Merge remote-tracking branch 'qatar/master'
authorMichael Niedermayer <michaelni@gmx.at>
Fri, 25 Nov 2011 00:38:21 +0000 (01:38 +0100)
committerMichael Niedermayer <michaelni@gmx.at>
Fri, 25 Nov 2011 03:34:44 +0000 (04:34 +0100)
* qatar/master:
  snow: split snow in snowdec and snowenc
  tiffenc: deprecate using compression_level
  swscale: fix failing fate tests.
  swscale: add support for planar RGB input.
  h264: add support for decoding planar RGB images.
  Clean up swscale pixfmt macros using av_pix_fmt_descriptors[].
  pixfmt: add planar RGB formats.

Conflicts:
libavcodec/h264.c
libavcodec/snow.c
libavcodec/utils.c
libavutil/avutil.h
libavutil/pixdesc.c
libavutil/pixfmt.h
libswscale/swscale.c
libswscale/swscale_internal.h
libswscale/swscale_unscaled.c
libswscale/utils.c

Merged-by: Michael Niedermayer <michaelni@gmx.at>
17 files changed:
1  2 
libavcodec/Makefile
libavcodec/h264.c
libavcodec/snow.c
libavcodec/snow.h
libavcodec/snowdec.c
libavcodec/snowenc.c
libavcodec/tiffenc.c
libavcodec/utils.c
libavcodec/version.h
libavutil/avutil.h
libavutil/pixdesc.c
libavutil/pixdesc.h
libavutil/pixfmt.h
libswscale/swscale.c
libswscale/swscale_internal.h
libswscale/swscale_unscaled.c
libswscale/utils.c

@@@ -365,19 -341,15 +365,19 @@@ OBJS-$(CONFIG_SIPR_DECODER)            
  OBJS-$(CONFIG_SMACKAUD_DECODER)        += smacker.o
  OBJS-$(CONFIG_SMACKER_DECODER)         += smacker.o
  OBJS-$(CONFIG_SMC_DECODER)             += smc.o
- OBJS-$(CONFIG_SNOW_DECODER)            += snow.o rangecoder.o
- OBJS-$(CONFIG_SNOW_ENCODER)            += snow.o rangecoder.o motion_est.o \
-                                           ratecontrol.o h263.o             \
-                                           mpegvideo.o error_resilience.o   \
-                                           ituh263enc.o mpegvideo_enc.o     \
-                                           mpeg12data.o
+ OBJS-$(CONFIG_SNOW_DECODER)            += snowdec.o snow.o rangecoder.o
+ OBJS-$(CONFIG_SNOW_ENCODER)            += snowenc.o snow.o rangecoder.o    \
+                                           motion_est.o ratecontrol.o       \
+                                           h263.o mpegvideo.o               \
+                                           error_resilience.o ituh263enc.o  \
+                                           mpegvideo_enc.o mpeg12data.o
  OBJS-$(CONFIG_SOL_DPCM_DECODER)        += dpcm.o
 +OBJS-$(CONFIG_SONIC_DECODER)           += sonic.o
 +OBJS-$(CONFIG_SONIC_ENCODER)           += sonic.o
 +OBJS-$(CONFIG_SONIC_LS_ENCODER)        += sonic.o
  OBJS-$(CONFIG_SP5X_DECODER)            += sp5xdec.o mjpegdec.o mjpeg.o
  OBJS-$(CONFIG_SRT_DECODER)             += srtdec.o ass.o
 +OBJS-$(CONFIG_SRT_ENCODER)             += srtenc.o ass_split.o
  OBJS-$(CONFIG_SUNRAST_DECODER)         += sunrast.o
  OBJS-$(CONFIG_SVQ1_DECODER)            += svq1dec.o svq1.o h263.o \
                                            mpegvideo.o error_resilience.o
@@@ -716,7 -677,9 +716,7 @@@ SKIPHEADERS-$(CONFIG_VDPAU)            
  SKIPHEADERS-$(CONFIG_XVMC)             += xvmc.h
  SKIPHEADERS-$(HAVE_W32THREADS)         += w32pthreads.h
  
- TESTPROGS = cabac dct fft fft-fixed h264 iirfilter rangecoder snow
 -EXAMPLES = api
 -
 -TESTPROGS = cabac dct fft fft-fixed h264 iirfilter rangecoder
++TESTPROGS = cabac dct fft fft-fixed h264 iirfilter rangecoder snowenc
  TESTPROGS-$(HAVE_MMX) += motion
  TESTOBJS = dctref.o
  
Simple merge
Simple merge
Simple merge
index 0000000,2b6f6e1..9fd25a5
mode 000000,100644..100644
--- /dev/null
@@@ -1,0 -1,546 +1,546 @@@
 - * This file is part of Libav.
+ /*
+  * Copyright (C) 2004 Michael Niedermayer <michaelni@gmx.at>
+  *
 - * Libav is free software; you can redistribute it and/or
++ * This file is part of FFmpeg.
+  *
 - * Libav is distributed in the hope that it will be useful,
++ * 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.
+  *
 - * License along with Libav; if not, write to the Free Software
++ * 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
+  */
+ #include "libavutil/intmath.h"
+ #include "libavutil/log.h"
+ #include "libavutil/opt.h"
+ #include "avcodec.h"
+ #include "dsputil.h"
+ #include "dwt.h"
+ #include "snow.h"
+ #include "rangecoder.h"
+ #include "mathops.h"
+ #include "mpegvideo.h"
+ #include "h263.h"
+ #undef NDEBUG
+ #include <assert.h>
+ static av_always_inline void predict_slice_buffered(SnowContext *s, slice_buffer * sb, IDWTELEM * old_buffer, int plane_index, int add, int mb_y){
+     Plane *p= &s->plane[plane_index];
+     const int mb_w= s->b_width  << s->block_max_depth;
+     const int mb_h= s->b_height << s->block_max_depth;
+     int x, y, mb_x;
+     int block_size = MB_SIZE >> s->block_max_depth;
+     int block_w    = plane_index ? block_size/2 : block_size;
+     const uint8_t *obmc  = plane_index ? obmc_tab[s->block_max_depth+1] : obmc_tab[s->block_max_depth];
+     int obmc_stride= plane_index ? block_size : 2*block_size;
+     int ref_stride= s->current_picture.linesize[plane_index];
+     uint8_t *dst8= s->current_picture.data[plane_index];
+     int w= p->width;
+     int h= p->height;
+     if(s->keyframe || (s->avctx->debug&512)){
+         if(mb_y==mb_h)
+             return;
+         if(add){
+             for(y=block_w*mb_y; y<FFMIN(h,block_w*(mb_y+1)); y++){
+ //                DWTELEM * line = slice_buffer_get_line(sb, y);
+                 IDWTELEM * line = sb->line[y];
+                 for(x=0; x<w; x++){
+ //                    int v= buf[x + y*w] + (128<<FRAC_BITS) + (1<<(FRAC_BITS-1));
+                     int v= line[x] + (128<<FRAC_BITS) + (1<<(FRAC_BITS-1));
+                     v >>= FRAC_BITS;
+                     if(v&(~255)) v= ~(v>>31);
+                     dst8[x + y*ref_stride]= v;
+                 }
+             }
+         }else{
+             for(y=block_w*mb_y; y<FFMIN(h,block_w*(mb_y+1)); y++){
+ //                DWTELEM * line = slice_buffer_get_line(sb, y);
+                 IDWTELEM * line = sb->line[y];
+                 for(x=0; x<w; x++){
+                     line[x] -= 128 << FRAC_BITS;
+ //                    buf[x + y*w]-= 128<<FRAC_BITS;
+                 }
+             }
+         }
+         return;
+     }
+     for(mb_x=0; mb_x<=mb_w; mb_x++){
+         add_yblock(s, 1, sb, old_buffer, dst8, obmc,
+                    block_w*mb_x - block_w/2,
+                    block_w*mb_y - block_w/2,
+                    block_w, block_w,
+                    w, h,
+                    w, ref_stride, obmc_stride,
+                    mb_x - 1, mb_y - 1,
+                    add, 0, plane_index);
+     }
+ }
+ static inline void decode_subband_slice_buffered(SnowContext *s, SubBand *b, slice_buffer * sb, int start_y, int h, int save_state[1]){
+     const int w= b->width;
+     int y;
+     const int qlog= av_clip(s->qlog + b->qlog, 0, QROOT*16);
+     int qmul= qexp[qlog&(QROOT-1)]<<(qlog>>QSHIFT);
+     int qadd= (s->qbias*qmul)>>QBIAS_SHIFT;
+     int new_index = 0;
+     if(b->ibuf == s->spatial_idwt_buffer || s->qlog == LOSSLESS_QLOG){
+         qadd= 0;
+         qmul= 1<<QEXPSHIFT;
+     }
+     /* If we are on the second or later slice, restore our index. */
+     if (start_y != 0)
+         new_index = save_state[0];
+     for(y=start_y; y<h; y++){
+         int x = 0;
+         int v;
+         IDWTELEM * line = slice_buffer_get_line(sb, y * b->stride_line + b->buf_y_offset) + b->buf_x_offset;
+         memset(line, 0, b->width*sizeof(IDWTELEM));
+         v = b->x_coeff[new_index].coeff;
+         x = b->x_coeff[new_index++].x;
+         while(x < w){
+             register int t= ( (v>>1)*qmul + qadd)>>QEXPSHIFT;
+             register int u= -(v&1);
+             line[x] = (t^u) - u;
+             v = b->x_coeff[new_index].coeff;
+             x = b->x_coeff[new_index++].x;
+         }
+     }
+     /* Save our variables for the next slice. */
+     save_state[0] = new_index;
+     return;
+ }
+ static void decode_q_branch(SnowContext *s, int level, int x, int y){
+     const int w= s->b_width << s->block_max_depth;
+     const int rem_depth= s->block_max_depth - level;
+     const int index= (x + y*w) << rem_depth;
+     int trx= (x+1)<<rem_depth;
+     const BlockNode *left  = x ? &s->block[index-1] : &null_block;
+     const BlockNode *top   = y ? &s->block[index-w] : &null_block;
+     const BlockNode *tl    = y && x ? &s->block[index-w-1] : left;
+     const BlockNode *tr    = y && trx<w && ((x&1)==0 || level==0) ? &s->block[index-w+(1<<rem_depth)] : tl; //FIXME use lt
+     int s_context= 2*left->level + 2*top->level + tl->level + tr->level;
+     if(s->keyframe){
+         set_blocks(s, level, x, y, null_block.color[0], null_block.color[1], null_block.color[2], null_block.mx, null_block.my, null_block.ref, BLOCK_INTRA);
+         return;
+     }
+     if(level==s->block_max_depth || get_rac(&s->c, &s->block_state[4 + s_context])){
+         int type, mx, my;
+         int l = left->color[0];
+         int cb= left->color[1];
+         int cr= left->color[2];
+         int ref = 0;
+         int ref_context= av_log2(2*left->ref) + av_log2(2*top->ref);
+         int mx_context= av_log2(2*FFABS(left->mx - top->mx)) + 0*av_log2(2*FFABS(tr->mx - top->mx));
+         int my_context= av_log2(2*FFABS(left->my - top->my)) + 0*av_log2(2*FFABS(tr->my - top->my));
+         type= get_rac(&s->c, &s->block_state[1 + left->type + top->type]) ? BLOCK_INTRA : 0;
+         if(type){
+             pred_mv(s, &mx, &my, 0, left, top, tr);
+             l += get_symbol(&s->c, &s->block_state[32], 1);
+             cb+= get_symbol(&s->c, &s->block_state[64], 1);
+             cr+= get_symbol(&s->c, &s->block_state[96], 1);
+         }else{
+             if(s->ref_frames > 1)
+                 ref= get_symbol(&s->c, &s->block_state[128 + 1024 + 32*ref_context], 0);
+             pred_mv(s, &mx, &my, ref, left, top, tr);
+             mx+= get_symbol(&s->c, &s->block_state[128 + 32*(mx_context + 16*!!ref)], 1);
+             my+= get_symbol(&s->c, &s->block_state[128 + 32*(my_context + 16*!!ref)], 1);
+         }
+         set_blocks(s, level, x, y, l, cb, cr, mx, my, ref, type);
+     }else{
+         decode_q_branch(s, level+1, 2*x+0, 2*y+0);
+         decode_q_branch(s, level+1, 2*x+1, 2*y+0);
+         decode_q_branch(s, level+1, 2*x+0, 2*y+1);
+         decode_q_branch(s, level+1, 2*x+1, 2*y+1);
+     }
+ }
+ static void dequantize_slice_buffered(SnowContext *s, slice_buffer * sb, SubBand *b, IDWTELEM *src, int stride, int start_y, int end_y){
+     const int w= b->width;
+     const int qlog= av_clip(s->qlog + b->qlog, 0, QROOT*16);
+     const int qmul= qexp[qlog&(QROOT-1)]<<(qlog>>QSHIFT);
+     const int qadd= (s->qbias*qmul)>>QBIAS_SHIFT;
+     int x,y;
+     if(s->qlog == LOSSLESS_QLOG) return;
+     for(y=start_y; y<end_y; y++){
+ //        DWTELEM * line = slice_buffer_get_line_from_address(sb, src + (y * stride));
+         IDWTELEM * line = slice_buffer_get_line(sb, (y * b->stride_line) + b->buf_y_offset) + b->buf_x_offset;
+         for(x=0; x<w; x++){
+             int i= line[x];
+             if(i<0){
+                 line[x]= -((-i*qmul + qadd)>>(QEXPSHIFT)); //FIXME try different bias
+             }else if(i>0){
+                 line[x]=  (( i*qmul + qadd)>>(QEXPSHIFT));
+             }
+         }
+     }
+ }
+ static void correlate_slice_buffered(SnowContext *s, slice_buffer * sb, SubBand *b, IDWTELEM *src, int stride, int inverse, int use_median, int start_y, int end_y){
+     const int w= b->width;
+     int x,y;
+     IDWTELEM * line=0; // silence silly "could be used without having been initialized" warning
+     IDWTELEM * prev;
+     if (start_y != 0)
+         line = slice_buffer_get_line(sb, ((start_y - 1) * b->stride_line) + b->buf_y_offset) + b->buf_x_offset;
+     for(y=start_y; y<end_y; y++){
+         prev = line;
+ //        line = slice_buffer_get_line_from_address(sb, src + (y * stride));
+         line = slice_buffer_get_line(sb, (y * b->stride_line) + b->buf_y_offset) + b->buf_x_offset;
+         for(x=0; x<w; x++){
+             if(x){
+                 if(use_median){
+                     if(y && x+1<w) line[x] += mid_pred(line[x - 1], prev[x], prev[x + 1]);
+                     else  line[x] += line[x - 1];
+                 }else{
+                     if(y) line[x] += mid_pred(line[x - 1], prev[x], line[x - 1] + prev[x] - prev[x - 1]);
+                     else  line[x] += line[x - 1];
+                 }
+             }else{
+                 if(y) line[x] += prev[x];
+             }
+         }
+     }
+ }
+ static void decode_qlogs(SnowContext *s){
+     int plane_index, level, orientation;
+     for(plane_index=0; plane_index<3; plane_index++){
+         for(level=0; level<s->spatial_decomposition_count; level++){
+             for(orientation=level ? 1:0; orientation<4; orientation++){
+                 int q;
+                 if     (plane_index==2) q= s->plane[1].band[level][orientation].qlog;
+                 else if(orientation==2) q= s->plane[plane_index].band[level][1].qlog;
+                 else                    q= get_symbol(&s->c, s->header_state, 1);
+                 s->plane[plane_index].band[level][orientation].qlog= q;
+             }
+         }
+     }
+ }
+ #define GET_S(dst, check) \
+     tmp= get_symbol(&s->c, s->header_state, 0);\
+     if(!(check)){\
+         av_log(s->avctx, AV_LOG_ERROR, "Error " #dst " is %d\n", tmp);\
+         return -1;\
+     }\
+     dst= tmp;
+ static int decode_header(SnowContext *s){
+     int plane_index, tmp;
+     uint8_t kstate[32];
+     memset(kstate, MID_STATE, sizeof(kstate));
+     s->keyframe= get_rac(&s->c, kstate);
+     if(s->keyframe || s->always_reset){
+         ff_snow_reset_contexts(s);
+         s->spatial_decomposition_type=
+         s->qlog=
+         s->qbias=
+         s->mv_scale=
+         s->block_max_depth= 0;
+     }
+     if(s->keyframe){
+         GET_S(s->version, tmp <= 0U)
+         s->always_reset= get_rac(&s->c, s->header_state);
+         s->temporal_decomposition_type= get_symbol(&s->c, s->header_state, 0);
+         s->temporal_decomposition_count= get_symbol(&s->c, s->header_state, 0);
+         GET_S(s->spatial_decomposition_count, 0 < tmp && tmp <= MAX_DECOMPOSITIONS)
+         s->colorspace_type= get_symbol(&s->c, s->header_state, 0);
+         s->chroma_h_shift= get_symbol(&s->c, s->header_state, 0);
+         s->chroma_v_shift= get_symbol(&s->c, s->header_state, 0);
+         s->spatial_scalability= get_rac(&s->c, s->header_state);
+ //        s->rate_scalability= get_rac(&s->c, s->header_state);
+         GET_S(s->max_ref_frames, tmp < (unsigned)MAX_REF_FRAMES)
+         s->max_ref_frames++;
+         decode_qlogs(s);
+     }
+     if(!s->keyframe){
+         if(get_rac(&s->c, s->header_state)){
+             for(plane_index=0; plane_index<2; plane_index++){
+                 int htaps, i, sum=0;
+                 Plane *p= &s->plane[plane_index];
+                 p->diag_mc= get_rac(&s->c, s->header_state);
+                 htaps= get_symbol(&s->c, s->header_state, 0)*2 + 2;
+                 if((unsigned)htaps > HTAPS_MAX || htaps==0)
+                     return -1;
+                 p->htaps= htaps;
+                 for(i= htaps/2; i; i--){
+                     p->hcoeff[i]= get_symbol(&s->c, s->header_state, 0) * (1-2*(i&1));
+                     sum += p->hcoeff[i];
+                 }
+                 p->hcoeff[0]= 32-sum;
+             }
+             s->plane[2].diag_mc= s->plane[1].diag_mc;
+             s->plane[2].htaps  = s->plane[1].htaps;
+             memcpy(s->plane[2].hcoeff, s->plane[1].hcoeff, sizeof(s->plane[1].hcoeff));
+         }
+         if(get_rac(&s->c, s->header_state)){
+             GET_S(s->spatial_decomposition_count, 0 < tmp && tmp <= MAX_DECOMPOSITIONS)
+             decode_qlogs(s);
+         }
+     }
+     s->spatial_decomposition_type+= get_symbol(&s->c, s->header_state, 1);
+     if(s->spatial_decomposition_type > 1U){
+         av_log(s->avctx, AV_LOG_ERROR, "spatial_decomposition_type %d not supported", s->spatial_decomposition_type);
+         return -1;
+     }
+     if(FFMIN(s->avctx-> width>>s->chroma_h_shift,
+              s->avctx->height>>s->chroma_v_shift) >> (s->spatial_decomposition_count-1) <= 0){
+         av_log(s->avctx, AV_LOG_ERROR, "spatial_decomposition_count %d too large for size", s->spatial_decomposition_count);
+         return -1;
+     }
+     s->qlog           += get_symbol(&s->c, s->header_state, 1);
+     s->mv_scale       += get_symbol(&s->c, s->header_state, 1);
+     s->qbias          += get_symbol(&s->c, s->header_state, 1);
+     s->block_max_depth+= get_symbol(&s->c, s->header_state, 1);
+     if(s->block_max_depth > 1 || s->block_max_depth < 0){
+         av_log(s->avctx, AV_LOG_ERROR, "block_max_depth= %d is too large", s->block_max_depth);
+         s->block_max_depth= 0;
+         return -1;
+     }
+     return 0;
+ }
+ static av_cold int decode_init(AVCodecContext *avctx)
+ {
+     avctx->pix_fmt= PIX_FMT_YUV420P;
+     ff_snow_common_init(avctx);
+     return 0;
+ }
+ static void decode_blocks(SnowContext *s){
+     int x, y;
+     int w= s->b_width;
+     int h= s->b_height;
+     for(y=0; y<h; y++){
+         for(x=0; x<w; x++){
+             decode_q_branch(s, 0, x, y);
+         }
+     }
+ }
+ static int decode_frame(AVCodecContext *avctx, void *data, int *data_size, AVPacket *avpkt){
+     const uint8_t *buf = avpkt->data;
+     int buf_size = avpkt->size;
+     SnowContext *s = avctx->priv_data;
+     RangeCoder * const c= &s->c;
+     int bytes_read;
+     AVFrame *picture = data;
+     int level, orientation, plane_index;
+     ff_init_range_decoder(c, buf, buf_size);
+     ff_build_rac_states(c, 0.05*(1LL<<32), 256-8);
+     s->current_picture.pict_type= AV_PICTURE_TYPE_I; //FIXME I vs. P
+     if(decode_header(s)<0)
+         return -1;
+     ff_snow_common_init_after_header(avctx);
+     // realloc slice buffer for the case that spatial_decomposition_count changed
+     ff_slice_buffer_destroy(&s->sb);
+     ff_slice_buffer_init(&s->sb, s->plane[0].height, (MB_SIZE >> s->block_max_depth) + s->spatial_decomposition_count * 8 + 1, s->plane[0].width, s->spatial_idwt_buffer);
+     for(plane_index=0; plane_index<3; plane_index++){
+         Plane *p= &s->plane[plane_index];
+         p->fast_mc= p->diag_mc && p->htaps==6 && p->hcoeff[0]==40
+                                               && p->hcoeff[1]==-10
+                                               && p->hcoeff[2]==2;
+     }
+     ff_snow_alloc_blocks(s);
+     if(ff_snow_frame_start(s) < 0)
+         return -1;
+     //keyframe flag duplication mess FIXME
+     if(avctx->debug&FF_DEBUG_PICT_INFO)
+         av_log(avctx, AV_LOG_ERROR, "keyframe:%d qlog:%d\n", s->keyframe, s->qlog);
+     decode_blocks(s);
+     for(plane_index=0; plane_index<3; plane_index++){
+         Plane *p= &s->plane[plane_index];
+         int w= p->width;
+         int h= p->height;
+         int x, y;
+         int decode_state[MAX_DECOMPOSITIONS][4][1]; /* Stored state info for unpack_coeffs. 1 variable per instance. */
+         if(s->avctx->debug&2048){
+             memset(s->spatial_dwt_buffer, 0, sizeof(DWTELEM)*w*h);
+             predict_plane(s, s->spatial_idwt_buffer, plane_index, 1);
+             for(y=0; y<h; y++){
+                 for(x=0; x<w; x++){
+                     int v= s->current_picture.data[plane_index][y*s->current_picture.linesize[plane_index] + x];
+                     s->mconly_picture.data[plane_index][y*s->mconly_picture.linesize[plane_index] + x]= v;
+                 }
+             }
+         }
+         {
+         for(level=0; level<s->spatial_decomposition_count; level++){
+             for(orientation=level ? 1 : 0; orientation<4; orientation++){
+                 SubBand *b= &p->band[level][orientation];
+                 unpack_coeffs(s, b, b->parent, orientation);
+             }
+         }
+         }
+         {
+         const int mb_h= s->b_height << s->block_max_depth;
+         const int block_size = MB_SIZE >> s->block_max_depth;
+         const int block_w    = plane_index ? block_size/2 : block_size;
+         int mb_y;
+         DWTCompose cs[MAX_DECOMPOSITIONS];
+         int yd=0, yq=0;
+         int y;
+         int end_y;
+         ff_spatial_idwt_buffered_init(cs, &s->sb, w, h, 1, s->spatial_decomposition_type, s->spatial_decomposition_count);
+         for(mb_y=0; mb_y<=mb_h; mb_y++){
+             int slice_starty = block_w*mb_y;
+             int slice_h = block_w*(mb_y+1);
+             if (!(s->keyframe || s->avctx->debug&512)){
+                 slice_starty = FFMAX(0, slice_starty - (block_w >> 1));
+                 slice_h -= (block_w >> 1);
+             }
+             for(level=0; level<s->spatial_decomposition_count; level++){
+                 for(orientation=level ? 1 : 0; orientation<4; orientation++){
+                     SubBand *b= &p->band[level][orientation];
+                     int start_y;
+                     int end_y;
+                     int our_mb_start = mb_y;
+                     int our_mb_end = (mb_y + 1);
+                     const int extra= 3;
+                     start_y = (mb_y ? ((block_w * our_mb_start) >> (s->spatial_decomposition_count - level)) + s->spatial_decomposition_count - level + extra: 0);
+                     end_y = (((block_w * our_mb_end) >> (s->spatial_decomposition_count - level)) + s->spatial_decomposition_count - level + extra);
+                     if (!(s->keyframe || s->avctx->debug&512)){
+                         start_y = FFMAX(0, start_y - (block_w >> (1+s->spatial_decomposition_count - level)));
+                         end_y = FFMAX(0, end_y - (block_w >> (1+s->spatial_decomposition_count - level)));
+                     }
+                     start_y = FFMIN(b->height, start_y);
+                     end_y = FFMIN(b->height, end_y);
+                     if (start_y != end_y){
+                         if (orientation == 0){
+                             SubBand * correlate_band = &p->band[0][0];
+                             int correlate_end_y = FFMIN(b->height, end_y + 1);
+                             int correlate_start_y = FFMIN(b->height, (start_y ? start_y + 1 : 0));
+                             decode_subband_slice_buffered(s, correlate_band, &s->sb, correlate_start_y, correlate_end_y, decode_state[0][0]);
+                             correlate_slice_buffered(s, &s->sb, correlate_band, correlate_band->ibuf, correlate_band->stride, 1, 0, correlate_start_y, correlate_end_y);
+                             dequantize_slice_buffered(s, &s->sb, correlate_band, correlate_band->ibuf, correlate_band->stride, start_y, end_y);
+                         }
+                         else
+                             decode_subband_slice_buffered(s, b, &s->sb, start_y, end_y, decode_state[level][orientation]);
+                     }
+                 }
+             }
+             for(; yd<slice_h; yd+=4){
+                 ff_spatial_idwt_buffered_slice(&s->dwt, cs, &s->sb, w, h, 1, s->spatial_decomposition_type, s->spatial_decomposition_count, yd);
+             }
+             if(s->qlog == LOSSLESS_QLOG){
+                 for(; yq<slice_h && yq<h; yq++){
+                     IDWTELEM * line = slice_buffer_get_line(&s->sb, yq);
+                     for(x=0; x<w; x++){
+                         line[x] <<= FRAC_BITS;
+                     }
+                 }
+             }
+             predict_slice_buffered(s, &s->sb, s->spatial_idwt_buffer, plane_index, 1, mb_y);
+             y = FFMIN(p->height, slice_starty);
+             end_y = FFMIN(p->height, slice_h);
+             while(y < end_y)
+                 ff_slice_buffer_release(&s->sb, y++);
+         }
+         ff_slice_buffer_flush(&s->sb);
+         }
+     }
+     emms_c();
+     ff_snow_release_buffer(avctx);
+     if(!(s->avctx->debug&2048))
+         *picture= s->current_picture;
+     else
+         *picture= s->mconly_picture;
+     *data_size = sizeof(AVFrame);
+     bytes_read= c->bytestream - c->bytestream_start;
+     if(bytes_read ==0) av_log(s->avctx, AV_LOG_ERROR, "error at end of frame\n"); //FIXME
+     return bytes_read;
+ }
+ static av_cold int decode_end(AVCodecContext *avctx)
+ {
+     SnowContext *s = avctx->priv_data;
+     ff_slice_buffer_destroy(&s->sb);
+     ff_snow_common_end(s);
+     return 0;
+ }
+ AVCodec ff_snow_decoder = {
+     .name           = "snow",
+     .type           = AVMEDIA_TYPE_VIDEO,
+     .id             = CODEC_ID_SNOW,
+     .priv_data_size = sizeof(SnowContext),
+     .init           = decode_init,
+     .close          = decode_end,
+     .decode         = decode_frame,
+     .capabilities   = CODEC_CAP_DR1 /*| CODEC_CAP_DRAW_HORIZ_BAND*/,
+     .long_name = NULL_IF_CONFIG_SMALL("Snow"),
+ };
index 0000000,4926eef..6f00941
mode 000000,100644..100644
--- /dev/null
@@@ -1,0 -1,1915 +1,2036 @@@
 - * This file is part of Libav.
+ /*
+  * Copyright (C) 2004 Michael Niedermayer <michaelni@gmx.at>
+  *
 - * Libav is free software; you can redistribute it and/or
++ * This file is part of FFmpeg.
+  *
 - * Libav is distributed in the hope that it will be useful,
++ * 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.
+  *
 - * License along with Libav; if not, write to the Free Software
++ * 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
+  */
+ #include "libavutil/intmath.h"
+ #include "libavutil/log.h"
+ #include "libavutil/opt.h"
+ #include "avcodec.h"
+ #include "dsputil.h"
+ #include "dwt.h"
+ #include "snow.h"
+ #include "rangecoder.h"
+ #include "mathops.h"
+ #include "mpegvideo.h"
+ #include "h263.h"
+ #undef NDEBUG
+ #include <assert.h>
+ #define QUANTIZE2 0
+ #if QUANTIZE2==1
+ #define Q2_STEP 8
+ static void find_sse(SnowContext *s, Plane *p, int *score, int score_stride, IDWTELEM *r0, IDWTELEM *r1, int level, int orientation){
+     SubBand *b= &p->band[level][orientation];
+     int x, y;
+     int xo=0;
+     int yo=0;
+     int step= 1 << (s->spatial_decomposition_count - level);
+     if(orientation&1)
+         xo= step>>1;
+     if(orientation&2)
+         yo= step>>1;
+     //FIXME bias for nonzero ?
+     //FIXME optimize
+     memset(score, 0, sizeof(*score)*score_stride*((p->height + Q2_STEP-1)/Q2_STEP));
+     for(y=0; y<p->height; y++){
+         for(x=0; x<p->width; x++){
+             int sx= (x-xo + step/2) / step / Q2_STEP;
+             int sy= (y-yo + step/2) / step / Q2_STEP;
+             int v= r0[x + y*p->width] - r1[x + y*p->width];
+             assert(sx>=0 && sy>=0 && sx < score_stride);
+             v= ((v+8)>>4)<<4;
+             score[sx + sy*score_stride] += v*v;
+             assert(score[sx + sy*score_stride] >= 0);
+         }
+     }
+ }
+ static void dequantize_all(SnowContext *s, Plane *p, IDWTELEM *buffer, int width, int height){
+     int level, orientation;
+     for(level=0; level<s->spatial_decomposition_count; level++){
+         for(orientation=level ? 1 : 0; orientation<4; orientation++){
+             SubBand *b= &p->band[level][orientation];
+             IDWTELEM *dst= buffer + (b->ibuf - s->spatial_idwt_buffer);
+             dequantize(s, b, dst, b->stride);
+         }
+     }
+ }
+ static void dwt_quantize(SnowContext *s, Plane *p, DWTELEM *buffer, int width, int height, int stride, int type){
+     int level, orientation, ys, xs, x, y, pass;
+     IDWTELEM best_dequant[height * stride];
+     IDWTELEM idwt2_buffer[height * stride];
+     const int score_stride= (width + 10)/Q2_STEP;
+     int best_score[(width + 10)/Q2_STEP * (height + 10)/Q2_STEP]; //FIXME size
+     int score[(width + 10)/Q2_STEP * (height + 10)/Q2_STEP]; //FIXME size
+     int threshold= (s->m.lambda * s->m.lambda) >> 6;
+     //FIXME pass the copy cleanly ?
+ //    memcpy(dwt_buffer, buffer, height * stride * sizeof(DWTELEM));
+     ff_spatial_dwt(buffer, width, height, stride, type, s->spatial_decomposition_count);
+     for(level=0; level<s->spatial_decomposition_count; level++){
+         for(orientation=level ? 1 : 0; orientation<4; orientation++){
+             SubBand *b= &p->band[level][orientation];
+             IDWTELEM *dst= best_dequant + (b->ibuf - s->spatial_idwt_buffer);
+              DWTELEM *src=       buffer + (b-> buf - s->spatial_dwt_buffer);
+             assert(src == b->buf); // code does not depend on this but it is true currently
+             quantize(s, b, dst, src, b->stride, s->qbias);
+         }
+     }
+     for(pass=0; pass<1; pass++){
+         if(s->qbias == 0) //keyframe
+             continue;
+         for(level=0; level<s->spatial_decomposition_count; level++){
+             for(orientation=level ? 1 : 0; orientation<4; orientation++){
+                 SubBand *b= &p->band[level][orientation];
+                 IDWTELEM *dst= idwt2_buffer + (b->ibuf - s->spatial_idwt_buffer);
+                 IDWTELEM *best_dst= best_dequant + (b->ibuf - s->spatial_idwt_buffer);
+                 for(ys= 0; ys<Q2_STEP; ys++){
+                     for(xs= 0; xs<Q2_STEP; xs++){
+                         memcpy(idwt2_buffer, best_dequant, height * stride * sizeof(IDWTELEM));
+                         dequantize_all(s, p, idwt2_buffer, width, height);
+                         ff_spatial_idwt(idwt2_buffer, width, height, stride, type, s->spatial_decomposition_count);
+                         find_sse(s, p, best_score, score_stride, idwt2_buffer, s->spatial_idwt_buffer, level, orientation);
+                         memcpy(idwt2_buffer, best_dequant, height * stride * sizeof(IDWTELEM));
+                         for(y=ys; y<b->height; y+= Q2_STEP){
+                             for(x=xs; x<b->width; x+= Q2_STEP){
+                                 if(dst[x + y*b->stride]<0) dst[x + y*b->stride]++;
+                                 if(dst[x + y*b->stride]>0) dst[x + y*b->stride]--;
+                                 //FIXME try more than just --
+                             }
+                         }
+                         dequantize_all(s, p, idwt2_buffer, width, height);
+                         ff_spatial_idwt(idwt2_buffer, width, height, stride, type, s->spatial_decomposition_count);
+                         find_sse(s, p, score, score_stride, idwt2_buffer, s->spatial_idwt_buffer, level, orientation);
+                         for(y=ys; y<b->height; y+= Q2_STEP){
+                             for(x=xs; x<b->width; x+= Q2_STEP){
+                                 int score_idx= x/Q2_STEP + (y/Q2_STEP)*score_stride;
+                                 if(score[score_idx] <= best_score[score_idx] + threshold){
+                                     best_score[score_idx]= score[score_idx];
+                                     if(best_dst[x + y*b->stride]<0) best_dst[x + y*b->stride]++;
+                                     if(best_dst[x + y*b->stride]>0) best_dst[x + y*b->stride]--;
+                                     //FIXME copy instead
+                                 }
+                             }
+                         }
+                     }
+                 }
+             }
+         }
+     }
+     memcpy(s->spatial_idwt_buffer, best_dequant, height * stride * sizeof(IDWTELEM)); //FIXME work with that directly instead of copy at the end
+ }
+ #endif /* QUANTIZE2==1 */
+ #if CONFIG_SNOW_ENCODER
+ static av_cold int encode_init(AVCodecContext *avctx)
+ {
+     SnowContext *s = avctx->priv_data;
+     int plane_index;
+     if(avctx->strict_std_compliance > FF_COMPLIANCE_EXPERIMENTAL){
+         av_log(avctx, AV_LOG_ERROR, "This codec is under development, files encoded with it may not be decodable with future versions!!!\n"
+                "Use vstrict=-2 / -strict -2 to use it anyway.\n");
+         return -1;
+     }
+     if(avctx->prediction_method == DWT_97
+        && (avctx->flags & CODEC_FLAG_QSCALE)
+        && avctx->global_quality == 0){
+         av_log(avctx, AV_LOG_ERROR, "The 9/7 wavelet is incompatible with lossless mode.\n");
+         return -1;
+     }
+     s->spatial_decomposition_type= avctx->prediction_method; //FIXME add decorrelator type r transform_type
+     s->mv_scale       = (avctx->flags & CODEC_FLAG_QPEL) ? 2 : 4;
+     s->block_max_depth= (avctx->flags & CODEC_FLAG_4MV ) ? 1 : 0;
+     for(plane_index=0; plane_index<3; plane_index++){
+         s->plane[plane_index].diag_mc= 1;
+         s->plane[plane_index].htaps= 6;
+         s->plane[plane_index].hcoeff[0]=  40;
+         s->plane[plane_index].hcoeff[1]= -10;
+         s->plane[plane_index].hcoeff[2]=   2;
+         s->plane[plane_index].fast_mc= 1;
+     }
+     ff_snow_common_init(avctx);
+     ff_snow_alloc_blocks(s);
+     s->version=0;
+     s->m.avctx   = avctx;
+     s->m.flags   = avctx->flags;
+     s->m.bit_rate= avctx->bit_rate;
+     s->m.me.temp      =
+     s->m.me.scratchpad= av_mallocz((avctx->width+64)*2*16*2*sizeof(uint8_t));
+     s->m.me.map       = av_mallocz(ME_MAP_SIZE*sizeof(uint32_t));
+     s->m.me.score_map = av_mallocz(ME_MAP_SIZE*sizeof(uint32_t));
+     s->m.obmc_scratchpad= av_mallocz(MB_SIZE*MB_SIZE*12*sizeof(uint32_t));
+     h263_encode_init(&s->m); //mv_penalty
+     s->max_ref_frames = FFMAX(FFMIN(avctx->refs, MAX_REF_FRAMES), 1);
+     if(avctx->flags&CODEC_FLAG_PASS1){
+         if(!avctx->stats_out)
+             avctx->stats_out = av_mallocz(256);
+     }
+     if((avctx->flags&CODEC_FLAG_PASS2) || !(avctx->flags&CODEC_FLAG_QSCALE)){
+         if(ff_rate_control_init(&s->m) < 0)
+             return -1;
+     }
+     s->pass1_rc= !(avctx->flags & (CODEC_FLAG_QSCALE|CODEC_FLAG_PASS2));
+     avctx->coded_frame= &s->current_picture;
+     switch(avctx->pix_fmt){
+ //    case PIX_FMT_YUV444P:
+ //    case PIX_FMT_YUV422P:
+     case PIX_FMT_YUV420P:
+     case PIX_FMT_GRAY8:
+ //    case PIX_FMT_YUV411P:
+ //    case PIX_FMT_YUV410P:
+         s->colorspace_type= 0;
+         break;
+ /*    case PIX_FMT_RGB32:
+         s->colorspace= 1;
+         break;*/
+     default:
+         av_log(avctx, AV_LOG_ERROR, "pixel format not supported\n");
+         return -1;
+     }
+ //    avcodec_get_chroma_sub_sample(avctx->pix_fmt, &s->chroma_h_shift, &s->chroma_v_shift);
+     s->chroma_h_shift= 1;
+     s->chroma_v_shift= 1;
+     ff_set_cmp(&s->dsp, s->dsp.me_cmp, s->avctx->me_cmp);
+     ff_set_cmp(&s->dsp, s->dsp.me_sub_cmp, s->avctx->me_sub_cmp);
+     s->avctx->get_buffer(s->avctx, &s->input_picture);
+     if(s->avctx->me_method == ME_ITER){
+         int i;
+         int size= s->b_width * s->b_height << 2*s->block_max_depth;
+         for(i=0; i<s->max_ref_frames; i++){
+             s->ref_mvs[i]= av_mallocz(size*sizeof(int16_t[2]));
+             s->ref_scores[i]= av_mallocz(size*sizeof(uint32_t));
+         }
+     }
+     return 0;
+ }
+ //near copy & paste from dsputil, FIXME
+ static int pix_sum(uint8_t * pix, int line_size, int w)
+ {
+     int s, i, j;
+     s = 0;
+     for (i = 0; i < w; i++) {
+         for (j = 0; j < w; j++) {
+             s += pix[0];
+             pix ++;
+         }
+         pix += line_size - w;
+     }
+     return s;
+ }
+ //near copy & paste from dsputil, FIXME
+ static int pix_norm1(uint8_t * pix, int line_size, int w)
+ {
+     int s, i, j;
+     uint32_t *sq = ff_squareTbl + 256;
+     s = 0;
+     for (i = 0; i < w; i++) {
+         for (j = 0; j < w; j ++) {
+             s += sq[pix[0]];
+             pix ++;
+         }
+         pix += line_size - w;
+     }
+     return s;
+ }
+ //FIXME copy&paste
+ #define P_LEFT P[1]
+ #define P_TOP P[2]
+ #define P_TOPRIGHT P[3]
+ #define P_MEDIAN P[4]
+ #define P_MV1 P[9]
+ #define FLAG_QPEL   1 //must be 1
+ static int encode_q_branch(SnowContext *s, int level, int x, int y){
+     uint8_t p_buffer[1024];
+     uint8_t i_buffer[1024];
+     uint8_t p_state[sizeof(s->block_state)];
+     uint8_t i_state[sizeof(s->block_state)];
+     RangeCoder pc, ic;
+     uint8_t *pbbak= s->c.bytestream;
+     uint8_t *pbbak_start= s->c.bytestream_start;
+     int score, score2, iscore, i_len, p_len, block_s, sum, base_bits;
+     const int w= s->b_width  << s->block_max_depth;
+     const int h= s->b_height << s->block_max_depth;
+     const int rem_depth= s->block_max_depth - level;
+     const int index= (x + y*w) << rem_depth;
+     const int block_w= 1<<(LOG2_MB_SIZE - level);
+     int trx= (x+1)<<rem_depth;
+     int try= (y+1)<<rem_depth;
+     const BlockNode *left  = x ? &s->block[index-1] : &null_block;
+     const BlockNode *top   = y ? &s->block[index-w] : &null_block;
+     const BlockNode *right = trx<w ? &s->block[index+1] : &null_block;
+     const BlockNode *bottom= try<h ? &s->block[index+w] : &null_block;
+     const BlockNode *tl    = y && x ? &s->block[index-w-1] : left;
+     const BlockNode *tr    = y && trx<w && ((x&1)==0 || level==0) ? &s->block[index-w+(1<<rem_depth)] : tl; //FIXME use lt
+     int pl = left->color[0];
+     int pcb= left->color[1];
+     int pcr= left->color[2];
+     int pmx, pmy;
+     int mx=0, my=0;
+     int l,cr,cb;
+     const int stride= s->current_picture.linesize[0];
+     const int uvstride= s->current_picture.linesize[1];
+     uint8_t *current_data[3]= { s->input_picture.data[0] + (x + y*  stride)*block_w,
+                                 s->input_picture.data[1] + (x + y*uvstride)*block_w/2,
+                                 s->input_picture.data[2] + (x + y*uvstride)*block_w/2};
+     int P[10][2];
+     int16_t last_mv[3][2];
+     int qpel= !!(s->avctx->flags & CODEC_FLAG_QPEL); //unused
+     const int shift= 1+qpel;
+     MotionEstContext *c= &s->m.me;
+     int ref_context= av_log2(2*left->ref) + av_log2(2*top->ref);
+     int mx_context= av_log2(2*FFABS(left->mx - top->mx));
+     int my_context= av_log2(2*FFABS(left->my - top->my));
+     int s_context= 2*left->level + 2*top->level + tl->level + tr->level;
+     int ref, best_ref, ref_score, ref_mx, ref_my;
+     assert(sizeof(s->block_state) >= 256);
+     if(s->keyframe){
+         set_blocks(s, level, x, y, pl, pcb, pcr, 0, 0, 0, BLOCK_INTRA);
+         return 0;
+     }
+ //    clip predictors / edge ?
+     P_LEFT[0]= left->mx;
+     P_LEFT[1]= left->my;
+     P_TOP [0]= top->mx;
+     P_TOP [1]= top->my;
+     P_TOPRIGHT[0]= tr->mx;
+     P_TOPRIGHT[1]= tr->my;
+     last_mv[0][0]= s->block[index].mx;
+     last_mv[0][1]= s->block[index].my;
+     last_mv[1][0]= right->mx;
+     last_mv[1][1]= right->my;
+     last_mv[2][0]= bottom->mx;
+     last_mv[2][1]= bottom->my;
+     s->m.mb_stride=2;
+     s->m.mb_x=
+     s->m.mb_y= 0;
+     c->skip= 0;
+     assert(c->  stride ==   stride);
+     assert(c->uvstride == uvstride);
+     c->penalty_factor    = get_penalty_factor(s->lambda, s->lambda2, c->avctx->me_cmp);
+     c->sub_penalty_factor= get_penalty_factor(s->lambda, s->lambda2, c->avctx->me_sub_cmp);
+     c->mb_penalty_factor = get_penalty_factor(s->lambda, s->lambda2, c->avctx->mb_cmp);
+     c->current_mv_penalty= c->mv_penalty[s->m.f_code=1] + MAX_MV;
+     c->xmin = - x*block_w - 16+3;
+     c->ymin = - y*block_w - 16+3;
+     c->xmax = - (x+1)*block_w + (w<<(LOG2_MB_SIZE - s->block_max_depth)) + 16-3;
+     c->ymax = - (y+1)*block_w + (h<<(LOG2_MB_SIZE - s->block_max_depth)) + 16-3;
+     if(P_LEFT[0]     > (c->xmax<<shift)) P_LEFT[0]    = (c->xmax<<shift);
+     if(P_LEFT[1]     > (c->ymax<<shift)) P_LEFT[1]    = (c->ymax<<shift);
+     if(P_TOP[0]      > (c->xmax<<shift)) P_TOP[0]     = (c->xmax<<shift);
+     if(P_TOP[1]      > (c->ymax<<shift)) P_TOP[1]     = (c->ymax<<shift);
+     if(P_TOPRIGHT[0] < (c->xmin<<shift)) P_TOPRIGHT[0]= (c->xmin<<shift);
+     if(P_TOPRIGHT[0] > (c->xmax<<shift)) P_TOPRIGHT[0]= (c->xmax<<shift); //due to pmx no clip
+     if(P_TOPRIGHT[1] > (c->ymax<<shift)) P_TOPRIGHT[1]= (c->ymax<<shift);
+     P_MEDIAN[0]= mid_pred(P_LEFT[0], P_TOP[0], P_TOPRIGHT[0]);
+     P_MEDIAN[1]= mid_pred(P_LEFT[1], P_TOP[1], P_TOPRIGHT[1]);
+     if (!y) {
+         c->pred_x= P_LEFT[0];
+         c->pred_y= P_LEFT[1];
+     } else {
+         c->pred_x = P_MEDIAN[0];
+         c->pred_y = P_MEDIAN[1];
+     }
+     score= INT_MAX;
+     best_ref= 0;
+     for(ref=0; ref<s->ref_frames; ref++){
+         init_ref(c, current_data, s->last_picture[ref].data, NULL, block_w*x, block_w*y, 0);
+         ref_score= ff_epzs_motion_search(&s->m, &ref_mx, &ref_my, P, 0, /*ref_index*/ 0, last_mv,
+                                          (1<<16)>>shift, level-LOG2_MB_SIZE+4, block_w);
+         assert(ref_mx >= c->xmin);
+         assert(ref_mx <= c->xmax);
+         assert(ref_my >= c->ymin);
+         assert(ref_my <= c->ymax);
+         ref_score= c->sub_motion_search(&s->m, &ref_mx, &ref_my, ref_score, 0, 0, level-LOG2_MB_SIZE+4, block_w);
+         ref_score= ff_get_mb_score(&s->m, ref_mx, ref_my, 0, 0, level-LOG2_MB_SIZE+4, block_w, 0);
+         ref_score+= 2*av_log2(2*ref)*c->penalty_factor;
+         if(s->ref_mvs[ref]){
+             s->ref_mvs[ref][index][0]= ref_mx;
+             s->ref_mvs[ref][index][1]= ref_my;
+             s->ref_scores[ref][index]= ref_score;
+         }
+         if(score > ref_score){
+             score= ref_score;
+             best_ref= ref;
+             mx= ref_mx;
+             my= ref_my;
+         }
+     }
+     //FIXME if mb_cmp != SSE then intra cannot be compared currently and mb_penalty vs. lambda2
+   //  subpel search
+     base_bits= get_rac_count(&s->c) - 8*(s->c.bytestream - s->c.bytestream_start);
+     pc= s->c;
+     pc.bytestream_start=
+     pc.bytestream= p_buffer; //FIXME end/start? and at the other stoo
+     memcpy(p_state, s->block_state, sizeof(s->block_state));
+     if(level!=s->block_max_depth)
+         put_rac(&pc, &p_state[4 + s_context], 1);
+     put_rac(&pc, &p_state[1 + left->type + top->type], 0);
+     if(s->ref_frames > 1)
+         put_symbol(&pc, &p_state[128 + 1024 + 32*ref_context], best_ref, 0);
+     pred_mv(s, &pmx, &pmy, best_ref, left, top, tr);
+     put_symbol(&pc, &p_state[128 + 32*(mx_context + 16*!!best_ref)], mx - pmx, 1);
+     put_symbol(&pc, &p_state[128 + 32*(my_context + 16*!!best_ref)], my - pmy, 1);
+     p_len= pc.bytestream - pc.bytestream_start;
+     score += (s->lambda2*(get_rac_count(&pc)-base_bits))>>FF_LAMBDA_SHIFT;
+     block_s= block_w*block_w;
+     sum = pix_sum(current_data[0], stride, block_w);
+     l= (sum + block_s/2)/block_s;
+     iscore = pix_norm1(current_data[0], stride, block_w) - 2*l*sum + l*l*block_s;
+     block_s= block_w*block_w>>2;
+     sum = pix_sum(current_data[1], uvstride, block_w>>1);
+     cb= (sum + block_s/2)/block_s;
+ //    iscore += pix_norm1(&current_mb[1][0], uvstride, block_w>>1) - 2*cb*sum + cb*cb*block_s;
+     sum = pix_sum(current_data[2], uvstride, block_w>>1);
+     cr= (sum + block_s/2)/block_s;
+ //    iscore += pix_norm1(&current_mb[2][0], uvstride, block_w>>1) - 2*cr*sum + cr*cr*block_s;
+     ic= s->c;
+     ic.bytestream_start=
+     ic.bytestream= i_buffer; //FIXME end/start? and at the other stoo
+     memcpy(i_state, s->block_state, sizeof(s->block_state));
+     if(level!=s->block_max_depth)
+         put_rac(&ic, &i_state[4 + s_context], 1);
+     put_rac(&ic, &i_state[1 + left->type + top->type], 1);
+     put_symbol(&ic, &i_state[32],  l-pl , 1);
+     put_symbol(&ic, &i_state[64], cb-pcb, 1);
+     put_symbol(&ic, &i_state[96], cr-pcr, 1);
+     i_len= ic.bytestream - ic.bytestream_start;
+     iscore += (s->lambda2*(get_rac_count(&ic)-base_bits))>>FF_LAMBDA_SHIFT;
+ //    assert(score==256*256*256*64-1);
+     assert(iscore < 255*255*256 + s->lambda2*10);
+     assert(iscore >= 0);
+     assert(l>=0 && l<=255);
+     assert(pl>=0 && pl<=255);
+     if(level==0){
+         int varc= iscore >> 8;
+         int vard= score >> 8;
+         if (vard <= 64 || vard < varc)
+             c->scene_change_score+= ff_sqrt(vard) - ff_sqrt(varc);
+         else
+             c->scene_change_score+= s->m.qscale;
+     }
+     if(level!=s->block_max_depth){
+         put_rac(&s->c, &s->block_state[4 + s_context], 0);
+         score2 = encode_q_branch(s, level+1, 2*x+0, 2*y+0);
+         score2+= encode_q_branch(s, level+1, 2*x+1, 2*y+0);
+         score2+= encode_q_branch(s, level+1, 2*x+0, 2*y+1);
+         score2+= encode_q_branch(s, level+1, 2*x+1, 2*y+1);
+         score2+= s->lambda2>>FF_LAMBDA_SHIFT; //FIXME exact split overhead
+         if(score2 < score && score2 < iscore)
+             return score2;
+     }
+     if(iscore < score){
+         pred_mv(s, &pmx, &pmy, 0, left, top, tr);
+         memcpy(pbbak, i_buffer, i_len);
+         s->c= ic;
+         s->c.bytestream_start= pbbak_start;
+         s->c.bytestream= pbbak + i_len;
+         set_blocks(s, level, x, y, l, cb, cr, pmx, pmy, 0, BLOCK_INTRA);
+         memcpy(s->block_state, i_state, sizeof(s->block_state));
+         return iscore;
+     }else{
+         memcpy(pbbak, p_buffer, p_len);
+         s->c= pc;
+         s->c.bytestream_start= pbbak_start;
+         s->c.bytestream= pbbak + p_len;
+         set_blocks(s, level, x, y, pl, pcb, pcr, mx, my, best_ref, 0);
+         memcpy(s->block_state, p_state, sizeof(s->block_state));
+         return score;
+     }
+ }
+ static void encode_q_branch2(SnowContext *s, int level, int x, int y){
+     const int w= s->b_width  << s->block_max_depth;
+     const int rem_depth= s->block_max_depth - level;
+     const int index= (x + y*w) << rem_depth;
+     int trx= (x+1)<<rem_depth;
+     BlockNode *b= &s->block[index];
+     const BlockNode *left  = x ? &s->block[index-1] : &null_block;
+     const BlockNode *top   = y ? &s->block[index-w] : &null_block;
+     const BlockNode *tl    = y && x ? &s->block[index-w-1] : left;
+     const BlockNode *tr    = y && trx<w && ((x&1)==0 || level==0) ? &s->block[index-w+(1<<rem_depth)] : tl; //FIXME use lt
+     int pl = left->color[0];
+     int pcb= left->color[1];
+     int pcr= left->color[2];
+     int pmx, pmy;
+     int ref_context= av_log2(2*left->ref) + av_log2(2*top->ref);
+     int mx_context= av_log2(2*FFABS(left->mx - top->mx)) + 16*!!b->ref;
+     int my_context= av_log2(2*FFABS(left->my - top->my)) + 16*!!b->ref;
+     int s_context= 2*left->level + 2*top->level + tl->level + tr->level;
+     if(s->keyframe){
+         set_blocks(s, level, x, y, pl, pcb, pcr, 0, 0, 0, BLOCK_INTRA);
+         return;
+     }
+     if(level!=s->block_max_depth){
+         if(same_block(b,b+1) && same_block(b,b+w) && same_block(b,b+w+1)){
+             put_rac(&s->c, &s->block_state[4 + s_context], 1);
+         }else{
+             put_rac(&s->c, &s->block_state[4 + s_context], 0);
+             encode_q_branch2(s, level+1, 2*x+0, 2*y+0);
+             encode_q_branch2(s, level+1, 2*x+1, 2*y+0);
+             encode_q_branch2(s, level+1, 2*x+0, 2*y+1);
+             encode_q_branch2(s, level+1, 2*x+1, 2*y+1);
+             return;
+         }
+     }
+     if(b->type & BLOCK_INTRA){
+         pred_mv(s, &pmx, &pmy, 0, left, top, tr);
+         put_rac(&s->c, &s->block_state[1 + (left->type&1) + (top->type&1)], 1);
+         put_symbol(&s->c, &s->block_state[32], b->color[0]-pl , 1);
+         put_symbol(&s->c, &s->block_state[64], b->color[1]-pcb, 1);
+         put_symbol(&s->c, &s->block_state[96], b->color[2]-pcr, 1);
+         set_blocks(s, level, x, y, b->color[0], b->color[1], b->color[2], pmx, pmy, 0, BLOCK_INTRA);
+     }else{
+         pred_mv(s, &pmx, &pmy, b->ref, left, top, tr);
+         put_rac(&s->c, &s->block_state[1 + (left->type&1) + (top->type&1)], 0);
+         if(s->ref_frames > 1)
+             put_symbol(&s->c, &s->block_state[128 + 1024 + 32*ref_context], b->ref, 0);
+         put_symbol(&s->c, &s->block_state[128 + 32*mx_context], b->mx - pmx, 1);
+         put_symbol(&s->c, &s->block_state[128 + 32*my_context], b->my - pmy, 1);
+         set_blocks(s, level, x, y, pl, pcb, pcr, b->mx, b->my, b->ref, 0);
+     }
+ }
+ static int get_dc(SnowContext *s, int mb_x, int mb_y, int plane_index){
+     int i, x2, y2;
+     Plane *p= &s->plane[plane_index];
+     const int block_size = MB_SIZE >> s->block_max_depth;
+     const int block_w    = plane_index ? block_size/2 : block_size;
+     const uint8_t *obmc  = plane_index ? obmc_tab[s->block_max_depth+1] : obmc_tab[s->block_max_depth];
+     const int obmc_stride= plane_index ? block_size : 2*block_size;
+     const int ref_stride= s->current_picture.linesize[plane_index];
+     uint8_t *src= s-> input_picture.data[plane_index];
+     IDWTELEM *dst= (IDWTELEM*)s->m.obmc_scratchpad + plane_index*block_size*block_size*4; //FIXME change to unsigned
+     const int b_stride = s->b_width << s->block_max_depth;
+     const int w= p->width;
+     const int h= p->height;
+     int index= mb_x + mb_y*b_stride;
+     BlockNode *b= &s->block[index];
+     BlockNode backup= *b;
+     int ab=0;
+     int aa=0;
+     b->type|= BLOCK_INTRA;
+     b->color[plane_index]= 0;
+     memset(dst, 0, obmc_stride*obmc_stride*sizeof(IDWTELEM));
+     for(i=0; i<4; i++){
+         int mb_x2= mb_x + (i &1) - 1;
+         int mb_y2= mb_y + (i>>1) - 1;
+         int x= block_w*mb_x2 + block_w/2;
+         int y= block_w*mb_y2 + block_w/2;
+         add_yblock(s, 0, NULL, dst + ((i&1)+(i>>1)*obmc_stride)*block_w, NULL, obmc,
+                     x, y, block_w, block_w, w, h, obmc_stride, ref_stride, obmc_stride, mb_x2, mb_y2, 0, 0, plane_index);
+         for(y2= FFMAX(y, 0); y2<FFMIN(h, y+block_w); y2++){
+             for(x2= FFMAX(x, 0); x2<FFMIN(w, x+block_w); x2++){
+                 int index= x2-(block_w*mb_x - block_w/2) + (y2-(block_w*mb_y - block_w/2))*obmc_stride;
+                 int obmc_v= obmc[index];
+                 int d;
+                 if(y<0) obmc_v += obmc[index + block_w*obmc_stride];
+                 if(x<0) obmc_v += obmc[index + block_w];
+                 if(y+block_w>h) obmc_v += obmc[index - block_w*obmc_stride];
+                 if(x+block_w>w) obmc_v += obmc[index - block_w];
+                 //FIXME precalculate this or simplify it somehow else
+                 d = -dst[index] + (1<<(FRAC_BITS-1));
+                 dst[index] = d;
+                 ab += (src[x2 + y2*ref_stride] - (d>>FRAC_BITS)) * obmc_v;
+                 aa += obmc_v * obmc_v; //FIXME precalculate this
+             }
+         }
+     }
+     *b= backup;
+     return av_clip(((ab<<LOG2_OBMC_MAX) + aa/2)/aa, 0, 255); //FIXME we should not need clipping
+ }
+ static inline int get_block_bits(SnowContext *s, int x, int y, int w){
+     const int b_stride = s->b_width << s->block_max_depth;
+     const int b_height = s->b_height<< s->block_max_depth;
+     int index= x + y*b_stride;
+     const BlockNode *b     = &s->block[index];
+     const BlockNode *left  = x ? &s->block[index-1] : &null_block;
+     const BlockNode *top   = y ? &s->block[index-b_stride] : &null_block;
+     const BlockNode *tl    = y && x ? &s->block[index-b_stride-1] : left;
+     const BlockNode *tr    = y && x+w<b_stride ? &s->block[index-b_stride+w] : tl;
+     int dmx, dmy;
+ //  int mx_context= av_log2(2*FFABS(left->mx - top->mx));
+ //  int my_context= av_log2(2*FFABS(left->my - top->my));
+     if(x<0 || x>=b_stride || y>=b_height)
+         return 0;
+ /*
+ 1            0      0
+ 01X          1-2    1
+ 001XX        3-6    2-3
+ 0001XXX      7-14   4-7
+ 00001XXXX   15-30   8-15
+ */
+ //FIXME try accurate rate
+ //FIXME intra and inter predictors if surrounding blocks are not the same type
+     if(b->type & BLOCK_INTRA){
+         return 3+2*( av_log2(2*FFABS(left->color[0] - b->color[0]))
+                    + av_log2(2*FFABS(left->color[1] - b->color[1]))
+                    + av_log2(2*FFABS(left->color[2] - b->color[2])));
+     }else{
+         pred_mv(s, &dmx, &dmy, b->ref, left, top, tr);
+         dmx-= b->mx;
+         dmy-= b->my;
+         return 2*(1 + av_log2(2*FFABS(dmx)) //FIXME kill the 2* can be merged in lambda
+                     + av_log2(2*FFABS(dmy))
+                     + av_log2(2*b->ref));
+     }
+ }
+ static int get_block_rd(SnowContext *s, int mb_x, int mb_y, int plane_index, const uint8_t *obmc_edged){
+     Plane *p= &s->plane[plane_index];
+     const int block_size = MB_SIZE >> s->block_max_depth;
+     const int block_w    = plane_index ? block_size/2 : block_size;
+     const int obmc_stride= plane_index ? block_size : 2*block_size;
+     const int ref_stride= s->current_picture.linesize[plane_index];
+     uint8_t *dst= s->current_picture.data[plane_index];
+     uint8_t *src= s->  input_picture.data[plane_index];
+     IDWTELEM *pred= (IDWTELEM*)s->m.obmc_scratchpad + plane_index*block_size*block_size*4;
+     uint8_t *cur = s->scratchbuf;
+     uint8_t tmp[ref_stride*(2*MB_SIZE+HTAPS_MAX-1)];
+     const int b_stride = s->b_width << s->block_max_depth;
+     const int b_height = s->b_height<< s->block_max_depth;
+     const int w= p->width;
+     const int h= p->height;
+     int distortion;
+     int rate= 0;
+     const int penalty_factor= get_penalty_factor(s->lambda, s->lambda2, s->avctx->me_cmp);
+     int sx= block_w*mb_x - block_w/2;
+     int sy= block_w*mb_y - block_w/2;
+     int x0= FFMAX(0,-sx);
+     int y0= FFMAX(0,-sy);
+     int x1= FFMIN(block_w*2, w-sx);
+     int y1= FFMIN(block_w*2, h-sy);
+     int i,x,y;
+     ff_snow_pred_block(s, cur, tmp, ref_stride, sx, sy, block_w*2, block_w*2, &s->block[mb_x + mb_y*b_stride], plane_index, w, h);
+     for(y=y0; y<y1; y++){
+         const uint8_t *obmc1= obmc_edged + y*obmc_stride;
+         const IDWTELEM *pred1 = pred + y*obmc_stride;
+         uint8_t *cur1 = cur + y*ref_stride;
+         uint8_t *dst1 = dst + sx + (sy+y)*ref_stride;
+         for(x=x0; x<x1; x++){
+ #if FRAC_BITS >= LOG2_OBMC_MAX
+             int v = (cur1[x] * obmc1[x]) << (FRAC_BITS - LOG2_OBMC_MAX);
+ #else
+             int v = (cur1[x] * obmc1[x] + (1<<(LOG2_OBMC_MAX - FRAC_BITS-1))) >> (LOG2_OBMC_MAX - FRAC_BITS);
+ #endif
+             v = (v + pred1[x]) >> FRAC_BITS;
+             if(v&(~255)) v= ~(v>>31);
+             dst1[x] = v;
+         }
+     }
+     /* copy the regions where obmc[] = (uint8_t)256 */
+     if(LOG2_OBMC_MAX == 8
+         && (mb_x == 0 || mb_x == b_stride-1)
+         && (mb_y == 0 || mb_y == b_height-1)){
+         if(mb_x == 0)
+             x1 = block_w;
+         else
+             x0 = block_w;
+         if(mb_y == 0)
+             y1 = block_w;
+         else
+             y0 = block_w;
+         for(y=y0; y<y1; y++)
+             memcpy(dst + sx+x0 + (sy+y)*ref_stride, cur + x0 + y*ref_stride, x1-x0);
+     }
+     if(block_w==16){
+         /* FIXME rearrange dsputil to fit 32x32 cmp functions */
+         /* FIXME check alignment of the cmp wavelet vs the encoding wavelet */
+         /* FIXME cmps overlap but do not cover the wavelet's whole support.
+          * So improving the score of one block is not strictly guaranteed
+          * to improve the score of the whole frame, thus iterative motion
+          * estimation does not always converge. */
+         if(s->avctx->me_cmp == FF_CMP_W97)
+             distortion = ff_w97_32_c(&s->m, src + sx + sy*ref_stride, dst + sx + sy*ref_stride, ref_stride, 32);
+         else if(s->avctx->me_cmp == FF_CMP_W53)
+             distortion = ff_w53_32_c(&s->m, src + sx + sy*ref_stride, dst + sx + sy*ref_stride, ref_stride, 32);
+         else{
+             distortion = 0;
+             for(i=0; i<4; i++){
+                 int off = sx+16*(i&1) + (sy+16*(i>>1))*ref_stride;
+                 distortion += s->dsp.me_cmp[0](&s->m, src + off, dst + off, ref_stride, 16);
+             }
+         }
+     }else{
+         assert(block_w==8);
+         distortion = s->dsp.me_cmp[0](&s->m, src + sx + sy*ref_stride, dst + sx + sy*ref_stride, ref_stride, block_w*2);
+     }
+     if(plane_index==0){
+         for(i=0; i<4; i++){
+ /* ..RRr
+  * .RXx.
+  * rxx..
+  */
+             rate += get_block_bits(s, mb_x + (i&1) - (i>>1), mb_y + (i>>1), 1);
+         }
+         if(mb_x == b_stride-2)
+             rate += get_block_bits(s, mb_x + 1, mb_y + 1, 1);
+     }
+     return distortion + rate*penalty_factor;
+ }
+ static int get_4block_rd(SnowContext *s, int mb_x, int mb_y, int plane_index){
+     int i, y2;
+     Plane *p= &s->plane[plane_index];
+     const int block_size = MB_SIZE >> s->block_max_depth;
+     const int block_w    = plane_index ? block_size/2 : block_size;
+     const uint8_t *obmc  = plane_index ? obmc_tab[s->block_max_depth+1] : obmc_tab[s->block_max_depth];
+     const int obmc_stride= plane_index ? block_size : 2*block_size;
+     const int ref_stride= s->current_picture.linesize[plane_index];
+     uint8_t *dst= s->current_picture.data[plane_index];
+     uint8_t *src= s-> input_picture.data[plane_index];
+     //FIXME zero_dst is const but add_yblock changes dst if add is 0 (this is never the case for dst=zero_dst
+     // const has only been removed from zero_dst to suppress a warning
+     static IDWTELEM zero_dst[4096]; //FIXME
+     const int b_stride = s->b_width << s->block_max_depth;
+     const int w= p->width;
+     const int h= p->height;
+     int distortion= 0;
+     int rate= 0;
+     const int penalty_factor= get_penalty_factor(s->lambda, s->lambda2, s->avctx->me_cmp);
+     for(i=0; i<9; i++){
+         int mb_x2= mb_x + (i%3) - 1;
+         int mb_y2= mb_y + (i/3) - 1;
+         int x= block_w*mb_x2 + block_w/2;
+         int y= block_w*mb_y2 + block_w/2;
+         add_yblock(s, 0, NULL, zero_dst, dst, obmc,
+                    x, y, block_w, block_w, w, h, /*dst_stride*/0, ref_stride, obmc_stride, mb_x2, mb_y2, 1, 1, plane_index);
+         //FIXME find a cleaner/simpler way to skip the outside stuff
+         for(y2= y; y2<0; y2++)
+             memcpy(dst + x + y2*ref_stride, src + x + y2*ref_stride, block_w);
+         for(y2= h; y2<y+block_w; y2++)
+             memcpy(dst + x + y2*ref_stride, src + x + y2*ref_stride, block_w);
+         if(x<0){
+             for(y2= y; y2<y+block_w; y2++)
+                 memcpy(dst + x + y2*ref_stride, src + x + y2*ref_stride, -x);
+         }
+         if(x+block_w > w){
+             for(y2= y; y2<y+block_w; y2++)
+                 memcpy(dst + w + y2*ref_stride, src + w + y2*ref_stride, x+block_w - w);
+         }
+         assert(block_w== 8 || block_w==16);
+         distortion += s->dsp.me_cmp[block_w==8](&s->m, src + x + y*ref_stride, dst + x + y*ref_stride, ref_stride, block_w);
+     }
+     if(plane_index==0){
+         BlockNode *b= &s->block[mb_x+mb_y*b_stride];
+         int merged= same_block(b,b+1) && same_block(b,b+b_stride) && same_block(b,b+b_stride+1);
+ /* ..RRRr
+  * .RXXx.
+  * .RXXx.
+  * rxxx.
+  */
+         if(merged)
+             rate = get_block_bits(s, mb_x, mb_y, 2);
+         for(i=merged?4:0; i<9; i++){
+             static const int dxy[9][2] = {{0,0},{1,0},{0,1},{1,1},{2,0},{2,1},{-1,2},{0,2},{1,2}};
+             rate += get_block_bits(s, mb_x + dxy[i][0], mb_y + dxy[i][1], 1);
+         }
+     }
+     return distortion + rate*penalty_factor;
+ }
+ static int encode_subband_c0run(SnowContext *s, SubBand *b, IDWTELEM *src, IDWTELEM *parent, int stride, int orientation){
+     const int w= b->width;
+     const int h= b->height;
+     int x, y;
+     if(1){
+         int run=0;
+         int runs[w*h];
+         int run_index=0;
+         int max_index;
+         for(y=0; y<h; y++){
+             for(x=0; x<w; x++){
+                 int v, p=0;
+                 int /*ll=0, */l=0, lt=0, t=0, rt=0;
+                 v= src[x + y*stride];
+                 if(y){
+                     t= src[x + (y-1)*stride];
+                     if(x){
+                         lt= src[x - 1 + (y-1)*stride];
+                     }
+                     if(x + 1 < w){
+                         rt= src[x + 1 + (y-1)*stride];
+                     }
+                 }
+                 if(x){
+                     l= src[x - 1 + y*stride];
+                     /*if(x > 1){
+                         if(orientation==1) ll= src[y + (x-2)*stride];
+                         else               ll= src[x - 2 + y*stride];
+                     }*/
+                 }
+                 if(parent){
+                     int px= x>>1;
+                     int py= y>>1;
+                     if(px<b->parent->width && py<b->parent->height)
+                         p= parent[px + py*2*stride];
+                 }
+                 if(!(/*ll|*/l|lt|t|rt|p)){
+                     if(v){
+                         runs[run_index++]= run;
+                         run=0;
+                     }else{
+                         run++;
+                     }
+                 }
+             }
+         }
+         max_index= run_index;
+         runs[run_index++]= run;
+         run_index=0;
+         run= runs[run_index++];
+         put_symbol2(&s->c, b->state[30], max_index, 0);
+         if(run_index <= max_index)
+             put_symbol2(&s->c, b->state[1], run, 3);
+         for(y=0; y<h; y++){
+             if(s->c.bytestream_end - s->c.bytestream < w*40){
+                 av_log(s->avctx, AV_LOG_ERROR, "encoded frame too large\n");
+                 return -1;
+             }
+             for(x=0; x<w; x++){
+                 int v, p=0;
+                 int /*ll=0, */l=0, lt=0, t=0, rt=0;
+                 v= src[x + y*stride];
+                 if(y){
+                     t= src[x + (y-1)*stride];
+                     if(x){
+                         lt= src[x - 1 + (y-1)*stride];
+                     }
+                     if(x + 1 < w){
+                         rt= src[x + 1 + (y-1)*stride];
+                     }
+                 }
+                 if(x){
+                     l= src[x - 1 + y*stride];
+                     /*if(x > 1){
+                         if(orientation==1) ll= src[y + (x-2)*stride];
+                         else               ll= src[x - 2 + y*stride];
+                     }*/
+                 }
+                 if(parent){
+                     int px= x>>1;
+                     int py= y>>1;
+                     if(px<b->parent->width && py<b->parent->height)
+                         p= parent[px + py*2*stride];
+                 }
+                 if(/*ll|*/l|lt|t|rt|p){
+                     int context= av_log2(/*FFABS(ll) + */3*FFABS(l) + FFABS(lt) + 2*FFABS(t) + FFABS(rt) + FFABS(p));
+                     put_rac(&s->c, &b->state[0][context], !!v);
+                 }else{
+                     if(!run){
+                         run= runs[run_index++];
+                         if(run_index <= max_index)
+                             put_symbol2(&s->c, b->state[1], run, 3);
+                         assert(v);
+                     }else{
+                         run--;
+                         assert(!v);
+                     }
+                 }
+                 if(v){
+                     int context= av_log2(/*FFABS(ll) + */3*FFABS(l) + FFABS(lt) + 2*FFABS(t) + FFABS(rt) + FFABS(p));
+                     int l2= 2*FFABS(l) + (l<0);
+                     int t2= 2*FFABS(t) + (t<0);
+                     put_symbol2(&s->c, b->state[context + 2], FFABS(v)-1, context-4);
+                     put_rac(&s->c, &b->state[0][16 + 1 + 3 + quant3bA[l2&0xFF] + 3*quant3bA[t2&0xFF]], v<0);
+                 }
+             }
+         }
+     }
+     return 0;
+ }
+ static int encode_subband(SnowContext *s, SubBand *b, IDWTELEM *src, IDWTELEM *parent, int stride, int orientation){
+ //    encode_subband_qtree(s, b, src, parent, stride, orientation);
+ //    encode_subband_z0run(s, b, src, parent, stride, orientation);
+     return encode_subband_c0run(s, b, src, parent, stride, orientation);
+ //    encode_subband_dzr(s, b, src, parent, stride, orientation);
+ }
+ static av_always_inline int check_block(SnowContext *s, int mb_x, int mb_y, int p[3], int intra, const uint8_t *obmc_edged, int *best_rd){
+     const int b_stride= s->b_width << s->block_max_depth;
+     BlockNode *block= &s->block[mb_x + mb_y * b_stride];
+     BlockNode backup= *block;
+     int rd, index, value;
+     assert(mb_x>=0 && mb_y>=0);
+     assert(mb_x<b_stride);
+     if(intra){
+         block->color[0] = p[0];
+         block->color[1] = p[1];
+         block->color[2] = p[2];
+         block->type |= BLOCK_INTRA;
+     }else{
+         index= (p[0] + 31*p[1]) & (ME_CACHE_SIZE-1);
+         value= s->me_cache_generation + (p[0]>>10) + (p[1]<<6) + (block->ref<<12);
+         if(s->me_cache[index] == value)
+             return 0;
+         s->me_cache[index]= value;
+         block->mx= p[0];
+         block->my= p[1];
+         block->type &= ~BLOCK_INTRA;
+     }
+     rd= get_block_rd(s, mb_x, mb_y, 0, obmc_edged);
+ //FIXME chroma
+     if(rd < *best_rd){
+         *best_rd= rd;
+         return 1;
+     }else{
+         *block= backup;
+         return 0;
+     }
+ }
+ /* special case for int[2] args we discard afterwards,
+  * fixes compilation problem with gcc 2.95 */
+ static av_always_inline int check_block_inter(SnowContext *s, int mb_x, int mb_y, int p0, int p1, const uint8_t *obmc_edged, int *best_rd){
+     int p[2] = {p0, p1};
+     return check_block(s, mb_x, mb_y, p, 0, obmc_edged, best_rd);
+ }
+ static av_always_inline int check_4block_inter(SnowContext *s, int mb_x, int mb_y, int p0, int p1, int ref, int *best_rd){
+     const int b_stride= s->b_width << s->block_max_depth;
+     BlockNode *block= &s->block[mb_x + mb_y * b_stride];
+     BlockNode backup[4]= {block[0], block[1], block[b_stride], block[b_stride+1]};
+     int rd, index, value;
+     assert(mb_x>=0 && mb_y>=0);
+     assert(mb_x<b_stride);
+     assert(((mb_x|mb_y)&1) == 0);
+     index= (p0 + 31*p1) & (ME_CACHE_SIZE-1);
+     value= s->me_cache_generation + (p0>>10) + (p1<<6) + (block->ref<<12);
+     if(s->me_cache[index] == value)
+         return 0;
+     s->me_cache[index]= value;
+     block->mx= p0;
+     block->my= p1;
+     block->ref= ref;
+     block->type &= ~BLOCK_INTRA;
+     block[1]= block[b_stride]= block[b_stride+1]= *block;
+     rd= get_4block_rd(s, mb_x, mb_y, 0);
+ //FIXME chroma
+     if(rd < *best_rd){
+         *best_rd= rd;
+         return 1;
+     }else{
+         block[0]= backup[0];
+         block[1]= backup[1];
+         block[b_stride]= backup[2];
+         block[b_stride+1]= backup[3];
+         return 0;
+     }
+ }
+ static void iterative_me(SnowContext *s){
+     int pass, mb_x, mb_y;
+     const int b_width = s->b_width  << s->block_max_depth;
+     const int b_height= s->b_height << s->block_max_depth;
+     const int b_stride= b_width;
+     int color[3];
+     {
+         RangeCoder r = s->c;
+         uint8_t state[sizeof(s->block_state)];
+         memcpy(state, s->block_state, sizeof(s->block_state));
+         for(mb_y= 0; mb_y<s->b_height; mb_y++)
+             for(mb_x= 0; mb_x<s->b_width; mb_x++)
+                 encode_q_branch(s, 0, mb_x, mb_y);
+         s->c = r;
+         memcpy(s->block_state, state, sizeof(s->block_state));
+     }
+     for(pass=0; pass<25; pass++){
+         int change= 0;
+         for(mb_y= 0; mb_y<b_height; mb_y++){
+             for(mb_x= 0; mb_x<b_width; mb_x++){
+                 int dia_change, i, j, ref;
+                 int best_rd= INT_MAX, ref_rd;
+                 BlockNode backup, ref_b;
+                 const int index= mb_x + mb_y * b_stride;
+                 BlockNode *block= &s->block[index];
+                 BlockNode *tb =                   mb_y            ? &s->block[index-b_stride  ] : NULL;
+                 BlockNode *lb = mb_x                              ? &s->block[index         -1] : NULL;
+                 BlockNode *rb = mb_x+1<b_width                    ? &s->block[index         +1] : NULL;
+                 BlockNode *bb =                   mb_y+1<b_height ? &s->block[index+b_stride  ] : NULL;
+                 BlockNode *tlb= mb_x           && mb_y            ? &s->block[index-b_stride-1] : NULL;
+                 BlockNode *trb= mb_x+1<b_width && mb_y            ? &s->block[index-b_stride+1] : NULL;
+                 BlockNode *blb= mb_x           && mb_y+1<b_height ? &s->block[index+b_stride-1] : NULL;
+                 BlockNode *brb= mb_x+1<b_width && mb_y+1<b_height ? &s->block[index+b_stride+1] : NULL;
+                 const int b_w= (MB_SIZE >> s->block_max_depth);
+                 uint8_t obmc_edged[b_w*2][b_w*2];
+                 if(pass && (block->type & BLOCK_OPT))
+                     continue;
+                 block->type |= BLOCK_OPT;
+                 backup= *block;
+                 if(!s->me_cache_generation)
+                     memset(s->me_cache, 0, sizeof(s->me_cache));
+                 s->me_cache_generation += 1<<22;
+                 //FIXME precalculate
+                 {
+                     int x, y;
+                     memcpy(obmc_edged, obmc_tab[s->block_max_depth], b_w*b_w*4);
+                     if(mb_x==0)
+                         for(y=0; y<b_w*2; y++)
+                             memset(obmc_edged[y], obmc_edged[y][0] + obmc_edged[y][b_w-1], b_w);
+                     if(mb_x==b_stride-1)
+                         for(y=0; y<b_w*2; y++)
+                             memset(obmc_edged[y]+b_w, obmc_edged[y][b_w] + obmc_edged[y][b_w*2-1], b_w);
+                     if(mb_y==0){
+                         for(x=0; x<b_w*2; x++)
+                             obmc_edged[0][x] += obmc_edged[b_w-1][x];
+                         for(y=1; y<b_w; y++)
+                             memcpy(obmc_edged[y], obmc_edged[0], b_w*2);
+                     }
+                     if(mb_y==b_height-1){
+                         for(x=0; x<b_w*2; x++)
+                             obmc_edged[b_w*2-1][x] += obmc_edged[b_w][x];
+                         for(y=b_w; y<b_w*2-1; y++)
+                             memcpy(obmc_edged[y], obmc_edged[b_w*2-1], b_w*2);
+                     }
+                 }
+                 //skip stuff outside the picture
+                 if(mb_x==0 || mb_y==0 || mb_x==b_width-1 || mb_y==b_height-1){
+                     uint8_t *src= s->  input_picture.data[0];
+                     uint8_t *dst= s->current_picture.data[0];
+                     const int stride= s->current_picture.linesize[0];
+                     const int block_w= MB_SIZE >> s->block_max_depth;
+                     const int sx= block_w*mb_x - block_w/2;
+                     const int sy= block_w*mb_y - block_w/2;
+                     const int w= s->plane[0].width;
+                     const int h= s->plane[0].height;
+                     int y;
+                     for(y=sy; y<0; y++)
+                         memcpy(dst + sx + y*stride, src + sx + y*stride, block_w*2);
+                     for(y=h; y<sy+block_w*2; y++)
+                         memcpy(dst + sx + y*stride, src + sx + y*stride, block_w*2);
+                     if(sx<0){
+                         for(y=sy; y<sy+block_w*2; y++)
+                             memcpy(dst + sx + y*stride, src + sx + y*stride, -sx);
+                     }
+                     if(sx+block_w*2 > w){
+                         for(y=sy; y<sy+block_w*2; y++)
+                             memcpy(dst + w + y*stride, src + w + y*stride, sx+block_w*2 - w);
+                     }
+                 }
+                 // intra(black) = neighbors' contribution to the current block
+                 for(i=0; i<3; i++)
+                     color[i]= get_dc(s, mb_x, mb_y, i);
+                 // get previous score (cannot be cached due to OBMC)
+                 if(pass > 0 && (block->type&BLOCK_INTRA)){
+                     int color0[3]= {block->color[0], block->color[1], block->color[2]};
+                     check_block(s, mb_x, mb_y, color0, 1, *obmc_edged, &best_rd);
+                 }else
+                     check_block_inter(s, mb_x, mb_y, block->mx, block->my, *obmc_edged, &best_rd);
+                 ref_b= *block;
+                 ref_rd= best_rd;
+                 for(ref=0; ref < s->ref_frames; ref++){
+                     int16_t (*mvr)[2]= &s->ref_mvs[ref][index];
+                     if(s->ref_scores[ref][index] > s->ref_scores[ref_b.ref][index]*3/2) //FIXME tune threshold
+                         continue;
+                     block->ref= ref;
+                     best_rd= INT_MAX;
+                     check_block_inter(s, mb_x, mb_y, mvr[0][0], mvr[0][1], *obmc_edged, &best_rd);
+                     check_block_inter(s, mb_x, mb_y, 0, 0, *obmc_edged, &best_rd);
+                     if(tb)
+                         check_block_inter(s, mb_x, mb_y, mvr[-b_stride][0], mvr[-b_stride][1], *obmc_edged, &best_rd);
+                     if(lb)
+                         check_block_inter(s, mb_x, mb_y, mvr[-1][0], mvr[-1][1], *obmc_edged, &best_rd);
+                     if(rb)
+                         check_block_inter(s, mb_x, mb_y, mvr[1][0], mvr[1][1], *obmc_edged, &best_rd);
+                     if(bb)
+                         check_block_inter(s, mb_x, mb_y, mvr[b_stride][0], mvr[b_stride][1], *obmc_edged, &best_rd);
+                     /* fullpel ME */
+                     //FIXME avoid subpel interpolation / round to nearest integer
+                     do{
+                         dia_change=0;
+                         for(i=0; i<FFMAX(s->avctx->dia_size, 1); i++){
+                             for(j=0; j<i; j++){
+                                 dia_change |= check_block_inter(s, mb_x, mb_y, block->mx+4*(i-j), block->my+(4*j), *obmc_edged, &best_rd);
+                                 dia_change |= check_block_inter(s, mb_x, mb_y, block->mx-4*(i-j), block->my-(4*j), *obmc_edged, &best_rd);
+                                 dia_change |= check_block_inter(s, mb_x, mb_y, block->mx+4*(i-j), block->my-(4*j), *obmc_edged, &best_rd);
+                                 dia_change |= check_block_inter(s, mb_x, mb_y, block->mx-4*(i-j), block->my+(4*j), *obmc_edged, &best_rd);
+                             }
+                         }
+                     }while(dia_change);
+                     /* subpel ME */
+                     do{
+                         static const int square[8][2]= {{+1, 0},{-1, 0},{ 0,+1},{ 0,-1},{+1,+1},{-1,-1},{+1,-1},{-1,+1},};
+                         dia_change=0;
+                         for(i=0; i<8; i++)
+                             dia_change |= check_block_inter(s, mb_x, mb_y, block->mx+square[i][0], block->my+square[i][1], *obmc_edged, &best_rd);
+                     }while(dia_change);
+                     //FIXME or try the standard 2 pass qpel or similar
+                     mvr[0][0]= block->mx;
+                     mvr[0][1]= block->my;
+                     if(ref_rd > best_rd){
+                         ref_rd= best_rd;
+                         ref_b= *block;
+                     }
+                 }
+                 best_rd= ref_rd;
+                 *block= ref_b;
+                 check_block(s, mb_x, mb_y, color, 1, *obmc_edged, &best_rd);
+                 //FIXME RD style color selection
+                 if(!same_block(block, &backup)){
+                     if(tb ) tb ->type &= ~BLOCK_OPT;
+                     if(lb ) lb ->type &= ~BLOCK_OPT;
+                     if(rb ) rb ->type &= ~BLOCK_OPT;
+                     if(bb ) bb ->type &= ~BLOCK_OPT;
+                     if(tlb) tlb->type &= ~BLOCK_OPT;
+                     if(trb) trb->type &= ~BLOCK_OPT;
+                     if(blb) blb->type &= ~BLOCK_OPT;
+                     if(brb) brb->type &= ~BLOCK_OPT;
+                     change ++;
+                 }
+             }
+         }
+         av_log(s->avctx, AV_LOG_ERROR, "pass:%d changed:%d\n", pass, change);
+         if(!change)
+             break;
+     }
+     if(s->block_max_depth == 1){
+         int change= 0;
+         for(mb_y= 0; mb_y<b_height; mb_y+=2){
+             for(mb_x= 0; mb_x<b_width; mb_x+=2){
+                 int i;
+                 int best_rd, init_rd;
+                 const int index= mb_x + mb_y * b_stride;
+                 BlockNode *b[4];
+                 b[0]= &s->block[index];
+                 b[1]= b[0]+1;
+                 b[2]= b[0]+b_stride;
+                 b[3]= b[2]+1;
+                 if(same_block(b[0], b[1]) &&
+                    same_block(b[0], b[2]) &&
+                    same_block(b[0], b[3]))
+                     continue;
+                 if(!s->me_cache_generation)
+                     memset(s->me_cache, 0, sizeof(s->me_cache));
+                 s->me_cache_generation += 1<<22;
+                 init_rd= best_rd= get_4block_rd(s, mb_x, mb_y, 0);
+                 //FIXME more multiref search?
+                 check_4block_inter(s, mb_x, mb_y,
+                                    (b[0]->mx + b[1]->mx + b[2]->mx + b[3]->mx + 2) >> 2,
+                                    (b[0]->my + b[1]->my + b[2]->my + b[3]->my + 2) >> 2, 0, &best_rd);
+                 for(i=0; i<4; i++)
+                     if(!(b[i]->type&BLOCK_INTRA))
+                         check_4block_inter(s, mb_x, mb_y, b[i]->mx, b[i]->my, b[i]->ref, &best_rd);
+                 if(init_rd != best_rd)
+                     change++;
+             }
+         }
+         av_log(s->avctx, AV_LOG_ERROR, "pass:4mv changed:%d\n", change*4);
+     }
+ }
+ static void encode_blocks(SnowContext *s, int search){
+     int x, y;
+     int w= s->b_width;
+     int h= s->b_height;
+     if(s->avctx->me_method == ME_ITER && !s->keyframe && search)
+         iterative_me(s);
+     for(y=0; y<h; y++){
+         if(s->c.bytestream_end - s->c.bytestream < w*MB_SIZE*MB_SIZE*3){ //FIXME nicer limit
+             av_log(s->avctx, AV_LOG_ERROR, "encoded frame too large\n");
+             return;
+         }
+         for(x=0; x<w; x++){
+             if(s->avctx->me_method == ME_ITER || !search)
+                 encode_q_branch2(s, 0, x, y);
+             else
+                 encode_q_branch (s, 0, x, y);
+         }
+     }
+ }
+ static void quantize(SnowContext *s, SubBand *b, IDWTELEM *dst, DWTELEM *src, int stride, int bias){
+     const int w= b->width;
+     const int h= b->height;
+     const int qlog= av_clip(s->qlog + b->qlog, 0, QROOT*16);
+     const int qmul= qexp[qlog&(QROOT-1)]<<((qlog>>QSHIFT) + ENCODER_EXTRA_BITS);
+     int x,y, thres1, thres2;
+     if(s->qlog == LOSSLESS_QLOG){
+         for(y=0; y<h; y++)
+             for(x=0; x<w; x++)
+                 dst[x + y*stride]= src[x + y*stride];
+         return;
+     }
+     bias= bias ? 0 : (3*qmul)>>3;
+     thres1= ((qmul - bias)>>QEXPSHIFT) - 1;
+     thres2= 2*thres1;
+     if(!bias){
+         for(y=0; y<h; y++){
+             for(x=0; x<w; x++){
+                 int i= src[x + y*stride];
+                 if((unsigned)(i+thres1) > thres2){
+                     if(i>=0){
+                         i<<= QEXPSHIFT;
+                         i/= qmul; //FIXME optimize
+                         dst[x + y*stride]=  i;
+                     }else{
+                         i= -i;
+                         i<<= QEXPSHIFT;
+                         i/= qmul; //FIXME optimize
+                         dst[x + y*stride]= -i;
+                     }
+                 }else
+                     dst[x + y*stride]= 0;
+             }
+         }
+     }else{
+         for(y=0; y<h; y++){
+             for(x=0; x<w; x++){
+                 int i= src[x + y*stride];
+                 if((unsigned)(i+thres1) > thres2){
+                     if(i>=0){
+                         i<<= QEXPSHIFT;
+                         i= (i + bias) / qmul; //FIXME optimize
+                         dst[x + y*stride]=  i;
+                     }else{
+                         i= -i;
+                         i<<= QEXPSHIFT;
+                         i= (i + bias) / qmul; //FIXME optimize
+                         dst[x + y*stride]= -i;
+                     }
+                 }else
+                     dst[x + y*stride]= 0;
+             }
+         }
+     }
+ }
+ static void dequantize(SnowContext *s, SubBand *b, IDWTELEM *src, int stride){
+     const int w= b->width;
+     const int h= b->height;
+     const int qlog= av_clip(s->qlog + b->qlog, 0, QROOT*16);
+     const int qmul= qexp[qlog&(QROOT-1)]<<(qlog>>QSHIFT);
+     const int qadd= (s->qbias*qmul)>>QBIAS_SHIFT;
+     int x,y;
+     if(s->qlog == LOSSLESS_QLOG) return;
+     for(y=0; y<h; y++){
+         for(x=0; x<w; x++){
+             int i= src[x + y*stride];
+             if(i<0){
+                 src[x + y*stride]= -((-i*qmul + qadd)>>(QEXPSHIFT)); //FIXME try different bias
+             }else if(i>0){
+                 src[x + y*stride]=  (( i*qmul + qadd)>>(QEXPSHIFT));
+             }
+         }
+     }
+ }
+ static void decorrelate(SnowContext *s, SubBand *b, IDWTELEM *src, int stride, int inverse, int use_median){
+     const int w= b->width;
+     const int h= b->height;
+     int x,y;
+     for(y=h-1; y>=0; y--){
+         for(x=w-1; x>=0; x--){
+             int i= x + y*stride;
+             if(x){
+                 if(use_median){
+                     if(y && x+1<w) src[i] -= mid_pred(src[i - 1], src[i - stride], src[i - stride + 1]);
+                     else  src[i] -= src[i - 1];
+                 }else{
+                     if(y) src[i] -= mid_pred(src[i - 1], src[i - stride], src[i - 1] + src[i - stride] - src[i - 1 - stride]);
+                     else  src[i] -= src[i - 1];
+                 }
+             }else{
+                 if(y) src[i] -= src[i - stride];
+             }
+         }
+     }
+ }
+ static void correlate(SnowContext *s, SubBand *b, IDWTELEM *src, int stride, int inverse, int use_median){
+     const int w= b->width;
+     const int h= b->height;
+     int x,y;
+     for(y=0; y<h; y++){
+         for(x=0; x<w; x++){
+             int i= x + y*stride;
+             if(x){
+                 if(use_median){
+                     if(y && x+1<w) src[i] += mid_pred(src[i - 1], src[i - stride], src[i - stride + 1]);
+                     else  src[i] += src[i - 1];
+                 }else{
+                     if(y) src[i] += mid_pred(src[i - 1], src[i - stride], src[i - 1] + src[i - stride] - src[i - 1 - stride]);
+                     else  src[i] += src[i - 1];
+                 }
+             }else{
+                 if(y) src[i] += src[i - stride];
+             }
+         }
+     }
+ }
+ static void encode_qlogs(SnowContext *s){
+     int plane_index, level, orientation;
+     for(plane_index=0; plane_index<2; plane_index++){
+         for(level=0; level<s->spatial_decomposition_count; level++){
+             for(orientation=level ? 1:0; orientation<4; orientation++){
+                 if(orientation==2) continue;
+                 put_symbol(&s->c, s->header_state, s->plane[plane_index].band[level][orientation].qlog, 1);
+             }
+         }
+     }
+ }
+ static void encode_header(SnowContext *s){
+     int plane_index, i;
+     uint8_t kstate[32];
+     memset(kstate, MID_STATE, sizeof(kstate));
+     put_rac(&s->c, kstate, s->keyframe);
+     if(s->keyframe || s->always_reset){
+         ff_snow_reset_contexts(s);
+         s->last_spatial_decomposition_type=
+         s->last_qlog=
+         s->last_qbias=
+         s->last_mv_scale=
+         s->last_block_max_depth= 0;
+         for(plane_index=0; plane_index<2; plane_index++){
+             Plane *p= &s->plane[plane_index];
+             p->last_htaps=0;
+             p->last_diag_mc=0;
+             memset(p->last_hcoeff, 0, sizeof(p->last_hcoeff));
+         }
+     }
+     if(s->keyframe){
+         put_symbol(&s->c, s->header_state, s->version, 0);
+         put_rac(&s->c, s->header_state, s->always_reset);
+         put_symbol(&s->c, s->header_state, s->temporal_decomposition_type, 0);
+         put_symbol(&s->c, s->header_state, s->temporal_decomposition_count, 0);
+         put_symbol(&s->c, s->header_state, s->spatial_decomposition_count, 0);
+         put_symbol(&s->c, s->header_state, s->colorspace_type, 0);
+         put_symbol(&s->c, s->header_state, s->chroma_h_shift, 0);
+         put_symbol(&s->c, s->header_state, s->chroma_v_shift, 0);
+         put_rac(&s->c, s->header_state, s->spatial_scalability);
+ //        put_rac(&s->c, s->header_state, s->rate_scalability);
+         put_symbol(&s->c, s->header_state, s->max_ref_frames-1, 0);
+         encode_qlogs(s);
+     }
+     if(!s->keyframe){
+         int update_mc=0;
+         for(plane_index=0; plane_index<2; plane_index++){
+             Plane *p= &s->plane[plane_index];
+             update_mc |= p->last_htaps   != p->htaps;
+             update_mc |= p->last_diag_mc != p->diag_mc;
+             update_mc |= !!memcmp(p->last_hcoeff, p->hcoeff, sizeof(p->hcoeff));
+         }
+         put_rac(&s->c, s->header_state, update_mc);
+         if(update_mc){
+             for(plane_index=0; plane_index<2; plane_index++){
+                 Plane *p= &s->plane[plane_index];
+                 put_rac(&s->c, s->header_state, p->diag_mc);
+                 put_symbol(&s->c, s->header_state, p->htaps/2-1, 0);
+                 for(i= p->htaps/2; i; i--)
+                     put_symbol(&s->c, s->header_state, FFABS(p->hcoeff[i]), 0);
+             }
+         }
+         if(s->last_spatial_decomposition_count != s->spatial_decomposition_count){
+             put_rac(&s->c, s->header_state, 1);
+             put_symbol(&s->c, s->header_state, s->spatial_decomposition_count, 0);
+             encode_qlogs(s);
+         }else
+             put_rac(&s->c, s->header_state, 0);
+     }
+     put_symbol(&s->c, s->header_state, s->spatial_decomposition_type - s->last_spatial_decomposition_type, 1);
+     put_symbol(&s->c, s->header_state, s->qlog            - s->last_qlog    , 1);
+     put_symbol(&s->c, s->header_state, s->mv_scale        - s->last_mv_scale, 1);
+     put_symbol(&s->c, s->header_state, s->qbias           - s->last_qbias   , 1);
+     put_symbol(&s->c, s->header_state, s->block_max_depth - s->last_block_max_depth, 1);
+ }
+ static void update_last_header_values(SnowContext *s){
+     int plane_index;
+     if(!s->keyframe){
+         for(plane_index=0; plane_index<2; plane_index++){
+             Plane *p= &s->plane[plane_index];
+             p->last_diag_mc= p->diag_mc;
+             p->last_htaps  = p->htaps;
+             memcpy(p->last_hcoeff, p->hcoeff, sizeof(p->hcoeff));
+         }
+     }
+     s->last_spatial_decomposition_type  = s->spatial_decomposition_type;
+     s->last_qlog                        = s->qlog;
+     s->last_qbias                       = s->qbias;
+     s->last_mv_scale                    = s->mv_scale;
+     s->last_block_max_depth             = s->block_max_depth;
+     s->last_spatial_decomposition_count = s->spatial_decomposition_count;
+ }
+ static int qscale2qlog(int qscale){
+     return rint(QROOT*log(qscale / (float)FF_QP2LAMBDA)/log(2))
+            + 61*QROOT/8; ///< 64 > 60
+ }
+ static int ratecontrol_1pass(SnowContext *s, AVFrame *pict)
+ {
+     /* Estimate the frame's complexity as a sum of weighted dwt coefficients.
+      * FIXME we know exact mv bits at this point,
+      * but ratecontrol isn't set up to include them. */
+     uint32_t coef_sum= 0;
+     int level, orientation, delta_qlog;
+     for(level=0; level<s->spatial_decomposition_count; level++){
+         for(orientation=level ? 1 : 0; orientation<4; orientation++){
+             SubBand *b= &s->plane[0].band[level][orientation];
+             IDWTELEM *buf= b->ibuf;
+             const int w= b->width;
+             const int h= b->height;
+             const int stride= b->stride;
+             const int qlog= av_clip(2*QROOT + b->qlog, 0, QROOT*16);
+             const int qmul= qexp[qlog&(QROOT-1)]<<(qlog>>QSHIFT);
+             const int qdiv= (1<<16)/qmul;
+             int x, y;
+             //FIXME this is ugly
+             for(y=0; y<h; y++)
+                 for(x=0; x<w; x++)
+                     buf[x+y*stride]= b->buf[x+y*stride];
+             if(orientation==0)
+                 decorrelate(s, b, buf, stride, 1, 0);
+             for(y=0; y<h; y++)
+                 for(x=0; x<w; x++)
+                     coef_sum+= abs(buf[x+y*stride]) * qdiv >> 16;
+         }
+     }
+     /* ugly, ratecontrol just takes a sqrt again */
+     coef_sum = (uint64_t)coef_sum * coef_sum >> 16;
+     assert(coef_sum < INT_MAX);
+     if(pict->pict_type == AV_PICTURE_TYPE_I){
+         s->m.current_picture.mb_var_sum= coef_sum;
+         s->m.current_picture.mc_mb_var_sum= 0;
+     }else{
+         s->m.current_picture.mc_mb_var_sum= coef_sum;
+         s->m.current_picture.mb_var_sum= 0;
+     }
+     pict->quality= ff_rate_estimate_qscale(&s->m, 1);
+     if (pict->quality < 0)
+         return INT_MIN;
+     s->lambda= pict->quality * 3/2;
+     delta_qlog= qscale2qlog(pict->quality) - s->qlog;
+     s->qlog+= delta_qlog;
+     return delta_qlog;
+ }
+ static void calculate_visual_weight(SnowContext *s, Plane *p){
+     int width = p->width;
+     int height= p->height;
+     int level, orientation, x, y;
+     for(level=0; level<s->spatial_decomposition_count; level++){
+         for(orientation=level ? 1 : 0; orientation<4; orientation++){
+             SubBand *b= &p->band[level][orientation];
+             IDWTELEM *ibuf= b->ibuf;
+             int64_t error=0;
+             memset(s->spatial_idwt_buffer, 0, sizeof(*s->spatial_idwt_buffer)*width*height);
+             ibuf[b->width/2 + b->height/2*b->stride]= 256*16;
+             ff_spatial_idwt(s->spatial_idwt_buffer, width, height, width, s->spatial_decomposition_type, s->spatial_decomposition_count);
+             for(y=0; y<height; y++){
+                 for(x=0; x<width; x++){
+                     int64_t d= s->spatial_idwt_buffer[x + y*width]*16;
+                     error += d*d;
+                 }
+             }
+             b->qlog= (int)(log(352256.0/sqrt(error)) / log(pow(2.0, 1.0/QROOT))+0.5);
+         }
+     }
+ }
+ static int encode_frame(AVCodecContext *avctx, unsigned char *buf, int buf_size, void *data){
+     SnowContext *s = avctx->priv_data;
+     RangeCoder * const c= &s->c;
+     AVFrame *pict = data;
+     const int width= s->avctx->width;
+     const int height= s->avctx->height;
+     int level, orientation, plane_index, i, y;
+     uint8_t rc_header_bak[sizeof(s->header_state)];
+     uint8_t rc_block_bak[sizeof(s->block_state)];
+     ff_init_range_encoder(c, buf, buf_size);
+     ff_build_rac_states(c, 0.05*(1LL<<32), 256-8);
+     for(i=0; i<3; i++){
+         int shift= !!i;
+         for(y=0; y<(height>>shift); y++)
+             memcpy(&s->input_picture.data[i][y * s->input_picture.linesize[i]],
+                    &pict->data[i][y * pict->linesize[i]],
+                    width>>shift);
+     }
+     s->new_picture = *pict;
+     s->m.picture_number= avctx->frame_number;
+     if(avctx->flags&CODEC_FLAG_PASS2){
+         s->m.pict_type =
+         pict->pict_type= s->m.rc_context.entry[avctx->frame_number].new_pict_type;
+         s->keyframe= pict->pict_type==AV_PICTURE_TYPE_I;
+         if(!(avctx->flags&CODEC_FLAG_QSCALE)) {
+             pict->quality= ff_rate_estimate_qscale(&s->m, 0);
+             if (pict->quality < 0)
+                 return -1;
+         }
+     }else{
+         s->keyframe= avctx->gop_size==0 || avctx->frame_number % avctx->gop_size == 0;
+         s->m.pict_type=
+         pict->pict_type= s->keyframe ? AV_PICTURE_TYPE_I : AV_PICTURE_TYPE_P;
+     }
+     if(s->pass1_rc && avctx->frame_number == 0)
+         pict->quality= 2*FF_QP2LAMBDA;
+     if(pict->quality){
+         s->qlog= qscale2qlog(pict->quality);
+         s->lambda = pict->quality * 3/2;
+     }
+     if(s->qlog < 0 || (!pict->quality && (avctx->flags & CODEC_FLAG_QSCALE))){
+         s->qlog= LOSSLESS_QLOG;
+         s->lambda = 0;
+     }//else keep previous frame's qlog until after motion estimation
+     ff_snow_frame_start(s);
+     s->m.current_picture_ptr= &s->m.current_picture;
+     s->m.last_picture.f.pts = s->m.current_picture.f.pts;
+     s->m.current_picture.f.pts = pict->pts;
+     if(pict->pict_type == AV_PICTURE_TYPE_P){
+         int block_width = (width +15)>>4;
+         int block_height= (height+15)>>4;
+         int stride= s->current_picture.linesize[0];
+         assert(s->current_picture.data[0]);
+         assert(s->last_picture[0].data[0]);
+         s->m.avctx= s->avctx;
+         s->m.current_picture.f.data[0] = s->current_picture.data[0];
+         s->m.   last_picture.f.data[0] = s->last_picture[0].data[0];
+         s->m.    new_picture.f.data[0] = s->  input_picture.data[0];
+         s->m.   last_picture_ptr= &s->m.   last_picture;
+         s->m.linesize=
+         s->m.   last_picture.f.linesize[0] =
+         s->m.    new_picture.f.linesize[0] =
+         s->m.current_picture.f.linesize[0] = stride;
+         s->m.uvlinesize= s->current_picture.linesize[1];
+         s->m.width = width;
+         s->m.height= height;
+         s->m.mb_width = block_width;
+         s->m.mb_height= block_height;
+         s->m.mb_stride=   s->m.mb_width+1;
+         s->m.b8_stride= 2*s->m.mb_width+1;
+         s->m.f_code=1;
+         s->m.pict_type= pict->pict_type;
+         s->m.me_method= s->avctx->me_method;
+         s->m.me.scene_change_score=0;
+         s->m.flags= s->avctx->flags;
+         s->m.quarter_sample= (s->avctx->flags & CODEC_FLAG_QPEL)!=0;
+         s->m.out_format= FMT_H263;
+         s->m.unrestricted_mv= 1;
+         s->m.lambda = s->lambda;
+         s->m.qscale= (s->m.lambda*139 + FF_LAMBDA_SCALE*64) >> (FF_LAMBDA_SHIFT + 7);
+         s->lambda2= s->m.lambda2= (s->m.lambda*s->m.lambda + FF_LAMBDA_SCALE/2) >> FF_LAMBDA_SHIFT;
+         s->m.dsp= s->dsp; //move
+         ff_init_me(&s->m);
+         s->dsp= s->m.dsp;
+     }
+     if(s->pass1_rc){
+         memcpy(rc_header_bak, s->header_state, sizeof(s->header_state));
+         memcpy(rc_block_bak, s->block_state, sizeof(s->block_state));
+     }
+ redo_frame:
+     if(pict->pict_type == AV_PICTURE_TYPE_I)
+         s->spatial_decomposition_count= 5;
+     else
+         s->spatial_decomposition_count= 5;
+     s->m.pict_type = pict->pict_type;
+     s->qbias= pict->pict_type == AV_PICTURE_TYPE_P ? 2 : 0;
+     ff_snow_common_init_after_header(avctx);
+     if(s->last_spatial_decomposition_count != s->spatial_decomposition_count){
+         for(plane_index=0; plane_index<3; plane_index++){
+             calculate_visual_weight(s, &s->plane[plane_index]);
+         }
+     }
+     encode_header(s);
+     s->m.misc_bits = 8*(s->c.bytestream - s->c.bytestream_start);
+     encode_blocks(s, 1);
+     s->m.mv_bits = 8*(s->c.bytestream - s->c.bytestream_start) - s->m.misc_bits;
+     for(plane_index=0; plane_index<3; plane_index++){
+         Plane *p= &s->plane[plane_index];
+         int w= p->width;
+         int h= p->height;
+         int x, y;
+ //        int bits= put_bits_count(&s->c.pb);
+         if (!s->memc_only) {
+             //FIXME optimize
+             if(pict->data[plane_index]) //FIXME gray hack
+                 for(y=0; y<h; y++){
+                     for(x=0; x<w; x++){
+                         s->spatial_idwt_buffer[y*w + x]= pict->data[plane_index][y*pict->linesize[plane_index] + x]<<FRAC_BITS;
+                     }
+                 }
+             predict_plane(s, s->spatial_idwt_buffer, plane_index, 0);
+             if(   plane_index==0
+                && pict->pict_type == AV_PICTURE_TYPE_P
+                && !(avctx->flags&CODEC_FLAG_PASS2)
+                && s->m.me.scene_change_score > s->avctx->scenechange_threshold){
+                 ff_init_range_encoder(c, buf, buf_size);
+                 ff_build_rac_states(c, 0.05*(1LL<<32), 256-8);
+                 pict->pict_type= AV_PICTURE_TYPE_I;
+                 s->keyframe=1;
+                 s->current_picture.key_frame=1;
+                 goto redo_frame;
+             }
+             if(s->qlog == LOSSLESS_QLOG){
+                 for(y=0; y<h; y++){
+                     for(x=0; x<w; x++){
+                         s->spatial_dwt_buffer[y*w + x]= (s->spatial_idwt_buffer[y*w + x] + (1<<(FRAC_BITS-1))-1)>>FRAC_BITS;
+                     }
+                 }
+             }else{
+                 for(y=0; y<h; y++){
+                     for(x=0; x<w; x++){
+                         s->spatial_dwt_buffer[y*w + x]=s->spatial_idwt_buffer[y*w + x]<<ENCODER_EXTRA_BITS;
+                     }
+                 }
+             }
+             /*  if(QUANTIZE2)
+                 dwt_quantize(s, p, s->spatial_dwt_buffer, w, h, w, s->spatial_decomposition_type);
+             else*/
+                 ff_spatial_dwt(s->spatial_dwt_buffer, w, h, w, s->spatial_decomposition_type, s->spatial_decomposition_count);
+             if(s->pass1_rc && plane_index==0){
+                 int delta_qlog = ratecontrol_1pass(s, pict);
+                 if (delta_qlog <= INT_MIN)
+                     return -1;
+                 if(delta_qlog){
+                     //reordering qlog in the bitstream would eliminate this reset
+                     ff_init_range_encoder(c, buf, buf_size);
+                     memcpy(s->header_state, rc_header_bak, sizeof(s->header_state));
+                     memcpy(s->block_state, rc_block_bak, sizeof(s->block_state));
+                     encode_header(s);
+                     encode_blocks(s, 0);
+                 }
+             }
+             for(level=0; level<s->spatial_decomposition_count; level++){
+                 for(orientation=level ? 1 : 0; orientation<4; orientation++){
+                     SubBand *b= &p->band[level][orientation];
+                     if(!QUANTIZE2)
+                         quantize(s, b, b->ibuf, b->buf, b->stride, s->qbias);
+                     if(orientation==0)
+                         decorrelate(s, b, b->ibuf, b->stride, pict->pict_type == AV_PICTURE_TYPE_P, 0);
+                     encode_subband(s, b, b->ibuf, b->parent ? b->parent->ibuf : NULL, b->stride, orientation);
+                     assert(b->parent==NULL || b->parent->stride == b->stride*2);
+                     if(orientation==0)
+                         correlate(s, b, b->ibuf, b->stride, 1, 0);
+                 }
+             }
+             for(level=0; level<s->spatial_decomposition_count; level++){
+                 for(orientation=level ? 1 : 0; orientation<4; orientation++){
+                     SubBand *b= &p->band[level][orientation];
+                     dequantize(s, b, b->ibuf, b->stride);
+                 }
+             }
+             ff_spatial_idwt(s->spatial_idwt_buffer, w, h, w, s->spatial_decomposition_type, s->spatial_decomposition_count);
+             if(s->qlog == LOSSLESS_QLOG){
+                 for(y=0; y<h; y++){
+                     for(x=0; x<w; x++){
+                         s->spatial_idwt_buffer[y*w + x]<<=FRAC_BITS;
+                     }
+                 }
+             }
+             predict_plane(s, s->spatial_idwt_buffer, plane_index, 1);
+         }else{
+             //ME/MC only
+             if(pict->pict_type == AV_PICTURE_TYPE_I){
+                 for(y=0; y<h; y++){
+                     for(x=0; x<w; x++){
+                         s->current_picture.data[plane_index][y*s->current_picture.linesize[plane_index] + x]=
+                             pict->data[plane_index][y*pict->linesize[plane_index] + x];
+                     }
+                 }
+             }else{
+                 memset(s->spatial_idwt_buffer, 0, sizeof(IDWTELEM)*w*h);
+                 predict_plane(s, s->spatial_idwt_buffer, plane_index, 1);
+             }
+         }
+         if(s->avctx->flags&CODEC_FLAG_PSNR){
+             int64_t error= 0;
+             if(pict->data[plane_index]) //FIXME gray hack
+                 for(y=0; y<h; y++){
+                     for(x=0; x<w; x++){
+                         int d= s->current_picture.data[plane_index][y*s->current_picture.linesize[plane_index] + x] - pict->data[plane_index][y*pict->linesize[plane_index] + x];
+                         error += d*d;
+                     }
+                 }
+             s->avctx->error[plane_index] += error;
+             s->current_picture.error[plane_index] = error;
+         }
+     }
+     update_last_header_values(s);
+     ff_snow_release_buffer(avctx);
+     s->current_picture.coded_picture_number = avctx->frame_number;
+     s->current_picture.pict_type = pict->pict_type;
+     s->current_picture.quality = pict->quality;
+     s->m.frame_bits = 8*(s->c.bytestream - s->c.bytestream_start);
+     s->m.p_tex_bits = s->m.frame_bits - s->m.misc_bits - s->m.mv_bits;
+     s->m.current_picture.f.display_picture_number =
+     s->m.current_picture.f.coded_picture_number   = avctx->frame_number;
+     s->m.current_picture.f.quality                = pict->quality;
+     s->m.total_bits += 8*(s->c.bytestream - s->c.bytestream_start);
+     if(s->pass1_rc)
+         if (ff_rate_estimate_qscale(&s->m, 0) < 0)
+             return -1;
+     if(avctx->flags&CODEC_FLAG_PASS1)
+         ff_write_pass1_stats(&s->m);
+     s->m.last_pict_type = s->m.pict_type;
+     avctx->frame_bits = s->m.frame_bits;
+     avctx->mv_bits = s->m.mv_bits;
+     avctx->misc_bits = s->m.misc_bits;
+     avctx->p_tex_bits = s->m.p_tex_bits;
+     emms_c();
+     return ff_rac_terminate(c);
+ }
+ static av_cold int encode_end(AVCodecContext *avctx)
+ {
+     SnowContext *s = avctx->priv_data;
+     ff_snow_common_end(s);
+     if (s->input_picture.data[0])
+         avctx->release_buffer(avctx, &s->input_picture);
+     av_free(avctx->stats_out);
+     return 0;
+ }
+ #define OFFSET(x) offsetof(SnowContext, x)
+ #define VE AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM
+ static const AVOption options[] = {
+     { "memc_only",      "Only do ME/MC (I frames -> ref, P frame -> ME+MC).",   OFFSET(memc_only), AV_OPT_TYPE_INT, { 0 }, 0, 1, VE },
+     { NULL },
+ };
+ static const AVClass snowenc_class = {
+     .class_name = "snow encoder",
+     .item_name  = av_default_item_name,
+     .option     = options,
+     .version    = LIBAVUTIL_VERSION_INT,
+ };
+ AVCodec ff_snow_encoder = {
+     .name           = "snow",
+     .type           = AVMEDIA_TYPE_VIDEO,
+     .id             = CODEC_ID_SNOW,
+     .priv_data_size = sizeof(SnowContext),
+     .init           = encode_init,
+     .encode         = encode_frame,
+     .close          = encode_end,
+     .long_name = NULL_IF_CONFIG_SMALL("Snow"),
+     .priv_class     = &snowenc_class,
+ };
+ #endif
++
++
++#ifdef TEST
++#undef malloc
++#undef free
++#undef printf
++
++#include "libavutil/lfg.h"
++#include "libavutil/mathematics.h"
++
++int main(void){
++    int width=256;
++    int height=256;
++    int buffer[2][width*height];
++    SnowContext s;
++    int i;
++    AVLFG prng;
++    s.spatial_decomposition_count=6;
++    s.spatial_decomposition_type=1;
++
++    av_lfg_init(&prng, 1);
++
++    printf("testing 5/3 DWT\n");
++    for(i=0; i<width*height; i++)
++        buffer[0][i] = buffer[1][i] = av_lfg_get(&prng) % 54321 - 12345;
++
++    ff_spatial_dwt(buffer[0], width, height, width, s.spatial_decomposition_type, s.spatial_decomposition_count);
++    ff_spatial_idwt(buffer[0], width, height, width, s.spatial_decomposition_type, s.spatial_decomposition_count);
++
++    for(i=0; i<width*height; i++)
++        if(buffer[0][i]!= buffer[1][i]) printf("fsck: %6d %12d %7d\n",i, buffer[0][i], buffer[1][i]);
++
++    printf("testing 9/7 DWT\n");
++    s.spatial_decomposition_type=0;
++    for(i=0; i<width*height; i++)
++        buffer[0][i] = buffer[1][i] = av_lfg_get(&prng) % 54321 - 12345;
++
++    ff_spatial_dwt(buffer[0], width, height, width, s.spatial_decomposition_type, s.spatial_decomposition_count);
++    ff_spatial_idwt(buffer[0], width, height, width, s.spatial_decomposition_type, s.spatial_decomposition_count);
++
++    for(i=0; i<width*height; i++)
++        if(FFABS(buffer[0][i] - buffer[1][i])>20) printf("fsck: %6d %12d %7d\n",i, buffer[0][i], buffer[1][i]);
++
++    {
++    int level, orientation, x, y;
++    int64_t errors[8][4];
++    int64_t g=0;
++
++        memset(errors, 0, sizeof(errors));
++        s.spatial_decomposition_count=3;
++        s.spatial_decomposition_type=0;
++        for(level=0; level<s.spatial_decomposition_count; level++){
++            for(orientation=level ? 1 : 0; orientation<4; orientation++){
++                int w= width  >> (s.spatial_decomposition_count-level);
++                int h= height >> (s.spatial_decomposition_count-level);
++                int stride= width  << (s.spatial_decomposition_count-level);
++                DWTELEM *buf= buffer[0];
++                int64_t error=0;
++
++                if(orientation&1) buf+=w;
++                if(orientation>1) buf+=stride>>1;
++
++                memset(buffer[0], 0, sizeof(int)*width*height);
++                buf[w/2 + h/2*stride]= 256*256;
++                ff_spatial_idwt(buffer[0], width, height, width, s.spatial_decomposition_type, s.spatial_decomposition_count);
++                for(y=0; y<height; y++){
++                    for(x=0; x<width; x++){
++                        int64_t d= buffer[0][x + y*width];
++                        error += d*d;
++                        if(FFABS(width/2-x)<9 && FFABS(height/2-y)<9 && level==2) printf("%8"PRId64" ", d);
++                    }
++                    if(FFABS(height/2-y)<9 && level==2) printf("\n");
++                }
++                error= (int)(sqrt(error)+0.5);
++                errors[level][orientation]= error;
++                if(g) g=av_gcd(g, error);
++                else g= error;
++            }
++        }
++        printf("static int const visual_weight[][4]={\n");
++        for(level=0; level<s.spatial_decomposition_count; level++){
++            printf("  {");
++            for(orientation=0; orientation<4; orientation++){
++                printf("%8"PRId64",", errors[level][orientation]/g);
++            }
++            printf("},\n");
++        }
++        printf("};\n");
++        {
++            int level=2;
++            int w= width  >> (s.spatial_decomposition_count-level);
++            //int h= height >> (s.spatial_decomposition_count-level);
++            int stride= width  << (s.spatial_decomposition_count-level);
++            DWTELEM *buf= buffer[0];
++            int64_t error=0;
++
++            buf+=w;
++            buf+=stride>>1;
++
++            memset(buffer[0], 0, sizeof(int)*width*height);
++            for(y=0; y<height; y++){
++                for(x=0; x<width; x++){
++                    int tab[4]={0,2,3,1};
++                    buffer[0][x+width*y]= 256*256*tab[(x&1) + 2*(y&1)];
++                }
++            }
++            ff_spatial_dwt(buffer[0], width, height, width, s.spatial_decomposition_type, s.spatial_decomposition_count);
++            for(y=0; y<height; y++){
++                for(x=0; x<width; x++){
++                    int64_t d= buffer[0][x + y*width];
++                    error += d*d;
++                    if(FFABS(width/2-x)<9 && FFABS(height/2-y)<9) printf("%8"PRId64" ", d);
++                }
++                if(FFABS(height/2-y)<9) printf("\n");
++            }
++        }
++
++    }
++    return 0;
++}
++#endif /* TEST */
Simple merge
@@@ -158,10 -159,13 +159,13 @@@ void avcodec_align_dimensions2(AVCodecC
      case PIX_FMT_YUV444P9BE:
      case PIX_FMT_YUV444P10LE:
      case PIX_FMT_YUV444P10BE:
-     case PIX_FMT_GBR24P:
+     case PIX_FMT_GBRP9LE:
+     case PIX_FMT_GBRP9BE:
+     case PIX_FMT_GBRP10LE:
+     case PIX_FMT_GBRP10BE:
          w_align= 16; //FIXME check for non mpeg style codecs and use less alignment
          h_align= 16;
 -        if(s->codec_id == CODEC_ID_MPEG2VIDEO || s->codec_id == CODEC_ID_MJPEG || s->codec_id == CODEC_ID_AMV || s->codec_id == CODEC_ID_THP || s->codec_id == CODEC_ID_H264)
 +        if(s->codec_id == CODEC_ID_MPEG2VIDEO || s->codec_id == CODEC_ID_MJPEG || s->codec_id == CODEC_ID_AMV || s->codec_id == CODEC_ID_THP || s->codec_id == CODEC_ID_H264 || s->codec_id == CODEC_ID_PRORES)
              h_align= 32; // interlaced is rounded up to 2 MBs
          break;
      case PIX_FMT_YUV411P:
Simple merge
   */
  
  #define LIBAVUTIL_VERSION_MAJOR 51
- #define LIBAVUTIL_VERSION_MINOR 28
 -#define LIBAVUTIL_VERSION_MINOR 19
++#define LIBAVUTIL_VERSION_MINOR 29
  #define LIBAVUTIL_VERSION_MICRO  0
  
  #define LIBAVUTIL_VERSION_INT   AV_VERSION_INT(LIBAVUTIL_VERSION_MAJOR, \
@@@ -560,32 -578,8 +578,34 @@@ const AVPixFmtDescriptor av_pix_fmt_des
              {0,5,3,0,15},       /* G */
              {0,5,5,0,15},       /* B */
          },
+         .flags = PIX_FMT_RGB,
      },
-         .flags = PIX_FMT_BE,
 +    [PIX_FMT_RGBA64BE] = {
 +        .name = "rgba64be",
 +        .nb_components= 4,
 +        .log2_chroma_w= 0,
 +        .log2_chroma_h= 0,
 +        .comp = {
 +            {0,5,1,0,15},       /* R */
 +            {0,5,3,0,15},       /* G */
 +            {0,5,5,0,15},       /* B */
 +            {0,5,7,0,15},       /* A */
 +        },
++        .flags = PIX_FMT_RGB | PIX_FMT_BE,
 +    },
 +    [PIX_FMT_RGBA64LE] = {
 +        .name = "rgba64le",
 +        .nb_components= 4,
 +        .log2_chroma_w= 0,
 +        .log2_chroma_h= 0,
 +        .comp = {
 +            {0,5,1,0,15},       /* R */
 +            {0,5,3,0,15},       /* G */
 +            {0,5,5,0,15},       /* B */
 +            {0,5,7,0,15},       /* B */
 +        },
++        .flags = PIX_FMT_RGB,
 +    },
      [PIX_FMT_RGB565BE] = {
          .name = "rgb565be",
          .nb_components= 3,
              {0,5,3,0,15},       /* G */
              {0,5,5,0,15},       /* R */
          },
+         .flags = PIX_FMT_RGB,
      },
 +    [PIX_FMT_BGRA64BE] = {
 +        .name = "bgra64be",
 +        .nb_components= 4,
 +        .log2_chroma_w= 0,
 +        .log2_chroma_h= 0,
 +        .comp = {
 +            {0,5,1,0,15},       /* B */
 +            {0,5,3,0,15},       /* G */
 +            {0,5,5,0,15},       /* R */
 +            {0,5,7,0,15},       /* A */
 +        },
 +        .flags = PIX_FMT_BE,
 +    },
 +    [PIX_FMT_BGRA64LE] = {
 +        .name = "bgra64le",
 +        .nb_components= 4,
 +        .log2_chroma_w= 0,
 +        .log2_chroma_h= 0,
 +        .comp = {
 +            {0,5,1,0,15},       /* B */
 +            {0,5,3,0,15},       /* G */
 +            {0,5,5,0,15},       /* R */
 +            {0,5,7,0,15},       /* A */
 +        },
 +    },
      [PIX_FMT_BGR565BE] = {
          .name = "bgr565be",
          .nb_components= 3,
              {0,1,2,0,7},        /* A */
          },
      },
 +    [PIX_FMT_GBR24P] = {
 +        .name = "gbr24p",
 +        .nb_components= 3,
 +        .comp = {
 +            {1,0,1,0,7},        /* B */
 +            {0,0,1,0,7},        /* G */
 +            {2,0,1,0,7},        /* R */
 +        },
++        .flags = PIX_FMT_PLANAR | PIX_FMT_RGB,
++    },
+     [PIX_FMT_GBRP] = {
+         .name = "gbrp",
+         .nb_components= 3,
+         .log2_chroma_w= 0,
+         .log2_chroma_h= 0,
+         .comp = {
+             {0,0,1,0,7},        /* G */
+             {1,0,1,0,7},        /* B */
+             {2,0,1,0,7},        /* R */
+         },
+         .flags = PIX_FMT_PLANAR | PIX_FMT_RGB,
+     },
+     [PIX_FMT_GBRP9LE] = {
+         .name = "gbrp9le",
+         .nb_components= 3,
+         .log2_chroma_w= 0,
+         .log2_chroma_h= 0,
+         .comp = {
+             {0,1,1,0,8},        /* G */
+             {1,1,1,0,8},        /* B */
+             {2,1,1,0,8},        /* R */
+         },
+         .flags = PIX_FMT_PLANAR | PIX_FMT_RGB,
+     },
+     [PIX_FMT_GBRP9BE] = {
+         .name = "gbrp9be",
+         .nb_components= 3,
+         .log2_chroma_w= 0,
+         .log2_chroma_h= 0,
+         .comp = {
+             {0,1,1,0,8},        /* G */
+             {1,1,1,0,8},        /* B */
+             {2,1,1,0,8},        /* R */
+         },
+         .flags = PIX_FMT_BE | PIX_FMT_PLANAR | PIX_FMT_RGB,
+     },
+     [PIX_FMT_GBRP10LE] = {
+         .name = "gbrp10le",
+         .nb_components= 3,
+         .log2_chroma_w= 0,
+         .log2_chroma_h= 0,
+         .comp = {
+             {0,1,1,0,9},        /* G */
+             {1,1,1,0,9},        /* B */
+             {2,1,1,0,9},        /* R */
+         },
+         .flags = PIX_FMT_PLANAR | PIX_FMT_RGB,
+     },
+     [PIX_FMT_GBRP10BE] = {
+         .name = "gbrp10be",
+         .nb_components= 3,
+         .log2_chroma_w= 0,
+         .log2_chroma_h= 0,
+         .comp = {
+             {0,1,1,0,9},        /* G */
+             {1,1,1,0,9},        /* B */
+             {2,1,1,0,9},        /* R */
+         },
+         .flags = PIX_FMT_BE | PIX_FMT_PLANAR | PIX_FMT_RGB,
+     },
+     [PIX_FMT_GBRP16LE] = {
+         .name = "gbrp16le",
+         .nb_components= 3,
+         .log2_chroma_w= 0,
+         .log2_chroma_h= 0,
+         .comp = {
+             {0,1,1,0,15},       /* G */
+             {1,1,1,0,15},       /* B */
+             {2,1,1,0,15},       /* R */
+         },
+         .flags = PIX_FMT_PLANAR | PIX_FMT_RGB,
+     },
+     [PIX_FMT_GBRP16BE] = {
+         .name = "gbrp16be",
+         .nb_components= 3,
+         .log2_chroma_w= 0,
+         .log2_chroma_h= 0,
+         .comp = {
+             {0,1,1,0,15},       /* G */
+             {1,1,1,0,15},       /* B */
+             {2,1,1,0,15},       /* R */
+         },
+         .flags = PIX_FMT_BE | PIX_FMT_PLANAR | PIX_FMT_RGB,
      },
  };
  
Simple merge
@@@ -154,18 -150,16 +154,33 @@@ enum PixelFormat 
      PIX_FMT_YUV422P9BE, ///< planar YUV 4:2:2, 18bpp, (1 Cr & Cb sample per 2x1 Y samples), big-endian
      PIX_FMT_YUV422P9LE, ///< planar YUV 4:2:2, 18bpp, (1 Cr & Cb sample per 2x1 Y samples), little-endian
      PIX_FMT_VDA_VLD,    ///< hardware decoding through VDA
-     PIX_FMT_GBR24P,    ///< planar GBR, 24bpp, 8G, 8B, 8R.
 +
++#ifdef AV_PIX_FMT_ABI_GIT_MASTER
 +    PIX_FMT_RGBA64BE,  ///< packed RGBA 16:16:16:16, 64bpp, 16R, 16G, 16B, 16A, the 2-byte value for each R/G/B/A component is stored as big-endian
 +    PIX_FMT_RGBA64LE,  ///< packed RGBA 16:16:16:16, 64bpp, 16R, 16G, 16B, 16A, the 2-byte value for each R/G/B/A component is stored as little-endian
 +    PIX_FMT_BGRA64BE,  ///< packed RGBA 16:16:16:16, 64bpp, 16B, 16G, 16R, 16A, the 2-byte value for each R/G/B/A component is stored as big-endian
 +    PIX_FMT_BGRA64LE,  ///< packed RGBA 16:16:16:16, 64bpp, 16B, 16G, 16R, 16A, the 2-byte value for each R/G/B/A component is stored as little-endian
++#endif
+     PIX_FMT_GBRP,      ///< planar GBR 4:4:4 24bpp
+     PIX_FMT_GBRP9BE,   ///< planar GBR 4:4:4 27bpp, big endian
+     PIX_FMT_GBRP9LE,   ///< planar GBR 4:4:4 27bpp, little endian
+     PIX_FMT_GBRP10BE,  ///< planar GBR 4:4:4 30bpp, big endian
+     PIX_FMT_GBRP10LE,  ///< planar GBR 4:4:4 30bpp, little endian
+     PIX_FMT_GBRP16BE,  ///< planar GBR 4:4:4 48bpp, big endian
+     PIX_FMT_GBRP16LE,  ///< planar GBR 4:4:4 48bpp, little endian
++
++#ifndef AV_PIX_FMT_ABI_GIT_MASTER
++    PIX_FMT_RGBA64BE=0x123,  ///< packed RGBA 16:16:16:16, 64bpp, 16R, 16G, 16B, 16A, the 2-byte value for each R/G/B/A component is stored as big-endian
++    PIX_FMT_RGBA64LE,  ///< packed RGBA 16:16:16:16, 64bpp, 16R, 16G, 16B, 16A, the 2-byte value for each R/G/B/A component is stored as little-endian
++    PIX_FMT_BGRA64BE,  ///< packed RGBA 16:16:16:16, 64bpp, 16B, 16G, 16R, 16A, the 2-byte value for each R/G/B/A component is stored as big-endian
++    PIX_FMT_BGRA64LE,  ///< packed RGBA 16:16:16:16, 64bpp, 16B, 16G, 16R, 16A, the 2-byte value for each R/G/B/A component is stored as little-endian
++#endif
      PIX_FMT_NB,        ///< number of pixel formats, DO NOT USE THIS if you want to link with shared libav* because the number of formats might differ between versions
  };
  
 +#define PIX_FMT_Y400A PIX_FMT_GRAY8A
++#define PIX_FMT_GBR24P PIX_FMT_GBRP
 +
  #if AV_HAVE_BIGENDIAN
  #   define PIX_FMT_NE(be, le) PIX_FMT_##be
  #else
  #define PIX_FMT_YUV422P16 PIX_FMT_NE(YUV422P16BE, YUV422P16LE)
  #define PIX_FMT_YUV444P16 PIX_FMT_NE(YUV444P16BE, YUV444P16LE)
  
 +#define PIX_FMT_RGBA64 PIX_FMT_NE(RGBA64BE, RGBA64LE)
 +#define PIX_FMT_BGRA64 PIX_FMT_NE(BGRA64BE, BGRA64LE)
+ #define PIX_FMT_GBRP9     PIX_FMT_NE(GBRP9BE ,    GBRP9LE)
+ #define PIX_FMT_GBRP10    PIX_FMT_NE(GBRP10BE,    GBRP10LE)
+ #define PIX_FMT_GBRP16    PIX_FMT_NE(GBRP16BE,    GBRP16LE)
  #endif /* AVUTIL_PIXFMT_H */
@@@ -1968,6 -1807,91 +1968,91 @@@ static void rgb24ToUV_half_c(int16_t *d
      }
  }
  
 -static void planar_rgb_to_y(uint8_t *dst, const uint8_t *src[4], int width)
++static void planar_rgb_to_y(uint16_t *dst, const uint8_t *src[4], int width)
+ {
+     int i;
+     for (i = 0; i < width; i++) {
+         int g = src[0][i];
+         int b = src[1][i];
+         int r = src[2][i];
 -        dst[i] = ((RY * r + GY * g + BY * b + (33 << (RGB2YUV_SHIFT - 1))) >> RGB2YUV_SHIFT);
++        dst[i] = (RY*r + GY*g + BY*b + (0x801<<(RGB2YUV_SHIFT-7))) >> (RGB2YUV_SHIFT-6);
+     }
+ }
+ static void planar_rgb16le_to_y(uint8_t *_dst, const uint8_t *_src[4], int width)
+ {
+     int i;
+     const uint16_t **src = (const uint16_t **) _src;
+     uint16_t *dst = (uint16_t *) _dst;
+     for (i = 0; i < width; i++) {
+         int g = AV_RL16(src[0] + i);
+         int b = AV_RL16(src[1] + i);
+         int r = AV_RL16(src[2] + i);
+         dst[i] = ((RY * r + GY * g + BY * b + (33 << (RGB2YUV_SHIFT - 1))) >> RGB2YUV_SHIFT);
+     }
+ }
+ static void planar_rgb16be_to_y(uint8_t *_dst, const uint8_t *_src[4], int width)
+ {
+     int i;
+     const uint16_t **src = (const uint16_t **) _src;
+     uint16_t *dst = (uint16_t *) _dst;
+     for (i = 0; i < width; i++) {
+         int g = AV_RB16(src[0] + i);
+         int b = AV_RB16(src[1] + i);
+         int r = AV_RB16(src[2] + i);
+         dst[i] = ((RY * r + GY * g + BY * b + (33 << (RGB2YUV_SHIFT - 1))) >> RGB2YUV_SHIFT);
+     }
+ }
 -static void planar_rgb_to_uv(uint8_t *dstU, uint8_t *dstV, const uint8_t *src[4], int width)
++static void planar_rgb_to_uv(uint16_t *dstU, uint16_t *dstV, const uint8_t *src[4], int width)
+ {
+     int i;
+     for (i = 0; i < width; i++) {
+         int g = src[0][i];
+         int b = src[1][i];
+         int r = src[2][i];
 -        dstU[i] = (RU * r + GU * g + BU * b + (257 << RGB2YUV_SHIFT)) >> (RGB2YUV_SHIFT + 1);
 -        dstV[i] = (RV * r + GV * g + BV * b + (257 << RGB2YUV_SHIFT)) >> (RGB2YUV_SHIFT + 1);
++        dstU[i] = (RU*r + GU*g + BU*b + (0x4001<<(RGB2YUV_SHIFT-7))) >> (RGB2YUV_SHIFT-6);
++        dstV[i] = (RV*r + GV*g + BV*b + (0x4001<<(RGB2YUV_SHIFT-7))) >> (RGB2YUV_SHIFT-6);
+     }
+ }
+ static void planar_rgb16le_to_uv(uint8_t *_dstU, uint8_t *_dstV, const uint8_t *_src[4], int width)
+ {
+     int i;
+     const uint16_t **src = (const uint16_t **) _src;
+     uint16_t *dstU = (uint16_t *) _dstU;
+     uint16_t *dstV = (uint16_t *) _dstV;
+     for (i = 0; i < width; i++) {
+         int g = AV_RL16(src[0] + i);
+         int b = AV_RL16(src[1] + i);
+         int r = AV_RL16(src[2] + i);
+         dstU[i] = (RU * r + GU * g + BU * b + (257 << RGB2YUV_SHIFT)) >> (RGB2YUV_SHIFT + 1);
+         dstV[i] = (RV * r + GV * g + BV * b + (257 << RGB2YUV_SHIFT)) >> (RGB2YUV_SHIFT + 1);
+     }
+ }
+ static void planar_rgb16be_to_uv(uint8_t *_dstU, uint8_t *_dstV, const uint8_t *_src[4], int width)
+ {
+     int i;
+     const uint16_t **src = (const uint16_t **) _src;
+     uint16_t *dstU = (uint16_t *) _dstU;
+     uint16_t *dstV = (uint16_t *) _dstV;
+     for (i = 0; i < width; i++) {
+         int g = AV_RB16(src[0] + i);
+         int b = AV_RB16(src[1] + i);
+         int r = AV_RB16(src[2] + i);
+         dstU[i] = (RU * r + GU * g + BU * b + (257 << RGB2YUV_SHIFT)) >> (RGB2YUV_SHIFT + 1);
+         dstV[i] = (RV * r + GV * g + BV * b + (257 << RGB2YUV_SHIFT)) >> (RGB2YUV_SHIFT + 1);
+     }
+ }
  static void hScale16To19_c(SwsContext *c, int16_t *_dst, int dstW, const uint8_t *_src,
                             const int16_t *filter,
                             const int16_t *filterPos, int filterSize)
@@@ -2146,12 -2061,16 +2230,16 @@@ static av_always_inline void hyscale(Sw
                                       uint8_t *formatConvBuffer,
                                       uint32_t *pal, int isAlpha)
  {
 -    void (*toYV12)(uint8_t *, const uint8_t *, int, uint32_t *) = isAlpha ? c->alpToYV12 : c->lumToYV12;
 +    void (*toYV12)(uint8_t *, const uint8_t *, const uint8_t *, const uint8_t *, int, uint32_t *) = isAlpha ? c->alpToYV12 : c->lumToYV12;
      void (*convertRange)(int16_t *, int) = isAlpha ? NULL : c->lumConvertRange;
+     const uint8_t *src = src_in[isAlpha ? 3 : 0];
  
      if (toYV12) {
-         toYV12(formatConvBuffer, src, src2, src3, srcW, pal);
 -        toYV12(formatConvBuffer, src, srcW, pal);
++        toYV12(formatConvBuffer, src, src_in[1], src_in[2], srcW, pal);
          src= formatConvBuffer;
+     } else if (c->readLumPlanar && !isAlpha) {
+         c->readLumPlanar(formatConvBuffer, src_in, srcW);
+         src = formatConvBuffer;
      }
  
      if (!c->hyscale_fast) {
@@@ -2189,9 -2104,15 +2277,15 @@@ static av_always_inline void hcscale(Sw
                                       const int16_t *hChrFilterPos, int hChrFilterSize,
                                       uint8_t *formatConvBuffer, uint32_t *pal)
  {
+     const uint8_t *src1 = src_in[1], *src2 = src_in[2];
      if (c->chrToYV12) {
 -        uint8_t *buf2 = formatConvBuffer + FFALIGN(srcW * FFALIGN(c->srcBpc, 8) >> 3, 16);
 -        c->chrToYV12(formatConvBuffer, buf2, src1, src2, srcW, pal);
 +        uint8_t *buf2 = formatConvBuffer + FFALIGN(srcW*2+78, 16);
-         c->chrToYV12(formatConvBuffer, buf2, src0, src1, src2, srcW, pal);
++        c->chrToYV12(formatConvBuffer, buf2, src_in[0], src1, src2, srcW, pal);
+         src1= formatConvBuffer;
+         src2= buf2;
+     } else if (c->readChrPlanar) {
 -        uint8_t *buf2 = formatConvBuffer + FFALIGN(srcW * FFALIGN(c->srcBpc, 8) >> 3, 16);
++        uint8_t *buf2 = formatConvBuffer + FFALIGN(srcW*2+78, 16);
+         c->readChrPlanar(formatConvBuffer, buf2, src_in, srcW);
          src1= formatConvBuffer;
          src2= buf2;
      }
@@@ -2860,7 -2753,6 +2966,7 @@@ static av_cold void sws_init_swScale_c(
          case PIX_FMT_RGB565BE: c->chrToYV12 = rgb16beToUV_c; break;
          case PIX_FMT_RGB555LE: c->chrToYV12 = rgb15leToUV_c; break;
          case PIX_FMT_RGB555BE: c->chrToYV12 = rgb15beToUV_c; break;
-         case PIX_FMT_GBR24P  : c->chrToYV12 = gbr24pToUV_c;  break;
++//        case PIX_FMT_GBR24P  : c->chrToYV12 = gbr24pToUV_c;  break;
          }
      }
  
      case PIX_FMT_RGB48LE: c->lumToYV12 = rgb48LEToY_c; break;
      case PIX_FMT_BGR48BE: c->lumToYV12 = bgr48BEToY_c; break;
      case PIX_FMT_BGR48LE: c->lumToYV12 = bgr48LEToY_c; break;
-     case PIX_FMT_GBR24P : c->lumToYV12 = gbr24pToY_c ; break;
++//    case PIX_FMT_GBR24P : c->lumToYV12 = gbr24pToY_c ; break;
      }
      if (c->alpPixBuf) {
          switch (srcFormat) {
@@@ -422,13 -420,23 +423,23 @@@ typedef struct SwsContext 
      yuv2packed2_fn yuv2packed2;
      yuv2packedX_fn yuv2packedX;
  
 -    void (*lumToYV12)(uint8_t *dst, const uint8_t *src,
 +    void (*lumToYV12)(uint8_t *dst, const uint8_t *src, const uint8_t *src2, const uint8_t *src3,
                        int width, uint32_t *pal); ///< Unscaled conversion of luma plane to YV12 for horizontal scaler.
 -    void (*alpToYV12)(uint8_t *dst, const uint8_t *src,
 +    void (*alpToYV12)(uint8_t *dst, const uint8_t *src, const uint8_t *src2, const uint8_t *src3,
                        int width, uint32_t *pal); ///< Unscaled conversion of alpha plane to YV12 for horizontal scaler.
      void (*chrToYV12)(uint8_t *dstU, uint8_t *dstV,
 -                      const uint8_t *src1, const uint8_t *src2,
 +                      const uint8_t *src1, const uint8_t *src2, const uint8_t *src3,
                        int width, uint32_t *pal); ///< Unscaled conversion of chroma planes to YV12 for horizontal scaler.
+     /**
+       * Functions to read planar input, such as planar RGB, and convert
+       * internally to Y/UV.
+       */
+     /** @{ */
+     void (*readLumPlanar)(uint8_t *dst, const uint8_t *src[4], int width);
+     void (*readChrPlanar)(uint8_t *dstU, uint8_t *dstV, const uint8_t *src[4], int width);
+     /** @} */
      /**
       * Scale one horizontal line of input data using a bilinear filter
       * to produce one line of output data. Compared to SwsContext->hScale(),
@@@ -521,99 -529,37 +532,44 @@@ SwsFunc ff_yuv2rgb_init_altivec(SwsCont
  SwsFunc ff_yuv2rgb_get_func_ptr_bfin(SwsContext *c);
  void ff_bfin_get_unscaled_swscale(SwsContext *c);
  
 +#if FF_API_SWS_FORMAT_NAME
 +/**
 + * @deprecated Use av_get_pix_fmt_name() instead.
 + */
 +attribute_deprecated
  const char *sws_format_name(enum PixelFormat format);
 +#endif
  
- //FIXME replace this with something faster
- #define is16BPS(x)      (           \
-            (x)==PIX_FMT_GRAY16BE    \
-         || (x)==PIX_FMT_GRAY16LE    \
-         || (x)==PIX_FMT_BGR48BE     \
-         || (x)==PIX_FMT_BGR48LE     \
-         || (x)==PIX_FMT_RGB48BE     \
-         || (x)==PIX_FMT_RGB48LE     \
-         || (x)==PIX_FMT_BGRA64BE    \
-         || (x)==PIX_FMT_BGRA64LE    \
-         || (x)==PIX_FMT_RGBA64BE    \
-         || (x)==PIX_FMT_RGBA64LE    \
-         || (x)==PIX_FMT_YUV420P16LE \
-         || (x)==PIX_FMT_YUV422P16LE \
-         || (x)==PIX_FMT_YUV444P16LE \
-         || (x)==PIX_FMT_YUV420P16BE \
-         || (x)==PIX_FMT_YUV422P16BE \
-         || (x)==PIX_FMT_YUV444P16BE \
-     )
- #define isNBPS(x)       (           \
-            (x)==PIX_FMT_YUV420P9LE  \
-         || (x)==PIX_FMT_YUV420P9BE  \
-         || (x)==PIX_FMT_YUV422P9LE  \
-         || (x)==PIX_FMT_YUV422P9BE  \
-         || (x)==PIX_FMT_YUV444P9BE  \
-         || (x)==PIX_FMT_YUV444P9LE  \
-         || (x)==PIX_FMT_YUV422P10BE \
-         || (x)==PIX_FMT_YUV422P10LE \
-         || (x)==PIX_FMT_YUV444P10BE \
-         || (x)==PIX_FMT_YUV444P10LE \
-         || (x)==PIX_FMT_YUV420P10LE \
-         || (x)==PIX_FMT_YUV420P10BE \
-         || (x)==PIX_FMT_YUV422P10LE \
-         || (x)==PIX_FMT_YUV422P10BE \
-     )
- #define is9_OR_10BPS isNBPS //for ronald
- #define isBE(x) ((x)&1)
- #define isPlanar8YUV(x) (           \
-            (x)==PIX_FMT_YUV410P     \
-         || (x)==PIX_FMT_YUV420P     \
-         || (x)==PIX_FMT_YUVA420P    \
-         || (x)==PIX_FMT_YUV411P     \
-         || (x)==PIX_FMT_YUV422P     \
-         || (x)==PIX_FMT_YUV444P     \
-         || (x)==PIX_FMT_YUV440P     \
-         || (x)==PIX_FMT_NV12        \
-         || (x)==PIX_FMT_NV21        \
-     )
- #define isPlanarYUV(x)  (           \
-         isPlanar8YUV(x)             \
-         || (x)==PIX_FMT_YUV420P9LE  \
-         || (x)==PIX_FMT_YUV422P9LE  \
-         || (x)==PIX_FMT_YUV444P9LE  \
-         || (x)==PIX_FMT_YUV420P10LE \
-         || (x)==PIX_FMT_YUV422P10LE \
-         || (x)==PIX_FMT_YUV444P10LE \
-         || (x)==PIX_FMT_YUV420P16LE \
-         || (x)==PIX_FMT_YUV422P10LE \
-         || (x)==PIX_FMT_YUV422P16LE \
-         || (x)==PIX_FMT_YUV444P16LE \
-         || (x)==PIX_FMT_YUV420P9BE  \
-         || (x)==PIX_FMT_YUV422P9BE  \
-         || (x)==PIX_FMT_YUV444P9BE  \
-         || (x)==PIX_FMT_YUV420P10BE \
-         || (x)==PIX_FMT_YUV422P10BE \
-         || (x)==PIX_FMT_YUV444P10BE \
-         || (x)==PIX_FMT_YUV420P16BE \
-         || (x)==PIX_FMT_YUV422P10BE \
-         || (x)==PIX_FMT_YUV422P16BE \
-         || (x)==PIX_FMT_YUV444P16BE \
-     )
+ #define is16BPS(x) \
+     (av_pix_fmt_descriptors[x].comp[0].depth_minus1 == 15)
  
- #define isPlanar(x)  (              \
-         isPlanarYUV(x)              \
-         ||  (x)==PIX_FMT_GBR24P     \
-     )
+ #define is9_OR_10BPS(x) \
+     (av_pix_fmt_descriptors[x].comp[0].depth_minus1 == 8 || \
+      av_pix_fmt_descriptors[x].comp[0].depth_minus1 == 9)
  
- #define isYUV(x)        (           \
-            (x)==PIX_FMT_UYVY422     \
-         || (x)==PIX_FMT_YUYV422     \
-         || isPlanarYUV(x)           \
-     )
++#define isNBPS(x) is9_OR_10BPS(x)
++
+ #define isBE(x) \
+     (av_pix_fmt_descriptors[x].flags & PIX_FMT_BE)
+ #define isYUV(x) \
+     (!(av_pix_fmt_descriptors[x].flags & PIX_FMT_RGB) && \
+      av_pix_fmt_descriptors[x].nb_components >= 2)
+ #define isPlanarYUV(x) \
+     ((av_pix_fmt_descriptors[x].flags & PIX_FMT_PLANAR) && \
+      isYUV(x))
+ #define isRGB(x) \
+     (av_pix_fmt_descriptors[x].flags & PIX_FMT_RGB)
 -
+ #if 0 // FIXME
+ #define isGray(x) \
+     (!(av_pix_fmt_descriptors[x].flags & PIX_FMT_PAL) && \
+      av_pix_fmt_descriptors[x].nb_components <= 2)
+ #else
  #define isGray(x)       (           \
             (x)==PIX_FMT_GRAY8       \
 -        || (x)==PIX_FMT_Y400A      \
 +        || (x)==PIX_FMT_GRAY8A      \
          || (x)==PIX_FMT_GRAY16BE    \
          || (x)==PIX_FMT_GRAY16LE    \
      )
          || (x)==PIX_FMT_MONOBLACK   \
          || (x)==PIX_FMT_MONOWHITE   \
      )
++
 +#define isRGBinBytes(x) (           \
 +           (x)==PIX_FMT_RGB48BE     \
 +        || (x)==PIX_FMT_RGB48LE     \
 +        || (x)==PIX_FMT_RGBA64BE    \
 +        || (x)==PIX_FMT_RGBA64LE    \
 +        || (x)==PIX_FMT_RGBA        \
 +        || (x)==PIX_FMT_ARGB        \
 +        || (x)==PIX_FMT_RGB24       \
 +    )
 +#define isBGRinBytes(x) (           \
 +           (x)==PIX_FMT_BGR48BE     \
 +        || (x)==PIX_FMT_BGR48LE     \
 +        || (x)==PIX_FMT_BGRA64BE    \
 +        || (x)==PIX_FMT_BGRA64LE    \
 +        || (x)==PIX_FMT_BGRA        \
 +        || (x)==PIX_FMT_ABGR        \
 +        || (x)==PIX_FMT_BGR24       \
 +    )
++
  #define isAnyRGB(x)     (           \
              isRGBinInt(x)           \
          ||  isBGRinInt(x)           \
 +        ||  (x)==PIX_FMT_GBR24P     \
      )
- #define isALPHA(x)      (           \
-            (x)==PIX_FMT_BGRA64BE    \
-         || (x)==PIX_FMT_BGRA64LE    \
-         || (x)==PIX_FMT_RGBA64BE    \
-         || (x)==PIX_FMT_RGBA64LE    \
-         || (x)==PIX_FMT_BGR32       \
-         || (x)==PIX_FMT_BGR32_1     \
-         || (x)==PIX_FMT_RGB32       \
-         || (x)==PIX_FMT_RGB32_1     \
-         || (x)==PIX_FMT_PAL8        \
-         || (x)==PIX_FMT_GRAY8A      \
-         || (x)==PIX_FMT_YUVA420P    \
-     )
++
+ #define isALPHA(x) \
+     (av_pix_fmt_descriptors[x].nb_components == 2 || \
+      av_pix_fmt_descriptors[x].nb_components == 4)
++#if 1
 +#define isPacked(x)         (       \
 +           (x)==PIX_FMT_PAL8        \
 +        || (x)==PIX_FMT_YUYV422     \
 +        || (x)==PIX_FMT_UYVY422     \
 +        || (x)==PIX_FMT_Y400A       \
 +        ||  isRGBinInt(x)           \
 +        ||  isBGRinInt(x)           \
 +    )
- #define usePal(x) ((av_pix_fmt_descriptors[x].flags & PIX_FMT_PAL) || (x) == PIX_FMT_GRAY8A)
++#else
+ #define isPacked(x) \
+     (av_pix_fmt_descriptors[x].nb_components >= 2 && \
+      !(av_pix_fmt_descriptors[x].flags & PIX_FMT_PLANAR))
++#endif
+ #define isPlanar(x) \
+     (av_pix_fmt_descriptors[x].nb_components >= 2 && \
+      (av_pix_fmt_descriptors[x].flags & PIX_FMT_PLANAR))
+ #define usePal(x) ((av_pix_fmt_descriptors[x].flags & PIX_FMT_PAL) || (x) == PIX_FMT_Y400A)
  
  extern const uint64_t ff_dither4[2];
  extern const uint64_t ff_dither8[2];
Simple merge
@@@ -144,7 -137,13 +144,14 @@@ const static FormatEntry format_entries
      [PIX_FMT_YUV444P9LE]  = { 1 , 1 },
      [PIX_FMT_YUV444P10BE] = { 1 , 1 },
      [PIX_FMT_YUV444P10LE] = { 1 , 1 },
 +    [PIX_FMT_GBR24P]      = { 1 , 0 },
+     [PIX_FMT_GBRP]        = { 1 , 0 },
+     [PIX_FMT_GBRP9LE]     = { 1 , 0 },
+     [PIX_FMT_GBRP9BE]     = { 1 , 0 },
+     [PIX_FMT_GBRP10LE]    = { 1 , 0 },
+     [PIX_FMT_GBRP10BE]    = { 1 , 0 },
+     [PIX_FMT_GBRP16LE]    = { 1 , 0 },
+     [PIX_FMT_GBRP16BE]    = { 1 , 0 },
  };
  
  int sws_isSupportedInput(enum PixelFormat pix_fmt)