*
* You should have received a copy of the GNU Lesser General Public
* License along with this library; if not, write to the Free Software
- * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*
* 4MV & hq & b-frame encoding stuff by Michael Niedermayer <michaelni@gmx.at>
*/
#include <limits.h>
#ifdef USE_FASTMEMCPY
-#include "fastmemcpy.h"
+#include "libvo/fastmemcpy.h"
#endif
//#undef NDEBUG
//#include <assert.h>
#ifdef CONFIG_ENCODERS
-static void encode_picture(MpegEncContext *s, int picture_number);
+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);
DCTELEM *block, int n, int qscale);
static void dct_unquantize_mpeg2_intra_c(MpegEncContext *s,
DCTELEM *block, int n, int qscale);
+static void dct_unquantize_mpeg2_intra_bitexact(MpegEncContext *s,
+ DCTELEM *block, int n, int qscale);
static void dct_unquantize_mpeg2_inter_c(MpegEncContext *s,
DCTELEM *block, int n, int qscale);
static void dct_unquantize_h263_intra_c(MpegEncContext *s,
}
#ifdef CONFIG_ENCODERS
-void ff_write_quant_matrix(PutBitContext *pb, int16_t *matrix){
+void ff_write_quant_matrix(PutBitContext *pb, uint16_t *matrix){
int i;
if(matrix){
}
#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;
+
+ assert(p<=end);
+ if(p>=end)
+ return end;
+
+ for(i=0; i<3; i++){
+ uint32_t tmp= *state << 8;
+ *state= tmp + *(p++);
+ if(tmp == 0x100 || p==end)
+ return p;
+ }
+
+ while(p<end){
+ if (p[-1] > 1 ) p+= 3;
+ else if(p[-2] ) p+= 2;
+ else if(p[-3]|(p[-1]-1)) p++;
+ else{
+ p++;
+ break;
+ }
+ }
+
+ p= FFMIN(p, end)-4;
+ *state= be2me_32(unaligned32(p));
+
+ return p+4;
+}
+
/* init common dct for both encoder and decoder */
int DCT_common_init(MpegEncContext *s)
{
s->dct_unquantize_mpeg1_intra = dct_unquantize_mpeg1_intra_c;
s->dct_unquantize_mpeg1_inter = dct_unquantize_mpeg1_inter_c;
s->dct_unquantize_mpeg2_intra = dct_unquantize_mpeg2_intra_c;
+ if(s->flags & CODEC_FLAG_BITEXACT)
+ s->dct_unquantize_mpeg2_intra = dct_unquantize_mpeg2_intra_bitexact;
s->dct_unquantize_mpeg2_inter = dct_unquantize_mpeg2_inter_c;
#ifdef CONFIG_ENCODERS
dst->type= FF_BUFFER_TYPE_COPY;
}
+#ifdef CONFIG_ENCODERS
static void copy_picture_attributes(MpegEncContext *s, AVFrame *dst, AVFrame *src){
int i;
}
}
}
+#endif
/**
* allocates a Picture
int i;
// edge emu needs blocksize + filter length - 1 (=17x17 for halfpel / 21x21 for h264)
- CHECKED_ALLOCZ(s->allocated_edge_emu_buffer, (s->width+64)*2*17*2); //(width + edge + align)*interlaced*MBsize*tolerance
- s->edge_emu_buffer= s->allocated_edge_emu_buffer + (s->width+64)*2*17;
+ CHECKED_ALLOCZ(s->allocated_edge_emu_buffer, (s->width+64)*2*21*2); //(width + edge + align)*interlaced*MBsize*tolerance
+ s->edge_emu_buffer= s->allocated_edge_emu_buffer + (s->width+64)*2*21;
//FIXME should be linesize instead of s->width*2 but that isnt known before get_buffer()
CHECKED_ALLOCZ(s->me.scratchpad, (s->width+64)*4*16*2*sizeof(uint8_t))
//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(partitioned_frame); //FIXME don't set in encode_header
#undef COPY
}
+#endif
/**
* sets the given MpegEncContext to common defaults (same for encoding and decoding).
{
int y_size, c_size, yc_size, i, mb_array_size, mv_table_size, x, y;
- if(s->avctx->thread_count > MAX_THREADS || (16*s->avctx->thread_count > s->height && s->height)){
+ s->mb_height = (s->height + 15) / 16;
+
+ if(s->avctx->thread_count > MAX_THREADS || (s->avctx->thread_count > s->mb_height && s->mb_height)){
av_log(s->avctx, AV_LOG_ERROR, "too many threads\n");
return -1;
}
s->flags2= s->avctx->flags2;
s->mb_width = (s->width + 15) / 16;
- s->mb_height = (s->height + 15) / 16;
s->mb_stride = s->mb_width + 1;
s->b8_stride = s->mb_width*2 + 1;
s->b4_stride = s->mb_width*4 + 1;
MPV_encode_defaults(s);
- if(avctx->pix_fmt != PIX_FMT_YUVJ420P && avctx->pix_fmt != PIX_FMT_YUV420P){
- av_log(avctx, AV_LOG_ERROR, "only YUV420 is supported\n");
- return -1;
- }
-
- if(avctx->codec_id == CODEC_ID_MJPEG || avctx->codec_id == CODEC_ID_LJPEG){
- if(avctx->strict_std_compliance>FF_COMPLIANCE_INOFFICIAL && avctx->pix_fmt != PIX_FMT_YUVJ420P){
+ 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;
}
- }else{
- if(avctx->strict_std_compliance>FF_COMPLIANCE_INOFFICIAL && avctx->pix_fmt != PIX_FMT_YUV420P){
- av_log(avctx, AV_LOG_ERROR, "colorspace not supported\n");
+ 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){
+ 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->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);
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((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->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))){
}
if(avctx->b_frame_strategy && (avctx->flags&CODEC_FLAG_PASS2)){
- av_log(avctx, AV_LOG_ERROR, "b_frame_strategy must be 0 on the second pass");
- return -1;
+ 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);
s->intra_only = 1; /* force intra only for jpeg */
s->mjpeg_write_tables = avctx->codec->id != CODEC_ID_JPEGLS;
s->mjpeg_data_only_frames = 0; /* write all the needed headers */
- s->mjpeg_vsample[0] = 1<<chroma_v_shift;
- s->mjpeg_vsample[1] = 1;
- s->mjpeg_vsample[2] = 1;
- s->mjpeg_hsample[0] = 1<<chroma_h_shift;
- s->mjpeg_hsample[1] = 1;
- s->mjpeg_hsample[2] = 1;
+ 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 (mjpeg_init(s) < 0)
return -1;
avctx->delay=0;
{
MpegEncContext *s = avctx->priv_data;
-#ifdef STATS
- print_stats();
-#endif
-
ff_rate_control_uninit(s);
MPV_common_end(s);
* @param color color of the arrow
*/
static void draw_line(uint8_t *buf, int sx, int sy, int ex, int ey, int w, int h, int stride, int color){
- int t, x, y, fr, f;
+ int x, y, fr, f;
sx= clip(sx, 0, w-1);
sy= clip(sy, 0, h-1);
if(ABS(ex - sx) > ABS(ey - sy)){
if(sx > ex){
- t=sx; sx=ex; ex=t;
- t=sy; sy=ey; ey=t;
+ SWAP(int, sx, ex);
+ SWAP(int, sy, ey);
}
buf+= sx + sy*stride;
ex-= sx;
}
}else{
if(sy > ey){
- t=sx; sx=ex; ex=t;
- t=sy; sy=ey; ey=t;
+ SWAP(int, sx, ex);
+ SWAP(int, sy, ey);
}
buf+= sx + sy*stride;
ey-= sy;
const int width = s->avctx->width;
const int height= s->avctx->height;
const int mv_sample_log2= 4 - pict->motion_subsample_log2;
- const int mv_stride= (s->mb_width << mv_sample_log2) + 1;
+ const int mv_stride= (s->mb_width << mv_sample_log2) + (s->codec_id == CODEC_ID_H264 ? 0 : 1);
s->low_delay=0; //needed to see the vectors without trashing the buffers
avcodec_get_chroma_sub_sample(s->avctx->pix_fmt, &h_chroma_shift, &v_chroma_shift);
}
alloc_picture(s, (Picture*)pic, 1);
}else{
- int offset= 16;
i= ff_find_unused_picture(s, 0);
pic= (AVFrame*)&s->picture[i];
alloc_picture(s, (Picture*)pic, 0);
- if( pic->data[0] + offset == pic_arg->data[0]
- && pic->data[1] + offset == pic_arg->data[1]
- && pic->data[2] + offset == pic_arg->data[2]){
+ 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;
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] + offset;
+ 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);
AVCodecContext *c= avcodec_alloc_context();
AVFrame input[FF_MAX_B_FRAMES+2];
const int scale= s->avctx->brd_scale;
- int i, j, out_size;
+ 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);
- ImgReSampleContext *resample;
int64_t best_rd= INT64_MAX;
int best_b_count= -1;
- const int lambda2= s->lambda2;
+
+ 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 *ABS(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)/2 ) >> FF_LAMBDA_SHIFT;
c->width = s->width >> scale;
c->height= s->height>> scale;
if (avcodec_open(c, codec) < 0)
return -1;
- resample= img_resample_init(c->width, c->height, s->width, s->height); //FIXME use sws
-
for(i=0; i<s->max_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;
+
+ if(pre_input_ptr)
+ 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;
+ }
avcodec_get_frame_defaults(&input[i]);
input[i].data[0]= av_malloc(ysize + 2*csize);
input[i].linesize[1]=
input[i].linesize[2]= c->width/2;
- if(!i || s->input_picture[i-1])
- img_resample(resample, &input[i], i ? s->input_picture[i-1] : s->next_picture_ptr);
+ if(!i || s->input_picture[i-1]){
+ 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; j<s->max_b_frames+1; j++){
c->error[0]= c->error[1]= c->error[2]= 0;
input[0].pict_type= I_TYPE;
- input[0].quality= 2 * FF_QP2LAMBDA;
+ input[0].quality= 1 * FF_QP2LAMBDA;
out_size = avcodec_encode_video(c, outbuf, outbuf_size, &input[0]);
- rd += (out_size * lambda2) >> FF_LAMBDA_SHIFT;
+// rd += (out_size * lambda2) >> FF_LAMBDA_SHIFT;
for(i=0; i<s->max_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= s->last_lambda_for[input[i+1].pict_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;
+ 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;
+ rd += (out_size * lambda2) >> (FF_LAMBDA_SHIFT - 3);
}
rd += c->error[0] + c->error[1] + c->error[2];
av_freep(&outbuf);
avcodec_close(c);
av_freep(&c);
- img_resample_close(resample);
for(i=0; i<s->max_b_frames+2; i++){
av_freep(&input[i].data[0]);
}
}
for(i=0; i<s->max_b_frames+1; i++){
- if(s->input_picture[i]==NULL || s->input_picture[i]->b_frame_score - 1 > s->mb_num/40) break;
+ 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);
copy_picture(&s->new_picture, s->reordered_input_picture[0]);
- if(s->reordered_input_picture[0]->type == FF_BUFFER_TYPE_SHARED){
+ 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;
- pic->reference = s->reordered_input_picture[0]->reference;
-
- alloc_picture(s, pic, 0);
-
copy_picture_attributes(s, (AVFrame*)pic, (AVFrame*)s->reordered_input_picture[0]);
s->current_picture_ptr= pic;
s->current_picture_ptr= s->reordered_input_picture[0];
for(i=0; i<4; i++){
- s->new_picture.data[i]+=16;
+ s->new_picture.data[i]+= INPLACE_OFFSET;
}
}
copy_picture(&s->current_picture, s->current_picture_ptr);
AVFrame *pic_arg = data;
int i, stuffing_count;
- if(avctx->pix_fmt != PIX_FMT_YUV420P && avctx->pix_fmt != PIX_FMT_YUVJ420P){
- av_log(avctx, AV_LOG_ERROR, "this codec supports only YUV420P\n");
- return -1;
- }
-
for(i=0; i<avctx->thread_count; i++){
int start_y= s->thread_context[i]->start_mb_y;
int end_y= s->thread_context[i]-> end_mb_y;
//emms_c();
//printf("qs:%f %f %d\n", s->new_picture.quality, s->current_picture.quality, s->qscale);
MPV_frame_start(s, avctx);
-
- encode_picture(s, s->picture_number);
+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;
if (s->out_format == FMT_MJPEG)
mjpeg_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->qscale < s->avctx->qmax){
+ s->next_lambda= s->lambda*(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;
+ }
+// av_log(NULL, AV_LOG_ERROR, "R:%d ", s->next_lambda);
+ for(i=0; i<avctx->thread_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);
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
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;
MPV_motion_lowres(s, dest_y, dest_cb, dest_cr, 1, s->next_picture.data, op_pix);
}
}else{
+ op_qpix= s->me.qpel_put;
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;
+ op_qpix= s->me.qpel_avg;
}
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);
add_dequant_dct(s, block[3], 3, dest_y + dct_offset + block_size, dct_linesize, s->qscale);
if(!(s->flags&CODEC_FLAG_GRAY)){
- add_dequant_dct(s, block[4], 4, dest_cb, uvlinesize, s->chroma_qscale);
- add_dequant_dct(s, block[5], 5, dest_cr, uvlinesize, s->chroma_qscale);
+ if (s->chroma_y_shift){
+ add_dequant_dct(s, block[4], 4, dest_cb, uvlinesize, s->chroma_qscale);
+ add_dequant_dct(s, block[5], 5, dest_cr, uvlinesize, s->chroma_qscale);
+ }else{
+ dct_linesize >>= 1;
+ dct_offset >>=1;
+ add_dequant_dct(s, block[4], 4, dest_cb, dct_linesize, s->chroma_qscale);
+ add_dequant_dct(s, block[5], 5, dest_cr, dct_linesize, s->chroma_qscale);
+ add_dequant_dct(s, block[6], 6, dest_cb + dct_offset, dct_linesize, s->chroma_qscale);
+ add_dequant_dct(s, block[7], 7, dest_cr + dct_offset, dct_linesize, s->chroma_qscale);
+ }
}
} else if(s->codec_id != CODEC_ID_WMV2){
add_dct(s, block[0], 0, dest_y , dct_linesize);
put_dct(s, block[3], 3, dest_y + dct_offset + block_size, dct_linesize, s->qscale);
if(!(s->flags&CODEC_FLAG_GRAY)){
- put_dct(s, block[4], 4, dest_cb, uvlinesize, s->chroma_qscale);
- put_dct(s, block[5], 5, dest_cr, uvlinesize, s->chroma_qscale);
+ if(s->chroma_y_shift){
+ put_dct(s, block[4], 4, dest_cb, uvlinesize, s->chroma_qscale);
+ put_dct(s, block[5], 5, dest_cr, uvlinesize, s->chroma_qscale);
+ }else{
+ dct_offset >>=1;
+ dct_linesize >>=1;
+ put_dct(s, block[4], 4, dest_cb, dct_linesize, s->chroma_qscale);
+ put_dct(s, block[5], 5, dest_cr, dct_linesize, s->chroma_qscale);
+ put_dct(s, block[6], 6, dest_cb + dct_offset, dct_linesize, s->chroma_qscale);
+ put_dct(s, block[7], 7, dest_cr + dct_offset, dct_linesize, s->chroma_qscale);
+ }
}
}else{
s->dsp.idct_put(dest_y , dct_linesize, block[0]);
}
}
-static void encode_mb(MpegEncContext *s, int motion_x, int motion_y)
+static 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[6][64];
- DCTELEM orig[6][64];
+ 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[6];
+ 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; i<6; i++) skip_dct[i]=0;
+ for(i=0; i<mb_block_count; i++) skip_dct[i]=0;
if(s->adaptive_quant){
const int last_qp= s->qscale;
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 * 8 * wrap_c) + mb_x * 8;
- ptr_cr = s->new_picture.data[2] + (mb_y * 8 * wrap_c) + mb_x * 8;
+ 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, 8, mb_x*8, mb_y*8, s->width>>1, s->height>>1);
+ 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, 8, mb_x*8, mb_y*8, s->width>>1, s->height>>1);
+ 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;
}
dct_offset= wrap_y;
wrap_y<<=1;
+ if (s->chroma_format == CHROMA_422)
+ wrap_c<<=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];
dct_offset= wrap_y;
wrap_y<<=1;
+ if (s->chroma_format == CHROMA_422)
+ wrap_c<<=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){
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(!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);
- memcpy(orig[0], s->block[0], sizeof(DCTELEM)*64*6);
+ 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;i<6;i++) {
+ for(i=0;i<mb_block_count;i++) {
if(!skip_dct[i]){
int overflow;
s->block_last_index[i] = s->dct_quantize(s, s->block[i], i, s->qscale, &overflow);
s->block_last_index[i]= -1;
}
if(s->avctx->quantizer_noise_shaping){
- for(i=0;i<6;i++) {
+ for(i=0;i<mb_block_count;i++) {
if(!skip_dct[i]){
s->block_last_index[i] = dct_quantize_refine(s, s->block[i], weight[i], orig[i], i, s->qscale);
}
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; i<6; i++)
+ for(i=4; i<mb_block_count; i++)
dct_single_coeff_elimination(s, i, s->chroma_elim_threshold);
if(s->flags & CODEC_FLAG_CBP_RD){
- for(i=0;i<6;i++) {
+ for(i=0;i<mb_block_count;i++) {
if(s->block_last_index[i] == -1)
s->coded_score[i]= INT_MAX/256;
}
//non c quantize code returns incorrect block_last_index FIXME
if(s->alternate_scan && s->dct_quantize != dct_quantize_c){
- for(i=0; i<6; i++){
+ for(i=0; i<mb_block_count; i++){
int j;
if(s->block_last_index[i]>0){
for(j=63; j>0; j--){
}
}
+static 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){
d->tex_pb= s->tex_pb;
}
d->block= s->block;
- for(i=0; i<6; i++)
+ 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;
flush_put_bits(&dst->pb);
}
-static void estimate_qp(MpegEncContext *s, int dry_run){
- if (!s->fixed_qscale)
+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){
s->lambda= s->current_picture.quality;
//printf("%d %d\n", s->avctx->global_quality, s->current_picture.quality);
update_qscale(s);
+ return 0;
}
-static void encode_picture(MpegEncContext *s, int picture_number)
+static int encode_picture(MpegEncContext *s, int picture_number)
{
int i;
int bits;
}
if(s->flags & CODEC_FLAG_PASS2){
- estimate_qp(s, 1);
+ 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)
}
}
- estimate_qp(s, 0);
+ 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
for(i=1;i<64;i++){
int j= s->dsp.idct_permutation[i];
- s->intra_matrix[j] = clip_uint8((ff_mpeg1_default_intra_matrix[i] * s->qscale) >> 3) & 0xFF;
+ s->intra_matrix[j] = 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);
merge_context_after_encode(s, s->thread_context[i]);
}
emms_c();
+ return 0;
}
-#endif //CONFIG_ENCODERS
-
static void denoise_dct_c(MpegEncContext *s, DCTELEM *block){
const int intra= s->mb_intra;
int i;
}
}
-#ifdef CONFIG_ENCODERS
-
static int dct_quantize_trellis_c(MpegEncContext *s,
DCTELEM *block, int n,
int qscale, int *overflow){
DCTELEM *block, int16_t *weight, DCTELEM *orig,
int n, int qscale){
int16_t rem[64];
- DCTELEM d1[64] __align16;
+ DECLARE_ALIGNED_16(DCTELEM, d1[64]);
const int *qmat;
const uint8_t *scantable= s->intra_scantable.scantable;
const uint8_t *perm_scantable= s->intra_scantable.permutated;
}
}
+static void dct_unquantize_mpeg2_intra_bitexact(MpegEncContext *s,
+ DCTELEM *block, int n, int qscale)
+{
+ int i, level, nCoeffs;
+ const uint16_t *quant_matrix;
+ int sum=-1;
+
+ if(s->alternate_scan) nCoeffs= 63;
+ else nCoeffs= s->block_last_index[n];
+
+ if (n < 4)
+ block[0] = block[0] * s->y_dc_scale;
+ else
+ block[0] = block[0] * s->c_dc_scale;
+ quant_matrix = s->intra_matrix;
+ for(i=1;i<=nCoeffs;i++) {
+ int j= s->intra_scantable.permutated[i];
+ level = block[j];
+ if (level) {
+ if (level < 0) {
+ level = -level;
+ level = (int)(level * qscale * quant_matrix[j]) >> 3;
+ level = -level;
+ } else {
+ level = (int)(level * qscale * quant_matrix[j]) >> 3;
+ }
+ block[j] = level;
+ sum+=level;
+ }
+ }
+ block[63]^=sum&1;
+}
+
static void dct_unquantize_mpeg2_inter_c(MpegEncContext *s,
DCTELEM *block, int n, int qscale)
{
MPV_encode_init,
MPV_encode_picture,
MPV_encode_end,
- .pix_fmts= (enum PixelFormat[]){PIX_FMT_YUVJ420P, -1},
+ .pix_fmts= (enum PixelFormat[]){PIX_FMT_YUVJ420P, PIX_FMT_YUVJ422P, -1},
};
#endif //CONFIG_ENCODERS