bias=0; + s->inter_quant_bias=-(1<<(QUANT_BIAS_SHIFT-2)); //(a - x/4)/x + } + + if(avctx->intra_quant_bias != FF_DEFAULT_QUANT_BIAS) + s->intra_quant_bias= avctx->intra_quant_bias; + if(avctx->inter_quant_bias != FF_DEFAULT_QUANT_BIAS) + s->inter_quant_bias= avctx->inter_quant_bias; + + avcodec_get_chroma_sub_sample(avctx->pix_fmt, &chroma_h_shift, &chroma_v_shift); + + if(avctx->codec_id == CODEC_ID_MPEG4 && s->avctx->time_base.den > (1<<16)-1){ + av_log(avctx, AV_LOG_ERROR, "timebase not supported by mpeg 4 standard\n"); + return -1; + } + s->time_increment_bits = av_log2(s->avctx->time_base.den - 1) + 1; + + switch(avctx->codec->id) { + case CODEC_ID_MPEG1VIDEO: + s->out_format = FMT_MPEG1; + s->low_delay= !!(s->flags & CODEC_FLAG_LOW_DELAY); + avctx->delay= s->low_delay ? 0 : (s->max_b_frames + 1); + break; + case CODEC_ID_MPEG2VIDEO: + s->out_format = FMT_MPEG1; + s->low_delay= !!(s->flags & CODEC_FLAG_LOW_DELAY); + avctx->delay= s->low_delay ? 0 : (s->max_b_frames + 1); + s->rtp_mode= 1; + break; + case CODEC_ID_LJPEG: + case CODEC_ID_MJPEG: + s->out_format = FMT_MJPEG; + s->intra_only = 1; /* force intra only for jpeg */ + s->mjpeg_vsample[0] = 2; + s->mjpeg_vsample[1] = 2>>chroma_v_shift; + s->mjpeg_vsample[2] = 2>>chroma_v_shift; + s->mjpeg_hsample[0] = 2; + s->mjpeg_hsample[1] = 2>>chroma_h_shift; + s->mjpeg_hsample[2] = 2>>chroma_h_shift; + if (!(ENABLE_MJPEG_ENCODER || ENABLE_LJPEG_ENCODER) + || ff_mjpeg_encode_init(s) < 0) + return -1; + avctx->delay=0; + s->low_delay=1; + break; + case CODEC_ID_H261: + if (!ENABLE_H261_ENCODER) return -1; + if (ff_h261_get_picture_format(s->width, s->height) < 0) { + av_log(avctx, AV_LOG_ERROR, "The specified picture size of %dx%d is not valid for the H.261 codec.\nValid sizes are 176x144, 352x288\n", s->width, s->height); + return -1; + } + s->out_format = FMT_H261; + avctx->delay=0; + s->low_delay=1; + break; + case CODEC_ID_H263: + if (h263_get_picture_format(s->width, s->height) == 7) { + av_log(avctx, AV_LOG_INFO, "The specified picture size of %dx%d is not valid for the H.263 codec.\nValid sizes are 128x96, 176x144, 352x288, 704x576, and 1408x1152. Try H.263+.\n", s->width, s->height); + return -1; + } + s->out_format = FMT_H263; + s->obmc= (avctx->flags & CODEC_FLAG_OBMC) ? 1:0; + avctx->delay=0; + s->low_delay=1; + break; + case CODEC_ID_H263P: + s->out_format = FMT_H263; + s->h263_plus = 1; + /* Fx */ + s->umvplus = (avctx->flags & CODEC_FLAG_H263P_UMV) ? 1:0; + s->h263_aic= (avctx->flags & CODEC_FLAG_AC_PRED) ? 1:0; + s->modified_quant= s->h263_aic; + s->alt_inter_vlc= (avctx->flags & CODEC_FLAG_H263P_AIV) ? 1:0; + s->obmc= (avctx->flags & CODEC_FLAG_OBMC) ? 1:0; + s->loop_filter= (avctx->flags & CODEC_FLAG_LOOP_FILTER) ? 1:0; + s->unrestricted_mv= s->obmc || s->loop_filter || s->umvplus; + s->h263_slice_structured= (s->flags & CODEC_FLAG_H263P_SLICE_STRUCT) ? 1:0; + + /* /Fx */ + /* These are just to be sure */ + avctx->delay=0; + s->low_delay=1; + break; + case CODEC_ID_FLV1: + s->out_format = FMT_H263; + s->h263_flv = 2; /* format = 1; 11-bit codes */ + s->unrestricted_mv = 1; + s->rtp_mode=0; /* don't allow GOB */ + avctx->delay=0; + s->low_delay=1; + break; + case CODEC_ID_RV10: + s->out_format = FMT_H263; + avctx->delay=0; + s->low_delay=1; + break; + case CODEC_ID_RV20: + s->out_format = FMT_H263; + avctx->delay=0; + s->low_delay=1; + s->modified_quant=1; + s->h263_aic=1; + s->h263_plus=1; + s->loop_filter=1; + s->unrestricted_mv= s->obmc || s->loop_filter || s->umvplus; + break; + case CODEC_ID_MPEG4: + s->out_format = FMT_H263; + s->h263_pred = 1; + s->unrestricted_mv = 1; + s->low_delay= s->max_b_frames ? 0 : 1; + avctx->delay= s->low_delay ? 0 : (s->max_b_frames + 1); + break; + case CODEC_ID_MSMPEG4V1: + s->out_format = FMT_H263; + s->h263_msmpeg4 = 1; + s->h263_pred = 1; + s->unrestricted_mv = 1; + s->msmpeg4_version= 1; + avctx->delay=0; + s->low_delay=1; + break; + case CODEC_ID_MSMPEG4V2: + s->out_format = FMT_H263; + s->h263_msmpeg4 = 1; + s->h263_pred = 1; + s->unrestricted_mv = 1; + s->msmpeg4_version= 2; + avctx->delay=0; + s->low_delay=1; + break; + case CODEC_ID_MSMPEG4V3: + s->out_format = FMT_H263; + s->h263_msmpeg4 = 1; + s->h263_pred = 1; + s->unrestricted_mv = 1; + s->msmpeg4_version= 3; + s->flipflop_rounding=1; + avctx->delay=0; + s->low_delay=1; + break; + case CODEC_ID_WMV1: + s->out_format = FMT_H263; + s->h263_msmpeg4 = 1; + s->h263_pred = 1; + s->unrestricted_mv = 1; + s->msmpeg4_version= 4; + s->flipflop_rounding=1; + avctx->delay=0; + s->low_delay=1; + break; + case CODEC_ID_WMV2: + s->out_format = FMT_H263; + s->h263_msmpeg4 = 1; + s->h263_pred = 1; + s->unrestricted_mv = 1; + s->msmpeg4_version= 5; + s->flipflop_rounding=1; + avctx->delay=0; + s->low_delay=1; + break; + default: + return -1; + } + + avctx->has_b_frames= !s->low_delay; + + s->encoding = 1; + + /* init */ + if (MPV_common_init(s) < 0) + return -1; + + if(s->modified_quant) + s->chroma_qscale_table= ff_h263_chroma_qscale_table; + s->progressive_frame= + s->progressive_sequence= !(avctx->flags & (CODEC_FLAG_INTERLACED_DCT|CODEC_FLAG_INTERLACED_ME|CODEC_FLAG_ALT_SCAN)); + s->quant_precision=5; + + ff_set_cmp(&s->dsp, s->dsp.ildct_cmp, s->avctx->ildct_cmp); + ff_set_cmp(&s->dsp, s->dsp.frame_skip_cmp, s->avctx->frame_skip_cmp); + + if (ENABLE_H261_ENCODER && s->out_format == FMT_H261) + ff_h261_encode_init(s); + if (s->out_format == FMT_H263) + h263_encode_init(s); + if (ENABLE_MSMPEG4_ENCODER && s->msmpeg4_version) + ff_msmpeg4_encode_init(s); + if (s->out_format == FMT_MPEG1) + ff_mpeg1_encode_init(s); + + /* init q matrix */ + for(i=0;i<64;i++) { + int j= s->dsp.idct_permutation[i]; + if(s->codec_id==CODEC_ID_MPEG4 && s->mpeg_quant){ + s->intra_matrix[j] = ff_mpeg4_default_intra_matrix[i]; + s->inter_matrix[j] = ff_mpeg4_default_non_intra_matrix[i]; + }else if(s->out_format == FMT_H263 || s->out_format == FMT_H261){ + s->intra_matrix[j] = + s->inter_matrix[j] = ff_mpeg1_default_non_intra_matrix[i]; + }else + { /* mpeg1/2 */ + s->intra_matrix[j] = ff_mpeg1_default_intra_matrix[i]; + s->inter_matrix[j] = ff_mpeg1_default_non_intra_matrix[i]; + } + if(s->avctx->intra_matrix) + s->intra_matrix[j] = s->avctx->intra_matrix[i]; + if(s->avctx->inter_matrix) + s->inter_matrix[j] = s->avctx->inter_matrix[i]; + } + + /* precompute matrix */ + /* for mjpeg, we do include qscale in the matrix */ + if (s->out_format != FMT_MJPEG) { + convert_matrix(&s->dsp, s->q_intra_matrix, s->q_intra_matrix16, + s->intra_matrix, s->intra_quant_bias, avctx->qmin, 31, 1); + convert_matrix(&s->dsp, s->q_inter_matrix, s->q_inter_matrix16, + s->inter_matrix, s->inter_quant_bias, avctx->qmin, 31, 0); + } + + if(ff_rate_control_init(s) < 0) + return -1; + + return 0; +} + +int MPV_encode_end(AVCodecContext *avctx) +{ + MpegEncContext *s = avctx->priv_data; + + ff_rate_control_uninit(s); + + MPV_common_end(s); + if ((ENABLE_MJPEG_ENCODER || ENABLE_LJPEG_ENCODER) && s->out_format == FMT_MJPEG) + ff_mjpeg_encode_close(s); + + av_freep(&avctx->extradata); + + return 0; +} + +static int get_sae(uint8_t *src, int ref, int stride){ + int x,y; + int acc=0; + + for(y=0; y<16; y++){ + for(x=0; x<16; x++){ + acc+= FFABS(src[x+y*stride] - ref); + } + } + + return acc; +} + +static int get_intra_count(MpegEncContext *s, uint8_t *src, uint8_t *ref, int stride){ + int x, y, w, h; + int acc=0; + + w= s->width &~15; + h= s->height&~15; + + for(y=0; ydsp.sad[0](NULL, src + offset, ref + offset, stride, 16); + int mean= (s->dsp.pix_sum(src + offset, stride) + 128)>>8; + int sae = get_sae(src + offset, mean, stride); + + acc+= sae + 500 < sad; + } + } + return acc; +} + + +static int load_input_picture(MpegEncContext *s, AVFrame *pic_arg){ + AVFrame *pic=NULL; + int64_t pts; + int i; + const int encoding_delay= s->max_b_frames; + int direct=1; + + if(pic_arg){ + pts= pic_arg->pts; + pic_arg->display_picture_number= s->input_picture_number++; + + if(pts != AV_NOPTS_VALUE){ + if(s->user_specified_pts != AV_NOPTS_VALUE){ + int64_t time= pts; + int64_t last= s->user_specified_pts; + + if(time <= last){ + av_log(s->avctx, AV_LOG_ERROR, "Error, Invalid timestamp=%"PRId64", last=%"PRId64"\n", pts, s->user_specified_pts); + return -1; + } + } + s->user_specified_pts= pts; + }else{ + if(s->user_specified_pts != AV_NOPTS_VALUE){ + s->user_specified_pts= + pts= s->user_specified_pts + 1; + av_log(s->avctx, AV_LOG_INFO, "Warning: AVFrame.pts=? trying to guess (%"PRId64")\n", pts); + }else{ + pts= pic_arg->display_picture_number; + } + } + } + + if(pic_arg){ + if(encoding_delay && !(s->flags&CODEC_FLAG_INPUT_PRESERVED)) direct=0; + if(pic_arg->linesize[0] != s->linesize) direct=0; + if(pic_arg->linesize[1] != s->uvlinesize) direct=0; + if(pic_arg->linesize[2] != s->uvlinesize) direct=0; + +// av_log(AV_LOG_DEBUG, "%d %d %d %d\n",pic_arg->linesize[0], pic_arg->linesize[1], s->linesize, s->uvlinesize); + + if(direct){ + i= ff_find_unused_picture(s, 1); + + pic= (AVFrame*)&s->picture[i]; + pic->reference= 3; + + for(i=0; i<4; i++){ + pic->data[i]= pic_arg->data[i]; + pic->linesize[i]= pic_arg->linesize[i]; + } + alloc_picture(s, (Picture*)pic, 1); + }else{ + i= ff_find_unused_picture(s, 0); + + pic= (AVFrame*)&s->picture[i]; + pic->reference= 3; + + alloc_picture(s, (Picture*)pic, 0); + + if( pic->data[0] + INPLACE_OFFSET == pic_arg->data[0] + && pic->data[1] + INPLACE_OFFSET == pic_arg->data[1] + && pic->data[2] + INPLACE_OFFSET == pic_arg->data[2]){ + // empty + }else{ + int h_chroma_shift, v_chroma_shift; + avcodec_get_chroma_sub_sample(s->avctx->pix_fmt, &h_chroma_shift, &v_chroma_shift); + + for(i=0; i<3; i++){ + int src_stride= pic_arg->linesize[i]; + int dst_stride= i ? s->uvlinesize : s->linesize; + int h_shift= i ? h_chroma_shift : 0; + int v_shift= i ? v_chroma_shift : 0; + int w= s->width >>h_shift; + int h= s->height>>v_shift; + uint8_t *src= pic_arg->data[i]; + uint8_t *dst= pic->data[i]; + + if(!s->avctx->rc_buffer_size) + dst +=INPLACE_OFFSET; + + if(src_stride==dst_stride) + memcpy(dst, src, src_stride*h); + else{ + while(h--){ + memcpy(dst, src, w); + dst += dst_stride; + src += src_stride; + } + } + } + } + } + copy_picture_attributes(s, pic, pic_arg); + pic->pts= pts; //we set this here to avoid modifiying pic_arg + } + + /* shift buffer entries */ + for(i=1; iencoding_delay+1*/; i++) + s->input_picture[i-1]= s->input_picture[i]; + + s->input_picture[encoding_delay]= (Picture*)pic; + + return 0; +} + +static int skip_check(MpegEncContext *s, Picture *p, Picture *ref){ + int x, y, plane; + int score=0; + int64_t score64=0; + + for(plane=0; plane<3; plane++){ + const int stride= p->linesize[plane]; + const int bw= plane ? 1 : 2; + for(y=0; ymb_height*bw; y++){ + for(x=0; xmb_width*bw; x++){ + int off= p->type == FF_BUFFER_TYPE_SHARED ? 0: 16; + int v= s->dsp.frame_skip_cmp[1](s, p->data[plane] + 8*(x + y*stride)+off, ref->data[plane] + 8*(x + y*stride), stride, 8); + + switch(s->avctx->frame_skip_exp){ + case 0: score= FFMAX(score, v); break; + case 1: score+= FFABS(v);break; + case 2: score+= v*v;break; + case 3: score64+= FFABS(v*v*(int64_t)v);break; + case 4: score64+= v*v*(int64_t)(v*v);break; + } + } + } + } + + if(score) score64= score; + + if(score64 < s->avctx->frame_skip_threshold) + return 1; + if(score64 < ((s->avctx->frame_skip_factor * (int64_t)s->lambda)>>8)) + return 1; + return 0; +} + +static int estimate_best_b_count(MpegEncContext *s){ + AVCodec *codec= avcodec_find_encoder(s->avctx->codec_id); + AVCodecContext *c= avcodec_alloc_context(); + AVFrame input[FF_MAX_B_FRAMES+2]; + const int scale= s->avctx->brd_scale; + int i, j, out_size, p_lambda, b_lambda, lambda2; + int outbuf_size= s->width * s->height; //FIXME + uint8_t *outbuf= av_malloc(outbuf_size); + int64_t best_rd= INT64_MAX; + int best_b_count= -1; + + assert(scale>=0 && scale <=3); + +// emms_c(); + p_lambda= s->last_lambda_for[P_TYPE]; //s->next_picture_ptr->quality; + b_lambda= s->last_lambda_for[B_TYPE]; //p_lambda *FFABS(s->avctx->b_quant_factor) + s->avctx->b_quant_offset; + if(!b_lambda) b_lambda= p_lambda; //FIXME we should do this somewhere else + lambda2= (b_lambda*b_lambda + (1<> FF_LAMBDA_SHIFT; + + c->width = s->width >> scale; + c->height= s->height>> scale; + c->flags= CODEC_FLAG_QSCALE | CODEC_FLAG_PSNR | CODEC_FLAG_INPUT_PRESERVED /*| CODEC_FLAG_EMU_EDGE*/; + c->flags|= s->avctx->flags & CODEC_FLAG_QPEL; + c->mb_decision= s->avctx->mb_decision; + c->me_cmp= s->avctx->me_cmp; + c->mb_cmp= s->avctx->mb_cmp; + c->me_sub_cmp= s->avctx->me_sub_cmp; + c->pix_fmt = PIX_FMT_YUV420P; + c->time_base= s->avctx->time_base; + c->max_b_frames= s->max_b_frames; + + if (avcodec_open(c, codec) < 0) + return -1; + + for(i=0; imax_b_frames+2; i++){ + int ysize= c->width*c->height; + int csize= (c->width/2)*(c->height/2); + Picture pre_input, *pre_input_ptr= i ? s->input_picture[i-1] : s->next_picture_ptr; + + avcodec_get_frame_defaults(&input[i]); + input[i].data[0]= av_malloc(ysize + 2*csize); + input[i].data[1]= input[i].data[0] + ysize; + input[i].data[2]= input[i].data[1] + csize; + input[i].linesize[0]= c->width; + input[i].linesize[1]= + input[i].linesize[2]= c->width/2; + + if(pre_input_ptr && (!i || s->input_picture[i-1])) { + pre_input= *pre_input_ptr; + + if(pre_input.type != FF_BUFFER_TYPE_SHARED && i) { + pre_input.data[0]+=INPLACE_OFFSET; + pre_input.data[1]+=INPLACE_OFFSET; + pre_input.data[2]+=INPLACE_OFFSET; + } + + s->dsp.shrink[scale](input[i].data[0], input[i].linesize[0], pre_input.data[0], pre_input.linesize[0], c->width, c->height); + s->dsp.shrink[scale](input[i].data[1], input[i].linesize[1], pre_input.data[1], pre_input.linesize[1], c->width>>1, c->height>>1); + s->dsp.shrink[scale](input[i].data[2], input[i].linesize[2], pre_input.data[2], pre_input.linesize[2], c->width>>1, c->height>>1); + } + } + + for(j=0; jmax_b_frames+1; j++){ + int64_t rd=0; + + if(!s->input_picture[j]) + break; + + c->error[0]= c->error[1]= c->error[2]= 0; + + input[0].pict_type= I_TYPE; + input[0].quality= 1 * FF_QP2LAMBDA; + out_size = avcodec_encode_video(c, outbuf, outbuf_size, &input[0]); +// rd += (out_size * lambda2) >> FF_LAMBDA_SHIFT; + + for(i=0; imax_b_frames+1; i++){ + int is_p= i % (j+1) == j || i==s->max_b_frames; + + input[i+1].pict_type= is_p ? P_TYPE : B_TYPE; + input[i+1].quality= is_p ? p_lambda : b_lambda; + out_size = avcodec_encode_video(c, outbuf, outbuf_size, &input[i+1]); + rd += (out_size * lambda2) >> (FF_LAMBDA_SHIFT - 3); + } + + /* get the delayed frames */ + while(out_size){ + out_size = avcodec_encode_video(c, outbuf, outbuf_size, NULL); + rd += (out_size * lambda2) >> (FF_LAMBDA_SHIFT - 3); + } + + rd += c->error[0] + c->error[1] + c->error[2]; + + if(rd < best_rd){ + best_rd= rd; + best_b_count= j; + } + } + + av_freep(&outbuf); + avcodec_close(c); + av_freep(&c); + + for(i=0; imax_b_frames+2; i++){ + av_freep(&input[i].data[0]); + } + + return best_b_count; +} + +static void select_input_picture(MpegEncContext *s){ + int i; + + for(i=1; ireordered_input_picture[i-1]= s->reordered_input_picture[i]; + s->reordered_input_picture[MAX_PICTURE_COUNT-1]= NULL; + + /* set next picture type & ordering */ + if(s->reordered_input_picture[0]==NULL && s->input_picture[0]){ + if(/*s->picture_in_gop_number >= s->gop_size ||*/ s->next_picture_ptr==NULL || s->intra_only){ + s->reordered_input_picture[0]= s->input_picture[0]; + s->reordered_input_picture[0]->pict_type= I_TYPE; + s->reordered_input_picture[0]->coded_picture_number= s->coded_picture_number++; + }else{ + int b_frames; + + if(s->avctx->frame_skip_threshold || s->avctx->frame_skip_factor){ + if(s->picture_in_gop_number < s->gop_size && skip_check(s, s->input_picture[0], s->next_picture_ptr)){ + //FIXME check that te gop check above is +-1 correct +//av_log(NULL, AV_LOG_DEBUG, "skip %p %"PRId64"\n", s->input_picture[0]->data[0], s->input_picture[0]->pts); + + if(s->input_picture[0]->type == FF_BUFFER_TYPE_SHARED){ + for(i=0; i<4; i++) + s->input_picture[0]->data[i]= NULL; + s->input_picture[0]->type= 0; + }else{ + assert( s->input_picture[0]->type==FF_BUFFER_TYPE_USER + || s->input_picture[0]->type==FF_BUFFER_TYPE_INTERNAL); + + s->avctx->release_buffer(s->avctx, (AVFrame*)s->input_picture[0]); + } + + emms_c(); + ff_vbv_update(s, 0); + + goto no_output_pic; + } + } + + if(s->flags&CODEC_FLAG_PASS2){ + for(i=0; imax_b_frames+1; i++){ + int pict_num= s->input_picture[0]->display_picture_number + i; + + if(pict_num >= s->rc_context.num_entries) + break; + if(!s->input_picture[i]){ + s->rc_context.entry[pict_num-1].new_pict_type = P_TYPE; + break; + } + + s->input_picture[i]->pict_type= + s->rc_context.entry[pict_num].new_pict_type; + } + } + + if(s->avctx->b_frame_strategy==0){ + b_frames= s->max_b_frames; + while(b_frames && !s->input_picture[b_frames]) b_frames--; + }else if(s->avctx->b_frame_strategy==1){ + for(i=1; imax_b_frames+1; i++){ + if(s->input_picture[i] && s->input_picture[i]->b_frame_score==0){ + s->input_picture[i]->b_frame_score= + get_intra_count(s, s->input_picture[i ]->data[0], + s->input_picture[i-1]->data[0], s->linesize) + 1; + } + } + for(i=0; imax_b_frames+1; i++){ + if(s->input_picture[i]==NULL || s->input_picture[i]->b_frame_score - 1 > s->mb_num/s->avctx->b_sensitivity) break; + } + + b_frames= FFMAX(0, i-1); + + /* reset scores */ + for(i=0; iinput_picture[i]->b_frame_score=0; + } + }else if(s->avctx->b_frame_strategy==2){ + b_frames= estimate_best_b_count(s); + }else{ + av_log(s->avctx, AV_LOG_ERROR, "illegal b frame strategy\n"); + b_frames=0; + } + + emms_c(); +//static int b_count=0; +//b_count+= b_frames; +//av_log(s->avctx, AV_LOG_DEBUG, "b_frames: %d\n", b_count); + + for(i= b_frames - 1; i>=0; i--){ + int type= s->input_picture[i]->pict_type; + if(type && type != B_TYPE) + b_frames= i; + } + if(s->input_picture[b_frames]->pict_type == B_TYPE && b_frames == s->max_b_frames){ + av_log(s->avctx, AV_LOG_ERROR, "warning, too many b frames in a row\n"); + } + + if(s->picture_in_gop_number + b_frames >= s->gop_size){ + if((s->flags2 & CODEC_FLAG2_STRICT_GOP) && s->gop_size > s->picture_in_gop_number){ + b_frames= s->gop_size - s->picture_in_gop_number - 1; + }else{ + if(s->flags & CODEC_FLAG_CLOSED_GOP) + b_frames=0; + s->input_picture[b_frames]->pict_type= I_TYPE; + } + } + + if( (s->flags & CODEC_FLAG_CLOSED_GOP) + && b_frames + && s->input_picture[b_frames]->pict_type== I_TYPE) + b_frames--; + + s->reordered_input_picture[0]= s->input_picture[b_frames]; + if(s->reordered_input_picture[0]->pict_type != I_TYPE) + s->reordered_input_picture[0]->pict_type= P_TYPE; + s->reordered_input_picture[0]->coded_picture_number= s->coded_picture_number++; + for(i=0; ireordered_input_picture[i+1]= s->input_picture[i]; + s->reordered_input_picture[i+1]->pict_type= B_TYPE; + s->reordered_input_picture[i+1]->coded_picture_number= s->coded_picture_number++; + } + } + } +no_output_pic: + if(s->reordered_input_picture[0]){ + s->reordered_input_picture[0]->reference= s->reordered_input_picture[0]->pict_type!=B_TYPE ? 3 : 0; + + copy_picture(&s->new_picture, s->reordered_input_picture[0]); + + if(s->reordered_input_picture[0]->type == FF_BUFFER_TYPE_SHARED || s->avctx->rc_buffer_size){ + // input is a shared pix, so we can't modifiy it -> alloc a new one & ensure that the shared one is reuseable + + int i= ff_find_unused_picture(s, 0); + Picture *pic= &s->picture[i]; + + pic->reference = s->reordered_input_picture[0]->reference; + alloc_picture(s, pic, 0); + + /* mark us unused / free shared pic */ + if(s->reordered_input_picture[0]->type == FF_BUFFER_TYPE_INTERNAL) + s->avctx->release_buffer(s->avctx, (AVFrame*)s->reordered_input_picture[0]); + for(i=0; i<4; i++) + s->reordered_input_picture[0]->data[i]= NULL; + s->reordered_input_picture[0]->type= 0; + + copy_picture_attributes(s, (AVFrame*)pic, (AVFrame*)s->reordered_input_picture[0]); + + s->current_picture_ptr= pic; + }else{ + // input is not a shared pix -> reuse buffer for current_pix + + assert( s->reordered_input_picture[0]->type==FF_BUFFER_TYPE_USER + || s->reordered_input_picture[0]->type==FF_BUFFER_TYPE_INTERNAL); + + s->current_picture_ptr= s->reordered_input_picture[0]; + for(i=0; i<4; i++){ + s->new_picture.data[i]+= INPLACE_OFFSET; + } + } + copy_picture(&s->current_picture, s->current_picture_ptr); + + s->picture_number= s->new_picture.display_picture_number; +//printf("dpn:%d\n", s->picture_number); + }else{ + memset(&s->new_picture, 0, sizeof(Picture)); + } +} + +int MPV_encode_picture(AVCodecContext *avctx, + unsigned char *buf, int buf_size, void *data) +{ + MpegEncContext *s = avctx->priv_data; + AVFrame *pic_arg = data; + int i, stuffing_count; + + for(i=0; ithread_count; i++){ + int start_y= s->thread_context[i]->start_mb_y; + int end_y= s->thread_context[i]-> end_mb_y; + int h= s->mb_height; + uint8_t *start= buf + (size_t)(((int64_t) buf_size)*start_y/h); + uint8_t *end = buf + (size_t)(((int64_t) buf_size)* end_y/h); + + init_put_bits(&s->thread_context[i]->pb, start, end - start); + } + + s->picture_in_gop_number++; + + if(load_input_picture(s, pic_arg) < 0) + return -1; + + select_input_picture(s); + + /* output? */ + if(s->new_picture.data[0]){ + s->pict_type= s->new_picture.pict_type; +//emms_c(); +//printf("qs:%f %f %d\n", s->new_picture.quality, s->current_picture.quality, s->qscale); + MPV_frame_start(s, avctx); +vbv_retry: + if (encode_picture(s, s->picture_number) < 0) + return -1; + + avctx->real_pict_num = s->picture_number; + avctx->header_bits = s->header_bits; + avctx->mv_bits = s->mv_bits; + avctx->misc_bits = s->misc_bits; + avctx->i_tex_bits = s->i_tex_bits; + avctx->p_tex_bits = s->p_tex_bits; + avctx->i_count = s->i_count; + avctx->p_count = s->mb_num - s->i_count - s->skip_count; //FIXME f/b_count in avctx + avctx->skip_count = s->skip_count; + + MPV_frame_end(s); + + if (ENABLE_MJPEG_ENCODER && s->out_format == FMT_MJPEG) + ff_mjpeg_encode_picture_trailer(s); + + if(avctx->rc_buffer_size){ + RateControlContext *rcc= &s->rc_context; + int max_size= rcc->buffer_index/3; + + if(put_bits_count(&s->pb) > max_size && s->lambda < s->avctx->lmax){ + s->next_lambda= FFMAX(s->lambda+1, s->lambda*(s->qscale+1) / s->qscale); + if(s->adaptive_quant){ + int i; + for(i=0; imb_height*s->mb_stride; i++) + s->lambda_table[i]= FFMAX(s->lambda_table[i]+1, s->lambda_table[i]*(s->qscale+1) / s->qscale); + } + s->mb_skipped = 0; //done in MPV_frame_start() + if(s->pict_type==P_TYPE){ //done in encode_picture() so we must undo it + if(s->flipflop_rounding || s->codec_id == CODEC_ID_H263P || s->codec_id == CODEC_ID_MPEG4) + s->no_rounding ^= 1; + } + if(s->pict_type!=B_TYPE){ + s->time_base= s->last_time_base; + s->last_non_b_time= s->time - s->pp_time; + } +// av_log(NULL, AV_LOG_ERROR, "R:%d ", s->next_lambda); + for(i=0; ithread_count; i++){ + PutBitContext *pb= &s->thread_context[i]->pb; + init_put_bits(pb, pb->buf, pb->buf_end - pb->buf); + } + goto vbv_retry; + } + + assert(s->avctx->rc_max_rate); + } + + if(s->flags&CODEC_FLAG_PASS1) + ff_write_pass1_stats(s); + + for(i=0; i<4; i++){ + s->current_picture_ptr->error[i]= s->current_picture.error[i]; + avctx->error[i] += s->current_picture_ptr->error[i]; + } + + if(s->flags&CODEC_FLAG_PASS1) + assert(avctx->header_bits + avctx->mv_bits + avctx->misc_bits + avctx->i_tex_bits + avctx->p_tex_bits == put_bits_count(&s->pb)); + flush_put_bits(&s->pb); + s->frame_bits = put_bits_count(&s->pb); + + stuffing_count= ff_vbv_update(s, s->frame_bits); + if(stuffing_count){ + if(s->pb.buf_end - s->pb.buf - (put_bits_count(&s->pb)>>3) < stuffing_count + 50){ + av_log(s->avctx, AV_LOG_ERROR, "stuffing too large\n"); + return -1; + } + + switch(s->codec_id){ + case CODEC_ID_MPEG1VIDEO: + case CODEC_ID_MPEG2VIDEO: + while(stuffing_count--){ + put_bits(&s->pb, 8, 0); + } + break; + case CODEC_ID_MPEG4: + put_bits(&s->pb, 16, 0); + put_bits(&s->pb, 16, 0x1C3); + stuffing_count -= 4; + while(stuffing_count--){ + put_bits(&s->pb, 8, 0xFF); + } + break; + default: + av_log(s->avctx, AV_LOG_ERROR, "vbv buffer overflow\n"); + } + flush_put_bits(&s->pb); + s->frame_bits = put_bits_count(&s->pb); + } + + /* update mpeg1/2 vbv_delay for CBR */ + if(s->avctx->rc_max_rate && s->avctx->rc_min_rate == s->avctx->rc_max_rate && s->out_format == FMT_MPEG1 + && 90000LL * (avctx->rc_buffer_size-1) <= s->avctx->rc_max_rate*0xFFFFLL){ + int vbv_delay; + + assert(s->repeat_first_field==0); + + vbv_delay= lrintf(90000 * s->rc_context.buffer_index / s->avctx->rc_max_rate); + assert(vbv_delay < 0xFFFF); + + s->vbv_delay_ptr[0] &= 0xF8; + s->vbv_delay_ptr[0] |= vbv_delay>>13; + s->vbv_delay_ptr[1] = vbv_delay>>5; + s->vbv_delay_ptr[2] &= 0x07; + s->vbv_delay_ptr[2] |= vbv_delay<<3; + } + s->total_bits += s->frame_bits; + avctx->frame_bits = s->frame_bits; + }else{ + assert((pbBufPtr(&s->pb) == s->pb.buf)); + s->frame_bits=0; + } + assert((s->frame_bits&7)==0); + + return s->frame_bits/8; +} + +static inline void dct_single_coeff_elimination(MpegEncContext *s, int n, int threshold) +{ + static const char tab[64]= + {3,2,2,1,1,1,1,1, + 1,1,1,1,1,1,1,1, + 1,1,1,1,1,1,1,1, + 0,0,0,0,0,0,0,0, + 0,0,0,0,0,0,0,0, + 0,0,0,0,0,0,0,0, + 0,0,0,0,0,0,0,0, + 0,0,0,0,0,0,0,0}; + int score=0; + int run=0; + int i; + DCTELEM *block= s->block[n]; + const int last_index= s->block_last_index[n]; + int skip_dc; + + if(threshold<0){ + skip_dc=0; + threshold= -threshold; + }else + skip_dc=1; + + /* are all which we could set to zero are allready zero? */ + if(last_index<=skip_dc - 1) return; + + for(i=0; i<=last_index; i++){ + const int j = s->intra_scantable.permutated[i]; + const int level = FFABS(block[j]); + if(level==1){ + if(skip_dc && i==0) continue; + score+= tab[run]; + run=0; + }else if(level>1){ + return; + }else{ + run++; + } + } + if(score >= threshold) return; + for(i=skip_dc; i<=last_index; i++){ + const int j = s->intra_scantable.permutated[i]; + block[j]=0; + } + if(block[0]) s->block_last_index[n]= 0; + else s->block_last_index[n]= -1; +} + +static inline void clip_coeffs(MpegEncContext *s, DCTELEM *block, int last_index) +{ + int i; + const int maxlevel= s->max_qcoeff; + const int minlevel= s->min_qcoeff; + int overflow=0; + + if(s->mb_intra){ + i=1; //skip clipping of intra dc + }else + i=0; + + for(;i<=last_index; i++){ + const int j= s->intra_scantable.permutated[i]; + int level = block[j]; + + if (level>maxlevel){ + level=maxlevel; + overflow++; + }else if(levelavctx->mb_decision == FF_MB_DECISION_SIMPLE) + av_log(s->avctx, AV_LOG_INFO, "warning, clipping %d dct coefficients to %d..%d\n", overflow, minlevel, maxlevel); +} + +static void get_vissual_weight(int16_t *weight, uint8_t *ptr, int stride){ + int x, y; +//FIXME optimize + for(y=0; y<8; y++){ + for(x=0; x<8; x++){ + int x2, y2; + int sum=0; + int sqr=0; + int count=0; + + for(y2= FFMAX(y-1, 0); y2 < FFMIN(8, y+2); y2++){ + for(x2= FFMAX(x-1, 0); x2 < FFMIN(8, x+2); x2++){ + int v= ptr[x2 + y2*stride]; + sum += v; + sqr += v*v; + count++; + } + } + weight[x + 8*y]= (36*ff_sqrt(count*sqr - sum*sum)) / count; + } + } +} + +static av_always_inline void encode_mb_internal(MpegEncContext *s, int motion_x, int motion_y, int mb_block_height, int mb_block_count) +{ + int16_t weight[8][64]; + DCTELEM orig[8][64]; + const int mb_x= s->mb_x; + const int mb_y= s->mb_y; + int i; + int skip_dct[8]; + int dct_offset = s->linesize*8; //default for progressive frames + uint8_t *ptr_y, *ptr_cb, *ptr_cr; + int wrap_y, wrap_c; + + for(i=0; iskipdct; + + if(s->adaptive_quant){ + const int last_qp= s->qscale; + const int mb_xy= mb_x + mb_y*s->mb_stride; + + s->lambda= s->lambda_table[mb_xy]; + update_qscale(s); + + if(!(s->flags&CODEC_FLAG_QP_RD)){ + s->qscale= s->current_picture_ptr->qscale_table[mb_xy]; + s->dquant= s->qscale - last_qp; + + if(s->out_format==FMT_H263){ + s->dquant= av_clip(s->dquant, -2, 2); + + if(s->codec_id==CODEC_ID_MPEG4){ + if(!s->mb_intra){ + if(s->pict_type == B_TYPE){ + if(s->dquant&1 || s->mv_dir&MV_DIRECT) + s->dquant= 0; + } + if(s->mv_type==MV_TYPE_8X8) + s->dquant=0; + } + } + } + } + ff_set_qscale(s, last_qp + s->dquant); + }else if(s->flags&CODEC_FLAG_QP_RD) + ff_set_qscale(s, s->qscale + s->dquant); + + wrap_y = s->linesize; + wrap_c = s->uvlinesize; + ptr_y = s->new_picture.data[0] + (mb_y * 16 * wrap_y) + mb_x * 16; + ptr_cb = s->new_picture.data[1] + (mb_y * mb_block_height * wrap_c) + mb_x * 8; + ptr_cr = s->new_picture.data[2] + (mb_y * mb_block_height * wrap_c) + mb_x * 8; + + if(mb_x*16+16 > s->width || mb_y*16+16 > s->height){ + uint8_t *ebuf= s->edge_emu_buffer + 32; + ff_emulated_edge_mc(ebuf , ptr_y , wrap_y,16,16,mb_x*16,mb_y*16, s->width , s->height); + ptr_y= ebuf; + ff_emulated_edge_mc(ebuf+18*wrap_y , ptr_cb, wrap_c, 8, mb_block_height, mb_x*8, mb_y*8, s->width>>1, s->height>>1); + ptr_cb= ebuf+18*wrap_y; + ff_emulated_edge_mc(ebuf+18*wrap_y+8, ptr_cr, wrap_c, 8, mb_block_height, mb_x*8, mb_y*8, s->width>>1, s->height>>1); + ptr_cr= ebuf+18*wrap_y+8; + } + + if (s->mb_intra) { + if(s->flags&CODEC_FLAG_INTERLACED_DCT){ + int progressive_score, interlaced_score; + + s->interlaced_dct=0; + progressive_score= s->dsp.ildct_cmp[4](s, ptr_y , NULL, wrap_y, 8) + +s->dsp.ildct_cmp[4](s, ptr_y + wrap_y*8, NULL, wrap_y, 8) - 400; + + if(progressive_score > 0){ + interlaced_score = s->dsp.ildct_cmp[4](s, ptr_y , NULL, wrap_y*2, 8) + +s->dsp.ildct_cmp[4](s, ptr_y + wrap_y , NULL, wrap_y*2, 8); + if(progressive_score > interlaced_score){ + s->interlaced_dct=1; + + dct_offset= wrap_y; + wrap_y<<=1; + if (s->chroma_format == CHROMA_422) + wrap_c<<=1; + } + } + } + + s->dsp.get_pixels(s->block[0], ptr_y , wrap_y); + s->dsp.get_pixels(s->block[1], ptr_y + 8, wrap_y); + s->dsp.get_pixels(s->block[2], ptr_y + dct_offset , wrap_y); + s->dsp.get_pixels(s->block[3], ptr_y + dct_offset + 8, wrap_y); + + if(s->flags&CODEC_FLAG_GRAY){ + skip_dct[4]= 1; + skip_dct[5]= 1; + }else{ + s->dsp.get_pixels(s->block[4], ptr_cb, wrap_c); + s->dsp.get_pixels(s->block[5], ptr_cr, wrap_c); + if(!s->chroma_y_shift){ /* 422 */ + s->dsp.get_pixels(s->block[6], ptr_cb + (dct_offset>>1), wrap_c); + s->dsp.get_pixels(s->block[7], ptr_cr + (dct_offset>>1), wrap_c); + } + } + }else{ + op_pixels_func (*op_pix)[4]; + qpel_mc_func (*op_qpix)[16]; + uint8_t *dest_y, *dest_cb, *dest_cr; + + dest_y = s->dest[0]; + dest_cb = s->dest[1]; + dest_cr = s->dest[2]; + + if ((!s->no_rounding) || s->pict_type==B_TYPE){ + op_pix = s->dsp.put_pixels_tab; + op_qpix= s->dsp.put_qpel_pixels_tab; + }else{ + op_pix = s->dsp.put_no_rnd_pixels_tab; + op_qpix= s->dsp.put_no_rnd_qpel_pixels_tab; + } + + if (s->mv_dir & MV_DIR_FORWARD) { + MPV_motion(s, dest_y, dest_cb, dest_cr, 0, s->last_picture.data, op_pix, op_qpix); + op_pix = s->dsp.avg_pixels_tab; + op_qpix= s->dsp.avg_qpel_pixels_tab; + } + if (s->mv_dir & MV_DIR_BACKWARD) { + MPV_motion(s, dest_y, dest_cb, dest_cr, 1, s->next_picture.data, op_pix, op_qpix); + } + + if(s->flags&CODEC_FLAG_INTERLACED_DCT){ + int progressive_score, interlaced_score; + + s->interlaced_dct=0; + progressive_score= s->dsp.ildct_cmp[0](s, dest_y , ptr_y , wrap_y, 8) + +s->dsp.ildct_cmp[0](s, dest_y + wrap_y*8, ptr_y + wrap_y*8, wrap_y, 8) - 400; + + if(s->avctx->ildct_cmp == FF_CMP_VSSE) progressive_score -= 400; + + if(progressive_score>0){ + interlaced_score = s->dsp.ildct_cmp[0](s, dest_y , ptr_y , wrap_y*2, 8) + +s->dsp.ildct_cmp[0](s, dest_y + wrap_y , ptr_y + wrap_y , wrap_y*2, 8); + + if(progressive_score > interlaced_score){ + s->interlaced_dct=1; + + dct_offset= wrap_y; + wrap_y<<=1; + if (s->chroma_format == CHROMA_422) + wrap_c<<=1; + } + } + } + + s->dsp.diff_pixels(s->block[0], ptr_y , dest_y , wrap_y); + s->dsp.diff_pixels(s->block[1], ptr_y + 8, dest_y + 8, wrap_y); + s->dsp.diff_pixels(s->block[2], ptr_y + dct_offset , dest_y + dct_offset , wrap_y); + s->dsp.diff_pixels(s->block[3], ptr_y + dct_offset + 8, dest_y + dct_offset + 8, wrap_y); + + if(s->flags&CODEC_FLAG_GRAY){ + skip_dct[4]= 1; + skip_dct[5]= 1; + }else{ + s->dsp.diff_pixels(s->block[4], ptr_cb, dest_cb, wrap_c); + s->dsp.diff_pixels(s->block[5], ptr_cr, dest_cr, wrap_c); + if(!s->chroma_y_shift){ /* 422 */ + s->dsp.diff_pixels(s->block[6], ptr_cb + (dct_offset>>1), dest_cb + (dct_offset>>1), wrap_c); + s->dsp.diff_pixels(s->block[7], ptr_cr + (dct_offset>>1), dest_cr + (dct_offset>>1), wrap_c); + } + } + /* pre quantization */ + if(s->current_picture.mc_mb_var[s->mb_stride*mb_y+ mb_x]<2*s->qscale*s->qscale){ + //FIXME optimize + if(s->dsp.sad[1](NULL, ptr_y , dest_y , wrap_y, 8) < 20*s->qscale) skip_dct[0]= 1; + if(s->dsp.sad[1](NULL, ptr_y + 8, dest_y + 8, wrap_y, 8) < 20*s->qscale) skip_dct[1]= 1; + if(s->dsp.sad[1](NULL, ptr_y +dct_offset , dest_y +dct_offset , wrap_y, 8) < 20*s->qscale) skip_dct[2]= 1; + if(s->dsp.sad[1](NULL, ptr_y +dct_offset+ 8, dest_y +dct_offset+ 8, wrap_y, 8) < 20*s->qscale) skip_dct[3]= 1; + if(s->dsp.sad[1](NULL, ptr_cb , dest_cb , wrap_c, 8) < 20*s->qscale) skip_dct[4]= 1; + if(s->dsp.sad[1](NULL, ptr_cr , dest_cr , wrap_c, 8) < 20*s->qscale) skip_dct[5]= 1; + if(!s->chroma_y_shift){ /* 422 */ + if(s->dsp.sad[1](NULL, ptr_cb +(dct_offset>>1), dest_cb +(dct_offset>>1), wrap_c, 8) < 20*s->qscale) skip_dct[6]= 1; + if(s->dsp.sad[1](NULL, ptr_cr +(dct_offset>>1), dest_cr +(dct_offset>>1), wrap_c, 8) < 20*s->qscale) skip_dct[7]= 1; + } + } + } + + if(s->avctx->quantizer_noise_shaping){ + if(!skip_dct[0]) get_vissual_weight(weight[0], ptr_y , wrap_y); + if(!skip_dct[1]) get_vissual_weight(weight[1], ptr_y + 8, wrap_y); + if(!skip_dct[2]) get_vissual_weight(weight[2], ptr_y + dct_offset , wrap_y); + if(!skip_dct[3]) get_vissual_weight(weight[3], ptr_y + dct_offset + 8, wrap_y); + if(!skip_dct[4]) get_vissual_weight(weight[4], ptr_cb , wrap_c); + if(!skip_dct[5]) get_vissual_weight(weight[5], ptr_cr , wrap_c); + if(!s->chroma_y_shift){ /* 422 */ + if(!skip_dct[6]) get_vissual_weight(weight[6], ptr_cb + (dct_offset>>1), wrap_c); + if(!skip_dct[7]) get_vissual_weight(weight[7], ptr_cr + (dct_offset>>1), wrap_c); + } + memcpy(orig[0], s->block[0], sizeof(DCTELEM)*64*mb_block_count); + } + + /* DCT & quantize */ + assert(s->out_format!=FMT_MJPEG || s->qscale==8); + { + for(i=0;iblock_last_index[i] = s->dct_quantize(s, s->block[i], i, s->qscale, &overflow); + // FIXME we could decide to change to quantizer instead of clipping + // JS: I don't think that would be a good idea it could lower quality instead + // of improve it. Just INTRADC clipping deserves changes in quantizer + if (overflow) clip_coeffs(s, s->block[i], s->block_last_index[i]); + }else + s->block_last_index[i]= -1; + } + if(s->avctx->quantizer_noise_shaping){ + for(i=0;iblock_last_index[i] = dct_quantize_refine(s, s->block[i], weight[i], orig[i], i, s->qscale); + } + } + } + + if(s->luma_elim_threshold && !s->mb_intra) + for(i=0; i<4; i++) + dct_single_coeff_elimination(s, i, s->luma_elim_threshold); + if(s->chroma_elim_threshold && !s->mb_intra) + for(i=4; ichroma_elim_threshold); + + if(s->flags & CODEC_FLAG_CBP_RD){ + for(i=0;iblock_last_index[i] == -1) + s->coded_score[i]= INT_MAX/256; + } + } + } + + if((s->flags&CODEC_FLAG_GRAY) && s->mb_intra){ + s->block_last_index[4]= + s->block_last_index[5]= 0; + s->block[4][0]= + s->block[5][0]= (1024 + s->c_dc_scale/2)/ s->c_dc_scale; + } + + //non c quantize code returns incorrect block_last_index FIXME + if(s->alternate_scan && s->dct_quantize != dct_quantize_c){ + for(i=0; iblock_last_index[i]>0){ + for(j=63; j>0; j--){ + if(s->block[i][ s->intra_scantable.permutated[j] ]) break; + } + s->block_last_index[i]= j; + } + } + } + + /* huffman encode */ + switch(s->codec_id){ //FIXME funct ptr could be slightly faster + case CODEC_ID_MPEG1VIDEO: + case CODEC_ID_MPEG2VIDEO: + mpeg1_encode_mb(s, s->block, motion_x, motion_y); break; + case CODEC_ID_MPEG4: + mpeg4_encode_mb(s, s->block, motion_x, motion_y); break; + case CODEC_ID_MSMPEG4V2: + case CODEC_ID_MSMPEG4V3: + case CODEC_ID_WMV1: + if (ENABLE_MSMPEG4_ENCODER) + msmpeg4_encode_mb(s, s->block, motion_x, motion_y); + break; + case CODEC_ID_WMV2: + if (ENABLE_WMV2_ENCODER) + ff_wmv2_encode_mb(s, s->block, motion_x, motion_y); + break; + case CODEC_ID_H261: + if (ENABLE_H261_ENCODER) + ff_h261_encode_mb(s, s->block, motion_x, motion_y); + break; + case CODEC_ID_H263: + case CODEC_ID_H263P: + case CODEC_ID_FLV1: + case CODEC_ID_RV10: + case CODEC_ID_RV20: + h263_encode_mb(s, s->block, motion_x, motion_y); break; + case CODEC_ID_MJPEG: + if (ENABLE_MJPEG_ENCODER) + ff_mjpeg_encode_mb(s, s->block); + break; + default: + assert(0); + } +} + +static av_always_inline void encode_mb(MpegEncContext *s, int motion_x, int motion_y) +{ + if (s->chroma_format == CHROMA_420) encode_mb_internal(s, motion_x, motion_y, 8, 6); + else encode_mb_internal(s, motion_x, motion_y, 16, 8); +} + +void ff_copy_bits(PutBitContext *pb, uint8_t *src, int length) +{ + const uint16_t *srcw= (uint16_t*)src; + int words= length>>4; + int bits= length&15; + int i; + + if(length==0) return; + + if(words < 16){ + for(i=0; i>(16-bits)); +} + +static inline void copy_context_before_encode(MpegEncContext *d, MpegEncContext *s, int type){ + int i; + + memcpy(d->last_mv, s->last_mv, 2*2*2*sizeof(int)); //FIXME is memcpy faster then a loop? + + /* mpeg1 */ + d->mb_skip_run= s->mb_skip_run; + for(i=0; i<3; i++) + d->last_dc[i]= s->last_dc[i]; + + /* statistics */ + d->mv_bits= s->mv_bits; + d->i_tex_bits= s->i_tex_bits; + d->p_tex_bits= s->p_tex_bits; + d->i_count= s->i_count; + d->f_count= s->f_count; + d->b_count= s->b_count; + d->skip_count= s->skip_count; + d->misc_bits= s->misc_bits; + d->last_bits= 0; + + d->mb_skipped= 0; + d->qscale= s->qscale; + d->dquant= s->dquant; +} + +static inline void copy_context_after_encode(MpegEncContext *d, MpegEncContext *s, int type){ + int i; + + memcpy(d->mv, s->mv, 2*4*2*sizeof(int)); + memcpy(d->last_mv, s->last_mv, 2*2*2*sizeof(int)); //FIXME is memcpy faster then a loop? + + /* mpeg1 */ + d->mb_skip_run= s->mb_skip_run; + for(i=0; i<3; i++) + d->last_dc[i]= s->last_dc[i]; + + /* statistics */ + d->mv_bits= s->mv_bits; + d->i_tex_bits= s->i_tex_bits; + d->p_tex_bits= s->p_tex_bits; + d->i_count= s->i_count; + d->f_count= s->f_count; + d->b_count= s->b_count; + d->skip_count= s->skip_count; + d->misc_bits= s->misc_bits; + + d->mb_intra= s->mb_intra; + d->mb_skipped= s->mb_skipped; + d->mv_type= s->mv_type; + d->mv_dir= s->mv_dir; + d->pb= s->pb; + if(s->data_partitioning){ + d->pb2= s->pb2; + d->tex_pb= s->tex_pb; + } + d->block= s->block; + for(i=0; i<8; i++) + d->block_last_index[i]= s->block_last_index[i]; + d->interlaced_dct= s->interlaced_dct; + d->qscale= s->qscale; +} + +static inline void encode_mb_hq(MpegEncContext *s, MpegEncContext *backup, MpegEncContext *best, int type, + PutBitContext pb[2], PutBitContext pb2[2], PutBitContext tex_pb[2], + int *dmin, int *next_block, int motion_x, int motion_y) +{ + int score; + uint8_t *dest_backup[3]; + + copy_context_before_encode(s, backup, type); + + s->block= s->blocks[*next_block]; + s->pb= pb[*next_block]; + if(s->data_partitioning){ + s->pb2 = pb2 [*next_block]; + s->tex_pb= tex_pb[*next_block]; + } + + if(*next_block){ + memcpy(dest_backup, s->dest, sizeof(s->dest)); + s->dest[0] = s->rd_scratchpad; + s->dest[1] = s->rd_scratchpad + 16*s->linesize; + s->dest[2] = s->rd_scratchpad + 16*s->linesize + 8; + assert(s->linesize >= 32); //FIXME + } + + encode_mb(s, motion_x, motion_y); + + score= put_bits_count(&s->pb); + if(s->data_partitioning){ + score+= put_bits_count(&s->pb2); + score+= put_bits_count(&s->tex_pb); + } + + if(s->avctx->mb_decision == FF_MB_DECISION_RD){ + MPV_decode_mb(s, s->block); + + score *= s->lambda2; + score += sse_mb(s) << FF_LAMBDA_SHIFT; + } + + if(*next_block){ + memcpy(s->dest, dest_backup, sizeof(s->dest)); + } + + if(score<*dmin){ + *dmin= score; + *next_block^=1; + + copy_context_after_encode(best, s, type); + } +} + +static int sse(MpegEncContext *s, uint8_t *src1, uint8_t *src2, int w, int h, int stride){ + uint32_t *sq = ff_squareTbl + 256; + int acc=0; + int x,y; + + if(w==16 && h==16) + return s->dsp.sse[0](NULL, src1, src2, stride, 16); + else if(w==8 && h==8) + return s->dsp.sse[1](NULL, src1, src2, stride, 8); + + for(y=0; y=0); + + return acc; +} + +static int sse_mb(MpegEncContext *s){ + int w= 16; + int h= 16; + + if(s->mb_x*16 + 16 > s->width ) w= s->width - s->mb_x*16; + if(s->mb_y*16 + 16 > s->height) h= s->height- s->mb_y*16; + + if(w==16 && h==16) + if(s->avctx->mb_cmp == FF_CMP_NSSE){ + return s->dsp.nsse[0](s, s->new_picture.data[0] + s->mb_x*16 + s->mb_y*s->linesize*16, s->dest[0], s->linesize, 16) + +s->dsp.nsse[1](s, s->new_picture.data[1] + s->mb_x*8 + s->mb_y*s->uvlinesize*8,s->dest[1], s->uvlinesize, 8) + +s->dsp.nsse[1](s, s->new_picture.data[2] + s->mb_x*8 + s->mb_y*s->uvlinesize*8,s->dest[2], s->uvlinesize, 8); + }else{ + return s->dsp.sse[0](NULL, s->new_picture.data[0] + s->mb_x*16 + s->mb_y*s->linesize*16, s->dest[0], s->linesize, 16) + +s->dsp.sse[1](NULL, s->new_picture.data[1] + s->mb_x*8 + s->mb_y*s->uvlinesize*8,s->dest[1], s->uvlinesize, 8) + +s->dsp.sse[1](NULL, s->new_picture.data[2] + s->mb_x*8 + s->mb_y*s->uvlinesize*8,s->dest[2], s->uvlinesize, 8); + } + else + return sse(s, s->new_picture.data[0] + s->mb_x*16 + s->mb_y*s->linesize*16, s->dest[0], w, h, s->linesize) + +sse(s, s->new_picture.data[1] + s->mb_x*8 + s->mb_y*s->uvlinesize*8,s->dest[1], w>>1, h>>1, s->uvlinesize) + +sse(s, s->new_picture.data[2] + s->mb_x*8 + s->mb_y*s->uvlinesize*8,s->dest[2], w>>1, h>>1, s->uvlinesize); +} + +static int pre_estimate_motion_thread(AVCodecContext *c, void *arg){ + MpegEncContext *s= arg; + + + s->me.pre_pass=1; + s->me.dia_size= s->avctx->pre_dia_size; + s->first_slice_line=1; + for(s->mb_y= s->end_mb_y-1; s->mb_y >= s->start_mb_y; s->mb_y--) { + for(s->mb_x=s->mb_width-1; s->mb_x >=0 ;s->mb_x--) { + ff_pre_estimate_p_frame_motion(s, s->mb_x, s->mb_y); + } + s->first_slice_line=0; + } + + s->me.pre_pass=0; + + return 0; +} + +static int estimate_motion_thread(AVCodecContext *c, void *arg){ + MpegEncContext *s= arg; + + ff_check_alignment(); + + s->me.dia_size= s->avctx->dia_size; + 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; //for block init below + ff_init_block_index(s); + for(s->mb_x=0; s->mb_x < s->mb_width; s->mb_x++) { + s->block_index[0]+=2; + s->block_index[1]+=2; + s->block_index[2]+=2; + s->block_index[3]+=2; + + /* compute motion vector & mb_type and store in context */ + if(s->pict_type==B_TYPE) + ff_estimate_b_frame_motion(s, s->mb_x, s->mb_y); + else + ff_estimate_p_frame_motion(s, s->mb_x, s->mb_y); + } + s->first_slice_line=0; + } + return 0; +} + +static int mb_var_thread(AVCodecContext *c, void *arg){ + MpegEncContext *s= arg; + int mb_x, mb_y; + + ff_check_alignment(); + + for(mb_y=s->start_mb_y; mb_y < s->end_mb_y; mb_y++) { + for(mb_x=0; mb_x < s->mb_width; mb_x++) { + int xx = mb_x * 16; + int yy = mb_y * 16; + uint8_t *pix = s->new_picture.data[0] + (yy * s->linesize) + xx; + int varc; + int sum = s->dsp.pix_sum(pix, s->linesize); + + varc = (s->dsp.pix_norm1(pix, s->linesize) - (((unsigned)(sum*sum))>>8) + 500 + 128)>>8; + + s->current_picture.mb_var [s->mb_stride * mb_y + mb_x] = varc; + s->current_picture.mb_mean[s->mb_stride * mb_y + mb_x] = (sum+128)>>8; + s->me.mb_var_sum_temp += varc; + } + } + return 0; +} + +static void write_slice_end(MpegEncContext *s){ + if(s->codec_id==CODEC_ID_MPEG4){ + if(s->partitioned_frame){ + ff_mpeg4_merge_partitions(s); + } + + ff_mpeg4_stuffing(&s->pb); + }else if(ENABLE_MJPEG_ENCODER && s->out_format == FMT_MJPEG){ + ff_mjpeg_encode_stuffing(&s->pb); + } + + align_put_bits(&s->pb); + flush_put_bits(&s->pb); + + if((s->flags&CODEC_FLAG_PASS1) && !s->partitioned_frame) + s->misc_bits+= get_bits_diff(s); +} + +static int encode_thread(AVCodecContext *c, void *arg){ + MpegEncContext *s= arg; + int mb_x, mb_y, pdif = 0; + int i, j; + MpegEncContext best_s, backup_s; + uint8_t bit_buf[2][MAX_MB_BYTES]; + uint8_t bit_buf2[2][MAX_MB_BYTES]; + uint8_t bit_buf_tex[2][MAX_MB_BYTES]; + PutBitContext pb[2], pb2[2], tex_pb[2]; +//printf("%d->%d\n", s->resync_mb_y, s->end_mb_y); + + ff_check_alignment(); + + for(i=0; i<2; i++){ + init_put_bits(&pb [i], bit_buf [i], MAX_MB_BYTES); + init_put_bits(&pb2 [i], bit_buf2 [i], MAX_MB_BYTES); + init_put_bits(&tex_pb[i], bit_buf_tex[i], MAX_MB_BYTES); + } + + s->last_bits= put_bits_count(&s->pb); + s->mv_bits=0; + s->misc_bits=0; + s->i_tex_bits=0; + s->p_tex_bits=0; + s->i_count=0; + s->f_count=0; + s->b_count=0; + s->skip_count=0; + + for(i=0; i<3; i++){ + /* init last dc values */ + /* note: quant matrix value (8) is implied here */ + s->last_dc[i] = 128 << s->intra_dc_precision; + + s->current_picture.error[i] = 0; + } + s->mb_skip_run = 0; + memset(s->last_mv, 0, sizeof(s->last_mv)); + + s->last_mv_dir = 0; + + switch(s->codec_id){ + case CODEC_ID_H263: + case CODEC_ID_H263P: + case CODEC_ID_FLV1: + s->gob_index = ff_h263_get_gob_height(s); + break; + case CODEC_ID_MPEG4: + if(s->partitioned_frame) + ff_mpeg4_init_partitions(s); + break; + } + + s->resync_mb_x=0; + s->resync_mb_y=0; + s->first_slice_line = 1; + s->ptr_lastgob = s->pb.buf; + for(mb_y= s->start_mb_y; mb_y < s->end_mb_y; mb_y++) { +// printf("row %d at %X\n", s->mb_y, (int)s); + s->mb_x=0; + s->mb_y= mb_y; + + ff_set_qscale(s, s->qscale); + ff_init_block_index(s); + + for(mb_x=0; mb_x < s->mb_width; mb_x++) { + int xy= mb_y*s->mb_stride + mb_x; // removed const, H261 needs to adjust this + int mb_type= s->mb_type[xy]; +// int d; + int dmin= INT_MAX; + int dir; + + if(s->pb.buf_end - s->pb.buf - (put_bits_count(&s->pb)>>3) < MAX_MB_BYTES){ + av_log(s->avctx, AV_LOG_ERROR, "encoded frame too large\n"); + return -1; + } + if(s->data_partitioning){ + if( s->pb2 .buf_end - s->pb2 .buf - (put_bits_count(&s-> pb2)>>3) < MAX_MB_BYTES + || s->tex_pb.buf_end - s->tex_pb.buf - (put_bits_count(&s->tex_pb )>>3) < MAX_MB_BYTES){ + av_log(s->avctx, AV_LOG_ERROR, "encoded frame too large\n"); + return -1; + } + } + + s->mb_x = mb_x; + s->mb_y = mb_y; // moved into loop, can get changed by H.261 + ff_update_block_index(s); + + if(ENABLE_H261_ENCODER && s->codec_id == CODEC_ID_H261){ + ff_h261_reorder_mb_index(s); + xy= s->mb_y*s->mb_stride + s->mb_x; + mb_type= s->mb_type[xy]; + } + + /* write gob / video packet header */ + if(s->rtp_mode){ + int current_packet_size, is_gob_start; + + current_packet_size= ((put_bits_count(&s->pb)+7)>>3) - (s->ptr_lastgob - s->pb.buf); + + is_gob_start= s->avctx->rtp_payload_size && current_packet_size >= s->avctx->rtp_payload_size && mb_y + mb_x>0; + + if(s->start_mb_y == mb_y && mb_y > 0 && mb_x==0) is_gob_start=1; + + switch(s->codec_id){ + case CODEC_ID_H263: + case CODEC_ID_H263P: + if(!s->h263_slice_structured) + if(s->mb_x || s->mb_y%s->gob_index) is_gob_start=0; + break; + case CODEC_ID_MPEG2VIDEO: + if(s->mb_x==0 && s->mb_y!=0) is_gob_start=1; + case CODEC_ID_MPEG1VIDEO: + if(s->mb_skip_run) is_gob_start=0; + break; + } + + if(is_gob_start){ + if(s->start_mb_y != mb_y || mb_x!=0){ + write_slice_end(s); + + if(s->codec_id==CODEC_ID_MPEG4 && s->partitioned_frame){ + ff_mpeg4_init_partitions(s); + } + } + + assert((put_bits_count(&s->pb)&7) == 0); + current_packet_size= pbBufPtr(&s->pb) - s->ptr_lastgob; + + if(s->avctx->error_rate && s->resync_mb_x + s->resync_mb_y > 0){ + int r= put_bits_count(&s->pb)/8 + s->picture_number + 16 + s->mb_x + s->mb_y; + int d= 100 / s->avctx->error_rate; + if(r % d == 0){ + current_packet_size=0; +#ifndef ALT_BITSTREAM_WRITER + s->pb.buf_ptr= s->ptr_lastgob; +#endif + assert(pbBufPtr(&s->pb) == s->ptr_lastgob); + } + } + + if (s->avctx->rtp_callback){ + int number_mb = (mb_y - s->resync_mb_y)*s->mb_width + mb_x - s->resync_mb_x; + s->avctx->rtp_callback(s->avctx, s->ptr_lastgob, current_packet_size, number_mb); + } + + switch(s->codec_id){ + case CODEC_ID_MPEG4: + ff_mpeg4_encode_video_packet_header(s); + ff_mpeg4_clean_buffers(s); + break; + case CODEC_ID_MPEG1VIDEO: + case CODEC_ID_MPEG2VIDEO: + ff_mpeg1_encode_slice_header(s); + ff_mpeg1_clean_buffers(s); + break; + case CODEC_ID_H263: + case CODEC_ID_H263P: + h263_encode_gob_header(s, mb_y); + break; + } + + if(s->flags&CODEC_FLAG_PASS1){ + int bits= put_bits_count(&s->pb); + s->misc_bits+= bits - s->last_bits; + s->last_bits= bits; + } + + s->ptr_lastgob += current_packet_size; + s->first_slice_line=1; + s->resync_mb_x=mb_x; + s->resync_mb_y=mb_y; + } + } + + if( (s->resync_mb_x == s->mb_x) + && s->resync_mb_y+1 == s->mb_y){ + s->first_slice_line=0; + } + + s->mb_skipped=0; + s->dquant=0; //only for QP_RD + + if(mb_type & (mb_type-1) || (s->flags & CODEC_FLAG_QP_RD)){ // more than 1 MB type possible or CODEC_FLAG_QP_RD + int next_block=0; + int pb_bits_count, pb2_bits_count, tex_pb_bits_count; + + copy_context_before_encode(&backup_s, s, -1); + backup_s.pb= s->pb; + best_s.data_partitioning= s->data_partitioning; + best_s.partitioned_frame= s->partitioned_frame; + if(s->data_partitioning){ + backup_s.pb2= s->pb2; + backup_s.tex_pb= s->tex_pb; + } + + if(mb_type&CANDIDATE_MB_TYPE_INTER){ + s->mv_dir = MV_DIR_FORWARD; + s->mv_type = MV_TYPE_16X16; + s->mb_intra= 0; + s->mv[0][0][0] = s->p_mv_table[xy][0]; + s->mv[0][0][1] = s->p_mv_table[xy][1]; + encode_mb_hq(s, &backup_s, &best_s, CANDIDATE_MB_TYPE_INTER, pb, pb2, tex_pb, + &dmin, &next_block, s->mv[0][0][0], s->mv[0][0][1]); + } + if(mb_type&CANDIDATE_MB_TYPE_INTER_I){ + s->mv_dir = MV_DIR_FORWARD; + s->mv_type = MV_TYPE_FIELD; + s->mb_intra= 0; + for(i=0; i<2; i++){ + j= s->field_select[0][i] = s->p_field_select_table[i][xy]; + s->mv[0][i][0] = s->p_field_mv_table[i][j][xy][0]; + s->mv[0][i][1] = s->p_field_mv_table[i][j][xy][1]; + } + encode_mb_hq(s, &backup_s, &best_s, CANDIDATE_MB_TYPE_INTER_I, pb, pb2, tex_pb, + &dmin, &next_block, 0, 0); + } + if(mb_type&CANDIDATE_MB_TYPE_SKIPPED){ + s->mv_dir = MV_DIR_FORWARD; + s->mv_type = MV_TYPE_16X16; + s->mb_intra= 0; + s->mv[0][0][0] = 0; + s->mv[0][0][1] = 0; + encode_mb_hq(s, &backup_s, &best_s, CANDIDATE_MB_TYPE_SKIPPED, pb, pb2, tex_pb, + &dmin, &next_block, s->mv[0][0][0], s->mv[0][0][1]); + } + if(mb_type&CANDIDATE_MB_TYPE_INTER4V){ + s->mv_dir = MV_DIR_FORWARD; + s->mv_type = MV_TYPE_8X8; + s->mb_intra= 0; + for(i=0; i<4; i++){ + s->mv[0][i][0] = s->current_picture.motion_val[0][s->block_index[i]][0]; + s->mv[0][i][1] = s->current_picture.motion_val[0][s->block_index[i]][1]; + } + encode_mb_hq(s, &backup_s, &best_s, CANDIDATE_MB_TYPE_INTER4V, pb, pb2, tex_pb, + &dmin, &next_block, 0, 0); + } + if(mb_type&CANDIDATE_MB_TYPE_FORWARD){ + s->mv_dir = MV_DIR_FORWARD; + s->mv_type = MV_TYPE_16X16; + s->mb_intra= 0; + s->mv[0][0][0] = s->b_forw_mv_table[xy][0]; + s->mv[0][0][1] = s->b_forw_mv_table[xy][1]; + encode_mb_hq(s, &backup_s, &best_s, CANDIDATE_MB_TYPE_FORWARD, pb, pb2, tex_pb, + &dmin, &next_block, s->mv[0][0][0], s->mv[0][0][1]); + } + if(mb_type&CANDIDATE_MB_TYPE_BACKWARD){ + s->mv_dir = MV_DIR_BACKWARD; + s->mv_type = MV_TYPE_16X16; + s->mb_intra= 0; + s->mv[1][0][0] = s->b_back_mv_table[xy][0]; + s->mv[1][0][1] = s->b_back_mv_table[xy][1]; + encode_mb_hq(s, &backup_s, &best_s, CANDIDATE_MB_TYPE_BACKWARD, pb, pb2, tex_pb, + &dmin, &next_block, s->mv[1][0][0], s->mv[1][0][1]); + } + if(mb_type&CANDIDATE_MB_TYPE_BIDIR){ + s->mv_dir = MV_DIR_FORWARD | MV_DIR_BACKWARD; + s->mv_type = MV_TYPE_16X16; + s->mb_intra= 0; + s->mv[0][0][0] = s->b_bidir_forw_mv_table[xy][0]; + s->mv[0][0][1] = s->b_bidir_forw_mv_table[xy][1]; + s->mv[1][0][0] = s->b_bidir_back_mv_table[xy][0]; + s->mv[1][0][1] = s->b_bidir_back_mv_table[xy][1]; + encode_mb_hq(s, &backup_s, &best_s, CANDIDATE_MB_TYPE_BIDIR, pb, pb2, tex_pb, + &dmin, &next_block, 0, 0); + } + if(mb_type&CANDIDATE_MB_TYPE_FORWARD_I){ + s->mv_dir = MV_DIR_FORWARD; + s->mv_type = MV_TYPE_FIELD; + s->mb_intra= 0; + for(i=0; i<2; i++){ + j= s->field_select[0][i] = s->b_field_select_table[0][i][xy]; + s->mv[0][i][0] = s->b_field_mv_table[0][i][j][xy][0]; + s->mv[0][i][1] = s->b_field_mv_table[0][i][j][xy][1]; + } + encode_mb_hq(s, &backup_s, &best_s, CANDIDATE_MB_TYPE_FORWARD_I, pb, pb2, tex_pb, + &dmin, &next_block, 0, 0); + } + if(mb_type&CANDIDATE_MB_TYPE_BACKWARD_I){ + s->mv_dir = MV_DIR_BACKWARD; + s->mv_type = MV_TYPE_FIELD; + s->mb_intra= 0; + for(i=0; i<2; i++){ + j= s->field_select[1][i] = s->b_field_select_table[1][i][xy]; + s->mv[1][i][0] = s->b_field_mv_table[1][i][j][xy][0]; + s->mv[1][i][1] = s->b_field_mv_table[1][i][j][xy][1]; + } + encode_mb_hq(s, &backup_s, &best_s, CANDIDATE_MB_TYPE_BACKWARD_I, pb, pb2, tex_pb, + &dmin, &next_block, 0, 0); + } + if(mb_type&CANDIDATE_MB_TYPE_BIDIR_I){ + s->mv_dir = MV_DIR_FORWARD | MV_DIR_BACKWARD; + s->mv_type = MV_TYPE_FIELD; + s->mb_intra= 0; + for(dir=0; dir<2; dir++){ + for(i=0; i<2; i++){ + j= s->field_select[dir][i] = s->b_field_select_table[dir][i][xy]; + s->mv[dir][i][0] = s->b_field_mv_table[dir][i][j][xy][0]; + s->mv[dir][i][1] = s->b_field_mv_table[dir][i][j][xy][1]; + } + } + encode_mb_hq(s, &backup_s, &best_s, CANDIDATE_MB_TYPE_BIDIR_I, pb, pb2, tex_pb, + &dmin, &next_block, 0, 0); + } + if(mb_type&CANDIDATE_MB_TYPE_INTRA){ + s->mv_dir = 0; + s->mv_type = MV_TYPE_16X16; + s->mb_intra= 1; + s->mv[0][0][0] = 0; + s->mv[0][0][1] = 0; + encode_mb_hq(s, &backup_s, &best_s, CANDIDATE_MB_TYPE_INTRA, pb, pb2, tex_pb, + &dmin, &next_block, 0, 0); + if(s->h263_pred || s->h263_aic){ + if(best_s.mb_intra) + s->mbintra_table[mb_x + mb_y*s->mb_stride]=1; + else + ff_clean_intra_table_entries(s); //old mode? + } + } + + if((s->flags & CODEC_FLAG_QP_RD) && dmin < INT_MAX){ + if(best_s.mv_type==MV_TYPE_16X16){ //FIXME move 4mv after QPRD + const int last_qp= backup_s.qscale; + int qpi, qp, dc[6]; + DCTELEM ac[6][16]; + const int mvdir= (best_s.mv_dir&MV_DIR_BACKWARD) ? 1 : 0; + static const int dquant_tab[4]={-1,1,-2,2}; + + assert(backup_s.dquant == 0); + + //FIXME intra + s->mv_dir= best_s.mv_dir; + s->mv_type = MV_TYPE_16X16; + s->mb_intra= best_s.mb_intra; + s->mv[0][0][0] = best_s.mv[0][0][0]; + s->mv[0][0][1] = best_s.mv[0][0][1]; + s->mv[1][0][0] = best_s.mv[1][0][0]; + s->mv[1][0][1] = best_s.mv[1][0][1]; + + qpi = s->pict_type == B_TYPE ? 2 : 0; + for(; qpi<4; qpi++){ + int dquant= dquant_tab[qpi]; + qp= last_qp + dquant; + if(qp < s->avctx->qmin || qp > s->avctx->qmax) + continue; + backup_s.dquant= dquant; + if(s->mb_intra && s->dc_val[0]){ + for(i=0; i<6; i++){ + dc[i]= s->dc_val[0][ s->block_index[i] ]; + memcpy(ac[i], s->ac_val[0][s->block_index[i]], sizeof(DCTELEM)*16); + } + } + + encode_mb_hq(s, &backup_s, &best_s, CANDIDATE_MB_TYPE_INTER /* wrong but unused */, pb, pb2, tex_pb, + &dmin, &next_block, s->mv[mvdir][0][0], s->mv[mvdir][0][1]); + if(best_s.qscale != qp){ + if(s->mb_intra && s->dc_val[0]){ + for(i=0; i<6; i++){ + s->dc_val[0][ s->block_index[i] ]= dc[i]; + memcpy(s->ac_val[0][s->block_index[i]], ac[i], sizeof(DCTELEM)*16); + } + } + } + } + } + } + if(mb_type&CANDIDATE_MB_TYPE_DIRECT){ + int mx= s->b_direct_mv_table[xy][0]; + int my= s->b_direct_mv_table[xy][1]; + + backup_s.dquant = 0; + s->mv_dir = MV_DIR_FORWARD | MV_DIR_BACKWARD | MV_DIRECT; + s->mb_intra= 0; + ff_mpeg4_set_direct_mv(s, mx, my); + encode_mb_hq(s, &backup_s, &best_s, CANDIDATE_MB_TYPE_DIRECT, pb, pb2, tex_pb, + &dmin, &next_block, mx, my); + } + if(mb_type&CANDIDATE_MB_TYPE_DIRECT0){ + backup_s.dquant = 0; + s->mv_dir = MV_DIR_FORWARD | MV_DIR_BACKWARD | MV_DIRECT; + s->mb_intra= 0; + ff_mpeg4_set_direct_mv(s, 0, 0); + encode_mb_hq(s, &backup_s, &best_s, CANDIDATE_MB_TYPE_DIRECT, pb, pb2, tex_pb, + &dmin, &next_block, 0, 0); + } + if(!best_s.mb_intra && s->flags2&CODEC_FLAG2_SKIP_RD){ + int coded=0; + for(i=0; i<6; i++) + coded |= s->block_last_index[i]; + if(coded){ + int mx,my; + memcpy(s->mv, best_s.mv, sizeof(s->mv)); + if(best_s.mv_dir & MV_DIRECT){ + mx=my=0; //FIXME find the one we actually used + ff_mpeg4_set_direct_mv(s, mx, my); + }else if(best_s.mv_dir&MV_DIR_BACKWARD){ + mx= s->mv[1][0][0]; + my= s->mv[1][0][1]; + }else{ + mx= s->mv[0][0][0]; + my= s->mv[0][0][1]; + } + + s->mv_dir= best_s.mv_dir; + s->mv_type = best_s.mv_type; + s->mb_intra= 0; +/* s->mv[0][0][0] = best_s.mv[0][0][0]; + s->mv[0][0][1] = best_s.mv[0][0][1]; + s->mv[1][0][0] = best_s.mv[1][0][0]; + s->mv[1][0][1] = best_s.mv[1][0][1];*/ + backup_s.dquant= 0; + s->skipdct=1; + encode_mb_hq(s, &backup_s, &best_s, CANDIDATE_MB_TYPE_INTER /* wrong but unused */, pb, pb2, tex_pb, + &dmin, &next_block, mx, my); + s->skipdct=0; + } + } + + s->current_picture.qscale_table[xy]= best_s.qscale; + + copy_context_after_encode(s, &best_s, -1); + + pb_bits_count= put_bits_count(&s->pb); + flush_put_bits(&s->pb); + ff_copy_bits(&backup_s.pb, bit_buf[next_block^1], pb_bits_count); + s->pb= backup_s.pb; + + if(s->data_partitioning){ + pb2_bits_count= put_bits_count(&s->pb2); + flush_put_bits(&s->pb2); + ff_copy_bits(&backup_s.pb2, bit_buf2[next_block^1], pb2_bits_count); + s->pb2= backup_s.pb2; + + tex_pb_bits_count= put_bits_count(&s->tex_pb); + flush_put_bits(&s->tex_pb); + ff_copy_bits(&backup_s.tex_pb, bit_buf_tex[next_block^1], tex_pb_bits_count); + s->tex_pb= backup_s.tex_pb; + } + s->last_bits= put_bits_count(&s->pb); + + if (s->out_format == FMT_H263 && s->pict_type!=B_TYPE) + ff_h263_update_motion_val(s); + + if(next_block==0){ //FIXME 16 vs linesize16 + s->dsp.put_pixels_tab[0][0](s->dest[0], s->rd_scratchpad , s->linesize ,16); + s->dsp.put_pixels_tab[1][0](s->dest[1], s->rd_scratchpad + 16*s->linesize , s->uvlinesize, 8); + s->dsp.put_pixels_tab[1][0](s->dest[2], s->rd_scratchpad + 16*s->linesize + 8, s->uvlinesize, 8); + } + + if(s->avctx->mb_decision == FF_MB_DECISION_BITS) + MPV_decode_mb(s, s->block); + } else { + int motion_x = 0, motion_y = 0; + s->mv_type=MV_TYPE_16X16; + // only one MB-Type possible + + switch(mb_type){ + case CANDIDATE_MB_TYPE_INTRA: + s->mv_dir = 0; + s->mb_intra= 1; + motion_x= s->mv[0][0][0] = 0; + motion_y= s->mv[0][0][1] = 0; + break; + case CANDIDATE_MB_TYPE_INTER: + s->mv_dir = MV_DIR_FORWARD; + s->mb_intra= 0; + motion_x= s->mv[0][0][0] = s->p_mv_table[xy][0]; + motion_y= s->mv[0][0][1] = s->p_mv_table[xy][1]; + break; + case CANDIDATE_MB_TYPE_INTER_I: + s->mv_dir = MV_DIR_FORWARD; + s->mv_type = MV_TYPE_FIELD; + s->mb_intra= 0; + for(i=0; i<2; i++){ + j= s->field_select[0][i] = s->p_field_select_table[i][xy]; + s->mv[0][i][0] = s->p_field_mv_table[i][j][xy][0]; + s->mv[0][i][1] = s->p_field_mv_table[i][j][xy][1]; + } + break; + case CANDIDATE_MB_TYPE_INTER4V: + s->mv_dir = MV_DIR_FORWARD; + s->mv_type = MV_TYPE_8X8; + s->mb_intra= 0; + for(i=0; i<4; i++){ + s->mv[0][i][0] = s->current_picture.motion_val[0][s->block_index[i]][0]; + s->mv[0][i][1] = s->current_picture.motion_val[0][s->block_index[i]][1]; + } + break; + case CANDIDATE_MB_TYPE_DIRECT: + s->mv_dir = MV_DIR_FORWARD | MV_DIR_BACKWARD | MV_DIRECT; + s->mb_intra= 0; + motion_x=s->b_direct_mv_table[xy][0]; + motion_y=s->b_direct_mv_table[xy][1]; + ff_mpeg4_set_direct_mv(s, motion_x, motion_y); + break; + case CANDIDATE_MB_TYPE_DIRECT0: + s->mv_dir = MV_DIR_FORWARD | MV_DIR_BACKWARD | MV_DIRECT; + s->mb_intra= 0; + ff_mpeg4_set_direct_mv(s, 0, 0); + break; + case CANDIDATE_MB_TYPE_BIDIR: + s->mv_dir = MV_DIR_FORWARD | MV_DIR_BACKWARD; + s->mb_intra= 0; + s->mv[0][0][0] = s->b_bidir_forw_mv_table[xy][0]; + s->mv[0][0][1] = s->b_bidir_forw_mv_table[xy][1]; + s->mv[1][0][0] = s->b_bidir_back_mv_table[xy][0]; + s->mv[1][0][1] = s->b_bidir_back_mv_table[xy][1]; + break; + case CANDIDATE_MB_TYPE_BACKWARD: + s->mv_dir = MV_DIR_BACKWARD; + s->mb_intra= 0; + motion_x= s->mv[1][0][0] = s->b_back_mv_table[xy][0]; + motion_y= s->mv[1][0][1] = s->b_back_mv_table[xy][1]; + break; + case CANDIDATE_MB_TYPE_FORWARD: + s->mv_dir = MV_DIR_FORWARD; + s->mb_intra= 0; + motion_x= s->mv[0][0][0] = s->b_forw_mv_table[xy][0]; + motion_y= s->mv[0][0][1] = s->b_forw_mv_table[xy][1]; +// printf(" %d %d ", motion_x, motion_y); + break; + case CANDIDATE_MB_TYPE_FORWARD_I: + s->mv_dir = MV_DIR_FORWARD; + s->mv_type = MV_TYPE_FIELD; + s->mb_intra= 0; + for(i=0; i<2; i++){ + j= s->field_select[0][i] = s->b_field_select_table[0][i][xy]; + s->mv[0][i][0] = s->b_field_mv_table[0][i][j][xy][0]; + s->mv[0][i][1] = s->b_field_mv_table[0][i][j][xy][1]; + } + break; + case CANDIDATE_MB_TYPE_BACKWARD_I: + s->mv_dir = MV_DIR_BACKWARD; + s->mv_type = MV_TYPE_FIELD; + s->mb_intra= 0; + for(i=0; i<2; i++){ + j= s->field_select[1][i] = s->b_field_select_table[1][i][xy]; + s->mv[1][i][0] = s->b_field_mv_table[1][i][j][xy][0]; + s->mv[1][i][1] = s->b_field_mv_table[1][i][j][xy][1]; + } + break; + case CANDIDATE_MB_TYPE_BIDIR_I: + s->mv_dir = MV_DIR_FORWARD | MV_DIR_BACKWARD; + s->mv_type = MV_TYPE_FIELD; + s->mb_intra= 0; + for(dir=0; dir<2; dir++){ + for(i=0; i<2; i++){ + j= s->field_select[dir][i] = s->b_field_select_table[dir][i][xy]; + s->mv[dir][i][0] = s->b_field_mv_table[dir][i][j][xy][0]; + s->mv[dir][i][1] = s->b_field_mv_table[dir][i][j][xy][1]; + } + } + break; + default: + av_log(s->avctx, AV_LOG_ERROR, "illegal MB type\n"); + } + + encode_mb(s, motion_x, motion_y); + + // RAL: Update last macroblock type + s->last_mv_dir = s->mv_dir; + + if (s->out_format == FMT_H263 && s->pict_type!=B_TYPE) + ff_h263_update_motion_val(s); + + MPV_decode_mb(s, s->block); + } + + /* clean the MV table in IPS frames for direct mode in B frames */ + if(s->mb_intra /* && I,P,S_TYPE */){ + s->p_mv_table[xy][0]=0; + s->p_mv_table[xy][1]=0; + } + + if(s->flags&CODEC_FLAG_PSNR){ + int w= 16; + int h= 16; + + if(s->mb_x*16 + 16 > s->width ) w= s->width - s->mb_x*16; + if(s->mb_y*16 + 16 > s->height) h= s->height- s->mb_y*16; + + s->current_picture.error[0] += sse( + s, s->new_picture.data[0] + s->mb_x*16 + s->mb_y*s->linesize*16, + s->dest[0], w, h, s->linesize); + s->current_picture.error[1] += sse( + s, s->new_picture.data[1] + s->mb_x*8 + s->mb_y*s->uvlinesize*8, + s->dest[1], w>>1, h>>1, s->uvlinesize); + s->current_picture.error[2] += sse( + s, s->new_picture .data[2] + s->mb_x*8 + s->mb_y*s->uvlinesize*8, + s->dest[2], w>>1, h>>1, s->uvlinesize); + } + if(s->loop_filter){ + if(s->out_format == FMT_H263) + ff_h263_loop_filter(s); + } +//printf("MB %d %d bits\n", s->mb_x+s->mb_y*s->mb_stride, put_bits_count(&s->pb)); + } + } + + //not beautiful here but we must write it before flushing so it has to be here + if (ENABLE_MSMPEG4_ENCODER && s->msmpeg4_version && s->msmpeg4_version<4 && s->pict_type == I_TYPE) + msmpeg4_encode_ext_header(s); + + write_slice_end(s); + + /* Send the last GOB if RTP */ + if (s->avctx->rtp_callback) { + int number_mb = (mb_y - s->resync_mb_y)*s->mb_width - s->resync_mb_x; + pdif = pbBufPtr(&s->pb) - s->ptr_lastgob; + /* Call the RTP callback to send the last GOB */ + emms_c(); + s->avctx->rtp_callback(s->avctx, s->ptr_lastgob, pdif, number_mb); + } + + return 0; +} + +#define MERGE(field) dst->field += src->field; src->field=0 +static void merge_context_after_me(MpegEncContext *dst, MpegEncContext *src){ + MERGE(me.scene_change_score); + MERGE(me.mc_mb_var_sum_temp); + MERGE(me.mb_var_sum_temp); +} + +static void merge_context_after_encode(MpegEncContext *dst, MpegEncContext *src){ + int i; + + MERGE(dct_count[0]); //note, the other dct vars are not part of the context + MERGE(dct_count[1]); + MERGE(mv_bits); + MERGE(i_tex_bits); + MERGE(p_tex_bits); + MERGE(i_count); + MERGE(f_count); + MERGE(b_count); + MERGE(skip_count); + MERGE(misc_bits); + MERGE(error_count); + MERGE(padding_bug_score); + MERGE(current_picture.error[0]); + MERGE(current_picture.error[1]); + MERGE(current_picture.error[2]); + + if(dst->avctx->noise_reduction){ + for(i=0; i<64; i++){ + MERGE(dct_error_sum[0][i]); + MERGE(dct_error_sum[1][i]); + } + } + + assert(put_bits_count(&src->pb) % 8 ==0); + assert(put_bits_count(&dst->pb) % 8 ==0); + ff_copy_bits(&dst->pb, src->pb.buf, put_bits_count(&src->pb)); + flush_put_bits(&dst->pb); +} + +static int estimate_qp(MpegEncContext *s, int dry_run){ + if (s->next_lambda){ + s->current_picture_ptr->quality= + s->current_picture.quality = s->next_lambda; + if(!dry_run) s->next_lambda= 0; + } else if (!s->fixed_qscale) { + s->current_picture_ptr->quality= + s->current_picture.quality = ff_rate_estimate_qscale(s, dry_run); + if (s->current_picture.quality < 0) + return -1; + } + + if(s->adaptive_quant){ + switch(s->codec_id){ + case CODEC_ID_MPEG4: + ff_clean_mpeg4_qscales(s); + break; + case CODEC_ID_H263: + case CODEC_ID_H263P: + case CODEC_ID_FLV1: + ff_clean_h263_qscales(s); + break; + } + + s->lambda= s->lambda_table[0]; + //FIXME broken + }else + s->lambda= s->current_picture.quality; +//printf("%d %d\n", s->avctx->global_quality, s->current_picture.quality); + update_qscale(s); + return 0; +} + +static int encode_picture(MpegEncContext *s, int picture_number) +{ + int i; + int bits; + + s->picture_number = picture_number; + + /* Reset the average MB variance */ + s->me.mb_var_sum_temp = + s->me.mc_mb_var_sum_temp = 0; + + /* we need to initialize some time vars before we can encode b-frames */ + // RAL: Condition added for MPEG1VIDEO + if (s->codec_id == CODEC_ID_MPEG1VIDEO || s->codec_id == CODEC_ID_MPEG2VIDEO || (s->h263_pred && !s->h263_msmpeg4)) + ff_set_mpeg4_time(s, s->picture_number); //FIXME rename and use has_b_frames or similar + + s->me.scene_change_score=0; + +// s->lambda= s->current_picture_ptr->quality; //FIXME qscale / ... stuff for ME ratedistoration + + if(s->pict_type==I_TYPE){ + if(s->msmpeg4_version >= 3) s->no_rounding=1; + else s->no_rounding=0; + }else if(s->pict_type!=B_TYPE){ + if(s->flipflop_rounding || s->codec_id == CODEC_ID_H263P || s->codec_id == CODEC_ID_MPEG4) + s->no_rounding ^= 1; + } + + if(s->flags & CODEC_FLAG_PASS2){ + if (estimate_qp(s,1) < 0) + return -1; + ff_get_2pass_fcode(s); + }else if(!(s->flags & CODEC_FLAG_QSCALE)){ + if(s->pict_type==B_TYPE) + s->lambda= s->last_lambda_for[s->pict_type]; + else + s->lambda= s->last_lambda_for[s->last_non_b_pict_type]; + update_qscale(s); + } + + s->mb_intra=0; //for the rate distortion & bit compare functions + for(i=1; iavctx->thread_count; i++){ + ff_update_duplicate_context(s->thread_context[i], s); + } + + ff_init_me(s); + + /* Estimate motion for every MB */ + if(s->pict_type != I_TYPE){ + s->lambda = (s->lambda * s->avctx->me_penalty_compensation + 128)>>8; + s->lambda2= (s->lambda2* (int64_t)s->avctx->me_penalty_compensation + 128)>>8; + if(s->pict_type != B_TYPE && s->avctx->me_threshold==0){ + if((s->avctx->pre_me && s->last_non_b_pict_type==I_TYPE) || s->avctx->pre_me==2){ + s->avctx->execute(s->avctx, pre_estimate_motion_thread, (void**)&(s->thread_context[0]), NULL, s->avctx->thread_count); + } + } + + s->avctx->execute(s->avctx, estimate_motion_thread, (void**)&(s->thread_context[0]), NULL, s->avctx->thread_count); + }else /* if(s->pict_type == I_TYPE) */{ + /* I-Frame */ + for(i=0; imb_stride*s->mb_height; i++) + s->mb_type[i]= CANDIDATE_MB_TYPE_INTRA; + + if(!s->fixed_qscale){ + /* finding spatial complexity for I-frame rate control */ + s->avctx->execute(s->avctx, mb_var_thread, (void**)&(s->thread_context[0]), NULL, s->avctx->thread_count); + } + } + for(i=1; iavctx->thread_count; i++){ + merge_context_after_me(s, s->thread_context[i]); + } + s->current_picture.mc_mb_var_sum= s->current_picture_ptr->mc_mb_var_sum= s->me.mc_mb_var_sum_temp; + s->current_picture. mb_var_sum= s->current_picture_ptr-> mb_var_sum= s->me. mb_var_sum_temp; + emms_c(); + + if(s->me.scene_change_score > s->avctx->scenechange_threshold && s->pict_type == P_TYPE){ + s->pict_type= I_TYPE; + for(i=0; imb_stride*s->mb_height; i++) + s->mb_type[i]= CANDIDATE_MB_TYPE_INTRA; +//printf("Scene change detected, encoding as I Frame %d %d\n", s->current_picture.mb_var_sum, s->current_picture.mc_mb_var_sum); + } + + if(!s->umvplus){ + if(s->pict_type==P_TYPE || s->pict_type==S_TYPE) { + s->f_code= ff_get_best_fcode(s, s->p_mv_table, CANDIDATE_MB_TYPE_INTER); + + if(s->flags & CODEC_FLAG_INTERLACED_ME){ + int a,b; + a= ff_get_best_fcode(s, s->p_field_mv_table[0][0], CANDIDATE_MB_TYPE_INTER_I); //FIXME field_select + b= ff_get_best_fcode(s, s->p_field_mv_table[1][1], CANDIDATE_MB_TYPE_INTER_I); + s->f_code= FFMAX(s->f_code, FFMAX(a,b)); + } + + ff_fix_long_p_mvs(s); + ff_fix_long_mvs(s, NULL, 0, s->p_mv_table, s->f_code, CANDIDATE_MB_TYPE_INTER, 0); + if(s->flags & CODEC_FLAG_INTERLACED_ME){ + int j; + for(i=0; i<2; i++){ + for(j=0; j<2; j++) + ff_fix_long_mvs(s, s->p_field_select_table[i], j, + s->p_field_mv_table[i][j], s->f_code, CANDIDATE_MB_TYPE_INTER_I, 0); + } + } + } + + if(s->pict_type==B_TYPE){ + int a, b; + + a = ff_get_best_fcode(s, s->b_forw_mv_table, CANDIDATE_MB_TYPE_FORWARD); + b = ff_get_best_fcode(s, s->b_bidir_forw_mv_table, CANDIDATE_MB_TYPE_BIDIR); + s->f_code = FFMAX(a, b); + + a = ff_get_best_fcode(s, s->b_back_mv_table, CANDIDATE_MB_TYPE_BACKWARD); + b = ff_get_best_fcode(s, s->b_bidir_back_mv_table, CANDIDATE_MB_TYPE_BIDIR); + s->b_code = FFMAX(a, b); + + ff_fix_long_mvs(s, NULL, 0, s->b_forw_mv_table, s->f_code, CANDIDATE_MB_TYPE_FORWARD, 1); + ff_fix_long_mvs(s, NULL, 0, s->b_back_mv_table, s->b_code, CANDIDATE_MB_TYPE_BACKWARD, 1); + ff_fix_long_mvs(s, NULL, 0, s->b_bidir_forw_mv_table, s->f_code, CANDIDATE_MB_TYPE_BIDIR, 1); + ff_fix_long_mvs(s, NULL, 0, s->b_bidir_back_mv_table, s->b_code, CANDIDATE_MB_TYPE_BIDIR, 1); + if(s->flags & CODEC_FLAG_INTERLACED_ME){ + int dir, j; + for(dir=0; dir<2; dir++){ + for(i=0; i<2; i++){ + for(j=0; j<2; j++){ + int type= dir ? (CANDIDATE_MB_TYPE_BACKWARD_I|CANDIDATE_MB_TYPE_BIDIR_I) + : (CANDIDATE_MB_TYPE_FORWARD_I |CANDIDATE_MB_TYPE_BIDIR_I); + ff_fix_long_mvs(s, s->b_field_select_table[dir][i], j, + s->b_field_mv_table[dir][i][j], dir ? s->b_code : s->f_code, type, 1); + } + } + } + } + } + } + + if (estimate_qp(s, 0) < 0) + return -1; + + if(s->qscale < 3 && s->max_qcoeff<=128 && s->pict_type==I_TYPE && !(s->flags & CODEC_FLAG_QSCALE)) + s->qscale= 3; //reduce clipping problems + + if (s->out_format == FMT_MJPEG) { + /* for mjpeg, we do include qscale in the matrix */ + s->intra_matrix[0] = ff_mpeg1_default_intra_matrix[0]; + for(i=1;i<64;i++){ + int j= s->dsp.idct_permutation[i]; + + s->intra_matrix[j] = av_clip_uint8((ff_mpeg1_default_intra_matrix[i] * s->qscale) >> 3); + } + convert_matrix(&s->dsp, s->q_intra_matrix, s->q_intra_matrix16, + s->intra_matrix, s->intra_quant_bias, 8, 8, 1); + s->qscale= 8; + } + + //FIXME var duplication + s->current_picture_ptr->key_frame= + s->current_picture.key_frame= s->pict_type == I_TYPE; //FIXME pic_ptr + s->current_picture_ptr->pict_type= + s->current_picture.pict_type= s->pict_type; + + if(s->current_picture.key_frame) + s->picture_in_gop_number=0; + + s->last_bits= put_bits_count(&s->pb); + switch(s->out_format) { + case FMT_MJPEG: + if (ENABLE_MJPEG_ENCODER) + ff_mjpeg_encode_picture_header(s); + break; + case FMT_H261: + if (ENABLE_H261_ENCODER) + ff_h261_encode_picture_header(s, picture_number); + break; + case FMT_H263: + if (ENABLE_WMV2_ENCODER && s->codec_id == CODEC_ID_WMV2) + ff_wmv2_encode_picture_header(s, picture_number); + else if (ENABLE_MSMPEG4_ENCODER && s->h263_msmpeg4) + msmpeg4_encode_picture_header(s, picture_number); + else if (s->h263_pred) + mpeg4_encode_picture_header(s, picture_number); + else if (ENABLE_RV10_ENCODER && s->codec_id == CODEC_ID_RV10) + rv10_encode_picture_header(s, picture_number); + else if (ENABLE_RV20_ENCODER && s->codec_id == CODEC_ID_RV20) + rv20_encode_picture_header(s, picture_number); + else if (s->codec_id == CODEC_ID_FLV1) + ff_flv_encode_picture_header(s, picture_number); + else + h263_encode_picture_header(s, picture_number); + break; + case FMT_MPEG1: + mpeg1_encode_picture_header(s, picture_number); + break; + case FMT_H264: + break; + default: + assert(0); + } + bits= put_bits_count(&s->pb); + s->header_bits= bits - s->last_bits; + + for(i=1; iavctx->thread_count; i++){ + update_duplicate_context_after_me(s->thread_context[i], s); + } + s->avctx->execute(s->avctx, encode_thread, (void**)&(s->thread_context[0]), NULL, s->avctx->thread_count); + for(i=1; iavctx->thread_count; i++){ + merge_context_after_encode(s, s->thread_context[i]); + } + emms_c(); + return 0; +} + +void denoise_dct_c(MpegEncContext *s, DCTELEM *block){ + const int intra= s->mb_intra; + int i; + + s->dct_count[intra]++; + + for(i=0; i<64; i++){ + int level= block[i]; + + if(level){ + if(level>0){ + s->dct_error_sum[intra][i] += level; + level -= s->dct_offset[intra][i]; + if(level<0) level=0; + }else{ + s->dct_error_sum[intra][i] -= level; + level += s->dct_offset[intra][i]; + if(level>0) level=0; + } + block[i]= level; + } + } +} + +int dct_quantize_trellis_c(MpegEncContext *s, + DCTELEM *block, int n, + int qscale, int *overflow){ + const int *qmat; + const uint8_t *scantable= s->intra_scantable.scantable; + const uint8_t *perm_scantable= s->intra_scantable.permutated; + int max=0; + unsigned int threshold1, threshold2; + int bias=0; + int run_tab[65]; + int level_tab[65]; + int score_tab[65]; + int survivor[65]; + int survivor_count; + int last_run=0; + int last_level=0; + int last_score= 0; + int last_i; + int coeff[2][64]; + int coeff_count[64]; + int qmul, qadd, start_i, last_non_zero, i, dc; + const int esc_length= s->ac_esc_length; + uint8_t * length; + uint8_t * last_length; + const int lambda= s->lambda2 >> (FF_LAMBDA_SHIFT - 6); + + s->dsp.fdct (block); + + if(s->dct_error_sum) + s->denoise_dct(s, block); + qmul= qscale*16; + qadd= ((qscale-1)|1)*8; + + if (s->mb_intra) { + int q; + if (!s->h263_aic) { + if (n < 4) + q = s->y_dc_scale; + else + q = s->c_dc_scale; + q = q << 3; + } else{ + /* For AIC we skip quant/dequant of INTRADC */ + q = 1 << 3; + qadd=0; + } + + /* note: block[0] is assumed to be positive */ + block[0] = (block[0] + (q >> 1)) / q; + start_i = 1; + last_non_zero = 0; + qmat = s->q_intra_matrix[qscale]; + if(s->mpeg_quant || s->out_format == FMT_MPEG1) + bias= 1<<(QMAT_SHIFT-1); + length = s->intra_ac_vlc_length; + last_length= s->intra_ac_vlc_last_length; + } else { + start_i = 0; + last_non_zero = -1; + qmat = s->q_inter_matrix[qscale]; + length = s->inter_ac_vlc_length; + last_length= s->inter_ac_vlc_last_length; + } + last_i= start_i; + + threshold1= (1<=start_i; i--) { + const int j = scantable[i]; + int level = block[j] * qmat[j]; + + if(((unsigned)(level+threshold1))>threshold2){ + last_non_zero = i; + break; + } + } + + for(i=start_i; i<=last_non_zero; i++) { + const int j = scantable[i]; + int level = block[j] * qmat[j]; + +// if( bias+level >= (1<<(QMAT_SHIFT - 3)) +// || bias-level >= (1<<(QMAT_SHIFT - 3))){ + if(((unsigned)(level+threshold1))>threshold2){ + if(level>0){ + level= (bias + level)>>QMAT_SHIFT; + coeff[0][i]= level; + coeff[1][i]= level-1; +// coeff[2][k]= level-2; + }else{ + level= (bias - level)>>QMAT_SHIFT; + coeff[0][i]= -level; + coeff[1][i]= -level+1; +// coeff[2][k]= -level+2; + } + coeff_count[i]= FFMIN(level, 2); + assert(coeff_count[i]); + max |=level; + }else{ + coeff[0][i]= (level>>31)|1; + coeff_count[i]= 1; + } + } + + *overflow= s->max_qcoeff < max; //overflow might have happened + + if(last_non_zero < start_i){ + memset(block + start_i, 0, (64-start_i)*sizeof(DCTELEM)); + return last_non_zero; + } + + score_tab[start_i]= 0; + survivor[0]= start_i; + survivor_count= 1; + + for(i=start_i; i<=last_non_zero; i++){ + int level_index, j; + const int dct_coeff= FFABS(block[ scantable[i] ]); + const int zero_distoration= dct_coeff*dct_coeff; + int best_score=256*256*256*120; + for(level_index=0; level_index < coeff_count[i]; level_index++){ + int distoration; + int level= coeff[level_index][i]; + const int alevel= FFABS(level); + int unquant_coeff; + + assert(level); + + if(s->out_format == FMT_H263){ + unquant_coeff= alevel*qmul + qadd; + }else{ //MPEG1 + j= s->dsp.idct_permutation[ scantable[i] ]; //FIXME optimize + if(s->mb_intra){ + unquant_coeff = (int)( alevel * qscale * s->intra_matrix[j]) >> 3; + unquant_coeff = (unquant_coeff - 1) | 1; + }else{ + unquant_coeff = ((( alevel << 1) + 1) * qscale * ((int) s->inter_matrix[j])) >> 4; + unquant_coeff = (unquant_coeff - 1) | 1; + } + unquant_coeff<<= 3; + } + + distoration= (unquant_coeff - dct_coeff) * (unquant_coeff - dct_coeff) - zero_distoration; + level+=64; + if((level&(~127)) == 0){ + for(j=survivor_count-1; j>=0; j--){ + int run= i - survivor[j]; + int score= distoration + length[UNI_AC_ENC_INDEX(run, level)]*lambda; + score += score_tab[i-run]; + + if(score < best_score){ + best_score= score; + run_tab[i+1]= run; + level_tab[i+1]= level-64; + } + } + + if(s->out_format == FMT_H263){ + for(j=survivor_count-1; j>=0; j--){ + int run= i - survivor[j]; + int score= distoration + last_length[UNI_AC_ENC_INDEX(run, level)]*lambda; + score += score_tab[i-run]; + if(score < last_score){ + last_score= score; + last_run= run; + last_level= level-64; + last_i= i+1; + } + } + } + }else{ + distoration += esc_length*lambda; + for(j=survivor_count-1; j>=0; j--){ + int run= i - survivor[j]; + int score= distoration + score_tab[i-run]; + + if(score < best_score){ + best_score= score; + run_tab[i+1]= run; + level_tab[i+1]= level-64; + } + } + + if(s->out_format == FMT_H263){ + for(j=survivor_count-1; j>=0; j--){ + int run= i - survivor[j]; + int score= distoration + score_tab[i-run]; + if(score < last_score){ + last_score= score; + last_run= run; + last_level= level-64; + last_i= i+1; + } + } + } + } + } + + score_tab[i+1]= best_score; + + //Note: there is a vlc code in mpeg4 which is 1 bit shorter then another one with a shorter run and the same level + if(last_non_zero <= 27){ + for(; survivor_count; survivor_count--){ + if(score_tab[ survivor[survivor_count-1] ] <= best_score) + break; + } + }else{ + for(; survivor_count; survivor_count--){ + if(score_tab[ survivor[survivor_count-1] ] <= best_score + lambda) + break; + } + } + + survivor[ survivor_count++ ]= i+1; + } + + if(s->out_format != FMT_H263){ + last_score= 256*256*256*120; + for(i= survivor[0]; i<=last_non_zero + 1; i++){ + int score= score_tab[i]; + if(i) score += lambda*2; //FIXME exacter? + + if(score < last_score){ + last_score= score; + last_i= i; + last_level= level_tab[i]; + last_run= run_tab[i]; + } + } + } + + s->coded_score[n] = last_score; + + dc= FFABS(block[0]); + last_non_zero= last_i - 1; + memset(block + start_i, 0, (64-start_i)*sizeof(DCTELEM)); + + if(last_non_zero < start_i) + return last_non_zero; + + if(last_non_zero == 0 && start_i == 0){ + int best_level= 0; + int best_score= dc * dc; + + for(i=0; iout_format == FMT_H263){ + unquant_coeff= (alevel*qmul + qadd)>>3; + }else{ //MPEG1 + unquant_coeff = ((( alevel << 1) + 1) * qscale * ((int) s->inter_matrix[0])) >> 4; + unquant_coeff = (unquant_coeff - 1) | 1; + } + unquant_coeff = (unquant_coeff + 4) >> 3; + unquant_coeff<<= 3 + 3; + + distortion= (unquant_coeff - dc) * (unquant_coeff - dc); + level+=64; + if((level&(~127)) == 0) score= distortion + last_length[UNI_AC_ENC_INDEX(0, level)]*lambda; + else score= distortion + esc_length*lambda; + + if(score < best_score){ + best_score= score; + best_level= level - 64; + } + } + block[0]= best_level; + s->coded_score[n] = best_score - dc*dc; + if(best_level == 0) return -1; + else return last_non_zero; + } + + i= last_i; + assert(last_level); + + block[ perm_scantable[last_non_zero] ]= last_level; + i -= last_run + 1; + + for(; i>start_i; i -= run_tab[i] + 1){ + block[ perm_scantable[i-1] ]= level_tab[i]; + } + + return last_non_zero; +} + +//#define REFINE_STATS 1 +static int16_t basis[64][64]; + +static void build_basis(uint8_t *perm){ + int i, j, x, y; + emms_c(); + for(i=0; i<8; i++){ + for(j=0; j<8; j++){ + for(y=0; y<8; y++){ + for(x=0; x<8; x++){ + double s= 0.25*(1<intra_scantable.scantable; + const uint8_t *perm_scantable= s->intra_scantable.permutated; +// unsigned int threshold1, threshold2; +// int bias=0; + int run_tab[65]; + int prev_run=0; + int prev_level=0; + int qmul, qadd, start_i, last_non_zero, i, dc; + uint8_t * length; + uint8_t * last_length; + int lambda; + int rle_index, run, q = 1, sum; //q is only used when s->mb_intra is true +#ifdef REFINE_STATS +static int count=0; +static int after_last=0; +static int to_zero=0; +static int from_zero=0; +static int raise=0; +static int lower=0; +static int messed_sign=0; +#endif + + if(basis[0][0] == 0) + build_basis(s->dsp.idct_permutation); + + qmul= qscale*2; + qadd= (qscale-1)|1; + if (s->mb_intra) { + if (!s->h263_aic) { + if (n < 4) + q = s->y_dc_scale; + else + q = s->c_dc_scale; + } else{ + /* For AIC we skip quant/dequant of INTRADC */ + q = 1; + qadd=0; + } + q <<= RECON_SHIFT-3; + /* note: block[0] is assumed to be positive */ + dc= block[0]*q; +// block[0] = (block[0] + (q >> 1)) / q; + start_i = 1; + qmat = s->q_intra_matrix[qscale]; +// if(s->mpeg_quant || s->out_format == FMT_MPEG1) +// bias= 1<<(QMAT_SHIFT-1); + length = s->intra_ac_vlc_length; + last_length= s->intra_ac_vlc_last_length; + } else { + dc= 0; + start_i = 0; + qmat = s->q_inter_matrix[qscale]; + length = s->inter_ac_vlc_length; + last_length= s->inter_ac_vlc_last_length; + } + last_non_zero = s->block_last_index[n]; + +#ifdef REFINE_STATS +{START_TIMER +#endif + dc += (1<<(RECON_SHIFT-1)); + for(i=0; i<64; i++){ + rem[i]= dc - (orig[i]<0); + assert(w<(1<<6)); + sum += w*w; + } + lambda= sum*(uint64_t)s->lambda2 >> (FF_LAMBDA_SHIFT - 6 + 6 + 6 + 6); +#ifdef REFINE_STATS +{START_TIMER +#endif + run=0; + rle_index=0; + for(i=start_i; i<=last_non_zero; i++){ + int j= perm_scantable[i]; + const int level= block[j]; + int coeff; + + if(level){ + if(level<0) coeff= qmul*level - qadd; + else coeff= qmul*level + qadd; + run_tab[rle_index++]=run; + run=0; + + s->dsp.add_8x8basis(rem, basis[j], coeff); + }else{ + run++; + } + } +#ifdef REFINE_STATS +if(last_non_zero>0){ +STOP_TIMER("init rem[]") +} +} + +{START_TIMER +#endif + for(;;){ + int best_score=s->dsp.try_8x8basis(rem, weight, basis[0], 0); + int best_coeff=0; + int best_change=0; + int run2, best_unquant_change=0, analyze_gradient; +#ifdef REFINE_STATS +{START_TIMER +#endif + analyze_gradient = last_non_zero > 2 || s->avctx->quantizer_noise_shaping >= 3; + + if(analyze_gradient){ +#ifdef REFINE_STATS +{START_TIMER +#endif + for(i=0; i<64; i++){ + int w= weight[i]; + + d1[i] = (rem[i]*w*w + (1<<(RECON_SHIFT+12-1)))>>(RECON_SHIFT+12); + } +#ifdef REFINE_STATS +STOP_TIMER("rem*w*w")} +{START_TIMER +#endif + s->dsp.fdct(d1); +#ifdef REFINE_STATS +STOP_TIMER("dct")} +#endif + } + + if(start_i){ + const int level= block[0]; + int change, old_coeff; + + assert(s->mb_intra); + + old_coeff= q*level; + + for(change=-1; change<=1; change+=2){ + int new_level= level + change; + int score, new_coeff; + + new_coeff= q*new_level; + if(new_coeff >= 2048 || new_coeff < 0) + continue; + + score= s->dsp.try_8x8basis(rem, weight, basis[0], new_coeff - old_coeff); + if(scoreavctx->quantizer_noise_shaping < 3 && i > last_non_zero + 1) + break; + + if(level){ + if(level<0) old_coeff= qmul*level - qadd; + else old_coeff= qmul*level + qadd; + run2= run_tab[rle_index++]; //FIXME ! maybe after last + }else{ + old_coeff=0; + run2--; + assert(run2>=0 || i >= last_non_zero ); + } + + for(change=-1; change<=1; change+=2){ + int new_level= level + change; + int score, new_coeff, unquant_change; + + score=0; + if(s->avctx->quantizer_noise_shaping < 2 && FFABS(new_level) > FFABS(level)) + continue; + + if(new_level){ + if(new_level<0) new_coeff= qmul*new_level - qadd; + else new_coeff= qmul*new_level + qadd; + if(new_coeff >= 2048 || new_coeff <= -2048) + continue; + //FIXME check for overflow + + if(level){ + if(level < 63 && level > -63){ + if(i < last_non_zero) + score += length[UNI_AC_ENC_INDEX(run, new_level+64)] + - length[UNI_AC_ENC_INDEX(run, level+64)]; + else + score += last_length[UNI_AC_ENC_INDEX(run, new_level+64)] + - last_length[UNI_AC_ENC_INDEX(run, level+64)]; + } + }else{ + assert(FFABS(new_level)==1); + + if(analyze_gradient){ + int g= d1[ scantable[i] ]; + if(g && (g^new_level) >= 0) + continue; + } + + if(i < last_non_zero){ + int next_i= i + run2 + 1; + int next_level= block[ perm_scantable[next_i] ] + 64; + + if(next_level&(~127)) + next_level= 0; + + if(next_i < last_non_zero) + score += length[UNI_AC_ENC_INDEX(run, 65)] + + length[UNI_AC_ENC_INDEX(run2, next_level)] + - length[UNI_AC_ENC_INDEX(run + run2 + 1, next_level)]; + else + score += length[UNI_AC_ENC_INDEX(run, 65)] + + last_length[UNI_AC_ENC_INDEX(run2, next_level)] + - last_length[UNI_AC_ENC_INDEX(run + run2 + 1, next_level)]; + }else{ + score += last_length[UNI_AC_ENC_INDEX(run, 65)]; + if(prev_level){ + score += length[UNI_AC_ENC_INDEX(prev_run, prev_level)] + - last_length[UNI_AC_ENC_INDEX(prev_run, prev_level)]; + } + } + } + }else{ + new_coeff=0; + assert(FFABS(level)==1); + + if(i < last_non_zero){ + int next_i= i + run2 + 1; + int next_level= block[ perm_scantable[next_i] ] + 64; + + if(next_level&(~127)) + next_level= 0; + + if(next_i < last_non_zero) + score += length[UNI_AC_ENC_INDEX(run + run2 + 1, next_level)] + - length[UNI_AC_ENC_INDEX(run2, next_level)] + - length[UNI_AC_ENC_INDEX(run, 65)]; + else + score += last_length[UNI_AC_ENC_INDEX(run + run2 + 1, next_level)] + - last_length[UNI_AC_ENC_INDEX(run2, next_level)] + - length[UNI_AC_ENC_INDEX(run, 65)]; + }else{ + score += -last_length[UNI_AC_ENC_INDEX(run, 65)]; + if(prev_level){ + score += last_length[UNI_AC_ENC_INDEX(prev_run, prev_level)] + - length[UNI_AC_ENC_INDEX(prev_run, prev_level)]; + } + } + } + + score *= lambda; + + unquant_change= new_coeff - old_coeff; + assert((score < 100*lambda && score > -100*lambda) || lambda==0); + + score+= s->dsp.try_8x8basis(rem, weight, basis[j], unquant_change); + if(score last_non_zero){ + last_non_zero= best_coeff; + assert(block[j]); +#ifdef REFINE_STATS +after_last++; +#endif + }else{ +#ifdef REFINE_STATS +if(block[j]){ + if(block[j] - best_change){ + if(FFABS(block[j]) > FFABS(block[j] - best_change)){ + raise++; + }else{ + lower++; + } + }else{ + from_zero++; + } +}else{ + to_zero++; +} +#endif + for(; last_non_zero>=start_i; last_non_zero--){ + if(block[perm_scantable[last_non_zero]]) + break; + } + } +#ifdef REFINE_STATS +count++; +if(256*256*256*64 % count == 0){ + printf("after_last:%d to_zero:%d from_zero:%d raise:%d lower:%d sign:%d xyp:%d/%d/%d\n", after_last, to_zero, from_zero, raise, lower, messed_sign, s->mb_x, s->mb_y, s->picture_number); +} +#endif + run=0; + rle_index=0; + for(i=start_i; i<=last_non_zero; i++){ + int j= perm_scantable[i]; + const int level= block[j]; + + if(level){ + run_tab[rle_index++]=run; + run=0; + }else{ + run++; + } + } + + s->dsp.add_8x8basis(rem, basis[j], best_unquant_change); + }else{ + break; + } + } +#ifdef REFINE_STATS +if(last_non_zero>0){ +STOP_TIMER("iterative search") +} +} +#endif + + return last_non_zero; +} + +int dct_quantize_c(MpegEncContext *s, + DCTELEM *block, int n, + int qscale, int *overflow) +{ + int i, j, level, last_non_zero, q, start_i; + const int *qmat; + const uint8_t *scantable= s->intra_scantable.scantable; + int bias; + int max=0; + unsigned int threshold1, threshold2; + + s->dsp.fdct (block); + + if(s->dct_error_sum) + s->denoise_dct(s, block); + + if (s->mb_intra) { + if (!s->h263_aic) { + if (n < 4) + q = s->y_dc_scale; + else + q = s->c_dc_scale; + q = q << 3; + } else + /* For AIC we skip quant/dequant of INTRADC */ + q = 1 << 3; + + /* note: block[0] is assumed to be positive */ + block[0] = (block[0] + (q >> 1)) / q; + start_i = 1; + last_non_zero = 0; + qmat = s->q_intra_matrix[qscale]; + bias= s->intra_quant_bias<<(QMAT_SHIFT - QUANT_BIAS_SHIFT); + } else { + start_i = 0; + last_non_zero = -1; + qmat = s->q_inter_matrix[qscale]; + bias= s->inter_quant_bias<<(QMAT_SHIFT - QUANT_BIAS_SHIFT); + } + threshold1= (1<=start_i;i--) { + j = scantable[i]; + level = block[j] * qmat[j]; + + if(((unsigned)(level+threshold1))>threshold2){ + last_non_zero = i; + break; + }else{ + block[j]=0; + } + } + for(i=start_i; i<=last_non_zero; i++) { + j = scantable[i]; + level = block[j] * qmat[j]; + +// if( bias+level >= (1<= (1<threshold2){ + if(level>0){ + level= (bias + level)>>QMAT_SHIFT; + block[j]= level; + }else{ + level= (bias - level)>>QMAT_SHIFT; + block[j]= -level; + } + max |=level; + }else{ + block[j]=0; + } + } + *overflow= s->max_qcoeff < max; //overflow might have happened + + /* we need this permutation so that we correct the IDCT, we only permute the !=0 elements */ + if (s->dsp.idct_permutation_type != FF_NO_IDCT_PERM) + ff_block_permute(block, s->dsp.idct_permutation, scantable, last_non_zero); + + return last_non_zero; +} + +AVCodec h263_encoder = { + "h263", + CODEC_TYPE_VIDEO, + CODEC_ID_H263, + sizeof(MpegEncContext), + MPV_encode_init, + MPV_encode_picture, + MPV_encode_end, + .pix_fmts= (enum PixelFormat[]){PIX_FMT_YUV420P, -1}, +}; + +AVCodec h263p_encoder = { + "h263p", + CODEC_TYPE_VIDEO, + CODEC_ID_H263P, + sizeof(MpegEncContext), + MPV_encode_init, + MPV_encode_picture, + MPV_encode_end, + .pix_fmts= (enum PixelFormat[]){PIX_FMT_YUV420P, -1}, +}; + +AVCodec flv_encoder = { + "flv", + CODEC_TYPE_VIDEO, + CODEC_ID_FLV1, + sizeof(MpegEncContext), + MPV_encode_init, + MPV_encode_picture, + MPV_encode_end, + .pix_fmts= (enum PixelFormat[]){PIX_FMT_YUV420P, -1}, +}; + +AVCodec rv10_encoder = { + "rv10", + CODEC_TYPE_VIDEO, + CODEC_ID_RV10, + sizeof(MpegEncContext), + MPV_encode_init, + MPV_encode_picture, + MPV_encode_end, + .pix_fmts= (enum PixelFormat[]){PIX_FMT_YUV420P, -1}, +}; + +AVCodec rv20_encoder = { + "rv20", + CODEC_TYPE_VIDEO, + CODEC_ID_RV20, + sizeof(MpegEncContext), + MPV_encode_init, + MPV_encode_picture, + MPV_encode_end, + .pix_fmts= (enum PixelFormat[]){PIX_FMT_YUV420P, -1}, +}; + +AVCodec mpeg4_encoder = { + "mpeg4", + CODEC_TYPE_VIDEO, + CODEC_ID_MPEG4, + sizeof(MpegEncContext), + MPV_encode_init, + MPV_encode_picture, + MPV_encode_end, + .pix_fmts= (enum PixelFormat[]){PIX_FMT_YUV420P, -1}, + .capabilities= CODEC_CAP_DELAY, +}; + +AVCodec msmpeg4v1_encoder = { + "msmpeg4v1", + CODEC_TYPE_VIDEO, + CODEC_ID_MSMPEG4V1, + sizeof(MpegEncContext), + MPV_encode_init, + MPV_encode_picture, + MPV_encode_end, + .pix_fmts= (enum PixelFormat[]){PIX_FMT_YUV420P, -1}, +}; + +AVCodec msmpeg4v2_encoder = { + "msmpeg4v2", + CODEC_TYPE_VIDEO, + CODEC_ID_MSMPEG4V2, + sizeof(MpegEncContext), + MPV_encode_init, + MPV_encode_picture, + MPV_encode_end, + .pix_fmts= (enum PixelFormat[]){PIX_FMT_YUV420P, -1}, +}; + +AVCodec msmpeg4v3_encoder = { + "msmpeg4", + CODEC_TYPE_VIDEO, + CODEC_ID_MSMPEG4V3, + sizeof(MpegEncContext), + MPV_encode_init, + MPV_encode_picture, + MPV_encode_end, + .pix_fmts= (enum PixelFormat[]){PIX_FMT_YUV420P, -1}, +}; + +AVCodec wmv1_encoder = { + "wmv1", + CODEC_TYPE_VIDEO, + CODEC_ID_WMV1, + sizeof(MpegEncContext), + MPV_encode_init, + MPV_encode_picture, + MPV_encode_end, + .pix_fmts= (enum PixelFormat[]){PIX_FMT_YUV420P, -1}, +};